llllliillllllJIIHIIUHI‘IIIIIIHllxlll‘llllllllllillllllll

31293 01555 243

This is to certify that the

dissertation entitled

Economics of the Namibian Millet Subsector

presented by

Stefan K. Keyler

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in Agricultural Economics

(iofiflzi

 

 

 

Major professor

 

MSU is an Affirmative Action/Equal Opportunity Institution 0-12771

 

LIBRARY

Michigan State
University

 

 

 

PLACE IN RETURN BOX to romavo thio Chockout from your rocord.
TO AVOID F INES rotum on or boforo doto duo.

DATE DUE DATE DUE DATE DUE

       
   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

WWW

MSU lo An Affirmotivo AdioniEquoi OpportuMy Inotitubn

 

 

 

 

 

 

ECONOMICS OF THE NAMIBIAN MILLET SUBSECTOR
By

Stefan K. Keyler

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Agricultural Economics

1 996

ABSTRACT
Tl-IE ECONOMICS OF THE NAMIBIAN MILLET SUBSECTOR
By

Stefan K. Keyler

Facing a national grain deficit and recognizing that more than half of
Namibia's population concentrates in the primary millet production zones
Ovambo and Kavango (further called study zone), Namibian policymakers
decided to prioritize the development of the millet sector. During the 1992/93
production year household, trader, and price surveys were conducted in the
study zone. Based on the collected data this study uses descriptive and
regression analyses to identify the main performance determinants of the millet
sector. To predict the effects of grain policy/technology changes on household
food security, millet’s competitiveness, and national grain import requirements,
three models are developed simulating potential reductions of millet production
and marketing costs and comparing various millet processing options.

Research findings demonstrate that few farmers are millet self-sufficient
and only ten percent of the millet production is commercially traded. Most
people in the study zone prefer millet over maize. But millet’s scarcity and high
price leaves imported maize to dominate the commercial food market and
attracts millet imports from Angola.

Research findings indicate that millet yields can be increased if fertilizer

and improved agronomic practices are introduced. But contrary to expectations,

Okashana 1, a millet variety introduced in 1988/89, does not significantly
increase farrners' yields. However, Okashana's capability to mature within three
months can protect farmers from crop failures in short rainy seasons. Due to
higher annual rainfall and the availability of more fertile land Kavango has a
greater potential for commercial millet production than Ovambo.

Results from the simulation models reveal the following: (a) Small scale
hammerrnill technology on the community level is the most cost effective millet
processing solution; (b) Even under improved millet production and marketing
conditions a quarter of rural households remain food insecure; (c) If current
population growth rates persist, the study zone will further depend on grain
imports.

Because many small holders lack access to adequate production
resources even an improved millet sector can not solve all problems of rural
poverty and food insecurity. This finding leads to the recommendation that
policymakers have to look beyond the farm level toward income and employment

creation in rural non-farm activities.

Copyright by
Stefan Karl Keyler

1 996

To my wife Steffi and our children Annika and Florian with gratitude.

Acknowledgment

I would like to express my gratitude to Professor Carl Eicher, my major
professor for seven years, for his academic and personal guidance and
inspiration throughout my graduate program up to the last page of this
dissertation. I am grateful for the support of Professor Torn Reardon, my thesis
adviser. He was crucial for the completion of this study. Since his arrival at
Michigan State University four years ago, he was willing to engage with me in
countless discussions to make this work an intellectually challenging experience.

I would also like to thank the other members of my committee, Professors
James Bonnen, Steven Hanson, and Carl Liedholm for their support and
understanding guidance throughout various stages of my graduate program.

I am also thankful to the Professors Alan Schmid and James Shaffer for the
special twist they added to this program that not only influenced this research but
will have a continuous impact on my future thinking in economics and daily life.
Also to mention is the moral support and warm friendship I received from Patricia
Eisele, other staff members, and many graduate students that not only helped
me in critical times but encouraged me throughout my stay at MSU.

I am very grateful to the Evangelic Foundation for Graduate Studies at
Villigst in Germany for financing my course program in the USA. I would also

wish to express my gratitude to the lntemational Center of Research in the

vi

Semi-Arid Tropics (ICRISAT) and especially Dr. David Rohrbach for providing
me with the opportunity to conduct my field research in Namibia with funds from
the German Ministry of Economic Collaboration (BMZ). Thanks also to Eckehard
Weiss from the German Agency for Technical Collaboration (GTZ) for his un-
bureaucratic but professional support during times when backstopping with
ICRISAT was not possible.

Special thanks are extended to the Director of Planning Bemd Rothkegel
and his colleague Christof Brock from the Namibian Ministry of Agriculture who
hosted me in Windhoek and provided assistance throughout all phases of the
organization and conduct of my field research. My recognition also goes to my
research assistant, colleague, and friend Paul Nashitati who accompanied me
during the 18 months of survey work. Above all, special appreciation to the
enumerators and data entry persons, the extension officers, and all small holders
and millet traders for their contribution to this study.

Finally, I remain indebted to my and my wife’s families for their continuous
moral support and visits from Europe. My greatest debt and appreciation are to
my wife Steffi, who offered patience, encouragement, love, and care throughout

this wonderful USA/Namibia chapter in our and our children’s life.

vii

TABLE OF CONTENTS

Page

LIST OF TABLES ............................................... xi

LIST OF FIGURES .............................................. xiv

LIST OF ACRONYMS ........................................... xvi
CHAPTER

1. INTRODUCTION .............................................. 1

1.1. Background .......................................... 1

1.2. Problem statement ..................................... 4

1.3. Research objectives ................................... 8

1.4. Working hypotheses ................................... 9

1.5. Methodology ......................................... 11

1.6. Study outline ........................................ 15

2. DATA COLLECTION AND ORGANIZATION ....................... 17

2.1. Zone description ..................................... 17

2.1.1. Ovambo zone .................................. 18

2.1.2. Kavango zone ................................. 18

2.2. Data collection methods ............................... 19

2.2.2. Market level surveys ............................ 22

2.3. Research data ....................................... 23

2.3.1. Data description ................................ 23

2.3.2. Data organization ............................... 26

3. MILLET PRODUCTION ........................................ 29

3.1. Production system .................................... 29

3.1. 1. Importance of millet ............................. 29

3.1.2. Millet and other crops ............................ 30

3.1.3. Rainfall pattern ................................. 31

3.1.4. Cropping calendar .............................. 35

3.1.5. Field numbers and land use pattern ................ 35

3.1.7. Labor input .................................... 38

3.1.8. Plowing and draft power .......................... 40

3.1.9. Local millet varieties and the new variety Okashana 1 . .41

3.1.10. Storage practices and losses ..................... 44

3.2.Systemperfonnance.................. ................ 45

3.2.1. Hectares under cultivation ........................ 45

3.2.2. Millet yields ................................... 47

viii

3.2.3. Grains self-sufficiency levels ...................... 50

3.2.4. Cost and returns ................................ 53

3.2.5. Contribution to annual household income ............ 56

3.3. Determinants of millet producers’ performance ............ 62
3.3.1. Descriptive analysis ............................. 62

3.3.2. Regression analysis of millet area cultivated .......... 67

3.3.3. Regression analysis of millet yields ................. 88

3.3.4. Correlations analysis ............................ 99

3.4. Chapter summary ................................... 103
3.4.1. Characteristics of the millet production system ....... 104

3.4.2. Performance of the millet production system ......... 105

3.4.3. Millet yield determinants ......................... 108

3.4.4. Yield determinants and household characteristics ..... 110

4. MILLET MARKETING ........................................ 113
4.1. Millet prices ........................................ 114
4.1.1. Formation of millet prices in general ............... 114

4.1.2. Millet price structure and variation during 1993 ....... 117

4.1.3. Price competitiveness of millet versus maize ......... 122

4.2. Commercial millet marketing .......................... 127
4.2.1. Characteristics of commercial millet traders .......... 128

4.2.2. Traders' millet acquisition after harvest ............. 132
4.2.2.1. Acquisition period ....................... 133

4.2.2.2. Grain traders' millet suppliers .............. 133

4.2.2.3. Transportation ......................... 136

4.2.2.4. Amounts traded ........................ 137

4.2.2.5. Payment modes ........................ 138

4.2.2.6. Reference prices ....................... 139

4.2.2.7. Supply prices .......................... 141

4.2.2.8. Millet imports from Angola ................ 143

4.2.3. Millet traders' value adding and marketing activities . . . 145
4.2.3.1. Storage practices ....................... 146

4.2.3.2. Cleaning, processing and packaging ........ 148

4.2.3.3. Traders' selling period ................... 151

4.2.3.4. Characteristics of millet consumers ......... 153

4.2.3.5. Setting retail prices ...................... 155

4.2.3.6. Profit margins .......................... 156

4.2.4. Prospects and limits of commercial millet trade ....... 159
4.2.4.1. Aggregate size of commercial millet market . . . 160

4.2.4.2. Trader views about the prospect of millet ..... 163

4.2.4.3. Maize market in the study zones ........... 166

4.2.4.4. Grain traders' expectations from government . . 169

4.3. Farm level millet sales ................................ 170
4.3.1. Frequency and extent of farm level millet sales ....... 170

4.3.2. Limitations of millet marketing .................... 176
4.3.2.1. Marketing versus production constraints ..... 177
4.3.2.2. Grain reserves against drought ............ 178
4.3.2.3. Transportation limitations ................. 181
4.3.2.4. Lack of local markets .................... 186
4.3.3. Regression analysis of farmers' marketing decision . . . 188
4.4. Chapter summary ................................... 200
4.4.1. Farm level millet marketing ...................... 200
4.4.2. Commercial millet trade ......................... 201
4.4.3. Price position of millet .......................... 203
5. MILLET CONSUMPTION ..................................... 206
5.1. Grain preference and consumption structure ............. 206
5.1.1. Cereal consumption choices based on preference . . . . 207
5.1.1.1. Preference structure under limited choices . . . 207
5.1.1.2. Preference structure without limits .......... 208
5.1.1.3. Determinants of cereal preferences ......... 212
5.1.2. Food staple consumption pattern .................. 213
5.1.2.1. Seasonal staple grain consumption shifts . . . . 214
5.1.2.2. Other seasonal diet changes .............. 217
5.2. Constraints on millet consumption and coping strategies . . . 219
5.2.1. Limits to millet consumption on the household level . . . 219
5.2.1.3. Household level food expenditures ......... 220
5.2.1.2. Obstacles to millet purchases .............. 223
5.2.1.3. Accessibility of millet in comparison ......... 224
5.2.2. Coping strategies of farmers in grain shortage ....... 226
5.2.2.1. Strategies to cope with millet shortage ....... 226
5.2.2.2. Farmer reactions to emptying maize reserves . 229
5.3. Determinants of farm level grain purchases .............. 229
5.3.1. Regression analysis of millet purchases ............ 230
5.3.2. Regression analysis of maize meal purchases ....... 240
5.4. Chapter summary ................................... 249
5.4.1. Grain consumption pattern ....................... 249
5.4.2. Limitations to millet purchases .................... 250

6. SIMULATION OF MILLET PRODUCTION
AND MARKET IMPROVEMENTS ............................ 253
6.1. Simulation of potential millet sector improvements ........ 254
6.1.1. Simulation of increases in millet yields .............. 257
6.1.2. Simulation of improved marketing conditions ......... 259
6.1.3. Other factors influencing the millet sector performance . 261
6.2. Millet procurement model ............................. 262
6.2.1. Model description .............................. 263
6.2.2. Model results ................................. 272
6.2.2.1. Current competitiveness (Base Scenario 1) ..... 274
6.2.2.2. After sector improvements (Scenarios 2 - 4) ..... 275

6.3. Household grain balance model ........................ 278
6.3.1. Model description .............................. 280

6.3.2. Model results ................................. 284
6.3.2.1. Current food security conditions (Base Scenario 1) 285

6.3.2.2. Effects of sector improvements (Scenarios2 -4) . 290

6.3.2.3. Effect of non-farm employment (Scenario 5) . . . . 290

6.4. Grain import substitution model ....................... 293
6.4.1. Model description .............................. 294
6.4.2. Model results ................................. 297

6.4.2.1. Current import requirements (Base Scenario 1) . . 297
6.4.2.2. Effects of sector improvements (Scenarios2-4) . 300
6.4.2.3. Projections for the year 2005 (Scenarios1 -4) . . . 302

6.5. Chapter Summary ................................... 307

6.5.1. Purpose of the simulation models ................. 307

6.5.2. Results of the millet procurement model ............ 309

6.5.3. Results of the household grain balance model ....... 311

6.5.4. Results of the grain import substitution model ........ 313

7. SUMMARY AND CONCLUSIONS .............................. 316
7.1. Background ........................................ 316

7.2. Objectives and data sources .......................... 318

7.3. Conduct and performance of the millet subsector ......... 320

7.3.1. Current and new technologies .................... 320

7.3.2. Household income and grain self-sufficiency ......... 321

7.3.3. Marketable surplus of millet ..................... 323

7.3.4. Consumer grain preferences and consumer prices . . . . 324

7.3.5. Constraints on commercial millet marketing .......... 326

7.3.6. Angolan millet imports .......................... 326

7.4. Simulation results of policy and technological changes . . . . 327

7.4.1. Millet processing and competitiveness with maize ..... 329

7.4.2. Household food security ........................ 331

7.4.3. Grain imports ................................. 331

7.4.4. Constraints on the improvement of the millet subsector 332

7.5. Implications for millet policy .......................... 333

7.5.1. Crop production ............................... 333

7.5.2. Marketing in rural areas ......................... 334

7.5.3. Millet processing .............................. 336

7.6. Attacking rural poverty ............................... 336

7.7. Implications for future research ........................ 337
APPENDIX A ................................................. 339
LIST OF REFERENCES ........................................ 344

xi

Table

Table 2-1.
Table 2-2.
Table 3-1.
Table 3-2.
Table 3-3.
Table 3-4.
Table 3-5.
Table 3-6.
Table 3—7.

Table 3-8.
Table 3-9.

Table 3-10.
Table 3-11.
Table 3—12.
Table 3-13.
Table 3-14.
Table 3-15.
Table 3-16.
Table 3-17.
Table 3-18.
Table 3-19.
Table 3-20.
Table 3-21.

Table 4-1.
Table 4-2.
Table 4-3.
Table 4-4.
Table 4-5.
Table 4-6.
Table 4-7.
Table 4-8.
Table 4-9.

Table 4-10.
Table 4-11.
Table 4-12.
Table 4-13.
Table 4—14.
Table 4-15.
Table 4-16.
Table 4-17.

LIST OF TABLES

Page
List of Namibian Millet Subsector Surveys, 1992193 .......... 20
Data generated by the Millet Subsector Research Project ..... 24
Rainfall of the 1992193 production season in comparison ...... 34
Household distribution by hectares of cultivated land ......... 46
Hectares of grain cultivated per adult equivalent ............ 47
Millet yields of the 1992193 production season .............. 48
Aggregate grain production in Ovambo and Kavango ........ 51
Household distribution by grain self-sufficiency levels ........ 51
Household average grain consumption requirements and
production during the year 1992193 ...................... 52
The share of millet on total household grain production ....... 53
Cost and returns of millet production during the year 1992193 . . 54
Occupational distribution of the rural labor force ............. 57
Average household income and shares of income sources . . . . 58
Social poverty map for Ovambo and Kavango .............. 61
Field preparation constraints during the year 1992193 ........ 64
Land cultivation constraints during the year 1992193 ......... 66
Characteristics of the millet hectare regression variables ...... 69
Correlations of the CASHINCOME to other model variables . . . 79
Estimates of the millet hectare equations .................. 81
Correlations of EXPOKASH and millet selling frequencies ..... 87
Characteristics of the millet yield regression variables ........ 89
Estimates of the millet yield equations .................... 96
Main determinants of millet production ................... 100
Average food staple consumer prices from April 1993 ....... 117
Business activities of rural and urban millet traders ......... 131
Millet suppliers of rural and urban grain traders ............ 135
Traders' references to determine millet acquisition prices . . . . 140
Acquisition prices of millet traders in 1991/92 and 1992193 . . 142
Millet acquisition prices of selected grain traders ........... 144
Characterization of typical millet and maize meal buyers ..... 154
Traders' millet acquisition costs and profit margins in 1993 . . . 157
Examples for the profitability of millet trade in 1993 ......... 159
Millet supply to commercial grain market in 1993 ........... 161
Millet selling frequencies of rural households .............. 172
Farm households' pre-conditions for selling millet ........... 173
People or trading locations farmers sell millet .............. 176
Millet marketing constraints of rural household ............. 178
Millet reserve targets of rural households ................. 179
Marketing locations for larger amounts of millet ............ 182
Transportation needs for millet marketing stated by farmers . . 184

xii

Table

Table 4-18.
Table 4—19.

Table 4-20.
Table 4-21.

Table 5-1.
Table 5—2.
Table 5-3.
Table 54.
Table 5-5.

Table 5-6.
Table 5-7.
Table 5—8.

Table 5-9

Table 5-16.
Table 5-11.

Table 5-12.
Table 5-13.
Table 5-14.
Table 5-15.
Table 5.16.
Table 5-17.

Table 6-1.
Table 6-2.

Table 6-3.
Table 6-4.

Table 6-5.

Page
Cost of hiring transportation to sell/purchase millet ......... 185
Percent of farmers lacking access to local grain markets and
their attitudes regarding the need of such markets ......... 186
Characteristics of the variables in the millet selling equation . . 190
Estimates of the millet selling equation ................... 197
Consumer preference structure between millet and maize . . . . 208
Household’s preference for different cereals .............. 209
Determinants of staple food preferences stated by farmers . . . 213
Seasonality in grain consumption during the year 1992193 . . . 215
Determinants of millet and maize consumption during the
1992193 hungry season ............................... 216
Seasonality of farm households' food consumption ......... 218
Farmers' monthly cash spending across various items ....... 221
Monthly cash spending across various items by household
expenditure categories ............................... 222
Household food constraints on millet purchases ........... 224
Comparison between farmers' access to millet and maize . . . . 225
Households' access to various food grains during the 1992193
hungry season ...................................... 226
Farmers' coping strategies during millet and maize shortages . 227
Rural households' purchase sources for millet and maize . . . . 228
Characteristics of variables in the millet purchase equation . . . 231
Estimates of the millet purchase equation ................ 237
Variables in the maize meal purchasing equation .......... 241
Estimates of the maize meal purchase equation ........... 247
Millet meal procurement costs under current and improved
technology conditions ................................ 267
Comparison of millet meal procurement costs and prices of
maize meal ........................................ 273
Model Results: Percent of food insecure farm households . . . 286
Model Results: study zones grain import needs, grain
production, and production losses ...................... 299
Potential improvements of millet subsector parameters and
their application across the model scenarios .............. 308

xiii

Figure

Figure 1-1.
Figure 2-1.
Figure 3-1.
Figure 3-2.
Figure 3-3.
Figure 4-1.

Figure 4-2.
Figure 4-3.
Figure 4-4.
Figure 4-5.
Figure 6-1.
Figure 6-2.
Figure 6-3.
Figure 6-4.
Figure 6-5.
Figure 6—6.
Figure 6-7.
Figure 6-8.

Figure 6-9.

Figure 6-10.
Figure 6-11.
Figure 6-1 2.
Figure 6-1 3.
Figure 6-14.
Figure 6-15.

LIST OF FIGURES

Page

Overview of Namibia's population distribution ............... 2
Research organization and information flow ................ 27
Long term annual rainfall by study zone ................... 33
Millet cropping calenders of Ovambo and Kavango .......... 36
Millet prices regressed by maize prices and location ......... 71
Availability of unprocessed millet and maize meal at retail

level between February 1993 and January 1994 ........... 119
Ovambo: millet prices on informal and commercial markets
from February 1993 to January 1994 .................... 121
Ovambo: millet prices on informal and commercial markets
from February 1993 to January 1994 .................... 121
Ovambo: price comparison between millet and maize ....... 125
Kavango: price comparison between millet and maize ....... 126
Overview of the millet procurement model ................ 264
Millet procurement cost calculations for Ovambo/Scenario 1 . . 271
The household grain balance model for Ovambo ........... 282
Scenario 1/Ovambo: Food insecure households under

1992193 production and marketing costs ................. 287
Scenario 11Kavango: Food insecure households under

1992193 production and marketing costs ................. 287
Scenario 210vambo: Food insecure households under

reduced production and 1992193 marketing costs .......... 288
Scenario 21Kavango: Food insecure households under

reduced production and 1992193 marketing costs .......... 288
Scenario 310vambo: Food insecure households under

1992193 production and reduced marketing costs .......... 289
Scenario 31Kavango: Food insecure households under

1992193 production and reduced marketing costs .......... 289
Scenario 4lOvambo: Food insecure households under

reduced production and marketing costs ................. 290
Scenario 41Kavango: Food insecure households under

reduced production and marketing costs ................. 290
Scenario 510vambo: Food insecure households under

increased income from off-farm employment .............. 292
Scenario 51Kavango: Food insecure households under

increased income from off-farm employment .............. 292
Potential grain import substitution based on population

estimates for the year 1995 of Ovambo and Kavango ....... 300
Potential grain import substitution based on population

estimates for the year 2005 of Ovambo and Kavango ....... 304

xiv

Figure Page

Figure 6-16. Projections of aggregate grain consumption and import
requirements for the study zones until the year 2020 ........ 306
Figure 6—17. Projections of the potential millet subsector development
path and resulting grain import savings for the study zones . . . 307

FIH

GTZ
ICRISAT
MAWRD
NAB
NDC
NGO
NEWFIU
SWAPO

LIST OF ACRONYMS

Food insecure household

Deutsche Gesellschaft ftir Technische Zusammenarbeit, GmbH
lntemational Crops Research Institute for Semi-Arid Tropics
Ministry of Agriculture, Water, and Rural Development

Namibia Agronomic Board

Namibian Development Corporation Ltd. (former FNDC or ENOK)
Non government organization

Namibia Early Warning & Food Information Unit

South West Africa People’s Organization

1. INTRODUCTION

1.1. Background

After independence from South Africa in 1991, the new government of
Namibia increased its efforts to improve the living conditions in communal areas.
The focus of this study lies on the most populous of these areas, the two millet
production zones, Ovambo and Kavango (further called study zones), with an
estimated population of 680,000 or 53 percent of Namibia’s total population.
(Figure 1-1) While health and education in both zones have shown relatively fast
progress through the installation of hospitals and schools, the progress in the
subsistence-oriented sector of agricultural has been slow. Two of the factors
causing the slow progress in agriculture are evident: first, the agricultural
potential is generally limited by the dry and drought-prone climate. Second,
private or public institutions that provide support to farmers through extension '
and/or provision of production inputs like fertilizer or credit are either inadequate

or missing.‘

 

‘ In their Draft of “National Agricultural Policy”, November 1994, (MAWRD,
1994a) the Policy Task Force and the Directorate of Planning of the Ministry of
Agriculture, Water, Rural Development (MAWRD) acknowledged that the
Agricultural Extension Service located in areas of former homelands is still
lacking well-trained manpower, a coherent extension message regarding crop
production and communal livestock husbandry, and a suitable system to
disseminate information to the large number of farmers living in remote areas.
The Draft also recognizes the fact that a credit system that serves the need of
the communal farmers is missing and that the current activities of Namibia
Development Corporation Ldt. (NDC), the former colonial development
organization FNDC, are still rather project-oriented instead of addressing the
needs of rural communities.

2

Figure 1-1. Location of the study zones Ovambo and Kavango

 

\

- - ’V'I'GQEA ........... ZAMBIA

// / ‘~~-- _.......;
4yV7/7Z}?é7///zW-¢< M

BOTSWANA

  
  

 

 

‘‘‘‘‘

\ SOUTH AFRICA

 

 

 

 

 

3

WIthin the study zones’ cropping system, about 80 to 90 percent of the
calories produced by the communal farmers stem from pearl millet production."2
However,during the last two decades, the production of millet could not keep
pace with population growth. The decline of the average annual rainfall by at
least 100 mm over the last 20 years and the increased occurrence of droughts
during the last 10 years has further aggravated the situation. The expansion of
crop land and cutting of bushes and trees for firewood has led, especially in
Ovambo, to soil degradation and the desertification of large areas. Today, most
of the rural households produce millet that covers consumption needs for only
four to six months after the grain harvest (Hay, et al., 1990, p. 20).

Before independence, the continuous increase in Ovambo and Kavango’s
grain deficit was not perceived as a problem by South African decision makers.
The male labor force from the study zones was expected to earn income for food
purchases either through work in the mines of southern Namibia or through work
for the South African troops deployed along the Angolan boarder. In fact, many
communal households that earned money from such or similar employments

balanced their grain requirements with maize meal that they purchased from the

 

‘ According to Hay, the term millet refers to pearl or bulrush millet
(Pennisetum glaucum:sys., P. typhoideum, P. americanum). (Hay, et al., 1991)

2 This estimation is based on household level yield data acquired through
household surveys of the Millet Subsector Research Project, 1992193.

4
urban-based retail system developed during the presence of the South African

Army.1

1.2. Problem statement

With the achievement of independence, the grain deficit in the millet
production zones became more apparent and critical. The withdrawal of the
South African troops led to a decline of employment opportunities in northern
Namibia, while decreased demand for diamonds and uranium on the world
market reduced the need for seasonal migration labor in Namibia's mines. As a
result, many communal households lost their cash income sources and
experienced problems paying for their basic food needs. The influx of 33,000
returnees from Angola to Ovambo and the severe drought during the 1991192

production season aggravated the economic situation.

 

‘ The maize meal that was and is bought by the communal population
originates predominantly from commercial farms in South Africa and Namibia.
The marketing of imported and nationally produced maize are controlled and
facilitated by the Namibian Agronomic Board. Besides determining maize
producer prices in collaboration with the Namibian government, the NAB
guarantees to purchase all maize production, holds stocks, supplies millers and
controls maize imports to maintain prices and margins. However, the customs
union agreement with South Africa obligates Namibia to first import maize from
South Africa and to only seek other sources on the international market if South
Africa is unable to provide enough maize from its own production. The latter
happened after the severe drought in 1992. According to the Directorate of
Planning, MAWRD, total imports of white maize for consumption amounted from
September 1992 until September 1993 to a total of 112,000 tons. Of these,
about 63 percent came from the USA and 37 percent came from South Africa.

5

Under the pressure of a continuing grain deficit1 and recognizing that the
majority of Namibia's population is concentrated in the millet producing zones,
agricultural policymakers decided to give priority to the development of the millet
subsector. This decision coincided with statements from Namibia's Chief
Agricultural Research Officer made in an ICRISAT publication that ‘Okashana 1',
the pearl millet variety introduced in 1989, had enabled Namibian farmers to
double their harvest from about 300 kglha, to more than 600 kglha, and that with
improved cultivation practices, Namibia could produce 2.4 tons per hectare in the
future. It was further stated that:

“... with better yields, the country faces the problem of processing
and marketing millet flour, which has to compete with the popular
maize flour from South Africa” (ICRISAT, 1991).

Based on this information, expectations among officials from the
Namibian Ministry of Agriculture, Water, and Rural Development (further called

MAWRD) were raised that increased millet production in Ovambo and Kavango

could solve many of Namibia’s pressing problems.2

 

‘ The hope of the newly elected government that commercial maize
production would increase within a few years after independence and as such
lead to reduced grain imports was soon invalidated. Between 1991 and 1994,
national grain production continued to cover only between 50 and 70 percent of
Namibia’s domestic grain need for human consumption. During this period,
commercial wheat production under irrigation increased from 6 to 12,000 tons,
while commercial maize production stagnated around 40,000 tons. With an
estimated annual production of about 50,000 tons, millet continued to provide
about 45 percent of total domestic grain supply or 25 to 35 percent of Namibia’s
grain need for human consumption.

2 Interview with the Deputy Permanent Secretary of MAWRD in
September 1992 and discussions with the Head of the Directorate of Planning in

6

On the local level, it was expected that the number of food-insecure
households could be reduced by increasing their own production. This
expectation was mainly based on the claim that the use of Okashana 1 alone
increases farmers’ yields to 600 kglha. Additionally, it was expected that many
households could sell surplus millet to the commercial food staple market (i.e.,
create income and employment opportunities for a large portion of Namibia’s
population).

On the regional level, it was assumed that the aggregate grain deficit
could soon be eradicated and that the establishment of new marketing channels
and processing facilities would generate new employment opportunities. It was
anticipated, additionally, that millet meal made from surplus production could
become price competitive against imported maize within and outside the millet
production zones.

On the national level, government officials hoped that increased grain
self-sufficiency in northern Namibia would reduce both national grain imports
(i.e., improve the nation’s trade balance) and the need for food aid. It was also
expected that more employment and income opportunities in the northern
agricultural sector could reduce the migration pressure from the communal north
to southern population centers.

Responding to the govemment’s desire to demonstrate quick

improvements in the communal millet subsector, the Namibian Agronomic Board

 

1992 and 1993.

7

(NAB) proposed that its mandate to control the pricing and marketing .of
commercial grain production should be expanded to include millet. The NAB
argued that with guaranteed producer prices, it would be possible to (a) acquire
about 1000 tons of millet each year,1 (b) process this millet with largeoscale
processing units to millet meal, and (c) sell this meal profitably to urban
consumers outside the millet production zones. The government, however,
opted to postpone its decision to implement price and market controls for millet.2

While the approval of the NAB proposal was postponed, MAWRD decided
that knowledge about prevailing costs and returns of millet production and

marketing was necessary to formulate an effective grain market policy for the

 

‘ It is estimated that 1000 tons cover less than 2 percent of the study
zones' total millet production. Therefore, it appears, the argument to offer a
stipulated price to all communal farmers was used to justify a marketing scheme
that would benefit only a very few communal surplus farmers.

2 Three main reasons led to the postponement of the govemment’s
decision to implement a controlled millet market in northern Namibia: First, a
formal marketing scheme that includes (a) the physical infrastructure for intake
points and processing facilities, (b) the organization of transport and other
' marketing activities, and (c) the recurring decision about annual producer prices
would be difficult to administer (this includes the problem that millet imports from
Angola could not be distinguished from domestic millet). Second, experiences of
neighboring countries indicated that formal marketing schemes for food grains
bear the risk of straining government finances, not only during the time of
implementation, but also over a long period of time. Third, from the specifics of
the NAB proposal, it became obvious that only a few and mainly very large
surplus producers could benefit from such scheme. It turned out that a specific
set of millet communal businessmen and surplus producers was originally meant
to be targeted who had already collaborated in a pilot marketing scheme
conducted by the NAB and the former colonial oriented FNDC (First National
Development Corporation), operating today under NDC (Namibian Development
Corporation). -

8
communal north. It was also acknowledged that an analysis about millet’s long
term competitiveness in the commercial food market was required. In August
1992, MAWRD began the “Millet Subsector Research Project’ in collaboration
with ICRISAT. The project’s main goals were (a) to collect basic information
about the current structure and conduct of the millet subsector, (b) to inform
grain market policymakers about alternative ways to support millet producers,
and (c) to evaluate the prospects for the commercialization of the millet

production.

1.3. Research objectives
The general objective of this dissertation research is to analyze household
and market-level data collected by the ‘Millet Subsector Research Project’ to

address the following questions:

1. What are the current patterns of millet production, marketing, and
consumption at the household level; i.e., what are millet subsector

participants doing?

2. Which factors determine household behavior at the three subsector
stages; i.e., what drives households' decisions regarding millet production,

millet marketing, and grain consumption?

3. What is the scope for millet yield increases, unit cost reductions for millet

production, and unit cost reductions for millet marketing in the future?

9

4. How would these millet subsector improvements affect the following three
millet subsector performance measures: (a) the number of food insecure
households, (b) the price competitiveness of locally produced and
processed millet against imported maize, and (c) the need to import maize

into the study zones ?

5. What recommendations can be drawn from the answers of the previous
questions for grain market policy, and what priority areas can be identified

for future food policy-relevant research?

1.4. Working hypotheses

The first hypothesis guiding this research is that adopting the millet variety
Okashana 1 without changes in production practices neither increases yields to
600 kglha, nor does it increase millet production enough to close the gap
between northern farmers’ grain production and grain consumption needs.

The second hypothesis concerns the commercialization prospects of
millet. It is hypothesized that providing incentives through guaranteed prices, as
suggested by the NAB, will not automatically convert the majority of subsistence-
oriented farmers to commercial producers. Two sub-hypotheses are used to
support this argument. First, limitations in both millet production and marketing
prevent farm households from responding to price incentives. Second, the fact
that most farm households are net grain buyers indicates that once more millet
can be produced, most of it will be consumed on the household level and not

sold.

1 0

The third hypotheSis argues, contrary to claims from the Namibian
Agronomic Board and the Namibian Development Corporation, that a
commercial millet market already exists, albeit, with severe deficiencies. The
proof of this hypothesis is important because it could lead to requests that the
government support the development of private traders rather than replacing
them through a market order that grants the right of grain market control in the
study zones to the NAB and the NDC as its executive branch.

The fourth hypothesis is that improved marketing conditions for locally
produced crops increase the likelihood of surplus farmers selling their
marketable millet surpluses; i.e., there are millet surplus farmers who would sell
millet already today if (a) the marketing of their surplus grain became physically
feasible, (b) their transaction costs for millet marketing could be reduced and/or
(c) the reductions in millet marketing costs made locally produced millet prices
competitive against imported maize.

The fifth hypothesis claims that introducing mechanical millet processing,
i.e., reducing labor requirements for manual pounding and providing millet in a
form that is 'ready-to-cook', makes the purchase of commercially offered millet
meal more attractive to potential millet buyers. But it is also hypothesized that
the type of processing technology used and the location of millet processing
facilities will affect the relative price of millet meal to such an extent that they
eventually become main factors influencing purchase decisions between millet

and maize meal.

1 1

The sixth hypothesis argues that the development and dissemination of
improved production technologies and the simultaneous improvement of millet
marketing conditions generally increase household food security, but do not fully
eradicate the existence of food insecure farm households in the study zones.

The seventh, and last, hypothesis states that even if potential increases in
millet production are achieved, they will not be large enough to provide
aggregate grain self-sufficiency in the study zones in the long-run, i.e., middle
and long-term population growth in the millet production zones Ovambo and
Kavango will overcompensate for potential millet production increases.

In sum, the study examines the production, marketing, and consumption
of millet in northern Namibia. The answers to the research questions and the
confirmation or rejection of the stated hypotheses will provide information about
(a) the conditions under which the current sector operates, (b) the factors that
propel or limit the development of the existing system, and (c) important
outcomes that government can expect from the introduction of a better millet

production technology and an improved grain marketing system.

1 .5. Methodology

This research concentrates on the fann-household level within the millet
subsector. Accordingly, most research questions are behavioral questions.
They ask either about rural households’ current behavior and its determinants or
about how changes in rural households’ opportunity set might affect their

behavior and the performance of the millet subsector as a whole.

12

Because most farm households from the study zones supply the majority
of their inputs that they use in farming and retain a substantial part of their output
for home consumption, farmers’ production and consumption must be examined
within the context of the farm household. However, because of limited data
available a comprehensive application of agricultural household decision theory
was not possible.

For this study, three methods are employed to answer the first three
research questions stated above:

The first method targets the first research question and is primarily of a
descriptive nature. Descriptive analysis of quantitative and qualitative data from
household and market level surveys will be used to establish a base of
information about the current circumstances of subsector participants at the
different sector stages.

The second method is used to answer the second research question. The
driving forces in household production, marketing, and consumption behavior are
determined with the help of regression analysis of cross sectional data collected
during the production season of 1992193. For a better interpretation of the
regression results, attitudinal statements of household representatives regarding
their motivations, goals, and limits are additionally employed.

The third method uses spreadsheet simulation models to address the
following three questions about the effects potential improvements in millet
production and marketing have on: (1) the price competitiveness of millet meal

against meal made from maize imports; (2) the number of food insecure farm

13
households in the study zones; and (3) study zones' short-term and long-term
grain import requirements. Each of these models relates with simple
mathematical equations the millet subsector parameters: (a) hectare yields,
(b) unit costs of production, and (c) unit marketing costs with one of the following
subsector performance measures: (a) procurement costs per unit millet meal ,
(b) number of food insecure households, and (c) amount of grain import required.

Within the established models, the millet subsector parameters take the
role of independent variables, while the subsector performance measures are
dependent variables that have to be estimated. To predict how much the millet
subsector parameters might change due to the availability of improved
production technologies and due to improvements in the communal grain
marketing system, informed estimates are made on the basis of primary cross
sectional and secondary data collected during the Namibian Millet Subsector
. Research Project, 1992193.

In order to relate millet subsector conditions with millets’ price
competitiveness (Question 1), simple cost accounting is used to calculate current
and potential millet meal procurement costs, i.e., accrued unit cost of millet
production, processing, and marketing are calculated. Based on the assumption
that market competition keeps profit margins of grain market participants close to
zero, the calculated procurement costs are treated as potential millet meal
market prices. Millet meal prices are finally compared with both currently
observed maize meal prices and with reduced maize meal prices, simulating

potential price declines of grain imports in the future.

14

The second model relates millet production and marketing parameters
with the number of food insecure households. (Question 2) For this purpose, the
model identifies households as food insecure if they neither produce enough
grain to cover their annual grain requirements nor earn enough cash income to
purchase the amount of grain necessary to balance their grain production deficit.
Based on this definition, sample households' food security levels are estimated
by comparing their annual grain consumption requirements with the sum of their
annual grain production and the amount of potential grain purchases they could
have made with their annual cash earnings. Finally, it is assumed that the
proportion of food insecure households among the surveyed household sample
reflects the actual distribution of food insecure households in the study zones.

The third model relates millet subsector conditions with aggregate grain
import requirements of the study zones (Question 3). For this purpose, study
zones’ grain import requirements are calculated by subtracting (a) the estimated
amount of subsistence production of millet and (b) the estimated amount of
purchased millet from (c) the aggregate grain consumption requirement of the
study zones.1

After modeling the mathematical relations between millet subsector
parameters and subsector performance measures in three separate models
and determining potential changes of millet subsector parameters under

improved conditions, each model is executed under four different scenarios:

 

‘ The factors named under a, b, and c are extrapolated from the values
estimated for the household sample of the Millet Subsector Household Surveys.

15

- Scenario 1 uses millet production and marketing conditions of the year
1993;

- Scenario 2 assumes, ceteris paribus, that millet production conditions

have been improved according to the estimates made before;

- Scenario 3 assumes, ceteris paribus, that millet marketing conditions have

been improved according to the estimates made before; and

- Scenario 4 assumes, ceteris paribus, that the improvements simulated in

Scenario 2 and 3 are achieved simultaneously.

1.6. Study outline

The study is organized in three sections. The first section consists of this
opening chapter and the second chapter. Chapter II shows where and how
primary data were collected and how these data are organized to match the
methods used during this study.

The second section of this study covers Chapter 3, 4, and 5. Each of
these chapters concentrates on one millet subsector stage. Chapter 3 covers
millet production by focusing on farmers' millet production practices, performance
criteria of the millet production system, and main determinants of this
performance. Chapter 4 deals with millet marketing issues by concentrating on
price positions of millet and maize in the informal and commercial market, millet
marketing in the commercial food market, and millet marketing practices of farm

households. Finally, Chapter 5 centers on farm households' grain consumption

16

pattern and their purchasing behavior. The performance of each sector stage is
assessed through descriptive analysis of household level behavior, while the
main determinants of rural households' behavior are identified through
regression analysis and through explanatory statements that originate directly
from survey participants.

The third section of the study contains chapter 6 and 7. Based on
findings from the previous chapters, three simple simulation models are
' developed in Chapter 6. The purpose of these simulations is to estimate the
effects that potential changes in millet production technology and/or marketing
conditions might have on local household food security levels, regional price
competitiveness of millet against imported maize, and national grain import
requirements. Finally, Chapter 7 summarizes the main findings of the study,
their policy implications, and suggests areas of focus for future food policy

relevant research.

2. DATA COLLECTION AND ORGANIZATION

2.1. Zone description

The two main millet production zones, Ovambo and Kavango, were
surveyed by the ‘Millet Subsector Research Project’ during the 1992/93
production year. Each of these zones covers the area of an ethnic group with its
own tradition and language. Due to Namibia’s colonial past, the zones' logistical
infrastructure (transport and telecommunication) as well as its social
infrastructure (education and health) are poorly developed. The agricultural
sector is dominated by low input levels and low yields.

Because Ovambo and Kavango were used for almost a century as cheap
labor reserves by the white South Africans, many northern farm households are
still used to sending each year seasonal migrating labor south to seek
employment either in the mining or fishing industries or in commercial agriculture.

In the north, both zones share long borders with Angola established
during the colonial period at the end of last century. Tribal culture and dialects
still reach from Namibia far into the southern Angolan territory. Although this
territory provides a better base for rain fed agriculture due to better soils and
higher annual rainfall, many of its areas are even less populated than those of
the study zones. This is probably caused by the very poor infrastructure of roads
provided by the Angolan government to its southern provinces. It is therefore not
surprising that despite the civil unrest over the last years in the Angolan territory,

many of the study zone farmers herd part of their cattle across the border. On

17

1 8
the other side, Angolan farmers who produce agricultural surpluses, but can not
reach Angolan consumer markets are often trying to sell these to the study

zones in exchange for western consumer goods imported from South Africa.

2.1.1. Ovambo zone

Ovambo, located in the northern center of Namibia, has an estimated
population of 671,000, of which 90 percent live in rural areas. Although the zone
comprises about 44 percent of Namibia’s total population, it covers only about
6.3 percent of Namibia's total area. Accordingly, the zone's population density is
the highest in the country.

The vegetation of Ovambo is mainly bush savanna with many areas that
are experiencing severe and ongoing desertification. The zone has no perennial
rivers that would permit extended crop production under irrigation. With an
average annual rainfall between 360 and 470 mm and 37 to 53 days of rainfall
during the year, the agro-climate of Ovambo is very poor. The food grains,

millet, and to a lesser extent, sorghum, are suited to this drought-prone climate.

2.1.2. Kavango zone

Kavango is located directly east of Ovambo. WIth an estimated
population of 136,000 (roughly 95 percent of those live in rural areas), this zone
comprises 9 percent of Namibia's population. With an area similar to the one of

Ovambo, Kavango's average population density is much less. However, about

19
90 percent of the population is concentrated in a strip 10 to 20 kilometers wide,
which follows the perennial Kavango river, the border to Angola.

With an average annual rainfall between 540 and 620 mm and 55 to 66
rain days per year, Kavango’s agro-climate is better than that of Ovambo. The
vegetation varies mainly between bush and forest. However, Kavango’s annual
rainfall is still so poor that again, only millet is grown as the main food grain.
Because in most cases the accumulated soil moisture dries up very early, maize
does not fully ripen. It is therefore only produced in small patches and eaten,
like a vegetable, in a green state before the millet harvest is available.

Although production conditions are better in Kavango, only a little
immigration occurred from Ovambo in the past. This is probably due to former
apartheid laws imposed by South Africa prohibiting movements from one
homeland to another. Still today, barriers to moving exist due to the differences
in culture, language, political perception, and tribal organization fostered by the

former apartheid regime.

2.2. Data collection methods

The data for this study originate primarily from four different surveys
carried out by the ‘Millet Subsector Research Project’ in 1992/93. Two of the
four surveys focused on data collection about the current conditions in the millet

subsector from the household level, while the remaining two gathered

20

 

 

 

 

 

 

 

Table 2-1. List of Namibian Millet Subsector Surveys, 1992193
Survey Sample Date and Information Gathered
Size Sector
Activities
HH Survey 320 HHs, December 1992: family composition; food prices; input
(Phase 1) 16 com- Field Preparation used; field preparation practices; plot
munities data; storage practice; crop sales; taste
preferences; investment needs.
HH sunny 320 HHs, April 1993: non-farm income; consumption pattern;
(Phase 2) 16 com- Grain Growrng food prices; labor Input into crop
munities Phase production; production limitations; crop
sales; livestock sales; expectations
from government.
Price 19 areas, February 1993 to retail prices of maize meal, millet
Monitoring 45 retail January 1994 (unprocessed), bread, rice, sorghum
outlets,
Survey every two June 1993 to producer prices of millet in quantities of
weeks January 1994 buckets (20 liter) and 50 kg bags
HH sun/9y 320 HHs, August 1993: non-farm income, consumption pattern.
(Phase 3) 16 com- Millet Harvest and food prices, labor input into crop
munities Threshing production, millet harvest data,
production limitations. crop sales.
livestock sales, main expenditures,
investment priorities
Fi9|d M93- 320 HHs, August through size of plots used for grain production
summent 16 com November 1993 during 1992/93 season
munities
Survey
Trade.- 59 Millet September business information; millet acquisition
Survey Traders through November and selling information; marketing

1993: Marketing
Season

margins; topology of millet sellers;
characterization of millet and maize
consumers; cost of value adding
activities; maize trade; perceptions
about grain market development

Data Source: Namibian Millet Subsector Research Project. 1992-93

21
information from millet traders about the existing marketing system and

monitored staple food prices in rural and urban areas during the year 1993.

2.2.1. Household level surveys

The purpose of the household level surveys was to gather quantitative
and qualitative data about general household characteristics, cash income from
local and off-farm employment, general crop and livestock production patterns,
millet production practices including information about labor input and the use of
draft power, harvest results and hectare yields, grain marketing practices,
general food consumption pattern, amounts and prices of food purchases,
investment goals, needs, and expectations from the government.

The household level surveys consisted of three household surveys and
one field measurement survey. All four surveys aimed at the same sample of
320 rural households. Each of the three household surveys was carried out
during a particular millet production phase. The first survey was launched during
the period of field preparation in December 1992. In April 1993, the second
survey was conducted during the actual grain growing period, when field
management activities, such as weeding, insect and bird prevention, prevailed.
The third household survey, conducted in August 1993, covered the time after
millet harvest and threshing, when the results of the growing season became
evident. The field measurement survey stretched from August through
November 1993. During this period, the land that sample households had used

during the 1992/93 season to cultivate grain was measured.

22
2.2.2. Market level surveys

The two market level surveys differed in their purposes. The trader
survey aimed (a) to prove that a commercial but private millet market already
exists, and (b) to identify its strengths and weaknesses. So information was
gathered about commercial traders’ supply, transaction costs, value adding, and
selling of millet. The objective of the price monitoring survey was to identify the
structure and variation of millet and maize prices in the informal and formal grain
market over the period of one production season.

The millet trader survey was conducted between September and
November 1993. Because rural households had just finished grain threshing,
millet traded most vigorously at that time. For this survey, a total of 59 millet
traders have been identified and interviewed in rural and urban areas of the
study zones.

During the price monitoring survey (February 1993 through January
1994), retail prices for maize meal, unprocessed millet, bread, rice, and sorghum
were gathered every two weeks in 19 different areas from a total of 45 market
outlets. From June 1993 through January 1994, when millet was traded by rural
producers, millet producer prices were also gathered in the 19 areas covered by

the price monitoring survey.

23
2.3. Research date
This section explains briefly what data have been collected during the
Millet Subsector Project and how the study transforms theses data into

information that answers the research questions.

2.3.1. Data description

As demonstrated in the previous section, data have been collected on the
household and market level across several areas of the study zones at different
times during the 1992/93 production season. Table 2-2 presents an overview
about the data collected during the ’Millet Subsector Research Project'.

For this study, the available data are grouped into three basic types:
descriptive, quantitative, and attitudinal data whose characteristics are as
follows:

Descriptive data are not only of qualitative, but also of quantitative nature.
However, their information content is more valuable for the description of general
behavioral patterns than for quantitative methods that try to explain what drives
behavior or what could be expected if certain determinants of behavior are
changed.

Examples for descriptive data are the type of inputs used to produce
millet; the months when households tend to run short of their own grain
production; household grain production per capita; years during which
households experienced grain deficit production; the type of containers used to

store, transport and measure grain; the food staples different household member

Table 2-2.

Data generated by the Millet Subsector Research Project,

 

COLLECTION
LEVEL

DATA TYPE

 

1 992193
VARIABLE SET PERIODICITYI
- VARIABILITY

Household two times during production

Characteristics season

Household Wealth! two times during production

Assets season

Production Practices one time during the
respective phase of
production

Amount of Seasonal one time at end of production

Rainfall season

Input Use one time during respective
phase of production

Crop Output one time alter harvest

Attitudes about one time after ploughing and

Production De- after harvest

terminants

Food Consumption two times during hunger
season and after grain
harvest

Grain Transport Cost one time

Local Millet Selling two times during production
season

Millet Producer Prices

Commercial Millet
Marketing

Food Staple Prices

Millet 8. Maize
Purchases

Millet Processing Cost

alter grain harvest every to

weeks for six month

one time at after grain
harvest

three times during production

season;

every two weeks for eleven

month

two times during hunger

season and after grain
harvest

one time

household level

household level

household level

district level

household level

household level

household level

household level

household level

household level

at the 16 survey
villages and to
urban markets

grain retail -,
wholesale level

household level
16 survey villages

and two urban
markets

household level

large and middle
scale operation

bivariate, continuous

bivariate, continuous

qualitative. bivariate,
categorical

continuous

bivariate, categorical

confinuous

ootegorica

qualitative, categorical

continuous

qualitative. categorical

confinuous

qualitative. categorical,
continuous

continuous
continuous

qualitative, continuous

qualitative, continuous

 

Data Source: Namibian Millet Subsector Research Project. 1992-93

25
types prefer; the portion that millet trade provides to the total gross income of a
business; etc..

Quantitative data are defined in this study as data that can either be used
in cross section regression analyses or in spreadsheet simulations. This means,
that in some cases, also qualitative information, mainly from bivariate variables
and previously defined as descriptive data, will be considered as quantitative
data. Examples for this data category are the sex of the head of household; the
ownership of plowing equipment; the use of animal power for field preparation;
the distance to the nearest grain marketing place; the transportation cost to the
next marketplace; prices that households pay for food staples; the unit costs of
processing according to the types of used technology: etc...

Attitudinal data are based on survey participants’ statements that try to
explain the outcomes of their own behavior or the outcomes of activities done by
others. Although such statements might be grounded on very subjective or
judgmental perceptions in an individual case, their explanatory value is
significantly increased on the aggregate level. Examples of attitudinal data are
reasons given for exceptionally good grain harvests; factors identified as main
limitations to grain production or marketing; ranking of various investment
options based on the goal to improve the general standard of living;
characterization and differentiation of typical millet and maize meal consumers;
explanations of the existing price structure of commercially traded food grains;

etc. .

26
2.3.2. Data organization

Chapters 3 through 5 investigate three different millet subsector levels:
production, marketing, and consumption. (Figure 2-1) These chapters transform
the above described data through descriptive analyses and regression analyses
into knowledge about behavioral patterns and knowledge about behavioral
determinants. Part of this gained knowledge can be used to formulate policy
recommendations in the final Chapter 7. The other part has to be further
transformed through the application of three simulation models in Chapter 6
before grain market policymakers can be advised in Chapter 7. The gained
knowledge about behavioral patterns quantifies performance parameters of the
different millet subsector stages and thus directs policy- makers” attention
towards the most deficient aspects of the millet subsector.

The gained knowledge about behavioral determinant identifies not only
the factors that determine the performance of the subsector in general but
informs decision makers about those factors that can be influenced effectively by
grain market policy.

Part of both types of knowledge gained throughout Chapters 3, 4, and 5 is
passed to Chapter 6. On its basis, three spreadsheet models are established
that simulate improved millet production and/or marketing conditions and
estimate their effects on (1) millet meals' price competitiveness against meal

made from maize imports, (2) the percent of households that are food insecure

27

Figure 2-1. Research organization and information flow

RESULTS POLICY ADVICE

DATA ANALYSIS
(output) (evaluation)

(transformation)

 

Unpufl

 

Chapter 3 to 5 :
Chapter 7:

identification of :

production, marketing.
and consumption

 

 

 

 

 
    
 

 
  
     
 
  
 
   

 
  

 

 
 
   
  

 
  

 

 

    

 

 
 
  
 

 

  
 
 
 
  
   
   
  

 

 

descriptive subsector

5'“ - descriptive analysis behavioral limitations
attitudinal patterns ”‘9 and

data - regression analysis

determinants perfofm‘nce
quantitative determinants
data
behavioral
patterns and determinants
priorities for :

simulated Improvements
and their effects on:

 

 

 

 

agricultural policy

  

- millets’ price competitiveness
potential

 

  

and

     

assumed
subsector

 

- household level food security
future reseach

parameters

   

performance

  

 

changes

 

28
in the study zones; and (3) the grain import requirements the study zones face
in the short and long term.
Finally, in Chapter 7, the results from Chapter 3 through Chapter 6 are
summarized before recommendations for agricultural sector policy are

formulated and further research topics are suggested.

3. MILLET PRODUCTION
Chapter 3 is organized in three sections. The first section briefly
describes the prevailing millet production system. Section 2 assesses the
system performance in terms of factor productivity and contribution to household
income. The final section identifies the main determinants of millet producers’

performance.

3.1. Production system

To describe the prevailing system of communal millet production, this
section focuses briefly on the following topics: general importance of millet, millet
and other crops, rainfall pattern, cropping calendar, land use, soil fertility and
fertilizer use, labor input, plowing equipment and draft power, use of local and

improved seed, storage practices and losses.

