E
«Egg , .. am
; r... ‘ .9. .tnuhq

. u: “Exam..-

i flan»

: ...
«2.: . .
59...; . , ‘ i.
. . L: . ..
.. p! . , 1..

. , . . ‘ . .413!

2.. . .

.19).. . ‘ A . .. 3;...

3.}; . , . , I
1.... .

n . yr
‘ .. :
a :5. 5...». :flfim‘.
5.3 .5}: a a

105.3 I. I.

aw i m V ’
Vt. : In 39...? tam. an
. 3 .. I .u. I...» v! . .I
V . . . vinx. “. .nm a.” x . a. 1.
... .2... :3 . . , , a: ‘n H. .. as .-
‘ ‘ , ., t u. z.
, V . . , . . , 0 31¢.
. .xxrahwz... . . ‘. ‘ . . «wave.
:Yt'FAoO’C . , . .
ii. ,

r4
1 2.... .. .
,. Era: L...
I, .33....

ifisin

..
3 .5339
"shah-4 .2.
. .. .1 . ‘ .
. , 4 s
.2... u..........uߴ.l._u . V . , .mfn?
. . . , . J xiii; 4..."
. ‘ , , a... Ch
. .. a... xv.
131 2.. 3? ER: 3. . t.
3: 33:2. 3., .311 "In 7.4
:27 23...}: .. «3.9.?
a «ha. a. x:
. annuity“... 4.
.P «0.... ‘
\.< ‘
.xvmuzw . :

t. . 1 ,. x
. x $3! mum. 3%
X ll‘n‘. Z lulr‘nl.|
33.“: 3.2
. r1}. . A»!
v: I:
V AHHEEIHE a:
. kg? .
. ‘ . . . . a Inn‘hu
. , . V ‘ . :5...
iii . . . . ax...
. ‘ V ‘ V ‘ , . P??§Mu.

\ at“

‘v.

.v . v2.1!» :1);

I41wN~1 .~.(.‘%
r .{I

_ : . . 113.2.

, 2,545.2. . .. 3.3.3
\. In . :. Va . ‘ l V\6\<IV if I.
$5. . n ....

 

A V A .zhmwm
-251 km «.344 h!

.11.? .l. t. .I'Yifio}.g
unlsvlihtk. I \

.5113!
+1 .1:

 

bolts mvensmr momma

") ’lllllllilllllln Illlll an ml"

Ufl U\) 3 1293 01405 3296

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

This is to certify that the

thesis entitled

Determinants of Farm Household Participation
in Rwanda's Banana Wine Subsector

presented by

Pamela Riley Miklavcic

has been accepted towards fulfillment
of the requirements for

MS degree in Aqricultural Economics

    

 

Major professor

 

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

LIBRARY
M"mitten State
University

 

 

 

PLACE IN RETURN BOX to remove thin checkout from your record.
TO AVOID FINES return on or bdoro dot. duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU IoAn Nat-mum Action/Equal Opportunity lnotltwon
Wan-9.1

DETERMINANTS OF FARM HOUSEHOLD PARTICIPATION
IN RWANDA’S BANANA WINE SUBSECTOR

By

Pamela Riley Mildavcic

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

MASTER OF SCIENCE

Department of Agricultural Economics

1 995

ABSTRACT

DETERMINANTS OF FARM HOUSEHOLD PARTICIPATION
IN RWANDA’S BANANA WINE SUBSECTOR

By

Pamela Riley Miklavéié

This thesis examines the determinants of farm household participation in Rwanda’s
banana wine subsector (BWS) and demonstrates the usefulness of a combined subsectoral
and household-level analysis. I rely on empirical and theoretical findings to identify
variables which influence BWS participation. A two-stage Heckman analysis is used, with
an Inverse Mill’s Ratio correcting for possible selectivity bias. Landholdings, income, and
zonal location are the most significant determinants of banana and wine production.
Although most rural Rwandans participate in the BWS, poor households participate more
regularly in banana sales, whereas richer households produce and sell larger quantities of
banana wine. The policy implications are: 1) efforts to strengthen the BWS would assist
those already most food secure while 2) research into alternate uses for beer bananas,
including their possible substitution with fruit or cooking bananas, would ensure the
continued growth of this erosion-controlling crop while simultaneously identifying new

banana-oriented ME opportunities for rural Rwandan households.

To Uncle Brooks, for inspiring me;
To Mom and Dad - and all my siblings - for encouraging me;

To Milan, for being there every step of the way.

iii

ACKNOWLEDGMENTS

Thank you to my adviser, Dr. Thomas Reardon, for his patience, enthusiasm, and
guidance throughout the last two years. To my other committee members, Dr. Michael
Weber and Dr. Donald Mead, for their advice and support at Michigan State University.
To Dr. Daniel Clay for his suggestions during the last year and a half, and for
participation as an alternate committee member during Dr. Weber’s sabbatical. Also,
thank you to Dr. James Shaffer, for challenging my logic while in Rwanda and to Dr.
Carl Liedholm, for spurring my initial interest in microenterprise development.

Thank you to all the staff at the Division of Agricultural Statistics in Kigali,
Rwanda for their advice and assistance during my six-week visit in 1993, especially my
counterpart, Vincent Habimfura, for his insightful perspective, enthusiasm, and friendship
during our project association. I hope that you will one day be able to return to Rwanda,
and see your country restored to peace and tranquillity.

Thanks to the computer staff at MSU - Chris Wolf, Jeff Wilson, Elizabeth
Bartilson, Lisa Besko, and Margaret Beaver. Also, thanks to classmates Duncan
Boughton, Lisa Daniels, Cynthia Donovan, Jaakko Kangasniemi, Mattias Lundberg,
Cynthia Phillips, and Kyosti Pietola for their help when this project seemed daunting.

Most of all, thank you to my husband, Milan, who taught me, perhaps more than

anyone, about the patience and diligence required of good research.

iv

TABLE OF CONTENTS

PAGE
LIST OF FIGURES .......................................... viii
LIST OF TABLES ............................................ ix
LIST OF ABBREVIATIONS ..................................... xii
CHAPTER 1 ................................................. 1
INTRODUCTION AND METHODS ................................. 1
1.1 Background ............................................... 1
1.2 Problem Statement .......................................... 2
1.2.1 What determines household participation in
Rwanda’s banana wine subsector? ......................... 2
1.2.2 How can subsectoral findings strengthen
household-level research? .............................. 4
1.3 Research Questions ......................................... 7
1.4 Research Approach .......................................... 8
1.5 Data ................................................... 10
1.6 Stratification of Sample . . .................................... 13
1.6.1 Agroclirnatic zones .................................. 13
1.6.2 Banana sellers and non-banana sellers ..................... 13
1.6.3 Large wine sellers and small wine sellers ................... 15
1.6.4 Rural and semi-rural households ......................... 15
1.7 Derived Data ............................................. 16
1.7.1 Wine production .................................... 16
1.7.2 Household wine consumption calculation and
wine purchases ..................................... 17
1.8 Description of Rwanda ...................................... 18
1.9 Organization of the Study .................................... 19
CHAPTER 2 ................................................ 20
DESCRIPTION OF THE BANANA WINE SUBSECTOR -
BANANA PRODUCTION AND TRANSACTIONS ..................... 20
2.1 Introduction ............................................. 20
2.2 General Structure of the Banana Wine Subsector .................... 21
2.3 Banana Production ......................................... 26
2.3.1 National and zonal findings ............................ 26

PAGE

2.3.2 Constraints to banana production ........................ 30

2.3.3 Qualitative observations ............................... 32

2.3.4 Data patterns and inferences from them .................... 33

2.4 Banana Transactions ........................................ 33

2.4.1 National and zonal findings ............................ 33

2.4.1.1 SALES ............................... 33

2.4.1.2 PURCHASES ........................... 42

2.4.2 Constraints to banana transactions ........................ 44

2.4.3 Qualitative observations ............................... 45

2.4.4 Banana sellers versus non-banana sellers ................... 45

2.4.4.1 BWS CHARACTERISTICS ................. 46

2.4.4.2 HOUSEHOLD CHARACTERISTICS .......... 48

2.4.4.3 DEMOGRAPHIC CHARACTERISTICS ........ 48

2.4.5 Data patterns and inferences from them .................... 51

2.5 Conclusion .............................................. 51

CHAPTER 3 ................................................ 53
DESCRIPTION OF THE BANANA WINE SUBSECTOR -

WINE PRODUCTION AND TRANSACTIONS ........................ 53

3.1 Introduction ............................................. 53

3.2 Wine Production .......................................... 54

3.2.1 National and zonal findings ............................ 54

3.2.2 Constraints to wine production .......................... 59

3.2.3 Qualitative observations ............................... 60

3.2.4 Data patterns and inferences from them .................... 60

3.3 Wine Transactions ......................................... 61

3.3.1 National and zonal findings ............................ 61

3.3.1.1 SALES ............................... 61

3.3.1.2 PURCHASES ........................... 68

3.3.2 Constraints to wine sales .............................. 72

3.3.3 Qualitative observations ............................... 72

3.3.4 Small versus large wine sellers .......................... 73

3.3.4.1 BWS CHARACTERISTICS ................. 74

3.3.4.2 HOUSEHOLD CHARACTERISTICS .......... 76

3.3.4.3 DEMOGRAPHIC CHARACTERISTICS ........ 78

3.3.5 Data patterns and inferences from them .................... 78

3.4 Kigali’s Unusual Market Structure Versus the Rest of Rwanda ........... 80

3.5 Wholesalers and Transporters .................................. 81

3.6 Conclusion .............................................. 82

CHAPTER 4 ................................................ 83

DETERMINANTS OF HOUSEHOLD-LEVEL PARTICIPATION ........... 83

4.1 Research Questions ........................................ 83

vi

PAGE

4.2 Model Specification ........................................ 83
4.2.1 Conceptual model ................................... 83

4.2.2 Economic framework ................................ 86

4.2.2.1 PRODUCTION MODEL ................... 86

4.2.2.2 MARKETED SURPLUS MODEL ............. 87

4.3 Regression Specification ..................................... 88
4.3.1 Heckman’s two-stage procedure ......................... 91

4.3.2 The PROBIT models ................................ 92

4.3.3 The OLS models ................................... 93

4.3.4 Variable justification and hypotheses ...................... 95

4.3.4.1 PRICES ............................... 95

4.3.4.2 FARM CHARACTERISTICS ................ 96

4.3.4.3 FACTORS OF PRODUCTION ............... 99

4.4 Descriptive Statistics ...................................... 100
4.5 Regression Results ........................................ 101
4.5.1 Participation in banana production ...................... 101

4.5.2 Participation in banana sales ........................... 110

4.5.3 Participation in wine manufacturing ..................... 112

4.5.4 Participation in wine sales ............................ 113

4.5.5 Supplementary findings .............................. 115

4.6 Conclusion ............................................. 117
CHAPTER 5 ............................................... 119
POLICY IMPLICATIONS AND CONCLUSIONS ..................... 119
5.1 Policy Implications ........................................ 119
5.2 Methodological Conclusions ................................. 122
5.3 Further Research ......................................... 123
5.4 Summary .............................................. 124
APPENDIX A .............................................. 126
1993 SURVEY ON PARTICIPATION IN THE BANANA WINE SUBSECTOR 126
APPENDIX B .............................................. 138
DESCRIPTION OF RWANDA’S FIVE AGROCLIMATIC ZONES ......... 138
BIBLIOGRAPHY ............................................ 142

vii

Figure 1.1:

Figure 1.2:

Figure 1.3:
Figure 2.1:
Figure 2.2:
Figure 2.3:
Figure 2.4:

Figure 4.]:

LIST OF FIGURES

PAGE
Flow chart of the banana wine subsector .................... 6
Schematic of the paper’s subsectoral and
household-level components ............................. 9
A map of Rwanda’s prefectures and zones .................. 14
Map of Rwanda’s banana and wine trade flows ............... 22
Flow chart of the wine subsector in rural areas ............... 23
Flow chart of the BWS in semi-rural areas .................. 24
Simplified view of the BWS options open to HHS ............. 26

Conceptual schematic of the proposed model ................ 84

viii

Table 2.1:

Table 2.2:

Table 2.3:

Table 2.4:

Table 2.5:

Table 2.6:

Table 2.7:

Table 2.8:

Table 2.9:

Table 2.10:

Table 2.11:

Table 2.12:

Table 2.13:

Table 2.14:

Table 2.15:

Table 2.16:

LIST OF TABLES

PAGE
Banana producers by zone (%) .......................... 27
Banana production by zone (F RW) ....................... 28
Banana production as share of household income (FRW) ........ 29
Zonal GINIs for banana production by
banana producer quintile (FRW) ......................... 3O
Zonal banana production by value category (FRW) ............ 31
Banana sellers by zone (%) ............................ 34
Banana sales by zone (FRW) ........................... 35
Banana sales as share of household income (%) .............. 36
Zonal correlations of household
revenues and banana seller quintiles (FRW) ................. 37
Zonal GINIs for banana sales by banana
seller quintile (F RW) ................................ 39
Zonal correlations of banana sales and
banana producer quintiles (FRW) ........................ 39
Marketed surplus of banana production (%) ................. 40
Producer price of bananas by zone (F RW/KG) ............... 40
Zonal banana sales by value category (FRW) ................ 41
Banana buyers by zone (%) ............................ 42
Banana purchases by zone (FRW) ........................ 43

ix

Table 2.17:

Table 2.18:

Table 2.19:

Table 2.20:

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:

PAGE
Zonal GINIS for banana purchases by

banana buyer quintile (F RW) ........................... 44
BWS characteristics of B85 and NBSs (N=757) .............. 47
Household characteristics of BSs and NBSs (N=757) ........... 49
Demographic characteristics of 885 and NBSs (N=757) ......... 50
Wine producers by zone (%) ........................... 55
Wine production by zone (F RW) ........................ 56
Wine production as share of household income (F RW) ......... 57

Zonal GINIS for wine production by wine

producer quintile (F RW) .............................. 58
Zonal wine production by value category (F RW) ............. 59
Wine sellers by zone (%) .............................. 62
Wine sales by zone (FRW) ............................ 62
Marketed surplus of wine (%) .......................... 63
Wine sales as share of household income (FRW) ............. 64
Zonal correlations of total household revenues and

wine seller quintiles (FRW) ............................ 65
Zonal GINIS for wine sales by wine seller quintile (FRW) ....... 66
Zonal wine sales by value category (FRW) ................. 67
Producer price of wine by zone (FRW/LITER) ............... 68
Wine buyers by zone (%) ............................. 69
Wine purchases by wine buyers by zone (FRW) .............. 70

Table 3.16:

Table 3.17:

Table 3.18:

Table 3.19:

Table 4.1:

Table 4.2:

Table 4.3:

Table 4.4:

PAGE
Zonal GINIS for wine purchases by

wine buyer quintile (FRW) ............................ 71
BWS characteristics of SWSs and LWSs (N=129) ............. 75
Household characteristics of SWSs and LWSs (N=129) ......... 77
Demographic traits of SWSs and LWSs (N=129) ............. 79
Descriptive statistics (household-level, N=853) .............. 102
Correlations of dependent variables ...................... 103
Decision to participate: PROBIT results ................... 104
Participation level: OLS results ......................... 106

xi

BS

BWS

DSA

F RW

GDP

HH

LDC

LWS

MSU

NBS

OLS

SS

SWS

LIST OF ABBREVIATIONS

Adult Equivalent

Banana Seller

Banana Wine Subsector
Division of Agricultural Statistics
Rwandan Franc

Gross Domestic Product
Household

Developing Country
Large Wine Seller
Microenterprise

Michigan State University
Non-Banana Seller
Ordinary Least Squares
Subsector

Small Wine Seller

xii

CHAPTER ONE

INTRODUCTION AND METHODS

1.1 Background

During the last 15 years, increasing research has been done on rural, Small-scale
industries in developing countries (LDCs) (see Chuta and Liedholm, 1979; Liedholm and
Mead, 1987; McPherson, 1992; Parker 1995). In a study of 18 LDCs, Kilby and
Liedholm found that the primary occupation of 14-48% of the rural labor force is in
activities outside of agriculture (1986). According to Nowak (1989), this figure is
probably higher since Kilby and Liedholrn’s survey did not include secondary occupations
or female workers.