3.1.1. Importance of millet

Without question, millet is the food staple of farmers in Ovambo and
Kavango. Of the 320 farmers interviewed during the household surveys of the
Millet Subsector Research Project, 95 percent identified millet as their most
important food crop.

Grain yield data from the 1992/93 production season show that millet
contributes, on average, 85 percent of households’ total grain production, while
the balance is supplied from sorghum and maize. Of those farm households that

do sell part of their agricultural production (Ovambo: 50 percent;

29

30

Kavango: 56 percent)‘, the largest group called millet their most important cash
crop2 (Ovambo: 21 percent, Kavango: 35 percent). It is estimated that of the
roughly 85,000 tons of millet produced in the study zones during the 1992/93
production year about 10 percent were commercially marketed. Although
Kavango’s population is only about a fourth of that from Ovambo and although
Kavango's 1992193 production was depressed due to poor rainfall half of the
marketed millet came from Kavango. This is an indication that the potential for
increases in millet market production lies in Kavango rather than in Ovambo.

According to an expert in breeding grains in semi-arid tropics, it seems
unlikely that pearl millet’s predominance in the study zones will change in the
future:

“...no other cereal is as well adapted to the climate and soil

conditions nor will provide the same stability of production as pearl

millet” (Bidinger, 1993, p.18).
3.1.2. Millet and other crops

Although farmers’ attention is mostly focused on millet production, they do
also grow sorghum and maize. However, the aggregate share of these two

grains on the total grain production in the 1992/93 season was only 8 percent in

Ovambo and 16 percent in Kavango. In addition to sorghum and maize, most

 

‘ Information presented for Ovambo and Kavango in parentheses reflects
findings from the Millet Subsector Research Project Surveys, 1992/93.

2 Estimates from the Grain Import Substitution Model presented in
Chapter 6 indicate that the amount of locally produced and marketed millet is
equivalent to 12 percent of the study zones' aggregate grain consumption
requirements and 27 percent of local millet production.

31
households intercrop beans (in Ovambo: 95 percent, in Kavango: 96 percent),
bambera nuts, watennelons, and pumpkins on their millet fields.

After the grains, millet, sorghum, and maize, beans are probably the
second major source of plant protein in rural household diet. Those farmers that
grew beans during the 1992/93 cropping season harvested, on average, 25 kg in
Ovambo and 40 kg in Kavango. But only a few farmers produce all four crops:
millet, sorghum, maize, and beans. In Ovambo, the most frequently used
combination of crops consists of millet, sorghum, and beans (59 percent of
households), while in Kavango, the largest group of households (30 percent)

produces a combination of millet, maize, and beans.

3.1.3. Rainfall pattern

The starting date, the amount, and the distribution of annual rainfall are
primary determinants of the agricultural production potential. The mean annual
rainfall figures from the last 50 years for the major millet production zones range
from 380 mm in western Ovambo to 620 mm in eastern Kavango.

The study zones’ beginning of rain season is defined as the date after
November 15 when rainfall is accumulated over 3 consecutive days to at least 20
mm and when thereafter no dry spell of more than 7 days occurs within 30 days
(Matanyaire, 1995). Because during the last few years dry spells occurred
frequently after mid November, many production seasons became extremely
short and thus produced yields that were close to those of drought years.

January and February are the wettest months and by the end of April the rain

32
season has usually ended.

The 12 year average of annual rainfall in the study zones reveals that at
the beginning of the 70s the average annual rainfall peaked the last time at
about 700 mm in Kavango and at about 525 mm in Ovambo. Since then, annual
rainfall continuously declined to the current low levels described in the previous
paragraph. (Figure 3-1) Whether the trend of the last two decades will further
continue or will come to an halt in the near future is unclear.

According to my analyses of monthly rainfall data, between 1969 and
19931 less than a third of all grain production seasons in the study zones can be
classified as good rainfall years (Ovambo: 20 percent, Kavango: 30 percent). In
both study zones about 30 percent of the production seasons can be classified
as middle or still productive rainfall seasons, while 50 percent of the rainfall
seasons in Ovambo and 40 percent in Kavango had to be classified as poor

rainfall seasons or droughts.

 

‘ Besides total seasonal precipitation, this analysis also took into
consideration the start, length, number of rainfall days, and dry spells longer than
a week.

Figure 3-1.

33

Long term annual rainfall by study zone
(12 year moving average)

 

 

 

 

    
 

 

 

 

Kavango

 

 

 

 

Ovambo

 

 

 

no
L
o
H
E
.. are
~
E
E
9‘
= 520
«'2
'5
F as
§
3.
t
O
2
o .140
(a
250
1903

1913

1923

1933

1943 1953 1993 1973 1993

1993

 

 

Data Source: Namibia Weather Bureau and my own calculations

34

Finally, Table 3-1 compares rainfall of the grain production season
1992/93 with the average rainfall of the last 23 production seasons. The data
demonstrate clearly that subzones from Kavango tend to receive more
accumulative precipitation than the subzones from Ovambo. However, contrary
to expectation, both rainfall data as well as sample farmers' statements about the
rainfall distribution of the 1992/93 production season revealed that only western
Ovambo had good rainfall distribution leading to an exceptionally good harvest in
1993. Further east, in central Ovambo, grain productivity was still acceptable
during the 1992193 season due to fairly good rainfall distribution. However, east
of central Ovambo, i.e., in eastern Ovambo and the whole of Kavango, the
rainfall season started very late, and during the rest of the season rainfall,

distribution was very poor.

Table 3-1. Rainfall of the 1992193 production season in comparison to the
average rainfall of the past 23 years, by subzone
(in mm of annual precipitation)

 

 

Ovambo Kavango
West Central East West Central East
1992193 season 287 266 na 552 481 571
1967-93 average 305 418 na 559 557 583

 

Data Source: Namibian Millet Subsector Project, 1992/93

35

3.1.4. Cropping calendar

Due to millet’s predominant position in the cropping system, most farm
activities are arranged around its production cycle. Once the rain season has
started and enough soil moisture has accumulated, rural households start with
field preparation. Due to earlier and more intensive rains, Kavango farmers plow
and plant usually between the second half of November and end of February,
while the majority of Ovambo households prepare their fields between the
beginning of January and the end of March. Weeding, as well as insect and bird
prevention, is mainly carried out between January and April. In both zones,
harvesting takes place between May and June while threshing is normally

finished at beginning of August (see Figure 3-2, next page).

3.1.5. Field numbers and land use pattern

Almost all Ovambo households produce their crops in one field that is
located close to or around their homestead. In Kavango, 50 percent of the farm
households produce their crops on two fields and another 23 percent have at
least three fields. The average distance between Kavango farmers’ homesteads
and their fields is about 5 to 7 kilometers. The variations in the number of fields

and distances are mainly caused by the different availability of cultivatable land.

Eco-0o E8... .31

 

   

gnaw £00.96 coin-om 8.0895 «0:5 cseEaz “00.590: 800

cools. soon I

 

 

 

9.232...

 

gamete:

 

 

 

 

assoc;

 

 

 

 

m 9.2.5... a 9.296:

 

 

 

«39.3

 

22.

 

uses

 

 

=a<

 

c052 Beacon Easel. LonEouoo LonEo>oz .3200 .onanum. 002<><x

 

 

 

 

 

 

 

 

9.235...

 

 

 

 

«sou—.3

 

2:1

 

snap

 

 

=5<

95335..

 

 

 

 

95.00;

 

 

 

.. .. seems. i
y ”i A am...%

i...

  

3:5... a 2226....

   

 

 
 

1.5.1..

 
 

5.5. I

goons. rained Ens—5.. LooEoooo ton—Ego: LBSoOLonEoEem, Gas—(>0

 

 

 

 

omega. ucu ano>O .8 23:23 5:3qu «2:2 .Né m¢=OE

 

 

37

In Ovambo, almost all cultivatable land is cleared and households have
settled close to it. On average, only 8 percent of all Ovambo sample households
stated they have cleared new land within the last five years. In Kavango, shifting
cultivation is still practiced. Forty eight percent of all Kavango households
cleared new land and 33 percent gave up at least one field during the same
period.

Farmers’ perception about the fertility of their fields varies considerably
across zones and subzones. The 25 percent of the sample farmers that classify
their soils as good in western Ovambo still compares favorably to the 14 percent
in eastern Ovambo and the 2 percent in central Ovambo. Contrary to Ovambo,
between 50 and 70 percent of all sample farmers from Kavango classify the
fertility of their soils as good. Taking into consideration that, due to population
growth, shifting cultivation has come to a halt in Ovambo, it is not surprising that
soil fertility is more depressed in this zone.

To counter the loss of soil fertility, most Ovambo farmers use either
manure (81 percent) and/or chemical fertilizer (9 percent). In Kavango, where
households still have the option to clear new land for crop production, much less
fertilizer is used. With the exception of households from western Kavango,
where much manure is available from cattle herding, only about 5 to 8 percent of
all Kavango households use manure on their fields and almost none use

chemical fertilizer.‘

 

‘ Neither primary nor secondary data are collected that indicate amounts
of manure and/or chemical fertilizer used by communal farmers per hectare.

38
3.1.7. Labor input

Within the current production system, family labor is one of the most
important variable inputs into millet production. From field preparation through
threshing, an average of 4.6 household members (51 percent of an average
sized household) are occupied with millet production in Ovambo and 3.4
household members in Kavango (43 percent). The time that an average
household spends for millet production (including harvest and threshing) is
estimated at 127 workdays for Ovambo (44 workdays per hectare) and only 76
workdays for Kavango (25 workdays per hectare).

The comparison between the two zones’ average number of workdays
spent per hectare of crop production are also far apart. In Ovambo, farm
households spent, on average, 73 workdays per hectare, while farm households
from Kavango spent only 49 workdays.

In both zones, women contribute 62 percent of the labor used for millet
production, while men contribute 33 percent, and children under 15 years
contribute 5 percent. Women’s over-proportional contribution to millet production

is explained mainly by four factors:

1) A high percentage of households have no male head of household that

could help with field work (Ovambo: 38 percent, Kavango: 16 percent).

(2) The analysis of seasonal migration patterns show that especially during
the months of millet production, significantly more men (Ovambo: 10

percent, Kavango 6 percent) leave their homesteads to seek off-farm

39

employment or to visit family relatives than women do

(Ovambo: 5 percent, Kavango: 2 percent).

(3) Men are traditionally less involved in crop production but more
concentrated on livestock husbandry and off-farm employment. Even in
cases where households have no livestock and no access to off-farm
employment, it is often accepted that men help only to plow and to

remove crop residuals from the fields after harvest.

(4) According to complaints from several tribal authorities, alcoholism has

become a severe problem among the male population in Kavango.

Labor exchange is common in both study zones. About 29 percent of
Kavango households hire laborers, which provide, on average, 39 percent millet
related work of these households. Only 14 percent of the Ovambo households
hire laborers. These laborers provide, on average, 13 percent of the
households’ millet-related work.

Again, women contribute most of the hired labor (60 to 70 percent). In
Ovambo, between 40 and 70 percent of the labor exchange takes place without
direct compensation through money or in kind. In Kavango only 10 to 15 percent
of the hired laborers work without payment. The wage rates that are paid for
agricultural laborers (cash or in kind) range mostly from NS 1.50 to NS 16.00 per
workday. In both zones, the average wage rate is roughly NS 5.00 per workday‘

and applies equally to both sexes.

 

‘ US $1.66 per day at the 1993 exchange rate

40
The data indicate that Kavango farmers face generally more labor
constraints for crop production than Ovambo farmers, and that those Kavango
farmers who can afford it balance their own labor shortage by hiring less wealthy

neighbors for money.

3.1.8. Plowing and draft power

While nearly 60 percent of the study zone farmers own a plow, less than
50 percent own both a plow and animals that could be used for plowing. Less
than two percent of the surveyed households have a tractor. Due to this
ownership structure, 20 percent of rural Ovambo households and 4 percent of
rural Kavango households prepare their fields with a hoe.

Of the households that use animal traction for field preparation (Ovambo:
53 percent, Kavango: 86 percent), a large portion hires draft animals from
neighbors (Ovambo: 42 percent, Kavango: 39 percent). Although very few farm
households own a tractor, still 30 percent of Ovambo households and 4 percent
of Kavango households were able to hire mechanical plowing during the 1992/93
season. The fact that the plowing service rendered by the governmental
extension service reaches not more than 2 percent of communal farmers implies
that most tractors hired for plowing are owned by private entrepreneurs. The fact
that the number of entrepreneurs with considerable investment capital is much
higher in Ovambo than in Kavango explains the zones’ difference in the number

of farmers that have access to mechanical plowing service.

41

A comparison of the relative cost for hired draft power indicates that hiring
of tractors for plowing at the beginning of the 1992/93 production season was
about three times cheaper in Ovambo and five times cheaper in Kavango than
using hired draft animals. This result explains the often heard request from
communal farmers that the government should extend its mechanical plowing
service or should at least encourage private entrepreneurs to rent out more
tractors.

One reason why the current demand for mechanical plowing service is not
covered by the private sector probably lies in the fact that most farmers are
holding relatively small amounts of arable land. This makes logistics at the time
when field preparation starts with the beginning of rain season difficult and drives

transportation costs high from one farmer to the next.

3.1.9. Local millet varieties and the new variety Okashana 1

Traditionally, communal farmers select their millet seed from their own
production. During the 1988189 cropping season, a new millet variety, known as
Okashana 1 was introduced by the Ressing Foundation, a Namibian foundation
financed by Rbssing Uranium Mines of Namibia. Because of the extensive
publicity about Okashana 1's high yields, government officials expect a

significant yield increase from the dissemination of this so called 'improved seed.’

Because Okashana 1 matures within three months, as compared to about

five months for local varieties, it reduces the risk of total crop failure during years

42
of short rainy seasons. Okashana 1 is also supposed to respond to chemical
fertilizer and to have a softer and more uniform kernel that eases manual and
mechanical processing.

On-farm research trials have been carried out during the 1992/93 season
to compare the performance of Okashana 1 with local varieties. At the Second
National Millet Workshop in Windhoek, November 1994, an ICRISAT scientist
concluded from the trial results that:

“...the potential gains due to improved management are much

larger compared to potential gains due to cultivar change”

(Matanyaire, 1994, p.9).

The on-fann trails demonstrated that improved field management that includes
the use of chemical fertilizer can increase hectare yields of Okashana 1 from
about 250 to 1650 kg (660 percent increase) and those of local millet varieties
from about 250 to 1150 kg (460 percent increase). Matanyaire, the leading
agronomist of the on-farrn trails concluded that the farmers in the study zones
already using fertilizer (mainly manure) may, if their management is improved
and the fertilizer is used in line with the up-coming guidelines, result in a millet
yield increase by at least 66 %.

Test results from the Mahanene research station demonstrated that under
optimal soil and field management conditions, the gap between yields of local
millet varieties and Okashana 1 narrows even more in comparison to gains from
improved management. During the 1992/93 crop season, the Mahanene station

achieved average yields of Okashana 1 at 3870 kglha and yields of local

43
varieties at 3630 kglha. Besides high doses of chemical fertilizer, further field
management practice included the use of chemical insecticides and fungicides
as well as occasional irrigation once the trails got under drought stress.

Most of the survey farmers that used Okashana 1 during the 1992/93
production season planted it only on a small portion of their fields. It was
estimated that Okashana 1 covered only 6 percent of the total area under millet
production in Ovambo and 8 percent in Kavango, respectively. Three reasons
probably caused this combined use of local varieties and Okashana 1:

(1) Farmers’ risk adverse behavior with regard to testing a yet-unknown
variety on all of their millet fields.

(2) Farmers’ goal to optimize the use of scarce labor and soil moisture by
planting the slower-maturing local varieties at the beginning of the rain season
and by planting Okashana 1 later if time is still available during weeding of the
first planted fields.

(3) Farmers’ perception about Okashana 1 as a pure cash crop and not as
crop for their own home consumption. This perception was probably caused by
radio announcements of the agricultural extension service advising farmers to
store the Okashana 1 separately from local millet varieties because it was
expected that Okashana 1's softer shell might lead to faster insect infestation
during storage. Instead of acquiring separate storage containers for Okashana
1, farmers might have opted to just grow as much Okashana 1 as they planned

to sell shortly after harvest.

44
Finally, the analysis of 1992193 millet yields from survey farmers (see
section 3.2.2. further below) reveals that Okashana 1 did not live up to the
claims of its promoters that it would double average yields from 300 to 600 kglha
on the farm level. The survey results rather indicate that local varieties out-
perform Okashana in good rainfall years and that Okashana 1 has a comparative

advantage only in areas with a very short rainy season.

3.1.10. Storage practices and losses

To prevent the infestation with insects, farmers try to keep freshly
harvested millet separate from millet that is already longer in storage. For this
purpose, the new harvest is put in clean storage containers, while older millet is
consumed first. The types of storage containers used by rural households varies
widely between Ovambo and Kavango households. In Ovambo, the majority of
households (90 percent) use large woven baskets that are made from branches
and sealed with dried clay. These storage containers can hold between one-half
and two tons of grain.

In Kavango, households use various types of storage containers and
combinations of these. The majority of Kavango households (60 percent) use
either 50 to 70 kg bags that are made from sisal or polyethylene, or they use
sealed 200 liter oil drums. The rest of the Kavango farmers use either granaries
that are lined and sealed with dried clay, or they leave the millet spadix on the

stem and store a bundle of stems inside or outside the house.

45

Households reported that millet can be stored from 1 to 8 years.
However, the average numbers of years that millet can be stored in baskets or
drums is closer to three years, and that of bags, granaries and storage on the
stem is closer to M0 years.

Grain losses through spoilage and insects seem to be of a lesser concern.
In Ovambo, 85 percent and in Kavango, even 93 percent of the farmers said
they have no or only little storage losses. Nevertheless, a third of Kavango
farmers and 82 percent of Ovambo farmers take measures to prevent insects
from infesting their millet. In both zones, roughly 12 percent of farmers use
chemicals against insects, while the rest of the farmers use traditional methods

like ash or leaves.

3.2. System performance

To assess the performance of the current millet production system, this
section concentrates on the following topics: hectares cultivated and crop yields,
grain self-sufficiency at household and zone level, returns to land and labor, and

contribution of millet production to total household income.

3.2.1. Hectares under cultivation

Field sizes of almost all sample households have been measured with the
assistance of the agricultural extension service . Due to limited resources, only
the fields that were cultivated during the 1992/93 production season were

measured. Pearl millet was the crop that covered most of the cultivated land. Of

46
the households that produced also sorghum and/or maize, the majority
intercropped these grains with millet. Because crops like beans cover relatively
little space, no additional measurements have been conducted.

During the 1992/93 cropping season, Ovambo households cultivated, on
average, 2.9 hectares (CV=0.87 and 0.41 hectare per adult-equivalent) of grain
with a median of 2.5 hectares. Kavango households cultivated on average 3.0
hectares (CV=1.58 and 0.48 hectare per adult-equivalent) but had a lower
median at 2.0 hectare.

As demonstrated in Table 3-2, the distribution of farm households by the
amount of land they cultivated with grain indicates that the largest portion of
farmers produce grain on 1 to 5 hectares of land (Ovambo: 76 percent, Kavango:
36 percent). However, the percentage of households producing less than one
hectare is also relatively large in both zones (Ovambo: 14 percent, Kavango: 19

percent), while very few farmers produce grain on more than 5 hectare.

Table 3-2. Household distribution by hectares of cultivated land in
1992193, by zone (in percent)

 

 

Hectares of Grain Cultivated
0.1-1 1-2 2-5 5-20 20+
Ovambo 14 20 56 9 1
Kavango 19 32 34 15 0

 

Source: Namibian Millet Subsector Project. 1992193

47
The calculation of the average number of hectares cultivated per adult
equivalent for each of the above presented hectare categories shows that
households cultivating less land tend also to have less land available per

household member. (Table 3—3.)

Table 3-3. Hectares of grain cultivated per adult equivalent. by
hectares cultivated per household and zone
(standard deviation in parentheses)

 

Hectares cultivated per household

 

0.1-1 1-2 2-5 5-20 20+
Ovambo 0.19 0.31 0.46 0.98 9.35
(0.18) (0.17) (0.23) (0.35) (8.27)
Kavango 0.15 0.28 0.57 2.24 na
(0.08) ((0.14) (0.27) (1.26) na

 

Source: Namibian Millet Subsector Project, 1992193

3.2.2. Millet yields

The yields of the 1992/93 production season were very low in most areas
of the study zones. In eastern Ovambo and whole Kavango millet yields were
less than 100 kglha. Only in western Ovambo yields averaged 590 kglha and in
central Ovambo still 240 kglha.

One major cause for the poor yield performance in eastern Ovambo and
in all of Kavango was the late arrival of rains during the 1992/93 survey year. In
these areas the short production season especially affected the performance of

local varieties that need about five months to mature. The average yields of

48
local varieties ranged only between 68 and 96 kglha. By contrast because of its
faster growing potential, Okashana 1 performed relatively better in these areas,
ranging from 127 to 235 kglha.

In the areas of western and central Ovambo with better rainfall conditions
during the 1992/93 production year, local varieties clearly outperformed
Okashana 1. The promoters of Okashana 1 claimed that it would double
average yields from 300 to 600 kglha on the farm level. However, due to very
good rainfall distribution in western Ovambo, local varieties yielded, on average,
644 kglha, while average yields of Okashana 1 were only about 234 kglha. In
central Ovambo, Okashana 1 yielded, on average, only 106 kglha compared to
local varieties’ average yield of 259 kglha. The main explanation for the better
performance of local varieties in good rainfall areas is their slower growth rate.
This permits a longer period of time than Okashana 1 to make use of the

available soil moisture and the plant nutrients.

Table 3-4. Millet yields of the 1992193 production season,
by subzone and variety (in kglha)

 

 

Ovambo Kavango
West Central East West Central East
"°°.a' . 644 259 70 68 78 96
Varieties
Okashana 1 234 106 127 235 148 n.a.

 

Data Source: Namibian Millet Subsector Project, 1992193

49

The two main reasons explaining why Okashana 1 did not reach the
suggested yield level of 600 kglha in western Ovambo are (1) the shorter growth
period that does not permit the same long use of available soil moisture and soil
nutrients than is possible for local varieties, and (2) that the cropping and soil
conditions under which Okashana 1 was tested to simulate farm level conditions
were still better than those actually occurring in the field; i.e., the achieved yields
of 600 kglha suggested and published by ICRISAT (ICRISAT, 1991) were not
based solely on a cultivar change.

The reason why Okashana 1 performed better than local varieties in
Kavango is that local varieties perform very poorly under short rainy seasons ‘
because they can not reach full maturity. The Okashana 1 can mature within 3
months; however, it needs higher soil fertility and moisture to perform better than
local varieties.

Comparing their 1992193 harvest with results from previous years most
farmers from Western Ovambo (81 percent) categorized their 1992/93 harvest as
“good” while majority of the farmers from central Ovambo (53 percent) classified
their harvest results as “middle”. Seventy percent of the farmers from eastern
Ovambo and Kavango perceived the 1992/93 harvest as “poor" in comparison to
previous years.

Of those farmers that categorized their millet harvest as “good", the
average yield was close to 600 kglha. Those farmers that categorized their
results as “middle” harvested, on average, about 250 kglha, while those who

called their results “poor” harvested, on average, 125 kglha.

50
The yield increase eventually used to simulate the application of an
improved technology package in the three spreadsheet models are based on
Matanyaire’s research results from on-fann trails. Matanyaire, ICRISAT advisor
to the Namibian government reported:

“ a quarter of the pearl millet farmers already using fertilizer (mainly
manure) may, if their management is improved and the fertilizer used
in line with the up-coming guidelines, result in their pearl millet grain
yields being increased by at least 66 %. Yield gains of up to 200 %
are also possible” (Matanyaire, 1994, p. 9).

3.2.3. Grains self-sufficiency levels

Estimates of communal farmers’ grain production have only existed since
Namibia achieved independence in 1991. According to these estimates, the
aggregate millet production of the 1992/93 year comes close to the five year
average of the production years 1990191 through 1994195. (Table 3-5) However,
household level data indicate that during the 1992/93 production season, very
few farmers produced enough millet to cover their grain need over the period of
one year.

On the basis of FAO recommendations regarding energy requirements for
different gender and age groups (FAO, 1974) the annual grain needs of survey
households were calculated. According to these calculations the average
Ovambo household needs 1690 kilogram of grain per annum while the average
Kavango household needs 1490 kilogram. (Table 3-6) With the exception of
households from western Ovambo that received extraordinarily good rainfall

during 1992193, no other area within the study zones had more than 8 percent of

51

 

 

 

 

 

 

 

Table 3-5. Aggregate grain production in Ovambo and Kavango
from 1990191 to 1994195
Ovambo Kavango
Season ‘000 kilogram ‘000 ‘000 kilogram ‘000
ha per mt ha per mt
hectare hectare
1990191 157 350 55 8 300 2
1991/92 150 100 15 12 100 1
1992193 143 240 34 17 260 4
1993194 272 200 54 22 420 9
1994195 275 120 34 12 210 3
Average 200 200 39 14 258 4
Data Source: Namibian Early Warning 8. Food Information System;
Crop Assessment Report, March 1996
Table 3-6. Household distribution by grain self-sufficiency levels in
1992193, by zone (in percent of households)
Grain Self- Ovambo Kavango
Sufficiency West Central East West Central East
less 30 % 20 65 88 69 69 64
31 to 70% 16 15 11 22 18 25
71 to 100 % 9 13 2 3 8 3
101 to 130 % 6 2 0 0 2 3
131 % plus 49 5 1 6 3 5
100 100 100 100 100 100

 

Data Source: Namibian Millet Subsector Project, 1992193

52
households reaching grain self-sufficiency. On average, 90 percent of all
communal households reached grain self-sufficiency levels that were below 70
percent. And 70 percent of all households reached grain self-sufficiency levels
that were below 30 percent.

To estimate the aggregate grain self-sufficiency levels of different areas,
the grain self-sufficiency levels of individual households were calculated for the
production year 1992/93 and than aggregated by production area. Due to
unusual good rainfall western Ovambo reached an aggregate grain self-
sufficiency level of 165 percent. The other production areas reached only grain

self-sufficiency levels up to 36 percent. (Table 3—7)

Table 3-7. Household average grain consumption requirements and
production during the year 1992193 (in kg per household)

 

 

 

Ovambo Kavango
West Central East West Central East
Requirement 1490 1780 1790 1550 1320 1600
Production 2460 640 270 340 380 520
Balance 970 -1140 -1520 -1210 -940 -1080
Self-sufficiency 165% 36% 15% 22% 29% 33%

 

Data Source: Namibian Millet Subsector Project, 1992193

The average contribution of millet to households’ total annual grain

production ranged between 83 and 91 percent across different production areas.

53

With 9 to 12 percent, sorghum was the second important grain in Ovambo, while

maize was the second important crop in Kavango, with 10 to 15 percent.

 

 

 

(Table 3-8)
Table 3-8. The share of millet on total household grain production
during the year 1992193 (in percent)
Ovambo Kavango
West Central East West Central East

Millet 89 91 88 83 84 86
Sorghum 1 1 9 12 2 6 4
Maize 0 0 0 15 10 10

100 100 100 100 100 100

 

Data Source: Namibian Millet Subsector Project, 1992193

3.2.4. Cost and returns

From a farm management perspective, millet production provides not only

part of farm households’ food requirements but also uses scarce household

resources. To better understand the cost and returns of the millet enterprise,

millet’s unit cost of production, its returns to land, and its returns to labor have

been calculated for households from Ovambo and Kavango and for households

that reached grain self-sufficiency for one year from the 1992/93 production and

those households that did not. (Table 3-9)

 

 

 

Table 3-9. Cost and returns of millet production during the year
1992193, by zone and households’ grain self-sufficiency
Ovambo Kavango
(n=200) (n=120)
Grain Self-Sufficient yes no yes no
Cost and Retums
Gross retumlvalue" N$ 2186 884 1871 394
Variable cost” N$ 1 152 918 1490 646
Net retumlvalue*** N$ 1034 -34 381 -252
Cost per unit N$lmt 580 1 142 1012 2083
Return to land**** N$lha 414 -13 100 -48
Return to labor N$1day 10.30 4.78 6.86 2.29

 

Source: Namibian Millet Subsector Project, 1992193

Q.

“i

O...

Millet was valued at informal market prices in Ovambo at 1.10 NSlkg and in Kavango

at 1.27 N$1kg.

Household labor that contributed roughly 85 percent of the total variable cost in Ovambo
and 70 percent in Kavango was valued at the current agricultural wage rate of NS 5 per day.
In the case of negative net retumslvalues valuation household members worked at
opportunities cost below the current wage rate of NS 5 per day.

The return to one hectare of land was calculated by dividing households’ net return/value by
the number of hectare cultivated per household.

55

At current production practices, the major cost factors of millet production
are household labor (Ovambo: 85 percent; Kavango: 70 percent), plowing
service (Ovambo and Kavango: 14 percent), hired labor (Ovambo: 1 percent;
Kavango: 13 percent), and seed (Ovambo: 1 percent; Kavango: 3 percent).

To calculate millet production cost household labor was valued at NS 5
per day, the average wage rate for agricultural labor in the study zones during
1992/93. The costs for hired labor and for plowing service were directly used
from sample household statements about their respective payments. Seed cost
was estimated by assuming an average need of 2 kg of millet per hectare valued
at informal market prices. The cost for manure was not directly included
because no market value exists and most farmers who used manure used it from
their own livestock. However, the amount of labor used for manure collection
and distribution on the fields was captured by the data about household and
hired labor spent for crop production. Due to the lack of price data of chemical
fertilizer and the exact amounts used by farmers, cost for chemical fertilizer could
not be included into the cost calculation.

In both zones, grain self-sufficient households had higher total variable
costs but lower unit cost of millet production than households with grain deficit.
While grain self-sufficient households’ higher variable costs are mostly explained
by their higher labor input, their lower unit costs can be attributed to significant
higher yield levels. The high difference in unit production costs between
Ovambo and Kavango is also caused by the different yields reached in the study

zones.

56
The calculation of returns to one day of household labor indicate that only
grain self-sufficient households generated more value through their own millet

production than they could earn by working on others’ fields.‘

3.2.5. Contribution to annual household income

Of the rural labor force (adults of 15 years and older), only 38 percent in
Ovambo and 65 percent in Kavango have crop production and/or livestock
herding as their principal employment. (Table 3-10) The relatively low
percentage in Ovambo is mainly caused by high proportions of young adults
(15 through 30 years) that attend school or vocational training since
independence (Ovambo: 62 percent, Kavango: 40 percent).

Major income sources for farm households in the study zones, besides
crop production, are migration employment, government pensions, local
employment, and livestock herding. Nearly half of all households in the study
zones earn income from migration work (Ovambo: 47 percent, Kavango:

46 percent). The annual income (consumption and cash income) of households

with migration employment is, on average, five to six times higher than the

 

‘ It is not uncommon in developing, as well as in developed, countries
that farm households accept returns to labor that are below the average wage
rate for hired labor. This kind of “self-exploitation" occurs especially often when
household members’ opportunity costs (internal opportunity costs) are below the
market‘s, e.g., in cases when their access to labor markets is constrained.

When farmers' returns are below, but close to, existing wage rates, they often
prefer to stay on the farm because they prefer their “independent lifestyle” and
their closeness to nature. Also the holding of land is often seen as an additional
security for which farmers are willing to forge a certain amount of higher income.

57

income of households without migration employment. (Table 3-11)

Table 3-10. Occupational distribution of the rural labor force, by zone and
gender“ (in percent of persons older than 14 years)

 

 

 

Ovambo Kavango
Female Male Female Male
(n = 433) (n = 598) (n = 236) (n =280)
‘Occupation
Farmer 45 29 75 52
Student 35 35 17 26
Employee" 13 24 6 19
Retired/Unemployed 7 12 2 3
Total 100 100 100 100

 

Source: Namibian Millet Subsector Project, 1992193

' The percentage calculation is based on data gathered about all adult members of the
households covered by the household surveys.
“ Includes both, local self-employment and migration employment.

Although the per capita grain production of households with migration
employment is higher than that of households without off-farm employment, the
contribution of grain production in percent of their total annual income is much
lower (Ovambo: 14 percent, Kavango: 6 percent). But with a share of 44 percent
in Ovambo and 41 percent in Kavango millet production is the most important
income source for households without migration employment. For those
households, the second most important income sources are government pension

payments to older people in Ovambo and local self-employment in Kavango.

58

Table 341. Average household income and shares of income sources, by
zone and employment status of household (in US Dollar“ and
percent of total annual income)

 

 

 

 

Ovambo Kavango

(n=199) (n=118)
Households with yes no yes no
Migration Income 47% 53% 46% 54%
Annual Income in US-S
- per Household 5310 890 3,880 680
-per Capita 910 110 710 100
Income Sources in %
- Grain Production 14 44 6 41
- Livestock Herding 1 4 1 3
- Local Employment" 3 9 9 37
- Migration Employment 76 0 79 0
- Government Pension 6 43 5 19

 

Data Source: Namibian Millet Subsector Project, 1992193
' During 1993 the exchange rate of the Namibian Dollar averaged at 3 NSIUSS
” Local employment is mainly based on the following occupations: laborers mainly working for the government (Ovambo

3.3 %, Kavango 3.0 96), teachers (Ov. 2.9 %, Kv. 2.8 96), skilled workers (Ov. 3.2 %, Kv. less than 1.0 96), police and
amied forces (both zones 1.2 it), and health support staff (Ov. 1.7 %, Kv. less than 1.0 %).

-ul-

59

Two major questions flow from these data: What explains such high
income discrepancies between those households with and those without off-farm
employment? Second, why are households thateam off-farm income engaged in
millet production instead of purchasing readily available maize meal in the
market?

The answer to the first question is based on the influence of the apartheid
system on the communal labor market. The communal areas have been treated
as reserves of cheap labor for the mines and other industries in Namibia and
South Africa. It was forbidden to leave one's own communal area, which was
formally called “homeland”. Only males that applied for a work permit and that
were classified in a category suitable for certain employment in demand were
granted a pass that allowed them to leave their area for a certain period of time
and migrate to a pre—defined employer. Those not allowed to leave work had to
find local employment opportunities.

It is not difficult to assume that, although free movement and equal
employment opportunities are officially granted to everyone since independence
in 1991, many of the old employment structures persisted. Especially those who
were formerly excluded from work outside their living areas face the frustration

that the decline in Namibia's mining sector after independence led to a significant

reduction in employment opportunities for industrial labor.

60

Chapter 5, which deals with grain consumption issues, will answer the
question why households with significant off-farm employment still pursue the
production of millet in more detail. In short, millet is preferred among the majority
of Ovambo and Kavango people over maize. Even many of those who have
been extensively exposed to maize meal while being migrant workers in the
south still prefer millet because of its distinct taste. Because until recently, millet
was not seen as a cash crop and therefore not available for purchase in the food
market, migration workers who wanted to eat millet after their return from
seasonal employment had to rely on the production from their own household.

The other reason why households still pursue the production of millet
although maize is readily available in the commercial food market is that South
Africa was not interested in introducing alternative income opportunities and/or
improved production crop technologies in the homelands.

In Table 3-12 the social poverty map for Ovambo and Kavango
demonstrates that rural poverty is highly differentiated in the study zones and
that various approaches are needed to combat it. It shows that in Ovambo
33 percent of rural households and in Kavango 51 percent cultivate less than

2 hectare of land and that their agricultural income is on average less than a
third of total household income. This is an indication that rural poor in the study
zones cannot be assisted by increasing the productivity of land use alone. It
means that development programs must be directed towards the various income

Opportunities that characterize the different categories of rural households, and

5:8 m: one .8 5.00 53:52 8.5 H 0.8 3:298 89

nmmmmw .aoloi 833m «0...: co§Eoz H350m 800

 

 

 

 

 

 

 

. one new 0mm 0mm
- omvd owmé ooo.~ ooo.~
Lo=00 .m.: nmmw
- 3. 3 m N?
. Fm an we on
.. vm 0N VN mm
- Fm _.N mu mN
.. mm he mv mm
. m m _. _.
I o o N o
.. 2. 2. 3 PM
. NF 9 2 2
. m0 N0 on mm
- 3. Q. ow F v
- E 00 mm on
- no em mm mm
0.0 0.3 Ndm mém 5dr
«canton.
+6" ourm mum Nip de
OOz<><¥

can one can now new case so

2.? 8% 8m...” 23 25 29.33.. on.
2:02... «02

5.8 as 82
on 8 mm on em 8.23.
9 S 9. en em 9:8... ones
a v e s 3 32:95..ch -
no em 8 B an 9.2.8.5... -
so «a R an 9. 38... Emu.
0:30.... «o ooocnom
o m a m a 235.85..
0 n P m m EoE>oEEoc3
o 3. mm nu cm 3.6038
c 3 a. 2 2 55.635 E8. to
2: mm 9. 9. 9. to; sea“.
5330.? Loan. ==u<
on mm mm m o cases .222
2 8 8 mm mm 58E
2 an we 8 E 528)...
o~omu<
3 as «.8 as. on. 3.2.33:
Eoflod
+3 8-... m.“ «a E... s. e. .23. 3.23.5
om2<>o

 

can. 2.32:3 no on.» .3 omega: use onEo>O Lou onE Egon _o_oom .N Tn Ens...

62
that these programs must look beyond the farm towards employment creation in

rural-based non-farm activities linked to agriculture.

3.3. Determinants of millet producers’ performance

In this section, three methods are used to identify the main determinants
of farm household grain production. First, the results of the analysis of attitudinal
data about the driving forces of millet production are described. Second, two
regression models are presented estimating (a) the number of hectares of grain
cultivated per household, and (b) the grain yields per hectare achieved. Third,
correlation analysis is used to investigate how various farmers' characteristics

are related to the main determinants of millet production.

3.3.1. Descriptive analysis
During the millet subsector household surveys, farmers were asked to
identify problems they expected and experienced during the 1992/93 millet
production season. Before field preparation in December 1992, "lack of draft
animals" caused by the severe 1991/92 drought was the most often expected
problem mentioned by Ovambo farmers, and the "continuation of the serious
drought from the previous production year 1991192" was the most often feared
problem of Kavango farmers. .
After the field preparation period in April 1993, farmers were asked about
the problems that actually occurred during field preparation. It revealed that

most farmers focused now less on the initial cause, drought and late rains, but

63
on those factors that kept them from effectively preparing their fields during the
short time-window after the rains had eventually started. In retrospect, farmers
in both study zones experienced shortages of labor, draft power, plowing
equipment, and seed, and had problems in getting the promised plowing
services from the agricultural extension service.

The figures in Table 3-13 demonstrate that Kavango farmers had
relatively more problems with field preparation under severe time pressure. The
reason for this discrepancy is that Kavango farmers are commonly used to
starting their field preparation in mid-November soon after the arrival of the first
rains.‘

At the end of the field preparation period in April 1993, most farm
households had not cultivated as much land for grain production as they had
intended initially (Ovambo: 74 percent, Kavango 96 percent). The reasons given
by the majority of farmers for not having reached their planting goals were:

(1) lack of land, (2) lack of family or paid labor, (3) lack of seed, and (4) lack of
draft power.

In Ovambo, the most severe constraints of cultivating adequate millet are:
(1) limited arable land and (2) a shortage of seed. While the first is more a

concern of farmers with little cash income and crop land less than 5 hectares, the

latter is more of a concern for farmers that earn migration income and have land

 

‘ The fact that the millet cropping season of Ovambo is about one-and-a-
half months shorter than the one in Kavango explains also why so many more
tractors are used in Ovambo for plowing compared to Kavango.

Table 3-13. Field preparation constraints during the year

1992193 (in percent of farmers")

 

 

OVAMBO (n = 200) KAVANGO (n = 120)
CONSTRAINTS expected actual expected actual
”9” °‘ "“9 11 19 4o 14
rains
labor shortage 7 7 6 42
lack of draft 23 14 7 36
power
'ac". °‘ 5 24 7 25
equrpment
problems with
plowing service 1 16 0 13
lack of seed 6 10 3 6

 

Data Source: Namibian Millet Subsector Research Project, 1992/93

* The percentage figures do not add up to 100 percent because sample
farmers could indicate more than one problem; i.e., the percentage
figures represent the number of farmers raising a certain concern.

holdings larger than 5 hectares. In Kavango the most severe constraints of

cultivating adequate millet are: (1) Lack of family labor, the most often stated

constraint affecting especially farmers (a) that do not earn migration income and

are, therefore, unable to balance their shortage by hiring field workers and (b)

that cultivate more than 5 hectares of land. (2) Seed shortage which affected

especially farmers with land holdings less than 5 hectares. This is probably due

to the fact that smaller farmers were more likely forced by the severe drought of

65
the previous season in 1991/92 to eat their last grain reserves. (3) Lack of draft
power was mainly a concern of farmers that cultivated more than 5 hectares.

Finally, a third of all Kavango farmers, independent from their cash
earnings and the size of their land holding, claimed they had problems in gaining
access to more land. This seems to contradict the statements made on previous
sections about Kavango’s abundance of arable land compared to Ovambo.
However, general availability of land is a necessary, but not sufficient condition,
for land accessibility.

Table 3-14 indicates that many Kavango farmers could not cultivate
enough land (access land) because they were short of labor and draft animals.
The other component of inadequate accessibility to land has to do with the fact
that shifting cultivation practices used in Kavango led to field locations far away
from farmers’ homesteads. This causes problems in transportation and travel
time.

In Ovambo, seed shortages and limited access to land were mayor
problems for many of those earning cash income and/or those with larger land
holdings. Those Ovambo farmers with cash income sources complained also
about the lack of field workers to hire. Those with larger field holdings had
problems with the cost of hiring draft power for plowing.

In Kavango, farmers with cash income from migration employment have
problems with accessing arable land, the availability of seed, and lack of family
labor. Although they have money to pay, many others complain that hired labor

is too expensive. Among those farmers that have larger land holdings, lack of

 

 

66

Table 3-14. Land cultivation constraints during the year 1992193,
by zone, migration income, and number of hectare
cultivated (In percent of farmers“)

 

 

 

 

 

 

OVAMBO KAVANGO
Households earning migration income
Percent share on all sample yes no yes no
farmers: 46 54 43 57
Cultivated not enough: 78 70 96 97
Constraints:
Limited land access 27 << 36 ‘ 31 31 -
Seed shortage 28 >> 16 33 34
Labor shortage
- lack of family labor 1 1 10 35 < 42
- lack of labor to hire 14 >> 4 10 9
- cost to hire labor 9 7 22 23
Lack of draft power
- lack of draft power
- cost of draft power 4 O 16 << 27
- government ploughing 8 6 12 14
service inadequate 7 3 6 6
Households cultivating more than 5 ha
Percent share on all sample yes no yes no
farmers: 8 92 12 88
Cultivated not enough: 75 73 100 96
Constraints:
Limited land access 19 << 33 29 31
Seed shortage 19 22 14 << 36
Labor shortage
- lack of family labor 0 << 13 43 38
- lack of labor to hire 0 4 14 9
- cost to hire labor 6 9 21 > 13
Lack of draft power
- lack of draft power 6 3 29 > 21
- cost of draft power 13 > 6 29 >> 11
- government ploughing _ 6 5 14 > 6

service inadequate

Data Source: Namibian Millet Subsector Research Project, 1992-93
' The percentage figures do not add up to 100 percent because sample farmers could
indicated more than one problem. i.e., The percent figures represent the number of
farmers raising a certain concern.

 

67
family labor is by far the most often mentioned constraint, while lack of own draft

power and high cost of hired draft power are following suit.

3.3.2. Regression analysis of millet area cultivated

The above analysis of attitudinal and descriptive data indicates that
farmers' millet production is influenced by a complex set of factors. The following
two sections attempt to test the significance of the effects of these factors
through econometric techniques. Due to the lack of historical information on key
variables affecting the land allocation patterns of households in the survey
sample, the estimated regressions are based on cross-section data gathered
during the millet subsector household surveys of the production year 1992/93.
The two main explanatory variables of grain production are: (a) the number of
hectares of millet planted and (b) the yield per hectare achieved by farm
households, are estimated in two separate equations and two consecutive
sections.

The cross-section regression model of the millet area planted by farm
households during the 1992/93 production season encompasses variables that
identify (a) the expected returns to resource investment in millet production and
(b) the production resources available to households. Because millet is the only
important cash crop in the study zones, relative returns to alternative cropping
enterprises are ignored. To identify differences in the magnitude of effects from
the independent variables in the two study zones, the area of millet cultivated is

estimated separately for Ovambo and Kavango.

68

In the following, the regression equation used for the amount of area

cultivated by rural households in the 1992/93 production year is presented

together with a brief description of its variables. The variables were all tested for

correlation and had met the test:

MILLET HA I f (MILLETPRICE, LANDCONSRT, HHLABSHRT, ADLEQUWRKDS,
DRFTANIMPLOW, PLOWHOE, PLOWTRAC, EXPOKASH, CASHINCOME)

where:

MILLETHA
MILLETPRICE
HHLABSI-IRT
ADLEOUWRKDS
DRFTANIMPLOW
PLOWHOE
PLOWTRAC
EXPOKASH
CASHINCOME

= area of millet planted (HA)

= local price for millet at the begin of planting (NS/KG)‘

2 household experienced labor shortage (DUMMY, 1 = yes))

. adult equivalent workdays spent for grain production (No. in '00)
= ownership of draft animals and plow (DUMMY, 1 = yes»

= use of has as sole plowing equipment (DUMMY, 1 = yes)

= use of tractor for plowing (DUMMY, 1 a yes)

= experience with Okashana 1 before 1992193 (DUMMY, 1 = yes)

= annual cash income from non-cropping activities (NS ’000)

Table 3-15 below presents both the averages and standard errors of the

continuous variables and the percentage distribution of the binomial variables

included in the regression model.

 

‘ 'Local' means that representatives of each survey household stated the
millet price they faced personally during field preparation time. In cases where
respondents could not supply a millet price, the average of the price data stated
by the sample farmers from the same community was substituted.

69

Table 3-15. Characteristics of the millet hectare regression
variables, by zone

 

 

OVAMBO (n = 114) KAVANGO (n = 50)
VARIABLE MEAN S.E. MEAN S.E.
MILLETHA (HA) 2.89 1.89 3.02 3.33
MILLETPRICE (NS/KG) 1.46 0.97 1 .49 0.21
LANDCONSRT (1 8 yes) 76 percent 8 1 92 percent = 1
HHLABSHRT (1 - yes) 10 percent = 1 24 percent = 1
ADLEQUWRKDS (No. of '00) 1.39 1.19 0.79 0.77
DRFTANIMPLOW (1 = yes) 30 percent = 1 36 percent = 1
PLOWHOE (1 . yes) 18 percent = 1 0 percent = 1
PLOWTRAC (1 = yes) 36 percent 8 1 12 percent = 1
CASHINCOME (NS ’000) 7.65 12.10 6.34 7.22

EXPOKASH (1 I yes)

43 percent 8 1

14 percent = 1

 

Data Source: Namibian Millet Subsector Research Project, 1992193

The independent variables used and the reasons for their inclusion in the

estimated equation are as follows:

(a) the local price for millet at the beginning of planting. (MILLETPRICE) In

the study zones, crop production is mostly carried out to provide food for

farmers’ household consumption. However, because 22 to 23 percent of the

interviewed farmers from Ovambo and Kavango identified millet as their most

important cash crop, it can be assumed that high millet prices at planting time

encourage farmers to cultivate more millet area for the market .

70

It is expected that the response to millet price changes will be stronger in
Kavango because in Ovambo only 30 percent of all interviewed farmers stated
they sell millet in some years. In comparison, 54 percent of the interviewed
Kavango farmers stated they sell millet in some years, and an additional 18
percent said they sell millet every year.

Also, those households that do not sell grain have more incentive to
cultivate lager millet fields if millet prices in their community are higher. Through
the cultivation of larger millet fields, these households try to prevent grain deficit
production that would force them eventually to purchase high priced millet from
their neighbors. To account for the effect that local producer and consumer
prices of millet have on the amount of land cultivated, the average of millet
prices stated by the sample farmers within one survey community at the
beginning of field preparation in December 1992 (MILLETPRICE) is included as
a determinant of the number of hectares produced per household. It is
hypothesized that the area of millet under cultivation is positively related to the
variable MILLETPRICE.

It can be assumed that the millet prices faced by farmers are dependent
on factors like local grain supply, price of major substitutes, distance to next
shop, roads, market, etc. Some of these various factors have been tested
regarding their relation to millet prices at the beginning of the field preparation
period. The tests indicate that two factors, geographic location of survey site
and maize meal prices are strongly correlated with millet consumer prices.

Figure 3-3 demonstrates the relations between millet prices, location of survey

 

71
sites, and maize prices by displaying both observed millet prices in geographical
order from west to east and the estimated regression line based on the following
variables: observed maize meal prices, ordinal numbers of survey sites (ordered
from west to east), and the square of the latter. The adjusted R-Square of this

regression was 0.38.

Figure 3-3. Millet prices regressed by maize prices and location (prices
from the 1992193 field preparation period)

 

1.2

A) j
3.. ' .’ l ’

(adjusted\ r-square 0. 38)

.. \/\A \ /\\ /
.,. w 57A

observed millet prices

I.s rogroeslonllne //// \\/ /

1.2

millet kilogram prices In December ‘02

 

 

1.9

 

ourvoy oltoo In geographical order from woot to ooot

Data Source: Namibian Millet Subsector Research Project. 1992193 and my own calculations

(b) the access to fertile and uncultivated land (LANDCONSRT). Within
communal farming areas, various factors influence farm households’ access to
uncultivated land. Traditionally, tribal chiefs assign land to individual heads of
households for payments in kind, and today also in the form of money. The

amount of payment is usually negotiated and depends to a large extent on the

CU

the

72
relationship between the head of household who wants to acquire land and the
local authority. Between 75 and 90 percent of the interviewed farmers stated
they face problems expanding their fields. In Ovambo, the most often mentioned
obstacles with regard to field expansion were “scarcity of arable land nearby”
and “lack of equipment” to clear land. In Kavango, the most frequently stated
problem claimed was lack of money to hire labor for de—bushing and clearing
new land.

To control for the effects of “lack” of equipment on the amount of
cultivated land, the three variables DRFTANIMPLOW, PLOWHOE, and
PLOWTRAC are included in the regression further below. To estimate the effect
of the residual problems farmers have with field expansion, the dummy variable
(LANDCONSRT) is included in the equation. The variable takes the value one if
representatives of a household stated that it has difficulty clearing new land. The
coefficient of the variable LANDCONSRT is expected to be negative as a result
of the negative effect that these problems have on the amount of hectares under
cultivation.

(c) farm households’ status of labor availability for crop production.
(HHLABSHRT) Within a farm household, the number of members available for
crop production can vary significantly over time; eg, if household members find
seasonal off-farm employment or if children and young adults that usually help
with fieldwork are admitted to distant boarding schools, household capacity to
cultivate land is significantly reduced. To account for the effect of labor scarcity

that individual households faced during the 1992/93 grain production season, a

73
dummy variable (HHLABSHRT) is included in the regression model. The
variable takes the value one if farmers stated that labor shortage prevented them
from plowing the amount of land they had intended before the actual start of the
1992193 production season. The coefficient of the variable HHLABSHRT is
expected to be negative as a result of the negative effect that labor scarcity has
on the amount of hectares under cultivation.‘

(d) the number of workdays household members spent for grain
production during the whole production season. (ADLEQUWRKDS) Even if
households are capable of plowing a larger number of hectares, labor scarcity for
tasks like weeding, prevention of crop damage from insects and birds,
harvesting, and threshing might limit the amount of land eventually cultivated. To
account for the effect that the use of household labor for the total millet
production process has on the amount of land cultivated, the estimated number
of adult-equivalent workdays that households actually spent for grain production
during the 1992/93 production season (ADLEQUWRKDS) is included as a
determinant of the number of hectares cultivated per household. The number of
ADLEQUWRKDS per household is based on farmers’ statements about the start

and end of the various phases of millet production (field preparation, weeding,

 

‘ To test whether labor shortage (HHLABSHRT) itself is caused by the
fact (a) that households earn income from migration employment or (b) that
households live in a certain zone, simple correlation tests are carried out. The
Pearson's coefficients indicate that households’ statements about labor shortage
for fieldwork are (a) negatively correlated with their annual migration income in
'000, albeit, at a non- significant level (coef.= -.0134; p=0.816; n= 304) and (b)
positively correlated with the Ovambo zone at a 89 percent confidence level
(coef..= .0923; p= .108; n= 304).