LDC-based microenterprises (MES), defined by Chuta and Liedholm (1979) as
having fewer than ten employees and selling more than 50% of production, account for
more than 20% of aggregate household income, especially in rural areas where 90% of
small businesses are located (see Liedholm, 1990; Liedholm and Mead, 1990).
Manufacturing is especially important, accounting for 67% of the rural MES studied in

southern and eastern Africa (Liedholm and Mead, 1992).

2

Despite these statistics, this thesis is one of the few which attempts to examine the
determinants of farm household participation in a specific manufacturing MB. In this

paper, I will study household participation in Rwanda’s banana wine subsector (BWS)‘.

1.2 Problem Statement

1.2.1 What determines household participation in Rwanda’s banana wine
subsector?

Throughout Africa, a frequently observed nonfarm activity is the home brewing
of indigenous alcoholic beverages, especially sorghum beer and banana wine (Haggblade,
1987). To date, little research has examined either of these activities - and, to my
knowledge, never in terms of participatory determinants. Although Haggblade (1984)
conducted studies of the indigenous beer industry in Botswana and the alcoholic beverage
subsector in Rwanda (1987), he was primarily concerned with how the subsector functions
- not what causes households to participate in the subsector. In addition, by focusing on
larger, more commercial BWS ventures in Rwanda - in which rural households do no;
participate - Haggblade excluded from his study the rural participatory element in which
I am most interested.

Banana wine, rather than sorghum beer, is the focus of this research because

bananas are a predominant crop which covered 26% of the increasingly cultivated

 

lAlthough referred to elsewhere as banana beer, and made with bananas locally known
as ’beer bananas’, I have chosen to use the term ’banana wine’. This is because wine is
generally recognized as being fruit-based and beer, grain-based. Nevertheless, I did
encounter a women’s cooperative in Gisenyi that produced both banana beer and banana
wine. Their fermentation process was much longer for the product they referred to as
banana wine, as was the shelf-life of the resultant product.

3
Rwandan agricultural land throughout both growing seasons in 1990 and supplied 27%

of total calorie output (N girumwami, 1992). Sorghum, on the other hand, covered only
13% of Rwanda’s cultivated land during one growing season, dropping to 2% during the
rest of the year.

Although fruit and cooking bananas are also grown in Rwanda, they are less
dominant than beer bananas and outside the scope of this paper. Whereas beer bananas
accounted for 71.9% of all banana production in 1990, cooking bananas accounted for
only 19.6% of banana production and fruit bananas for even less (8.5%). Although fruit

bananas are occasionally added to wine for flavor, this occurrence is rare. Therefore,

unless sgcified otherwise, bananas will always refer to beer bananas in this document and

wine always to banana wine.

Home-based banana wine brewing (and the cash income it generates) contributes
significantly to Rwanda’s informal economy. Among households selling wine, mean
gross wine sales ranged from 7% to 13% of annual household income in 1990 -
depending on the agroclimatic zone (Miklavcié and Habimfura, 1995). This was between
9% and 27% of non-crop income.

Since no researcher has yet studied the impact of household characteristics (e. g.,
demographics, household income, agroclimatic location, or farm size) on the rural
Rwandan household’s decision to become involved in the BWS, it is not clear what

potential might exist, through government policy acting on these determinants, to expand

4

banana wine involvement as a strategy for earning more incomez. Therefore, the core
objective of this thesis is to explore the determinants of participation in the BWS as a
function of the endogenous and exogenous characteristics of the household. In turn,
it is hoped that this objective will help me to strategically address, in future research, the
assessment of other household-based MES, to augment income and thereby improve food

security throughout the developing world.

1.2.2 How can subsectoral findings strengthen household-level research?

A second objective is to demonstrate how a subsector study can inform, and
thereby improve, the quality of a household-level study. Although subsectoral findings
have already been applied in the context of ME research (Boomgard et al, 1986),
household-level research continues to be criticized for losing track of the big picture.
Much contextual information associated with the question of interest is neglected,
especially that involving upstream and downstream linkages to the household. For
example, some subsectoral characteristics formed outside the household are ignored as
influences shaping the household’s decision to participate in the subsector (i.e, proximity

of market and transport services). In turn, policy recommendations that could have been

 

enriched by this added contextual knowledge are not.

The reverse view - that household behavioral work can be used to reinforce

subsector studies - has already been successfully demonstrated by Boughton et a1. (1994).

 

2Although the use of supplemental funds is touched on only briefly in this paper, I
am assuming that significantly augmented income would probably be used by HHS to
encourage the grth of additional income-diversifying MES.

5

For example, in the case of banana wine, accessibility of certain inputs influences whether
a household produces a given output (i.e., a household must have enough sorghum,
bananas, water, labor, and containers to manufacture wine). If a subsector study does not
examine such household-specific elements, it misses crucial explanatory information about
the subsector.

To date, it is rare to use a combined approach in a household-level study. By
identifying and better understanding these outside influences, like subsectoral
characteristics that act on a household’s decision to participate in the BWS, I hope to

strengthen my:

1) selection of regression variables for Chapter 4, and
2) overall conclusions and policy recommendations regarding
Rwanda’s BWS.

To carry out this subsector-informed household-level approach, I collaborated with
the Division of Agricultural Statistics (DSA) in Kigali, Rwanda to conduct both wine-
specific subsectoral and supplemental household-level research, each of which was used
to expand upon the extensive household production and transaction data collected by DSA
in 1990. Figure 1.1 is a flow chart showing the stages within the subSector that each
household can participate in. Each rural household is involved in zero, one, or more of

the subsectoral stages, to varying degrees. Its choice of activities depends not only on the

6

Figure 1.1: Flow chart of the banana wine subsector

BANANA PRODUCTION
(sometimes involving hired labor)

transportation

market transactions
(direct or through middleman)

BANANA SALES

transportation

WINE PRODUCTION
(hired labor)

transportation

market transactions
(direct or through middleman)

WINE SALES
(to clients or bar owners)

transportation to selling point

sales from bar to client

household’s production and other characteristics, but also on existing conditions elsewhere
in the subsector. In other words, upstream linkages to other parts of the subsector (like
regional banana production), and to transportation infrastructure influence a household’s
decision to produce bananas, as do downstream linkages (like banana and wine demand).

For example, a household will choose to produce wine for sale only if its income
from wine sales exceeds the opportunity cost of other farm and non-farm activities (e. g.,

potential income earned through hired out labor or another ME activity). In rural areas

7

far from towns, where transportation is scarce and markets are small, production with the
intent of selling will be lower than in places where both markets and transportation
opportunitiesiare much larger (e.g., places located closer to urban centers). This implies
that farmers with the same household characteristics, located in different areas and facing
different subsectoral opportunities, will make different decisions about entering the
subsector.

Through my analysis, I hope to establish the relevance of this combined subsector-

household approach for use in other areas of ME research.

1.3 Research Questions

Four research questions will be used to inform my household analysis:

1) What determines whether, and to what extent, a household
produces bananas?

2) What determines whether, and to what extent, a household
sells at least some of the bananas that it produces?

3) What determines whether, and to what extent, a household
manufactures banana wine?, and

4) What determines whether, and to what extent, a household

sells wine from its own production (i.e., marketed surplus)?

8

Each of the four questions deals with a stage of the subsector — banana production, banana
transactions, wine production, or wine transactions’. Rural households are usually
involved in two or more of these stages, such that it becomes important to understand why

each household decides to participate in some stages but not others.

1.4 Research Approach

Figure 1.2 is a schematic of the steps followed in this research. It shows how the
household-level and subsectoral methods are joined to inform the final regression analyses.

The household-level commnent uses household-level production and transaction
data to statistically describe: 1) sample household involvement in the BWS by
agroclimatic zone, and 2) characteristics of groups involved in the production and
distribution of bananas and wine (e.g., banana versus non-banana sellers and large versus
small wine sellers).

Paralleling the household-level component in Figure 1.2 is the subsectoral
commne t. A common focus of James Shaffer’s work (1973), the purpose of this
component is to describe the structure of the BWS in terms of: 1) the exchange of
bananas and wine, and 2) the role of households in these operations.

Together, the household-level and subsectoral components provide the necessary
information to select a number of farm, price, and climatic variables relevant for inclusion

in the regressions. Following a series of descriptive statistics (including means and

 

3Consumption is also a frequent component of subsectoral studies. However, since
I am primarily interested in the production and marketed surplus aspects of the BWS,
consumption is excluded, for the most part, from my analysis.

Figure 1.2:

Schematic of the paper’s subsectoral
and household-level components

SUBSECTORAL COMPONENT

 

 

RAPID APPRAISAL SUPPLEMENTAL
OF THE BWS SURVEY OF

 

 

 

WINE-SELLING

HHS

 

 

 

 

 

BANANA WINE
SUBSECTOR

 

DESCRIPTION OF THE

 

 

 

HOUSEHOLD-LEVEL COMPONENT

 

1990 CORE DATA

 

 

 

 

 

DESCRIPTION OF
HOUSEHOLDS
INVOLVED IN

BWS ACTIVITIES

***$****$***#**$****

 

 

 

 

 

 

 

SELECTION OF REGRESSION

VARIABLES

 

 

 

RUNNING OF REGRESSIONS
(USING 1990 DATA - 853 HHS)

 

10

standard deviations), these same variables are used to estimate a Heckman two-stage
regression. Heckman’s first stage consists of four PROBIT regressions, each of which
addresses the first four research questions listed in Section 1.4 - those pertaining to the
decision to participate in the different stages of the subsector. The second stage consists
of four ordinary least squares (OLS) regressions which address household l_e_v_els_ of

participation in each activity.

1.5 Data

Three sets of data are used in the analysis. The first set, used in the household-
level component, is quantitative. It consists of production and transaction data collected
in 1990 from 1209 rural households by the Division of Agricultural Statistics (Kigali,
Rwanda). Henceforth, this data will be referred to as the core dag, and will be
consistently treated with standardized weights throughout the remainder of the document.

The core data are used to describe aggregate household participation in the BWS
by agroclimatic zone, to describe household characteristics of specific groups involved in
the BWS, and in the regressions. Included are data on annual crop production, cropping
patterns, sales, and purchases of Q farm commodities. The core data also include
household demographics. Although 1209 households were interviewed, data were
complete for only 853 of the core households. Selectivity bias is not anticipated because
of the random nature of missing information among households.

The second and third data sets are henceforth called the supplemental data. They

are composed of strictly wine-oriented information collected in the summer of 1993.

11

The second set is data from a Rapid Appraisal (conducted in collaboration with
DSA) which included visits to numerous banana and wine exchange points throughout
Rwanda. The structure of several larger, more urban-oriented wine markets were studied
(Butare, Kigali, and Gitarama), as were a number of rural wine markets.

The third set is household-level information collected during the above Rapid
Appraisal. It was gathered from a subset of households who participated in the 1990
survey. An informal survey was used (see Appendix A) because it was necessary to ask
wine-specific questions (not asked in 1990) that allowed enough flexibility to explore
unforeseen wine-related topics“. This is used to expand our understanding of banana
wine as a household activity, and contains quantitative and qualitative material about
banana and wine production and sales which were impossible to deduce from the original
survey. Included are data on: 1) wine consumed at home and away from home, 2) labor
purchases, 3) input quantities, 4) prices, 5) historical involvement of the household in the
BWS, and 6) household projections about future participation.

The Rapid Appraisal household sample consisted of 43 of the core sample’s largest
wine sellers; they comprise the target sample, selected because they sold 20,000 F RW of
wine in 1990, almost equal to Rwanda’s 1990 GDP per capita (26,000 FRW). In other
words, the income from these sales was a significant contributor to the economic well-

being of each household, especially when compared to average Rwandans.

 

’Each supplemental interview lasted between 45 minutes and one hour, and followed
a semi-structured questionnaire devised and pre-tested during two weeks.

12

In addition:

1) This was a reasonable number of households to interview in

only a few weeks, and

2) This number seemed adequate to supply the level of

Supplementary information required for the overall analysis.

During three weeks, 25 of the 43 target households were surveyed’. Ten
additional interviews were also conducted, described as follows. Six were among the
smallest wine sellers in the original 1990 sample, selling between 1 and 2500 F RW worth
of wine. The remaining four were not part of the original study, but were visited because
of:

1) their status as very large wine sellers in Rwanda, and

2) their proximity to the other survey households.

Their inclusion in the survey is intended to Shed further light on the determinants of high

levels of banana wine production and sales“.

 

5All but one of the 25 sold more than 20,000 FRW of wine in 1990 - the last sold
18,950 FRW worth.

6 Originally, I had n_ot planned to visit the smallest or largest sellers. Rather, my

plan was to deduce the characteristics of BWS participants by questioning only the largest
sellers from the original study. Nevertheless, return visits to many of the 43 target HHS
were not possible. A few of the HHS were located but abandoned. Others were
inaccessible because of their location in the rebel-occupied northeastern part of the
country. Many others were simply too far away in terms of the dormant yet ongoing civil
war which made it necessary to return to the safety of Kigali each night. For these
reasons, I decided to expand the sample by adding these ten smallest and largest sellers,
all of whom were located near HHS from the target sample.

It is possible that this choice introduced some bias to the supplemental study. This
bias, however, is considered irrelevant since the data obtained from these interviews was
considered strictly accessory from the start. The purpose of this data was to inform

13

1.6 Stratification of Sample
1.6.1 Agroclimatic zones

Although both agroclimatic and prefectural data are available, I have stratified the
sample by agroclimatic zone. This is because agroclimatic conditions exerted a greater
influence on 1990 levels of banana production than did political boundaries”. By
stratifying in this manner, large banana-producing zones were compared to small ones in
ways focused on climatic influences, rather than political or administrative ones. This
stratification is especially useful in Chapter 5, where zonal distinctions are suggested as
influencing who should be considered as target populations for BWS policy
recommendations. Figure 1.3 is a map of Rwanda, showing its five zones and ten

prefectures (from Schnepf, 1992).

1.6.2 Banana sellers and non-banana sellers
To characterize banana sellers (BSs) and non-banana sellers (NBSS), the 757
households which produced bananas were divided into two groups - those who sold any
bananas (184 households, or 24% of banana producers) and those who did not (573

households, or 76% of banana producers).

 

selection of econometric variables and to provide a reliable context for the interpretation
of the final results. I never intended to examine their statistical significance.

7This statement, however, does not preempt the fact that inter-zonal trade also exerts
an influence on banana production.

8This statement is probably inaccurate as of April 1994, when civil war was renewed,
more than half a million people died, and all crop production and harvesting were
disrupted by political turmoil.

14

Figure 1.3: A map of Rwanda’s prefectures and zones*

    
      
      

. Byumba .

  

Central

1’
O /

W

:‘Bum.

I
E

6 1092111

Agroclimatic border
Prefecture border
I Prefecture capital

 

 

* - taken from Schnepf (1992)

15

1.6.3 Large wine sellers and small wine sellers

To study differences in characteristics of large wine sellers (LWSS) and small wine
sellers (SWSs), I compare the first and last deciles of those sample households which sold
wine. LWSS are the top 10% (64 households) that sold more than 13,700 FRW”. SWSs
are the bottom 10% (65 households) that sold less than 701 F RW of wine in 1990.
Although 65 is not a large sample size, I selected the top and bottom deciles as a means
to identify differences between LWSS and SWSs. Since 70% of my LWSS meet Chuta
and Liedholm’s definition of a ME (compared to only 3% of SWSs and 23% of all wine
sellers), these differences may contain strong policy implications regarding the support of

BWS activities as MES.

1.6.4 Rural and semi-rural households
All households included in this research are rural. That is, they are located in
small, primarily agricultural communities, lacking the kinds of off-farm opportunities
which are evident in urban settings. Nevertheless, some households are a good deal closer
to urban markets and better transportation services than others.
The 1993 Rapid Appraisal and supplemental survey demonstrated that the kinds
of BWS opportunities (market and transportation access, sales opportunities etc.) available

to more centrally-located, rural households are undisputably different from those available

 

9This upper limit intentionally differs from my Rapid Appraisal cut-off of 20,000
FRW. I wanted to compare equal numbers of HHS in this stratification yet did not want
to include less than 10% of the sample in each grouping. Had I maintained my 20,000
FRW cut-off, I feel that the stratified sample would have contained too few HHS.

16

to rural households in more remote areas. Although no two households are alike, and
precise rural/semi-rural data were not collected for this research, several qualitative BWS
distinctions were noted in 1993 between households which are nearer to and further away
from urban areas. I will therefore use the terms ’rural’ and ’semi-rural’ where appropriate
in this thesis, to emphasize the divergence of BWS opportunities in rural and semi-rural

areas.

1.7 Derived Data
1.7.1 Wine production
Wine production figures were not collected in 1990. They are therefore calculated

from the core data using the following two equations:

beer bananas produced (kgs)
+ beer bananas purchased (kgs)
+ beer bananas received as gifts (kgs)

- beer bananas given away as gifts (kgs)
- beer bananas sold (kgs)

 

BANANAS AVAILABLE FOR WINE PRODUCTION

BANANAS AVAILABLE FOR WINE PRODUCTION
— — LITERS OF BANANA WINE”.
3.4

 

Although these calculations are based exclusively on beer banana data”, a small bias

 

”Haggblade (1987) determined that 3.4 kilograms of bananas are required to
manufacture one liter of wine.

”There are three types of bananas in Rwanda - beer, cooking, and fruit bananas. As
a general rule, beer bananas are used only in the production of banana wine whereas
cooking bananas are never used in wine production. Occasionally, fruit bananas (eaten
directly, without having to be cooked) are added to banana wine for extra flavor. Because

17

may be introduced to the wine production figures since they are determined indirectly.
These wine production estimates are based on three assumptions:
1) Beer bananas are never consumed directly - all production
is either sold to other wine brewers or used to manufacture
banana wine.
2) Haggblade’s estimated conversion ratio of 3.4 kilograms of
bananas per liter of wine is accurate.
3) All wine brewers require the same average quantity of
bananas per liter of production, regardless of quality or
seasonal variation. In other words, no systematic bias

occurs in the wine-manufacturing process.

1.7.2 Household wine consumption calculation and wine purchases
Household wine consumption is calculated using the following formula (all figures

are stated in F RW value terms):

household wine production

+ wine received as a gift
+ wine purchases
- wine sales

- wine given as a gift

 

HOME CONSUMPTION OF WINE.

There may also be some bias in this figure because: 1) it includes household wine

production (already described as possibly biased), and 2) it is obtained indirectly.

 

their use is minimal, I am excluding them from my calculations.

18

1.8 Description of Rwanda

Rwanda is a small, landlocked country in central, eastern Africa. Located just
south of the equator, where little seasonal variation occurs, Rwanda has an average
temperature of 20 degrees Celsius and annual rainfall of 900-1300 mm. Rwanda’s East
zone is relatively drier and warmer than the rest of the country, with an average
temperature of 21-22 degrees Celsius and 800-1000 mm of annual rainfall. Bananas
thrive especially well in this zone, although they are grown throughout the nation.

There are two rainy seasons, one long and one short (March through May, and
October through November, respectively). Banana production is highest following these
rains, once in January and again in June and July.

Rwanda’s gross domestic product (GDP) was 193 billion F RW in 199012
(International Monetary Fund, 1993). Having a population of 7.18 million at the time
these data were collected, this indicates a 1990 GDP per capita of 26,880 FRW.

With 48% of the population under 14 years, Rwanda’s population density in 1988
was 254 inhabitants per square kilometer. Ninety percent of Rwanda’s population lives
in rural areas and is involved in agriculture. There are more than one million farms, with

more than half covering less than one hectare of land.

 

12In 1990, the official exchange rate was 120 FRW = $1.00. But because of the
market rate premium of 47.5%, 177 FRW was actually required to purchase $1.00.

19
1.9 Organization of the Study

The role of Chapters 2 and 3 is twofold. First, by describing the BWS in four
stages (Banana Production and Transactions in Chapter 2, Wine Production and
Transactions in Chapter 3), it gives the reader an overall understanding of the BWS
structure in Rwanda - its stages, participants, marketed surplus, direction and concentration
of national trade, and degree of participation by rural households. It accomplishes this
through maps, flow charts, and stratified comparisons of different groups within the
subsector. The second purpose of Chapters 2 and 3 is to underline a number of household
characteristics identified through the data as possibly influencing participation in the BWS.

The interesting characteristics identified in Chapters 2 and 3 are incorporated into
the regression model explained at the beginning of Chapter 4. Chapter 4 also presents the
results of Heckman’s two-stage procedure: four PROBIT and four OLS regressions,
which respond to the stated model by revealing the underlying determinants of
participation of rural households in the four primary stages of the BWS.

Chapter 5 discusses policy alternatives to expand the banana wine subsector, draws

methodological conclusions, and suggests related research topics for the future.

CHAPTER TWO

DESCRIPTION OF THE BANANA WINE SUBSECT OR -
BANANA PRODUCTION AND TRANSACTIONS

2.1 Introduction

The purpose of Chapters 2 and 3 is to provide a detailed portrait of the banana
wine subsector. Following a general description of the subsector, I have synthesized
subsector-wide and household-level information under the following headings: Banana
Production and Banana Transactions in Chapter 2; Wine Production and Wine
Transactions in Chapter 3.

Each section begins with a quantitative exploration of the national and zonal
significance of each activity. I include participation rates, production percentages, and
Gini coefficients to indicate the distributional impact of each activity. Each section
briefly reviews constraints facing the activity under discussion as well as a series of
qualitative observations. These observations provide interesting insights into the historical
development of the BWS and Rwandan attitudes towards it.

In the Banana Transactions section, I compare household characteristics of those
who sell some portion of their bananas with those who do not (BSs versus NBSS). I also
profile large and small wine-selling households under Wine Transactions (LWSS versus

SWSS). These profiles help to explain the household characteristics of different categories

20

21

of subsector participants. They also reveal additional determinants which may be

associated with participation levels.

 

The descriptive data of this and the following chapter allow me to make several
inferences about the BWS, which in turn enable me to recommend specific variables for

inclusion in the Chapter 4 regressions.

2.2 General Structure of the Banana Wine Subsector

Figure 2.1 shows the location of larger banana and wine markets in Rwanda, as
well as the trade within and between regions. This information was obtained from the
1993 supplemental survey. Note particularly the flow of bananas and wine moving
towards Kigali, as well as the trade routes between the East and South Central zones.

The flow charts in Figures 2.2 and 2.3 demonstrate the complexity of the subsector
in rural and semi-rural areas, for both large and small wine producers. Again obtained
from the supplemental survey, key participants and interactions are noted, excluding the
suppliers of minor wine inputs like sorghum and water'. The origin of containers, used

in the brewing and storage of wine, is also excluded from these diagramsz.

 

ISince it was learned during the interviews that sorghum is usually supplied by the
HH and only purchased in small quantities, its inclusion did not seem necessary. As for
water, it is generally transported by family members, or by hired labor as one of many
jobs for which they are paid. Like sorghum, the inclusion of water seemed irrelevant.

2Initially, it was hypothesized that containers for brewing and storing wine might
impose prohibitive start-up costs for entry to the SS. Our rapid appraisal indicated
otherwise. Wine is usually brewed in felled, dug-out tree sections, prepared by HH
members and used repeatedly. These containers impose negligible costs on HHS. As for
storage containers, vegetable oil containers are readily available - purchased at local
markets and shared between families. Container costs therefore seemed unimportant to the

22

Figure 2.1: Map of Rwanda’s banana and wine trade flows

 

 

BURUNDI

typical local markets where bananas and wine are sold alongside numerous other
commodities

- OVIBAR, the government-run banana wine facility

- exclusive banana market

- exclusive banana wine market

local wine transactions

- unidirectional banana transactions

- unidirectional wine transactions

ilk-Hue.

 

subsector’s overall flow.

23

Figure 2.2: Flow chart of the wine subsector in rural areas

 

PRODUCE BANANAS

******************************************

 

 

 

 

 

 

 

 

 

 

 

 

SMALL PRODUCERS* * LARGE PRODUCERS*
l
WHOLESALER* WHOLESALER*
transport bananas transport bananas
to local market to another region

 

 

 

 

 

 

 

H SELL TO WINE PRODUCER H

 

 

 

 

H transporter*

 

 

 

MANUFACTURE WINE

*****************************************

SMALL PRODUCER* * LARGE PRODUCER* J

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

f , I , 1 l
SELL AT SELL IN SELL INFOR- GIVE TO CONSUME AT
LOCAL HH—OWNED MALLY FROM NEIGHBORS HOME

CABARET* CABARET HOME -

 

 

 

* - activity requires labor
(at least one person
at 100 FRW/day)

 

 

 

 

 

 

,..J
H SELL TO CLIENT H

 

 

24

Figure 2.3: Flow chart of the wine subsector in semi-rural areas

 

PRODUCE BANANAS

******************************************

SMALL PRODUCERS* * LARGE PRODUCERS*

 

 

 

 

WHOLESALER*

transport bananas
to market

 

 

 

 

 

H SELL TO WINE PRODUCER H

 

 

H transporter* ll

 

 

 

 

MANUFACTURE WINE

*****************************************

 

 

 

p

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

SMALL PRODUCER* * LARGE PRODUCER* H
7 l
I , , , I . , 7 l ,

SELL AT SELL IN SELL INFOR- GIVE TO CONSUME

LOCAL HH-OWNED MALLY FROM NEIGHBORS AT HOME

CABARET* CABARET HOME 7

, , I s
- l l
SELL AT TRANSPORT SELL TO H H SELL TO
MARKET* SERVICES* URBAN H CLIENT
, - CABARET 7

 

 

 

* - activity requires labor (at least one person at 100 FRW/day)

25

In particular, note the presence of a second group of wholesalers in rural areas,
which is responsible for transporting bananas to other agroclimatic zones where the
bananas are manufactured into wine. This trade route is especially evident between the
East and South Central zones. Also note the additional market and n'ansport services
aVailable to wine traders in semi-rural areas, especially to traders in the vicinity of Kigali.

In both rural and semi-rural areas, note how small banana producers occasionally
sell a portion of their bananas as marketed surplus to wine producers. Also note how, in
both areas, large wine producers are more inclined to operate cabarets from their homes.

For the purpose of this thesis, the BWS can be thought of as divided into four
stages: Banana Production, Banana Transactions, Wine Production, and Wine
Transactions (see Figure 2.4)3. Since households can also function as wholesalers and
transporters, they are also included in this diagram and will be discussed at the end of
Chapter 3".

This simplification of the subsector into four parts is useful both for describing the
BWS and for presenting the econometric portion of the analysis. Except for transporting
and wholesaling, each stage parallels a primary research question (discussed in Section

1.4), and two regressions (one PROBIT, one OLS).

 

3Wine consumed by the household is also depicted in Figure 2.4, to emphasize the
point that all wine is not usually sold.

’This thesis will pot examine the determinants of rural HH-level participation in
wholesaling and transporting activities, although it would be interesting to focus on these
participants in future research.

26
Figure 2.4: Simplified view of the BWS options open to households

 

— NOT
PRODUCE
BEER BANANASS
HOUSEHOLD —
DECIDES To 7 SELL BANANAS
— PRODUCE
BEER BANANAS _. —SELL WINE

 

-—-PRODUCE
BANANA WINE -——

 

-—-CONSUME WINE
PRODUCED AT HOME

| ANY HOUSEHOLD CAN'ALSO SERVE AS A I

 

I WHOLESALER OR TRANSPORTER I

2.3 Banana Production
2.3.1 National and zonal findings
Eighty-nine percent of Rwanda’s households produced bananas in 1990. The
highest share of banana producers was located in the East, with decreasing shares
westward across the country. Although Rwanda’s climatic conditions encourage banana
production throughout the nation, her decreasing elevation eastward across the country has
endowed the East zone with slightly drier and warmer conditions which are especially

favorable to the production of bananas. Thus, more farmers are inclined to grow bananas.

 

’Households who choose not to produce bananas often purchase wine, such that they
participate in the subsector as wine buyers. Only one household who did not produce
bananas purchased bananas for home production of wine. In either case, this portion of
the diagram represents at most 10% of the population. Since this paper is primarily
interested in the 90% who participate in the BWS by growing bananas and/or producing
wine, this sub-group of households is considered outside the realm of my research.

27

Table 2.1 shows the share of producers located in each of Rwanda’s agroclimatic zones.

A high of 29% lived in the East; a low of 9% lived in the Northwest.

Table 2.1: Banana producers by zone (%)

 

 

 

 

 

AGROCLIMATIC ZONE PERCENT OF BANANA PRODUCERS
NORTHWEST 8.7

SOUTHWEST 15.1

NORTH CENTRAL 25.3

SOUTH CENTRAL 22.4

EAST 28.5

TOTAL 100.0

(N) (757)

)6, Significant at < .0000]

 

‘ Source: DSA HouseFola Survey, I530, Ki—gzr, Rwanda _

Likewise, the East zone was responsible for the largest share of banana production
value (3 8%); share of production value decreased westward. These findings are again
attributable to the increasingly favorable agroclimatic conditions for bananas eastward
across Rwanda. AS shown in Table 2.2, the North and South Central zones were jointly
responsible for 43% of total value, and the Northwest and Southwest zones for a
cumulative 19%. Among banana growers, the East zone’s household mean for banana
production was highest (9550 FRW), followed by the Northwest (7710 FRW).

As Share of total household income, banana production was a fairly steady
component among banana growers throughout Rwanda. It ranged between 15% of

income among banana producers in the North Central zone and 19% in the South Central

28

zone (see Table 2.3). The North Central zone’s lower share is probably attributable to
the dominance of other crops in this zone (such as yarns and beans), which result in
relatively high household incomes. In the South Central zone, however, where mean
household incomes are generally lower and major cash crops not as prevalent, the
importance of banana production increases. In the East zone, where incomes are highest

of all, banana production (although high) remains intermediate as a share of income.