74
harvesting, threshing) and the number of hours individual household members
worked, on average, per week on crop production during these production
phases. Regardless of sex, all workers below 15 years of age or older than. 55
years of age were attributed a weight of 0.5, while all workers of an age between
these two categories were given a weight of one. The breakdown and weighing
by age was based on the conventional equivalence system used by agricultural
research institutes that measure farm labor inputs in semi-arid tropics. It is
hypothesized that the area of millet under cultivation is positively related to the

variable ADLEQUWRKDS.‘

(e) farm households' ownership of a full set of draft animals and a plow
(DRFTANIMPLOW). The majority of the sample farmers used draft animals for
plowing during the 1992/93 cropping season. However, only farmers with a full
set of draft equipment/animals are assured of the timely availability of draft
Dower when rains start and plowing can commence. The farmers that hire draft
Power usually face a reduced amount of time for plowing because they have to

Wait until the owners of the animals have finished the preparation of their own

 

 

‘ To test whether the amount of workdays households spent for grain
prodl-I<:tion corresponds to the amount of labor available to them, a simple linear
step regression is carried out. In this regression, ADLEQUWRKDS is the
dePendent variable, while the independent variables are (a) household members

HHWIB), (b) household adults (HHAD), and (c) household adult equivalent

(HHAE). The variables HHMB and HHAE were excluded from the model

becatIse they yielded insignificant results (p= .62 and p= .59 respectively).
HOWever, the number of adults per household (HHAD) is strongly positively
mated to the number of workdays spent for crop production. It is estimated that
an additional adult increases the number of workdays by 18 (coef. 18.27,

F 0000, n= 317).

75
fields. It is hypothesized that less time for plowing, in combination with the
relative high cost for hiring draft animals has a negative effect on the amount of
land cultivated by those hiring draft animals. To account for the described
effects, a dummy variable (DRFTANIMPLOW) is included in the regression
model. The variable takes the value one if farm households own a full set of
draft animals with a plow. The coefficient of the variable DRFTANIMPLOW is
expected to be positive as a result of the positive effect that the ready availability
of animal draft power has on the amount of hectares cultivated with grains.

(f) the type of plowing equipment used that is different from draft animals
with a plow (PLOWHOE and PLOWTRAC)‘ . During the 1992/93 field
preparation period, about 20 percent of the surveyed households from Ovambo
and 4 percent of Kavango cultivated their fields with a hoe. On the other hand,
about 30 percent of the sampled households from Ovambo and 4 percent from
Kavango hired plowing services from private tractor owners. Accordingly, the

tW0 dummy variables PLOWHOE and PLOWTRAC are indicators of the type of
Plowing equipment farmers use instead of or in addition to plowing with draft
animals. PLOWHOE takes the value one for households that used solely the

hoe for field preparation and zero otherwise. PLOWTRAC equals one for

\

 

‘ It has to be noted that the ownership of a full set of draft animals and a

plow (DRFTANIMPLOW) does not mean that farmers did not use the hoe
(PLOWHOE) or did not hire a tractor. (PLOWTRAC) It is more an indication that
such ownership allowed farmers to start field preparation in time independently
“0"" the availability of other equipment. Accordingly, the three variables DRAFT,
PLOWHOE, and PLOWTRAC are not mutually exclusive options of field
Preparation technologies and do not require an n-1 dummy matrix to avoid

statistical problems.

76
households that used mechanical plowing service at least for part of their land
and zero otherwise. Due to insufficient cases of farmers from Kavango using
solely the hoe for field preparation, the variable PLOWHOE is excluded from the
regression of the millet area of Kavango. The coefficient of these variables
represents the marginal contribution of the corresponding type of plowing
equipment used. The coefficient estimated from PLOWHOE (used solely the
hoe) is expected to be negative because field preparation by hand is more labor
intensive and more time-consuming than plowing with draft animals. The
coefficient estimated from PLOWTRAC (use of a tractor) is expected to be
positive because mechanical plowing tends to be less labor intensive and less
time-consuming than plowing with animal traction. Thus, mechanical plowing
helps to overcome (1) the constraints arising from a short window of time for field
preparation, and (2) households‘ labor shortage during this particular period of
time.

(9) farmers' experience with the millet variety Okashana 1 (EXPOKASH).
Because the newly introduced variety Okashana 1 requires the same spacing as
local varieties, its use cannot explain household differences in the amount of
land under millet cultivation. However, farmers that cultivate more land are more
likely to have tried out Okashana 1, because of two reasons: First, for many
subsistence farmers, the switch from local millet varieties to an unknown, but
supposedly improved, variety bears an unpredictable risk. Therefore, as
observed during the 1992/93 production year, most of the farmers that tried

Okashana 1 planted it only on a small portion of the land under millet production.

77
It is hypothesized that larger farmers are more likely to try Okashana 1 because
they produce generally more food grains and thus can more easily bear the risk
of planting a new variety on part of their land (reduced risk adversity).

Second, Okashana 1 was from the beginning of its distribution not only
characterized as an early maturing variety that can cope with relatively short
production seasons, but was also promoted as a cash crop with a high yield
potential on fertile soils. It is hypothesized that farmers that are capable of
producing marketable surplus are more likely to use Okashana 1 than local
varieties. Therefore, a dummy variable (EXPOKASH) was included in the
regression that indicates whether farm households had grown Okashana 1 at
least once before the 1992/93 season (one) or not (zero). The coefficient of the
variable EXPOKASH is expected to be positive due to the positive relation
between adoption or trial of a new variety and (1) the size of land farmers
already cultivated, and (2) the perception of Okashana 1 is basically a cash crop
and the fact that those farmers that sell grain tend to have larger fields than
those who pursue solely subsistence

Two reasons explain why those farmers who actually Used Okashana 1
during the production season 1992193 are not the base of the dummy
EXPOKASH. First, it is assumed that not all farmers that tried Okashana 1 in the
past might have continued to use it until the 1992/93 survey period. Secondly,
after the drought of the 1991/92 production year, the Namibian government and
various NGOs distributed Okashana 1 seed widely among farmers to counter

potential seed shortages. Employing the actual Okashana 1 use during the .

78

1992193 production year as determinant of land cultivation would probably
under-represent those larger farmers that had already tried Okashana 1 before
the 1992/93 production season but did not continue to grow it. It would also
over-represent small farmers that would not have tried the new variety but had to
do so due to the seed shortage caused by a preceding drought.

(h) the level of farm households' annual cash income (CASHINCOME).
Cash income from activities other than crop production (e.g., migration or local
employment, livestock herding, and government pensions) might allow farmers
to acquire production inputs such as the right to produce on a certain land, seed,
field preparation equipment, fences against livestock, fertilizer, hired labor,
plowing services, etc.. To account for the effect that cash income has on the
area of millet planted per household, the amount of cash income that sample
households earned during the 1992/93 period (CASHINCOME) is included in the

equafionf

 

‘ Because the variable PLOWTRAC already controls the use of tractors
for plowing, the CASHINCOME variable is meant to estimate the residual impact
cash income has on the amount of land cultivated. Correlation tests show that
the use of tractors for plowing is strongly related to farmers' cash income, their
zone, and the geographic location of their community between west Ovambo
and east Kavango. The number of hectares cultivated does not seem to be
related to tractor use. The test results are presented below:

CASHINCOME HECTARES ZONE“ LOCATION“

PLOWTRAC Coefficient .2351 -.0090 .2857 -.2734
(Cases) ( 319) (247) ( 319) (319)
2-tailed Significance .000 .887 .000 .000

“ Ovambo = 1, Kavango = 0 ; *‘ ordinal numbers of survey sites ascending from west to east

79

The correlation estimates between the CASHINCOME variable and those
variables already included in the models indicate that for Ovambo, the variables
LANDCONSRT, PLOWHOE, PLOWTRAC, EXPOKASH, and for Kavango, the
variables DRFTANIMPLOW and PLOWTRAC are correlated with farmers' cash
income earnings (Table 3-16). These results explain why CASHINCOME can
only be included in the model to test its residual effects on hectares cultivated,
i.e., the effects that are not yet controlled for, like the right to produce on a

certain land, hired labor, fertilizer, seed, and fences against livestock.

Table 3-16. Correlations of the CASHINCOME to other model variables,
by zone (Pearson’s coefficients, 1-tailed Sig.)

 

Independent model variables

 

 

MILLET- LAND- HHLAB- ADL- DRFT- PLOW- PLOW- EXP-
PRICE CONST SHRT EOU— ANIM- HOE TRAC OKASH
R WRKDS PLOW
OVAMBO
coef.= .054 -.265 .048 .024 .018 -.1 1 1 .237 .197
p= .255 .001 .278 .384 .415 .089 .002 .008
KAVANGO
coef.= -.023 .042 -.024 .125 .144 -.092 .291 .077
p= .419 .355 .415 .133 .100 .206 .004 .288

 

Data Source: Namibian Millet Subsector Research Project, 1992193 and my own calculations

Both simple-linear and log-linear functional forms of the above described
equation were estimated for Ovambo and Kavango. The simple linear form was

preferred to the log-linear form on the basis of a careful analysis of the sign and

80
the significance of the estimated coefficients, the size of the coefficient of
determination adjusted for the degrees of freedom (adjusted R-Square), and the
residuals.

Table 3-17 presents the main statistics of the estimated simple-linear
equations. The levels of the adjusted R-Square indicate that the model explains
28 percent of the variability of the number of hectares under millet cultivation in
Ovambo and 41 percent of the respective variability in Kavango. These
moderate levels are not uncommon for cross-sectional data. They are
attributable to: (a) the categorical nature of most of the independent variables
considered, (b) the impact of the severe drought of the preceding crop year, and
(c) the late start of the 1992/93 cropping season under investigation. The severe
drought of 1991192 and its aftermath had reduced the number of available draft
animals or significantly weakened animals that had sustained the long dry
season. The late arrival of rains for the 1992/93 cropping season caused many
farmers to delay plowing and to cultivate less land than normal. Despite the
moderate R—Square values, the signs of all regression results are consistent with
expectations, and most of the considered explanatory factors demonstrate a
significant relation to the number of hectares .of millet cultivated. In the
following interpretation of the regression results, the base (control) case
represents a farm household that plowed its land with hired draft animals
(DRFTANIMPLOW=0, PLOWHOE=0, PLOWTRAC=0), that neither stated
constraints on clearing new land (LANDCONSRT=0), nor claimed labor shortage

as a cause for not having reached its planting goal (HHLABSHRT=0), and that

l—

fllv
vu—

wll

81

had not used the improved variety Okashana 1 before the 1992/93 season
(EXPOKASH=0).

Table 3-17. Estimates of the millet hectare equations, by zone

 

 

 

ovmeo KAVANGO
VARIABLES coer. S.E. SIGNIF. coer. S.E. SIGNIF.
Constant 2.33 0.16 0.00 - 1.29 2.76 0.64
MILLETPRICE 0.14 0.38 0.39 3.21 1.79 0.06
LANDCONSRT - 0.90 0.51 0.02 - 2.41 1.34 0.07
HHLABSHRT - 0.92 .12 0.08 - 0.57 0.87 0.51
ADLEQUWRKDS 0.35 .33 0.01 0.58 0.49 0.25
DRFTANIMPLOW 0.56 .43 0.09 0.67 0.84 0.43
PLOWHOE - 0.84 .36 0.06 - - —
PLOWTRAC 0.11 .01 0.75 3.04 1.19 0.01
EXPOKASH 0.87 .31 0.01 3.10 0.05 0.01
CASHINCOME 0.03 .56 0.06 0.05 1.07 0.26
Adjusted R-Square 0.28 0.43
Case Number 114 50

 

Data Source: Namibian Millet Subsector Project Surveys, 1992193

Holding all other factors constant, a rise of local kilogram prices of millet
(MILLETPRICE) by one Namibian dollar increases the area of millet cultivated
per household by 0.14 hectares (5 percent of the average amount of land
cultivated) in Ovambo but by 3.21 hectares (105 percent) in Kavango. In other
words, while Ovambo farmers respond to significant millet price increases only

with a marginal expansion of their fields, Kavango farmers respond to the same

82
price change by doubling their field sizes. Taking into account that only in
Kavango a larger number of farmers sell millet every year, while in Ovambo less
than a third of the farmers is actually able to sell millet every other year, the
differential responses to price incentives are not surprising.

The comparable small response in Ovambo can also be explained by the
increased population pressure in most parts of Ovambo that led to stationary
land cultivation practices and a more rigid structure of land distribution.‘ As
demonstrated in Figure 3-4, the MILLETPRICE variable was closely correlated to
the geographic location of the survey site and to the price of imported maize. In
December 1992, millet prices and maize meal were especially high in the remote
subzones western Ovambo and eastern Kavango. The correlation to location
might indicate that field sizes in remote areas are larger because there is less
land constraint. The correlation to prices of imported maize may indicate that
high maize prizes in western Ovambo and eastern Kavango, due to high
transportation costs and quasi-monopoly positions of rural grain traders, make
increased efforts in millet production necessary.

In Kavango, the average distance of 5 to 6 kilometers between farmers’
homesteads and their cultivated land indicates that many Kavango farmers still
have the option to find new arable land away from their densely populated

neighborhood. The small percentage of Kavango farmers that use manure to

 

‘ Under current production practices, farmers who try to increase their
millet production have traditionally opted to do this through field size expansion
rather than through intensification measures that are independent from the
amount of cultivated land.

 

83
increase land productivity (5 to 8 percent) helps explain why the expansion of
cropland is the predominant measure to increase total millet production.

As hypothesized, the indication of difficulties with field expansion has a
negative and significant effect on the amount of land cultivated per household.
The estimated marginal effect of the existence of problems with field size
expansion (LANDCONSRT) is a decrease of 0.9 hectares of millet area per
Ovambo household and a decrease of 2.41 hectares per Kavango household.
As explained previously, the most frequently mentioned constraints with respect
to field expansion were, in Ovambo, “scarcity of land near by” and “lack of
equipment” and, in Kavango, “lack of money to hire labor for de-bushing and
clearing new land.” The difference in the magnitude of the effects between
Ovambo and Kavango can be explained by the greater availability of arable land
in Kavango, which leaves Kavango households to be more affected by intrinsic
constraints regarding field expansion.

When sample farmers mentioned ‘labor scarcity during the phase of field
preparation’ as a problem and when all other factors are held constant, the
expected level of millet area is reduced per household in Ovambo by 0.92
hectares and in Kavango by 0.57 hectares, respectively. The coefficient of the
HHLABSHRT variable in the equation applied to Kavango farmers is
insignificant. The relatively strong and significant effect (8 percent) of labor
scarcity on the millet area in Ovambo is probably due to the large number of
Ovambo households that use either the hoe as the only field preparation

equipment (20 percent) or that use still the hoe in combination with draft power

34
from animals or tractors (58 percent). The comparable percentages of Kavango
households using the hoe are much lower (4 and 27 percent), indicating that in
Kavango, human labor is involved in field preparation to a much lesser extent
than it is in Ovambo.

The positive effect that the amount of labor used for grain production has
on the millet area supports the view that also the availability of household labor
for production phases after planting influences households’ land allocation
decisions. Ceteris paribus, an increase of labor spent on total millet production
during the 1992193 cropping season (ADLEQUWRKDS) by 10 adult-equivalent
workdays is associated with millet area increases per household of 0.35 hectares
in Ovambo and 0.58 hectares in Kavango. Although only significant at a 25
percent level, total labor availability for grain production demonstrates a larger
effect for Kavango, where labor shortage is more often stated than in Ovambo.

The ownership of draft animals, as well as the use of plowing equipment
that is different from a plow drawn by animals, has the expected effects on the
cultivated millet area. The estimated marginal effect of the ownership of a
complete set of draft equipment/animals (DRFTANIMPLOW), which allows field
preparation in a timely manner, is an increase in millet area per farm by 0.58
hectares in Ovambo and 0.67 hectares in Kavango. The coefficient of the
DRFTANIMPLOW variable is in the Ovambo equation significant at the 10
percent level, while the respective coefficient in the Kavango regression is

insignificant.

85

The effect of using only a hoe for field preparation on the millet area per
household (PLOWHOE) is negative and significant at the 6 percent level for
Ovambo. The expected area of millet per household decreases by 0.84
hectares for households that used human labor equipped with a hoe rather than
hiring a set of draft equipment/animals. Due to lack of household cases that use
only human labor in combination with a hoe in Kavango, the PLOWHOE variable
had to be excluded from the Kavango equation.

As hypothesized, the effect of using a tractor for plowing is positive on the
millet area per household in both zones. However, for Ovambo farmers, the use
of a tractor (PLOWTRAC) does not seem to be significantly related to the millet
area per household. This finding can be explained by the high cost for
mechanical plowing service and the uncertainty of its timely arrival that led two
thirds of the farmers that hired mechanical plowing service also to use a hoe or
draft animals. Contrary to Ovambo, the marginal effect of mechanical plowing is
quite large (3.04 hectares) and significant (0.01 percent) in Kavango. Because
only four percent of the interviewed Kavango farmers received plowing service
during the 1992/93 production season, and because the correlation coefficient of
0.35 between Kavango households’ annual cash income and tractor use is the
highest among the variables included in the Kavango equation, it is assumed
that mainly wealthy farmers with larger land holdings have access to mechanical

plowing.

86

As discussed, the variable indicating that farmers had used Okashana 1
before the 1992/93 production season (EXPOKASH) can not explain the
variability of the amount of hectares under millet cultivation. However, as
expected, the use of Okashana 1 before its wide distribution in 1992/93 seems to
be a good predictor of farmers with larger fields. For households that used
Okashana 1 before the 1992/93 production year, the expected area of millet
under cultivation is increased by 0.87 hectares in Ovambo and by 3.10 hectares
in Kavango. Both coefficient estimates were significant at the level of one
percent.

The most plausible explanations for the strong and significant relation
between experience with Okashana 1 before the 1992/93 production season and
hectare size are as follows: First, larger farmers are probably less risk adverse
regarding the trial of a new variety. Second, farmers who reached the level of
millet self-sufficiency and opted to produce millet beyond their own need
probably tried Okashana 1 as a cash crop. The second explanation is tested by
calculating the correlation coefficients between the EXPOKASH variable and
three bi-nominal variables (SELLNEVER, SELLSOMEYEARS, SELLEACHYEAR)
indicating the general millet selling frequency of surveyed farmers (Table 3-18).

The correlation results presented in Table 3-18 suggest that (a) farmers
who do not sell millet at all have less likely used Okashana 1 before the 1992/93
production season, and (b) farmers who do sell millet at least in some years
have more likely used Okashana 1 before the 1992/93 production season.

Because of the low number of farmers selling millet every year, no statistically

87
significant results could be obtained. The correlation results generally confirm
the explanation that those farmers that tend to sell millet had tried to sell

Okashana 1 as a cash crop before 1992193.

Table 3-18. Correlations of EXPOKASH and millet selling
frequencies, by zone (Pearsons’ coefficients)

 

Farmers' millet selling frequency

 

 

sell never sell some years sell each year
coef. = -.2030 coef. = .2030 coef. = na
OVAMBO n = 198 n = 198 n = 198
p = .004 p = .004 p = na
coef. = -.1506 coef. = .1316 coef. = .0061
KAVANGO n= 117 n= 117 n= 117
p= .105 p= .157 p= .948

 

Data Source: Namibian Millet Subsector Research Project, 1992193 and my own calculations

The variable CASHINCOME was included to estimate the effects that
purchases of fertilizer and seed, hired labor, and payments to tribal authorities
for permission to hold crop land have on the amount of cultivated land. The
obtained regression results indicate that the variable CASHINCOME affects the
number of cultivated hectares very little. In both study zones, the increase of the
annual cash income by N$ 1000 (CASHINCOME) has only a small but positive
effect on the estimated millet area. This finding might be explained partly by the
fact that alternative employment activities not only provide the financial means to
acquire or access the mentioned production inputs but also reduce (a) the

number of household members available for millet production and (b) the need to

88
produce millet as a cash crop.

An additional explanation for the relatively low estimated effect of annual
cash income on households’ millet area is the incorporation of other variables in
the model, like LANDCONSRT, PLOWHOE, PLOWTRAC, EXPOKASH, and
DRFTANIMPLOW, that are at least to some extent correlated with household
cash income and thus reduce the explanatory power of the CASHINCOME

variable itself.

3.3.3. Regression analysis of millet yields

The purpose of the following section is, first, to give an overview about the
cross-section model of households’ millet yields per hectare, second, to describe
the independent variables employed in the equation, and third, to present and
interpret the estimated coefficients of the model.

The cross-section model of millet yields per hectare achieved by farm
households during the 1992/93 production season encompasses variables that
reflect (a) natural production factors that are beyond human control, (b) direct
production inputs that are under farmers’ control, and (c) household
characteristics that might influence indirectly the availability and use of
production resources. To better identify zones’ differences in the magnitude of
effects from the independent variables, yields of millet are estimated separately
for Ovambo and Kavango. In the following, the regression equation used is
presented together with a brief description of its variables. The variables are all

tested for correlation and had met the test:

89

_,.c'_"

MILLETHAYLD = f (RAIN92193, AEWRKDSIHA, MNRALLFLDS,

USEOKASH, CASHINCOME, GENHHEAD)
where:
MILLETHAYLD = amount of millet harvested per hectare (KG)
AEWRKDSIHA = adult-equivalent workdays per hectare (No. in '0)
MNRALLFLDS = use of cattle manure on all fields (DUMMY, 1 = yes)
USEOKASH = use of Okashana 1 during 1992/93 season (DUMMY, 1 = yes))
CASHINCOME = annual cash income from non-cropping activities (N5 in '000)
GENHHEAD = gender of head of household (DUMMY, 1= male)

Table 3-19. Characteristics of the millet yield regression variables,

 

 

by zone
OVAMBO (n 8 147) KAVANGO (I1 I47)
VARIABLE MEAN S.E. MEAN S.E.
MILLETHAYLD (KG) 392 559 1 98 41 5
RAIN92193 (MM in '00) 3,62 1.79 3.21 1,81
AEWRKDSIHA (No. in '0) 6.8 7.4 5.5 8.1
MNRALLFLDS (1 = yes) 51 percent = 1 19 percent = 1
USEOKASl-l (1 = yes) 47 percent = 1 36 percent = 1
CASHINCOME (NS '000) 7.8 11.7 6.9 8.2
GENHHEAD (1 = male) 64 percent = 1 81 percent = 1

 

Data Source: Namibian Millet Subsector Research Project, 1992193

90

The independent variables used and the reasons for their inclusion in the
estimated equation are as follows:

(a) the local rainfall during the 1992/93 production season (RAIN92193).
As described in Part 1 of Chapter 3, rainfall is the most important input for
dryland crop production in Namibia. On average, the Kavango zone receives
more and earlier rain than Ovambo. However, during the 1992/93 production
season, rains arrived very late in Kavango, and during the short growing period
left, the distribution of the rain was poor. According to farmers’ statements, the
area of western Ovambo received the best rainfall season that it had for years.
To account for the effect that rainfall has on millet yields, the total precipitation
(millimeters in ‘00) of the 1992/93 production season measured by the rainfall
stations closest to the 16 survey sites (RAIN92193) is included in the equation of
millet yields. Although it is possible that poor distribution of local rains can nullify
or even reverse the effect of an adequate amount of rain in a production area, it
is hypothesized that millet yields are positively related to the variable RAIN92/93.

(b) the number of adult-equivalent workdays that household members
spent per hectare of planted millet (AEWRKDSIHA) during the whole 1992/93
production season. At the prevailing low state of production technology, human
labor is assumed to be one of the most important determinants of millet yields in
northern Namibia. The availability of enough labor to weed grain fields and to .
protect the crops against insects and birds often decides whether farmers’ efforts
during field preparation are redeemed at harvest time. To account for the

variation of labor input into millet production, the number of adult-equivalent

91
workdays that farmers spent per hectare of cultivated millet (AEWRKDSIHA)
during the 1992/93 production season is included in the equation as determinant
of millet yields. As for the preceding regression model of millet area per
household, the estimation of AEWRKDSIHA was based on farmers’ statements
about the start and end of the various phases of millet production and the
number of hours individual household members worked per week on crop
production during these production phases. Regardless of sex, all workers
below 15 years of age or older than 55 years of age were attributed a weight of
0.5, while all workers of an age between these two categories were given the
weight of one. The amount of total labor spent for millet production was
eventually divided by the number of hectares of millet cultivated per household.
The coefficient of the variable AEWRKDSIHA is expected to be positive as a
result of the positive effect that the availability of labor for millet production has
generally on field management and thus on the productivity of land.

(c) the use of manure and chemical fertilizer on millet fields
(MNRALLFLDS). As described in Part 1 of Chapter 3, the combined effects of
low clay and low organic matter content result in soils with low water holding
capacity and low fertility. To counter the loss of soil fertility through continuous
cultivation of millet on the same fields, Ovambo farmers use either manure (81
percent) and/or chemical fertilizer (9 percent). In Kavango, where households
still have the option to clear new land for crop production, much less fertilizer is
used. Only between 5 and 8 percent of all Kavango households use manure on

their fields and less than 1 percent use chemical fertilizer. However, of the cases

92
employed in the regression equation, only 51 percent of the Ovambo farmers
and only 19 percent of the Kavango farmers fertilized all of their fields with
manure or chemical fertilizer during the 1992/93 production year. Of those
Ovambo farmers who do not, or only partly fertilize their fields, the majority
explain their behavior with lack of animals and therefore lack of manure. Of the
respective Kavango farmers, the majority explained their behavior with ( 1) lack of
equipment to transport cattle manure to their far distant fields, and (2) the high
cost of chemical fertilizer. To account for the effect fertilization has on farm
households’ average millet yields, the dummy variable (MNRALLFLDS) is
included in the regression model. The variable takes the value one if farmers
stated they used manure and/or chemical fertilizer on all of their fields during the
1992/93 grain production. The coefficient of the variable MNRALLFLDS is
expected to be positive as a result of the positive effect that fertilizer use has on
crop yields.

(d) the use of the new millet variety Okashana 1 (USEOKASH) during the
1992193 season. The distribution of Okashana 1 at the start of the 1992193
production season through the Namibian Ministry of Agriculture and NGOs led to
increased use of this fast maturing millet variety. To measure the effect that the
use of Okashana 1 seed had on farm households’ overall millet yields, the
dummy variable USEOKASH is included in the equation. The variable takes the
value 1 if farmers had cultivated at least part of their millet fields with Okashana
1. It is hypothesized that the amount of millet harvested per hectare is positively

related to the variable USEOKASH.

93

(e) the amount of farm households’ annual cash income
(CASHINCOME). If farm households are limited in the amount of labor they
have available for crop production, cash earnings from non-crop production
related activities might be useful to compensate for this shortage through hiring
labor. Hired labor can be used directly for field work or production related
activities like fence making around crop fields or gathering of cattle manure for
fertilization. Higher cash incomes might also enable households to have timely
access to the service of draft power during field preparation time, which
eventually lengthens the actual period of plant growth and leads to higher yields
of millet. Additionally, the amount of cash income available might influence
households’ decision whether and how much chemical fertilizer to apply on crop
fields. To account for the effect that cash income has on the use of production
inputs that are not captured by other variables, the equation for the amount of
cash income earned during the 1992193 production year per household is
included in the equation under the variable name CASHINCOME. Based on the
explanations given above, it is hypothesized that millet yields per hectare are
positively related to the CASHINCOME variable.

(f) the gender of head of household (GENHHEAD) as an indicator of farm
hoeseholds' access to production inputs like hired labor or timely plowing service
that are not only determined by farmers' financial resources, but also by their
social status. Of the interviewed farm households, 38 percent in Ovambo and
16 percent in Kavango are female-headed. As demonstrated in Part 2 of

Chapter 3, significantly less women (Ovambo: 13 percent, Kavango 6 percent)

94

earn cash income from local or migration employment than men. (Ovambo: 24
percent, Kavango 19 percent) As a result, female headed households often lack
either a male household member who could contribute directly to crop production
through additional labor support or lack cash income earned by a male head of
household that could be used to acquire production inputs that affect yields.

It is hypothesized that in addition to the effects of reduced labor and cash
income already controlled by two other variables in the equation, the lower social
status endured by women in northern Namibia affects negatively the access of
female-headed households to production inputs. According to a UNICEF report
about the situation of children and women in Namibia, the difference in the socio-
cultural status of women becomes particularly evident in customary inheritance
laws where:

.. widows may be dispossessed of their deceased husbands’
estate through the ‘Iegal’ intervention of their husbands’ brothers or
male relatives. As married women, wives may use, but rarely own,
the family’s livestock or land, and may not enter into contractual
agreements without their husbands’ permission” (UNICEF/NISER,
1991, p.123).

To account for the effect that the gender of the household head has on
yields, the dummy variable GENHHEAD is included in the regression model.
The variable takes the value one if the farm household is headed by a male head
of household. The variable GENHHEAD is expected to be positively related to
millet yields due to the better access of male-headed households to production ,

inputs.

95

Both simple-linear and log-linear functional forms of the above described
equation were estimated for Ovambo and Kavango. The simple linear form was
preferred to the log-linear form on the basis of a careful analysis of the sign and
the significance of the estimated coefficients, the size of the coefficient of
determination adjusted for the degrees of freedom (adjusted R-Square), and the
residuals.

Table 3-20 presents the main statistics of the estimated simple-linear
equations. The adjusted R-Square levels indicate that the model explains
27 percent of the variability of millet yields in Ovambo and 60 percent of the
respective variability in Kavango. Similar to the regression model of millet area
per household, the moderate levels of the adjusted R-Square have to be
attributed to the categorical nature of half of the independent variables
considered and to the impact of the severe drought of the preceding year in
combination with a late start of the cropping season under investigation. Despite
the moderate R-Square values, almost all signs of the estimated coefficients are
according to expectation and most of the considered explanatory factors
demonstrate a significant relationship to the yields of millet.

For the following interpretation of the regression results, the base (control)
case represents a farm household during the 1992/93 production year that did
not or only partly fertilized its crop land (MNRALLFLDS=0), did not use
Okashana 1 (USEOKASH=0), and whose head of household was female

(GENHHEAD=0)

96

Table 3-20. Estimates of the millet yield equations

 

 

 

OVAMBO KAVANGO
VARIABLES COEF. S.E. SIGNIF. COEF. S.E. SIGNIF.
Constant - 455 121 0.00 - 505 185 0.01
RAIN92193 152 23 0.01 304 56 0.00
AEWRKDSIHA 14 5 0.01 30 6 0.00
MNRALLFLDS 156 81 0.06 4 104 0.97
USEOKASH 6 85 0.94 100 87 0.25
CASHINCOME - 1 3 0.81 9 5 0.07
GENHHEAD 189 85 0.03 265 121 0.03
Adjusted R-Square 0.27 0.60
Case Number 148 47

 

Data Source: Namibian Millet Subsector Project Surveys, 1992193

Holding all other factors constant, a rise of local rainfall (RAIN92193) by
100 millimeters increases millet yields by 152 kg in Ovambo and by 304 kg in
Kavango. The fact that, in Kavango, yields responded twice as much to rainfall
increases can be explained by the overall low rainfall levels in Kavango during
the 1992/93 season. The estimated coefficients of the RAIN92193 variable are
significant at the 1 percent level.

Compared to the described relationship between rainfall and yields the
effect of labor spent per hectare (AEWRKDSIHA) is relatively small. When all
other factors are held constant, the expected level of millet yield increase per 10

additional days of adult-equivalent labor spent per hectare is only 14 kg in

97
Ovambo and 30 kg in Kavango. For both study zones, the estimated coefficients
of the AEWRKDSIHA variable are significant at the 1 percent level.

As hypothesized, fertilization of all millet fields affects millet yields
positively. In Ovambo, the use of manure on all millet fields of a farm household
increases the hectare productivity by a respectable 156 kg (significant level at
6 percent). In Kavango, the marginal effect of manure use on all fields was only
4 kglha. However, this result was far from being significant.

The coefficients estimated from the USEOKASH variable were below
significance level for both study zones. This is probably due to the small number
of household cases that used Okashana 1, the large yield variation among those
farmers, and the fact that Okashana 1 was cultivated only on part of households’
millet fields. However, the magnitude of both coefficients is coherent with the
results from the descriptive analysis. In Ovambo, where rainfall was adequate,
the fast maturing millet variety Okashana 1 could not outperform local millet
varieties. The use of Okashana 1 increased estimated millet yields only by
6 kglha (2 percent of Ovambo farmers’ average yield). The almost similar
performance of local millet varieties is probably based on their slower growth
rate, which allowed longer use of the available soil moisture during this
production season. In Kavango, however, where due to late and poor rainfall,
yields were generally depressed, average yields increased by about 100 kglha
(33 percent of Kavango farmers average yield) if farmers used Okashana 1 in

addition to local varieties.

98

For Ovambo, households’ annual cash income shows little significant
relation to land productivity. The estimated coefficient of the CASHINCOME
variable is negative, but close to zero, and at an insignificant level. For
Kavango, the sign of the estimated CASHINCOME coefficient is according to
expectation. An increase of annual income by NS 1000 per household increases
millet yields by 9 kg. This result is significant at a 7 percent level. As for similar
findings from the preceding regression of the millet area, two factors might
explain the small effect of increased CASHINCOME income earnings on millet
yields. First, farmers that earn considerable amounts of cash income from non-
crop production related activities might be tempted rather to invest further in
these activities instead of investing in millet production inputs. Second, other
variables in the model, like MNRALLFLDS, USEOKASH and GENHHEAD, are at
least to some extent related to household income and thus reduce the
explanatory power of the CASHINCOME variable.

As hypothesized, the gender of the household head shows an effect on
households’ millet yields per hectare. Estimated yields increase by 189 kg in
Ovambo and 265 kg in Kavango if the head of household is male. As discussed
before, it can be assumed that the gender variable takes some explanatory
power from the variables that represent households’ labor input (AEWRKDSIHA)
and annual cash income (CASHINCOME). But the most plausible explanation
for such high effects is still the relatively low social position of female household
heads (especially of older women) that limits their access to important production

factors, like fertile land or timely plowing and weeding services.

99
3.3.4. Correlations analysis

Descriptive and regression analyses employed in previous sections of this
chapter helped to identify the main production determinants of millet. During the
discussion of the analysis results, hypotheses were raised as to how these
production determinants relate to farm household characteristics like ‘income
from migration”, ’gender of head of household’, 'ownership of cattle’, 'farrn
location', and 'grain marketing behavior'. The goal of this section is to determine
the characteristics of farmers that are already using production enhancing inputs
and those who don’t. Table 3-21 presents the factors for which a strong and
statistically significant relation to millet production on the farm level could be
determined with the regression analyses in the preceding sections.

The method chosen in this section to identify statistically significant
relationships is based on simple correlation testing.‘ For each study zone, the
correlation coefficients were estimated between the nine production
determinants and 30 variables that are grouped in the following categories:

(1) household characteristics, (2) household possessions,
(3) farm characteristics, (4) farm location, and (5) farmers' millet selling

frequencies.

 

‘ Theoretically, the most correct (econonometric) way to test on which
factors the production determinants are dependent is through the estimation of a
simultaneous system. However, this is laborious for more than one input use,
and the effort would not be justified, taking into account the limited accuracy of
available cross-section data. The same applies to the formulation of separate
regression models with the production determinants as dependent variables.

100

Table 3-21. Main determinants of millet production, by zone

 

 

Ovambo Kavango

millet prices X
land constraints X X
labor shortage X x
plowing equipment ownership X

plowing equipment used X X
1992193 rainfall X X
use of manure X

use Okashana X
gender of household head X X

 

Data Source: Namibian Millet Subsector Project Surveys, 1992193 and my own calculations

The correlation results revealed that most correlation coefficients
(Pearsons’) were low. The level of the highest coefficient reached only 0.32.
However, the signs of most coefficients were in line with previously made
hypotheses. The following paragraphs discusses relations between production
determinants and household characteristics that were found to be statistically
significant. The correlation coefficients and respective significant levels on which
the discussion is based are presented in Appendix A.

Use of the hoe for field preparation: Regression and correlation results
demonstrate that farmers using solely the hoe for field preparation have

considerably smaller fields. Farmers that use only the hoe for field preparation

1 01
are more often found in eastern Ovambo and eastern Kavango. Ovambo
households using the hoe tend to be female-headed, grain self-sufficient, have
no or little migration income, have few literate household members, and produce
low yields.

Use of draft animals for plowing: Because almost all Kavango farmers
use oxen for plowing, no significant relation between the respective variable and
other variables could be identified. Ovambo household using either oxen or
donkeys for field preparation tend to be: located in western Ovambo with better
soils, grain self-sufficient, produce millet for the market, sells millet in some
years, and are far away from commercial grain markets.

Use of tractor service for plowing: The Ovambo households using tractor
service for plowing are more likely located in central Ovambo and less likely in
eastern Ovambo. They tend to be male-headed. Respective Kavango
households do not show a geographic pattern. They seem less likely to be grain
self-sufficient. In both zones farmers that use tractor service earn cash income
from migration work, have more literate household members, own motor
vehicles, intend to produce marketable millet surplus, and cultivate more than 5
hectares.

Ownership of draft animals and a plow: Ovambo farmers owning draft
animals and a plow tend to be grain self-sufficient, cultivate more hectares than
average, are more likely to live in central Ovambo, and live less likely in eastern
Ovambo. The respective Kavango farmers cultivate more land so that the land

available per adult-equivalent is larger. Their 1992/93 millet production was

102
comparatively high.

Experience problems with land clearing: Farmers from eastern Kavango
tend to have more often problems with clearing new land than farmers from other
areas in Kavango. Since Ovambo is more densely populated than Kavango and
arable land is scarce farmers have to be high in social status and have to have
financial resources to acquire the right from local authorities to cultivate land.
Ovambo households having problems accessing new land for crop production
tend to be female-headed, have little or no cash income, have few members who.
are literate, have a current land holding that is very small, produce relatively little
grain, are located far away from markets, but sell millet in some years to balance
their most urgent cash needs.

Labor shortage: Labor shortage occurs western Ovambo, central and
eastern Kavango.

Millet prices: In Ovambo, millet prices were low in central areas close to
major urban centers like Oshakati or Ohangwenna. Those Ovambo farmers
facing higher millet prices tend to use chemical fertilizer, but do not market their
millet. Kavango farmers facing higher millet prices tend to have problems with
land clearing, are not motorized, market millet, and live in eastern and central
Kavango.

Use of manure or chemical fertilizer: In Kavango, the fertilization is
positively correlated to the amount of land cultivated. Ovambo farmers using
manure tend to be grain self-sufficient, own a motor vehicle, produce millet for

the market, cultivate more land, produced relatively much millet in the production

1 03
years 1991193 and 1992193, have already good soils, are more likely located in
western Ovambo and less likely in central Ovambo.

Use of chemical fertilizer: Ovambo farmers using chemical fertilizers tend
to be grain self-sufficient and live either in western or central Ovambo. Kavango
farmers using chemical fertilizer are located either in eastern or central Kavango,
live close to the road, and tend to sell millet in some years.

Experience with Okashana 1: Ovambo farmers that tried Okashana 1
within the first three years after its introduction in 1989 live in western Ovambo,
are male headed, are grain self-sufficient, earn cash income, have high
proportion of literate household members, have larger number of household
members, own a motor vehicle, produce part of their millet as cash crop,
cultivate comparatively more land, had high yields and a large millet production
during the 1992/93 season, have good soils, and sell millet in some years. The
respective Kavango farmers that cultivate more millet than the average farmer,

are located in central Kavango.

3.4. Chapter summary

The final part of Chapter 3 summarizes, first, the description of the millet
production system in northern Namibia, second, the assessment of the
performance of the millet production system, and third, the relations that

important millet production determinants have to farm household characteristics.

1 04
3.4.1. Characteristics of the millet production system

The analysis of millet production practices in Ovambo and Kavango
reveals a semi-subsistence agriculture that is making limited use of modern
inputs such as improved seed and fertilizer. Only about half of the farmers state
that agriculture is their main occupation. Low and uneven rainfall and lack of
water for irrigation limit crop diversification. While lack of livestock prevents
many farmers from using organic fertilizer, chemical fertilizer is only used by nine
percent of the farmers in Ovambo and less than one percent of farmers in
Kavango.

Pearl millet, the food staple in Ovambo and Kavango, forms also the
centerpiece of farming in both regions. Because of increasing population the
practice of shifting has been discontinued in Ovambo. Farmers practicing
shifting cultivation in Kavango have to clear new crop land further and further
away from their homestead. The observation indicates the need for higher yields
and labor productivity. The latter is due to the fact that only half of the adult
household members in Ovambo and only 40 percent of those in Kavango are
mainly employed in agriculture.

The fact that 33 percent of rural households in Ovambo and 51 percent in
Kavango cultivate less than 2 hectare and generate not more than a third of their
total income through agriculture indicates that development programs in the
study zones must look beyond the farm toward employment creation in rural-

based non-farm activities linked to agriculture.

1 05

Because less than half of the communal farm households own a full set of
plowing equipment/animals and because the majority of households cannot
afford to pay much for renting draft power, many farmers are unable to cultivate
more millet. Also, farmers' choice between different types of millet varieties is
limited. While local varieties need a long time to mature to make full use of the
low soil fertility levels, the high yield potential of the newly introduced variety,
Okashana 1, requires favorable soil conditions.

Figures about the amounts of millet traded in the commercial food market
indicate that Kavango has a higher potential for commercial millet production
than Ovambo. Although Kavango has only a forth of Ovambo’s population and
its millet production was depressed in 1992/93 due to poor rainfall, it contributed
about as much marketable millet surplus to the commercial grain market as

Ovambo.

3.4.2. Performance of the millet production system

Knowing the restrictions of the prevailing cropping system, the low yields
of the majority of farmers in Ovambo and Kavango is not surprising. Only about
10 to 14 percent of rural farm households cultivate more than 5 hectares of
grain. While the average of 3 hectares of cultivated land per household and 0.4
to 0.5 per adult-equivalent is in line with neighboring countries like Zimbabwe or
Zambia, the low average yields of 200 kg of millet per hectare in Ovambo and
250 kg in Kavango lead to a significant gap between grain production and

households’ annual food need.

106

The yields of the 1992/93 production season were very low in most areas
of the study zones. In eastern Ovambo and whole Kavango millet yields were
less than 100 kglha. In central Ovambo 240 kg of millet were harvested per
hectare. Only in western Ovambo, were rainfall was exceptional good, yields
averaged 590 kg.

One major cause for the poor yield performance in eastern Ovambo and
in all of Kavango was the late arrival of rains during the 1992/93 survey year.
The short production season especially affected the performance of local
varieties that need a relatively long time to mature. The average yields of local
varieties ranged only between 68 and 96 kglha. By contrast because of its faster
growing potential, Okashana 1 performed relatively better in these areas, ranging

from 127 to 235 kglha. In western Ovambo were the rainfall distribution was
very good local varieties clearly outperformed Okashana 1.

The promoters of Okashana 1 had claimed that it would double average
Yields from 300 to 600 kglha on the farm level. However, in western Ovambo,
lOcal varieties yielded, on average, 644 kglha, while average yields of Okashana
1 were only about 234 kglha. In central Ovambo, Okashana 1 yielded, on
average, only 106 kglha compared to local varieties’ average yield of 259 kglha.
The main explanation for the better performance of local varieties in good rainfall
areas is their slower growth rate. This permits a longer period of time than

Okashana 1 to make use of the available soil moisture and the plant nutrients.

1 07

The calculation of average cost and returns of millet production revealed
that those farm households that reached grain self-sufficiency from their 1992193
production were predominantly located in areas of good rainfall, but also had
higher variable cost for labor input than households that produced a grain deficit.
Calculation of returns to household labor indicate, that only grain self-sufficient
households were able to generate significantly more value per workday through
their own millet production (Ovambo: N$1Day 10.30, Kavango: NSIDay 6.86)
compared to what they could earn through work on the fields of other farmers
(NS/Day 5.00).

Total household income (consumption and cash income) and also the
income contribution from grain production vary mainly with the employment
status of individual household members. About half of all Ovambo and Kavango
households have at least one household member that earns cash income from
migration employment. The total income of such households is, on average, six
times more than the income of household with no migration employment. While
the share of grain production on total income averages only between 6 and 14
percent for households with migration employment, grain production represents
a share of 40 to 45 percent for households without migration employment. Thus,
for the latter, significantly poorer household category grain production represents
the most important source of income. Other important income sources for this
household category are government pensions to the elderly in Ovambo (average
share of 43 percent) and employment activities such as woodworking and beer

brewing in Kavango (average share of 37 percent).

1 08
3.4.3. Millet yield determinants

Congruent and complementary results regarding the determinants of
millet production are provided by (a) the descriptive analysis of farmers’
production problems and reasons for high yields levels, and (b) two simple
regressions estimating the area of millet and millet yields achieved per
household. Problems experienced by farmers during the field preparation period
of the 1992193 season and the reasons stated by farmers for not having
achieved their initial planting goals point to the following production constraints in
the study zones. The major constraints on land preparation in Ovambo are ( 1)
lack of arableffertile land, (2) lack of plowing equipment like hoes and/or plows,
and (3) latelnon-arrival of the plowing service organized by the MAWRD. In
Kavango, the three main obstacles to land preparation are (1) shortage of
household or hired labor, (2) lack of draft power, and (3) lack of plowing
equipment. In both study zones, crop damage through insects and birds present
a continued threat until grain harvest.

The regressions estimating millet area and millet yields per household for
the 1992/93 season confirm not only part of the findings from the descriptive
analysis, but also quantify the effects of various production determinants and
make differences between Ovambo and Kavango more explicit.

In Ovambo, the three variables effecting the area of millet cultivated most
are: (1) access to land; (2) household labor; (3) draft power in combination
with the used plowing equipment. These results are not surprising if one

considers that Ovambo faces increased population density in production areas,

1 09
that only half of Ovambo's adult population is occupied with agriculture, and that
hand hoeing is the most labor- intensive and time-consuming field preparation
technology available.

In contrast to Ovambo, the most important determinants in Kavango are:
(1) millet prices at the start of the production season; (2) the use of tractors for
plowing; (3) labor for land clearing. These results fit well in the context of
Kavango where natural resource levels such as seasonal rainfall and fertile land
are generally higher than in Ovambo. Large areas of uncleared and less
populated bushland allow producers to react to price incentives with new land
clearings. It is obvious that those few Kavango farmers who have tractors can
not only plow more land, but also have better access to uncleared land that is far
away from the more densely populated areas. Finally, it is understandable that
farm households with intrinsic land clearing constraints cultivate smaller millet
fields than those who can adjust their field sizes to changes in nutritional needs
and/or changes in producer prices.

The results from the regression of millet yields demonstrate that the
amount of seasonal rainfall is the most important determinant of land productivity
in both study zones. Another main determinant of millet yields in both study
zones is the gender of the household head. Regression results indicate that
female-headed households have significantly lower millet yields than male-
headed households. This can be explained by the low social status of female
heads of household that limits their access to important production factors like

fertile land or timely plowing and weeding services.

1 10

In Ovambo, where continuous cultivation of the same land with millet
leads to fast decline in natural soil fertility, the use of manure shows additionally
large effects on farmers’ yields of millet.

In Kavango, where the very short and dry season of 1992193 favored fast-
maturing crops, farmers' yields were increased if they used the newly introduced
millet variety, Okashana 1. However, these yield increases through Okashana 1
occurred at relatively low yield levels and, due to large variability of yield data,
only a significance level of 25 percent could be obtained for this result.

Although statistically significant, the relations between households' annual
cash income and households' number of workdays spent per hectare of millet

production show little effect on millet yields.

3.4.4. Yield determinants and household characteristics

Correlation analysis was used to determine the characteristics of farmers
that are already using production enhancing inputs and those who don’t. The
statistically significant correlations identified between yield determinants and
household characteristics were of low magnitude but their sign conformed to the
expectations from economic theory.

Those households that use only the hoe for field preparation are often
female-headed, have little or no cash income, have low education levels, but
reach grain self-sufficiency. Households that experience difficulties expanding

their crop land have similar characteristics.

1 1 1

In Kavango, almost all farm households use animal traction (either hired
or owned) for field preparation. In Ovambo, plowing with an animal span is
concentrated in western Ovambo, which is remote from markets but has
favorable production conditions so that cash crop production is common.
Farmers owning a plow and draft animals tend to be grain self-sufficient. They
also tend to have larger land holdings than other farmers. In western Ovambo
such farmers tend to fertilize all their land with manure.

The use of a tractor services is more common among households whose
main income stems from migration work. In Ovambo, the use of tractor services
concentrated in central areas. Farmers using chemical fertilizer are more likely
located in western and central Ovambo and western Kavango. The Kavango
farmers that use chemical fertilizer often produce millet as a cash crop and live
close to a road.

Early adopters of new production technologies like the newly introduced
seed, Okashana 1, tend to be male-headed households that earn cash income,
are better educated, own a motor vehicle, have large fields, are grain self-
sufficient, grow millet partly as a cash crop, and live in western Ovambo.

In summary, farmers with marketable millet surplus tend to cultivate more
than 5 hectare of millet, use manure or chemical fertilizer, possess a full set of
plowing equipment/animals and/or have enough cash income to hire tractor
service. The production areas that have more farmers with these characteristics

and that therefore are more likely to produce marketable millet surplus on the

1 12
aggregate level in the future are: western Ovambo, eastern Kavango, and

central Kavango.

4. MILLET MARKETING

The preceding chapter demonstrated that most farmers in Ovambo and
Kavango are geared rather towards subsistence than to cash crop production.
However, as northern Namibia moves from an agricultural to an urbanized, and
thus more manufacturing and service-oriented economy, the performance of the
rural and urban grain marketing system becomes increasingly important. This
chapter deals with three issues of the existing marketing system.

The first section compares the general price structure between staple
grains, presents the findings about consumer prices for locally produced millet,
and finally, focuses on the price competitiveness of locally produced,
unprocessed millet versus processed maize imported into the study zones. The
price data used in the first section stem mainly from the food price monitoring
survey carried out by the Millet Subsector Research Project, between February
1993 and January 1994. However, some of the used price data are based on
information gathered through the millet subsector household surveys.

Section 2 presents findings from the millet trader survey conducted at the
end of the 1992/93 production season. It concentrates mainly on the question of
how the commercial sector is involved in millet trade. It also introduces new
findings about millet imports from Angola to urban Ovambo. Seen as a whole,
Section 2 contradicts common beliefs among policymakers in Windhoek that
northern Namibia has no commercial millet market and that government
institutions have to fill this vacuum in the future. The section proves the

existence of private entrepreneurs that are heavily involved in millet trade. It also

113

1 14

indicates areas where support from outside is necessary in order to make locally
produced and commercially sold millet more attractive to northern consumers.

Finally, Section 3 focuses on the marketing behavior of millet producers.
By describing farmers' millet marketing practices, their limitations of selling millet
commercially are revealed. Regression analysis is used to identify determinants
of farmers’ grain selling decisions. The last point is especially important because
it reveals whether the lack of incentives, as argued by the Namibian Agronomic

Board, is the main reason that keeps farmers from producing millet commercially.

4.1. Millet prices

This section describes first how millet prices generally evolve in the study
zones. Thereafter, findings about seasonal and geographic millet prices are
presented that are based on data gathered during the 11-month-Iong price

monitoring survey that started in February 1993.

4.1.1. Formation of millet prices in general

Millet is often sold or bartered between rural neighbors, i.e. between rural
households that produce marketable millet surplus and those that are net food
buyers. This type of millet exchange that does not involve middlemen is also
called 'infonnal trade.’ When talking about 'inforrnal trade' at the aggregate level,

this study uses the term 'infonnal market.’

1 15

lnforrnal trade prices of millet are well-known for smaller, traditional
volume measures like recycled oil cans and tin buckets. Although there is no
official mechanism that regulates food prices in rural areas, institutions like rural
churches, as well as the traditional obligation to help neighbors who are in need,
keep informal millet prices at low levels compared to prices on the commercial
grain market.

Problems arise in the informal millet market after a very poor grain
harvest. Due to the fear that the next harvest might also be poor, households
that still have millet reserves reduce and eventually stop the selling of millet to
neighbors and conceal the amounts of millet left in order to not be further
approached for help. Such behavior leads eventually to the collapse of the
informal millet market and often leaves grain deficit households no other choice
than to rely entirely on commercial grain traders‘ in the area who are offering
unprocessed millet and/or maize meal at prices that are much higher than those
of informally traded millet.

Besides traditional containers, farmers also use sisal or polyethylene bags
for millet. Contrary to the expectation that larger quantities sell cheaper per kg

than smaller quantities, millet kg prices that are calculated from 50 kg bags are

 

‘ The term ‘commercial trader' includes traders from rural and urban
areas; traders for whom millet represents only a small fraction of activities and
traders for whom millet is the main trading good; traders that move between
different areas to exploit geographic price variations; and finally, traders that are
only operating in one location or area. This study defines all the grain that is
traded via intermediaries such as 'commercially traded grain'. The term
'commercial grain market’ aggregates all the activities of commercial grain trade
in a specified area.

1 16
mostly higher than those derived from smaller traditional measures. The main
reason for this is that millet packaged in bags is mainly bound for commercial
trade beyond the seller's direct neighborhood; i.e., with increasing social
distance to millet buyers, millet producers tend to base their millet pricing
decisions less on social obligations and more on prevailing prices in the
commercial staple food market.