Table 2.2: Banana production by zone (F RW)*

 

 

AGROCLIMATIC PERCENT OF I-H-I MEAN
ZONE BANANA (FRW)
PRODUCTION
NORTHWEST 9.4 771 l
SOUTHWEST 9.4 4463
NORTH CENTRAL 23.4 6601
SOUTH CENTRAL 19.7 6268
EAST 38.1 9546
TOTAL 100.0 7140
(N) (757)
F, significant at < .0000]

 

 

 

 

 

 

 

Source: DSA Household Survey, $30, Kigfir, Rwanda

* - Data refer to banana producers only

Banana production value was most concentrated (among a small number of
growers) in Rwanda’s western zones, probably the result of fewer households with larger
landholdings. For example, 20% of zonal growers grew nearly 70% of the banana

production value in the Northwest zone (Gini of .60), compared to 47% in the East (Gini

29

of .42). These values were respectively higher and lower than the national Gini of .47
(see Table 2.4). Households in the East zone typically farmed 28% more land than
farmers in the Northwest zone (83 versus 50 ares), although banana growers in each zone
respectively farmed 34% and 45% more land than their non-banana growing counterparts.
In other words, larger landholdings appear to encourage banana production. Households
in the East have generally larger landholdings, thus distributing banana production more

evenly among its zonal residents.

Table 2.3: Banana production as share of household income (FRW)*

 

AGROCLIMATIC ZONE BANANA PRODUCTION AS
MEAN PERCENT
OF HOUSEHOLD INCOME

 

NORTHWEST 14.9
SOUTHWEST 14.6
NORTH CENTRAL 14.5
SOUTH CENTRAL l 8.9
EAST 16. 1
RWANDA 1 6.0
(N) (757)

F, significant at < .001

Source: DSA Housefiolfi Survey, $33, Egan Rwana:

* - Data refer to banana producers only

 

 

 

Consistent with this uneven distribution is the finding that the Northwest zone also
exhibited the highest share of households with zero banana production (46%). I attribute

this finding to the fact that bananas do not grow well at high altitudes. Thus, households

30

in the Northwest zone are preoccupied by crops better suited to the Northwest’s climate.
In addition, farming land which is more constrained than in other zones, only a small
number of households have moved into banana production seriously, foregoing other crops
to do so. Table 2.5 examines zonal banana production by value category. Contrastingly,
only 1% of the East zone’s population produced no bananas; 88% produced more than

2000 FRW.

Table 2.4: Zonal GINIS for banana production by banana producer quintile (FRW)"'

 

 

 

 

 

 

 

 

 

 

AGROCLIMATIC GINI % BANANA PRODUCTION BY
ZONE COEFFICIENT BANANA PRODUCER QUINTILE
TOP BOTTOM
QUINTILE QUINTILE
NORTHWEST .60 69 .6
SOUTHWEST .55 60 2
NORTH CENTRAL .43 47 3
SOUTH CENTRAL .44 51 4
EAST .42 47 4
RWANDA .47 53 2
(N) (757) ..
ource: DSA Housefiolfi Survey, , g 1, wanda
* - F-test significant at <.00001, based on production means across all five zones by

producer quintile.

2.3.2 Constraints to banana production
Our supplemental survey shows that banana production is perceived as having
decreased from 1990 to 1993. The most frequently cited reasons are lack of rain and

increased disease among the banana trees. At the time of this supplemental survey, only

31

one respondent cited the war as a reason for decreased production. As noted in

Habimfura and Miklavcic’s supplemental survey conclusions (1995), lack of inputs

Table 2.5: Zonal banana production by value category (FRW)*

 

 

 

 

 

 

 

 

 

 

 

NO BANANA PRODUCTION 1-250 FRW 251-500 FRW
MEAN % OF ZONE MEAN % OF ZONE MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST .00 46.2% 128.08 6.9% 421.53 3.7%
SOUTHWEST .00 16.5% 188.69 5.4% 442.89 7.4%
NORTH CENTRAL .00 4.4% 144.91 2.6% 374.13 2.5%
SOUTH CENTRAL .00 3.0% 96.39 .7% 344.79 2.5%
EAST .00 1.2% 247.37 .2% 470.84 .6%
501-1000 FRW 1001-2000 FRW > 2000 FRW
MEAN % OF ZONE MEAN % OF ZONE MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST 865.92 3.2% 1552.05 3.7% 11126.93 36.3%
SOUTHWEST 767.27 12.8% 1445.64 12.2% 7459.50 45.7%
NORTH CENTRAL 831.49 3.9% 1662.03 12.7% 8197.81 73.8%
SOUTH CENTRAL 722.39 4.1% 1518.25 12.9% 7603.65 76.9%
EAST 701.45 2.8% 1452.79 7.3% 10580.37 87.9%
ZONAL MEAN
MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST 4148.91 100.0%
SOUTHWEST 3726.04 100.0%
NORTH CENTRAL 6311.27 100.0%
SOUTH CENTRAL 6081.08 100.0%
EAST 9429.72 100.0%

 

DSA Household Survey, 1990, Kigali, Rwanda

* - F-test significant at <.00001.

and small landholdings were also cited as reasons, as were a number of lesser constraints,
including inadequate training about how to best grow bananas, insufficient funds for

investing in banana maintenance, and a preference for other crops“.

 

6In 1990, the seven most preferred crops, in descending order and in addition to
bananas, were yams, beans, coffee, sorghum, potatos, maize, and manioc. Beans, coffee,

32

2.3.3 Qualitative observations

Many of the 35 households interviewed in the supplemental survey had obtained
their first banana trees from the State, although many others had either inherited the trees
or inherited the land on which the trees now grow. Despite the constraints cited above,
most of the households said they would like to grow more bananas, although more than
50% said this would be difficult because of land constraints7. Half would like to see
their children continue in banana and wine production; half would not. Rather, they
prefer that their children pursue income-generating activities outside of the BWS.

Although a small sample was used to acquire these observations, important
indicators of household attitudes towards the BWS are revealed. Obtained from primarily
large wine sellers, who participate in all stages of the subsector, I cannot help but
conclude that these attitudes are among the most positive (towards the BWS) which we
would have encountered anywhere in Rwanda. It is also plausible, however, that this
group is the most income-oriented because of their ME experience, such that they are
more likely to seek activities for their children which exceed income from BWS

participation.

 

sorghum, and manioc are dominant in the East; potatos and maize in the Northwest; and
yarns in the Central zones. Two exceptions, however, are that beans are also partial to
the North Central zone and coffee is partial to the Northwest. Forty-two percent of all
sorghum production is sold, compared to 10% of manioc, and 7% and under for all other
crops mentioned (excluding coffee).

7Twenty-five percent made a special point of stressing their concern about land
constraints which will undoubtedly worsen as land is divided among more and more
children.

33

2.3.4 Data patterns and inferences from them

Based on the above findings, agroclimatic location appears to be an important
determinant of banana production in Rwanda. As the Northwest results indicate, this
seems particularly relevant to participation levels.

Financial and land assets also appear to be determinants since insufficient assets
were repeatedly cited as constraining a household’s ability to participate in banana
production - either by inhibiting a household’s ability to finance proper maintenance or
through insufficient land for expansion. Although disease was cited as a primary reason
for decreased production in recent years, I do not have disease or time series production
data with which to explore this possibility. I do, however, have access to rainfall and
altitude statistics.

It may also be worthwhile to examine the impact of three other constraints - input

costs, other crops grown, and other sources of non-farm income.

2.4 Banana Transactions
2.4.1 National and zonal findings
2.4.1.1 SALES
Twenty-four percent of households who produced bananas in 1990 participated in
banana sales, yet less than 3% of Rwanda’s total banana production value was sold.
Among producers, the South Central zone had the highest percentage of Rwanda’s sellers
(28%), followed closely by the East zone (see Table 2.6). The smallest percentage of

sellers was in the Northwest (10%). These findings tell us that in zones where households

34

produce the smallest quantities of bananas, households are more inclined to sell a portion
of what they produce - probably because they do not have enough bananas, at any given
time, to manufacture a reasonable quantity of wine. In other words, many households are
selling smaller quantities of bananas. Also, our supplemental survey pointed to significant
banana markets around Kigali (on the border between the East and North Central zones)
and to a lesser extent Butare (in the South Central zone). The increased demand for

bananas initiated by these markets undoubtedly encourages banana sales.

Table 2.6: Banana sellers by zone (%)

 

 

AGROCLIMATIC ZONE PERCENT OF BANANA SELLERS ll
NORTHWEST 10.0
SOUTHWEST 21.0
NORTH CENTRAL 146
SOUTH CENTRAL 27.9
EAST 26.5
TOTAL 100.0
(N) (134)
)8, significant at < .001

 

 

 

 

Source: DSA Housefiold Survey, ISSS, Egan Rwanda

Nevertheless, the share of Rwanda’s total banana sales value was highest in the
East zone, followed by the Northwest zone. All other zones sold less than half the East
zone’s amount (see Table 2.7). In the East, where more households produce large
quantities of bananas than in other zones, there is a greater surplus of bananas over and

above that required by households to manufacture wine on a regular basis. This banana

35

surplus is therefore sold, both locally and also through truck-owning wholesalers

markets in both Kigali and Butare.

Table 2.7: Banana sales by zone (FRW)"'

t0

 

AGROCLIMATIC

 

 

 

 

 

* - Data refer to banana sellers only.

 

PERCENT OF HH MEAN

- ZONE BANANA SALES (FRW)
NORTHWEST 28.0 2356
SOUTHWEST 14.7 587
NORTH CENTRAL 9.0 518
SOUTH CENTRAL 15.9 478
EAST 32.4 1025
TOTAL 100.0 839
(N) (184)

F, significant at < .00001

 

 

Source: DSA Household Survey, ISSS, Egan Rwanda

Banana sales as a percentage of household income was small for banana sellers

throughout Rwanda. This percentage was between 2% and 3% in all zones but the

Northwest8 (see Table 2.8), thus reinforcing my observations that banana sales play a

relatively minor role in Rwanda’s economy.

 

8Banana and wine sales were initially calculated as percentages of both Loveridge’s
and Kangasniemi’s definitions of HH income. Loveridge’s definition (1990) netted out
purchased inputs; Kangasnierni’s definition (1994) did not. Banana sale means per zone,
as percentages of income, were identical for each definition. Maximum value percentages,
however, were slightly lower using the Kangasniemi definition. Kangasniemi’s HH
income definition will be used throughout the rest of this document.

36

Mean household revenues earned by banana sellers in 1990 were well-distributed
nationally. As the correlations in Table 2.9 Show, the top quintile of sellers did not earn
a disproportionate share of income. In fact, the opposite may actually be true. In the
Northwest zone, a negative correlation was calculated, implying that the lowest quintile
of sellers (those who sold the least from among banana sellers) consisted of the wealthiest
households. It can therefore be hypothesized that richer households are less inclined to
sell bananas on the market, probably because they produce adequate quantities of bananas
with which to manufacture wine and because they do not need the intermittent income
from bananas to finance unforeseen expenses. Nevertheless, it Should be kept in mind

that these correlation results are statistically insignificant.

Table 2.8: Banana sales as share of household income (%)*

 

 

 

 

 

 

 

 

 

AGROCLIMATIC ZONE BANANA SALES AS
MEAN PERCENT
OF HOUSEHOLD INCOME
NORTHWEST 6.3
SOUTHWEST 2.7
NORTH CENTRAL 1.9
SOUTH CENTRAL 2.5
EAST 2.2
RWANDA 2.8
(N) (134)
F, significant at < .001
‘ Source: DSA Household Survey, ISSS, King, Rwanda

* - Data refer to banana sellers only

37

Table 2.9: Zonal correlations of HH revenues and banana seller quintiles (FRW)"‘

 

 

 

 

 

 

 

 

 

 

AGROCLIMATIC CORRELATION % HOUSEHOLD REVENUES BY
ZONE BANANA SELLER QUINTILE
TOP BOTTOM
QUINTILE QUINTILE
NORTHWEST -.25 13 44
SOUTHWEST .09 29 10
NORTH CENTRAL .04 23 18
SOUTH CENTRAL -.06 19 23
EAST .12 35 19
RWANDA .11 29 17
(N) (134)
Source: DSA Household Survey, , g r, wanda _____ T
* - F-test insignificant, based on revenue means across all five zones by seller
quintile.

Banana sales were least concentrated among banana sellers in the North Central
zone (Gini is .34) and most concentrated in the East (Gini is .53) (see Table 2.10). In
other words, sales were more evenly distributed across sellers in the North Central zone,
unlike in the East zone where the same percentage of sellers were responsible for a much
larger share. This finding is unusual in that banana production and number of sellers are
both higher in the East, meaning that I would have expected the opposite to be true. I
can explain my frndings only by claiming that the North Central zone is preoccupied with
crops other than bananas.

Although statistically insignificant, the correlation between banana sales and
banana producer quintile is highest in the Northwest zone (see Table 2.11). In other

words, a disproportionate share of banana sales are made by this zone’s largest banana

38

producers. It is also interesting to note that although the Northwest zone produced only
9% of Rwanda’s production value of bananas, it sold 28% of the nation’s total. The East
zone, on the other hand, representing 38% of Rwanda’s banana production value, sold
only Slightly more at 32%. Banana sales are indeed concentrated in the Northwest.

The Northwest zone also sold the highest share of bananas produced (see Table
2.12). Among banana sellers, 35% of the Northwest zone’s household production value
was sold; compared to 17% in the South Central zone’. This finding may be attributable
to a number of larger wine manufacturers in the Northwest zone, who make it worthwhile
for banana producers to part with bananas on a more regular basis. For example, I visited
a very large women’s cooperative in Gisenyi, who weekly purchased large quantities of
bananas from nearby suppliers. They, in turn, manufactured both beer and wine from
bananas, for distribution throughout Rwanda. This cooperative claimed to always have
a steady supply of bananas, with their only constraint the availability of containers in
which to sell their product.

The producer price of bananas was lowest in the East zone and highest in the
Southwest (3.8 versus 5.5 FRW/KG) (see Table 2.13). Expectedly, the price was lowest
where the influx of bananas to specialized banana markets was greatest - and highest

where banana availability is lowest of all.

 

9Upon averaging in all banana growers, including those who sell no bananas, these
numbers drop to 4% and 2% respectively.

39

Table 2.10: Zonal GINIS for banana sales by banana seller quintile (FRW)”‘

 

 

 

 

 

 

 

 

 

AGROCLIMATIC GINI % BANANA SALES BY BANANA
ZONE COEFFICIENT SELLER QUINTILE
TOP BOTTOM
QUINTILE QUINTILE
NORTHWEST .51 50 2
SOUTHWEST .51 59 3
NORTH CENTRAL .34 40 5
SOUTH CENTRAL .52 63 3
EAST .53 60 3
RWANDA .56 64 2
(N) (184) 4
Source: DSA HouseFrold Survey, ISSS, Kigfir, Ewan: '-
* - F-test significant at <.00001, based on banana sale means across all five zones by

banana seller quintile.

Table 2.11: Zonal correlations of banana sales and banana producer quintiles (FRW)”’

 

 

 

 

 

 

 

 

 

 

 

AGROCLIMATIC GINI % BANANA SALES BY BANANA
ZONE COEFFICIENT PRODUCER QUINTILE
TOP BOTTOM
QUINTILE QUINTILE
NORTHWEST .47 43 0
SOUTHWEST .21 19 7
NORTH CENTRAL -.24 8 24
SOUTH CENTRAL .01 6 20
EAST .11 22 19
RWANDA .17 28 9
(N) (757)
ource: DSA Housel'iold Survey, ISSS, Kigar, Rwanda
* - F-test insignificant, based on banana sale means across all five zones by banana

producer quintile.

40
Table 2.12: Marketed surplus of banana production (%)*

 

 

AGROCLIMATIC ZONE BANANA SALES AS
MEAN PERCENT

OF BANANA PRODUCTION

 

NORTHWEST 34.5
SOUTHWEST 28.3
NORTH CENTRAL 22.9
SOUTH CENTRAL 17.2
EAST 21.0
RWANDA 23.1
(N) (184)

p F, significant at < .05
A Source: DSA Housefiold Survey, ISSS, Egan Rwanda

 

 

 

 

* - Data refer to banana sellers only.

Table 2.13: Producer price of bananas by zone (FRW/KG)*

 

 

 

 

AGROCLIMATIC ZONE PRODUCER PRICE OF BANANAS
(FRW/KG)
NORTHWEST 4.58
SOUTHWEST 5.48
NORTH CENTRAL 4.70
SOUTH CENTRAL 3.97
EAST 3.84
RWANDA 4.42
(N) (184)
F, significant at < .00001

 

 

 

; Source: DSA Household Survey, ISSS, figzr, Rwanda

* — Data refer to banana sellers only.

41

In closing, as shown in Table 2.14, banana sales are insignificant to a majority of
Rwanda’s population. Between 71% and 87% of the households in each zone did not
participate in banana sales, and only a small fraction of those who did earned more than
2000 FRW in 1990. Nevertheless, by comparing this table to Table 2.5, I can Show that
in the East, although 88% of the households grew a lot of bananas, only 3% sold a lot.
In the Northwest zone, however, where less than half this percentage of households grew

a lot of bananas, 6% (or twice as many) sold a lot.

Table 2.14: Zonal banana sales by value category (FRW)*

 

 

 

 

 

 

 

 

 

 

 

NO BANANA SALES 1-250 FRW 250-500 FRW
MEAN % OF ZONE MEAN % OF ZONE MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST .00 85.0% 137.55 2.7% 460.00 .5%
SOUTHWEST .00 71.7% 151.78 12.6% 319.71 6.2%
NORTH CENTRAL .00 86.6% 145.50 3.1% 408.15 6.9%
SOUTH CENTRAL .00 70.5% 148.53 15.4% 342.71 8.1%
EAST .00 77.6% 139.35 6.0% 395.42 5.8%
501-1000 FRW 1001-2000 FRW > 2000 FRW
MEAN % OF ZONE MEAN % OF ZONE MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST 777.88 2.1% 1492.39 3.9% 4725.87 5.8%
SOUTHWEST 641.55 3.1% 1439.35 4.8% 2351.55 1.6%
NORTH CENTRAL 799.94 1.2% 1195.50 2.3% . .0%
SOUTH CENTRAL 661.62 2.8% 1496.90 2.0% 3354.98 1.3%
EAST 818.66 4.9% 1618.43 3.1% 4258.61 2.5%
ZONAL MEAN
MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST 352.39 100.0%
SOUTHWEST 166.01 100.0%
NORTH CENTRAL 69.69 100.0%
SOUTH CENTRAL 140.83 100.0%
EAST 229.58 100.0%

 

DSA Household Survey, 1990, Kigali, Rwanda

* F-test significant at < .00001

42

2.4.1.2 PURCHASES

Almost all households use their own bananas to make wine and therefore purchase
bananas only infrequently. In fact, only 10% of rural Rwandan households purchased any
bananas in 1990. The greatest share of buyers was located in the South Central zone.
Although the East zone was a close second, Table 2.15 Shows that the western and North
Central zones had approximately half this share of buyers. The banana trade is
undoubtedly highest in the South Central zone because of the influx of bananas to the
markets at Kigali and Butare. In other words, the greatest number of purchases can be
expected to take place here. On the other hand, because banana markets and wine

production are more limited in the other zones, the demand for bananas is negatively

 

 

 

impacted.

Table 2.15: Banana buyers by zone (%)*
AGROCLIMATIC ZONE PERCENT OF BANANA BUYERS
NORTHWEST 17.5
SOUTHWEST 13.4
NORTH CENTRAL 15.0
SOUTH CENTRAL 29.5
EAST 24.6
TOTAL 100.0
(N) (90)

)6, significant at < .10

 

 

 

Source: DSA Housefiold Survey, 1553, mgzr, Rwanda 7

* - Data refer to banana buyers only.

43

Forty percent of the value of Rwanda’s banana purchases (in FRW) were made in
the Northwest zone. Although I do not have data available about the origin of the
bananas purchased (i.e., are bananas provided locally or do they come from another
zone?), the most likely explanation for this Northwest finding is that wine manufacturers
in this zone are not self-sufficient in bananas, unlike the East and North Central zones
where they are (see Table 2.16). As stated earlier, this is because the Northwest’s high

altitude makes it difficult to grow bananas.

Table 2.16: Banana purchases by zone (FRW)*

 

 

AGROCLIMATIC PERCENT OF HH MEAN

ZONE BANANA PURCHASES (FRW)
NORTHWEST 39.7 4967
SOUTHWEST 20.8 3408
NORTH CENTRAL 9.9 1449
SOUTH CENTRAL 17.6 1307
EAST 12.0 1067
TOTAL 100.0 2191
(N) (90)

 

 

 

 

 

 

* - Data refer to banana buyers only.

The majority of all banana purchases were made by a fairly concentrated number
of buyers (national Gini equalled .61). As shown in Table 2.17, the top 20% of buyers

purchased 68% of beer bananas. Purchases were similarly concentrated in each zone,

44

implying wine manufacturers in each zone who supplement their own levels of banana

production with market purchases.

Table 2.17: Zonal GINIS for banana purchases by banana buyer quintile (F RW)*

 

 

 

 

 

 

 

 

 

AGROCLIMATIC GINI % BANANA PURCHASES BY
ZONE COEFFICIENT BANANA BUYER QUINTILE
TOP BOTTOM
QUINTILE QUINTILE
NORTHWEST .52 53 1
SOUTHWEST .63 40 1
NORTH CENTRAL .52 59 2
SOUTH CENTRAL .47 53 2
EAST .57 65 3
RWANDA .61 68 l
(N) (90)
ource: DSA Housefiold Survey, , g r, wanda
* - F-test significant at .0000], based on banana sale means across all five zones by

banana producer quintile.

2.4.2 Constraints to banana transactions
The primary constraint to banana sales, as identified in the supplemental survey,
is the custom of using one’s home-grown bananas to produce banana wine. In other
words, by conforming to tradition, households are inhibited from selling otherwise surplus
bananas.
In addition, because bananas are fragile, survey participants claimed they are

difficult to transport to market or to be traded interzonally. Poor road systems and a

45

limited market for banana wine were also cited as constraints. The limited banana wine

market presumably inhibits the need for bananas by other households.

2.4.3 Qualitative observations

A majority of the households with whom we spoke during the supplemental survey
declared that bananas are an excellent and frequently relied upon source of rapid, small
income. Like a miniature savings account, the fruit of a banana tree can be cut down and
sold to cover small, unforeseen costs (such as those incurred by unexpected visits to the
local health clinic).

In other words, the intention behind intermittent banana sales differs from that
behind the sale of other crops. Whereas other crop income, like coffee, is used for
investment purposes (i.e., schooling, small business start-up costs) or to cover larger
expenditures (i.e., house repairs), banana sales are a more versatile source of income -

used in case of emergency.

2.4.4 Banana sellers versus non-banana sellers
Interesting distinctions are noted between core survey participants who produced
bananas in 1990, but who then decided to sell or not sell bananas. Below, I will describe
a number of traits which characterize these two populations. In particular, I will focus
on distinctions in levels of participation in the BWS, household characteristics, and

descriptions of the household head.

47

Table 2.18: BWS characteristics of BSs and NBSS (N=757)*

 

 

 

 

 

* - Data refer to banana producers only.

 

 

 

BWS TRAIT BANANA NON- F-TEST
(HH MEAN IN FRW) SELLERS BANANA SIGNIFICANCE
(N = 184) SELLERS
(N = 573)
1 WI
BANANA PRODUCTION 4880 7868 <.00001
BANANA SALES 839 0 <.00001
BANANA PURCHASES 252 263 insignificant
BANANA GIFTS GIVEN 29 38 insignificant
BANANA GIFTS 5 34 insignificant
RECEIVED
WINE PRODUCTION 8386 15995 <.00001
WINE SALES 3037 5916 <.00001
WINE PURCHASES 667 300 insignificant
WINE GIFTS GIVEN 429 1244 <.00001
WINE GIFTS RECEIVED 172 566 <.01
Source: ESE Housefiold Survey, ISSS, R151, Rwanda

 

Last, wine purchases were greater among BSS, probably because these households
have the same taste for wine yet have less wine available at home to consume. This

finding, however, is not statistically significant.

48
2.4.4.2 HOUSEHOLD CHARACTERISTICS

Although not statistically significant, NBSS had non-BWS household revenues 20%
higher in 1990 than BSs (see Table 2.19). This was 12% higher per adult equivalent
(AE) (11822 F RW versus 10583 FRW). These increased revenues were likely attributable
to the statistically Significant fact that NBSS had 41% more cultivated land (20% more
per AE, or 16.3 versus 13.6 ares per AE), thus earning 52% more from other agricultural
crops. In turn, these increased revenues are probably what made it possible for NBSS to
consume at a rate 32% higher than BSS and to purchase nearly twice as many inputs (a

majority of which was for hired labor - 99% and 98% of respective input expenditures).

2.4.4.3 DEMOGRAPHIC CHARACTERISTICS
The following characteristics apply to the household head and not to other
members of the family. This is based on my assumption that the household head is the
primary economic decision-maker in the family. Demographic information about the
average household head in each group is Shown in Table 2.20. No demographic
characteristics were determined significantly different between the two categories of

household heads.

49

Table 2.19: Household characteristics of BSs and NBSS (N=757)*

 

 

 

 

BWS TRAIT BANANA NON- F-TEST
(HH MEAN) SELLERS BANANA SIGNIFICANCE
(N = 184) SELLERS
(N = 573)
l —- ——-——--————-—————— r
NON-BWS REVENUES 37242 44621 insignificant
(FRW)
NON-BWS REVENUES 10583 11822 insignificant
PER AE (FRW)
CULTIVATED LAND 49 69 <.00001
(ARES)
CULTIVATED LAND 13.6 16.3 <.01
PER AE (ARES)
INPUT PURCHASES 1652 3061 <.05
(FRW)
LABOR PURCHASES 1628 2983 <.05
(FRW)
LABOR SALES (FRW) 13218 12433 insignificant
SALES FROM AG 6961 10575
PRODUCTION (FRW) <.001
HOME CONSUMPTION 20217 26752 <.001
(FRW)
ADULT EQUIVALENTS 4.1 4.5 <.05
IN HOUSEHOLD (#)

 

 

 

 

 

 

- ' it
ource: I ‘ 'ouse o . urvey, ISSS, KigSI-r, Rwanda

* - Data refer to banana-producing households only.

50

Table 2.20: Demographic characteristics of BSs and NBSS (N=757)*

 

 

 

 

 

 

 

HOUSEHOLD BANANA SELLERS NON-BANANA SELLERS
HEAD TRAITS (N = 184) (N = 573)
GENDER male 81 .9% male 7 8.3%
female 1 8.1% female 21 .7%
AGE mean = 45.2 mean = 46.3
MARITAL married 59.9% married 65.2%
STATUS widowed 14.5% widowed 17.6%
divorced 1.3% divorced --
separated 4.2% separated 2.3%
live together 18.8% live together 13.2%
single 1 .3% single 1 .9%
EDUCATION none 54.0% none 56.4%
some primary 32.6% some primary 30.8%
all primary 9.6% all primary 10.2%
post-primary 0.7% post-primary 1 . 1%
some secondary 2.6% some secondary 0.6%
all secondary 0.5% all secondary 0.9%
PRINCIPAL farmer 86.1% farmer 89.9%
ACTIVITY paid farm worker 0.7% paid farm worker 1.2%
paid laborer 4.2% paid laborer 0.8%
ind. artisan 3.0% ind. artisan 1.8%
salaried artisan 0.7% salaried artisan 1.3%
commerce 0.9% commerce 0.7%
civil servant 2.6% civil servant 2.1%
domestic help ~ -- domestic help 0.2%
student 0.4% student --
other 1 .4% other 0.5%
unemployed -- unemployed 1 .4%
ource: -—*——use—urvey¥;1‘"H , g r, wanda

*

significant.

 

 

All households produced bananas; no demographic interactions are statistically

 

51

2.4.5 Data patterns and inferences from them

Similar to banana production, agroclimatic location seems to be a relevant
influence on banana sales. Likewise, total household revenues, quantity of cultivated land,
and other crops grown appear to influence sales, as does the price of bananas.

The comparison of BSs with NBSS elicited a number of other potential
determinants, many of which may also influence (or be influenced by) the decision to
grow bananas. Among these are banana production, demand for banana wine, wine
production and sales, income from other crops, home consumption in 1990, input
purchases, and labor sales. No demographic characteristics related the household head

are implicated.

2.5 Conclusion

This chapter enables me to draw four conclusions about participation in banana
activities.

First, banana production is highest and most evenly distributed in the East; it is
more concentrated elsewhere in the country, especially in the Northwest zone.

Second, sales and purchases of bananas are very small throughout Rwanda, and
are mainly in areas outside the East zone. With the exception of Kigali and Butare, there
is little to no interzonal trade of bananas.

Third, banana sales are very concentrated.

Fourth, and most interestingly, while one would think that bigger and richer

farmers would sell more bananas, it is really the smaller and poorer farmers that do,

52

primarily because bananas provide a quick and easy source of small income.

CHAPTER THREE

DESCRIPTION OF THE BANANA WINE SUBSECTOR -
WINE PRODUCTION AND TRANSACTIONS

3.1 Introduction

This chapter on Wine Production and Wine Transactions follows the same format
as Chapter 2. Nevertheless, there are two additions.

First, following a subsectoral description of wine activities, I have included a
section which describes Kigali’s unusual market structure for wine. At the time of our
supplemental survey, this market was both active and growing. AS Rwanda’s population
continues to grow and urban areas expand in response to land constraints throughout the
country, I expect that the demand for purchased wine will also grow. This expectation
derives from Minot’s finding, using mid-1980’s data, that the urban sector’s level of
expenditure is 2.4 times that of the rural sector (1992). In other words, urban households
which are accustomed to consuming banana wine yet no longer able to manufacture q
adequate quantities on their own will eventually encourage urban centers to imitate the
trading patterns presently found in Kigali. This expectation warrants a careful explanation
of Kigali’s wine market.

Second, a description of the BWS is incomplete without some mention of the

wholesalers and transporters who participate. Although the emphasis of this thesis is on

53

54

banana and wine production and sales, and the household-level decision-making processes
involved in these activities, many households contribute to the subsector by serving as
transporters and wholesalers. Since the organizational structure of these downstream
services affects upstream levels of production and sales, I have chosen to include them
in this chapter.