Commercial food traders sell millet usually in the same bags in which they
acquired it. Traders determine their millet selling prices mainly on the basis of
their competitive position to other staple food traders; i.e., they are price takers.
Other factors that are used by traders to determine the millet retail price are their
individual millet supply prices, the general availability of millet in the traders’
vicinity, the prevailing price for maize meal, and traders' own profit expectations.

It is estimated that of all millet produced in the study zones only 10
percent are marketable surplus. This is by far not enough to cover the growing
demand for millet in the commercial food market . The three main reasons for

the growing demand for millet are as follows:

(1) low yields and production of millet over the last couple of years because
of drought;

(2) the scarcity of millet on the informal market two to three months after the
millet harvest;

(3) the high preference of the Ovambo and Kavango population for millet over

maize;

117

4.1.2. Millet price structure and variation during 1993

Locally produced millet and imported maize are the cheapest food staples
in northern Namibia. Table 4-1 demonstrates how far prices of these two grains
are apart from other cereals. The presented prices are average consumer
prices, i.e., prices that survey households actually paid at the end of the 'hungry
season' in April 1993. The price figures show that so-called 'modern foods’ like
bread, rice, and macaroni are between 50 and 150 percent more expensive than
maize meal, the dominant food staple in the commercial market. They also show
that, on average, informally traded millet is the cheapest of all calorie sources in

the study zones.

Table 4-1. Average food staple consumer prices from April 1993, by
zone (N5 per kg, percent of maize meal price)

 

 

OVAMBO KAVANGO
N$Ikg % NSIkg %
Infonnally traded millet (unprocessed) 1.10 92 1.27 92

Commercially traded maize (processed) 1.20 100 1.38 100

Commercially traded millet 1.36 1 13 1.87 136

(unprocessed)

Sorghum (unprocessed) 1.75 146 na na
Bread 2.00 167 2.24 162
Rice 3.30 275 2.97 215

 

Data Source: Namibian Millet Subsector Project Surveys, 1992193

1 18

The fact that informally traded millet is unprocessed does not reduce its
attractiveness to rural consumers with grain production deficits. First, almost all
rural farmers are processing their own millet manually and are, therefore, willing
to do the same with purchased millet. Second, the majority of the rural
population in the study zones prefers millet over maize (see also Chapter 5).
Third, although small, the price difference between informally traded millet and
maize meal is an incentive for many of the farm households that have only very
small cash resources.

The problem with informally traded millet is that it is often not available
when it is most needed, i.e., when many farmers face the emptying of their millet
reserves a few months after harvest. And because even commercially traded
millet is often unavailable in rural areas, no alternative is left for grain deficit
farmers than to buy maize meal. Figure 4-1 supports the argument about limited
availability of millet for purchase. It presents those time periods between
February 1993 and January 1994 during which both informally and commercially
traded millet were not available for purchase in the study zones (shadowed and
blackened time periods). The figure demonstrates that of the 19 areas where
the food price monitoring survey was conducted, only three areas, the urban
centers Oshakati and Ohangwenna in Ovambo, and Rundu in Kavango, had

millet available over a long period of time.

119

 

Sana-sass: axis-558:? ole—333E555:

 

 

 

 

 

 

 

 

 

Baa—«1.1.1....ilr « . 3 38.8655
55:9.
a»; 50 8.3.5.
8.5.
8.5..
$66.95. .2550 3535.
92.3.
flag 5.: 09.55.
8:95
52sz
Sabin
3:42.30 53 3.3.6
562320
53330
5:35.35. .3500 oos<>o
925.0
.922.
.250
. . . 5.20 on; 09.56
.558 a. t: 5.. 5&5..g§.£89.55.§.€ .2 63.8.58..." £9.92 .5 a...
.5.. .8.. .8.. so: .5.. do do 58 .93 a... .9... .93 .5.. .5.. .5.. .5... is are £2 it is is .aa a... flow 8.33555...
.5.... .. as... 2.52» 8:85: 8.8.2. 8213.5 a... 883..»

 

ooEcaEEoo oiEoo use 2.03:» an
.32 case... a... 82 basic... e823 .2... =82 .- .3E SEE e..- I..E 8382...... 8 3.3.2.-3.

 

TV 2:2".

 

120
Seasonal millet prices in Ovambo

Figure 4-2 and Figure 4-3 present the average millet prices for
(a) commercially traded millet and (b) informally traded millet collected in
Ovambo and Kavango during the food price monitoring survey. While
commercial prices were rendered between February 1993 and January 1994,
informal millet prices were only gathered between July 1993 and January 1994.
The reason for later gathering of informal producer prices was mainly based on
the observation that during the hungry season until mid-July, when farmers had
started to thresh their millet harvest, none or very little millet was traded among
households.

The observed millet prices in Ovambo stayed mostly close together and
moved in a narrow band between N$Ikg 1.10 and N$lkg 1.60. Within the
described price range, informally traded millet tended to be on the lower end
(N$lKg 1.20), while kg prices from 50 kg bags at retail level where close to about
N$IKg 1.50.

Seasonal millet prices in Kavango

In Kavango, prices of informally traded millet were relatively stable and
stayed, close to the prices of informally traded millet in Ovambo, averaging
N$lKg 1.15. However, the kg price of commercial traded millet varied with the
different seasons of the 1992/93 production year, reflecting the scarcity of millet
offered at the commercial millet market. In December 1992, millet retail prices

had been at a level of N$Ikg 2.00. When farmers' millet reserves from

121

Figure 4-2. Ovambo: millet prices on informal and commercial
markets from February 1993 to January 1994

 

 

 

 

 

4.00
E
g 3.30 .
O
2
'2
‘6 no -
o. .
h
2
0 commercial prices
0 1.00 -
s f —
— ~ ’
e \ ’ — I \ ’ \ , - \
E \ ’ — ’ — U
s 1.20 - -
2 '
infonnai prices
0.50 I I I I I I I I I I I I I I I I I I I I I I T 1
Feb.15 Apr.12 June 7 Aug.02 Sep.27 Nov."
Mums any 10 Jul.” Aug.” ccus Dec.20

Data Sources: Namibian Millet Subsector Research Project. 1992M!

Figure 4-3. Kavango: millet prices on informal and commercial
markets from February 1993 to January 1994

 

 

 

 

4.00
E /\\
3 3.30 . I ~ " , \
"f I \ commercial prices
o / \
9- 2.00 - A
5 l \ / " \
g / \V/ \
c 1.00 - "' — ’\ a ’
I
§
5 1.20 -
z
informal prices
50 I I I I I I I 1 I I I I 1 I I I I I I I I
Feb.15 Apr.12 June 7 1109.02 Sep.21 Nov.1 1
Mauls May 10 Jul.05 Aug.30 Oct.25 Dec.20

Data Sources: Namibian Milat Subsector Research Project. 1992193

122
subsistence production became more and more depleted in the beginning of
1993 and when the rain of the 1992/93 season had finally started, late in
January, indicating a severely depressed harvest later in the year, retail prices
soared until the beginning of May, reaching N$Ikg 3.80. Thereafter, millet prices
declined to the minimum of N$Ikg 2.00 at the end of July, when farmers had
finished threshing their harvest.

After the 1992/93 harvest, millet was only temporarily available at retail
level. Kilogram prices ranged between N$ 2.00 to N$ 3.00 N$. In October,
shipments of Angolan millet arrived via Ovambo in Rundu/Kavango. From then
on. retail prices declined to levels between N$Ikg 1.50 and N$lkg 2.00. Surplus
producers sold 50 kg bags at kg prices between N$ 2.20 and N$ 2.40 in rural
areas. This price range stayed constant even in November, when grain retailers
from Rundu acquired cheaper Angolan millet and sold it at kg prices below

N$ 2.00.

4.1.3. Price competitiveness of millet versus maize

The purpose of this section is to present the price competitiveness of
locally produced millet with imported maize as it occurred during the 1992/93
production year. This means that prices of unprocessed millet will be compared

with prices of processed maize, further called 'maize meal.’1 The reason for

 

‘ Maize, wheat, and sunflower are regulated by the Namibian Agronomic
Board. Almost all nationally produced maize is sold and transported via the NAB
to the main Namibian milling corporations: Namib Mills and Agro Mills. After

123
comparing prices of two grain types that are in different processing states is
based on the following two arguments:

First, it reflects the reality. Mechanical equipment to dehull and mill millet
to meal is not yet disseminated in rural areas of northern Namibia. Additionally,
as mentioned earlier, almost all farmers in the study zones are used to
processing their millet themselves. This leads to the assumption that at least
those farmers that (a) earn no cash income and (b) have few skills that would
help them find off-farm employment value their opportunity costs for millet
processing close to zero. in summary, the price difference between
unprocessed millet and maize meal is an important purchase criteria for grain
deficit households, because the convenience of purchasing grain in a 'ready to
cook' form is not yet highly valued in monetary terms.

Second, just adding a cost margin for processing to observed millet prices
in order to make them comparable to maize meal is not as easy as it sounds. By

adding millet processing as an option, various other issues have to be

 

being processed to meal and packaged into various bag sizes, maize is sold to
wholesalers and retailers all over Namibia. After Namibia's own annual grain
production is estimated, licenses for grain imports, mainly for white maize from
South Africa, are issued that enable the commercial traders (mainly the milling
corporations mentioned above) to take part in balancing the national grain
deficit.

Throughout the year 1993, maize meal was available in most areas of
Ovambo and Kavango. However, notable shortages occurred in some areas in
June and July during millet harvest time and in October. (see Figure 4-1) These
temporary shortages were probably induced by reduced consumer demand after
the harvest of farmers' subsistence crops and the reduced supply of maize meal
from the milling operations to northern Namibia's wholesale level at the same
time.

124
considered such as (a) the variation of production costs according to the
production technology employed and the production conditions of various areas;
(b) the variation of marketing cost (mainly transportation cost) according to the
state of the existing marketing system and the distances in different zones;
(c) the type and/or scale of millet processing technology used and where
processing operations are located; and finally, (d) whether millet meal has to
compete against maize meal at urban or rural food markets of the study zones.
To address these issues more adequately, a millet procurement cost model is
established in Chapter 6. This model allows the simulation of various production
and marketing conditions as well as the use of different processing technologies.

Because the introduction of millet processing in rural areas lies still in the
future, the existing level of price competition between unprocessed millet and
maize meal is presented in the paragraphs below.

Figures 4-4 demonstrates for Ovambo and Figure 4-5 for Kavango that
the price ratios between unprocessed millet and maize meal were more or less
stable during the course of 1993. They also reveal that under current production
and market conditions, unprocessed millet is considerably more expensive than
maize meal.

if available, informally traded millet constitutes the cheapest source of
calories for rural Ovambo and Kavango households. However, as already
described, after consecutive poor rainfall seasons, grain reserves of
mosthouseholds are so low that already a few weeks after grain harvest, informal

millet trade comes to a halt.

Namibian Dollar per kilogram

Namibian Dollar per kilogram

125

Figure 44. Ovambo: price comparison between unprocessed millet

and processed maize

 

 

 

 

 

 

4.00 '
3.30
2.00 '
1.00 '
1 .20
0.50 ' . I r T r r I . . r T T I r I I r r e r
Feb. Apr. Jun. Aug. Sep. Nov.
Mar. May. Jul. Aug. Oct. Dec
commerchlly ”mm ma’ze traded in make traded h
traded mllet traded let urban markets rural markets

Data Sources: Namb’sn Mht Subsector Research Project 1992/93

Figure 4-5. Kavango: price comparison between unprocessed millet and

processed maize

 

 

 

 

 

4.00 '
3.30
2.00 ‘
1.00
1 .20
0.50 ' I I I I - . t Y I I - . - r I 1 . ' Y I
Feb Apr. Jun Aug Sep. Nov.
Mar May. Jul Aug Oct. Dec.
commercially hformal maize traded 'ei ma'ze b'aded ‘ar
traded milet traded urban markets rural markets

Data Sources: Nambhn Mlet Subsector Research Project 1992193

 

 

126

Maize meal that is sold in larger package sizes at urban supermarkets
and wholesalers represents the second cheapest food for those farmers that
have relatively cheap opportunities to travel to and from urban areas. About 25
percent of rural Ovambo households and 15 percent of rural Kavango
households acquire maize meal from these distantretail outlets.

In urban centers of Ovambo, the commercially traded millet was only
slightly higher priced than maize meal. More than 80 percent of the millet offered
by urban Ovambo traders during 1993 originated from Angola. It can be
assumed that, similar to maize meal, not more than 25 percent of rural Ovambo
households buy millet at urban supermarkets and wholesalers.

Also in Ovambo, maize meal sold at small rural shops and rural
supermarkets was the most expensive food staple during 1993. Roughly 50
percent of all rural Ovambo households purchasing maize meal acquire it from
local shops and/or local supermarkets.

In Kavango, maize meal sold at small rural shops and supermarkets was
slightly more expensive than maize meal sold at the supermarket and wholesale
level in Rundu. About roughly 75 percent of all rural Kavango households that
buy maize meal acquire it from such local retail outlets

In Kavango, prices of commercially traded (unprocessed) millet were by
far the highest among all available food staples. During the first half of 1993,
millet cost about twice as much as maize meal. After the millet harvest, millet
prices were still 36 percent higher than maize meal prices sold in rural shops and

supermarkets. It can be assumed that under such conditions, only households

127
with high cash incomes could afford to buy larger quantities of millet from
commercial traders. From the comparison between millet and maize prices, two
important findings emerge:

First, at current production technology and market structure, millet is only
competitive against maize meal at informal markets within rural communities. lf
millet production could be increased without raising average unit cost, millet
would probably be available longer at the informal market and could thus
substitute for maize meal that is currently purchased to balance households'
grain production deficit.

Second, the prevailing prices at which Kavango surplus producers and
commercial food retailers offer millet are so expensive that only very few high
income households can afford to buy millet at the retail level. At such high price
levels it will eventually be attractive to transport Angolan millet via Ovambo to
Kavango. Only if, through an increased number of millet surplus producers in
Kavango and through an increased competition on the retail level, profit margins
of both surplus producers and commercial grain traders are reduced, will millet
'retail prices decline to more affordable levels. This will eventually make millet

shipments from Angola less profitable.

4.2. Commercial millet marketing‘
This section describes the characteristics of study zones' millet traders,
analyzes their trading practices, and discusses trade attitudes about the

prospects and limitations of commercial millet marketing. The findings presented

 

‘ In this study, the term 'commercial millet marketing' is broadly defined as
grain transactions from producers to consumers that involve at least one
commercial intermediary.

128
in this section are important due to two reasons: (a) because they provide, for
the first time, information about a group of entrepreneurs (whom policymakers
assumed don‘t exist) that are engaged in millet trade, and (b) because they
describe the market environment farmers face when they want to market their
millet or have to purchase grain.

The data for this section come from the millet trader survey conducted in
the study zones after the millet threshing period between October and November
1993. A total of 58 rural and urban millet traders were interviewed. It is
estimated that the survey covered 70 percent of the commercial millet trade
volume in urban Ovambo and Kavango and about 8 percent of rural Ovambo

and 21 percent of rural Kavango respectively.‘

4.2.1. Characteristics of commercial millet traders
Of the millet traders surveyed in Ovambo, 22 operated in rural areas and

9 traded millet in one of the three peri-urban townships Oshakati, Ondangwa, or

 

‘ There are not many urban grain traders in the study zones. They are
generally known and, therefore, easy to identify. From the extent of their other
business activities, it is also possible to make some rough estimates about their
millet trade volume, even if they refused to be interviewed or were not included in
the millet trader survey due to other reasons. It was not possible to establish a
sample frame of rural commercial grain traders in Ovambo and Kavango.
However, during the household surveys in 16 different communities, respondents
were asked to identify grain traders in their area. From the number of grain
traders identified for each survey site, a rough estimate was made for the whole
of Ovambo and Kavango. The resulting estimates are that Ovambo has about
260 rural millet traders and that Kavango has about 120 rural millet traders. The
average trade volume of the randomly selected (within geographic strata) and
interviewed millet traders was extrapolated on the basis of these estimates.

129

Ohangwenna. In Kavango, 23 rural traders were interviewed and 4 millet traders
from Rundu. Most of the urban millet traders are wealthy businessmen who own
wholesale operations or several supermarkets. Most of the smaller urban food
retailers identified as potential survey participants sold only maize meal and no
millet to their customers.

Four types of millet retail outlets could be identified in the rural areas:
(1) The majority of the rural outlets were privately owned ‘cuca shops' (small
retail shops trading basic necessities) and/or somewhat larger and better
stocked 'supermarkets'. Most of the owners of these rural retail outlets were
also larger farmers who provided part of the millet they sold to their customers
from their own fields.
(2) The second type of millet trading places did not differ physically from the
cuca shops or small supermarkets, however, they belonged to a chain of retail
outlets owned by a wholesaler or by the owner of a large supermarket from one
of the northern peri-urban areas.
(3) The third type of millet traders are rural churches. Some of these accept
annual membership fees to be paid in the form of millet. The conversion factor
between the official church fee and the accepted millet price values millet
relatively cheaply (usually N$ 1 per kilogram). Because churches resell the
acquired millet during the hungry season to members and non-members at the
same conversion factor, the impression that millet has this relative low value is

confirmed again, but this time to a larger group in public.

1 30
(4) The last type of millet traders are large, urban based wholesale and retail
businesses. In urban Ovambo, 7 of the 9 interviewed millet traders were
wholesalers and two were owners of large supermarkets. In Kavango, 3 of the 4
interviewed millet traders owned supermarkets, the fourth owned the only private
wholesale operation in Rundu. The fifth wholesale operation trading grain in
Kavango is the wholesale operation of the Namibian Development Corporation
(NDC). This operation is kept competitive in the market because many of its
employees are paid by the government, which permits its sales prices to be set
below actual costs.

Most of the interviewed millet traders were involved in trade branches
other than millet marketing. Table 4-2 presents the share that millet sales have
on rural and urban millet traders total business. The figures indicate that millet
trade has a larger share of total sales in rural trading operations than in urban.
Additionally, maize meal trade has a significantly larger share than millet among
Ovambo millet traders' sales volume (not mentioning those urban food retailers
who market only maize meal and no millet at all), compared to millet traders in
Kavango.

WIth very few exceptions, commercial millet trade does not have a long
tradition in Ovambo and Kavango. Although 84 percent of all interviewed
Ovambo traders had started their general trading business before 1982, about
65 percent started to trade millet just recently after the severe drought of the
1991/92 production season. During this year, millet prices and profit margins

were high. Of the interviewed Kavango business people, 52 percent started their

131

millet trade after the 1991/92 drought, although 74 percent had started their

general business already before that time.1

Table 4-2. Business activities of rural and urban millet traders,
by zone and location (in percent of total trade volume)

Rural Millet Traders

Urban Millet Traders

 

 

Ovambo Kavango Ovambo Kavango
Millet 19 27 1 3 1 3
Maize meal 24 12 31 6
Other food and
household goods 25 27 16 68
Consumer goods
8: services 13 25 24 0
Alcohol 13 7 10 9
Cooked food 1 2 6 4
Total 100 100 100 100

 

Data Source: Namibian Millet Subsector Project Surveys 1992/93

 

‘ At this point, it can not be determined whether only the high profit
margins due to the 1991/92 drought triggered the expansion of commercial millet

trade, or whether Namibia's independence in 1991 influenced this development

also.

1 32
Almost all millet traders confirmed during the interviews that the demand
for millet and also for maize is suppressed after good rainfall seasons,
includingafter the 1992/93 harvest, on which this study focuses, while after poor

production seasons, like the 1991/92 drought year, millet trade flourishes.‘

4.2.2. Traders' millet acquisition after harvest

The following section describes market transactions between millet
producers and grain traders from the viewpoint of the latter. The motivations
behind this section are (a) to familiarize the reader with grain traders' options to
acquire millet; (b) to reveal the problems commercial traders face trying to
aggregate millet supply from the dispersed locations of surplus farmers; and
(c) to lay the basis for farmers' description of their own millet marketing activities
presented in Section 4.3. below. The section itself is divided into eight
subsections that cover the following topics: (1) millet acquisition period, (2) types
of millet suppliers, (3) transportation, (4) amounts traded, (5) payment modes
and credit availability, (6) traders' reference prices, (7) comparison of supply

prices, and (8) competition from Angola.

 

‘ After good millet harvest years, most farmers tend to replenish their
depleted grain reserves against potential droughts. The few very large and

influential farmers producing marketable millet surpluses every cropping season
are financially hurt by the decline in millet market prices after good harvest years.
They tend to favor the NAB proposal suggesting guaranteed producer prices
every year.

1 33

4.2.2.1. Acquisition period

Most rural and urban traders start to buy millet between June and
September after millet harvest and threshing. At that time, traders have to judge
for themselves whether the harvest results were generally poor, i.e., whether the
demand for commercially traded millet will develop soon after the grain harvest
or whether farmers' harvest was good, i.e., whether they could sell millet only in
later months when the new grain harvest is close. Those traders with relatively
limited financial resources try to buy millet immediately after harvest when
farmers are happy about their new crops and are willing to accept lower producer
prices. Traders with larger financial capacities acquire their millet supply in
portions over a longer period of time. This helps them to better judge where they
can get the best offer and how much grain they are able to sell until the next
millet harvest arrives. However, by the end of December, almost all millet

traders have ended their millet acquisition from producers and other traders.

4.2.2.2. Grain traders' millet suppliers

in urban areas, larger producers or long distance traders from Angola
approach traders at their business site to offer millet for sale. If millet producers
can not organize transportation to the trader they want to sell to, they ask the
trader to transport millet from their fields to the retail outlet in town. The
transportation costs thus incurred by the traders are usually reflected in the millet

producer price they offer.

134

in rural areas, the transaction between millet producers and traders
occurs on three different levels: (1) at the homestead of the farmer who offers
millet for sale; (2) at the retail outlet of a rural trader; and (3) when the millet
producer sells his millet surplus him/herself and offers it either from his
homestead or from a central place in his or her community.

Most farmers offering their millet to traders approach them by chance and
inquire whether the trader is interested in buying. Sometimes the farmers know
already from other producers about specific traders' willingness to buy. The few
traders reaching out in rural areas to purchase millet tend to first approach the
farmers they know from previous production seasons. However, farmers also
invite traders sometimes to come to their place to buy their surplus.

The type of producers from which traders acquire millet is often correlated
with both the location of the traders' market outlet, i.e., whether the trader is
located in rural or urban areas, and the zone in which the trader is located, i.e.,
in Ovambo or in Kavango.

The rural millet traders from Ovambo claim that 65 percent of the acquired
millet comes from ‘marginal farmers’, who are either in urgent need of money or
in need of other goods. (Table 4-3.) Only 15 percent of their millet is supplied by
‘middle scale millet producers’ and only 18 percent by “large and wealthy
farmers’. Middle scale millet producers usually try to balance their millet deficit
production from one year with surplus production from previous years. They are
less inclined to sell their annual millet surplus. If they are in need of money, they

acquire the necessary resources most often through activities that are not

135
directly related to millet production, like plowing or transportation service,
woodwork or handcrafts, etc- Rural Ovambo traders acquire very little millet

from Angolan farmers near the border.

Table 4-3. Millet suppliers of rural and urban grain traders in
Ovambo and Kavango (percent of trade volume)

 

 

 

Rural Millet Traders Urban Millet Traders

Ovambo Kavango Ovambo Kavango
Large farmers" 18 61 12 87
Middle farmers' 15 16 2 13
Marginal farmers“ 65 20 3 0
Angola 2 3 84 0
Total 100 100 100 100

 

Data Source: Namibian Millet Subsector Project Surveys 1992/93

* see definitions in the last paragraph on the previous page

Rural millet traders in Kavango receive 61 percent of their millet from large
surplus producers. Another 20 percent of millet is received from marginal farm
households and only 16 percent comes from middle scale producers.

One of the most surprising results of the millet marketing survey is that
during the year 1993, millet traders from urban Ovambo acquired 84 percent of
their millet from Angola. This millet was legally shipped across the Angolan
border and custom fees were paid for most of it. Most of the residual millet
acquired by urban Ovambo traders came from large and wealthy farm

households.

1 36

Similar to traders in rural Kavango, wealthy farmers with large surpluses
supply about 87 percent of urban Kavango traders' millet. The remaining 13
percent of millet is received from middle scale farmers. The supply of larger
quantities of millet from Angola is of less importance to Kavango. This is
probably due to the fact that there are no direct transportation routes from
Kavango into southern Angola. The main routes from Kavango to Angola's
population and grain production centers are via Ovambo, which has a major

border transit in the town of Ohangwenna.

4.2.2.3. Transportation

According to statements from rural traders, they arrange transportation for
their millet suppliers in at least half of all millet transactions. The average radius
of rural traders' millet acquisition is in about 70 kilometers in both study zones.
ln Kavango, roughly 45 percent‘ and in Ovambo, roughly 70 percent of the millet
acquired by rural traders is transported by motor vehicles.

In contrast to rural areas, most of the millet acquired by urban traders is
transported to them by producers that own or organize transportation
themselves, or, as is the case for Angolan millet, transported by Angolan long-
distance traders. The average distance between urban traders in Kavango and

the origin of millet production is 120 kilometers. Because urban millet traders

 

‘ In Kavango, the use of an oxen sledge that has the form of a canoe is
often the only means by which farmers can transport grain from their fields in the

bush to potential market places.

1 37
from Ovambo acquired, in 1993, most of their millet from the Angolan province
Huila, Lubango, the average distance between them and the millet origin is
about 500 kilometers. Almost all of the millet transported to urban traders in
Ovambo and Kavango comes on trucks or other motor vehicles.

The average transportation cost per ton of millet produced and marketed
is N5 90 in Ovambo and N3 60 in Kavango. Transportation cost of farmers to
local traders are on average only a third of the cost to grain traders in urban
centers. Surprisingly cheap is the transportation of millet between Ovambo’s
central food market Oshakati and Rundu in Kavango. The good road connection
between Oshakati and Rundu (600 kilometer) allows the use of large 30 ton -

trucks at the cost of only N$ 65 per ton.

4.2.2.4. Amounts traded

After the 1992/93 grain harvest, most of the rural Ovambo millet traders
acquired between 1 and 5 tons of millet from producers. Only a few purchased
between 5 and 50 tons, and only one rural trader in our sample stocked more
than 150 tons of millet for resale. in rural Kavango, only a quarter of the survey
millet traders acquired less than 5 tons of millet. About one half of them acquired
between 15 and 50 tons of millet for resale. The last quarter acquired between
50 and 200 tons of millet.

In urban Ovambo, the amounts of millet acquired by traders during 1993
for resale varied significantly. At the end of the millet acquisition period,

commercial millet traders' stocks ranged from 10 tons to more than 600 tons of

138
millet. Also, among the four major grain traders in urban Kavango (Rundu), the
millet supply for selling varied significantly. The smallest amount acquired was
about 15 tons, while the largest amount that was in stock by the end of
December was 250 tons of millet.

With the exception of rural Kavango traders, most traders acquired after
the 1992/93 production season less millet than in previous years. The three
reasons most often given for this were (1) the harvest in the trading area was
good, and therefore, the demand for millet was expected to be low; (2) the prices
for millet from Angola had increased from the previous drought year 1992/93 with
the introduction of custom fees; (3) the traders' own millet production was less
than usual, or his/her financial situation did not allow the purchase of as much

millet as usual.

4.2.2.5. Payment modes

To see whether commercial millet trade is restricted through cash flow
problems on the side of the traders, questions were asked about how traders are
paying their millet suppliers and whether they have opportunities to borrow
operating funds for their grain marketing activities. Two-thirds of the interviewed
traders in rural Ovambo and one-third of those in rural Kavango pay their millet
suppliers immediately in cash, while the reminder pay their suppliers both ways,
with cash and/or in kind. Among the interviewed urban traders from both zones,

50 percent pay their suppliers in kind and 50 percent pay them in cash.

int-'4- - ‘ }”‘-x—

1 39

Only 30 percent of rural millet traders from Ovambo and 15 percent of
those from Kavango claimed they could get credit for their trade business if
needed. Only 25 percent of the urban millet traders in Ovambo and half of the
millet traders from Rundu claim they could receive business credit.

Surprisingly, most of those who claim they had access to credit named the
government as their potential source. it has to be assumed that the Namibian
Development Corporation (NDC), that has its own wholesale businesses in the
study zones, offered private traders operation loans in the past and that traders
confused this institution with the Namibian government. Only two rural millet
traders named the Namibian Development Corporation directly as a credit
source. Two other rural traders from Ovambo indicated commercial banks as a
potential credit source.

From the way traders responded to the question about the accessibility of
credit, it seems as if (a) business people in northern Namibia who already
accumulated considerable wealth are assured of getting loans but chose mostly
not to take them; (b) business people who have very limited funds to acquire
larger amounts of millet from producers for marketing tend to be excluded from

the support through commercial loans.

140
4.2.2.6. Reference prices

In rural areas, about 50 percent of the traders determine the price at
which they buy millet from producers alone, i.e., without negotiating. The other
half of the rural traders stated that their millet acquisition prices depend on their
individual negotiations with the supplier. All urban millet traders stated that they
have to bargain with their millet suppliers about millet supply prices.

The largest group of rural traders uses the millet bucket price, i.e., the
informal millet exchange price used among community members as reference for
the producer price they offer. Other rural traders use their targeted retail prices
together, minus accrued transaction costs like transportation and handling, to
determine the price they can pay to their suppliers. All urban millet traders in
Ovambo stated that they determine their millet acquisition prices according to the

millet retail prices of their competitors.

Table 4-4. Traders' references to determine millet acquisition prices
(percent)

 

Rural Millet Traders Urban Millet Traders
Ovambo Kavango Ovambo Kavango

 

Price Indicator

 

informal prices 50 33 0 33
competitors' prices 12 19 100 0
own cost 18 30 0 33
maize meal price 0 0 0 33
other 15 18 0 0
total 100 100 100 100

 

Data Source: Namibian Millet Subsector Project Surveys 1992/93

141

In urban Kavango, the grain market structure resembles typical features of
an oligopoly. Of the four millet traders, the largest grain trader and price leader
for millet and maize meal sales did not want to disclose his pricing strategy. One
of the other three claimed that it is essential to know the potential millet retail
price at which he could sell to his customers before he could decide about the
price he was willing to pay to his suppliers. it was explained that the demand for
millet in Rundu was closely linked to the price of its closest substitute, maize
meal. Millet can be sold with a premium that does not exceed 10 to 20 percent

of the offered maize meal.

4.2.2.7. Supply prices

To gain more knowledge about the prices millet traders pay to their
suppliers, traders were asked to state their acquisition prices from the last
production season 1991/92 and in the ongoing production season 1992/93. The
comparison of their answers by zone and by location (rural/urban) leads to the
following observations (see also Table 4-5.):

First, as expected, millet prices of the 1991/92 drought year were

generally higher than the millet prices from the average production year 1992/93.

Second ,~ on average, millet traders from Ovambo pay lower supply prices
than those from Kavango. This finding is surprising because Kavango has, on
average, better rainfall conditions than Ovambo, which led to the assumption

that Kavango's productivity in grain production is higher, and thus its millet prices

142
are more likely below those of Ovambo. However, Ovambo has a very well-
developed wholesale and retail system that channels grain imports efficiently to
deficit households in rural areas. As already mentioned, during the year 1993
year, urban Ovambo, which is relatively well connected to Angola by road,
received large amounts of commercial millet imports from Angola. As will be
discussed in later sections, these imports are very price competitive compared to
local producer prices and thus reduce the general price level of grains in

Ovambo.

Table 4-5. Acquisition prices of millet traders in 1991I92 and
1992/93, by zone and trader location" (NS per ton“)

 

 

Production Rural Millet Traders Urban Millet Traders
Year Ovambo Kavango Ovambo Kavango
1991I92 1020 1860 920 na

1 992I93 1030 1200 800 970

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

" Include millet transportation costs to traders' market outlets.
“ During 1993, the exchange rate of the Namibian Dollar averaged 3 N$lUS$.

Third, on average, urban traders face lower millet acquisition prices than
rural millet traders. This finding has several explanations: (a) As described
above, transportation costs per unit of distance are about two to three times
higher in rural areas than in urban zones. (b) Larger producers are more likely to
sell their millet surpluses in bulk to urban centers where commercial traders'

bargaining power is relatively strong. (c) Smaller rural grain traders have to offer

143
attractive prices to prevent potential millet sellers from bypassing them and
organizing transportation to the next urban center. (d) Relatively cheap millet
imports from Angola not only enter Namibia via the urban areas of Ovambo, but

are shipped effectively to Rundu in Kavango.

4.2.2.8. Millet imports from Angola

As demonstrated in Table 4-3, urban millet traders from Ovambo acquired
most of their millet from Angola. The comparison of (a) millet supply prices
(including transportation costs) of selected traders from urban Ovambo buying
more than 70 percent of their millet from Angola with (b) prices of selected
Kavango traders buying most of their millet from larger Namibian farmers leads

to the following results:‘ (see also Table 4-6.)

1) In 1993, Angolan millet supplied to urban Ovambo traders was between
20 and 140 percent cheaper than millet supplied by larger millet producers

in Kavango.

2) The price gap between Angolan millet and millet from Namibia increases
with the trade volume of individual traders; i.e, millet supplied in bulk from

Angola was very cheap in the 1993 post harvest season.

 

‘ It seems quite possible that even if supply prices for domestic and
Angolan millet were equal, urban traders from the study zones would prefer

buying millet in bulk from Angola than trying to purchase it from the dispersed
offerings of domestic surplus producers.

144

Due to the resumption of Angola's civil war in 1993, researchers were
unable to visit Angola and collect information on millet exports to Namibia.
However, Namibian traders explained that Angolan millet originates from the
province Huila, Lubango about 500 kilometers north of Oshakati in Ovambo.
The millet production zones are easy to reach via the long-distance roads from
Ovambo's boarder town Ohangwenna. Millet is produced on large-scale
commercial farms that are fully-mechanized and produce under more favorable

climatic conditions than exist in northern Namibia.

Table 4-6. Millet acquisition prices of selected grain traders,
by millet production origin and trade volume (NS per MT)

 

Millet supply from

 

 

Angola‘ Namibian producers”
Trader no. Trade Price per Trader Trade Price per
8. location volume ton in NS no. volume ton in N8

In tons 8. in tons

location

1. Urban 1. Urban
Ovambo 60 430 Kavango 200 1000
2. Urban 2. Rural
Ovambo 150 500 Kavango 40 800
3. Urban 3. Rural
Ovambo 70 900 Kavango 10 1150
4. Urban 4. Rural
Ovambo 50 700 Kavango 7 1200

 

Data Source: Namibian Millet Subsector Project Surveys 1992/93

* Average millet supply price of urban Ovambo traders acquiring more than 70 percent
of their millet from Angola.

" Average millet supply price of rural and urban Kavango traders acquiring most of their
millet from larger millet producers.

145

Three reasons explain the large millet exports from Angola to northern

Namibia at the end of 1993:

(1)

(2)

(3)

The war in Angola blocked access to grain markets in towns and cities so
that northern Namibia with its millet deficit production became a welcome

market alternative for millet producers.

Due to the poor economic state of Angola's economy, its currency was
undervalued to such an extent that Angolan agricultural products became

competitive on the northern Namibian staple food market.

The warfare limited south Angola's access to imported goods via Angolan
harbors. Thus, the access to consumer goods via trade with northern
Namibia became attractive. Many of the Namibian wholesalers bartered

Angolan millet for consumer goods that originated from South Africa.

Although it is not clear whether the supply of relatively cheap Angolan

millet will continue in the years to come, the findings about the millet imports in

1993 add a new and important piece of information necessary to judge the

development prospects of the Namibian millet subsector. Before the 1992/93

grain production year, only maize coming either from South Africa or from

Namibia's own commercial farmers was considered a main competitor of millet

produced in the communal north. Now, imported millet from neighboring Angola

is added to the competition. The main question, however, is still about whose

grain will eventually be bought by the increasing number of rural and urban

consumers in northern Namibia who are in need of cheap food staples.

146
4.2.3. Millet traders' value adding and marketing activities
This section describes commercial millet traders' value adding and selling
activities: (1) storage practices, (2) cleaning, processing and packaging,
(3) selling period, (4) characteristics of millet consumers, (5) setting of consumer

prices, and (6) profit margins.

4.2.3.1. Storage practices

With the exception of farmers who sell their own millet to consumers, most
commercial millet traders intend to store millet for less than one year, i.e., to buy
millet briefly after grain harvest and sell it before the next millet harvest starts. It
can be assumed that the short selling targets are prompted by traders' fears of
market price variations caused by rapid and unexpected changes in millet and
maize supplies. During the period of millet acquisition, commercial millet traders
face the risk of underestimating harvest yields and overestimating the demand of
those farm households that usually become net grain buyers before the next
millet harvest. Unexpected low demand for commercially traded millet might
force traders to carry over large millet stocks that can only be sold if poor
production seasons follow within one or two years. Throughout an extended
storage period, not storage losses, but the retention of valuable trading capital in
millet stocks is most damaging to the millet traders.

Most rural millet traders store millet in bags in houses with roofs of
corrugated iron. However, about a third of the traders also store millet in

traditional granaries. Most of the urban millet traders own large storage buildings

147
where they store, besides bags of millet, large quantities of maize meal, sugar,
alcohol, as well as other food and consumer goods for their trade business. The
average storage capacity of the surveyed millet traders was unexpectedly high.
Similar to rural farmers' tradition of demonstrating wealth by accumulating large
herds of livestock and positioning a large number of big granaries near their
homestead, the ownership of large storage facilities seems to be an important
status symbol among communal businessmen.

Rural millet traders have an average storage capacity of roughly 250 tons
of grain in Ovambo and about 530 tons in Kavango. Urban millet traders in
Ovambo reach an average storage capacity of about 2400 tons of grain and
about 2000 tons in Kavango. Many of the storage facilities of the surveyed
business Operations were underutilized. The high grain storage capacities in the
private sector indicate that at the current level of market supply, little need exists
to construct extra storage facilities for regional grain reserves against drought
years.

About half of the rural traders in Ovambo and a sixth of the rural traders in
Kavango claim that they have storage losses from insects. Among the urban
millet traders, almost none reported insects as a problem for grain storage. Most
owners of large storage facilities fumigate them at least once a year.

Storage costs for millet are difficult to acquire from traders. Most storage
buildings in which millet is stored are also used for other trading goods.
Additionally, traders do not consider the cost of constructing storage facilities

attributable to their millet business. However, some traders were able to

148
attribute the cost for unloading and handling, as well as the cost of security, to
their millet business. According to these estimates, labor and security costs are

between 15 to 20 Namibian dollars per traded ton of millet.

4.2.3.2. Cleaning, processing and packaging

All interviewed traders were asked specifically whether they add value to
millet during the trading process. Most traders said they didn't. The majority
judges only the quality of the millet when it arrives from the suppliers to
determine the acquisition price accordingly. If necessary, millet grain is refilled
into new bags. Together with the labor and security cost estimated above,
average millet marketing costs are estimated to be not higher than NS 100 per
ton.

Some urban traders in Ovambo have experimented with processing millet
with either small-scale hammer mills or with the roller mills they use to grind
maize to maize meal. But none of them seemed to have developed a
mechanical millet processing system that worked, so that at the time the
marketing survey was carried out, none of their processing operations were
running.

Some rural millet traders from Kavango who sell part of the millet
production of their own farm explained that they clean and grade their millet at
threshing time before they ship it to their market outlet. However, they

considered these activities more as a normal part of the production process,

149
rather than a part of their millet marketing activities. They would also not charge
extra fees for this service.

Among the interviewed traders, only two exceptions with regard to millet
processing could be identified: The first is the commercial NDC farm in
MuseselKavango, where processing equipment for millet grading, cleaning,
dehulling, milling and packaging is redesigned and adjusted for large scale millet
processing and marketing.‘ The second is the self-help “Katemo Farmers'
Marketing Cooperative” in Rundu/Kavango. In 1993, the cooperative imported,
with the help of the Canadian development aid, the same millet processing
technology from Zimbabwe, but at a smaller level than NDC had developed its
operation from. The cooperative competes with the NDC operation for the same
millet suppliers and customers. Its main problems are collecting and transporting
millet to its retail outlet in Rundu and effectively managing and selling the
aggregated millet reserves. Because many Kavango people still perceive the
NDC operation (formerly 'First National Development Corporation') as an
instrument of the former apartheid system, Katemo has declined so far any
technical collaboration with NDC.

The reasons for such limited activity in the field of millet processing are

not obvious. Lack of capital can not be the problem among larger grain traders

 

‘ From the proposal made by the Agronomic Board to use a single market
channel system for millet in northern Namibia, it became obvious that the large-

scale processing facility in MuseselKavango of the Namibian Development
Corporation was meant to become the only official processing facility for millet
produced in Kavango. The proposal suggested also that a similar operation
should be built in Oshakati with government funds.

150
in urban areas because many of them are expanding their businesses by
investing in new branches like gas stations, bakeries, car repair shops, etc..

Also insufficient demand for processed millet seems not to be the
problem. in one of the household survey areas of the Millet Subsector Research
Project, UNICEF had tested the adaptability of small-scale hammer mills in rural
areas. The purpose of these tests were to offer rural women relief from the
burden of the second millet processing step after dehulling, namely millet milling.
After the 1992/93 production season, in two of the four UNICEF test sites, an
additional small-scale hammer mill could be bought from the funds, which
organized groups of female mill operators had generated with the first small-
scale mill.‘

In the area of urban Ovambo, two successful millet service processing
operations could be identified during the millet trader survey in November 1993.
One of the two is the well-known roller mill operation in Oniipa, located 30
kilometers east of Ovambo's center town Oshakati. This mill has been operated
by a church for more than 20 years. The advantage of using a roller mill is that
both millet dehulling and millet milling can be done with the same machine. So
many farmers organize transportation to Oniipa in order to process their grain

that waiting periods are up to three weeks after harvest, due to processing

 

‘ An economic analysis of this pilot project is not possible because
UNICEF had provided the first small scale-milling machines, set the salaries for

the operators and prices for the milling service, paid the fuel, and organized the
maintenance of the machines without keeping adequate records of costs and
benefits.

1 51
backlogs.‘ The second identified millet processing service is located in Oshakati.
It is operated by a mechanic who owns, besides a car repair shop, an electrical
power driven hammer mill. Although hammer mill processing permits only the
milling of already manually dehulled millet, many urban customers are using this
service.

The example of the UNICEF project, together with those of the two
privately operated millet processing services in central Ovambo, indicates that
mechanical millet processing could be further developed in rural and urban areas
of the study zones in the future. However, further research about the suitability
of various technologies is necessary. Potential millet meal procurement costs
under various scenarios of millet production, marketing conditions, and
processing technologies are calculated with the help of a simple cost accounting
model in Chapter 6. The simulation results reveal whether millet meal can
eventually be offered at prices lower or at least close to those of maize meal
already marketed in Ovambo and Kavango.

Independently of the calculation of potential millet meal procurement
costs, UNICEF's maintenance problems with the small-scale hammer mills
introduced to the rural areas and the fact that commercial grain traders have
tried, but not continued to use, small-scale hammer mills for commercial millet

processing, are an indication that not only mechanical processing equipment

 

‘ According to information from Ms. Gomez, a former ICRISAT food
technology expert, the millet meal from Oniipa is of high quality and it might

justify the development of similar mills for the commercial sector.

1 52
should be disseminated once it proves to deliver price-competitive millet meal,

but also appropriate skills in maintenance and business accounting.

4.2.3.3. Traders' selling period

After traders have acquired millet from their suppliers, their main concern
is to sell it before the next grain harvest. The beginning and the duration of
commercial millet selling from traders to consumers varies between Ovambo and
Kavango and according to the traders' location in rural or urban areas. The peak
period of millet selling starts in Ovambo in August, while in Kavango, millet
selling starts mainly in October. In both zones, rural traders are the first ones to
sell millet. Urban traders follow about two months later. Parallel to the increase
in millet sales, maize selling picks up again. However, while most commercial
millet selling activities come to an end by mid-January, maize meal sales
continue until the next millet harvest arrives in May or June. Two explanations
can be given for the early end of the millet selling season. First, and most
important, traders millet reserves are usually depleted by the end of December.
Second, due to the start of the rain season (usually) in November in Kavango
and (usually) late December in Ovambo, the first vegetables from subsistence
production are available for consumption, which reduces the need to purchase
grain.

Traders explain the beginning of the peak selling period with the emptying
of rural households' millet stocks. In rural Kavango, half of the interviewed

traders claimed that farmers come during this period because they are in need of

153
seed. About two-thirds of the urban Ovambo traders claimed that, due to the
beginning of plowing, people need more food and drinks made from millet.
Finally, urban traders from both zones explain that migrant workers return in
December to the north to celebrate Christmas with their families. On their way
back, they tend to buy millet in Oshakati or Rundu to asure that during the
religious festivities their household has enough millet porridge and millet beer to
feed visiting relatives and friends.

The time gap between the start of commercial millet selling in Ovambo
and Kavango leads to the conclusion that, on average, rural Ovambo
households experience a larger grain production deficit than rural Kavango
households, which manifests itself in a significantly earlier beginning of millet

buying from commercial traders in Ovambo.

4.2.3.4. Characteristics of millet consumers

Although most of the interviewed commercial traders sell more or at least
as much maize meal as millet to their customers (see Table 4-2 above),
93 percent of them said that the‘ majority of their customers favor millet for
consumption over maize. Of the other traders, 5 percent said that consumers
were indifferent in their taste preference between millet and maize, and only
2 percent said that people prefer maize over millet.

Traders' differentiation between customers that actually buy millet versus
those that buy maize meal is presented in Table 4-7. A large majority of traders

claimed that all people in their zone prefer millet over maize (83 percent of the

154
answers), and that is why they buy millet. However, some of the traders were
more specific and reduced the set to all ‘rural’ people (10 percent) or ‘poor’
people who can not always afford to buy millet to balance their grain deficit, but
buy it if they want to make 'Oshikundo', a fermented millet drink, or millet beer

(3 percent).

Table 4-7. Characterization of typical millet and maize meal buyers
(percent of traders, n =58)

 

 

 

Consumer Characteristics Buy Millet Buy Maize
All people 83 0
All rural people 10 0
Poor people for bear brewing 3 0
Poor people in urgent need of food 0 16
Migrant workers 0 8
Town residents 0 8
Urban employees and rich people 0 14
Other 4 1 3
No one 0 38
Total 100 100

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

When asked to characterize consumers who buy maize meal, 38 percent
of the millet traders claimed that none of their maize meal buyers really prefer
maize over millet. The next largest group of traders (16 percent) claimed that
only those people who can not afford the high prices of commercially-traded

millet buy maize meal, i.e., the poorer section of the people. From these

155
answers, it became more obvious that for most people, the decision to buy maize
meal and not millet is a forced decision; i.e., maize is bought because it is more
affordable.

Few traders associated maize buyers with people that really prefer to eat
maize meal over that of millet. People who really prefer maize over millet were
mainly characterized as rich and urban-employed people (14 percent).

However, town residents in general and migrant workers were also mentioned as
typical maize meal consumers (each 8 percent). Only a few traders were able to
make a rough estimate about the population share that actually prefers to buy
maize meal. Urban millet traders estimated that between 10 and 30 percent of
the population belong to those who are rich and urban-employed people. Most
rural traders estimated that no one or less than 1 percent of the total population
prefer to eat solely maize without millet meal mixed into it.

The amount of millet that individual people buy from commercial traders
during a single purchase is generally larger than that which is traded directly and
informally among rural households. Most of the commercial traders named
quantities between 50 and 250 kg as the typical amount millet buyers acquire.
None of the traders indicated average selling quantities below a 50 kg bag. Still
a high portion of millet traders (Ovambo: 29 percent, Kavango: 25 percent)
claimed that most of their millet buyers acquire more than 250 kg of millet at a
time. These relatively large amounts and the notion that commercially traded

millet is, for many customers, too expensive indicate that millet traders target a

1 56
prosperous clientele that prefers the traditional taste of millet over the taste of

maize.

4.2.3.5. Setting retail prices

Of the interviewed traders, 90 percent stated they set their millet retail
price unilaterally. Another 7 percent claimed that within certain limits their selling
prices are determined through negotiations with the buyers. A third of the
traders said they would give discounts to buyers of larger quantities.

Grain traders use three main measures to set their milling selling prices.
First, traders' most often used reference for retail price determination is their
already accrued cost, i.e. the acquisition price traders paid to their millet
suppliers. Second, traders observe the general price development for millet in
the market, and third, they observe in particular the prices of their direct
competitors.

When asked why commercially traded millet is higher priced than maize
meal, traders named three major factors: (1) high production cost (especially for
labor, plowing, threshing, and transportation), (2) consumers' strong preference
for millet, and (3) the general scarcity of millet and the small number farmers

who are willing to sell.

4.2.3.6. Profit margins
Although millet traders make factors other than their own pricing practices

responsible for high millet retail prices, differences between their gross margins

1 57
and their marketing costs indicate that at least some of the commercial millet
traders take advantage of the general scarcity of millet.

During the millet marketing survey, only 26 percent of the traders were
willing to disclose both their 1993 millet acquisition/supply costs, including
average cost for transportation and handling, and their millet selling prices.
Because in most cases millet represents only a small fraction of traders’ total
retail operations, storage costs are negligibly low. Because most traders neither
calculate their marketing cost nor do they incur costs for processing grain, their

marketing cost had to be estimated.

Table 4-8. Traders' millet acquisition costs and profit margins in 1993,

 

 

by trade volume (N3 per MT)
Trade Volume Categories
1 to 50 Tons 51 to 600 Tons

Millet supply price" 1,075 545
Retail price 1 ,890 1,025
Marketing cost (estimated) 100 100
Profit margin 715 380
Profit in % of retail price 38 % 37 %

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

“ includes transportation to traders' market outlet and handling to put in storage

The comparison of millet acquisition cost in Table 4-8 between large and
small scale millet traders demonstrates that three factors affect millet acquisition

prices significantly:

158

(1) the source of millet (larger traders are supplied by cheaper sources),
(2) the quantity discount (large traders get quantity discounts), and

(3) the bargaining power (larger traders have more bargaining power).

The weighted average millet acquisition price per ton was about N3 545
for larger millet traders and N$ 1075 for small traders. The comparison of millet
traders' retail prices indicates that the cost advantages of large-scale traders are
passed on to their customers. The average millet retail price of large-scale
traders in urban areas is N$lkg 1.03. This is only 54 percent of rural traders’
average retail price of N$lkg 1.89.

Table 4-8 also demonstrates that millet traders profit margins are high.
On the average, small-scale traders in rural areas have net returns of N$ 715 per
ton, while large-scale traders' net returns average N$ 380 per ton. However,
both trader categories reach net returns that are 37 to 38 percent of their retail
prices.

Although small-scale traders net returns are almost twice the returns of

larger traders, the following points have to be considered before judging them:

(1) Small millet traders have, by nature, lower trade volumes so that their total

profits are comparatively low.

(2) Millet marketing has a larger share of rural traders total business. Thus,
millet trade absorbs a high portion of rural traders' scarce funds. Due to
the larger share of millet trade of small traders' total business, their

financial risk from trading millet is higher. Market price variations caused

159

by unpredictable harvest results or grain shipments from South Africa or
Angola might annul small traders' profits expectation and leave them with

little cash to replenish their stocks with other tradable goods.

Although profit margins of larger millet traders are, on average,
significantly less than of rural traders, commercial exchange of millet can be very
profitable. This point is demonstrated in Table 4-9 , which presents two

examples of larger millet trade operation in 1993.

Table 4-9. Examples for the profitability of millet trade in 1993

 

 

 

Example 1: Example 2:

Urban Ovambo Urban Kavango

Trade volume 1993 (MT) 150 75
Millet supply price" (N$/ MT) 500 800
Marketing cost“ (NS/MT) 1 00 1 00
Retail Price (N$IMT) 1,000 1 .400
Net Profit (N$/MT) 400 500
Total Net Profit 1993 (N$) 60,000 37,500

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

" transportation cost to traders' market outlet included ** estimated

4.2.4. Prospects and limits of commercial millet trade
To analyze the prospects and limits of the commercial millet market, this
section focuses on four issues: (1) the aggregate size of the existing commercial

millet market, (2) millet trader's expectations for future market developments,

160
(3) the effects of grain imports and seasonal harvest results, and (4) grain

traders’ expectations from government.

4.2.4.1. Aggregate size of commercial millet market

It is estimated that the millet trader survey of 1993 covered about 70
percent of the study zones' commercial millet trade in urban areas, about
8 percent of commercial millet trade in rural Ovambo. and 21 percent in rural
Kavango. Based on these estimates, the total volume of commercially traded
millet amounted in 1993 to roughly 8,900 tons. Traders' statements about the
millet amounts that various producers supplied after the 1992/93 harvest permit
the extrapolation of percentage shares of the main millet sources of the
commercial millet market in general (see Table 4-10).

The most important finding from this extrapolation is that in 1993, about
half of the commercially traded millet in the study zones was produced in
Kavango although (a) its population is only about a quarter of that from Ovambo
and (b) Kavango's 1992/93 production was depressed due to late and poor
rainfall.