The above-added sections are qualitative. Their content is drawn from our 1993

supplemental survey findings.

3.2 Wine Production
3.2.1 National and zonal findings

All households that produced bananas in 1990 manufactured banana wine. This
was 89% of all sample households‘. The highest percentage of wine producers was in
the East zone with decreasing shares westward across the country. Since a relatively
small proportion of bananas are ever transported from the East to the South Central zone
(remember that two constraints to sales are banana fragility and generally inadequate
transportation services), it should not be surprising that the percentage of wine producers
is highest where banana producers are most numerous. Having already shown that only
a small percentage of bananas are ever sold, I can now state that households have but one
alternative - to manufacture their bananas into wine. Table 3.1 shows the share of

producers located in each of Rwanda’s agroclimatic zones.

 

|This is one more than the number of sample HHS that produced bananas, implying
that the bananas used by this additional HH to manufacture wine were either received as
gifts or purchased.

55
Table 3.1: Wine producers by zone (%)

 

 

 

AGROCLIMATIC ZONE PERCENT OF WINE PRODUCERS
NORTHWEST g7

SOUTHWEST 150

NORTH CENTRAL 253

SOUTH CENTRAL 22.5

EAST 28.5

TOTAL 1000

(N) (758)

)6, significant at < .00001

 

 

 

 

Source: ES: Housefiold Survey, I555, Elgar, Rwanda

As shown in Table 3.2, the East zone led wine production at 39% of total value.
Similar to our banana production findings, share of wine production value decreased
westward. The increasingly favorable agroclimatic conditions and landholdings, which
were beneficial to the growth of bananas in the East zone, in turn promoted the
manufacturing of wine. The total value of wine production in 1990 (16 billion FRW),
throughout Rwanda, was nearly twice the production value of bananasz. In part, this
added value of 100% explains the nation-wide interest of banana growers in
manufacturing wine. The dominant role of wine in Rwandan culture also justifies
production levels.

As share of total income in wine-producing households, wine production ranged

between an imputed value of 24% in the Northwest zone and 43% in the South Central

 

2As before, we are talking about the production value of beer bananas - exclusive of
fruit and cooking bananas. All beer bananas are assumed to be sold to wine producers
or manufactured into wine by the grower. They are never consumed directly.

56
zone. This figure was 32% nationally (see Table 3.3). Among wine producers, the South

Central zone had the second highest household production mean (the East zone was first),
followed closely by all but the Southwest zone (see Table 3.2). AS in the description of
banana production, these findings are attributable to the fact that few lucrative crops
contribute to the South Central zone’s mean household income, unlike in the East and
North Central zones where climate and landholdings are more amenable to cash crop
growth. Thus, even though the mean household production of wine was slightly higher
among wine producers in the South Central zone, this higher mean only partially explains

the high percentage of total household income attributable to wine production.

Table 3.2: Wine production by zone (FRW)*

 

 

 

 

   

flu—E a:
AGROCLIMATIC PERCENT OF WINE HH MEAN

ZONE PRODUCTION (F RW)
NORTHWEST 7.6 12309
SOUTHWEST 8.8 8249
NORTH CENTRAL 21.9 12266
SOUTH CENTRAL 22.9 14353
EAST 38.8 19241
TOTAL 100.0 14120
(N) (758)

F, significant at < .0000]

 

 

 

  

 

 

Source: SSA Household Survey, I55 , g 1,

* - Data refer to wine-producing households only.

57

Similar to banana production, wine production was most concentrated over
households in Rwanda’s western zones. As Shown by the Gini coefficients in Table 3.4,
20% of the wine producers in the Northwest and Southwest zones respectively produced
64% and 76% of zonal production value (Ginis were .60 and .57). By contrast, 20%
produced only 49% of production value in the North Central zone. Since the national
Gini of .48 was found to be the same as that for banana production, I conclude that
bananas are indeed traded only minimally in Rwanda. Wine is produced by the same

people who grow bananas, and in roughly the same proportions.

Table 3.3: Wine production as share of household income (FRW)"'

 

 

 

 

 

= : “"*"—*—‘
AGROCLIMATIC ZONE WINE PRODUCTION AS
MEAN PERCENT
OF HOUSEHOLD INCOME

NORTHWEST 23.6
SOUTHWEST 26.6

NORTH CENTRAL 27.1

SOUTH CENTRAL 42.9

EAST 32.4 "
RWANDA 31.8

(N) (758)

 

 

 

 

* - Data refer to wine producers only

58

Table 3.4: Zonal GINIS for wine production by
wine producer quintile (FRW)”'

 

 

 

 

 

 

 

 

AGROCLIMATIC GINI % WINE PRODUCTION BY WINE
ZONE COEFFICIENT PRODUCER QUINTILE

TOP BOTTOM

QUINTILE QUINTILE
NORTHWEST .60 64 2
SOUTHWEST .57 76 1
NORTH CENTRAL .44 54 3
SOUTH CENTRAL .44 52 3
EAST .42 49 2
RWANDA .48 57 2

(N) (758)
ource: DSA Household Survey, I555, agar, Rwanda
* - F-test significant at <.00001, based on wine production means across all five zones

by wine producer quintile.

AS Table 3.5 shows, the numbers of households with higher levels of wine
production increased as one moved eastward in 1990, with the greatest percentage of
larger producers in the central and eastern zones. Although the overall household mean
was Slightly higher and production slightly more concentrated among wine producers in
the South Central zone, a greater percentage of households manufactured moderate
quantities in these other zones - again probably due to higher banana production levels
and larger landholdings in the East and North Central zones. Paralleling banana
production, the Northwest zone exhibited the highest Share of non-wine producing

households. Yet only 1% of households in the East zone produced no wine.

59

Table 3.5: Zonal wine production by value category (FRW)

 

 

 

 

 

 

 

 

 

 

 

 

NO WINE PRODUCTION 1-2500 FRW 2501-5000 FRW
MEAN % OF ZONE MEAN % OF ZONE MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST .00 46.2% 779.13 16.6% 3829.49 9.6%
SOUTHWEST .00 16.5% 1019.17 32.1% 3614.75 15.0%
NORTH CENTRAL .00 4.4% 1109.85 12.2% 3821.97 20.0%
SOUTH CENTRAL .00 2.3% 1202.00 11.8% 3909.70 11.7%
EAST .00 1.2% 1403.07 7.0% 3957.62 9.7%
5001-10000 FRW 10001-20000 FRW > 20000 FRW
MEAN % OF ZONE MEAN % OF ZONE MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST 7073.24 7.8% 13002.80 11.2% 48063.87 8.6%
SOUTHWEST 7708.36 14.6% 14859.09 12.7% 33068.83 9.1%
NORTH CENTRAL 7064.30 17.7% 15150.25 28.8% 30887.68 16.9%
SOUTH CENTRAL 7589.85 25.0% 13994.09 28.0% 35941.59 21.1%
EAST 7512.21 17.4% 14257.04 29.1% 36761.47 35.6%
ZONAL MEAN
MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST 6622.82 100.0%
SOUTHWEST 6886.84 100.0%
NORTH CENTRAL 11726.75 100.0%
SOUTH CENTRAL 14024.14 100.0%
EAST 19007.09 100.0%
DSA Household Survey, 1990, Kigali, Rwanda
* - F-test significant at <.00001.

3.2.2 Constraints to wine production
Habimfura and Miklavcic (1995) cite four primary constraints to wine production
in 1990. First, among the smaller wine producers in the supplemental survey, an ability
to grow more beer bananas was limited because of land constraints, thus limiting wine
production.
Second, among large wine producers in rural areas, participants in the

supplemental survey claimed that the local market for banana wine was too small to

60

justify expanding wine production. Most households manufacture their own wine from
time to time, and regularly exchange it with neighbors as gifts, thus limiting their demand.

Third, a lack of modern techniques and appropriate technologies, for improving
the efficiency of wine production and increasing the shelf-life of wine, were perceived by
Rwandans as inhibiting production.

Last, inadequate training and poor credit availability were cited as secondary
constraints. In other words, even if better fabrication techniques were available, the
organized dissemination of such knowledge is poor in Rwanda, as is access to credit with

which to buy modern equipment and implement new strategies.

3.2.3 Qualitative observations
In response to decreased banana production, the consensus from the supplemental

survey is that wine production has decreased in recent years.

3.2.4 Data patterns and inferences from them
The data reveal that wine production patterns and banana production patterns were
almost identical in 1990. Thus, it should come as no surprise that a number of
determinants identified as influencing banana production influence wine production.
Therefore, if I were to consider banana production a determinant of wine production, I
would have to consider the following as endogenous determinants: agroclimatic zone,

financial and land assets, other crops grown, and other sources of non-farm income.

61

In addition, the demand for wine (defined as wine purchases by both wholesalers
and cabaret clients) was cited as a determinant of wine production. Although this may
be a potential determinant for inclusion in the Chapter 4 regressions, data concerning the
availability of modern techniques and appropriate technologies, credit access, and training
opportunities are not available to me. Thus these determinants cannot be tested as

influences on wine production at this time.

3.3 Wine Transactions
3.3.1 National and zonal findings
3.3.1.1 SALES

Seventy-one percent of all sample households sold some wine in 1990; this is 80%
of the households that produced bananas and wine. The highest percentage of sellers was
in the East zone (33%), with share of sellers decreasing westward (see Table 3.6). The
smallest percentage of sellers was in the Northwest (8%). These numbers parallel those
for banana and wine producers, meaning that wine sellers were generally the same
households which also produced bananas and wine. Again, the East zone dominated
because of its larger landholdings, as well as warmer and less rainy climate. Nevertheless,
these figures also point to the fact that 20% of the households which produced wine did
not sell any wine. Rather, they consumed all of their own production (or gave it away

as gifts).

62

Thirty-eight percent of wine sales were by households in the East zone (see Table

3.7). The central zones accounted for 20% each of sales, with sales lowest in the western

zones. Once again, this pattern follows the production patterns of both bananas and wine.

Table 3.6: Wine sellers by zone (%)

AGROCLIMATIC ZONE

 

PERCENT OF WINE SELLERS

 

 

NORTHWEST
SOUTHWEST
NORTH CENTRAL
SOUTH CENTRAL
EAST

 

  

 

 

X2, Significant at < .0000]
ouse o . -—urv*~ana.72

7.8
9.2
26.3
23.6
33.1

100.0
(603)

Table 3.7: Wine sales by zone (FRW)"‘

rm~

 

 

 

AGROCLIMATIC PERCENT OF WINE
ZONE SALES
NORTHWEST 12.6
SOUTHWEST 9.1
NORTH CENTRAL 20.2
SOUTH CENTRAL 20.3
EAST 37.8
TOTAL 100.0 6550
(N) (603)

F, significant at <.001
Source: DSA Housefiold Survey, 1555, Kigzr, RwanE

* - Data refer to wine sellers only.

 

 

 

63

Households sold, as marketed surplus, only 44% of the wine value they produced
throughout Rwanda in 1990 (see Table 3.8). Although the western zones sold upwards
of 77% of production, this implies that upwards of 55% was consumed at home or given
away as gifts. In other words, local tradition so encourages home consumption of the
product, especially in the central and eastern zones, that more than half of the BWS’s
output was inhibited from ever reaching the market. This finding will be a significant

factor in establishing the BWS’S potential for ME expansion in the future.

Table 3.8: Marketed surplus of wine (%)"‘

 

AGROCLIMATIC ZONE WINE SALES AS
MEAN PERCENT j
OF BANANA PRODUCTION '
NORTHWEST 59.1
SOUTHWEST 76.9
NORTH CENTRAL 39.0 l
SOUTH CENTRAL 36.5
EAST 39.1
RWANDA 43.5
(N) (603)

 

 

F, significant at < .01

ource: ouse o . urvey, "I ' gir—Tvmn‘auf- "

* - Data refer to wine sellers only.

 

As a share of real household income among wine sellers, wine sales were
relatively high everywhere (see Table 3.9). It was highest in the western zones (21% in

the Southwest, 19% in the Northwest) and lowest in the East and North Central zones

64

(13% and 12% respectively). Although low compared to wine production’s share of
income, it must be kept in mind that these sales generated actual revenues for the
household - which in turn could be used for purchases and investments leading to
household economic growth. On the other hand, the 55% of wine which was not sold

only contributed to household income by supplying nutrients.

Table 3.9: Wine sales as share of household income (FRW)"‘

 

 

 

 

1 a: -— —»— -——-- ...__.————.
[AGROCLIMATIC ZONE WINE SALES AS
MEAN PERCENT )
OF HOUSEHOLD INCOME
NORTHWEST 18.6
SOUTHWEST 21.2
NORTH CENTRAL 11.7
SOUTH CENTRAL 15.7
EAST 13.0
RWANDA 14.5
(N) (603)
F, significant at < .001

 

 

 

ource: I . 'ouse o . urvey, ”Lg‘f: 'wanT—f

* - Data refer to wine sellers only.

Although responsible for only 13% of Rwanda’s wine sales, the Northwest zone
had wine sellers with the highest average earnings from wine (see Table 3.7). Averaging
10,520 FRW in wine revenues in 1990, these earnings imply a small number of large
wine sellers in the Northwest zone. I will support this statement through Gini coefficients

at the end of this section.

65

Total household revenues, from all sources of income, were well-distributed among
wine sellers throughout Rwanda (Table 3.10). In fact, the top 20% of Rwanda’s wine
sellers secured only 32% of aggregate household revenues earned by all of Rwanda’s wine
sellers (correlation coefficient was .17). The correlation coefficient ranged from .13 in
the Southwest zone to .17 in the Northwest zone, thus indicating that the largest wine

sellers are not disproportionately wealthier than their counterparts who sell less wine.

Table 3.10: Zonal correlations of HH revenues and wine seller quintiles (FRW)*

 

 

 

 

 

 

AGROCLIMATIC GINI % HOUSEHOLD REVENUES BY
ZONE COEFFICIENT WINE SELLER QUINTILE
TOP BOTTOM
QUINTILE QUINTILE ’
NORTHWEST .17 37 18 1
SOUTHWEST .13 28 11
NORTH CENTRAL .13 27 17
SOUTH CENTRAL .15 30 14
EAST .17 31 13
RWANDA .17 32 14
(N) (603) _ ___.._ _ _________ _ _ l
ource: ouse o urvey, ”0, ' g 1, 'owan
* - F-test significant at <.00001, based on household revenue means across all five

zones by wine seller quintile.

66

Wine sales were most evenly distributed among wine sellers in the East zone (Gini
was .45). They were most concentrated, however, in the Southwest zone (Gini was .64)
where 20% of the sellers were responsible for 76% of wine sales. Nationally, as shown
in Table 3.11, 20% of Rwanda’s wine sellers were responsible for 57% of its sales (Gini
was .50). In other words, as one moves eastward across Rwanda, larger landholdings and
a more suitable climate result in a more equitable distribution of banana and wine
production, in turn impacting the distribution of wine sales. Nevertheless, Table 3.12
shows that very few producers in each zone sold large quantities of banana wine (i.e.,

more than 20,000 FRW).

Table 3.11: Zonal GINIS for wine sales by wine seller quintile (FRW)“‘

    

 

 

   

 

 

 

 

 

 

 

 

AGROCLIMATIC CORRELATION % WINE SALES BY WINE SELLER
ZONE QUINTILE
TOP BOTTOM
QUINTILE QUINTILE
NORTHWEST .56 64 2
SOUTHWEST .64 76 1
NORTH CENTRAL .47 54 3
SOUTH CENTRAL .46 52 3
EAST .45 49 2
RWANDA .50 57 2
(N) (603)
Source: DSA Household Survey, , g 1, wanda
* - F-test significant at <.00001, based on wine sale means across all five zones by

wine seller quintile.

The producer price of wine was lowest in the East zone and highest in the

Southwest zone (24 versus 37 FRW/L). As shown in Table 3.13, price rose from east to

67

west and from north to south (in the west)3. The lower price eastward is likely due to
the higher volume of wine in this direction, indirectly the result of better growing
conditions for bananas. The increasingly higher price from north to south is supply-driven

in the western zones. The smaller wine supply in the south results in higher prices.

Table 3.12: Zonal wine sales by value category (FRW)*

 

 

 

 

 

 

 

 

 

 

 

NO WINE SALES 1-2500 FRW 2501-5000 FRW
MEAN % OF ZONE MEAN % OF ZONE MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST .00 61.6% 1474.28 11.2% 3692.02 10.2%
SOUTHWEST .00 59.6% 1002.61 22.5% 3691.17 8.6%
NORTH CENTRAL .00 20.9% 1250.71 32.8% 3390.41 15.6%
SOUTH CENTRAL .00 18.6% 1276.83 29.1% 3407.36 17.0%
EAST .00 8.6% 1240.60 25.8% 3721.25 17.0%
5001-10000 FRW 10001-20000 FRW > 20000 FRW
MEAN % OF ZONE MEAN % OF ZONE MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST 6800.34 7.4% 16500.74 3.2% 38657.42 6.4%
SOUTHWEST 7175.49 2.3% 13450.09 4.8% 54144.43 2.4%
NORTH CENTRAL 7195.52 22.8% 13106.53 6.8% 47000.00 1.1%
SOUTH CENTRAL 7006.54 24.5% 14089.32 8.3% 29949.62 2.5%
EAST 7248.02 22.5% 13087.40 19.0% 24917.31 7.1%
ZONAL MEAN
MEAN % OF ZONE
AGROCLIMATIC
ZONE
NORTHWEST 4034.61 100.0%
SOUTHWEST 2627.87 100.0%
NORTH CENTRAL 3981.34 100.0%
SOUTH CENTRAL 4596.60 100.0%
EAST 6829.20 100.0%

 

DSA Household Survey, 1990, Kigali, Rwanda

* - F-test significant at .001

Table 3.13: Producer price of wine by zone (FRW/LITER)*

 

3 According to the supplemental survey, wine prices varied by as much as 400 FRW

per jerrican during the good and bad growing seasons.

68

 

AGROCLIMATIC ZONE PRODUCER PRICE OF WINE
(FRW/LITER)
NORTHWEST 26.14
SOUTHWEST 37.37
NORTH CENTRAL 31.26
SOUTH CENTRAL 30.76
EAST 23.98
RWANDA 28.89
(N) (603)
F, significant at < .0000]

 

 

 

Source: US: Housefioid Survey, i990, fiigfil, RwanE

* - Data refer to wine sellers only.

3.3.1.2 PURCHASES

The mean annual household consumption of wine from household production was
more than 7900 FRW‘. This figure is based on earlier production and sales calculations.
Consumption was highest in the East zone, where wine production was highest.

Thirty-four percent of Rwandan households actually purchased wine in 1990. The
greatest share of wine buyers was in the East zone, decreasing westward (see Table 3.14).
The highest household revenues were found in the East, thus implying that more money
was available for the purchase of wine. Nevertheless, the fewest number of buyers lived
in the Northwest zone, where household revenues for 1990 were also quite high. This

finding may be attributable to poorer households in the Northwest having a preference for

 

4To put this quantity in perspective, one bottle of wine costs 40 FRW. Thus, the
mean HH consumption of wine in 1990 was 198 bottles (a little less than 4 bottles a
week).

69

sorghum beer, the price of which was 65% lower than for banana wine. Although a
similar price differential for these beverages exists in the East (the price of sorghum beer
was 40% lower), traditional behavior and fewer poor households may contribute to the

popularity of banana wine purchases among the East zone’s households.

Table 3.14: Wine buyers by zone (%)*

PERCENT OF WINE BUYERS

AGROCLIMATIC ZONE

 

 

 

ll NORTHWEST 12.2
SOUTHWEST 144
NORTH CENTRAL 227
SOUTH CENTRAL 23.5
EAST 27.2
TOTAL 100.0
(N) (289)
LW x2 insignificant
°“’°e‘ ouse 0 7_urv"'“*"""0‘""‘f‘-;t" ' -

* - Data refer to wine buyers only.

Mean annual purchases of banana wine were highest in the Southwest zone, and
decreased eastward. In fact, Southwest purchases were more than twice the national
average, accounting for 35% of wine purchases in 1990 (not statistically significant - see
Table 3.15). Since the Southwest zone produced the least amount of wine in 1990 (and
had fewer producers than the east or central zones), these figures show that the zones with

the highest and lowest levels of wine purchases do not correspond to the high production

zones for wine. In other words, households which produce a lot of wine on a regular

70

basis have less need to purchase wine.

Wine purchases were most concentrated in the Southwest zone, where the top
quintile of buyers made 94% of all wine purchases in that zone (see Table 3.16).
Although concentrated throughout the country, purchase concentration was lowest in the
East zone (Gini equalled .53), where total household revenues were highest and wine

production greatest. In other words, households in the East were more evenly distributed

in both their demand for purchased wine and ability to pay for it.

Table 3.15: Wine purchases by wine buyers by zone (F RW)“

 

 

 

 

 

* - Data refer to wine buyers only.

AGROCLIMATIC PERCENT OF WINE HH MEAN

ZONE PURCHASES (FRW)
NORTHWEST 7.8 683
SOUTHWEST 34.5 2568
NORTH CENTRAL 22.3 1050
SOUTH CENTRAL 20.0 911
EAST 15.4 607
TOTAL 100.0 1071
(N) (289)

F insignificant

 

 

 

 

_ 0, Egan Rwanda

71

Table 3.16: Zonal GINIS for wine purchases by wine buyer quintile (FRW)"‘

   
   
    
 

 

 

 

 

   
 
 
    
 
  

 

 

AGROCLIMATIC GINI % WINE PURCHASES BY WINE
ZONE COEFFICIENT BUYER QUINTILE
TOP
QUINTILE
NORTHWEST .54 58 2
SOUTHWEST .77 94 0
NORTH CENTRAL .58 69 1
SOUTH CENTRAL .58 67 1
EAST .53 57 3
RWANDA .64 75
(N) (239)
Source: ouse o urvey, , g 1, wan
* - F-test significant at .0000], based on wine purchase means across all five zones

by wine buyer quintile.

The supplemental survey nevertheless indicated that wine purchases were highly

correlated with wine sales, meaning that despite how much they produced, wine sellers

still frequented bars where they purchased additional wines. This is not surprising since

households only manufacture wine two to three times a month, and the shelf-life is only

two to three days.

The supplemental survey also indicated that many household heads regularly

purchased wine using wine credit instead of cash.

 

5In no instance did I come across a sample HH who resold some quantity of the wine

which it purchased.

72

3.3.2 Constraints to wine sales
The primary constraint to wine sales, cited in Habimfura and Miklavcic (1995),
is the short shelf-life associated with wine. Conservable for only two to three days
following production, it is often difficult to sell wine before it goes bad. This problem
is compounded by limited rural markets for wine and poor road systems between regions.
In addition, poor wine quality was cited as a constraint to sales, as were inadequate
training about how to produce better quality wine and a preference for other alcoholic

beverages.

3.3.3 Qualitative observations

All wine producers visited for the supplemental survey had been manufacturing
wine for sale for more than twenty years. Among the large sellers, wine production was
considered "successful" mostly because of access to healthy trees, good tree maintenance,
and inexpensive hired labor (used to care for trees and manufacture wine)‘. Nevertheless,
most households also claimed that revenues from wine had dropped in recent years
because of decreased banana production and wine output, the result of inadequate rainfall
and disease.

Most households in the supplemental survey (i.e., mainly big sellers) said that wine
sales (like banana sales) provide frequent, accessible revenues to meet daily expenses, like

food purchases (e.g., vegetable oil) and clinic expenditures. Cash crops (like coffee)

 

“Only one of the small sellers considered wine selling monetarily "successful".

73

provide bulk revenues once or twice a year. These revenues are more frequently used for
bigger expenses, like house repairs or starting new businesses.

Only 25% of the households had put wine revenues into savings, to increase future
wine production or to educate a child. Twenty percent had used wine revenues to start
additional income-generating activities (usually animal husbandry)7. Although no
households claim to have used their savings to start additional household-level MES,
fungibility must be kept in mind. In other words, the savings generated through wine

revenues helped households to use cash crop earnings to meet other expenses.

3.3.4 Small versus large wine sellers
In this section, I will stratify the core sample to examine differences between small
wine sellers (SWSs, who sold less than 701 FRW of wine) and large wine sellers (LWSS,
who sold more than 13,700 FRW of wine). These are the top and bottom deciles of the
sample, comprised of 65 and 64 households each. In particular, I will focus on
distinctions in levels of participation in the BWS, household characteristics, and
descriptions of the household head. As in my stratification of B85 and NBSS, I will use

these characteristics to inform my Chapter 4 regressions.

 

7Yet in only one case did a HH consider other activities more profitable than wine
sales.

74
3.3.4.1 BWS CHARACTERISTICS

BWS characteristics are shown in Table 3.17. On average, LWSS produced seven
times the value of bananas as SWSs. The value of bananas produced comprised 22% of
total household revenues in the former case, but only 9% in the latter. Although banana
purchases by LWSs were 23 times greater than for SWSs, banana sales, banana purchases,
and banana gifts (both given and received) were virtually the same in each group. All
banana transaction and gift findings, however, are statistically insignificant. Except for
purchases by large wine sellers, these statistics show that banana activities outside of
production are not important components of the BWS, even when stratifying the sample
by its largest and smallest wine sellers.

The gross value of wine production among LWSS was, on average, eight times
greater than for SWSs. Whereas LWSs doubled the original value of their bananas (i.e.,
the wine produced by these households was worth two times the original banana value),
the value added of wine production was only 80% for SWSs. In other words, LWSS
appear better able than SWSs to benefit from economies of scale by producing large
quantities of wine. Wine production accounted for 45% of total household revenues for
LWSS and 16% for SWSs.

GrOss wine sales were a significant source of income for LWSs, but only minor
for SWSs (27% versus 2% of total household revenues). This tells us that households
which sold less than 701 FRW of wine in 1990 probably relied on a number of other
crops and crop sales for their livelihoods than did households which sold more than

13,700 FRW of wine. The Chapter 4 regressions will tell us what an impact these other

75

crops really had on production and sales. Nevertheless, LWSS earned 55 times more

income from wine sales than did SWSs, implying that LWS household incomes are

significantly larger on average than are the incomes of SWSs. This income difference is

probably influenced by different size landholdings.

Table 3.17: BWS characteristics of SWSs and LWSS (N=129)*

BWS TRAIT
(HH MEAN IN FRW)

 

 

 

BANANA SALES 252
BANANA PURCHASES 69
BANANA GIFTS GIVEN 13

BANANA GIFTS 2

RECEIVED
WINE PRODUCTION 5216
WINE SALES 452
WINE PURCHASES 375
WINE GIFTS GIVEN 277
WINE GIFTS RECEIVED 77

“—‘T‘T—"ce. I ' . ____...__,__

* - Data refer to banana producers only.

WINE
SELLERS

(N=65)

 

 

'ouse—o"; urvey, "Im' 1, '

 

SELLERS
(N =_64)

ll BANANA PRODUCTION I 2883

20014

103

1603

28

41

41016

24904

1 690

3251

1339

F-TEST
SIGNIFICANCE

—T—

insignificant
<. 0 1

insignificant

insignificant

<.00001

<.00001
insignificant

<.00001

<.01

 

 

** - Number in parentheses indicates percent of total household revenues.

76

Wine gifts given and received were likewise higher. Although instances of wine
purchases for resale were noted near Kigali during the supplemental survey, it is unlikely
that the purchases made by the core sample were likewise for resale. Informal
cOnversation and observation lead me to believe that an ability to frequent local cabarets
is viewed as a status symbol in Rwanda. Also, accounting for a statistically insignificant
2% of total household revenues among LWSs, I can only assume that these purchases
were the result of an increased ability to pay for wine on a regular basis - through

generally higher household incomes.

3.3.4.2 HOUSEHOLD CHARACTERISTICS

Non-BWS household incomes were more than two times greater for LWSS than
for SWSS in 1990 (see Table 3.18). Nevertheless, after subtracting out the value of wine
production from each category, LWSS earnings 50% dropped below those for SWSs. This
was two times greater per AR (18258 FRW versus 7971 F RW). LWSs also had twice as
much cultivated land (53% more per AE), earned seven times more from other crops, and
had a home consumption rate (inclusive of all farm and non-farm goods) two and half
times greater than SWSs. LWSS also purchased seven times as many farm inputs (a
majority of which was for hired labor), and provided 14% more labor off-farm
(statistically insignificant). In summary, it can be concluded that LWSS are generally

richer than SWSs, and are thus better able to participate in the BWS.

77
Table 3.18: Household characteristics of SWSs and LWSS (N=129)*

 

 

 

   

 

BWS TRAIT SMALL LARGE F-TEST
(HH MEAN) WINE WINE SIGNIFICANCE
SELLERS SELLERS
(N = 65) (N = 64)
_ __ _______ _ _ L .
NON-BWS HH 32707 65577 < 01
REVENUES (FRW)
NON-BWS REVENUES 7971 18258 <.00001
PER AE (FRW)
CULTIVATED LAND 52 105 <.00001 ll
(ARES)
CULTIVATED LAND 14.0 21.4 <.05
PER AE (ARES)
INPUT PURCHASES 1114 7834 <.00001
(FRW)
LABOR PURCHASES 1132 7662 <.00001
(FRW)
LABOR SALES (FRW) 13440 17194 insignificant
SALES FROM AG 3987 27830 <.00001
PRODUCTION (FRW)
HOME CONSUMPTION 15713 41864 <.00001
(FRW)
ADULT EQUIVALENTS 4.1 5.3 <.001
IN HOUSEHOLD (#)

 

 

 

 

 

 

 

‘ ‘ouse o . urvey, 1990, Egan Rwanda

* - Data refer to banana-producing households only.

 

l

78
3.3.4.3 DEMOGRAPHIC CHARACTERISTICS

As shown in Table 3.19, a slightly higher percentage of LWSS than SWSs had
male household heads in 1990 (82% male versus 77% female - not statistically
significant). Also, 71% of LWSS were married, compared to 60% of SWSs. These
findings imply that men may be better geared, at the present time, towards participation
in larger ME activities like banana wine production and sales. Although statistically
insignificant, and therefore omitted from my analysis in Chapter 4, this is supported by
my observation (taken from informal conversations throughout Rwanda) that women, to
date, are not allowed as much autonomy as women in other parts of Africa. They also
receive less education. In addition, married couples may be better committed to taking
on the additional time and responsibility required to manufacture and sell large quantities

of wine.

3.3.5 Data patterns and inferences from them

Participation in wine sales is influenced by many of the same determinants as
banana production, banana sales, and wine production. These determinants include
agroclimatic location, total household revenues, cultivated land, other crops grown, banana
and wine production, and banana and wine prices.