Another important finding is that about 93 percent of Kavango's
commercial millet trade takes part in rural areas, while in Ovambo, the respective
portion is only 58 percent; i.e., compared to Ovambo, rural grain markets in
Kavango seem to be less integrated with Kavango’s center market in Rundu.

This means the potential exists that in the future, better market integration in

161
Kavango could lead to an increased offer of millet in Rundu, which might well be

suited to being sold in Ovambo.

Table 4-10. Millet supply to commercial grain market in 1993, by zone
and market Iocation“ (tons)

 

 

 

 

 

 

Market Location: OVAMBO
Rural Urban Total Percent
Supplier Type:
Marginal farmer 1,660 50 1710 39%
Middle farmer 390 30 420 9%
Large farmer 470 220 690 16%
Angela 50 1,560 1,610 36%
Total 2,570 1,860 4,430 100%
Percent 58% 42% 100% -
Market Location: KAVANGO
Rural Urban Total Percent
Supplier Type:
Marginal farmer 840 0 840 19%
Middle farmer 670 40 710 16%
Large farmer 2,560 260 2,820 63%
Angola 130 0 130 3%
Total 4,200 300 4,500 100%
Percent 93% 7% 100% -

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

" extrapolated from trade data acquired during the millet trader survey of November 1993

162

For the 1992/93 production season, aggregate millet production of the
study zones is estimated on the basis of the production levels reached by the
320 sample farmers. According to these estimates, Ovambo farmers produced a
total of about 70,000 tons of millet, and Kavango farmers produced about 15,000
tons during the 1992/93 production year. The estimated 7,200 tons of
commercially marketed millet surplus from the study zones and the 1,700 tons of
imported millet from Angola make together 10 percent of Ovambo and
Kavango's total millet production during the 1992/93 production year.‘

Assuming that the poor geographical distribution of millet yields during the
1992/93 cropping season is not an exception, the question arises whether a 10
percent coverage of total millet production by commercial trade is enough to
adequately transfer millet between surplus and deficit grain production areas.
Because informal millet trade between neighbors is mainly bound within the
borders of rural communities, it does not contribute to the redistribution of millet
surpluses across long distances. Figure 4-1 in the first section of this chapter
demonstrated that during the year 1993, millet was only seldom available for
purchase in both informal and commercial grain markets. A part of Chapter 5,
about farm households' consumption pattern, will demonstrate that consumption,
at least during the year 1993, was constrained by limited access (physical) to
millet. In conclusion, commercial millet marketing might have to be expanded if

rural households' grain deficit is to be balanced with nationally produced millet.

 

‘ No comparable estimates of the commercial millet trade volume have
been made by other sources.

163

4.2.4.2. Trader views about the prospect of millet

The last paragraph questioned whether the trade volume of the
commercial millet in Ovambo and Kavango is sufficient to effectively supply
areas that are in grain deficit with millet surpluses produced in other areas. This
section presents commercial traders’ attitudes about the prospects and limits to
millet market expansion.

Of the millet traders interviewed in 1993, 93 percent claimed they would
like to expand their millet trading operation in the future. When asked about the
factors that make it difficult to achieve increases in millet trade, six major themes

evolved from their answers:

1. Transportation: Fifty percent of the commercial traders mentioned lack of
transportation as a limiting factor. Additionally, forty percent of the traders

complained about the poor conditions of feeder roads in rural areas.

2. Market transparency: Forty-three percent of the traders claimed that it is
difficult to locate those millet producers that have marketable surplus and

are willing to sell.

3. Operating funds: Forty-three percent of the traders claimed they do not
have enough operating funds to buy as much millet after harvest as they

would like and could sell in the hungry season.

4. Deficit production: Seven percent of the traders mentioned the general

production deficit of millet as limiting for their trade supply.

164

5. Price competition: Five percent of the traders claimed that maize meal is

supplied so cheaply that millet is not price competitive.

6. Processing: Five percent of the traders, mainly urban, indicated that
maize has the advantage of being already processed, while access to

adequate millet processing technology is still a problem.

For the future, most rural traders planned to increase their own millet
production to foster their millet trade operation. Field expansion and the use of
tractors for plowing and transportation were the methods of choice to achieve
such production increases. Investment in transportation vehicles was the
second most often mentioned goal for improving millet trade. Rural traders also
had the idea to ask the government for business loans to increase the amounts
of millet they could buy from farmers after harvest.

Among urban traders, none of whom was a farmer, improved advertising
and the establishment of contracts to supply government institutions with millet,
e.g., school feeding programs or hospitals, were most often mentioned.

Half of the interviewed Ovambo traders and 70 percent of the Kavango
traders believe that the millet market will grow faster in the future than the maize
meal market. The most often given explanation for this speculation was that
millet is the traditional staple food in northern Namibia, and that the majority of
people prefer it over maize. However, most of the traders who believed in a fast
growth of the commercial millet market mentioned at least one of the following

three conditions as prerequisites to make use of millet‘s market potential:

165

Supply: More millet has to be produced in Namibia and the access to it
must be improved.

Price: The supply price of Namibian millet has to become lower so that
millet's price competitiveness with maize meal is improved.

Processing: Better millet processing technology is necessary because

more and more consumers prefer the convenience of already-processed
maize.

Those traders who believed that the maize meal market will expand faster

stated the same issues used by the previous respondents as prerequisites for

the millet market development to confirm their opinion. They claim that:

1.

2.

Supply: Millet is not always available in the shops, but maize meal is.
Price: Maize meal is cheaper than millet.

Processing: Maize meal is already supplied in processed form and
people with higher living standards don't like to process grain manually.

Those traders who were indifferent about which of the two food staple

markets will develop faster pointed to two important factors that might eventually

determine the outcome:

1.

Rainfall: If better rainfall conditions allow increased millet production in
the future, millet supply for commercial trade will be guaranteed and
prices will decrease to competitive levels.

Preference: Customers preference will decide which grain will dominate
the food staple market in the future.

In summary, grain traders' opinions about the prospects and limits of the

commercial millet clarify that the future of the commercial millet sector lies, first,

in the capacity of Namibian farmers to produce more and cheaper millet, and

166
second, in millet's competitiveness against maize meal on the basis of price,

taste, and convenience of use.

4.2.4.3. Maize market in the study zones

This section describes some characteristics of the maize market of the
study zones and discusses millet traders' attitudes about the dynamics between
millet and maize trade, potential effects of import restrictions for Angolan millet,
and the effects of good millet harvest results on commercial grain trade in
general.

'Maize' or better 'maize meal' is by far the most sold food staple in
Ovambo and Kavango. The millet trader survey from 1993 revealed that
comparatively few food retailers in the study zones trade millet, but almost all
trade maize in the form of maize meal. Even those food retailers who traded
during 1993 both millet and maize sold between two and five times more maize
meal than they sold millet. The only exception to this pattern were rural traders
in Kavango. They sold, on average, seven times more millet than maize meal in
1993.

The largest food retailers from urban areas in Ovambo and Kavango
traded significant amounts of maize meal during the year 1993. The average
maize meal trading volume of six larger food retailers in Ovambo amounted to
320 tons in 1993’. During the same year, the three dominant food retailers from

Rundu traded, on average, 480 tons of maize meal. During 1993, rural

167
supermarkets and rural shops sold, on average, 26 tons of maize meal in
Ovambo and 5 tons in Kavango.‘

In Ovambo, urban maize traders received about a third of their maize
meal directly from Namibia's two large milling operations, 'Namib Mills' and 'Agrar
Mills', while in Rundu I Kavango, the main food retailers receive almost all their
maize meal from these mills. Most rural food retailers receive their maize meal
via wholesalers and not directly from the processing industry.

The three most important types of maize meal retailers of Ovambo and
Kavango are: (1) retail outlets of wholesalers in urban centers, (2) supermarkets
in urban and rural areas, and (3) small private shops in rural areas. Most of the
urban and rural food retailers have maize meal available throughout the year.
However, because consumer demand for maize meal tends to decline drastically
after millet harvest in June for a period of three to six months, wholesalers and
large supermarket chains reduce their maize meal stocks during this period to
about 30 percent of their normal level. Due to the reduced consumer demand
and the reduced supply from wholesalers, many of the rural food traders stock

only little or no maize meal after millet harvest.

 

‘ The large maize trade volume of individual traders in urban areas might
lead to the assumption that maize is mainly consumed in urban areas. However,
the number of rural shops and supermarkets selling maize meal to the rural
pepulation is also substantial. Therefore, from the location of the traders alone,
it can not be derived whether maize is more consumed in urban or rural areas.
As will be shown in Chapter 5 about households' grain consumption pattern, a
relatively high proportion of rural farm households purchases its maize meal
directly from urban retailers.

168

Millet traders were asked hypothetical questions about how their
customers would react to an increased maize offer (ceter paribus) in the market
and how they would react to a maize meal price decline (ceter paribus). Ninety-
five percent of the interviewed millet traders were convinced that an increased
maize supply without a decline in maize meal prices has no effect on their
current millet trade volume. Still, 87 percent of the traders claimed that even a
price decline in maize meal would not affect their millet trade. Most of the
traders argued that 'millet sells as long as it is available' and that 'people with
migration income will always buy millet.‘ The last argument especially indicates
that in northern Namibian, consumers perceive maize as an inferior good
compared to millet. Some traders considered a decline in millet sales but
claimed that 'a certain amount of millet will always be commercially traded.’ The
traders that assumed that millet price reductions would trigger a drastic reduction
in millet trade pointed to the fact that 'poorer customers have no other choice
than to buy the cheapest staple food available, no matter what taste preference
they have.‘

To the hypothetical question of how commercial millet trade would be
affected by an official ban of Angolan millet imports, most traders' answers were
ambivalent. In the first place, traders claimed a stop of millet imports would
aggravate the hunger situation in the study zones because less food would be
available. They also indicated that their own millet business could be hurt
because less millet would be available for trade. However, most traders also

mentioned that a decline of Angolan millet imports could also lead to increased

169
market prices for nationally produced millet. Most traders assumed that such
price incentives could prompt, in the long run, higher national millet production
and as such benefit the northern communal population if rainfall and technical
assistance would allow farmers to respond adequately.‘

When asked about the effect that a good millet harvest in the study zones
has on millet and maize trade, 72 percent of the interviewed traders responded
that commercial trade with millet and maize is generally depressed for several
months. Some of the interviewed traders said that millet sales to farmers usually
in grain deficit would decline. Others traders said that their commercial millet
trade would be unaffected because their urban clientele always needs millet.
The vast majority of the sample traders (93 percent) said that poor grain harvest
years like the drought year of the 1991/92 season effect their millet trade very

positively.

4.2.4.4. Grain traders' expectations from government
The interviewed grain traders made also statements about measures the
Namibian government could use to foster commercial millet trade in the future.

Traders' recommendations for improving commercial millet sector are as follows:

 

‘ Probably due to their own strong background in farming or relationships
to farmers, traders reacted from two standpoints. From the trader point of view,

they argued that government should not intervene in their trade business and
Angolan imports should continued to be allowed. But for the benefit of Namibian
farmers they were inclined to accept an import stop for Angolan millet to increase
producer prices.

170

First, to increase production, the government should: (a) improve its
plowing service program, (b) provide mechanical threshing equipment,

(c) render an effective agricultural extension service, and (d) provide a credit
scheme for farm investments.

Second, to improve the millet market government should: (a) provide
business loans to ease the acquisition of millet after harvest, (b) foster long
distance transport between areas with millet surplus production and millet deficit
areas, (c) encourage the organization of local millet markets that would reduce
the need of long-distance travel, and (d) help to identify suitable equipment for

mechanical processing of millet.

4.3. Farm level millet sales

This section covers three important aspects of millet marketing activities
on the household level. It first describes the frequency and extent of farmers'
millet sales. Second, it assesses those factors that farmers indicated as most
limiting to millet trade. And third, it presents the results from a simple probit
regression model that estimates the effects that various factors have on farm

households' millet selling decisions.

4.3.1. Frequency and extent of farm level millet sales
During the millet subsector household surveys, several attempts have

been made to quantify households millet sales between their harvests of

171
May/June 1992 and May/June 1993. The results to these inquiries were very
meager; i.e., only very few farmers stated that they sold millet during this period
and the quantities obtained from these farmers were very vague. The following
reasons might have caused such responses:

(1 ) After poor harvest results of the drought production year 1991/92, farmers
might have preferred to keep their millet to replenish their reserves;

(2) The concepts of recalling quantities of marketed millet over longer periods
of time might have been too difficult for the respondents so that they
stated only the most recent sales; and

(3) Because most farmers received at least some food aid during the
production year 1992/93, they might not have wanted to jeopardize further

handouts by revealing their own millet selling activities.

To acquire at least some indications about farm households' millet
marketing activities, survey farmers were asked more generally how often and
under which conditions they sell millet (Table 4-11). Responses to these
questions indicate that Kavango farmers sell millet significantly more often than
Ovambo farmers. Of the interviewed Kavango farmers, 18 percent claimed they
'sell millet every year.’ Less than one percent of the interviewed Ovambo
farmers made the same statement. Another 54 percent of the farmers in
Kavango 'sell millet in some years.’ This is still high in comparison to the
30 percent of the Ovambo farmers from the same category. Finally, the
comparison between different production zones indicates that relatively more

farmers whose production areas are closer to the fast growing population centers

172
(central Ovambo and central Kavango) sell millet than those located in peripheral

production zones.

Table 4-11. Millet selling frequencies of rural households, by zone and
subzone (percent)

 

 

 

Zone Ovambo Kavango '
(n=200) (n=120)

Subzone West Cen- East Mean West Cen- East Mean
tral tral

 

Selling Frequency

 

Never 76 55 80 70 34 18 32 28
Some years 24 45 20 30 42 80 40 54
Every year 0 0 0 0 24 2 28 1 8
Total 100 100 100 100 100 100 100 100

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

The households that sell millet every year or at least in some years were
asked to point out the conditions under which they usually sell millet. The rank
order of important selling conditions is the same for both study zones. Most of
the millet selling farmers state an 'urgent need for money or goods' as the
primary motivation for their millet sales.

However, especially farmers from Kavango often mention 'enough millet
reserves' as a precondition for grain sales. Of the millet-selling Kavango
farmers, 40 percent claim that they sell millet only if they have millet reserves
that last at least for two years. An additional 30 percent of Kavango farmers

require reserves that last at least one year. (Table 4-12)

173

Table 4-12. Farm households' pro-conditions for selling millet

 

 

 

(percent')
OVAMBO KAVANGO
(n=200) (n=120)
% of Households Selling Millet 30 72
Urgent need of money or goods 23 54
If reserves are enough“ 11 54
After a good millet harves *** 10 44
Being approached to sell food 6 28
After every harvest 0 18

 

Data Source: Namibian Millet Subsector Research Project, 1992-93

" Because farmers could identify several of their selling conditions at the same time
the
presented percentages do not sum up to 100.

“ Most farmers have a millet reserve/selling threshhold that makes them accumulate
millet reserves until they feel secure against drought. Only if this threshhold is
reached they consider selling some of their grain. (this excludes forced sales of

grain)

*“ This category could but does not have to be congruent with the previous category.
It is possible that a farmer had a good millet harvest but decide not to sell millet
because his/her reserve/sellng threshhold is not yet reached.

174

Again, compared to Ovambo, a significantly higher proportion of millet-
selling farmers in Kavango state they sell millet 'only after a good grain harvest,’
i.e., when the above-mentioned thresholds of grain-reserves are achieved.
Eventually, not all millet sales at the farm level depend solely on households'
grain supply. Half of the millet selling Kavango farmers and 25 percent of the
respective farmers from Ovambo are selling millet if they are 'approached by
other households who are in need of food.’

The quantities of millet sold during individual selling transaction are
especially small in Ovambo. In this zone, 85 percent of millet-selling households
sell not more than about 16.5 kg at a time, the amount traditionally used
measures can hold. Eleven percent sell millet mostly at quantities of a 50 kg
bag. In Kavango, the situation is reversed. Only 25 percent of millet selling
households sell millet to individual buyers at quantities below or equal to 16.5 kg,
while 61 percent sell millet mostly at the quantity of a 50 kg bag.

Of the millet selling households, only the minority plans to sell millet at the
time of planting (Ovambo: 9 percent, Kavango: 38 percent). An active search for
millet buyers is undertaken by only a small portion of millet selling producers
(Ovambo: 19 percent, Kavango: 40 percent). In both zones, active millet sellers
tell their neighbors and community members about their willingness to sell and/or
send messages to other communities. A few cases even broadcasted over
public radio their willingness to sell millet.

In order to search for potential millet buyers and to transport millet to the

homestead, most sellers walk (Ovambo: 58 percent, Kavango: 38 percent).

175
Those who don't walk use a variety of other means to reach buyers: in Ovambo,
bicycles (19 percent), private motor vehicles (6 percent), and donkey carts
(8 percent) are used. In Kavango, the oxen sledge is the predominant means of
transportation (40 percent), while bush taxis are used by about 29 percent of the
sellers, and private motor vehicles by 9 percent.

As stated before, the millet subsector household surveys were not able to
collect adequate information about how much farm households sell millet over a
specific period of time and in what proportions to different buyers. However,
farmers were willing to state to whom or at what location they tend to sell their
millet. When asked to whom they sell their millet, neighbors are named by the
vast majority millet producers (Ovambo: 100 percent, Kavango: 89 percent).
Traveling traders were the second most often mentioned category of millet
buyers (Ovambo: 22 percent, Kavango: 42). Selling millet at local markets
seems to be only important for Ovambo farmers. In Kavango, millet is also sold
to local shops, urban markets or retailers, and to the two competing
organizations, the self-help marketing cooperative 'Katemo' located in Rundu
and the Namibian Development Corporation, which owns several semi-

commercial farms in Kavango. (Table 4-13)

176

Table 4-13. People or trading locations farmers sell millet, by zone
(percent of millet selling households‘”)

 

Millet Selling Households
Ovambo (n=53) Kavango (n=85)

 

Neighbors 100 89
Traveling traders 22 42
Local markets 17 1 1
Local shops 4 36
Urban markets/retailers 6 24
Katemo Cooperative/Rundu 0 25
NDC farms in Kavango 0 12
Traders from Wrndhoek 0 5

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

"Because farmers could indicate several of their millet buyers at the same time, the
presented percentages do not add up to 100.

Among neighbors, millet is often bartered for meat. This and similar types
of barter occur usually briefly after the grain harvest and involve quantities that
range between 1 and 16 kg. Millet is sold to traveling traders mainly in exchange
for money or consumer goods. Again, these transactions take place mainly after
harvest. However, the quantities sold range from 16 kg up to several 50 kg bags

of millet.

4.3.2. Limitations of millet marketing

As mentioned in Chapter 1, the proposal of the Namibian Agronomic
Board to implement a millet marketing scheme in northern Namibia was based
on the assumption that informal and commercial millet prices are not high

enough; i.e., the proposal argued that govemment-stipulated producer prices are

177
necessary to encourage more millet production. The following section
demonstrates that millet prices are only one of several factors affecting farmers'
millet-selling behavior. This section first identifies various limiting factors of millet
marketing on the farm level and then addresses the more important of these

factors individually.

4.3.2.1. Marketing versus production constraints

The farmers included in the millet subsector household surveys were
asked to identify factors that make millet marketing difficult. The responses of
the interviewed farmers were analyzed separately for each study zone and for
(1) those farmers that ‘never sell millet’, (2) those who 'sell millet in some years’,
and (3) those who ‘sell millet every year’.

For farmers that never sell millet, ’lack of marketable millet surplus' and
'fear of drought' are the leading constraints on millet marketing.

With increasing millet selling frequencies, farmers mention also more true
marketing constraints. In both study zones, over half of the farmers that sell
millet in some years mention 'expensive transportation' as an obstacle to millet
marketing. ln Kavango, over 90 percent of these farmers complain additionally

about 'low market prices'. (Table 4-14)

178

Table 4-14. Millet marketing constraints of rural household, by farmers'
selling frequency_and zone (percent)

 

 

Ovambo (n 8 200) Kavango (n = 120)
never some every never some every
Households Selling years year years year
Frequency in Years 70 30 0 28 54 18
Constraints
not enough millet 83 68 n.a. 94 98 84
fear of drought 79 86 n.a. 94 97 88
expensive transport 22 56 n.a. 20 67 77
no own transport 6 29 n.a. 10 33 53
long distances 7 33 n.a. 0 34 65
no local market 18 11 n.a. 3 23 53
tradition not to sell 15 16 n.a. 32 8 33
search for buyers 12 9 n.a. 0 26 25
low millet prices 7 11 n.a. 0 94 29

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

Eventually, over 50 percent of the farmers that sell millet every year state
'lack of their own transportation', 'long trading distances', and 'lack of local

markets’ as selling constraints.

4.3.2.2. Grain reserves against drought
To analyze the effect that fear of drought has on households' millet-
marketing decisions, farmers were asked to state the amount of grain reserves

they intend to have in storage before they consider selling larger amounts of

179
millet. The responses regarding farmers' grain storage/selling-thresholds have
been converted into (a) amounts stored per household member and (b) the

period of time such reserves would last per household (see Table 4-15).

Table 4-15. Millet reserve targets of rural households, by zone and
farmers' millet selling frequency (kglcapita, yearslcapita)

 

 

 

Ovambo (n = 200) Kavango (n = 120)
Selling never some every never some every
Frequency years year years year
Reserve targets
- kg I capita 540 870 n.a. 260 220 460
- years I capita 4 6 ‘/z n.a 2 2 3 ‘/2

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

These conversions reveal that Ovambo farmers' fear of drought translates
into very high, almost unrealistic millet reserve targets. On average, Ovambo
households stated storage targets for millet that would last between 4 and 6 ‘/2
years. These high thresholds imply the following:

(1) The occurrence of drought is imminent and farmers take this into
consideration.

(2) Most Ovambo households are not used to selling millet in larger quantities
and therefore have the tendency to overestimate their reserve needs.

(3) Because of fear of drought, household reserves have first priority so that
even after a good production season, commercial millet sales are not as
large as they could be otherwise.

1 80

The storage/selling-thresholds of Kavango households seem to be more
in line with actual reserve needs. Kavango households that never or only rarely
sell millet prefer to have, on average, millet reserves that last 2 years. Kavango
households that sell millet every year perceive a need for higher storage/selling-
thresholds. Their target grain reserves last, on average, 3‘/2 years. It is
assumed that these higher storagelselling-thresholds are related to the social
obligations richer farmers have with respect to their extended family and their
poor neighbors.

Once the knowledge that those households selling larger amounts of
millet also tend to accumulate considerable amounts of millet reserves is pieced
together with the information about farmers' storage practices, interesting
implications evolve with regard to the prospects of further millet market
developments. As already presented in Chapter 3, farmers traditionally store
freshly harvested millet as drought reserve and use older grain reserves for
consumption. The same is true for marketable surplus, older grain surpluses are
offered to the market only when the new harvest arrives and storage containers
become scarce.

While the described behavior was justified in the past by household food
security considerations, it is not always reasonable under today‘s circumstances.
Contrary to the past, maize meal is today readily available in the food market,
even when millet can not be purchased. This means that larger and wealthy
farmers might be better off selling their surplus grain soon after harvest and

using at least part of it to make investments in non agricultural activities like

181
education or business, instead of letting their millet sit and rot in their granaries.
The commercial millet market currently suffers because even after good harvest
years, mostly older and thus lower-quality millet is offered. This prevents
consumers that have the cash income to purchase fresh millet from acquiring it.
In the middle and long run, such financially-strong customers are lost because
they switch to maize meal that is always available, convenient to use, and of
consistent quality.

In summary, the above analysis demonstrates that after good production
years, farmers ' priority is first to replenish their millet reserves; i.e., at least part
of the marketable millet surplus from one production season is not immediately
available to consumers in demand and/or areas with production shortfalls. While
this behavior is justifiable for farmers who have little access to cash income
during periods of drought, the same behavior by wealthier surplus farmers leads
to a reduced offer of marketable millet surplus in the commercial food market,

and thus to reduced development prospects for the commercial millet market.

4.3.2.3. Transportation limitations

The section about the general limitations to millet marketing at the
household level demonstrated that long distances between producers and
consumers, as well as lack of transportation and/or high transportation cost, are
important constraints for those farmers who sell millet in some years or even
every year. To analyze the effect that transportation constraints have on

households’ millet marketing decisions, sample farmers were asked to indicate

1 82
(a) the locations where they would buy and/or sell a 50 kg bag of millet,
(b) whether they needed transportation to reach these locations, and
(c) how much it would cost them to travel to these locations with and without a

bag. (see Table 4-16.)

Table 4-16. Marketing locations for larger amounts of millet,
by zone and type of transaction (percent)*

 

 

 

 

Ovambo Households Kavango Households
(n=200) (n=120)

Location Buying Selling Buying Selling
neighbors 35 35 32 32
local market 38 30 1 1
distant market 24 23 5 27
local shop 0 8 27 8
distant wholesaler 2 2 8
Katemo cooperative 0 0 17 5
FNDC 0 0 6 17
other 1 2 4 4
total 100 100 100 100

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

" Farmers were asked to state the locations where they buy and sell at least 50 kg of millet In
case the respondent claimed heIshe never buys or sells millet, the question was asked
hypothetically.

note: Traveling traders were not mentioned by respondents because the survey question asked
specifically for locations, while traveling traders from northern Namibia tend to be unpredictable
in the time and the location of their appearance.

183

In Ovambo, farmers' millet buying and selling locations do not differ from
each other. Neighbors and local markets rank first, with each used by more than
30 percent of the survey farmers for millet buying and selling. Distant markets
rank third. They are used by roughly a quarter of farmers for marketing and
purchasing millet in a 50 kg bag or more. Similar to Ovambo, about a third of all
Kavango farmers sell and buy 50 kg bags from their neighbors. Other places
differ in their importance for selling and buying. Beside from neighbors 27
percent of Kavango farmers buy millet from local shops and 17 percent buy from
the 'Katemo’ cooperative in Rundu. Besides to neighbors, Kavango farmers sell
millet to the open market in Rundu and to various semi-commercial NDC farms
scattered along the Kavango river.

Table 4-17 presents the percentage of households that (a) do not need
transportation to trade a bag of millet (buying and selling), (b) use their own
transportation, or (c) have to hire transportation service. Because of the low
population density in eastern Ovambo and western Kavango, millet buying and
selling points are rare. Accordingly, in these two zones, between 80 and 100
percent of all households need transportation for millet trade. In the other
subzones, between 30 and 60 percent of households need transportation to buy
or sell millet. On average, between 40 and 50 percent of the households from
both study zones hire transportation to reach a location where they can sell or

buy a larger amount of millet.

184

Table 4-17. Transportation needs for millet marketing stated by
farmers, by subzone, and transaction type (percent)

 

Zones Ovambo Households Kavango Households

Subzones West Centra East West Central East
I

(n=80) (n=40) (n=80) (n=40) (n=40) (n=40)

Transportation need for millet purchases

 

 

 

need no

transportation 51 41 1 23 59 63
use own

transportation 19 3 50 23 3 3
hire

transportation 30 56 49 54 38 34 ,
total 100 100 100 100 100 100

Transportation need for millet marketing

need no

transportation 51 54 0 8 50 71
use own

transportation 17 3 6 12 0 4
hire

transportation 32 43 36 80 50 25
total 100 100 100 100 100 100

 

Data Source: Namibian Millet Subsector Project Surveys 1992/93

Hiring transportation for grain marketing is expensive. (Table 4-18) At the
end of the 1992/93 production year, the price for transporting a 50 kg bag of
millet tolfrom the trading place ranged in Ovambo, between N$ 1.20 and
N$ 4.00, and in Kavango, between N$ 2.10 and N$ 8.10. The transportation

costs for the person that buys or sells a bag of millet are even higher. The

185
average travel costs for trading individuals ranged, in Ovambo, from N$ 5.50 to
N$ 20.00, and in Kavango, from N$ 5.70 to N$ 21.20. Based on an average
price of N$ 60 per 50 kg of millet during the 1992/93 production year, total
transportation costs ranged, in Ovambo, from 11 to 40 percent of the traded

millet value, and in Kavango, from 13 to 59 percent respectively.

Table 4-18. Cost of hiring transportation to sell/purchase millet,

 

by subzone (N3 per 50 kg)
Zones Ovambo Kavango
West Central East West Central East
Subzones
(n=80) (n=40) (n=80) (n=40) (n=40) (n=40)

 

Transportation costs to purchase millet
per 50 kg bag" 1.20 4.20 4.00 4.60 8.10 2.00

 

 

 

'°' "“39 5.50 9.20 5.50 12.20 21.20 6.40
person

‘0‘“ “"3”" 6.70 13.40 9.50 16.80 29.30 8.40
cost

Transportation costs to market millet

per 50 kg bag“ 2.30 2.60 3.10 4.50 5.00 2.10
'°' ““139 20.00 6.10 9.10 11.00 10.00 5.70
person

at: ““3”" 22.30 8.70 12.20 15.50 15.00 7.80

 

Data Source: Namibian Millet Subsector Project Surveys 1992/93

" cost for one way “ cost for return trip

1 86
The above analysis indicates that many rural Ovambo and Kavango
households have to hire transportation for millet trade and that their
transportation costs are, in some cases, very high. It can be assumed that such
high transportation costs reduce the profit expectations of potential millet sellers
and ,therefore, lead to reduced millet marketing activities beyond rural farmers'

immediate neighborhood.

4.3.2.4. Lack of local markets

Lack of local grain markets is often closely related to the need of traveling
far distances to sell or buy millet. In Ovambo and Kavango, between 40 and 60
percent of the surveyed households claimed that there is no local market in their

vicinity. (Table 19)

Table 4-19. Percent of farmers lacking access to local grain markets
and their attitudes regarding the need of such markets

 

 

 

(percent)
Zones Ovambo Households Kavango Households
Subzones West Centra East West Central East

(n=80) (n=40) (n=80) (n=40) (n=40) (n=39)

No Market Access: 49 50 59 63 50 51
Expressed Attitudes:
Yes, need market 26 13 53 63 50 49
Market will not work 13 37 6 O 0 2

 

Data Source: Namibian Millet Subsector Project Surveys 1992/93

1 87

Most of the interviewed farmers who lack access to local grain markets
expressed an urgent need for such places. These farmers expect from local
markets the following: (a) shorter travel distances (50 percent of all answers);
(b) easier trading of grain (30 percent of all answers); and (c) a general
improvement of living and income conditions (10 percent of all answers).

A few farmers said that a local grain market would not function (mainly in
western and central Ovambo). The two most often mentioned reasons of these
farmers are related to the small trade volume in their neighborhood. In western
Ovambo, farmers claimed more often that, due to the low purchasing power of
their neighbors, potential millet suppliers could not sell much millet and,
therefore, could not earn enough profits. In central Ovambo, opponents of local
markets claimed that new grain markets in their vicinity would fail because most
of their neighbors are not producing enough marketable millet.

The concerns about an insufficient trade volume of local markets bear
some validity. Local market places are not necessary in every millet production
area, nor is the organization of certain market days promising . However, lack of
local grain markets in areas with high demand for millet, or large millet surpluses
(or both), increases transaction cost and makes it impossible for individuals to
make their needs known to long-distance traders so that they can react. Lack of
local grain markets leads surplus producers either to store their millet in their

granaries where it is more difficult to access by deficit farmers‘, or to sell their

 

‘ As demonstrated in Table 4-12, a significant portion of farmers selling
millet (Ovambo: 6 percent, Kavango: 28 percent) claim they only sell if they are

1 88
millet at distant marketplaces, which, again, can hardly be reached by poorer
deficit farmers in their neighborhood. Consequently, grain deficit farmers, short
of local grain markets, have no other alternative than to buy imported maize that

is channeled via the urban-based retail system to rural shops.

4.3.3. Regression analysis of farmers' marketing decision

As explained above, data about actual amounts of millet marketed by
survey farmers in the past could not be acquired through the conducted millet
subsector household surveys. As a result, only farmers' categoric statements
about their general millet selling frequencies are available as a basis for the
quantitative analysis of household level millet marketing behavior. Of 320
households, 56 percent claimed they 'never sell millet,’ while 44 percent stated
they sell millet in 'some years' or 'every year'. The following presents, first, an
overview of the cross-section probit model that uses the categoric millet selling
frequencies as independent variables; second, describe the independent
variables employed in this model, and third, discuss the regression results.

When just one observation on each decision maker is available a probit
regression model can be used in combination with maximum likelihood methods

to predict whether a farm household is selling or not selling millet, given a set of

 

approached by potential buyers. This means farmers who are interested in
purchasing millet have first to visit potential millet sellers to investigate whether
they are willing to sell millet at all, and if so, how much and at what price. In the
widely-dispersed communities of northern Namibia this could mean that potential
buyers might have to invest significant efforts (search costs) until they identify
someone willing to sell his/her millet.

1 89
values for the explanatory variables. The measure used to evaluate the
estimated coefficients is the percent of successful predictions within the given
sample. Due to the nature of the probit model the estimated coefficients of the
probit model do not indicate the increase in the probability of farmers millet
selling, given a one-unit increase in the corresponding independent variable.
Rather, the estimated coefficients reflects the effect of a change in an
independent variable from an initial value on the probability of millet selling at this
initial value. For this study the initial values of the independent variables are set
according to a base (control) case that represents an average farmer in Ovambo
and Kavango.

The cross-section probit model encompasses explanatory variables that
identify (a) farmers' grain production performance in the past, (b) the use of
inputs as indicators that farmers try to increase their millet production to be able
to sell at least part of it, (c) farmers' access to grain markets, and (d) household
characteristics. Because millet producer prices from the past are neither
available on the household level nor on any higher aggregate level, economic
price incentives for millet sellers had to be ignored. The variables are all tested

for correlation and had met the test.

190

SELLMIL = f (092, SPYRS, ZONE, CEMKT, LOMKT, MANR, OKAS, CASH)

where:

SELLMIL
092
SPYRS
ZONE
CEMKT
LOMKT
MANR
OKAS
CASH

= farm household tends generally to sell millet (Dummy, 1 = yes)
= millet production during the 1991/92 drought year (mt)

= millet surplus production years between 1988192 (No. 1 - 5)

= located in Ovambo or Kavango (DUMMY, 1 = Ovambo)

= located close to large center market of zone (DUMMY, 1 = yes)
= have access to local market (DUMMY, 1 = yes)

= use of cattle manure on all fields (DUMMY, 1 = yes)

= experience with Okashana 1 before 1992/93 (DUMMY, 1 = yes)
= annual cash income from non-cropping activities (NS '000)

Table 4-20. Characteristics of the variables in the millet

sellingequation

 

Ovambo and Kavango (n I 268)
VARIABLE MEAN S.E.
SELLMIL (1 = yes) 44 percent = 1
092 (MT) 0.57 2.12
SPYRS (No. 0-5) =35%, 1=37 %, 2=16, 3=10, 4=5%,5=0.3%
ZONE (1 = Ovambo) 63 percent = 1
CEMKT (1 = yes) 25 percent =1
LOMKT (1 = yes) 47 percent = 1
MANR (1 = yes) 18 percent = 1
OKAS (1 = yes) 43 percent =1
CASH (NS '000) 7.65 12.10

 

Data Source: Namibian Millet Subsector Research Project, 1992/93

1 91

The independent variables used, and the reasons for their inclusion in the
estimated equation, are as follows:

(a) the amount of millet harvested in the 1991/92 production year (092).
When asked about the limiting factors of millet marketing, the majority of the
interviewed sample households responded that low millet yields prevent them
from selling millet. From these statements it is concluded that households'
overall millet production capacity affects their millet marketing decisions. To
account for this effect, households' millet production during the severe drought
year of 1991/92 (092) has been included in the probit model as a proxy for
households' overall millet production capacity. It is hypothesized that the
likelihood that a farm household is generally selling millet is positively related to
the variable 092.

(b) the number of years the household was self-sufficient in millet
production from 1988 through 1992 (SPYRS). Not only farm households'
general millet production capacity affects their millet marketing behavior, but also
the fact whether the household is actually capable of producing enough millet
over time to cover its grain consumption needs. It is assumed that if a household
is not able to produce marketable millet surpluses over a longer time period, the
likelihood that this household sells larger amounts of millet is reduced. To
account for the effect of grain surplus production on farmers' millet marketing
decisions, the variable SPYRS is included in the equation. The SPYRS variable
can have six different values, from 0 to 5, reflecting the number of years survey

respondents stated they experienced millet self-sufficiency during the period

1 92
from 1988 to 1992. Based on the assumption stated above that with an
increased number of surplus years farmers are more likely to sell millet, it is
hypothesized that the number of grain surplus years experienced by rural
households is positively related to the dependent variable SELLMIL.

(c) the production zone in which farm households produce (ZONE). As
demonstrated in Chapter 3, the overall millet production potential of Ovambo is
lower than that of Kavango. This is mainly caused by lower average rainfall in
Ovambo and its significantly higher population density, which leads to the use of
more marginal soils for crop production. It is assumed that lower millet yield
levels in Ovambo hinder farmers from millet selling, not only through the physical
lack of surplus production, but also by creating an attitudinal environment in
which millet sales have a much lower priority than household's own millet
consumption and its millet reserves against drought-related crop failures. To
control for the effect that the lower millet production potential of Ovambo has on
farmers' millet marketing behavior, the dummy variable ZONE is included in the
estimation. The variable takes the value one if the sample household is located
in Ovambo. It takes the value 0 if the sample household is from Kavango. The
coefficient of the variable ZONE is expected to be negative as a result of the
negative effect that the lower millet production potential of Ovambo has on
farmers' millet selling behavior.

(d) the farm location with respect to the central market (CEMKT). Farmers
often mention long distances and high transportation costs as constraints to

millet marketing. As discussed in Part 2 of this chapter, high transportation costs

- ssssss

1 93
reduce profit expectations of potential millet surplus producers, and, therefore,
might discourage farmers from commercial millet production. To account for the
effect that long distances and high transportation costs have on farmers' decision
to produce millet for sale, the dummy variable CEMKT has been included in the
regression. The variable has the value one if farmers live in the center region of
their zone and are therefore closer to urban grain markets where they can sell
larger amounts of grain for cash. Additionally, it is assumed that a closer
distance to urban centers increases households' exposure to modern consumer
goods which can only be acquired through monetary payments. This again
encourages the marketing of grain surpluses. It is therefore hypothesized that
households' millet selling is positively related to the variable CEMKT.‘

(e) farmers' access to a local grain market (LOMKT). Only 44 percent of
the interviewed sample farmers have access to local grain markets. The majority
of the farmers that have no local market in their vicinity state that they need such
a place to reduce their trading distances and to generally ease the exchange of
millet and of other goods. It is hypothesized that those farmers who have a local
marketplace close to their homestead are more likely to be engaged in millet
trade than those without. To test this hypothesis, the dummy variable LOMKT is

included in the regression model. The variable takes the value one if farm

 

‘ The question whether those millet selling farmers who are located further
away from center markets might compensate this fact by selling more millet to
their neighbors bears little validity. This is because farmers located closer to
center towns have also neighbors to sell to. The center of the above made
argument is the agdjtignal incentive the close center markets present for millet
selling.

1 94
households have access to a local grain market; otherwise, the variable is zero.
The coefficient of the LOMKT variable is expected to be positive as a result of
the positive effect that close by marketing opportunities have on households'
millet selling decision.

(f) the use of yield-enhancing production inputs to produce marketable
surplus (MANR and OKAS). The regression analyses of millet area and millet
hectare yields per household in Chapter 3 demonstrate that (a) the fact that
farmers used Okashana 1 before it was distributed for the 1992/93 production
season to relieve potential seed shortages, and (b) the use of cattle manure or
chemical fertilizer are positively correlated with farm household millet yields. It is
hypothesized that those farmers who invest (either financially or through more
labor) into productivity enhancing inputs like seed and fertilizer try to recapture at
least part of their production costs by marketing millet. To control for the effect
that the use of potential yield enhancing inputs has on farmers' millet selling
behavior, the two dummy variables MANR and OKAS are included in the model.
The variable MANR takes the value one if farmers claimed they fertilized their
field either with manure and/or chemical fertilizer. The OKAS variable indicates
whether farm households have grown Okashana 1 at least once before the
1992/93 season (one) or not (zero). The coefficients of both variables are
expected to be positive due to the positive relation between input use and
marketing intention.

(h) the amount of farm households' annual cash income (CASH). Farm

households’ annual cash income has shown only small positive effects on

195
farmers' cultivated millet area and on their millet hectare yields. The explanation
for this weak relationship between farm household production intensity and
annual cash income is based on the argument that households that earn larger
amounts of cash income from non-crop production related activities (further
called cash income) don't have to balance their consumption needs with grain
production. A similar argument is used to hypothesize about the relation
between farm households' millet selling behavior and their annual cash income.
It is assumed that households that earn cash income are less likely to produce
millet for commercial purpose but rather to make use of their labor force bound
to the rural homestead. To account for the effect that cash income has on
farmers' millet marketing behavior, the amount of cash income earned by sample
farmers during the 1992/93 production year (CASH) is included in the equation.
Based on the explanations given above, it is hypothesized that household level

millet sales are negatively related to the CASH variable.

Table-21 presents the main statistics of the estimated equation. The
model's prediction levels indicate that 75 percent of the 268 cases regressed
were correctly predicted to be either millet sellers (69 percent) or not (80
percent). All signs of the estimated coefficients are according to expectation and
the lowest significance level of the coefficients is 0.07.

For the following interpretation of the probit regression results, the base
(control) case represents a farm household during the 1992/93 production year

that did not produce millet surplus during any of the last five grain production

196
periods (SPYRS = 0), lives in Kavango (ZONE = 0), lives in production areas
that are far away from commercial centers (CEMKT = 0), has no access to local
markets (LOMKT = 0), does not fertilize its fields (MANR = 0), and has not used
Okashana 1 before the 1992/93 production season (OKAS = 0).

Holding all other factors constant, the zone in which the sample farmers
live (ZONE) affects most the probability that a farm household sells millet. If a
farmer is located in Ovambo, the likelihood that he markets millet is reduced by
58.5 percent compared to a Kavango farmer. This result is not surprising, taking
Ovambo's marginal grain production potential into account.

Also, under ceteris paribus conditions, the number of grain surplus years
(SPYRS) a household experienced between 1988 and 1992 has a significant
effect on the probability of millet selling. With each additional year of grain
surplus production during the five year period, the likelihood of millet selling
increases by 13.8 percent. This conforms with the hypothesis that farmers
prefer to first accumulate a certain amount of millet reserves over the course of

several production seasons before considering selling larger amounts of it.

197

 

 

Table 4-21. Estimates of the millet selling equation

Ovambo and Kavango
VARIABLES COEF. S.E. SIGNIF. PROB.‘
CONSTANT - 0.88 0.39 0.03 .-
092 0.20 0.06 0.00 4.6”
SPYRS 0.65 0.17 0.00 13.8
ZONE - 3.91 0.61 0.00 58.5
CEMKT 1.34 0.35 0.00 24.3
LOMKT 0.58 0.32 0.07 12.5
MANR 2.28 0.59 0.00 _ 32.4
OKAS 0.84 0.36 0.02 19.7
CASH - 0.05 0.02 0.02 1.2m

 

Prediction Levels: Non-Sellers = 80 %, Sellers 8 69 %, Overall = 75 %
Case Number = 268, Chi - Square = 106, DF. = 8, SIG. = 0.0000

 

Data Source: Namibian Millet Subsector Project , 1992/93

' lndicateshowthelncreaseofan Independent variablebyoneunit altersthe
probebllltythatthebaseceseisamllletseller. (inpercent)

" Millet production Ofthe1991/92 droughtyeerls increased byoneton.

“' AnnmlcashlncomefromnoncroppingactivitiesisincreasedbyNS1000.

1 98

Compared to the described relationship between number of surplus years
in the past and millet selling, the amount of millet harvested during the severe
drought season (092) shows little effect on farmers' millet marketing behavior. A
large increase of millet production by one ton during the 19991I92 production
season increases the likelihood that a farm household sells millet only by 4.6
percent. The high significance level of 0.00 indicates that households' capacity
to produce some grain under severe drought conditions has a positive effect on
their general millet selling behavior. However, as argued before, other factors
that reflect farmers' production capacity over a period longer than one year
demonstrate a stronger relation to farmers' millet selling behavior.

As hypothesized, the use of yield enhancing inputs is positively correlated
with farmers' millet selling behavior. If farmers fertilize their grain fields with
manure (MANR), the probability that they sell millet increases by 32.4 percent.
Similarly, farmers who have gained experience with the newly-introduced millet
variety Okashana 1 (OKAS) have an increased likelihood to be sellers by
19.7 percent. Having controlled for households' overall millet production
capacity (092) and the number years farmers were millet self-sufficient since the
year 1988 (SPYRS), it can be assumed that those farmers who make
investments in production inputs expect and try to recapture their investment
costs or efforts by marketing at least part of their millet production.

Both coefficients estimated from variables that represent farmers'
closeness to commercial grain markets (CEMKT) and informal millet markets

(LOMKT) indicate a positive relation to farmers' millet selling behavior. If millet

199
producers are living closer to the population centers of Ovambo or Kavango, the
likelihood that they are millet sellers increases by 24.3 percent. This can be
explained by the fact that farmers who are closer to urban markets are (a) more
exposed to commercial ideas, and (b) receive a higher stimulus to buy consumer
goods for which they have to earn money through millet sales. If millet
producers are located close to a local grain market, the likelihood that they are
selling millet is increased by 12.5 percent. The explanation for this is lower
transaction cost for millet marketing as well as exposure to the idea to
commercially produce millet. However, the closeness of a local grain market
might just indicate that the farmers in consideration live in an area with increased
millet production, so that the likelihood of having marketable surplus is generally
increased.

The coefficient Of the CASH variable reflecting households’ annual cash
income is relatively small and negative, but significant at the 2 percent level.
This result indicates that those farm households that earn larger amounts of
money from alternative income sources are less likely to sell millet compared to
those who have no other employment than to produce millet. An increase of
households' annual cash income by NS 1000 reduces the probability of millet
selling by only 1.2 percent. The result confirms the assumption that households
that earn cash income are less likely to produce millet for commercial purposes,

but rather to make use of their labor force staying at home.

200
4.4. Chapter summary
The following section summarizes the findings about farm level millet
marketing, commercial millet trade, and the price position of millet in the study

zones.

4.4.1. Farm level millet marketing

The most important finding about millet marketing on the household level
is that about 70 percent of the farmers in Ovambo never sell millet, while the
remaining sell millet only in small quantities every other year. In Kavango, millet
selling is much more common. About half of Kavango's farmers sell millet every
other year, and almost a fifth of them sell millet every year. Also the amounts of
millet that Kavango farmers market are larger than those from Ovambo farmers.
The differences in household level sales of millet lead to the fact that although
Kavango has only a forth of Ovambo's population the amount of its marketable
millet surplus from the 1992/93 production was the same as in Ovambo. This is
a clear indication that Kavango has a higher potential for commercial millet
production in the future than Ovambo.

Most of the farmers that market millet sell it informally to their neighbors.
For the farmers that sell millet to commercial intermediaries marketing
constraints are a problem. The most important marketing constraints of these
farmers are high transportation cost. Farmers from Kavango with marketable
surplus indicate that the lack of local markets leads to far distances to the central

market in Rundu. The results from the cross-section probit model also confirmed

201
that the proximity to urban and local grain markets increases the likelihood that
farmers engage in the production of millet for the market.

TO cope with the large annual yield variations and to reduce the threat to
household food security, most farmers try to keep millet reserves that last
between one and three years. This means that after good harvest years, most
farmers replenish their millet reserves first, and that most of the marketable
surplus from such seasons is not immediately available to farmers with grain
production shortfalls. Farmers' practice of storing freshly harvested millet as
reserves and consuming or Offering older millet stocks for sale leads often to

older, and thus lower-quality, supplies for the commercial millet market.

4.4.2. Commercial millet trade

In rural areas of Kavango millet dominates the commercial trade of grain.
Large farmers that use traditional, low input production methods supply most of
the marketable millet surplus in Kavango. In Ovambo, most commercially traded
millet is supplied by poor farmers who are in urgent need of cash. In Ovambo
commercial millet trade gained importance only since the severe drought in the
year 1991/92. During this year the demand for millet was so high that millet
imports from Angola were encouraged. During the production year of 1992/93
urban Ovambo traders received up to 80 percent of their millet supply from
Angola.

Although it is unclear whether millet imports from Angola will persist in the

long run, the fact that millet was imported from neighboring countries changes

 

202
the prospects of the Namibian millet sector significantly. So far, only maize
imports from South Africa were considered the major competitor of locally
produced millet. In the future millet imports from Angola, Zambia, and Zimbabwe
might secure commercial traders' grain supply in poor production years. Millet
imports from abroad might functions as price cap on farm level millet prices in
poor rainfall years and such discourage Namibian producers from investing in
new production technologies.

During the marketing season of 1993, most commercial millet traders sold
between three and six times more maize meal than millet. However, over 90
percent of them claim that their customers prefer the taste of millet over that of
maize. They explain that only those people buy maize meal that can not afford
the high prices of commercially traded millet.

Findings about millet traders' profit margins indicate that at least some of
them take advantage of local shortages of millet. The net returns of millet
traders average 37 percent from their millet retail prices. However, most urban
traders pass the cost advantage from cheaper bulk supplies on to their
customers. Their profit margins of N$ 380 per ton of traded millet are roughly
half of what rural traders earn. On the other hand, most rural millet traders have
only small trade volume and face higher risks from millet trade because the
invest large portion of their operational funds in millet.

Grain traders identified the following six obstacles to commercial millet
trade: (1) lack of marketable millet surplus, (2) lack of transportation equipment,

(3) lack of operation funds to purchase millet from producers, (4) lack of market

203
transparency; (5) lack of an appropriate millet processing technology, and
(6) comparatively low maize meal prices.

To increase millet production traders recommended that the government
expands its mechanical plowing service, implements an effective agricultural
extension service and provides a credit scheme for farm investments. To
improve the prospects of the commercial millet market, traders asked for
investment credits so that they can improve their transportation capacity and
venture into mechanical millet processing. Additionally, traders ask that the
government support long-distance trade between areas with marketable millet
surplus and millet deficit areas. Commercial millet traders also suggested that

the government initiates the organization of local markets.

4.4.3. Price position of millet

The analysis of the formation of millet and maize prices during the year
1993 by zone and by different marketing stages indicates that informally traded
millet is the cheapest food staple source for rural households in both study
zones. But informal markets are usually bound within the locality of a
community, and are therefore very volatile to low millet yields after a poor rainfall
season. As a result, millet was only seldom available on the informal millet
market.

The second cheapest food grain in the study zones is maize meal sold in
urban centers of both zones. Most of the maize is imported from South Africa

and is processed by Namibia’s two milling corporations south of the study zones.

204
At present only 25 percent of the rural households in Ovambo and 15 percent of
rural Kavango households acquire maize meal from center towns. ,

In Ovambo, commercially traded millet was the third expensive source
during the year 1993, just below the price of maize meal sold in rural areas.
Because, during the year 1993 food retailers Offered unprocessed millet only for
short periods of time, most grain deficit households in Ovambo had no
alternatives other than buying expensive maize meal in the nearest shop, or
traveling to the next town and buying larger amounts of maize meal at cheaper
kg prices.

In Kavango, the drought of 1991/92 had caused millet shortages at the
informal millet market and very high millet kg prices at commercial food retailers.
Maize meal sold in rural and urban areas was significantly cheaper than
unprocessed millet. Because millet imports from Angola had to be shipped via
Ovambo to central Kavango, and because the in Rundu concentrated grain
traders did not pass the cheaper prices of Angolan millet to their consumers,
commercial millet prices continued to stayed above maize meal prices even after
the 1992/93 millet harvest. After harvest, informally traded millet was
significantly lower priced than the millet and maize Offered on the commercial
grain market. However, the millet offer on the informal millet market was very
thin, so that most of the farmers in need of grain had to acquire one of the more

expensive alternatives.

205

The two lessons learned from the price analysis of millet and maize are:
(1) Locally produced millet is the cheapest calorie source for the rural
population in both study zones. However, informal millet trade among farmers is
volatile due to production shortages on the household level. Therefore,
increased millet production due to improved production technology would
probably benefit those households most that depend mainly on informal millet
sales from their neighbors.
(2) As long as commercial prices for millet in Kavango are kept so high that
only households with income from off-farm employment can afford it, the
transportation of Angolan millet to Kavango will continue to be profitable.
Households with grain shortages but little cash income will be further forced to

purchase maize meal.

5. MILLET CONSUMPTION

The preceeding marketing chapter demonstrated that while informally
traded millet constitutes the cheapest calorie source for rural households in
Ovambo and Kavango, prices of millet offered at the retail level are considerably
higher than those of its main competitor, maize meal. To shed light on the
question of why it is possible that grain traders offer millet at prices higher than
those of processed and more refined maize meal, Chapter 5 will focus on three
grain consumption issues: Section 1 of this chapter describes both the
preference structure and consumption habits of rural consumers with respect to
major food grains. Section 2 presents findings about the limiting factors on millet
consumption and farmers‘ coping strategies regarding shortages Of millet and
maize meal. Finally, section 3 employs two cross-section regression models
(OLS) to identify the main determinants of farmers' millet and maize meal

purchases during the 1992/93 production season.

5.1. Grain preference and consumption structure

Section 1 Of this chapter addresses two issues. It first analyzes which
staple grain is preferred by rural households in general and by individual
household member categories, as well as which factors determine this
preference structure. The second issue concentrates on rural households' diet,
with special reference to the main food staples, millet and maize, and the

consumption shift between the two during the production season 1992/93.

206

207

5.1.1. Cereal consumption choices based on preference

During the household surveys of the Millet Subsector Research Project,
1992/93, survey respondents were asked twice to state the type of food staple
they prefer. When the question was posed the first time, only a limited choice
between the two most commonly consumed food staples, millet and maize, was
presented to the respondents. This reduction of choices was based on the fact
that a large portion of rural households face cash constraints that limit their
consumption choices to the least expensive staple foods available. When
household representatives were asked the second time about their most
preferred food staple, a larger variety of food grains was given as a choice. The
goal of this preference question was to determine whether consumers would eat

other food staples if their cash constraints were removed.

5.1.1.1. Preference structure under limited choices

As Table 5-1 demonstrates, under a limited choice between millet and
maize, most respondents chose millet over maize. This finding confirms grain
traders' statements presented in Chapter 4, that the majority of northern
Namibia's population prefers millet over maize. However, the fact that already
20 percent of rural Ovambo households were indifferent in their preference
between millet and maize indicates that the traditional preference structure has

begun to weaken.

 

208

Table 5-1. Consumer preference structure between millet and maize,

 

 

by zone (percent)
Prefer Like Both Prefer Total
Millet Equally Maize
OVAMBO (n = 200) 77 20 3 100
KAVANGO (n = 120) 94 3 3 100

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

5.1.1.2. Preference structure without limits

To understand whether rural consumers would switch to another food
staple if it were offered at the price level of millet or maize, the following six
commonly eaten cereal types were offered to choose from, independently from
their price: (1) millet meal, (2) maize meal, (3) a mix of millet and maize meal,
(4) bread, (5) rice, and (6) macaroni. Except for children, who mostly prefer
bread, pure millet meal was most often chosen by all types of household
members.

In Ovambo, more than 50 percent of adults prefer pure (non-mixed) millet
meal over other staple foods (Table 5-2). Another 20 percent preferred a mix of
millet meal and maize meal. Rice and pure maize meal are each preferred by
about 10 percent of adults in Ovambo. In Kavango, more than 70 percent of

adults favor pure millet meal. Maize meal is preferred by roughly 20 percent.

 

209

Contrary to Ovambo, mixing millet with maize seems not to be preferred among

Kavango people. Rice, macaroni, or bread were mentioned by the remaining

10 percent of adults from Kavango.

 

 

 

 

 

 

Table 5-2. Household’s preference for different cereals, by household
member types and zone (percent)
Elderly Women Men Migrant Children
Worker"
Ovambo (n = 200)
Millet 66 45 51 61 20
Millet/Maize Mix 11 22 25 1 1 13
Maize 6 1 1 12 6 2
Rice 1 0 1 3 1 1 12
Macaroni 3 1 1 4
Bread 1 0 50
Total 100 100 100 100 100
Kavango (n= 120)
Millet/Maize Mix 0 3 0 14 4
Maize 20 18 20 0 1 0
Rice 0
Macaroni 0 5
Bread 0 50
Total 100 100 100 100 100
Data Source: Namibian Millet Subsector Project Surveys, 1992/93
" Migrant workers who leave their household for part of the year to work in Wrndhoek

or one of the southern mining towns.

 

21 0

The comparison between the two study zones demonstrates that in
Ovambo, people's taste preferences have a much wider range among various
cereal types than in Kavango. Besides maize and the mix of it with millet,
significant portions of Ovambo people developed, over time, a preference for
rice, macaroni, or bread. That taste preferences in Kavango concentrate on
millet and maize can mainly be explained by less urban influence in the zone and
therefore also less exposure to 'modern food staples'.

Having a much higher population density, Ovambo developed a much
stronger urban based retail system that is well-connected by roads and offers a
much wider variety of different food items to its urban and rural consumers. In
Kavango, cereal imports from other countries are also available at food shops,
but the retail system of Kavango reaches to a much lesser extent into rural
areas. Accordingly, most of Kavango's farm population had no chance to get
acquainted with different cereals and their tastes.

Besides differences in food staple preference by zone, age also seems to
be related to food staple preferences of various household members' types. In
Ovambo, a clear majority of older people prefer millet among different cereals.
The preference data of younger adults indicate that the mix of maize meal with
millet or porridge made solely from maize are becoming increasingly popular
among younger generations.

That more migrant workers (former and current) prefer millet than men in
general can be explained with two reasons: the first reason is again related to

age. The proportion of older men that have been raised with millet and therefore

 

211
like it better is probably higher among migrant workers than among adult men in
general. Second, survey respondents explained that the men working in the
south have to eat and are constantly fed with porridge made from maize so that
they are eventually longing for the taste difference that millet offers.

Further, in both study zones, about half of all children were said to prefer
bread over other non-wheat cereals. This high preference for bread can be
explained by children's early exposure to it. After independence in 1991,
businessmen from urban Ovambo and Kavango invested in large-scale bakeries
in peri-urban areas. At the same time, NGOs like UNICEF promoted the
establishment of small-scale bakeries in rural areas to enhance employment and
income opportunities of rural women. Survey respondents explained that young
children prefer bread because of its sweet taste and because parents often give
it to them out Of convenience.

Looking at the age pyramid of northern Namibia's population, it becomes
obvious that already today young adults and children represent the majority of
the communal population in northern Namibia. And it is probably not wrong to
assume that tomorrow’s household decision makers are the children and
adolescents of today. The shift in the food preference structure of this emerging
consumer group away from millet towards more refined food products has
become evident already. Once the preference structure has changed, the
substitution of maize for millet in the traditional, but fairly simple, dishes will be

easy.

212
5.1.1.3. Determinants of cereal preferences

Survey respondents identified five general factors associated with cereal
preferences: (1) tradition, (2) availability, (3) low price and easy accessibility,

(4) nutritious value, and (5) taste. Surprisingly, the convenience of being able to
purchase maize in the form of meal that is ready to cook has not been given as a
reason for preferring maize meal over other food types. The convenience of
preparation was only mentioned in relation to the ‘modern’ food types: bread,
rice, and macaroni‘ (Table 5-3).

In Ovambo, the preference for millet and the millet/maize mix was mainly
determined by tradition and availability. For those who preferred either rice,
macaroni, or bread, the main reasons were good taste and the perceived high
nutritional value. Reasons for pure maize preference were mainly based on its
availability, and secondly, on its low price and easy accessibility. In Kavango,
considerations of nutritious value, i.e., the positive effect of a specific food on
consumers' health or their perception of hunger, are the most often used

explanations for the consumption preference for all cereals.

 

‘ The fact that the comparatively easy preparation of maize meal to
porridge was not mentioned by the survey respondents can be seen as an
indicator that the rural population of the study zones perceives millet still as their
main grain staple for which manual processing is normal. It can be assumed
that, with the introduction of mechanical processing of millet, opportunity cost
considerations for processing will become a more important issue.

 

213

Table 5-3. Determinants of staple food preferences stated by farmers

 

 

 

(percent)
Tradi- Availa- Price & Nutritious Good Total
tion labillty Acces Value Taste
8
OVAMBO (n = 200)
Millet 44 27 7 15 7 100
Maize 44 23 8 1 0 1 5 1 00
Millet/Maize Mix 1 1 64 14 7 4 100
“Bum“: R5“: 3 2 19* 33 43 100
acaronr
KAVANGO (n = 1 20)
Millet 17 12 0 58 1 100
Maize 0 0 0 100 0 100
Millet/Maize Mix 7 29 10 43 12 100
an“: R3“: 0 5 5* 55 35 100
acaronr

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

* includes convenience and modern lifestyle

The most surprising of these findings is that 'good taste' plays a minor role
in determining the preference for the two main staple grains, millet and maize.
Only the preference for more 'modern foods' was relatively often guided by

considerations of taste.

5.1.2. Food staple consumption pattern
The following section briefly describes what food staples households

consume and how the composition of rural households' food diet changes with

214

the arrival of the local grain harvest.

5.1.2.1. Seasonal staple grain consumption shifts

To gain insights into what rural Ovambo and Kavango households eat
during the year, they were asked about their daily diet in April 1993, i.e., eight
months after the last grain harvest, and in August 1993, i.e., when threshing of
the 1992/93 harvest had just ended. The changes of households' diet that
occurred within the period of four months was profound. Especially significant
was the consumption shift away from maize towards millet after harvest.

In Ovambo, the percentage of households that ate pure millet porridge
increased from 29 percent in April 1993 to 63 percent in August 1993
(Table 5-4). The percentage of households that ate pure maize porridge or a
porridge made from a maize/millet mixture declined from a total of 71 percent in
April 1993 to 38 percent in August 1993.

In Kavango, the shift from maize towards millet was even more significant.
The percentage of households that ate pure millet porridge increased from
8 percent in April 1993 to 69 percent in August 1993. The percentage of
households that ate pure maize porridge or a porridge made from a maize/millet
mixture declined from a total of 92 percent in April 1993 to 31 percent in August
1 993.

215

Table 5-4. Seasonality in grain consumption during the year
1992I93, by household member types, by zone (percent)

 

 

 

OVAMBO (n = 200) KAVANGO (n = 120)
before after before after
harvest harvest harvest harvest
Millet 29 < 63 8 < 69
Millet/Maize Mix 20 < 34 14 +I- 13
Maize 51 > 4 78 > 18
Total 100 100 100 100

 

Source: Namibian Millet Subsector Project Surveys, 1992/93

The significant diet change away from maize porridge at a time when
millet is available confirms farmers' statements about their consumption
preferences. It indicates, additionally, that for the majority of communal
households, staple food consumption is mainly supply driven; i.e., as long as
millet from their own production is available, it is the preferred food staple. Wrth
few exceptions, only when millet stocks decrease notably do households start to
stretch millet porridge with maize meal. This continues until no millet from their
own production is left and households have to rely entirely on purchased maize
until the next harvest.

To make sure that the supply of a certain grain has a major impact on the
consumption decision of rural households, farmers were asked before the 1993
grain harvest why they ate one of the two main staples, millet and maize. The

answers confirmed the above stated assumption. Of those Ovambo households

216
that ate pure millet in April 1993, the largest group (46 percent) explained their
consumption decision with millet's' exclusive availability (Table 5-5). The term
exclusive means that maize meal was either not accessible or not affordable
because it had to be purchased with money. The other major determinant for
millet consumption given in Ovambo. was tradition (43 percent). In Kavango, the
determinants for millet consumption were similarly weighted. Forty percent of
Kavango households claimed that millet's exclusive availability is the main
reason why they eat millet. Another 30 percent claimed they ate millet at that

time because of their tradition.

Table 5-5. Determinants of millet and maize consumption during the
1992I93 hungry season, by zone (percent)

 

 

 

Exclusive Tradition Price 8. Other Total
Availability Access
OVAMBO (n = 200)
Millet 46 43 6 5 100
Maize 93 0 7 0 1 00
KAVANGO (n = 120)
Millet 39 30 0 31* 100
Maize 70 0 22 8 100

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

" mainly high nutritious value and good taste of millet

217
The majority of those households solely consuming maize stated its
exclusive availability as the main decision criterion (Ovambo: 93 percent,
Kavango: 70 percent). The other factors often mentioned as reasons for maize
consumption were its relatively low price and accessibility in the market

(Ovambo: 7 percent, Kavango: 22 percent).

5.1 .2.2. Other seasonal diet changes

Not only the consumption of millet and maize changes during different
phases of one production year. Table 5-6 shows how rural households' diet
composition changed from the time period of grain production in April 1993, to
the time period after millet harvest in August 1993.

The consumption of goat milk was increased in both zones in April
compared to August. In Ovambo, also, the consumption of fresh and dried
vegetable leaves was strongly increased during April. In August 1993, after the
cropping season, the consumption of beans and of meat was increased in both
zones. In the same period of time, cow milk and cabbage were more often
consumed in Ovambo, while the consumption of fish and pumpkins was
increased in Kavango.

The consumption of more expensive grain staples did change comparably
little between April and August. In Ovambo, about 30 percent and in Kavango,
13 percent of all households include bread in their daily diet. Rice is consumed

by about 3 percent of households in both zones. Five percent of Ovambo

218

Table 5-6. Seasonality of farm households' food consumption, by zone
(percent, ranked according to change)

 

 

OVAMBO (n = 200) KAVANGO (n = 120)
Food Item April' change August“ Food Item April change August“
Millet 30 33 63 Millet 8 61 69
Millet Drink 57 30 87 Millet Drink 12 25 37
Beans 20 19 38 Fish 0 24 24
Cow Milk 4 17 21 Beans 22 16 38
Cabbage 1 1 7 1 8 Pumpkins 3 1 5 18
Meat 54 10 64 Meat 38 13 51
Fish 0 4 4 Poultry 0 3 3
Bread 28 3 30 Cabbage 0 3 3
Pumpkins 4 2 6 Sorghum 1 2 3

Drink

Poultry 0 1 1 Sauce 2 1 3
Maize Drink 2 -2 1 Bread 13 1 13
TealCOffee 39 -3 36 Leaves 64 0 64
Rice 5 -4 1 Rice 3 0 3
Sauce 24 -5 19 Eggs 1 0 1
Eggs 6 -5 1 TeaICoffee 11 -1 10
Macaroni 8 -6 2 Macaroni 2 -2 O
Sorghum 19 -7 11 MilletIMaize 14 -2 12
Drink Mix
Maize 20 -1 6 4 Cow Milk 8 -5 3
MilletIMaize 51 -17 34 Maize Drink 1 1 -8 3
Mix
Goat Milk 22 -22 0 Goat Milk 15 -15 0
Leaves 89 -35 53 Maize 78 -60 18

 

Data Source: Namibian Millet Subsector Project, 1992/93

* eight months after the last grain harvest
** when threshing of the 1992/93 harvest had just ended

 

21 9
households and one percent of Kavango households included macaroni in their
diet.

Even without knowledge about the actual amounts of different food items
eaten per household, the comparison of rural households' diet composition, over
time, yields three important insights: first, rural households' diet shows quite a
variety of food items besides the main food staples, millet and maize. Second,
rural households' diet changes significantly within one production year. Third,
the consumption of millet and maize are among those diet components that are
most affected by the change from food production phase to the post-harvest

period.

5.2. Constraints on millet consumption and coping strategies
This section discusses factors that limit the consumption of millet in rural
Ovambo and Kavango and presents strategies used by farmers to cope with

shortages of millet and Of maize.

5.2.1. Limits to millet consumption on the household level

The majority of rural farmers from the study zones do not produce enough
millet to reach grain self-sufficiency from one harvest to the next. Recurring
scarcity of millet on the informal market at the community level, combined with
insufficient links between various production areas, reduces the profitability of

commercial millet marketing in rural areas. As a result, only a small portion of

220
households with considerable off-farm income can afford to travel to, and
purchase millet from, urban grain traders.

AS described above, the majority of rural households that eat maize
instead of millet explain that their consumption decision is based on (1) the non-
availability of millet, (2) the relatively low price of maize compared to
commercially traded millet, and (3) the easier access to maize than millet.

To provide information about the limits of millet consumption, the following
sections will first give a brief overview of the amount of money farm households
spend for food, second, describe general problems that grain deficit farmers
experience when they want to acquire millet, and third, analyze the problem of

non-accessibility of millet in more detail.

5.2.1.3. Household level food expenditures

To understand how large the share of households’ food expenditures is of
their total financial expenditures, survey respondents were asked after the
1992/93 harvest to state the three types of items for which they spend most
money during one month. Households were also asked to estimate the
approximate amount they spend on these important expenditure categories. The
following two tables show the average distribution of farmers' most important
expenditures in August 1993. Table 5-7 demonstrates the expenditure
distribution of all interviewed farmers from Ovambo and Kavango. Table 5-8 dis-
aggregates the expenditure analysis by grouping households in categories of

different total monthly expenditure levels.

221
As both tables demonstrate, the largest expenditure category on which
rural people spend money is food. A surprising result is that the proportion of
money spent for food does not decline with increasing household expenditures.
This points to the fact that households with higher incomes increase their food
expenditures, probably first by increasing the amount of food eaten and then by

substituting cheaper staple food categories with more expensive food items.

Table 5-7. Farmers' monthly cash spending across various items
in August 1993, by zone (percent)

OVAMBO (n = 188) KAVANGO (n = 91)

 

 

 

Food 53 67
Household Goods 23 0*
Clothing 13 . 1 5
Transportation 3 6
School 2 1
Other 6 1 1
Total 100 100

 

Data Source: Namibian Millet Subsector Project, 1992/93

" AS already explained elsewhere in the study, the rural population of Kavango
is to a much lesser extent linked to urban biased consumer markets; i.e., they

are more subsistence-oriented than Ovambo people and are, therefore, much

less used to purchasing household goods.

222

Table 5-8. Monthly cash spendings across various items by household
expenditure categories and zone (_percent)

 

 

 

 

 

 

Household 1 1 01 201 301 401 501 601 av-
Expenditure 0 - - - - - - < er-
Leve. in N3 100 200 300 400 500 600 32.
Percent of OVAMBO (n = 1 88)

Households 6 23 28 24 10 4 1 3 100
Expenditures

Food - 58 60 47 44 42 49 61 53
Household - 27 23 24 21 16 17 8 23
Goods

Clothes - 6 1 1 20 19 24 27 5 1 3
Transportation - 3 1 3 3 0 0 3 3
School - 3 3 3 0 4 0 1 2
Water - 0 1 2 11 0 0 0 2
Health - 1 2 2 3 5 0 8 2
Drinks - 2 0 0 0 0 0 0 1
Rent - 0 0 1 0 0 6 0 1
Remittances - 0 0 0 0 0 0 4 0
Hired Labor - 0 0 0 0 0 0 9 0
Total - 100 100 100 100 100 100 100 100
Percent of KAVANGO (n = 91)

Households 24 7 9 16 23 11 11 0 100
Expenditure

Food - 58 84 61 63 70 69 na 67
Household - 0 0 0 0 0 na 0
Goods

Clothes - 3 6 23 17 16 1 1 na 15
Transportation - 0 0 0 3 0 0 na 6
School - 39 5 2 1 3 1 na 1
Water - 0 0 0 0 0 0 na 0
Health - 0 0 5 5 4 7 na 4
Drinks - 0 5 2 5 3 6 na 4
Rent - 0 0 0 0 3 0 na 0
Remittances - 0 0 4 2 0 3 na 2
Hired Labor - 0 0 4 4 0 3 na 1
Total - 100 100 100 100 100 100 100 100

 

Data Source: Namibian Millet Subsector Project, 1992/93

223

In Ovambo, about 75 percent of households spend between 100 and 300
Namibian dollars per month for the three most important expenditure goods.
Only 6 percent of all households claimed they have no cash income to spend.
Of the money spent per month, the average share spent for food amounts to
about 53 percent across all Ovambo households. In Kavango, 24 percent of the
interviewed households claimed they have no cash expenditures. But
39 percent spend, on average, between 200 and 400 Namibian dollars a month
for the three most important expenditure goods. Those households that have
money spent, on average, 67 percent of it for food. The higher levels of food
expenditures in Kavango were caused probably by two circumstances: (1)
Kavango's relatively poor results of the 1992/93 grain harvest that triggered the
need of more food purchases, and (2) low expenditures of Kavango households

for household goods.

5.2.1.2. Obstacles to millet purchases

The interviewed households had to indicate factors that limit their buying
of millet. In Ovambo, the three most often indicated problems were 'lack of
money', 'high transportation cost to and from the place of purchase', and 'high
millet prices' (Table 5—9). In Kavango significantly higher proportions of people
experienced diffictu purchasing millet. In addition to the problems already
identified for Ovambo, in Kavango, cheaper and more easily accessible maize

were additional reasons that limit households' millet purchases.

224

Table 5-9. Household food constraints on millet purchases, by zone
and household food buying frequency (percent‘)

 

 

 

Ovambo Households Kavango Households
(n=200) (n=120)
Food buying frequency never some- often total never some- often total
times times

5 78 17 100 0 73 27 100
Constraints
not enough money 100 84 88 85 ~ 91 90 91
expensive transport 50 51 53 51 - 76 52 69
high millet prices 57 49 44 48 - 90 100 93
no transport 50 48 38 46 - 41 35 39
long distances 33 43 44 43 - 47 29 42
no local market 33 37 28 36 - 41 26 37
difficult to find seller 33 28 31 29 - 36 35 36
easier access of maize 17 13 44 19 - 91 97 92
unknown price 17 17 28 19 - 24 13 21
cheaper maize 17 1 1 41 17 - 91 94 92

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

* since several answers were allowed percent figures do not add up to 100

5.2.1.3. Accessibility of millet in comparison

As already demonstrated in Figure 4-1, between February 1993 and
January 1994, millet was to a much lesser extent available at rural food retailers
than maize meal. When survey households were asked whether millet or maize
is generally easier to purchase, responses were clearly in favor of maize. In
Ovambo, 57 percent of all households said maize meal is easier to buy than

millet. In Kavango, the respective figure was even 91 percent.

225

Table 5-10. Comparison between farmers' access to millet and
maize, by zone (percent)

 

 

 

OVAMBO KAVANGO

(n = 200) (n = 120)
Millet is more accessible 17 7
Both are equally accessible 26 2
Maize is more accessible 57 91
Total 100 100

 

Data Source: Namibian Millet Subsector Project, 1992/93

Looking at the figures about the accessibility of various food grains to rural
households in April 1993 (Table 5-11), it becomes clear why millet is relatively
difficult to buy. In this month, only 44 to 45 percent of all interviewed households
claimed that unprocessed millet was available for purchase. Millet meal was
only marginally available to rural households.

At the same time, almost 100 percent of the interviewed households
indicated that maize meal was available to buy. However, unprocessed maize
was in both zones difficult to access. Bread was available to about 86 percent of
all Ovambo households and to about 67 percent of all Kavango households,
respectively. In both zones, rice was available to the vast majority of all
households, while sorghum could only be bought in Ovambo by a quarter of the

interviewed households.

226

Table 5-11. Households' access to various food grains during the
1992I93 hungry season in April 1993, by zone (percent)

 

OVAMBO (n = 200) KAVANGO (n = 120)

 

Unprocessed Millet 45 44
Millet Meal 2 4

Unprocessed Maize 56 43
Maize Meal 99 100
Bread 99 87
Rice 94 82
Sorghum 24 0

 

Data Source: Namibian Millet Subsector Project, 1992/93

5.2.2. Coping strategies of farmers in grain shortage

To determine how rural households react to millet and maize shortages
and whether this reaction differs between the two grains, sample farmers were
asked what they do when the stocks of either grain empty. The results of this

inquiry are briefly presented in the following two sections.

5.2.2.1. Strategies to cope with millet shortage

Farmers use mainly four different strategies to cope with declining millet
reserves. They first change consumption habits, second, ask others for food,
third, work for food or money, and fourth, purchase the food that is out of stock
from various sources. However, the composition of these coping strategies

differs significantly between Ovambo and Kavango households (Table 5-12).

227

Table 5-12. Farmers' coping strategies during millet and maize
shortages, by zone (percent)

Millet Shortage Maize Shortage

 

Ovambo Kavango Ovambo Kavango
(n = 200) (n = 120) (n = 200) (n = 120)

 

Coping Strategy:
mix millet with maize 43 47 2 5
eat less 17 70 1 26
eat other food 17 60 5 40
ask relatives for grain 38 12 8 9
ask relatives for money 13 6 7 5
ask neighbors for help 26 8 7 13
work for food or money 1 24 2 29
buy the grain in shortage 58 19 92 44

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

In Ovambo, the first reaction to millet shortage is to stretch the existing
millet meal with maize meal or eat less millet and change to other food types.
The next step is to ask relatives or neighbors that still have millet reserves to
give part of it to the household. Relatives with migration income are asked to
provide money for the purchase of food. Eventually, 58 percent of all Ovambo
households buy millet.

Kavango households mostly reduce millet consumption and shift to other
food items of which some, like nuts and fruits, are collected in nature. Also the
stretching of millet meal with maize meal is prominent. In comparison to

Ovambo, fewer households ask relatives or neighbors for help. Also different is

 

/—‘et

228
that about a quarter of the interviewed Kavango households are willing to work
for others to receive food or money. Only a fifth of the sample households in
Kavango eventually buy millet.

The large difference in the number of Ovambo and Kavango farmers that
purchase millet is explained by the much higher proportions of Kavango farmers
that indicate obstacles to millet purchases like 'lack of money', 'expensive
transportation', and 'high millet prices' in comparison to maize meal prices. Most

of the farmers that purchase millet acquire it from various sources (Table 5-13).

Table 5-13. Rural households' purchase sources for millet and maize,
by zone (percent)

 

Millet Shortage Maize Shortage
Ovambo Kavango Ovambo Kavango
(n = 200) (n = 120) (n = 200) (n = 120)

 

 

.. 5. .. 92 44

Sources:
Neighbors 22 9 11 11
Local Shop 13 6 66 32
Local Market 5 1 22 2
Urban Market 3 0 8 0
Urban Supermarket 10 6 54 18
Urban Wholesaler 4 4 41 13
Traveling Traders 14 0 8 0
Angolan Traders 14 3 4 2
Angola directly 2 0 2 0

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

229
In both study zones, the largest group of farmers buys millet from
neighbors. Local shops and urban supermarkets are also important sources of
millet in both zones. In Ovambo, traveling traders that originate from northern

Namibia and from Angola also play an important role for millet supply.

5.2.2.2. Farmer reactions to emptying maize reserves

Of all surveyed households in the study zones, only aboutone percent
stated they do not eat maize meal. In Ovambo, the vast majority of households
(92 percent) go and buy maize meal once their own stocks are empty (see Table
5-12 above). In contrast, Kavango households respond to the emptying of their
maize meal reserves the same way as they respond to millet shortage. Only
about 44 percent of the interviewed Kavango households would buy maize meal
to replenish their stocks. A quarter of them reduce their maize meal
consumption. Almost half of them switch to other foods. And a third of the
sample farmers start to work to be paid in food or money to purchase food.

Table 5-1 3 above demonstrated that the largest group of farmers acquire
maize meal from local shops. But urban supermarkets and wholesalers are also
an important source of maize meal, especially in Ovambo. Unlike with millet,

traveling traders are a relatively unimportant source of maize.

5.3. Determinants of farm level grain purchases
To make any predictions about the commercialization prospects of the

communal millet sector, it is important to know under what circumstances

230
northern grain consumers would buy locally produced millet rather than maize
meal that is imported into the study zones. Accordingly, this final part of
Chapter 5 focuses on the main determinants of grain purchases on the
household level.

The analysis of household level consumption pattern and obstacles to
millet consumption above demonstrated that the grain buying decision of
farmers is influenced by a complex set of factors. In the following, the attempt is
made to test the significance of the effects of these factors through two OLS
regression models. Due to the lack of detailed time series data about farm
household grain purchases, both regressions are based on cross-section data of
farmers' food purchases madeat a certain point of time. The independent
variables are households’ purchases during the period of the first household
survey in December 1992 of (a) unprocessed millet in the regression model of
the following section 5.3.1. and (b). maize meal in the regression model of section

5.3.2..

5.3.1. Regression analysis of millet purchases

The following sections give an overview of the cross-section regression of
millet purchases per household in December 1992, discuss the independent
variables employed in the equation, and presents the regression results.

The cross-section model of farmers' millet purchases in December 1992
during field preparation encompasses variables that identify (a) households’

millet supply from own production, (b) economic determinants of purchasing

231
behavior like grain prices and cash incomes, and (c) factors that affect farmers’

market access to millet.

MILPRCH = f (MILRES, PMIL, PMAZ, CASH, CASHZ, CESZ, CEMKT, LOMKT)
where:

MILPRCH = amount of millet purchased during field preparation in 1992 (kg)

MILRES = household millet reserves from their own production (DUMMY, 1 = yes)
PMIL = price of millet quoted by household for December 1992 (NS/kg)

PMAZ = price of maize meal quoted by household for December 1992 (NS/kg)
CASH = annual cash income from non cropping activities (N5 '000)

CASH2 = squared annual cash income from non-cropping activities (N5 '000)
CESZ = resides in a subzone with urban markets (DUMMY, 1 = yes)

DISMKT = distance to next larger grain market (km '0)

LOMKT = have access to local market (DUMMY, 1 = yes)

Table 5-14. Characteristics of variables in the millet

 

 

purchase equation
Ovambo 8. Kavango (n=265)
VARIABLE MEAN S.E.
MILPRCH (kg) 8.92 21.99
MILRES (1 = yes) 40 percent = 1
PMIL (NS/kg) 1.41 0.73
PMAZ (NS/kg) 1.20 0.16
CASH (N5 '000) 6.70 9.67
CAsH2 (NS '000 2) 1.38 4.60
CESZ (1 = yes) 25 percent = 1
DISMKT (km '0) 81.55 62.49
LOMKT (1 = yes) 48 percent = 1

 

Data Source: Namibian Millet Subsector Research Project, 1992/93

232

The independent variables used and the reasons for their inclusion in the
estimated equation are as follows:

(a) the availability of millet reserves from their own production at the time
of grain purchases during field preparation in December 1992 (MILRES). After
the severe drought of the production year 1992/93, food aid in the form of maize
meal and canned food was distributed in the communal north of Namibia.
Farmers that tried to be classified as 'in need of handouts' were therefore very
reluctant to reveal their actual status of grain reserves. Due to the sensitivity of
the topic, questions in the direction of grain reserves had to be omitted in all
three household surveys during the 1992/93 production season. However,
during the April 1993 household survey, two months before the start of the millet
harvest, household representatives were asked what food items they were
currently eating. if millet was mentioned, it was inquired whether this millet came
from their own production or whether it was bought somewhere else. It is
hypothesized that those farmers that ate in April 1993 millet that was produced
on their own fields were less likely buyers of millet in December 1992 than those
who either ate purchased millet or purchased maize meal. To account for this
effect, the dummy variable MILRES was included in the regression model to
distinguish between those households that were probably in grain shortage in
December 1992 (MILRES = 0) and thus more likely to buy millet or maize, and
those households that had enough millet from their own production still in April

1993 (MILRES = 1). According to this reasoning, it is expected that the

233
coefficient of the MILRES variable is negatively correlated with household level
millet purchases.

(b) the price of millet and of maize meal that farmers faced during
December 1992 (PMIL and PMAZ). Not only the absolute price of millet, but also
the discrepancy between the price of millet and maize meal affects farmers'
decision whether to buy millet at all, and if so, how much. As discussed in the
first section of this chapter, the majority of the rural population prefer millet over
maize. From this preference structure, one can expect that households with
considerable amounts of cash income are willing to pay a premium for millet,
although it is not available in processed form as is the case with maize.

However, the larger the price gap between millet and maize is, the more likely is
it that consumers choose to buy the cheaper maize meal.

It is also assumed that the purchase decision of households with very
scarce cash sources will almost always be made in favor of the absolute cheaper
grain. Poor farmers probably take less into consideration that millet still has to
be processed manually. For them, money is the main constraint, while they
value their opportunity cost of labor close to zero.

The hypotheses that millet purchases are both negatively affected by
increases of millet prices and positively affected by a decline of maize meal
prices are not solely based on economic theory. As demonstrated in the second
section of this chapter, 'lack of money', 'high millet prices’, and 'comparatively low
maize meal prices' have been ranked among the most limiting factors for millet

purchases.

234

To control the effects that millet and maize meal prices have on the
decision to purchase millet, two variables are included in the model. These
variables reflect the local price of millet (PMIL) and of maize meal (PMAZ) the
farmers faced in December 1992.

(c) the amount of farm households' annual cash income and the squared
value of it (CASH and CASHZ). As demonstrated in Table 5-9, farmers mention
'lack of money' most often as an obstacle to millet purchases. This is not
surprising. In both study zones, more than 50 percent of rural households do not
earn cash income from local or migration employment. The average annual
cash income per capita of these households is about N5 100 or US$ 33 at the
exchange rate of 1993. Based on this information, it is hypothesized that millet
purchases are positively correlated with annual cash income represented by the
variable CASH.

Since the per capita consumption of food staples is naturally limited, and
because higher income households tend to substitute food staples with higher
value food like meat, the relation between annual cash income and grain
purchases is probably not linear. To enable the model to capture the fact that
the marginal increase of millet purchases is diminishing, while cash income
increases, the variable CASH2 is included. It is expected that the estimated
coefficient of the CASH2 variable is negative.

(d) the location in a center subzone with urban grain markets (CESZ).
The independent variables discussed so far concerned household level millet

supply and economic decision factors. The following three variables discussed

235
and included in the model focus on the accessibility of millet for purchase. The
dummy variable CESZ distinguishes between households that live in the area
near the urban trading centers of Ovambo and Kavango (value one) and
households that live in the periphery of the study zones (value zero). As
presented in the marketing Chapter 4, many Iocalfinfonnal millet markets
collapse within a few of months after harvest due to lack of local supply.
Therefore, most commercially offered millet is available in urban centers like
Oshakati I Ovambo, originating from Angola, and in Rundu I Kavango, coming
from a few larger Kavango farmers producing millet surplus.

It is assumed that farmers living close to the urban centers are more
knowledgeable about the millet offered at center markets and face lower
transportation costs to reach millet suppliers than farmers from the periphery.
Accordingly, it is hypothesized that farmers from subzones close to urban
population centers are more likely to buy millet. Therefore, the coefficient of the
CESZ variable is expected to be positive.

(e) the distance to the next larger grain market (DISMKT). As data about
staple grain sources of rural households have shown (Table 5-13 above),
farmers who are in millet deficit do not rely solely on local/Informal millet sources
but also purchase grain from urban supermarkets and wholesalers. However, as
price data have shown, maize meal available at these market outlets is usually
significantly cheaper than unprocessed millet. From this, the hypothesis is
derived that price-conscious farmers that live closer to larger sources of millet

and maize meal are less likely buying millet, but maize. The latter part of this

236
hypothesis will be tested in the regression model that estimates household level
maize meal purchases. The first part will be tested in this model through the
DISMKT variable. This variable contains the kilometer distances (measured in
10 kilometers per unit) between individual sample farmers and the nearest point
where they can purchase larger amounts of millet or maize. It is expected that
the coefficient of the DISMKT variable is positively related to the amount of millet
purchased. Again, the main reason for this hypothesis is that farmers who are
further away from urban grain traders often have less opportunity to buy cheaper
maize meal.

(f) the access to a local grain market (LOMKT). In both study zones,
about a third of the survey respondents pointed to missing local grain markets as
an obstacle to purchasing millet. As discussed in the millet marketing Chapter 4,
in communities without locally organized grain trade surplus, producers might
develop the tendency to bypass poor deficit farmers that can only afford to buy
grain in small quantities. The millet surplus producers tend to sell their produce
in bulk to distant grain traders in urban centers. Often, such a pattern leaves
deficit farmers no other choice than to purchase relatively high-priced maize
meal at local shops. To estimate the effect that the existence of a local grain
market has on millet purchases, the dummy variable LOMKT is included in the
model. If the surveyed farmers had access to a local grain market, the LOMKT
takes value one; othenNise, its value is zero. It is hypothesized that the

MILPRCH variable is positively related to the existence of local markets.

237

Table 5-15 presents the main statistics of the estimated equations for
millet purchases. The level of the adjusted R-Square indicates that the model
explains 30 percent of the variability of household level millet purchases. This
level of explanation, which is not bad for household level surveys, is mainly
attributable to the rough information about farm level millet purchases based on
a one-time interview. However, despite the low R-Square value, the signs of all
regression results are consistent with expectations, and all considered
explanatory factors demonstrate a significant relation to the amount of millet

purchased.

Table 6-16. Estimates of the millet purchase equation

 

 

 

Ovambo 8 Kavango

VARIABLES COEF. S.E. SIGNIF.
CONSTANT -14.87 6.15 0.01
MILRES -3.70 1.48 0.01
PMIL -3.32 1.04 0.00
PMAZ 13.97 4.88 0.00
CASH 0.48 0.15 0.00
CASH2 -0.007 0.003 0.02
CESZ 14.60 1.79 0.00
DISMKT 3.20 0.01 0.01
LOMKT 5.41 1.52 0.00
Adjusted R- 0.30

Case Number 256

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

238

In the following interpretation Of the regression results, the base case
represents a farm household that had no millet reserves from its own production
available in April 1993 (MILRES = 0), is located on the periphery rather than
Close to the population centers Of the zone (CESZ = 0), and has no access to a
local grain market (LOMKT = 0).

Holding all other factors constant, a rise of local prices of millet (PMIL) by
one Namibian dollar per kg leads to a decrease of farmers' monthly millet
purchases by 3.3 kg. This is a reduction of about 30 percent of the average
amount households bought during December 1992. A price decline of millet's
closest substitute, maize meal, by one Namibian dollar triggers, on average, a
decline of millet purchases by 14 kg. Both results are statistically significant and
conform with economic theory. The fact that farmers buy, on average, three
times more maize meal than millet probably explains why they react more
strongly to price Changes of maize meal than to changes of millet.

The variable that also influences strongly household level millet purchases
is CESZ. Those farmers living in a subzone close to a large center market of
their zone buy, on average, 14.6 kg more millet than those who live on the
periphery. This result reaches a significant level of 0.00 percent. It confirms the
assumption that living close to a market source where millet is available most
times of the year has a positive effect on the purchases of staple food that
consumers prefer. The other two explanations for this outcome are: first,
closeness to the urban center markets reduces farmers' transaction cost for grain

purchases significantly. Second, households living close to urban centers more

239
likely have household members that generate enough cash income from off-farm
employment, allowing the purchase of millet even if it is more expensive than
maize meal.

Another factor that shows larger effects on farm level millet purchases is
local grain markets. The marginal effect of a nearby market is a 5.4 kg increase
in households' millet purchase per month. This finding is especially important
because disseminating the idea of locally organized grain marketdays among
rural communities is a relatively simple measure, compared to changing factors
like grain market prices.

The hypothesis that farmers living further away from larger grain markets
are buying more millet but less maize is confirmed with respect to millet at a
significant level of 0.01 percent. Wrth every 10 kilometer distance, farmers
purchase about 0.32 kg more millet per month.

Finally, cash income from activities other than crop production does not
seem to be closely related to households’ millet purchasing behavior. Although
significant at 0.00 percent, the estimation results show that with every additional
N$ 1,000 annual cash earnings, farm households buy only half a kilogram more
of millet per month. But this result shows that the communal population doesn't
perceive millet as an inferior good whose consumption is automatically reduced

with increasing income.

240
5.3.2. Regression analysis of maize meal purchases

This section gives first an overview of the cross-section regression model
of maize meal purchases per household in December 1992. Then it describes
the independent variables employed in the equation. Finally, regression results
are presented and discussed.

The cross-section model of rural households' maize meal purchases in
December 1992 during field preparation‘ incorporates one additional variable
compared to the model estimating millet purchases. Besides variables that
reflect (a) households' millet supply from their own production, (b) economic
factors like grain prices and cash incomes, and (c) farmers’ access to millet, a
fourth variable type (d) was added, mirroring farmers' exposure to more 'modem'

or urban-biased attitudes that might influence their grain purchase decision.

MAZPRCH = f (MIL92, PMIL, PMAZ, CASH, CASH2, MILAV, WEEA, DISMKT, LOMKT, READ)
where:

MAZPRCH = amount of maize meal purchased during December 1992 (kg)

MIL92 = millet produced during the 1991/92 production season (mt)

PMIL = price of millet quoted by household for December 1992 (NS/kg)
PMAZ = price of maize meal quoted by household for December 1992 (NS/kg)
CASH = annual cash income from non-cropping activities (N5 '000)

CASH2 = squared annual cash income from non-cropping activities (NS '000)

 

‘ These purchases also include maize meal purchases brought by
migrant workers returning to their home. These usually buy maize meal in the
center town of their zone before they organize their transportation to their
homestead in the bush.

241

MILAV = household stated ’millet is available for purchase' (DUMMY, 1 = yes)
WEEA .= geographic west/east location (No. 1 - 6, 1= West, 6 =East)
DISMKT = distance to next larger grain market (km '0)

LOMKT = have access to local market (DUMMY, 1 = yes)

READ = number of household members that can read (persons)

Table 5-16. Variables in the maize meal purchasing

 

 

equafion
Ovambo 8. Kavango (n-266)

VARIABLE MEAN S.E.
MILPRCH (kg) 24.98 43.63
MIL92 (mt) 0.77 3.99
PMIL (NS/kg) 1.41 0.73
PMAZ (NS/k9) 1.20 0.16
CASH (NS '000) 6.70 9.67
CASH2 (NS ‘000 2) 1.38 4.60
MILAV (1 = yes) 53 percent = 1

WEEA (No. 1-6) 1, 3 =25 % and 2, 4, 5, 6, =12.5 %
DISMKT (km '0) 81.55 62.49
LOMKT (1 = yes) 48 percent = 1

READ (persons) 5.52 4.30

 

Data Source: Namibian Millet Subsector Research Project, 1992/93

The independent variables used and the reasons for their inclusion in the
estimated equation are as follows:

(a) the amount of millet produced during the production year 1991/92
(MIL92). As for the regression model estimating millet purchases, it is

hypothesized that farmers buy less grain on the informal and commercial market

242
the more millet they have in stock from their own grain production. Lack of data
about farmers' actual grain storage levels in December 1992 make it necessary
to use a variable closely related to farm level millet reserves. For this cross-
section regression, farmers' harvest results from May/June 1992 were used as a
proxy for farmers' supply of millet from their own production in December 1992
(MIL92). It is assumed that the coefficient estimated for MIL92 is negative.

(b) the price of millet and of maize meal that farmers faced during
December 1992 (PMIL and PMAZ). According to economic theory, both maize
meal and millet prices are supposed to affect farmers' grain purchasing decision.
To control for these effects, the price variables PMIL and PMAZ are included in
the model. Each variable reflects the respective grain price farmers faced in
December 1992. It is hypothesized that PMIL is positively and PMAZ is
negatively correlated with maize meal purchases.

(c) the amount of farm households' annual cash income and the squared
value of it (CASH and CASH2). Economic theory predicts that disposable cash
income affects household food consumption. It is believed that 'lack of money'
does not only limit household level millet purchases (see Table 5-9 above), but
also the purchase of maize meal. It is assumed that maize meal purchases are
positively correlated with annual cash income (CASH). Since grain consumption
per capita is naturally limited and because higher income households tend to
substitute higher value food for staple food, the variable CASH2 of the squared

annual cash income is added to enable the model to estimate the fact that the

243
marginal increase of millet purchases are diminishing with cash income
increases.

The next four variables included in the model focus on the accessibility of
maize meal or its substitute, millet:

(d) the availability of millet to farmers (MILAV). Not only the price of maize
meal's major substitute, millet, affects households’ grain purchasing decision.
The availability or accessibility of millet to individual households is also an
important decision factor. Only about half of all interviewed farmers stated they
had access to millet for purchase in December 1992. This implies that many of
the households in grain deficit had no alternative other than buying maize meal
to cover their food staple needs. To estimate the effect that the general
accessibility of millet has on farmers' maize meal purchases, the variable MILAV
was included in the model. If households stated in December 1992 'they had
access to millet' the value of MILAV is one; otherwise, it is zero. It is
hypothesized that the coefficient of the MILAV variable is negative.

(e) the geographic west/east location of farmers within the study zones
(WEEA). Infrastructure like roads and communication services are better
developed in western than in eastern parts of the study zones. A similar
distribution exists for farm households' levels of formal education and annual
cash income. It is difficult to determine what exactly caused these imbalances in
development. However, it can be speculated that several factors, like cultural,

historical and political events, as well as different levels of population density led

244
to this situation. Nevertheless, it is assumed that differences in infrastructure
affect marketing of locally produced millet and imported maize meal differently.

In areas with less technical infrastructure, the collection, advertising, and
eventual transporting of marketable millet surplus is more difficult than
distributing well- packaged maize meal that is shipped together with other goods
from urban centers to rural shops. From this, it is hypothesized that as maize
meal's comparative advantage over millet in marketing increases, the less
market infrastructure exists. To account for the fact that toward eastern areas of
the study zones market infrastructure worsens, the variable WEEA was included.
The variable uses ordinal numbers from 1 to 6 to reflect the west/east location of
the six subzones within the study zones, i.e., between the neighboring regions
Kaukofeld and Caprivi. It is hypothesized that the coefficient of the WEEA
variable is positively correlated with household level maize meal purchases.

(f) the distance to the next large grain market (DISMKT). Not only the
general level of market infrastructure, but also the distance to larger grain
markets affects the purchasing decision of farmers in grain deficit. As the
regression results about household level millet purchases demonstrate, the
further farmers have to travel to a larger market where they can buy relatively
cheaper maize meal, the more millet they buy. To test whether farmers who live
closer to grain markets buy more maize meal, the variable DISMKT used in the
regression model above is again included. The variable contains the kilometer
distances (measured in 10 kilometers per unit) between individual sample

farmers and the nearest point where they can purchase larger amounts of millet

245
or maize. It is expected that the coefficient of the DISMKT variable is negatively
related to the amount of maize meal purchased.

(9) the access to a local grain market (LOMKT). The existence of local
grain markets usually improves access to maize meal's main substitute, locally
produced millet. It is assumed that improved access to millet reduces farmers'
maize meal purchases. To test this hypothesis, the dummy variable LOMKT is
included in the model. If the surveyed farmers stated they have access to a local
grain market, the LOMKT takes value one; othenrvise, its value is zero.

(h) the number of household members that can read. (READ). Grain
traders interviewed during the millet marketing season emphasized that most of
their customers buying maize meal do this because they can not afford to
purchase the more expensive millet. However, some traders indicated that
urbanized people or people with more 'modem' attitudes are the only ones who
actually prefer the taste of maize meal over that of millet. To test whether the
level of urbanization effects maize meal purchases on the household level, the
proxy variable READ has been chosen to reflect farm households' exposure to
modern, i.e., non-rural, attitudes.

The READ variable contains the number of household members that can
read. The knowledge of reading can be gained either through formal school
education or through informal training in later life phases. It is hypothesized that
the number of household members that can read is positively correlated with

maize meal purchases.

246

The results of the regression of maize meal purchases are presented in
Table 5-16. The level of the adjusted R-Square indicates that the model
explains 24 percent of the variability of household level maize meal purchases.
The relatively low level of explanation can be explained by information gaps
about farm level maize meal purchases gained from a one-time interview and the
use of two categorical and two proxy variables. However, with one exception,
the signs of all regression results are consistent to expectation, and most
explanatory factors included in the model demonstrate a significant relation to
the maize meal amount purchased. The control case underlying the
interpretation of the regression results is a farm household that has no access to
a local grain market (LOMKT = 0).

The most unexpected result of the regression model is the large and
positive correlation between maize meal prices and maize meal purchases. The
increase of the maize meal price by one Namibian dollar per kg seems to
increase maize meal purchases by almost 30 kg per month and household
(significant at 1 percent). This result is contrary to economic theory but can be

explained by the monopolistic pricing behavior of grain traders in rural areas.

247

 

 

 

Table 5-17. Estimates of the maize meal purchase

equafion

Ovambo 6 Kavango

VARIABLES COEF. S.E. SIGNIF.
CONSTANT -21.01 11.94 0.08
MIL92 -0.40 0.28 0.16
PMIL 0.74 1.79 0.67
PMAZ 29.85 10.68 0.01
CASH 0.28 0.26 0.28
CASH2 0001 0.005 0.83
MILAV — 3.65 2.60 0.16
WEEA 2.79 0.85 0.00
DISMKT - 0.85 0.24 0.00
LOMKT - 6.42 2.68 0.02
READ 0.73 0.29 0.01
Adjusted R- 0.24
Case Number 223

 

Data Source: Namibian Millet Subsector Project Surveys, 1992193

When farmers’ millet reserves from their own production empty, they
usually turn to their neighbors to buy millet. However, after poor rainfall seasons
in one locality, most farmers face millet shortage, and informal millet markets
collapse soon after harvest. In such cases, the majority of farmers that can not
afford the high travel cost to distant grain markets have no other choice than to
buy maize meal at rural shops. There, rural traders tent to exploit their quasi-

monopolistic position by adjusting their prices upward.

248

Three other regression results confirm the above stated arguments. First,
maize meal purchases are reduced by 6.4 kg if farmers have access to local
grain markets. Second, maize meal purchases are reduced by 3.6 kg if farmers
stated that millet is generally available for purchase. Third, with each subzone
further to the east, i.e., the less infrastructure is available to support local millet
trade, 2.7 kg more maize meal is bought per household and month. Although
the second result is only significant at 16 percent, all three results point in the
same direction: maize meal purchases are increased when millet is not available
or difficult to access.

If the distance to larger grain markets with cheaper maize meal offerings
is increased by ten kilometers, maize meal purchases decline at about 0.8 kg per
month.

The number of household members that can read was used as a proxy
variable for households’ urbanization level. The variable demonstrates a small
but significant effect on maize meal consumption. Wrth each additional literate
per household, 0.72 kg more maize meal are purchased per month.

Both households’ millet harvest results of the 1991/92 production season
and millet prices affect maize meal purchase relatively little. The increase of the
1991/92 millet harvest by one ton reduces maize meal purchases only by 0.4 kg.
The increase of the millet price by one dollar per kg increases maize meal
purchases only by 0.7 kg. While the statistical significance is weak for the first

result, the second ls definitely insignificant.

249
Although the coefficient estimates of the CASH and CASH2 variables
have the predicted significance, the effect Of considerable cash increases are

very small and statistically insignificant.

5.4. Chapter summary
This section summarizes findings about rural households grain

consumption pattern and their limitations to purchase millet In the commercial

 

food market.

5.4.1- Grain consumption pattern

Millet is the most preferred grain of the rural people in Ovambo and
Kavango. Given a choice between pure millet and pure maize, 77 percent of the
interviewed households in Ovambo and 94 percent in Kavango stated they
prefer millet over maize. With more options to chose from only 50 percent of the
respondents from Ovambo and 70 percent in Kavango favored pure millet over
other staple grains. A mix of millet and maize was chosen by 25 percent of the
respondents from Ovambo and pure maize was chosen by 20 percent of the
respondents from Kavango.

The portion of adults preferring one of the more 'modem' grains, such as
rice, macaroni or wheat in the form of bread, was significantly higher in Ovambo
(17 percent) than in Kavango (10 percent). In both zones, bread is the number
one choice for rural children. Survey respondents explained that young children

get used to bread because it is often given to them for convenience.

250

'Tradition', 'availability' and the 'nutritious value’ were the most often
Characteristics that led consumer preference either towards millet or maize.
'Good taste' played only a minor role in farmers' decision to choose millet or
maize as their most preferred grain. Taste was only an important decision
criteria for respondents who favored either bread, rice, or macaroni over millet
and maize.

The investigation into the actual composition of rural households' diet
revealed that farmers' grain consumption is mostly supply driven; i.e., as long as
millet from their own production is available, it is the preferred food staple.
Households begin to purchase and consume maize meal only when millet
reserves are empty and millet is not available in the local market. Differences of
households’ diet before and after grain harvest confirm this finding. Two months
before the 1992/93 millet harvest, most farm households included maize in their
diet (Ovambo: 71 percent, Kavango: 92 percent). Two months after the millet
harvest only 37 percent of the households in Ovambo and 31 percent in

Kavango included maize in their diet.

5.4.2. Limitations to millet purchases

In 1992/93 rural household spent roughly 50 percent of their cash
expenditures on food in Ovambo and 70 percent in Kavango. Household
respondents from both zones stated that 'lack of money', 'expensive
transportation’, and 'high millet prices’ are the main limiting factors for purchasing

millet. Ninety percent of the respondents from Kavango mentioned 'better

251

access to maize meal' and ’cheaper maize meal prices' as an additional reasons
for not purchasing millet.

In Ovambo 57 percent and in Kavango 91 percent of the respondents
claimed that maize is more accessible than millet. These result are in line with
survey findings indicating that only 45 percent of the respondents could buy
millet around planting time while maize was accessible to most of the survey
participant.

In OVambo, most farmers who run out of millet reserves from own
production purchase millet (57 percent) or stretch their remaining millet with
maize meal (43 percent). In contrast, Kavango farmers respond to depleting
millet reserves by eating less millet (70 percent) and switching to other food
sources (60 percent). Only 19 percent of Kavango households would eventually
buy millet to replenish their food stocks. The fact that millet deficit households in
Kavango are less likely purchasing millet than households from Ovambo seems
to be in contradiction to the fact that a larger share of the survey respondents in
Kavango expressed preference for millet than those in Ovambo, and, that
Kavango farmers produce comparatively more marketable millet surplus than
Ovambo farmers. The results from the OLS cross-section regressions indicated
that factors such as ‘the relative price of millet’ and ‘the accessibility of millet on
the food market’ affect households millet purchasing decision more than (a) the
their food staple preference or the millet production capacity of their area.

The regression results reveal also that the existence of local grain

markets leads to a significant reduction of household level maize meal

252
purchases. i.e., Once local markets are available, farmers can more easily
exchange locally produced millet or millet supplied by traveling traders and thus
reduce the need to purchase expensive maize meal at local shops.
Finally, the finding that increased literacy at the household level is
positively correlated with maize meal purchases confirms the argument that

increased exposure to 'modem' lifestyles or attitudes will lead consumers to an

 

increased purchase of maize meal. It is therefore to fear that with continued

ll:

supply shortages of millet and increased exposure to urban life styles consumers
become used to the consumption of maize meal so that millet's advantage of
being the most preferred grain staple declines and millet surplus farmers start

loosing their market.

6. SIMULATION OF MILLET PRODUCTION
AND MARKET IMPROVEMENTS

The purpose of this chapter is to determine whether millet subsector
improvements can contribute to the fulfillment of policymakers' expectations; i.e.,
can agricultural policies and programs, aiming to increase farmers’ millet yields
and to reduce farmers’ millet marketing costs, lead to (a) millet that is price
competitive with maize imports, (b) increased household food security, and
(c) reduced grain imports?

To answer these questions, we shall analyze the prospects of the millet
subsector, assuming that millet yield levels could be increased and marketing
cost of millet could be reduced within the coming years. The assumptions about
how much millet yields will increase in the future are based on findings from on-
farrn yield trials conducted during the production season 1992/93. The
assumptions about how much millet marketing costs of farmers could be reduced
in the future are based on survey findings about marketing costs of farmers that
are already well connected to commercial grain markets. To estimate what
impact the assumed millet subsector improvements might have three
spreadsheet models that are based on data from the Millet Subsector Research
Project, 1992I93 were developed.

The first model calculates relative supply prices of millet for various millet
market Channels and processing technologies under current conditions and

under conditions that simulate potential yield increases and cost reductions for

253

254

millet marketing. The second model estimates the percent of rural households in
the study zones that are food insecure‘ under current millet subsector conditions
and under the assumption that millet production and marketing conditions are
improved. The third spreadsheet model estimates current import requirements
of grain in the study zones and the effect that millet yield increases and
marketing cost reductions will have on future grain imports.

Before the study focuses on each model individually the procedures and
the limits of estimating potential millet production and marketing improvements

are presented in the following section.

6.1. Simulation of potential millet sector improvements

In accordance with the first and second objectives of this study, the
preceding chapters analyzed both the current pattern of millet production,
marketing, and consumption and the main determinants of household behavior

at these sector stages. Important findings of these chapters are as follows:

(1) No other cereal is as well adapted to the climate and soil conditions of the
study zone as millet. However, most study zone farmers do not reach millet self-
sufficiency. Only very few farmers produce marketable millet surplus in Ovambo,
while a fifth of the farmers in Kavango sell millet regularly on the commercial

grain market. Because Kavango has usually better rainfall and less scarcity of

 

‘ For this study, farm households are defined as “Food Insecure
Households” (Fle) if they neither produce enough grain to cover their annual
household grain requirements nor earn enough cash income to purchase
enough grain to balance their grain production deficit.

255

arable land than Ovambo its potential to produce a marketable millet surplus in

the future is also higher than Ovambo.

(2) The marketing system for locally produced millet is insufficiently
developed. Local millet markets are missing and the costs to transport millet to
the central markets are especially high in Kavango where the population density

is lower and distances between communities are further than in Ovambo.

(3) Most consumers in the rural areas of the study zones prefer millet over
imported maize. However, millet is often not available in local markets so that
households that are net grain buyers have Often to purchase maize that is
imported from South Africa. The continued exposure to maize show already
effects on consumers preferences for food staples. In Ovambo where millet
shortages are frequent and maize is more often consumed than in Kavango,
roughly 20 percent of the survey respondents claimed they are indifferent
between millet and maize.

Based on these finding it seems logical to propose that the Namibian
government should foster its efforts to increase farmers’ millet yields and
improve the local millet markets in order to make marketable millet surplus
available to the many net grain buyers in the study zones and to reduce the
trend towards the consumption of imported grains. However, before designing
policies and programmes to improve Namibia’s millet subsector decision makers
should know how much millet yield can potentially increase and by how much
can current millet marketing cost decline in the future. Only when these

questions are answered an analysis can be made to determine the effects of

256
these improvements on important issues such as: competitiveness of local millet
with imported maize, household food security, and requirements to import grain.

It is Obvious that each of the addressed issues is worth a comprehensive
study by itself. HoWever, it seems very unlikely that within the near future
Namibian policymakers will have access to sufficient research capacity to
address these questions in a comprehensive manner. Since policymakers need
at least some information about expected gains from millet sector investments,
the attempt is made to predict on the basis of the available cross-section data
what can, and what can not, be expected from millet sector improvements.

One major limitation of using cross-section data from the production
season 1992/93 is that they are not suitable to estimate price elasticities. To
circumvent this problem, simple cost accounting will be used to determine the
prices at which millet could be offered under current and improved sector
scenarios. The underlying assumption of this approach is that market
competition is strong enough to keep prices close to actual costs, i.e., to keep
profit margins of farmers, processors, and traders low.

It is also necessary to emphasize that up to now, neither improved millet
production technologies have been developed for northern Namibia, nor has any
policy decision been made regarding the structure or organization of the
communal millet market in the future. Of course, without specific information
about the type of improvements to expect, it is even more difficult to make

assumptions about the magnitude of change to expect from such improvements.

257
The following two sections discuss briefly how the study deals with these
dilemmas.
6.1.1. Simulation of incrIeases in millet yields

The task of developing, testing, and disseminating an improved
production technology for millet lies in the hands of both Namibian and
international crop researchers,‘ as well as the extension service. The 1992/93
farm survey revealed that locally produced millet becomes competitive with
imported maize only if the introduction of improved production technologies
increases millet yields and reduces the average unit cost of millet production.
Only under such. conditions is it possible that millet out-competes imported
maize.

Chapter three demonstrates that the widely publicized millet variety
Okashana 1 has not significantly increased farm yields. The results from the
1992/93 on-fann trials reveal that relative large yield gains are possible due to
agronomic field management changes compared to small grain yield gains
realized by cultivar change. (Matanyaire, 1994, p. 9). The household survey
reveals that an improved technology will include several components, such as:

. improved seed, cheaper draft power, better field preparation equipment, better

 

‘ Since 1988, the international agricultural research institute ICRISAT has
collaborated with government and non-govemment institutions in the testing and
development of suitable millet cultivars for northern Namibia.

258
field management practices, chemical fertilizer, improved storage protection,
mechanical threshing and processing equipment.

Since the components for an improved technology package have not yet
been confirmed by researchers and extension workers, it is difficult to predict
whether or to what extent hectare yields might increase in the future and how
much unit costs might decline within the coming years. However, one can
assume that average hectare yields of communal farmers will be significantly
below the yield levels of Namibian crop research stations like “Mahanene” in
Ovambo and “Mashari” in Kavango. It is well known in neighboring countries
that communal farmer are unlikely to have similar soil fertility conditions as
maintained under controlled research conditions. Communal farmers also lack
irrigation and other farmer support services.

The yield increase eventually used to simulate the application of an
improved technology package in the three spreadsheet models are based on the
regression findings from this study and research results from on-farm trails. The
regression analysis of millet yields indicated that the use of manure or chemical
fertilizer increases millet yields by 157 kilogram per hectare on average.
Matanyaire, ICRISAT advisor to the Namibian government reported from his on-
farm trails:

.. a quarter of the pearl millet farmers already using fertilizer
(mainly manure) may, if their management is improved and the
fertilizer used in line with the up-coming guidelines, result in their
pearl millet grain yields being increased by at least 66 %. Yield
gains of up to 200 % are also possible” (Matanyaire, 1994, p. 9).

259

Based on these results the yield increases used in the simulation models
are 66 percent for farmers that had above average millet yields during the
production year 1992/93. Farmers with yield levels below the 1992/93 yield
levels are assumed to rise between 10 to 30 percent to the 1993 yield average of
their zone. Weighing the fact that due to late rains in the production year
1992/93 millet yields were fairly depressed in the production areas east of central
Ovambo and that most farmers had yields below the zone average the simulated
millet yield increase was determined at 20 percent for Ovambo and 30 percent

for Kavango.

6.1.2. Simulation of improved marketing conditions

Another difficulty stems from the attempt to estimate potential decreases
in millet marketing costs resulting from better market organization in rural areas.
Chapter 4 concluded that an increase in rural markets will reduce transportation
costs. In addition, the improvement or construction of feeder roads is likely to
reduce farmers' marketing costs, first, because traveling traders might more
easily and, therefore, more often reach into the rural periphery, and second,
because more private entrepreneurs might invest into vehicles to provide
transportation service to farmers. The increase of price competition among
transporters might also lead to significant transportation cost reductions.

Eventually, more markets and better roads, combined with market and
price information via public radio, might allow farmers and traders to better locate

centers of high millet demand or high millet surplus production. Although

260
traditional communication systems prove to be very efficient on the community or
even district level, they often fail over long distances. A reliable price information
network that does not discriminate between small and large-scale traders and
even informs consumers about price discrepancies between production areas
increases transparency and stimulates efficiency and faster price adjustments in
rural markets.

Despite the above mentioned options to improve millet marketing in
communal areas of Namibia, there are few indications of which measures the
MAWRD will eventually choose to ease the marketing of locally produced millet.
At present, it seems unlikely that the government will agree to the original NBA
proposal that includes the announcement of producer prices every year and the
mandate that marketable millet surpluses have to be delivered to large-scale
processing units in the major towns.

Nevertheless, to simulate potential improvements in the local grain
marketing structure, grain marketing costs Obtained from survey farmers (mainly
transportation costs) will be reduced by 55 percent for Ovambo and by
45 percent for Kavango. The resulting transportation costs come close to those
observed in the peri-urban areas of northern Namibia, where the market
infrastructure is generally better.

The reductions seem feasible because marketing costs for locally
produced grain are relatively high compared to production costs, and because
changes in the current market system, like the introduction of better grain market

information and organization, seems neither difficult to introduce nor extremely

261
costly. In some cases, even improvements in the physical market infrastructure
could be achieved with relatively low cost solutions. Zone differences in
marketing cost reductions are based on Ovambo's natural marketing advantage.
It has a higher population density and a better road net that has led to the

establishment of more rural markets than in Kavango.

6.1.3. Other factors influencing the millet sector performance

Simulating changes only within the millet subsector might lead to the
wrong perception that issues like grain price competitions, household food
security, and national grain imports are only influenced by factors that fall in the
realm of the Namibian Ministry of Agriculture. To avoid this perception, each of
the developed models is exposed to some exogenous changes as well. For the
“Millet Procurement Model”, presented in the next section, the possibility of a
significant reduction in maize import prices will be considered. This is meant to
test the robustness of millet procurement cost reduction in the long run and to
simulate potential reductions of maize prices on the world market. The
simulations of the “Household Grain Balance Model” include a scenario testing
the effects Of increased household income from off-farm employment without
simulating any improvements in the millet subsector. Finally, the simulations of
millet sector improvements in the “Grain Import Substitution Model” are carried
out twice. First, using the population estimates of the year 1995 to calculate

food grain requirements, and second, using population estimates of the year

262
2005. The second simulation demonstrates how population growth influences

grain imports.

6.2. Millet procurement model

This section tries to determine whether improvements of the millet
production, marketing, and processing can enable millet to become competitive
with maize. To investigate this question, a simple cost accounting model is
established that calculates procurement costs for millet meal on the village and
urban center level and under different cost scenarios for millet production,
marketing, and processing in the study zones. To test whether millet will be
competitive with maize in the long run, millet procurement costs will be compared
with maize meal prices that are 20 percent lower than current prices. From the
results of these price comparisons, conclusions are drawn regarding (a) the
effectiveness of production and marketing improvements on the price
competitiveness of millet, and (b) the appropriate choice of millet processing
technology.

The production and transportation costs data used in the model are from
the Namibian Millet Subsector Household Surveys, 1992/93. The processing
cost calculations are from a 1992 feasibility study of the Namibian Agronomic

Board (NAB) and Likwama Farmers' Cooperative from Caprivi.

263
6.2.1. Model description

The base or control Scenario 1 of the Procurement Model starts out by
adding today’s main cost components of millet meal procurement to rural and
urban consumers: production, marketing, and processing cost. Production costs
are mainly based on labor input valued at current agricultural wage rates,
expenses for seed, and received services, like plowing or weeding. Marketing
costs are mainly transportation costs to the pressing location and from there to
consumers, either in urban centers or rural communities. Some grain handling
costs are included in the marketing costs. Processing costs include the actual
processing cost of the simulated processing technology and extraction losses
calculated from the cost accrued so far.

The calculations will be done separately for the study zones due to the
differences in natural grain production potential and differences of distances
between the locations of production, processing, and point of retailing
(Figure 6-1).

Millet meal procurement costs will be separately calculated for the four
millet processing technologies currently available in Namibia: (1) large-scale,
(2) middle-scale, (3) small-scale full mechanical, and (4) small-scale semi-
mechanical. Each type of processing technology assumed influences the
location of processing. Large-scale processing units are more likely established
close to urban centers, while middle-scale technology is more suitable to rural
sub-centers. Small-scale processing units are assumed to operate only on the

community level. Because the location of processing and its distance to

264

83:2. :35
use 1.... .- 880 .5.—350
.2.. .3... 8...... a... 80....
12.. 0.0:. .o 5.3.388

.

as... s8
953.0... 8%..- ..oao: no...
use... 5... I 8.5.8..

.

28:8 :3... c.
2.382.. 0.8-490..

>f

 

 

 

 

 

82.8... =3... 83...... :3!
o:- IS. .0 886 2.2.8.6 use 1.... an 9.8 2.8.2....
.2.. .8... 3...... 2.. .8.... .2.. .8... 3...... s... .8....
.8... 81... .0 5.4.8an». .3... 6.9.. .0 i.e.-Eco
196. .08 82... ~80
n.a.-fies. 3%... .6383... use-fie... 6.9... season:
use: is! 3.88» 2...... s... I 8.938
.26. 3...... so Sax-.668 .5 c.
8.388.. 1.34.8.8 3.382.. 0.8-.3...»
_
0025.5.
.7
82...... :3... 83...... :3... 83.3... :3!
o... .8.: .a 8.8 Ease-.6 a... .2... .- 880 E02956 o..- .E... .- 386 2.8.93»
.2.. .00... 8...... o..- 803 .2: .3... 2...... o...- Suta .6... .3... «0...... a... 80.!
.ee... 81... .6 sign It: on]: .6 53.09.80 1.... 31:. .c 5.5.03.3
:26. .86 12s. «.8 sci. .08
953...... 8%... c3892.. Past-E 6%..- coaoaoEa 952...... 3%... 5.39.3...
363551.482...» 65:95:13.2...» 6595.61.46.53
88:8 :35 .26. .03.! .8 83.3568 .3.. c.
3.888.. :80. 9.3.86... 1.3.12.3 2.882.. 0.81.2.»
V a s
§>o
L»
.86 £889.. 2%

9.85... .8886...

.82.. .8586... 3.? 2.. .6 34.326 .3 2.5...

265
consumers affects marketing costs, millet meal procurement costs have to be
calculated individually for rural and urban consumer markets.

Once the procurement costs for millet meal are calculated for each
procurement/processing channel, they are compared with average maize meal
prices observed in 1993. The comparison is based on the nominal costs of millet
meal procurement and nominal maize meal prices, assuming inflated prices of
both will rise together in the future. Eventually, those procurement lprocessing
scenarios are identified where millet meal procurement costs are lower than
1993 maize meal prices.

To test whether solutions where millet meal can be more cheaply offered
than maize meal are robust enough in the long run, it is assumed that maize
meal prices observed in 1993 could decline up to 20 percent over the next years.
This assumption is based on potential price reductions due to increased
competition in the Namibian milling industry and possible reductions of grain
import prices due to improved production conditions in South Africa.

The following paragraphs describe how various procurement cost
components are derived. Table 6-1 summarizes the simulations of costs under
current and improved millet production and marketing conditions and various

processing technologies.

Millet production costs:
The three main components of the millet meal procurement cost

calculation are production, marketing, and processing costs. Because it can be

266
assumed that surplus producers sell millet commercially, production cost are
derived only from sample farmers that produced a marketable millet surplus
during the 1992/93 production year.

The main inputs used by communal farmers for millet production are land,
labor, draft power, manure, and seed. The use of communal land was not
valued as a cost factor. Manure was also valued at zero cost because it has no
market value and it comes from producers' own livestock. Labor input is
measured in workdays for field preparation, weeding, fertilization, bird chasing,
harvesting, and threshing. The calculation of workday equivalents was already
described in the production Chapter 3. One workday was valued at the 1992/93
average wage rate of N5 5 for agricultural laborers. lf farmers used only the hoe
and/or their own span of draft animals for field preparation, only the time they
spent in the fields was valued as cost. If households hired draft power, cash
payments made for this service were included in the cost calculation.

It is assumed that roughly 2.5 kg of millet seed are necessary to plant one
hectare of millet. One kilogram of millet seed is valued at NS 1.10, the average
price for millet in 1992/93 on informal millet markets.

Based on these production cost and average yields of millet during the
year 1992/93 production costs per ton of millet are estimated at N$ 670 for
Ovambo and N5 620 for Kavango. Since on-farm trials demonstrated that for
farmers that are already using fertilizer the introduction of improved seed and
production technology could lead to a yield increase of 66 percent it is simulated

that farmers that are currently producing a marketable surplus of millet have a

267

 

 

 

 

6.00m 02.2.0.9. :2...sz 6230— .8035 8.00095 .232 :2...sz ”00.390: 0.00
I o o I o o 2.5.6.50 .05. 55.2. -
$09 8. on 0.00. 2 Eu 22:00 :2... ES. -
0.08:0..00 o. coautoamcufi
0.000 002.000 .0 2.00.00 m. 0....-qu
$3 - 80 08. $8 - 80 8... .8228... :2 20820.... -
$8 - 8.. 8... $8 - 80 8.. .mozmzooe.....0m 0.82.05... -
.. 03 o: .. o: 03 325508.... 208.202... -
.. 8m 08 .. 8m 8m 5.5.85.5. 2800...... -
9200002..
«.000 002.000 .0 200.00 cm mammo— eczouhxm
I o o l o o .26. £c:EE00.0.000.=0Em -
.8.... - 8 m0. 2.... - 8 0... 0.2.80 mo..o~..=m.20om.2n...... -
$8 - cm. can .80 - no. 0 .u .250 082208-022 -
0300009... 0. :o..0to..0..0.#
$3 .. a... 8. $8- an. 2. 0.000 530.60...
8 #2.»... #232 .8 #232 #232
80. 80.
00:0 0:03.230 .00» 00:0 2.0.2.200 .00»
00.5. 00.5.08. 0.00.. .08... 00.6.9... 000.. 0:023:00 .Oaucw
omega. 2.80.5 0:0N

 

:00... 3...... 60.500... 050... 3% «z. 0..o~ .3 £536.30

30.05.00. 50.6.9... 0.... 2.0.5.0 .00.... 8000 2.050.509... .00... «2:2

.9.» 030....

268

yield increase of 66 percent. It is estimated that this yield increase reduces millet
production costs by about 40 percent to N$ 420 per ton in Ovambo and to N$

370 per ton in Kavango.

Millet marketing costs:

The analysis of transportation cost data acquired during the 1992/93 farm
household surveys indicates that farmers who sell their grain outside of their own
community are facing marketing costs of N$ 200 per ton in Ovambo and NS 300
per ton in Kavango. However, most farmers selling their millet within their
community have no explicit transportation or other marketing costs.

Those farmers that marketed millet at urban centers like Oshakati in
Ovambo or Rundu in Kavango faced, on average, 40 percent higher marketing
costs than those who traded millet at nearer sub-centers of their zone.

The average transportation costs identified for the production year
1992/93 are used directly for the base Scenario 1 of the accounting model. To
simulate improved marketing conditions in terms of reduced marketing costs, it is
assumed that current transportation costs can be reduced by 55 percent in
Ovambo and 45 percent in Kavango. As explained in the introduction of this
chapter, such high reductions seem to be feasible because the current situation
suffers under a poor market infrastructure and low competitive pressure on

transporters' price demands.

269

Processing costs:

The Namibian Agronomic Board estimated the cost of millet processing
with large scale processing equipment at N3 333 per ton. The Likwama farmers'
cooperative calculates the millet processing costs of its middle scale processing
operation at NS 440 per ton. The costs for small-scale processing of millet with a
hammerrnill and with a dehuller (full mechanical) or without a dehuller (semi-
mechanical) at the community level have to be inferred from observed service
fees of private operators. The costs for semi-mechanical processing are
estimated at N$ 500 per ton, assuming zero opportunity cost for farmers’ manual
dehulling. For full-mechanical processing the cost to process one ton of millet is
calculated at N$ 1000.

For the scenarios of improved marketing conditions, it is estimated that
increased competition among small scale processors will lead to the decline of
fees for milling and dehulling by 20 percent. This leads to estimated fees for
small-scale processing at the semi-mechanical level of N5 400 and at the full-
mechanical level of NS 800 under improved marketing conditions.

A study about mechanical millet processing in northern Namibia reported
that the extraction losses of used dehullers are close to 30 percent, but that this
could be reduced to about 20 percent if the extracted bran is recycled through
the dehulling process for a second time. (Dendy, 1993) This leads to the

simulation of extraction losses of 20 percent for both large and middle scale

270

millet processing. For small scale processing operations it is assumed that the

bran is returned to the customers who can use it for beer brewing or animal feed.

Example of procurement cost calculation:

Having described how individual components of the Procurement Model
are determined, Figure 6-2 demonstrates an example of how these costs are
combined for a particular zone and a particular scenario. The actual calculations
under different scenarios are done by a standard PC-spreadsheet program and
the results are obtained in table form. The main advantages of using this type of
software are its flexibility in adjusting calculations to specific cases and its
capacity to repeat calculations of various scenarios without high programming

and calculation costs.

271

s 853... 81...

 

gasses... gas... 3...... £2.25.
assist. 328.35.... assigns 358.2550
5.52.23 28...... as 32...... as .25.... as
280.82.. .0... 205.82.. .8... 28238. .8... 52.282. .8....
2.5.... .2.... 2.5.2.22. 2.5.22.5. 2.22.0.2...
2:258:53. 535......02 Eases; 52.2.5.5
3...... 2.23... 3.5.2 2.3.8.. .05....
350%.... 3.2.2.. 3.05.53.12.20 "2232.322:
8.... 5.2. 3.888... 3. :52 ”5882..
#:5528322...— 0.5—.2835.

v

# 2.0..00m . 0280.5 .8 2.2.0.3200 ~30 2.080.500... «0...:

:58... D2.... sari... 3....
38.2580 .o 50.5%..
8.. 52. 88 8m. 52. .38
28.232. .20.. 22532.1...
.8 .22. 50.0... 8" 5.2. 520...
€3.58 22.2.... 3.22%.“:
2. 5......

2.882.. 28. .5....

SN .33.. .2088...

.3... 2 505.2...

.~.0 0.59..

272
6.2.2. Model results

The results from the simulation models are presented for four different
sector scenarios: the base Scenario 1, which is based on current production and
marketing costs; scenario 2, that ceteris paribus is based on reduced production
costs; scenario 3, that ceteris paribus is based on reduced marketing costs; and
scenario 4, that ceteris paribus is based on reduced production and marketing
costs. For each scenario mentioned, the cost of six alternative procurement
channels is estimated separately for Ovambo and Kavango. Through these
procurement channels millet meal reaches the following:

(1) urban markets via large-scale processing operations located at urban
centers;

(2) urban markets via middle-scale processing operations located at subzone
centers;

(3) rural markets via large-scale processing operations located at urban
centers;

(4) rural markets via middle-scale processing operations located at subzone
centers;

(5) rural markets via small-scale semi-mechanical processing operations
located in communities;

(6) rural markets via small-scale full-mechanical processing operations
located at communities.

Table 6-2 presents the model results in two forms: first, as absolute
values of millet meal procurement cost per ton, and second, as percent deviation
from 1993 maize meal prices. Solutions that display negative percent values can
be considered as price-competitive against maize meal, assuming that both
millet and maize meal prices undergo the same rate of inflation. The most cost

effective solution within a scenario is printed bold if the millet procurement costs

273

 

 

5005...... . 00.0_ . .0500 0:00

 

 

 

 

 

s 00. 8: W...

.x. 00. 003 mm 5005...... . 0.00.5 . 0:350
I I WW 5005.500 . .050. .002
I I .5.. 5005.5: .050 . .000.

5...: .0 00.0 +. 0 2.0:00m

.x. 0.. 5005...... . 00.0. . .0500 0:0».

.\. 0 T 5005...... 0.0.25 . 05.300
I I 5005.500 . .050 . _000_
I I 5005...... =05... 08.

35.02.02 +. n 2.0:00w
$8- 003 5005...... 00.0. . .0500 0:00
*5. 003. 5005...... . 0.00.5 . 0:05.50
I I 5005.500 . .050 . .000.
I I 5005...... . .050 . .000.
52.00020; N 2.5000
x. m - on: 5005-..... . 00.0. . .0500 053
.x. n - on: 5005...... 2225 0:350
I I 5005.500 . __050 . .002
I I 2,.» . . . . . M 5005.5: ._050 . .0w0.
.000“. p 25:30
00:0. .005 30.05.03 0500000...
.05... 00.05. 00.0.2 0:0..0 :2.000.. .2558
0.02.02 c0050 . 0.02.0.2 .05.“... 0.02.02 :02... " 0.02.0.2 .0..m,m.

002<><¥ OmE<>O

 

.025 .005 00.05 000. .0 500.00 .:0. .00 02. 2.5000 .:05020>00 0:0 $0205.00.
05000020 52.002 0:00 .3 ._005 00.05 .0 002.0 0:0 0.000 50505020 .005 .055 .0 50.50500 .Né 0.00...

 

274

are 20 percent less than the 1993 maize meal prices, i.e., the solution would still

be competitive if maize meal prices decline by 20 percent in the long-term.

6.2.2.1 . Current competitiveness (Base Scenario 1)

Scenario 1 estimates millet meal procurement costs under 1993's
production, marketing, and processing conditions. The model estimates that
packaged millet meal could be offered at prices ranging between NS 1120 and
N$ 1737 per ton, depending on the study zone and the type of processing
technology used. The cheapest procurement options are (a) for rural consumers
in both zones, semi-mechanically processed millet meal (small-scale
hammennills) with N$ 1170 per ton in Ovambo and NS 1120 per ton in Kavango,
and (b) for urban consumers of both zones, middle-scale processing operations
(Ovambo: N3 1370, Kavango NS 1382).

The comparison of millet meal procurement costs with prices of 25 kg
bags of maize meal in local markets in 1992/93 reveals that under 1993's
conditions, millet meal could be offered to rural consumers at lower prices than
those of maize meal if semi-mechanical processing technology is used.

For Ovambo, estimated millet meal prices for semi-mechanically
processed millet are 19 percent below maize meal prices in rural communities.
However, while in Ovambo none of the remaining procurement options can out-

compete maize meal prices, some estimates come very close. Large and

275

middle-scale processing for urban consumers and middle-scale processing for

rural consumers come within a 5 percent difference of maize meal prices.

In rural Kavango, estimated prices for millet meal from semi-mechanical
processing with small-scale hammerrnills are 30 percent below maize meal
prices in rural communities. This large price advantage guarantees millet meal a
long-term competitive position versus maize.

The main conclusion is that if rural consumers continue to prefer millet
over maize and buy the cheapest grain available, the introduction of small-scale
hammerrnills will be economic viable in rural Ovambo and Kavango.

Small-scale milling services in rural communities will give rural farmers the
option to purchase semi-mechanically processed millet meal at prices that are
lower than current maize meal prices. Additionally, rural employment and
income opportunities could be created through the use of small-scale processing
technology. It is also possible that local grain markets, if not yet existent, will

emerge close to small-scale hammerrnill operations.

6.2.2.2. After sector improvements (Scenarlos 2 - 4)
Scenario 2 simulates the impact of the introduction of an improved millet

production technology which reduces production costs in Ovambo by 37 percent

276

and 40 percent in Kavango. The reduction in production costs reduces the cost
of millet meal by roughly 20 percent for both urban and rural consumers.
However, under the restriction that millet meal prices have to remain competitive
even if maize meal prices decline in the long run by 20 percent, the outcome is
more diverse. In Ovambo, only semi-mechanical small-scale processing in rural
communities reaches prices 36 percent lower than maize meal prices, and are
therefore, far beyond the required 20 percent. Middle-scale processing at the
subzone level still achieves cost advantages over maize meal that come very
close to the 20 percent threshold (rural markets: -18 percent, urban markets: -
19 percent).

In urban areas of Kavango, both middle and large-scale processing units
would reach long-term price competitiveness, while in rural Kavango, only
middle-scale millet processing and semi-mechanical small-scale processing
would provide millet meal at costs that are competitive in the long run. As in
Ovambo, cost advantages of semi-mechanical processed millet over maize meal
are very high.

The three most important findings from Scenario 2 are as follows: First,
local semi-mechanical millet processing with small-scale hammerrnills will
increase millet meal's price advantage in rural areas significantly once an
improved production technology is identified and implemented. Second, once
farmers' production costs are reduced, middle-scale processing operations at

subzone level have the potential to serve both rural and urban grain consumers

277

with price-competitive millet meal. Third, procurement of millet meal on the basis
of full mechanical small-scale processing with a small-scale dehuller and
hammerrnill is too costly for local markets.

In Scenario 3, millet transportation costs to processing facilities and
consumers are reduced between 50 and 66 percent to reflect potential
improvements in the millet marketing system of Ovambo and Kavango,
respectively. Since reducing transportation costs benefits millet procurement
options that are transportation intensive, procurement of millet meal via large-
scale processing in urban centers with retailing in rural communities was most
affected (Ovambo: 17 percent; Kavango: 22 percent).

The reduction of millet transportation costs makes 5 of the 6 millet meal
procurement channels in Ovambo competitive against maize meal prices and all
6 channels in Kavango. However, the introduction of the condition of a 20
percent cost advantage over maize meal leaves, in both study zones, only semi-
mechanical millet processing as a competitive option.

The results from the improved market system simulated in Scenario 3 lead
to similar conclusions as previous scenarios. The competitive position of local
semi-mechanical millet processing with small-scale hammerrnills can hardly be
challenged by other alternatives. Middle and large-scale processing operations
might not yield competitive millet meal prices if only marketing condition are

improved.

278

Scenario 4 simulates the joint impact of reducing millet production and
marketing costs. The result reveal that procurement costs are between 26 to 39
percent less than under cost conditions in 1993. The estimated cost for one ton
of millet meal from middle and large-scale processing units range between
N$ 957 and N$ 1057 in both zones. The cost for millet meal processed semi-
mechanically by small-scale hammermills leveled at around N$ 800 per ton of
meal. The cost for fully-mechanical small-scale processed millet at local market
level is about N$ 1200 per ton.

Except for full mechanical small-scale milling on the community level, all
other procurement options provide millet meal at costs that are 20 percent less of
what rural and urban consumers have to pay for maize meal today. Millet meal
that is dehulled manually by producers and milled by small-scale hammermills
can be competitive in both zones if maize meal prices decline by even 40
percent. Under the conditions of combined improvement in millet production and
marketing, large-scale processing in urban centers becomes slightly more cost
effective than middle-scale processing at subzone level.

The outcome of Scenario 4 eventually confirms findings from all previous
scenarios. The dissemination of small-scale hammermills in rural areas is the
most competitive solution for millet meal procurement in rural areas. In the case
that aggregate millet production levels increase to such an extent that also urban
consumer markets can be catered, both middle and large-scale processing

operations seem to be suitable.

279

6.3. Household grain balance model

This section investigates how household food security in the study zones
will be affected by the introduction of an improved production technology and
improved grain marketing. A Household Grain Balance Model is used to relate
millet production and marketing parameters with household food security.

Current definitions of “household food security” emphasize both farm
households' food production capacity and their capability to purchase food to
meet household needs. “Food Insecure Households” (Fle) are defined as
those who neither produce enough grain nor acquire enough income to meet
household grain requirements.

To be able to identify food insecure sample households the potential grain
balance was calculated in the base or control Scenario 1. The calculation
includes the following three figures: (a) households' grain consumption needs,
(b) households' subsistence production, and (c) potential grain purchases from
off-farm cash income.

Households' current food security levels are expressed in percent of
households' grain consumption needs covered through staple grain production
and potential staple grain purchases. It is assumed that the acquired average
percent figures from the survey sample reflect general food security conditions in

Ovambo and Kavango.1

 

‘ Due to lack of historical data and to simplify modeling, it is necessary to
assume that the Fle identified during the 1992/93 surveys are a fair

280

Three further scenarios simulate improvements (1) in millet production
technology, (2) in northern Namibia's grain marketing structure, and (3) in
improvements 1 and 2 together. The improvements are simulated by using
increased grain yields, reduced unit production costs, and reduced unit
marketing costs.

These simulations can help inform agricultural policymakers about the
potential reduction in household food insecurity through programs geared
towards improving grain production and marketing conditions in the principle
millet production zones. To provide policymakers with a comparison between
the reduction of Fle through grain sector improvements and potential
improvements in non-agricultural sectors Scenario 5 shows how improved
access to off-farm employment can increase rural incomes and reduce food ‘
insecurity.

The rest of this section is divided in two parts. The first part describes the
Household Grain Balance Model. Part two presents and discusses the model

results.

 

representation of grain deficit households in recent years in the study zones.
This assumption is supported by aggregate grain production estimates from the
Namibian Early Warning & Food lnforrnation System Unit. As Table 3-5 of
Chapter 3 already demonstrated, the harvest results of the 1992/93 production
year are not significantly different from other recent production years.

281
6.3.1 . Model description

The base or control Scenario 1 identifies the food insecure farmers of the
320 households surveyed in the 1992/93. These farmers did not produce enough
grain to cover the grain needs for the following year or earn insufficient cash
income to meet their household food security needs.

Figure 6-3 presents the information used to calculate the percentage of
FIHs in Ovambo. For each sample household, the difference between annual
grain need and grain availability is estimated. The calculation of households'
annual grain needs is based on FAO recommendations regarding energy
requirements for different gender and age groups. The amount of grain available
per household is estimated by adding to households' 1992I93 grain production
the amount of grain possibly bought with households' annual cash income.

The grain prices used to calculate potential purchases were the cheapest
grain prices mentioned by individual survey households during the April 1993
household surveys (maize meal was the cheapest calorie source during most at
that time).

The amount of cash income available for the purchases of staple grain is
based on the finding that, on average, 50 percent of Ovambo households'
available cash income and 60 percent of Kavango households‘ available cash
income is used to buy food (see Tables 5-8 and 5-9 above) and 80 percent of

the money spent on food is used to purchase grain.

282

Qin§¢08 Q‘NQuICOOO

332.58
5.838;

 

o ates-o- _
| I I
s Sign]
a EEm Roan-o 83
l I | 8a..— 950:,
_|l4I _ T
2:85
:30

r

0..

is: 5.: .0
85.2. 520

8-22 «8.2.. .333... 58.92:... 53.52 32:8 83

38.8.:- 23353 .6 .0 2339.30 .5
8 05238 one-£0 8303:. 2. $332.. IcoEofi .

v d N 83:00.

a
£82.. .318 can:
.8... .333

r _ _
_

cause.
8558.... flea-.82...
_ 128...:

to... £80

OOS<>O z. acaozmmzoz
mcaomaz. 000“. no kzmucma

.onEE'O com .258 3:22. Ema Bonanza: 2:

.0.-

 

12.33..
is: s. .o
85.2 see

.né 2:9...

283

After the base conditions are established and the percent figures of Fle
in Ovambo and Kavango are estimated, the following assumptions are made for

Scenarios 2, 3, 4 and 5:

(a) Household millet production can be increased by the implementation of

an improved technology package (Scenario 2);

(b) More and cheaper surplus millet is available to buy after the communal

marketing system has been improved (Scenario 3);

(c) Production and market improvements estimated in Scenarios 2 and 3 can

be achieved at the same time (Scenario 3);

(d) Improvements in non-agricultural sectors that lead to increased off-farm
employment opportunities for rural households can be simulated by
adding to each household an average annual income from non-

agricultural employment.

284

Scenario 2 assumes that improved millet production technologies can be
identified that increase the average millet yield in Ovambo 20 percent and in
Kavango 30 percent. The average increase was calculated on the assumption,
that (a) yields of farmers that produce below zone average can increase their
yields up to this zone average, and (b) farmers that are currently operating at
levels higher than zone average can increase their yields by two thirds, the
potential yield increase derived from on-fann trails conducted by agronomists
during the production year 1992/93.

In Scenario 3, improvements of the communal grain marketing system'are
assumed to lead to lower marketing costs and cheaper millet prices for
consumers. Transportation costs constitute the lion's share of total grain
marketing costs. Under current market conditions, transportation of millet from
producers to markets amounts, on average, to 30 and 50 percent of the original
production cost, i.e., 200 Rand per ton in Ovambo and 300 Rand in Kavango.

We assume improvements in the grain market structure will reduce the
marketing costs of farmers by 45 percent in Kavango and 55 percent in Ovambo.
Under the improved market conditions, a higher percentage of grain deficient
farmers are now enabled to purchase millet locally that is cheaper than maize
meal. The millet prices used to simulate the new conditions are taken from the
Millet Procurement Model developed in the previous section. In cases where

maize meal remains cheaper than millet, maize prices will be used.

285

Scenario 4 simulates both improvements simultaneously, production is
increased due to the adoption of improved production technologies and grain
purchases are increased because of improved grain marketingconditions.

Scenario 5 assumes that improvements in the manufacturing and service
sectors will generate more non-farm employment opportunities to the rural
population of Ovambo and Kavango. Under this assumption, the cash income of
each household is increased by N5 6000 per year.

This figure represents the current average annual salary of unskilled men
and women in occupations such as laborer, sales person, food processor,
security personnel, assistant clerk, and assistant nurse, that were reported in the

household surveys in 1992/93.

6.3.2. Model results

As described, the Household Grain Balance Model generates results for
five different scenarios: the base Scenario 1, representing today's conditions,
Scenarios 2, 3, and 4, simulating potential improvements in millet production
and/or marketing, and Scenario 5, simulating increased cash income from more
employment opportunities in non-agricultural sectors. The results of these
scenarios are summarized in Table 6-3. Figures 6-4 through 6-13 present the

simulation results graphically.

286

6.3.2.1. Current food security conditions (Base Scenario 1)

In Table 6-3 the results of the base Scenario 1 show that the share of

food insecure households is 38 percent (230,000 people) in Ovambo and 43

percent (56,000 people) in Kavango.

Table 6-3. Model Results: Percent of food insecure farm households,

by zone and scenario

 

 

Zones Ovambo Kavango
percent of ' 000 of percent of ' 000 of
rural HHS people rural HHS people

Scenario 1: current production, 38 % 229 43 % 56

current marketing, current income

Scenario 2: improved production,

current marketing, current income 29 % 175 32 % 41

Scenario 3: current production, 35 % 211 42 % 54

improved marketing, cunent income

Scenario 4: improved production, 23 % 139 27 % 35

improved marketing, current income

Scenario 5: current production, 3 % 18 0 % 0

current marketing, improved income

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93

287

Figure 6-4. Scenario ‘llOvambo: Food insecure households under 1992I93
production and marketing costs

 

ace ( .—-—

i
g , ‘—— Feed insecure ———> ! ‘— F00“ 'fQWV’fl—l’i»

a“.
.

l , '

. w»

l

 

 

Feed
Beauty

 

 

 

0% 20% MO“ .0”

Data Source: Namibian Millet Subsector Research Project and my own calculations

Figure 6-5. Scenario 1IKavango: Food insecure households under
1992193 production and marketing costs

 

 

 

Food
Security

8e"-
Sufficiency

E
5
g , ‘— Feed lneecure $ ! {Food secure-k!

 

 

 

Data Source: Namibian Millet Subsector Research Project and my own calculations

288

Figure 6-6. Scenario ZIOvambo: Food insecure households under

reduced production and 1992I93 marketing costs

 

. <

 

Food insecure ——'>

Foo'd’eecure

 

newsman-mm

 

Data Source: Namibian Millet Subsector Research Project and my own calculations

 

l
H l”

 

 

Feed
Securly

Self-
Sufficiency

Figure 6-7. Scenario 2IKavango: Food insecure households under

reduced production and 1992/93 marketing costs

 

A

. ‘— Food insecure ——-> ! ‘— Fe

 

,1

od epicure f-—>

 

mama-drum

 

Data Source: Namibian Millet Subsector Research Project and my own calculations

1“
e'.‘
.I'
r”
a

 

 

Feed
Security

Self-
Sufficiency

289

Figure 6-8. Scenario 3IOvambo: Food insecure households under
1992I93 production and reduced marketing costs

 

I,
I

‘ ‘—— Food Insecure ———‘> ‘— Food/gem}: ——>
#1

 

thmm

 

 

 

 

Data Source: Namibian Millet Subsector Research Project and my own calculations

Figure 6-9. Scenario 3IKavango: Food insecure households under .
1992I93 production and reduced marketing costs

 

 

‘ ‘————— Food Insecure ‘> I‘ Feed secure)“.

 

 

 

 

Data Source: Namibian Millet Subsector Research Project and my own calculations

290

Figure 6-10. Scenario ”Ovambo: Food insecure households under

reduced production and marketing costs

 

A

 

‘— Feed Insecure ——<>I‘——— Foed secure

 

mammm
3

 

 

‘-I 4

0%

 

Data Source: Namibian Millet Subsector Research Project and my own calculations

 

Figure 6-11. Scenario ”Kavango: Food insecure households under

reduced production and marketing costs

 

‘—Foed Insecure *— Focdsicure

 

WIMMM

 

    

Data Source: Namibian Millet Subsector Research Project and my own calculations

 

291
6.3.2.2. Effects of sector improvements (Scenarios 2 - 4)

The results of Scenario 2 and 3 show that improvements in millet
production or marketing can reduce the number of FIHs. However, these
reductions cannot eradicate all food security problems. Estimations from
Scenario 4 indicate that even with simultaneous improvements in millet
production and marketing, still a quarter of all rural households remain food
insecure.

The results of these scenarios are important for Namibian decision
makers because they show that policies designed to improve the food grain
sector in northern Namibia could have a considerable impact on the incidence of
FIHs. However, Scenarios 2 to 4 also show that improvements in the agricultural
sector alone cannot overcome severe rural poverty that is the product of policies
of the colonial government. The results show that political empowerment and
economic development of Namibia's rural majority, has to be pursued in
agriculture and non agricultural sector because grain production accounts for
less than 40 percent of total food security requirements (Ovambo: 38 percent

and 32 percent in Kavango).

6.3.2.3. Effect of non-farm employment (Scenario 5)
A surprisingly large reduction in the number of Fle stems from an
increase in cash income in Scenario 5. (Figure 6-12 and 6-13) The cash income

of each sample household is increased by the average wage of an

292

Figure 6-12. Scenario 5IOvambo: Food insecure households under
increased income from off-farm employment

 

Food insecure

zoo: '
/
o-r’p'i.
- ' 5

! ‘ [fl Food secure
40% so

Data Source: Namibian Millet Subsector Research Project and my own calculations

 

 

Feed
Security

 

 

Self-
Sufficiency

 

 

%

 

mm

Figure 6-13. Scenario 5IKavango: Food insecure households under
increased income from off-farm employment

 

\

 

 

. ‘ ,‘f Food secure <>

 

thmm
I

Feed
Security

 

 

Self-
Sufficiency

0%

 

Data Source: Namibian Millet Subsector Research Project and my own calculations

293
unskilled laborer (N$ 6000 per year that allows at 1992/93 prices the purchase of
the annual millet consumption requirement of an average household) to simulate
increased employment opportunities for farm households in the study zones. ln
Ovambo, the number of Fle decreases from 38 to 3 percent, and in Kavango,
from today's 43 percent to a level of zero.

The conditions of Scenario 5 reduce the average share of grain
production on total grain access of Fle to 6 or less percent and for food secure
farm households to 17 percent in Ovambo and 7 percent in Kavango. This
means that grain production loses its importance with regard to household food
security once more off-farm employment becomes available to the rural
population of Ovambo and Kavango. The results of Scenario 5 confirm the
conclusion from the preceding scenarios. Policies targeting solely improvement
in the grain staple sector are not enough to deal with the problem of food
insecurity in northern Namibia.

Although increased off-farm incomes affect farmers' food security more
than improvements in the millet subsector, the final conclusion is not that policies
trying to achieve the latter are void. On the contrary, the following five reasons
explain why it is necessary that improvements in the millet subsector be pursued
with regard to household food security:

(1) Although additional off-farm employment opportunities are promising,

Namibia is facing the prospect of reduced off-farm employment

opportunities due to the withdrawal of South African troops and the
decline in its mining industries.

(2)

(3)

(4)

(5)

294

Ovambo, and especially Kavango, are the two zones in Namibia with the
highest‘potential for rainfed crop production. Enough labor is available in
agriculture once the profitability of grain production is improved, i.e., the
development of the agricultural sector to make best use of available
physical and human resources seems to be of paramount importance for
the economic development of Namibia.

Scenario 4 demonstrated that a better performance of the millet subsector
can actually reduce the number of Fle significantly. Fle that can not
benefit from sector improvements have to be helped with other measures.

The MAWRD, as well as various national and international NGOs, are
already developing measures to enhance millet production and improve
local and regional grain marketing.

Improvements in the millet subsector can be seen as first steps toward
engaging a larger portion of communal farmers in commercial food
production without simultaneously increasing their food insecurity. At a
time when technologies for the production of higher value crops are
available, farmers might already have adopted knowledge how to
integrate cash crop production into their farming system.

6.4. Grain import substitution model

The purpose of the grain substitution model is to estimate the grain import

needs of the study zones under various scenarios. A base or control model is

developed which estimates the aggregate grain import needs of Namibia's main

millet production zones Ovambo and Kavango in 1995. Scenarios 2, 3 and 4

estimate grain import needs under the simulation of increased millet production

on the household level and/or reduced millet marketing cost. The rates of yield

increases and marketing cost reductions are similar to those used in the

preceding Millet Procurement Model and Household Grain Balance Model (see

sections 6.1.1. and 6.1.2. above).

295
6.4.1 . Model description

The basic equation used to derive the import need from the available data

is as follows:
Grain Grain Subsistence Purchases
import = consumption - grain - of local
requirement need production grain

The aggregate grain consumption need of 1995 is determined by
aggregating sample households' annual food grain requirements and
extrapolating those to the zone level. As done for the household grain balance
model, the calculations of grain requirements on the household level are based
on FAO recommendations regarding energy requirements for different gender
and age groups.

The urban population from Ovambo (10 percent) and Kavango (5 percent)
is included in the extrapolation process to reflect total grain need of the study
zones. To forecast the aggregate staple grain need of rural Ovambo and
Kavango for the year 2005, the following average population growth rates from
the years between 1983 and 1993 are used: Ovambo 3.1 %, Kavango 2.5 °/o
(Central Statistics Office, 1993).

The aggregate amount of subsistence grain production is calculated by
adding up all households' grain consumption needs that are covered by
households' own grain production. Aggregate purchases of locally produced

millet are calculated by adding up all millet purchases estimated for the sample

296

households individually. If the aggregate amount of individual millet purchases
surpasses the aggregate amount of marketable millet surplus production, the
latter figure is used.

To estimate whether a household purchased millet and how much, the
cheapest grain type available to individual households is determined first;
i.e., household representatives have stated the current price of available millet
and imported maize meal. For a sample case where millet is offered cheaper
than maize meal, it is assumed the household is spending the average portion of
its cash income available for grain purchases on millet (Ovambo: 45 percent,
Kavango 55 percent). If maize meal is cheaper than the available millet offered,
it is assumed the household balances its grain deficit with maize purchases.

Wrth each scenario, the price of available millet changes due to the
simulation of various combinations of millet sector improvements. If millet prices
become cheaper, more of the farm households that are buying maize meal to
balance their grain deficit switch over to commercially offered millet.

The basic assumptions and constraints applied throughout the simulations
are as follows:

(1) Production, purchase, and price data collected from the randomly
selected household sample are a fair representation of the overall
conditions in the study zones.

(2) Although the population of the study zones increases over time, current
grain production does not increase proportionally due to more land under
cultivation. This is due to the fact that most of the arable land is already
occupied, and since independence, the opportunity exists to move to
other regions in search of off-farm employment.

297

(3) Farm households in the study zones consume their subsistence grain
production first before they purchase food grain on the informal or
commercial grain market.

(4) Households that purchase grain purchase the cheapest available; i.e., if
improved production and/or marketing conditions lead to the availability of
cheaper millet than maize, consumers change their purchasing behavior
in favor of millet.

(5) The total amount of purchased grain from local production is not allowed
to be more than the amount of local grain surplus production.

(6) Under today's grain market structure (Scenarios 1 and 2), locally
produced millet surplus is not traded from one zone to the other. Surplus
millet that is not sold within its own zone is counted as loss.1

(7) Under the conditions of improved millet marketing (Scenarios 3 and 4),
inter-zone exchange of millet surplus becomes possible. Grain surplus
producers that do not sell their surpluses so far are willing to do so under
improved marketing conditions, i.e., when market transaction costs are
reduced and commercial production is more common.

The aggregate amount of grain losses for each zone is calculated as the
difference between total grain production and the sum of aggregate subsistence
grain production and aggregate purchases of locally produced grain.

In the case that all maize imports are replaced by local grain production,

the aggregate amount of exportable surplus is calculated by subtracting the

 

‘ Results from the regression analysis determining farm households'
selling behavior demonstrated that farmers closer to central markets are more
likely to sell millet. Up to the present, it is very common that farmers in remote
areas who produce larger amounts of surplus grain keep these in storage without
making any effort to sell. Although this grain degenerates over time and thus
loses its value for human consumption, its status value is still high. Similar to the
number of cattle owned, the number of full granaries, whether filled with old or
fresh grain, is a traditional indicator of wealth.

298
aggregate need of grain consumption from the total amount of local grain

production.

6.4.2. Model results

This section reports model estimates of study zones’ grain import
requirements in 1995, demonstrates the effects that millet subsector
improvements have on grain imports, and discusses the effects of population
growth in the study zones long-term grain import needs. The results of this

section are summarized in Table 6-4 and Figures 6-14 through 6-17.

6.4.2.1. Current import requirements (Base Scenario 1)

Under current millet production and marketing conditions, grain import
requirements for the major millet production zones are substantial. According to
the estimates of Scenario 1, up to 92,000 tons of grain have to be imported
annually to secure adequate staple grain supply. This is 57 percent of the
162,000 tons grain consumption requirement for Ovambo and Kavango in 1995
(Table 6-4 and Figure 6-14).

Two important features of the current production system are the following:
First, commercial grain production contributes only 12 percent to balance total
grain consumption need. As demonstrated in previous study parts, the main
explanations for this phenomenon are that most study zone farmers produce a

grain deficit and can therefore not sell surplus grain. Additionally, those who

299
produce marketable grain surpluses are usually not used to marketing grain and

discouraged by high market transaction cost.

Table 64. Model Results: study zones grain import needs, grain
production, and production losses (in '000 MT, °/o of
aggregate consumption need in parentheses)*

 

Scenario 1 Scenario 2 Scenario 3 Scenario 4

 

base year 1995 increased improved Incr. production
production marketing lmpr. marketing
Grain . MT 92 47 68 0
imports % (57) (29) (42) (0)
Subsistence MT 51 76 51 76
production % (31) (47) (31) (47)
Commercial MT 19 39 43 86
production 7. ( 12) (24) (27) (53)
Exportable MT 0 0 0 10
surplus % (0) (0) (0) (6)
Production MT 24 57 0 0
losses '1. (15) (35) (0) (0)

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93 and my own calculations

‘ based on the estimate of 162,000 tons of grain consumption requirements in the study
zones in 1995

Second, aggregate grain losses are comparatively high. They amount to
15 percent of current grain consumption requirements. Even worse is the fact
that these grain losses amount to 34 percent of Ovambo and Kavango‘s
aggregate grain production (subsistence and commercial). The production
losses are not so much explained by poor grain storage practices. But,

traditional perceptions permit wealthier farmers to display large grain reserves in

300
the storage baskets as a status symbol without making an effort to engage in
any marketing activities that could transfer millet surpluses to areas with
production deficits. Also farmers in remote who want to sell their marketable
millet surplus often face often such high marketing cost that prefer to keep it in
storage where only small amounts are sold or given away to poor neighbors,

while the rest loses its consumption value over time.

Figure 6-14. Potential grain import substitution based on population
estimates for the year 1995 of Ovambo and Kavango

 

 

     

SCENARIO 1 SCENARIO 2 SCENARIO 3 SCENARIO 4
0 current e improved e current 0 improved
production production production production
0 current 0 current 0 improved ' “PM!“
marketing marketing marketing marketing
250 ’1
" 200 |
2 ‘1 Study Zones’ 10“ Grain
8 Consumption Need of IT 162
P 150 —l l I m
c: mpo
'ii , I IMO“! ss,ooo MT
'- r |92,000 MT
(9 100 *l
0 i
a I
3 ' d
o 50 —l
0 j i '

 

 

 

: Marketable Surplus Production I Non-marketable Surplus Production
3 Coarse Grain Imports I Commercial Production
[1 Subsistence Production

Data Sources: Namibian Millet Subsector Research Project, 1992/93 and my own calculations

301
Individual food safety concerns are the third factor contributing to high
losses of marketable grain surplus. To reduce the risk of food shortages due to
consecutive drought years, farmers' storage/marketing thresholds tend to be
high. In the case that several good harvest years follow each other, grain
reserves are not used for a long time. Eventually, they have to be replaced with
grain from more recent harvests. By then, the old grain reserves have lost their

consumption and market value due to rot and insect infestation.

6.4.2.2. Effects of sector improvements (Scenarios 2 - 4)

The simulation of production and/or marketing improvements significantly
affects aggregate import needs in the study zones. The sole increase of hectare
yields on the household level simulated in Scenario 2 reduces the 1995 grain
import requirements by almost 50 percent, to 47,000 tons. The two effects
causing these large import reductions are increased subsistence production and
increased millet purchases on the household level. The latter is caused by the
reduced production. costs that are passed on to the consumers via reduced millet
prices on the informal and commercial grain market.

Although in Scenario 2 aggregate grain production is increased by 78,000
tons, the reduction of grain imports amounts only to 45,000 tons. The reason for
this disproportional effect on grain import is the insufficient development of the
current millet marketing system reflected in the scenario conditions. The poor
market development prohibits inter-zone grain exchange of local surplus

production and thus increases current grain losses to the level of 57,000 tons.

302

The sole improvement of millet marketing conditions simulated in
Scenario 3 leads to a reduction of grain import needs by 24,000 tons, or
26 percent of current import needs. This import reduction is achieved by
converting the surplus production that is currently not marketable into marketable
surplus. Although the import reductions achieved in Scenario 3 are much less
than those achieved in Scenario 2, the results are still impressive. The pure
marketing solution is appealing because it makes better use of already eXisting
resources rather than emphasizing the production of more grain, of which a
significant amount will be lost.

The results of the best world Scenario 4, which combines production
increases with marketing cost reductions, are encouraging. If such
improvements could be applied to cover today's grain consumption requirements
of 162,000 tons, even exportable surplus of 10,000 tons could be produced.

However, it seems very unlikely that the simulated sector improvements
will be achieved in the near future. It is also doubtful that all maize imports can
be crowded out by a more efficient communal grain sector. The two reasons
supporting the latter argument are the foreseeable change in consumers' food
preferences toward maize meal consumption, and increased grain consumption
need due to continuing population growth. The preference change towards
“modern food staples” among younger population sections was already
documented during the preceding consumption Chapter 5. To get a better

understanding of how population growth might counterbalance potential import

303
reductions, the following section presents simulation estimates that take

population growth into account.

6.4.2.3. Projections for the year 2005 (Scenarios 1 - 4)

Introducing population growth into the modeling of study zones' grain
import requirements alters the perspective from which millet subsector
improvements can be judged. So far, the simulation results of Scenario 4
implied that grain imports could be totally substituted with local grain production if
subsector improvements could be achieved instantly. However, experiences in
other countries have demonstrated that it takes between 10 to 15 years to
identify and disseminate improved and sustainable millet production technologies
while developing simultaneously an effective grain marketing system. Therefore,
it is necessary to estimate grain import requirements based on population figures
forecasted for the middle and long run.

Figure 6-15 presents import requirements based on the assumption that
the simulated millet subsector improvements could be achieved by the year
2005. In this example, the aggregate demand for grain staples is adjusted to
217,000 tons, assuming that recent trends in population growth continue in the
study zones.

Under these propositions, aggregate grain demand can not be covered by
local millet production in the future. Even if the “best world" Scenario 4 applies,
roughly 20 percent of the estimated grain requirements, or 47,000 tons of grain,

have to be imported in 10 years' time. Valuing these import requirements on the

304
basis of the 1993 import parity price of white maize, foreign currency in the value

of US $12.3 million are necessary to balance study zones' grain requirements.‘

Figure 6-15. Potential grain import substitution based on population
estimates for the year 2005 of Ovambo and Kavango

 

 

 

 

     

 

SCENARIO 1 SCENARIO 2 SCENARIO 3 SCENARIO 4
0 current 0 improved 0 current ' improved
production production production production
0 current 0 current 0 improved ' improved
marketing marketing marketing I'll-“OWN
300 j
I
E 250 '7 Study Zones' sees Grain
0 Consumption Need of MT 217
8 200 ~ Imports
- 47,000 MT
,5 Imports Imports
a
5 150 4 Imports 1°2'°°° “T 123.000 MT
0 147,000 MT
3 100 —
o
0 , , j
5° 2 . . , * g- . i. ‘ 3 f
0 P , ' ' A

 

 

 

I Non Marketable Surplus Production D Coarse Grain Imports
I Commercial Production D Subsistence Production

Data Sources: Namibian Millet Subsector Research Project, 1992/93 and my own calculations

* To forecast the aggregate staple grain requirements, the following average population
growth rates from 1983 to 1993 are used: Ovambo 3.1 percent, Kavango 2.5 percent.

 

‘ On the basis of weighted grain import figures from the USA and South
Africa the MAWRD calculated a parity price for white maize of NS 787.24 per
ton. At the exchange rate of 1993, this ton price converts to US-$ 262.

305

The fact that agricultural policies targeting the improvement of the millet
subsector will not be able to fully eliminate Ovambo and Kavango's grain imports
in the future might be used as an argument against such policies. To counter
such arguments, it is necessary to point out that future grain imports will be
significantly higher if no subsector improvements are achieved. Simulation
results of Scenario 1 (no sector improvements) for the year 2005 yield a grain
import need of 147,000 tons. This is three times the amoUnt estimated under
improved subsector conditions in Scenario 4.

Figure 6-16 demonstrates that, in the long run, substantial reductions of
grain imports are possible due to millet sector improvements. The thick solid line
reflects the increase of food grain requirements over time due to continuous
population growth in the study zones. The line of Scenario 1 indicates the
amount of food imports necessary if no grain sector improvements take place.
The three lines of Scenario 2 through 4 reflect the import needs under the
simulated scenarios of millet sector improvements. The line of Scenario 4 shows
the trend of grain import requirements for the study zones once all potential millet
subsector improvements are achieved.

Finally, to estimate the potential middle term effect that millet subsector
improvements in the study zones could have on national grain imports, it is
assumed that millet production increases and improvements in the grain
marketing structure are achievable within the next 10 years. Figure 6-17

demonstrates the possible development path from the 1995 position of

306

Scenario 1 to the improved subsector conditions of Scenario 4 reached by the

year2005.

Figure 6-16. Projections of aggregate grain consumption and import
requirements for the study zones until the year 2020
based on four millet subsector scenarios" (in '000 MT)

  
  

 

 

 

 

350 ’j
300 -i
Food Grain
E 250 '1 Requirements \
o
o
9 200 -
.E ' .—
e ,."
5 150 - ' , - —
a ——————
8 "’0‘ —"§G‘°a.£i) " ‘
O c- " - — .o’” l
’ l ’ ’ ’0', I
50 fl, ’..’..’.0 '
-""..’.. i
0 . " r T l l 1
1995 2000 2005 2010 2015 2020
— - - - — ........ n- sssssssssss
agongoa‘rg IMPORTS IMPORTS IMPORTS IMPORTS
FOOD ORAIN SCENARIO 1 SCENARIO 2 SCENARIO 3 SCENARIO 4
production production production production
e current 0 comm . ”SWIG 0 Improved
marketing marketing marketkig marketing

Data Sources: Namibian Millet Subsector Research Project, 1992/93 and my own calculations

The grain import savings achieved through these subsector developments

are presented in the shaded area between the two lines. Over the next 10 year

until 2005 cumulative import savings are 247,500 tons or US$ 64.8 million (using

the 1993 import parity price for white maize). Assuming that no further

productivity increases will be achieved in the millet subsector after the year 2005

307
but population growth rates persist, the cumulative import savings are estimated

to be 922.500 tons or US$ 241.7 million by the year 2020.

Figure 6-17. Projections of the potential millet subsector development path
and resulting grain import savings for the study zones

   
  

 

 

 

350 -]
|
300 ‘l
i
’5 250-1:
0 .
O
9 200
.5
5 150
§
e 100
O
0
50
! SERPJ‘P’
0 7"." I I I I
1995 2000 2005 2010 2015 2020
— -- ........... —'— I I I I I
AGGREGATE IMPORTS IMPORTS DEVELOPMENT IMPORT
GRAIN NEED SCENARIO 1 SCENARIO 4 PATH SAVINGS
0 current P “PM!“
production production
0 current P "PM“
marketing marketing

Data Sources: Namibian Millet Subsector Research Project, 1992/93 and my own calculations

6.5. Chapter Summary

The chapter summary describes briefly the purpose and structure of the
three simulation models established and presents the results of each individual

model.

308
6.5.1. Purpose of the simulation models

The purpose of the simulation models was to determine whether millet
subsector improvements such as increases of farmr " millet yields and
reductions of farmers’ millet marketing costs lead to an significant improvement
of the millet subsector performance.

The performance of the millet subsector is measured according to the
following parameters: First, the procurement costs of millet meal expressed in
percent of unit prices of maize meal observed at rural and urban grain markets in
Ovambo and Kavango or predicted for the future at the respective markets (Millet
Procurement Model). Second, the number of food insecure households
expressed in percent of all rural farm households in Ovambo and Kavango
(Household Grain Balance Model). Third, the amount of grain import
requirements necessary to balance the aggregate grain deficit in Ovambo and
Kavango, expressed in thousands of tons of staple grain imports (Grain Import
Substitution Model).

The simulation models are based on primary and secondary data from the
Millet Subsector Research Project, 1992/93. The most important millet
subsector parameters used in the simulations are as follows: The average millet
yield of farmers operating below the average in each zone Ovambo and
Kavango can be raised to these averages, while the yields of farmers operating
above current zone averages can be raised by 66 percent. It is assumed that
the production cost of millet surplus farmers can be reduced from N5 670 to N5

420 per ton in Ovambo and from N5 620 to N$ 370 per ton in Kavango. It is

309
estimated that improvements in the communal grain marketing system could
reduce farmers’ transportation/marketing costs of millet by 55 percent‘in Ovambo
and 45 percent in Kavango.
After modeling the mathematical relations between millet subsector
parameters and subsector performance measures for each of the three
simulation models each model is executed under four different scenarios

indicated in Table 6-5.

Table 6-5. Potential improvements of millet subsector parameters and
their application across the model scenarios

 

Scenario 1 Scenario 2 Scenario 3 Scenario 4
Improvements

 

'base run' 'best wortd'
Yield increases8r
production cost no yes no yes
reductions
Marketin cost
reductior?s no no yes yes

 

Data Source: Namibian Millet Subsector Project Surveys, 1992/93 and my own calculations

Each model is supplemented by simulating a parameter change external
to the millet subsector. For the Millet Procurement Model, the possibility of a
significant reduction in maize import prices is considered, demonstrating that
increased price competition from other production zones in the region could
reduce commercialization prospects for the Namibian millet subsector

significantly.

 

310

For the Household Grain Balance Model increased household cash
income indicates that policies promoting off-farm employment opportunities could
affect household food security at least as much as policies in the agricultural
sector.

For the Grain Import Substitution Model, the increase of staple food
demand due to population growth in the study zones is simulated to demonstrate
that expected grain import reductions on the national level based on millet sector

improvements might only be temporary.

6.5.2. Results of the millet procurement model

The millet meal procurement model attempts to determine which
procurement/processing channel is most cost effective and whether
improvements of millet production and marketing conditions make millet

competitive with maize.

WW (Scenario 1)

The estimation millet meal procurement costs under current, non-
improved cost conditions and their comparison with maize meal prices observed
in rural and urban food markets in 1993 demonstrates that if small scale
hammermills (semi-mechanical small scale processing) are introduced today
millet meal could be offered to rural consumers at prices lower than those of

maize meal (Ovambo: 19 percent, Kavango: 30 percent).

31 1
‘e._t:_ - A. -. .1.. . u - . 0.0 010: -.l° . . .- r- (Scenari02l3)
The simulation of reduced cost for millet production and to lesser extent

the simulation of reduced millet marketing cost make millet meal competitive for
most processing options. However, the outcome is more diverse once the
prerequisite for long-term competitiveness is used, i.e., if it is demanded that
millet meal prices have to remain competitive even if current maize meal prices
are reduced by 20 percent. Now, the only alternative that stays competitive in
both study zones is semi-mechanical small-scale processing with hammermills in

rural communities.

.I I0‘ no 0 :1- n ‘ 0 H .010: -_|- “-.I.‘ l' 0:010: (Scenario4)

Scenario 4, simulating reductions in millet production and marketing costs
simultaneously, leads to millet meal procurement costs that are significantly
lower than those under 1993's subsector conditions. The lowest cost achievable
for rural consumers stems from local processing with small-scale hammermills
where millet has to be dehulled manually. This option leads to millet meal prices
that undercut current maize meal prices in rural Ovambo by 43 percent, and in
rural Kavango, by 47 percent.

Except for full-mechanical small-scale milling with a dehuller, all other
processing options provide millet meal to rural and urban markets at costs that

are 20 percent less than 1993 maize meal prices.

312

900mm

Policy initiatives and agricultural programs to reduce millet production,
and marketing costs in Ovambo and Kavango increase millet’s competitive
position. By far the most cost effective policy initiative is to introduce semi-
mechanical millet processing with small-scale hammermills at community level.
This solution could be competitive under current subsector cost conditions.
Additional advantages from the introduction of small-scale processing technology
in rural areas are increased off-farm employment and income opportunities, as
well as incentives to organize local grain markets, if not yet existent, close to the

processing location.‘

6.5.3. Results of the household grain balance model

The Household Grain Balance Model investigates how improvements of
millet production and marketing conditions in Ovambo and Kavango affect
percentage of food insecure households (Fle) in these zones. The model
relates millet production and marketing costs mathematically with household
food security. The model identifies households as food insecure if they neither
produce enough grain to cover their annual grain requirements nor earn enough
cash income to purchase the amount of grain necessary to balance their grain

production deficit.

 

‘ Zimbabwe’s recent policy reforms have stimulated private investments
in a large number of small scale mills in rural and urban areas. Namibian policy
makers may want to examine Zimbabwe’s experience.

 

 

 

 

 

 

fam

31 3
Estimations of the base or control Scenario 1 reveal that 38 percent of the
farm households in Ovambo and 43 percent of those in Kavango are currently

food insecure.

l‘ej e _o ‘01.]... ‘||‘] e. I‘l.ll," a see. ‘ _°|si ‘ges

The results from Scenario 2 (improved production conditions), Scenario 3
(improved marketing conditions), and Scenario 4 (simultaneous improvements of
Scenarios 2 and 3) demonstrate that policies designed to improve the food grain
sector in northern Namibia could have a considerable impact on the number of
FIHs.

Under the conditions of the 'best world' Scenario 4, the percentage of
Fle is reduced to 23 percent in Ovambo and 27 percent in Kavango. However,
translating these figures into absolute numbers, still 139,000 people from
Ovambo and 35,000 from Kavango are food insecure. Scenarios 2 to 4 show

that improvements in the agricultural sector alone can not overcome severe rural

poverty.

i‘ej . '. -._ ..- .fi... -... . ..-. .. .- 1 J..- .

To simulate increased off-farm employment in rural areas, farm
households’ cash earnings are increased by N$ 6000 per year in Scenario 5.
This income increase reduces the percentage of Fle from 38 to 3 percent in
Ovambo and from 43 percent to zero in Kavango. This finding reveals that

political empowerment and economic development of Namibia’s rural majority

E Tl
‘ ..L

 

314
has to be pursued in agriculture and non agriculture because grain production

accounts for less than 40 percent of food security requirements.

6.5.4. Results of the grain import substitution model

The grain import substitution model estimates potential short and middle
term reductions in grain import requirements for Ovambo and Kavango under
improved millet production and marketing conditions. Study zones’ staple grain
requirements are estimated on the basis of FAO data about daily food
requirements of individuals and population estimates from the Central Statistics

Office in V\findhoek.

Ell l' |.. I' I

It is estimated that under the millet subsector conditions of the year 1993,
Ovambo and Kavango have an annual grain import requirement of 92,000 tons.

This amount represents 57 percent of the zones' annual food grain requirements.

Holding aggregate food grain demand at the level of 1993 while simulating
improvements in millet production and marketing conditions reduces grain import
needs dramatically. The simulation of increased production in Scenario 2
reduces aggregate import requirements from 92,000 to 47,000 tons of grain.
Reduced marketing costs simulated in Scenario 3 reduce grain import

requirements to 68,000 tons. Simulating cost reductions in millet production and

 

31 5
marketing simultaneously leads to 100 percent grain self-sufficiency for the study

zones and a marketable millet surplus of 10,000 tons for export.

EII'II _| [Ellillll I' |

Unfortunately, the encouraging results from above hold only if the annual
population growth rates of 3.1 percent in Ovambo and 2.5 percent in Kavango
would drop instantly to zero. Since this seems unrealistic, the simulations of
Scenario 2 to 4 are repeated, assuming that current population growth rates
persist in the middle and Iong-terrn and that it would last until the year 2005 to
implement all the millet subsector improvements discussed above.

Now the simulation results reveal a long-term dependency on food grain
imports for the study zones. If by the year 2005, the millet subsector could reach
the 'best world' Scenario 4, grain import needs would decline from the 1993 level
of 92,000 to 47,000 tons. Although dependent on grain imports, it is estimated
that if millet subsector improvements are implemented by 2005 a total of
2.5 million tons of grain imports could be saved during the next period of 25

years until 2020.

Conclusions

The population growth, even if it slows down over the next decades, will
probably over-compensate for improvements in the millet subsector.
Nevertheless, grain import savings due to millet subsector improvements will still
be very high in the long run and therefore worth being considered and

undertaken as soon as possible.

7. SUMMARY AND CONCLUSIONS
7.1. Background

The continuation of Namibia’s grain deficit after independence in 1991
pressured agricultural policy makers for action. Since more than half of
Namibia's 1.5 million people were living in the main millet producing zones of
Ovambo and Kavango in northern Namibia, the government decided to give
priority to the development of the millet subsector. At the same time, ICRISAT
announced that its pearl millet variety ‘ICTP 8203', which was distributed in
Namibia for the 1988/89 cropping season under the name Okashana 1, would
enable farmers to double their yield from about 300 to more than 600 kglha with
local agronomic practices. The head of the Namibian crop research also
suggested, that with improved cultivation practices, smallholder could produce
2.4 tons per hectare in the future (ICRISAT, 1991).

Under the impression that an increase in millet yields could soon eliminate
the grain deficit, the Ministry of Agriculture, Water, and Rural Development
(MAWRD) shifted its attention from production issues to policies to
commercialize the millet subsector. It was also assumed that by enabling
communal farmers to produce a surplus of millet that could be sold to
commercial traders, household food security would improve in the study zones,
farmers' cash incomes would increase, and the amount of foreign currency
needed for grain imports would decline significantly.

Policymakers in Windhoek also assumed that a commercial grain market

was not functioning in the millet production zones. Therefore, the Namibian

316

31 7
Agronomic Board (NAB) proposed to introduce guaranteed producer prices for
millet and create a one-channel millet marketing scheme in collaboration with the
Namibian Development Corporation (NDC). After careful review of all available
information the MAWRD eventually decided that more information about the
millet production and marketing system was necessary to formulate an effective
grain policy for the communal areas in northern Namibia including an
assessment of the long term competitiveness of millet in the national food
economy.

In August 1992, the MAWRD initiated the ‘Millet Subsector Research
Project, 1992/93’ in collaboration with lntemational Crop Research Institute for
Semi-Arid Tropics (ICRISAT) and with financial support from the German
Ministry of Economic collaboration BMZ. The BMZ financed 18 months of
fieldwork and 8 months of data analysis by a member of the Agricultural
Economics Department of the Michigan State University, USA. The project goals
were (a) to collect basic information about the current structure and conduct of
the millet subsector, (b) to evaluate the prospects for the commercialization of
the millet production, and (c) to recommend alternative ways to assist millet
producers in Ovambo and Kavango. (hereafter called study zone)

Between August 1992 and January 1994 four surveys were conducted in
the study zones on the household and market level. The data were processed
and preliminary research findings were presented at the first national millet
subsector workshop in Windhoek. Survey data were subsequently analyzed and

a baseline research report was made available at a second national workshop in

318

November 1994. A final report on millet processing alternatives was completed

in February 1995.

7.2. Objectives and data sources
The objective of this dissertation research was to analyze household and
market level data collected by the 'Millet Subsector Research Project' in order to

address five research questions:

(1) What are the current patterns of millet production, marketing, and
consumption in Namibia’s most populated regions Ovambo and Kavango?

(2) Which factors determine household production, marketing, and
consumption behavior?

(3) What yield increases and reductions in millet marketing cost can be
expected for the future based on certain hypothetical scenarios?

(4) How would these hypothetical improvements in the millet subsector affect
(a) the number of food insecure households, (b) the price
competitiveness of locally produced and processed millet against
imported maize, and (c) the need to import maize into the study zones ?

(5) What are recommendations for future grain marketing and food policy
research?

Most of the data for this study came from four surveys that were carried
out in Ovambo and Kavango between August 1992 and January 1993. One
household survey and one field measurement survey were conducted covering
the same sample of 320 farm households living in one of 16 survey communities
of the study zones. The household survey included three visits to each sample

household during different phases of the 1992/93 millet production season: field

31 9
preparation, field management, and grain harvest and threshing. Data were
gathered on household characteristics, cash income, agricultural production
patterns, millet production marketing practices, as well as food consumption and
purchase behavior. After harvest, the field measurement survey measured the
size of fields on which sample farmers produced millet during the 1992/93
season.

A millet market survey was conducted after the millet harvest in November
1993. This survey targeted commercial grain traders in rural and urban areas of
the study zones. The survey investigated whether a commercial millet market
exists, and if so, to determine its strengths and weaknesses.

From February 1993 to the end of January 1994, a food price monitoring
survey was conducted at 15 communities and at four urban grain markets in the
study zones. Prices of millet, maize and other food staples were collected every
two weeks at 59 retail outlets. At the beginning of the millet harvest in July 1993,
millet producer prices were included in the price monitoring survey.

Secondary data were acquired on rainfall patterns in the study zones,
grain production estimates since independence, crop yields on research stations
and in on-farm trials, and the costs of processing millet with various processing
technologies.

The data from the Millet Subsector Research Project cover the 1992/93
production year. The research findings of this dissertation are based on cross-
section analysis because prior to Namibia's independence, time series data were

not collected on grain production in communal areas.

320
7.3. Conduct and performance of the millet subsector
This section presents the five most important areas of research findings
concerning the conduct and performance of northern Namibian millet subsector
participants with respect to millet production, marketing and consumption. The
findings are obtained from the descriptive and regression analyses of the survey

data from the 1992/93 production season.

7.3.1. Current and new technologies

Survey farmers' average millet yields from the 1992/93 season are with
200 kilogram per hectare in Ovambo and 250 kilogram per hectare in Kavango
close to the average of the only available yield estimates conducted by the
MAWRD between 1991 and 1995. The average number of hectares cultivated
with grain was 2.9 hectare in Ovambo and 2.5 in Kavango. Aggregate estimates
of the MAWRD of the annual millet area indicate that in 1992/93 Kavango
farmers cultivated the usual amount of land while farmers from Ovambo
cultivated significantly less. The latter can be explained with the late arrival of
rains in 1992/93 compared to farmers’ experience from the 80s and weakened
draft animals in Ovambo, due to the severe drought in the production year
1 991/92.

Twenty percent of all farm households in Ovambo use manual labor for all
field preparation activities. Less than 1 percent of farmers owned tractors and
less than 5 percent used chemical fertilizer. Cattle manure was used in Ovambo

where livestock can be kept close to fenced millet fields. In Kavango shifting

321
cultivation is still a common practice. Fields are unfenced and livestock herds
are kept in distant areas.

Contrary to the claim of crop scientists, the survey farmers using
Okashana 1 neither increased yields significantly in the direction of 600 kilogram
per hectare nor the marketable surplus of millet in comparison to farmers not
using Okashana 1. On-fann trials by ICRISAT and Namibian crop scientists
during the 1992/93 production season revealed that improved management
practices, including the application of chemical fertilizer on traditional varieties,
produced significantly higher yields than the cultivar change to Okashana 1.
From these on-farm trails it was concluded that farmers using manure or
chemical fertilizer and improved agronomic practices could increase their yields
by at least 66 percent. The main advantage of Okashana 1 is its early maturity,
which is a valuable trait when the rainy season starts late and yields are
generally depressed. Survey data revealed that in very short rainy seasons
Okashana 1 yielded with an average of 125 kilogram per hectare about 40
percent more than local millet varieties that average only 90 kilogram per

hectare.

7.3.2. Household income and grain self-sufficiency

The household surveys revealed that the average annual per capita
income (cash plus imputed value of own production) was N5 480 in Ovambo
(US$ 160) and N5 380 in Kavango (US$ 130). The annual per capita income of

Ovambo households that earn off-farm income is N3 910 compared with N5 110

322

' for Ovambo households without off-farm earnings. The average annual income
per capita in Kavango households with off-farm employment is N5 710 versus
NS 100 for households without of farm income. Correlation analysis revealed
that households’ cash income is positively correlated with the use of mechanical
plowing service and negatively correlated with manual field preparation.
However, regression results indicated that cash income has little affect on
households' millet area or yields. The explanation for this is that most of the
households earning cash income do not invest money into their millet production
in order to turn it to a commercial enterprise but make investments to lesson the
burden of manual labor of their own household members.

Although millet is the staple food in the study zones, the surveys reveal it
contributes much less to rural households consumption and cash income than
originally assumed. For the 55 percent of rural households that have no off-farm
income millet contributes an average of 42 percent of farm income. For the 45
percent of farm households that earn income from off-farm employment, millet
contributes on average only 10 percent to their total income. This demonstrates
that activities outside of agriculture such as school education and off-farm
employment have become important.

The surveys reveal that 33 percent of the farm households in Ovambo
and 51 percent in Kavango cultivate less than 2 hectares and derive less than a
third of total household income from agriculture. This indicates that development
programs in the study zones must look beyond the farm toward income and

employment creation in rural non-farm activities.

323
During the 1992/93 production year, 75 percent of Ovambo households
and 68 percent of Kavango households in rural areas met less than 30 percent
of their household grain needs from farming. As a result most rural households
are net grain buyers. Many are too poor to purchase sufficient food to meet their
household food needs. An estimated 38 percent of rural Ovambo households

and 43 percent of rural Kavango households are food insecure.

7.3.3. Marketable surplus of millet

Seventy percent of the survey farmers in Ovambo sell neither millet to
traders nor to neighbors, 30 percent sell millet in years having a good season,
and less than one percent sell millet every year. By contrast, 18 percent of
Kavango farmers sell millet every year and 54 percent sell millet after good
harvest years. Although Kavango’s population is only a forth of the population in
Ovambo, the aggregate marketable surplus is larger in Kavango than in
Ovambo. The 1992/93 household survey does not provide enough data to
estimate the amount of millet that farmers barter or sell informally between each
other. But survey data about the frequency and the usual amounts per millet
sale indicate that the farmers who sell grain sell more often to their neighbors in
small quantities rather than selling to commercial traders. The analysis of the
millet trader survey indicate that from the estimated 70,000 tons of millet
produced in Ovambo during the year 1992/93 only 2,800 tons were sold to
commercial grain traders. In the same year about 4,400 tons were sold on the

commercial market in Kavango out of a total 15,000 tons millet production.

324

Based on FAO recommendations for the minimum energy requirements
per capita, Kavango and Ovambo needed 153,000 tons of grain for human
consumption in 1993. Only 85.000 tons or 55 percent of this requirement could
be covered through local millet production during 1992/93. The rest had to be
met with grain imports, mainly maize from South Africa. This reveals millet
production has to expand significantly if northern Namibia wants to become more
food self-reliant in the future.

The long-term potential to increase the marketable surplus of millet is
higher in Kavango than in Ovambo. Due to higher annual rainfall and the
availability of more fertile land, millet yields are already under current,
unimproved production practices higher in Kavango than in Ovambo. An
important constraint on the expansion of the millet cropping area in Kavango is
that the fertile land along the Kavango river is already under cultivation and most
arable land in other areas is difficult to access because rural roads are non

existent and drinking water is scarce.

7.3.4. Consumer grain preferences and consumer prices

Most rural and urban consumers from the study zones prefer millet over
maize. Seventy seven percent of the surveyed households in Ovambo and 94
percent in Kavango stated they prefer millet over maize. The farmers that do not
produce enough grain to cover their household needs try to purchase
unprocessed millet informally from neighbors, which is the cheapest source of

calories. However, millet is only available in informal markets during part of the

325
year and the millet offered by commercial grain traders is very expensive.
Therefore, many consumers are forced to purchase maize from domestic and
imported sources, mainly South Africa.

Although majority of the surveyed farmers perceived millet superior to
maize and regression analysis revealed that millet purchases increase with
higher cash income the long-term exposure to maize due to lack of millet in the
market is shifting rural consumers preference toward maize. In Ovambo were
the grain production deficit is larger than in Kavango already twenty percent of
the surveyed households claim they got so used to maize that they are
indifferent to the consumption of millet or maize. Also, the number of those who
claimed they like the taste of a millet/maize mix better than that of pure millet
was larger in Ovambo than in Kavango.

None of the survey respondents that stated a preference for maize
explained this with the fact that maize is available in meal form while millet can
only be acquired in unprocessed form. They rather explained their preference
with the continued consumption of maize that made them less used to the taste
of millet. But an increasing number of urban consumers is willing to endure long
waiting periods to have a local miller process their millet. This is an indication
that millet processing will become a more important issue as incomes increase in

rural areas.

326
7.3.5. Constraints on commercial millet marketing

Between 40 and 60 percent of rural farmers in the study zones hire
transportation to sell their millet. Survey date for the 1992/93 production year
indicate that Ovambo farmers that sell millet pay up to 40 percent and Kavango
farmers up to 60 percent of the value of traded millet for transportation. Such
high marketing cost are probably the reason why many farmers sell millet only
within the boundaries of their community.

Also, between 40 and 60 percent of the study zone farmers do not have
access to a local market. During the 1992/93 surveys most of those farmers
expressed the need for a place were millet is regularly traded. For the
households that are in need of grain but cannot acquire millet from neighbors
because of a local millet shortage such markets would allow to make their
demand known to farmers with marketable millet surplus and traders. For
farmers with marketable millet surplus local markets could provide a simple
alternative to the organization of long distance transport of grain to a far distant
market. Results from the cross-section Iogit model regarding farmers’ millet
marketing behavior also indicate that farmers’ proximity to grain markets

increases the likelihood of millet marketing significantly.

7.3.6. Angolan millet imports
Since the severe drought in 1991/92, grain traders from urban Ovambo
started to import millet from Angola. During the 1992/93 production year, Angola

supplied more than a third of Ovambo’s total commercial millet trade but only

327
three percent of the commercial millet trade in Kavango. The Angolan millet
supply between 20 and 140 percent cheaper than the millet offered by Namibian
surplus producers to commercial grain traders. It originates from the province
Huila, Lubango about 500 kilometers north of Ovambo were large-scale
commercial farms produce under more favorable climatic conditions than exist in
northern Namibia. The following three factors might have contributed to the
millet imports from Angola: (1) The war in Angola blocked access to grain
markets in towns so that northern Namibia with its millet deficit production
became a welcome market alternative. (2) Due to the poor economic state of
Angola's economy, its currency was undervalued to such an extent that Angolan
agricultural products became competitive in Namibia. (3) The warfare limited
south Angola's access to imported goods via Angolan harbors. Thus, the access
to consumer goods via trade with northern Namibia became attractive.

Although it is undetermined whether the supply of relatively cheap
Angolan millet will continue in the future the fact that countries other than South
Africa are supplying grain to northern Namibia indicates that the prospects for
Namibia’s millet subsector will increasingly depend on its competitiveness with
grain imports from South Africa and neighboring countries such as Angola,

Zambia, and Zimbabwe.

7.4. Simulation results of policy and technological changes
Three simulation models were used in Chapter 6 to quantify the effects of

the development and dissemination of improved millet production technologies

328

and the improved functioning of the local millet market system on three food
policy issues: (1) household food security in the communal north,
(2) competitiveness of locally produced and processed millet versus maize meal,
and (3) substitution of national maize imports and saving of foreign exchange.

To be able to simulate policy and technological change, the potential
increase in farmers’ millet yields and the potential reduction in farmers’ millet
marketing cost had to be estimated. Primary data collected during the 1992/93
household surveys reveal that average millet yields were 240 kglha in Ovambo
and 260 kglha in Kavango. The regression analysis of millet yields indicated that
the use of manure or chemical fertilizer increases millet yields by 157 kilogram
per hectare on average. This result is in agreement with secondary data from
lCRlSAT's 1992/93 on-fan'n cropping trials which indicate that farmers using
manure, could increase their yields by up to two thirds if they also adopted
improved agronomic practices. From these results it was estimated that yield
increases in the future could reduce the average production costs of farmers
producing a marketable millet surplus by 40 percent or from N$ 670 to NS 420
per ton in Ovambo and from N5 620 to NS 370 per ton in Kavango. Since the
majority of farmers do not have access to manure it was estimated that the
introduction of improved production technology would increase average yields by
20 percent in Ovambo and 30 percent in Kavango in the long-term.

The household and market level surveys from the 1992/93 production
year demonstrated that farmers’ average millet marketing costs add 40 percent

to the initial millet production costs in Ovambo and 60 percent in Kavango.

329
However, the surveys also indicated that farmers who have access to local
markets pay about half the amount for marketing of what farmers pay without
access to a local market. We further assume that farmers average cost of millet
marketing could be reduced by roughly 50 percent if market improvements were
made such as organizing local market days, improving roads, and providing

market information to farmers and traders.

7.4.1. Millet processing and competitiveness with maize

One of the most important issues with regard to the commercialization
prospects of the millet subsector is whether locally produced and processed
millet can become competitive with maize that is imported from South Africa and
processed by large Namibian milling companies outside the study zones. Until
the end of 1993 only few organizations gained some expertise with millet
processing with hammermills and the marketing of millet meal. The Namibian
Development Corporation had developed a large scale processing operation in
Kavango, two farmer cooperatives had used middle scale millet processing, and
two NGOs had tested and introduced small scale processing on the community
level.

Based on data from the above mentioned enterprises the “Millet
Procurement Model” compared alternative millet procurement
channel/processing options to identify the most cost effective way to provide
processed millet to rural and urban areas in the north. For the model it was

assumed that large and middle scale hammermills are located in urban or rural

330
centers respectively and that the hammermills are combined with mechanical
dehullers that remove the grain’s shell before the actual milling process. It was
also assumed that small scale hammermills are operated in rural communities
and that the dehulling of the millet could either be done manually or
mechanically.

The base or control scenario of the Millet Procurement Model compared
various procurement channels and millet processing options based on 1992/93
production and marketing cost. The results of the model indicated that under .
current subsector conditions locally produced and processed millet could
become price competitive with imported maize meal, if small scale hammerrnill
technology were used at the community level. The model estimates
demonstrated that with the introduction of small scale hammermills at the local
level millet meal prices would be 19 percent lower than the 1993 maize meal
prices in rural Ovambo and 30 percent lower than the 1993 maize meal prices in
rural Kavango. The simulation of reduced production and marketing costs for
millet producers led to a significant increase in competitiveness of millet that is
locally processed with small scale hammerrnill technology (Ovambo: -43 percent,
Kavango: -47 percent). The simulated reductions in production and marketing
cost also makes the procurement of millet via middle and large scale millet
processing operations competitive with processed maize but to a lesser extent

than community based small scale hammermills.

331
7.4.2. Household food security

On the basis of demographic and production data from the 1992/93
household surveys and FAQ recommendations for the minimum energy
requirements per capita, the “Household Food Balance Model” estimated that
38 percent of rural households in Ovambo and 43 percent in Kavango were food
insecure between 1992/93 and the 1993/94 harvest. The simulation of an
increase in millet yields and reduction of marketing costs revealed that policies
that improve millet sector performance in the future could reduce food insecure
households from 38 to 23 percent in Ovambo and from 43 to 27 percent in
Kavango.

An hypothetical increase of household income by an average salary of an
unskilled laborer (NS 6000 per year that allows at 1992/93 prices the purchase of
the annual millet consumption requirement of an average household) would
reduce the percentage of food insecure households from 38 to 3 percent in
Ovambo, and from 43 to 0 percent in Kavango. This finding underlines the
importance of non-agricultural employment opportunities for reducing household

food insecurity.

7.4.3. Grain imports

The “Grain Import Substitution Model” estimated that the 1995 grain
import requirement in Ovambo and Kavango was 92,000 tons or 57 percent of
annual food grain requirements. Under the assumption that it will take

approximately 10 years to carry out millet production and marketing

332
improvements and current population growth rates will persist until 2005
(Ovambo: 3.1 percent, Kavango: 2.5 percent) the simulation results reveal a
potential decline of grain import requirements from 92,000 tons in 1995 to 47,000
tons in Ovambo and Kavango by the year 2005. This potential reduction in grain
imports could generate gross savings of US$ 12.3 million in the year 2005 (on
the basis of the 1993 parity price of imported white maize) and cumulative gross

savings of US$ 64.8 million from 1995 to 2005.

7.4.4. Constraints on the improvement of the millet subsector

Although the simulation results indicate that increased yields, reduced
marketing costs, and cost effective processing technologies could help to
achieve some of policy makers' goals for northern Namibia, the general

prospects for the millet subsector are not optimistic for the following reasons:

(1) Uncertain rainfall will generate yield instabilities.

(2) Farmers will increase their investments into staple grain production only, if
it is perceived to be profitable on a recurring basis. The heavy migration
to Vifindhoek since independence indicates that many would prefer
temporary employment in the capital to farming in communal areas.

(3) Okashana 1 has not lived up to the claims of crop scientists. Moreover,
improved millet technologies suitable to the varying production conditions
in the study zones are not yet identified at this time.

(4) In Kavango, large areas of rain fed land for crop production are
unaccessible for small holders because of the lack of rural roads and a
reliable source of drinking water.

333

(5) Grain imports from neighboring countries, such as millet from Angola have
added a new dimension to the national food security policy debate.
Keeping the boarders open is of obvious benefit to Namibian traders and
to net food buyers in Ovambo and Kavango.

(6) Consumers' preference will likely shift from millet to maize as younger
consumers become used to maize meal because of the lack of a reliable
supply of locally processed millet.

7.5. Implications for millet policy

The research findings reveal that changes in production, marketing, and
processing can improve the overall performance of the millet sector and thus
improve the living conditions of the rural population in Ovambo and Kavango.
However, the constraints described in the preceding section demonstrate the
difficult task of decision makers to design a grain policy that on the one side
fosters the millet sector development while on the other does not aggravate
problems of rural poverty and food security that even an improved millet sector is
unable to solve. The four areas to be addressed by policy makers are the

following:

7.5.1. Crop production

Although Okashana 1 has not lived up to the initial expectations of crOp
scientists, it can be of value as a niche variety during short rainy seasons. But
more research is needed on existing millet Iandraces and other early flowering
varieties and agronomic practices that are suitable to the varying conditions in

the study zones.

334

Because Kavango’s potential for commercial millet production is
significantly higher than in Ovambo, the government should give priority to
develop local grain markets, improve roads and the supply of drinking water in
Kavango.

The survey findings that millet contributes much less to households'
consumption and cash income in the study zones, than initially assumed leads to
the conclusion that alternative crops should be tested in the study zones and
made available to farmers as soon as possible. Crops that might have potential
in the study zones are sorghum, groundnut, cowpea, phaselus beans,
pigeonpeas, and bambara nut.

The simulation results of the Grain Import Substitution Model reveal that
at the current rate of population growth, the study zones do not have the
potential to become grain self-sufficient in the medium. From this it is concluded
that government policy should focus programs (a) on increasing grain production
in alternative production zones, with more favorable rainfall conditions, such as
Caprivi in the east and the commercial farming area south of the study zones,
and (b) on rural economic development to generate income and employment in

the study zones.

7.5.2. Marketing in rural areas
The Namibian government should encourage and facilitate the

development of a more market-oriented millet subsector. The government

335
should not pursue the proposed one channel millet marketing scheme and large-
scale processing scheme.

In Kavango were many farmers are producing a marketable surplus, the
most severe millet marketing constraints are: a lack of local markets and
vehicles, high cost of hired transportation, and long distances to markets.
Additionally, traders from urban trading centers complained about difficulties in
locating and reaching farmers with marketable millet surplus in remote areas.
Rural communities without local agricultural markets should be informed about
and encouraged to organize local market days according to their needs to
exchange agricultural products among community members and attract long-
distance traders. The initiation of local market days will not require any
government investment. When local markets become very popular and the trade
volume of perishable products increases significantly, investments in physical
market structures may become necessary.

The construction of feeder roads will make remote grain production areas
(especially in Kavango) more accessible. The encouragement of private taxis
and buses help exploit price differences between various areas.

To encourage commercial millet trade between rural communities over
long distances (including trade between Ovambo and Kavango), traders need to
be better informed about the prices and crop situation in different markets. A
feasibility study of a market information service should be carried out. The
1992/93 food price monitoring survey reveals that the gathering of price data

could be accomplished in a relatively cheap and reliable manner. Since most

336
communal farmers listen to the radio on a regular basis, grain market information

could be included in radio broadcasts.

7.5.3. Millet processing

In most rural households women have to process millet manually. This is
a laborious and time consuming work. In the towns Oshakati and Rundu, an
increasing number of consumers are requesting maize processors to process
their millet. This is an indication that the demand for millet processing will grow
as household incomes increase in rural areas and women's opportunity cost for
labor begin to rise.

The government should encourage private traders to invest in small scale
millet processing technology. Government could provide training for rural
entrepreneurs and farm cooperatives in how to set up and maintain an efficient
processing business. The Millet Procurement Model has shown, small scale
hammermills can be competitive with imported and processed maize in villages.
Some rural development projects have demonstrated that the hammen'nill
technology is adaptable to rural needs and provides an affordable alternative to
manual processing. In addition, small-scale mills generate more rural

employment than middle and large-scale processing units.

 

337
7.6. Attacking rural poverty

Although the expansion of millet production could contribute significantly
to the development of northern Namibia, additional policies are needed to
address rural poverty and unemployment. Study findings clearly demonstrate
that improvements in the millet subsector cannot offer a solution to rural poverty
for a significant fraction of the rural population in the study zones. A third of rural
households in Ovambo and 51 percent in Kavango cultivate less than 2 hectare
and derive less than 33 percent of their total income from agriculture. Therefore,
benefits of improved millet production technology will accrue primarily to
households with adequate productive resources to produce a marketable
surplus.

Rural households with limited resources have to find non-agricultural
income sources. While the men of resource poor households tend to look for off-
farrn employment opportunities in non communal regions, women are largely
engaged in activities complementary to agriculture such as raising small animals,
dairy processing, handicraft, and trade. Simulation results from the Household
Food Balance Model revealed that roughly a quarter of the study zone
population would still be food insecure, even if millet production were expanded
and the real price of millet were reduced. The implication of these findings is that
policies are urgently needed to generate alternative jobs and incomes in the

study zones.

338
7.7. Implications for future research

Three areas warrant further research attention: First, to better determine
the prospects for millet, more research is needed on millet technology
development and farm management research which could be conducted jointly
by the national extension service and the newly established agricultural research
department of the MAWRD.

Second, national and international researchers should focus on strategies
to enhance the productivity in agriculture and rural economic development
programs to (a) create off-farm jobs for men close to their rural homesteads and
(b) link women’s income generating activities with market institutions that can
respond efficiently to consumer demand. Such programs require careful
monitoring, evaluation, and re—design in a learning by doing mode of operations.

The third area of research is regional grain trade in the light of recent
millet imports from neighboring countries such as Angola, Zambia, and
Zimbabwe. Research is needed to investigate the size of this trade potential to
determine how improved trade links could affect Namibia's regional and national

food security.

APPENDIX A

339

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 1 . Correlation coefficients of the PLOWHOE variable
(use of hoe as sole plowing equipment; DUMMY, 1 = yes)
OVAMBO KAVANGO
Correlated Variable Coef. Sig. Correlated Variable Coef. . I Sig. .
hh male headed -.2444 .000 eastern Kavango .2107 l .021
hh grain self-sufficiency .2259 .001
hh earns migration income -.2010 .004
migation employment _' 1982 .006
largest lncome source
:leerrcnebretrgflicgrate "1418 '047
hectare yields -.1742 .030
larger ha category -.3239 .000
1992/93 grain production -.1820 .010
eastern Ovambo .1906 .007

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 2. Correlation coefficients of the PLOWANIM variable
(use mainly animal spann for plowing; DUMMY, 1 = yes)
OVAMBO KAVANGO
Correlated Variable Coef. Sig. Correlated Variable Coef. Sig.
hh grain self-sufficiency .2259 .001
own tractor -.1463 .039
swirling“: .0.
1991/92 grain production -.1514 .048
hh judge soil as good .1655 .019
western Ovambo .2377 .001
central Ovambo -.2811 .000
sells millet in some years .1566 .027
sells millet never .1566 .027
close to market -.1382 .0.51

 

 

 

340

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 3. Correlation coefficients of the PLOWTRAC variable
(use mainly tractor for plowing; DUMMY, 1 = yes)
OVAMBO KAVANGO
Correlated Variable Coef. Sig. Correlated Variable Coef. Sig.
hh male headed .1847 .009 hh grain self-sufficiency 1776 .054
hh earns migration income .2163 .002 hh earns migration income .2018 .028
hh cash earnings .2244 .001 hh cash earnings .2231 .015
migation employment migation employment
largest income source '1612 '027 largest income source “2159 '025
percent of hh percent of hh
members literate 2177 '002 members literate 2832 '002
. hectare cultivated in
own a motor vehicle .2504 .000 1992/93 .2214 .035
own a tractor .2308 .000 more than 5 ha cultivated .1915 .037
plan selling of grain _
at planting time .1592 .024
larger ha category .2271 .004
central Ovambo .3469 .000
eastern Ovambo —.1961 .005
Table 4. Correlation coefficients of the DRFTANIMPLOW variable
(ownership of draft animals and plow; DUMMY, 1 = yes)
OVAMBO KAVANGO
Correlated Variable Coef. Sig. Correlated Variable Coef. Sig.
. . hectare cultivated in -
hh grain self-sufficrency .1727 .014 1992/93 .2820 .007
hectare per adult-
larger ha category .2271 .004 equivalent .2438 .020
central Ovambo .3469 .000 1992/93 grain production .2871 .002
eastern Ovambo -.1961 .005

 

 

 

341

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 5. Correlation coefficients of the LANDCONSRT variable
(household has problem to cultivate new land; DUMMY, 1 =
yes)

OVAMBO KAVANGO
Correlated Variable Coef. Sig. Correlated Variable Coef. Sig.
hectare cultivated in
hh male headed -.1553 .030 1992/93 .2820 .007
hh cash earnings -.2432 .001 hectare per adult-equivalent .2438 .020
percent of hh _ . .
members literate .1456 .045 1992/93 grain production .2871 .002
hh adult-equivalent -.2390 .001
hectare cultivated in
1992/93 -.2313 .004
hectare ”9' ad““‘ -.1970 .014
equrvalent
larger ha category -.3032 .000
more than 5 ha cultivated -.3180 .000
1992I93 grain production -.2351 .001
sells‘millet in some years .1712 .017
sells millet never -.1712 .017
close to market -.2184 .002
Table 6. Correlation coefficients of the HHLABSHRT variable
(household experienced labor shortage; DUMMY, 1 = yes)
OVAMBO KAVANGO
Correlated Variable Coef. Sig. Correlated Variable Coef. Sig.
western Ovambo .1880 .009 western Kavango .2237 .019

 

 

 

 

 

 

 

 

 

 

342

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 7. Correlation coefficients of the MILLETPRICE variable
(local price for millet at the begin of planting; NSIKG)
OVAMBO KAVANGO
Correlated Variable Coef. Sig. Correlated Variable Coef. Sig.
. . difficulties with land -
uses chemical fertllzer .3420 .000 clearing .2888 .040
central Ovambo -.2720 .000 owns moter vehicle ‘3150 .022
sell millet some years - .1916 .008 western Kavango .2842 .039
sell millet never .1916 .008 sell millet never .3992 .027
Table 8. Correlation coefficients of the MNRALLFLDS variable
(use fertilizer (manure S chemical) on all fields; DUMMY, 1 =
Y»)
OVAMBO KAVANGO
Correlated Variable Coef. Sig. Correlated Variable Coef. Sig.
. . hectare cultivated in
hh grain self-sufficrency .1905 .007 1992/93 .2628 .012
. hectare per adult-
own a motor vehlcle .1408 .047 equivalent .3307 .001
plan selling of grain .
at planting time .1858 .008 more than 5 ha cultivated .2654 .004
larger ha category .1890 .018
1991/92 grain production .1610 .035
1992/93 grain production .1408 .047
hh judge soil as good .1887 .007
western Ovambo .2123 .003
central Ovambo -.2700 .000
sells millet in some years .1986 .005
sells millet never -.1986 .005

 

 

 

343

Table 9. Correlation coefficients of the CHEMUSE variable
(use chemical fertilizer; DUMMY, 1 = yes)

 

 

 

 

 

 

 

 

 

OVAMBO KAVANGO
Correlated Variable Coef. Sig. Correlated Variable Coef. Sig.
hh grain self-sufficiency .1404 .052 close to roads .2011 .030
eastern Ovambo -.2429 .001 western Kavango -.2610 .004
sells millet in some years .1833 .050
sells millet never -.2580 .005

 

 

 

 

 

Table 10. Correlation coefficients of theEXPOKASH variable
(experience with Okashana 1 before 1992I93; DUMMY, 1 = yes)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

OVAMBO KAVANGO

Correlated Variable Coef. Sig. Correlated Variable Coef. Sig.

hh male headed .1804 .011 hectare cultivated in 1992/93 .2608 .013
hh grain self-sufficiency .3175 .000 hectare per adult-equivalent .2651 .012
hh cash earnings .1655 .020 1992/93 grain production .1941 .035
$2123“ agate .2689 .000 central Kavango .2160 .019
hh adult-equivalent .1512 .033 close to market -.3050 .001
own a motor vehicle .2177 .002

sagasffi'i'rggtgg'a'" .1499 .035

larger ha category .2590 .001

hectare yields .2278 .004

1992/93 grain production .1531 .031

hh judge soil as good .1430 .044

western Ovambo .3537 .000

eastern Ovambo -.3038 .000

sells millet in some years .2030 .004

 

sells millet never .-.2030 .004

 

 

 

 

 

 

LIST OF REFERENCES

 

LIST OF REFERENCES

Bidinger, F .R. (1993). “Research on Small Farmer Crops in Northern Namibia.”
Report to the Division of Agricultural Research, Ministry of Agricultural,
Water and Rural Development, Government of the Republic Namibia.
Windhoek, Namibia.

Central Statistics Office, (1993). “Statistical Abstract 1993, No.2", National
Planning Commission, Windhoek, Namibia.

Dendy, David, (1993) "Report on a visit to Namibia to assist in the improvement
of pearl millet processing and products at the Namibia Development
Corporation, Musese, Kavango,” Natural Resources Institute, Overseas
Development Administration, 1993, p.13 and p.15.

FAO, (1974). “Handbook on Nutritional Requirements.” Nutritional Studies No.
28, Rome.

Hay, R., John, P., Tanner, C. (1990). “Household Food Security in Northern
Namibia." lntemational Development Center Food Studies Group,
University of Oxford, United Kingdom.

ICRISAT (1991), Newsletter of the lntemational Crops Research Institute for
Semi-Arid Tropics, SAT-NEWS, No.7, Jul-Sep 1991.

Matanyaire, CM. (1994). “Prospects for Increasing Pearl Millet Production in
Namibia.” Paper presented at the Second Namibian Millet Workshop,
Windhoek, Namibia.

MAWRD, (1994 a). “National Agriculture Policy." Draft for Public Consultation,
Ministerial Policy Task Force and Directorate of Planning, Ministry of
Agriculture, Water, and Rural Development, Windhoek, Namibia.

MAWRD, (1994 b). “Crop and Food Security Bulletin No: 4/94,” Namibia Early
Warning 8 Food lnfonnation Unit, Directorate of Planning, Ministry of
Agriculture, Water, and Rural Development, Windhoek, Namibia.

UNICEF I NISER (1991). “A Situation Analysis of Children and Women in
Namibia,” Windhoek, Namibia.

344

 

STRT

III

53

 

E UNIV. LIBRARIES
IIIIIIIIIIIIIIIIIIIIIIIII'I
15552437

I
C)

., tit-WWII...—