The comparison of LWSS and SWSS points to a number of other influences as well
- particularly with regard to participation levels. Included are number of AEs in the
household, home consumption levels, value of purchased inputs, supply of off-farm labor,

and household head education level.

79
Table 3.19: Demographic traits of SWSs and LWSS (N=129)*

fi

 

HOUSEHOLD SMALL WINE LARGE WINE SELLERS

 

 

 

 

HEAD SELLERS (N = 64)
TRAITS (N = 65)
GENDER male 77.1% male 81.7%
female 22.9% female 1 8.3%
AGE mean 47.8 years mean 45.5 years
MARITAL married 59.6% married 70.9%
STATUS widowed 18.2% widowed 12.2%
divorced 1 .3% divorced --
separated 2.1% separated --
live together 13.3% live together 14.9%
single 5.5% single 2.0%
EDUCATION none 67.4% none 41 .8%

some primary 29.6% some primary 37.7%
all primary 0.5% all primary 15.9%
post-primary 0.8% post-primary 2.0%
all secondary 0.5% all secondary 2.6%

 

 

 

 

PRINCIPAL farmer 84.3% farmer 86.5%
ACTIVITY paid farm worker 0.8% paid farm worker 2.0%
ind. artisan 6.1% ind. artisan 0.8%
commerce -- commerce 3.5%
civil servant 2.3% civil servant 2.6%
student 1 .3% student --
other 2.4% other 1 .4%
unemployed 2.9% unemployed 3.3%

D ‘ ouse 0c °‘i‘,g1,"'_“ancl

 

 

 

All households produced bananas; no demographic interactions are statistically
significant.

80

3.4 Kigali’s Unusual Market Structure Versus the Rest of Rwanda

Immediately outside Kigali, two markets have developed for the exclusive trade
of banana wine. One is for more expensive wine (i.e., less diluted with water), the other
for lower quality wine (i.e., more water added). These markets are unlike any other found
in Rwanda because they deal exclusively in wine.

Cabaret owners from Kigali travel fifteen miles to these markets two days a week.
They make advance, round-trip reservations with transport services for their wine barrels
(large, plastic containers) and arrange for their own travel to and from the market by taxi.
At the market, they sample numerous wines brought in by producers and wholesalers,
buying what they prefer and filling their large barrels. Once the barrels are full, they are
carried back to Kigali, where their shelf-life is 2-3 days. Barrels empty before the next
market day.

The key point of this description is that an expanding wine market is in the
process of developing near Kigali. This market is very unlike other semi-rural areas in
Rwanda, where wine is carried to a market to be exchanged among all kinds of other
commodities and sold for immediate resale at market wine stalls. Nevertheless, the Rapid
Appraisal indicated that more and more people are beginning to participate in this wine
market. Since there are many other urban areas throughout Rwanda, where a concerted
effort could undoubtedly result in the same kind of market success, it may just be possible

to initiate other such centrally-located markets by improving infrastructural services.

81

3.5 Wholesalers and Transporters

The BWS involves many more participants than those who produce and directly
sell bananas and wine. Numerous wholesalers, transporters, and bar owners are also
involved. As Figures 2.2 and 2.3 show, a number of wholesalers and transporters,
involved in trading the relatively small quantitites of bananas which ar_e traded, are very
important to the subsector. In rural areas, wholesalers consist of: 1) local residents who
purchase bananas from neighbors, transport bananas via modest means to local markets
(i.e., by bicycle or wheelbarrow), and resell them for a small profit, and 2) wealthy truck
owners who buy, transport, and resell bananas from one region to another as part of a
regular but small, very lucrative transportation business. This is commonly seen between
the East zone (i.e., Kibungo prefecture), where banana production is highest and the South
Central zone (i.e., Butare), where banana production is much lower but household wine
production mean highest. These truckers follow established routes for purchasing bananas
from Eastern banana growers and sell to regular buyers at the other end.

Unlike rural areas, where two kinds of wholesalers are found, semi-rural areas
have only one kind of wholesaler. The semi-rural wholesaler buys, transports, and resells
bananas, but at larger, more urban-oriented markets.

There is also another kind of transporter, who participates in no banana
transactions, but rather delivers the bananas purchased by wine producers at local markets.
In rural areas, these services are done on a small-scale, but in semi-rural locations, they
are the highest paying segment of the subsector. In 1990, at the banana market 15 miles

outside of Kigali, truck owners with established delivery routes to urban areas were

82

grossing $200.00 a day on market day, two days a week. Although the overall trade of

bananas is small in Rwanda, a few participants are earning large incomes from it.

3.6 Conclusion

From this chapter, I am able to summarize a number of key points concerning
participation in wine activities.

First, all households who grow bananas manufacture at least some wine; virtually
all households who do not grow bananas do not produce wine. In other words, wine is
rarely, if ever, produced exclusively from purchased bananas.

Second, wine production is highest where banana production is highest - and
lowest where banana production is lowest. In other words, wine production is highest and
most evenly distributed in the East; it is more concentrated in the western half of the
country.

Third, a majority of households which produce wine also sell wine, although most
of the wine that is produced is actually consumed at home. Very few producers in each
zone sell large quantities of wine.

Fourth, although wine sales are extensive throughout Rwanda, wine purchases are
less extensive. Wine purchases are most frequently made by large wine sellers whose
purchases frequently offset the value of their sales.

Fifth, larger and richer households usually produce and sell larger quantities of
wine. This complements my findings that banana sales are most frequently made by

smaller, poorer households.

CHAPTER FOUR

DETERMINANTS OF HOUSEHOLD-LEVEL PARTICIPATION

4.1 Research Questions

This chapter addresses four questions. First, what determines whether (and to what
extent) a household produces bananas? Second, of households that grow bananas, what
determines whether (and to what extent) a household sells at least some of the bananas
that it produces? Third, what determines whether (and to what extent) a household
manufactures banana wine? Fourth, what determines whether (and to what extent) a

household sells wine from its own production?

4.2 Model Specification
4.2.1 Conceptual model
The conceptual model proposed in this thesis is adapted from Randall Schnepf’s
1992 household nutrition model for Rwanda (Schnepf, 1992). I have adapted Schnepf’s
model to identify the influences acting on a household’s decision to produce and/or sell
both beer bananas and banana wine.
AS shown in Figure 4.1, the resources of a household consist of land, labor, other

capital assets (savings, crop production, and equipment value), and education. Resource

83

84

use is constrained by farm characteristics (distance to market, child-to-adult ratio,
landholdings, and characteristics of the household’s primary decision-maker), agroclimatic
influences (like altitude, rainfall, and soil quality), and incentives and disincentives (credit

access, extension services, infrastructure, prices).

Figure 4.1: Conceptual schematic of the proposed model

 

HOUSEHOLD RESOURCES:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

- Land
- Labor ~ Note: Banana and
- Other Capital wine activities
assets are highlighted by
- Education double lines, and
H ' capitalized, bold
letters.
RESOURCE ALLOCATION:
(constrained by)
- Farm Characteristics
- Climatic Influences
- Incentives and
Disincentives
H
ON-PARK OPP-PARK
ACTIVITIES ACTIVITIES:
- - skilled
l labor
- unskilled
u AGRICULTURAL II I non-ac Household: labor
, ==== - meals
- childcare
- food/water
other 11 833R
crops BANANAS
-———{other activities]
-' 3) mm mu ::
.1 PRODUCTION H I

 

 

H

2)__|| ems:

 

Home Consumption
or Gifts of Wine

 

 

 

 

 

- directly from home
- indirectly (local market)
H - indirectly (bar for beer)

ll ,
ll ms II

 

 

 

 

 

fl

INVIBTIINT:
- )ag activity
- non-ag activity
- other

 

Food IBudget] l Non- Food Budget

 

 

 

 

   

 

85

Resources can be allocated in four ways, a combination of which is used by each
household: towards agricultural, non-agricultural, household, and off-farm activities.
Non-agricultural activities refer to household operations such as wine production, whereas
off-farm activities refer to skilled or unskilled labor hired out to pier farms. Off-farm
activities are important to this model because they shrink the labor pool available for
home production of wine and provide alternative sources of cash (hence serving as
substitutes for banana or wine sales).

In turn, the above resources and constraints influence the household’s decisions
about:

(a) what quantity of bananas to grow (refer to Figure 4.1, step

1).
(b) whether to sell the bananas and/or use them in banana wine
production (many households do both, as respectively
shown in steps 2 and 3),

(c) how much of the household’s wine production to sell1 (step
4), and

(d) ‘ how to use the revenues obtained through the sale of
bananas and wine (step 5).

This chapter focuses on the determinants which influence points (a) through (c).

 

1The above resources and constraints also influence how much wine a household
purchases from sources external to the household. Because these purchases involve
production by sources other than the household, they are not included in this diagram.
Recall that they were discussed in Section 3.3.1.2.

86

4.2.2 Economic framework

Using Figure 4.1, I have developed the following economic framework against
which to model the regression component of my analysis. Questions 1 and 3, which
examine the determinants of household-level participation (and degree of participation)
in banana and wine production, can be addressed with standard production fimction
theory; questions 2 and 4, which examine sales, with the marketed surplus literature.
Both aspects of my theoretical framework are based on Strauss’s examination of marketed
surplus among agricultural households in Sierra Leone (1984a). As in Rwanda, Sierra
Leonean households are semi-subsistence households. That is, they produce at least some
of the goods which they consume. According to Strauss, "the most important property
of this model is that it is recursive. The household behaves as though its production and

consumption decisions are separable" (1984b, p. 79).

4.2.2.1 PRODUCTION MODEL
Strauss (1984a) specifies production of commodity X as:
(1) X. = X.- (13.2, k)-
Subscript i specifies the commodity, p a vector of prices, 2 a vector of farm characteristics
including fixed inputs, and k a vector of production technology parameters. Since
banana- and wine-specific production technology parameters are not contained in my data
set, I am instead examining the impact of two so-called factors of production - total inputs

purchased by the household and a series of climatic influences (as contained in my

87

dummy variables for agroclimatic zone). The vector of prices and vector of farm

characteristics remain the same.

4.2.2.2 MARKETED SURPLUS MODEL
Strauss (1984a) specifies marketed surplus of good i as:
(2) MSi = Xi - X", where
household consumption of i is defined as:
(3) X5 = X." [P, n. A + P~T(m) + “(2,0,01-
p represents a vector of prices, pN the price of labor, I] the household characteristics
affecting taste, T the time available to a household for work and leisure, m the household
characteristics determining T, A the household’s exogenous income, and H the household’s
profits from the output of good i. In other words, household consumption is determined
by prices, tastes, and income. In turn, household consumption influences the quantities of
goods made available for sale.
Strauss’s production and marketed surplus models are derived from the following
utility function:
(4) U (X°, L), where X‘ = (X,°' XN'“).
Xc represents a vector of goods consumed by the household and L represents household
leisure. This utility function is subject to three constraints: a production function
(described previously), a time constraint, and a budget constraint. The time and budget

constraints are combined as follows:

(5) A+ X pi(xi'xic)+pN(T'L'LT)=Oa

88

such that
(6) A + ()3 P1X. - PNLT) + PNT = 2 PIXf + PNL-
In equation (6), the left-hand side represents full income (the sum of exogenous income,
profits from household enterprises, and value of household time) and the right-hand side
represents consumption (value of goods plus leisure consumed). This separation of
income versus expenditure defines the recursive property noted in Section 4.2.2.
From these equations, Strauss derives the following marketed surplus function:
(7) MSi = MS,(p,n,m,z,k,A).

For my analysis, I am retaining the same price vector. As for household
characteristics, categorized by Strauss’s model as either influencing taste or time, I have
chosen to compile these categories with farm characteristics, 2, because of a lack of data
pertaining to time and taste shifters. Production technologies again include inputs and
climatic influences. Thus, my marketed surplus reduced form becomes:

(7’) M8. = MS.(P.z,k).

the contents of which are identical to those contained in my production function.

4.3 Regression Specification
Four reduced forms are used to represent each of the four primary research
questions:
(1) Y1 = BIIPI + [31221 + 1313K) + U1
(2) Y2 = 13er2 + 132222 1’ 323K2 + U2

(3) Y3 = B31133 + 133223 + B33K3 + U3

89
(4) Y4 = [3,,P4 + [3,224 + [3,,K4 + U4

Yl represents the resultant output of banana production decisions by the household,
Y2 the output of banana sales decisions, Y3 the output of wine production decisions, and
Y4 the output of wine sales decisions. Each decision consists'of two parts: whether to
participate in the cited activity (examined with a PROBIT model), and to what extent
(examined with OLS, using only the subset of households which participated). P, Z, and
K are the respective sets of price, farm, and factor of production variables to be tested in
each regression, and U the unmeasured residual.

The above reduced forms are obtained from the following structural forms, in
which X represents a joint vector of P, Z, and K variables, and the definitions of Y,, Y2,
Y3, Y4, and U are stated in the preceding paragraph:

(a) Y] = 51X. + THY; + 7nY3 + YI3Y4 + U1
(b) Y2 = Bzxz + Y21Y1 + 722Y3 + THY. + U2
(C) Y3 = Bsxs + Y31Y1 + Y32Y2 + 'Ys3Y4 + U3
(d) Y4 = B4X4 + Y41Y1 + Y42Y2 + Y43Y4 + U4

As the structural forms show, each dependent variable is partially determined by
three endogenous variables, identified as dependent variables in the other three equations
of the series. In other words, one subsectoral decision depends on decisions made
elsewhere in the subsector. For example, the quantity of bananas a HH decides to grow
is determined by a number of exogenous variables (household income, landholdings,
education etc.) as well as by the quantities of bananas it sells, wine it produces, and wine

it sells. I therefore make one general assumption that all household decisions to

9O

participate in different stages of the subsector are made simultaneously. This assumption
allows for the structural forms to be rendered into reduced forms through the following

matrix calculation:

 

 

 

 

 

 

 

 

1 “1'11 'le 'Yrs Y1 BIXI U1
'721 1 '722 '723 Y2 = Bzxz + U2
‘731 '732 1 "Y33 Y3 I33X3 U3
'Y41 “742 '743 1 Y4 B4X4 U4

In turn,
Y1 BIXI U1
Y2 = A] Bzxz + U2
Y3 53X; U3
Y4 B4X4 U4

 

 

 

 

 

 

 

Ultimately, the following reduced form is obtained:
Y = Bx + U.
This is the form shown in equations (1) through (4) in the first paragraph of this section.
This reduced form is also applied to three secondary pairs of regressions. Using
the same independent variables, I use this same form to examine their influence on wine
sales as a percentage of non-BWS household income, wine sales as a percentage of wine
production, and wine purchases. Wine sales as a percentage of income controls for sales
and income disparity among households. Wine sales as a percentage of production
controls for different-sized landholdings. Wine purchases enables a closer look at the

characteristics which encourage households to buy wine.

91

Heckman’s treatment of OLS is incorporated into this model because it adjusts for
censored dependent variables. Defined as underlying dependent variables whose models
are linear but that are not fully observable (Goldberger, 1991), the censored data in my
thesis result from the fact that not all sample households participated in each BWS
activity in 1990. Thus, the subset of households used to examine the determinants of
levels of participation may be biased because they do not include households which did
not participate (Heckman, 1979). Heckman’s methodology, also known as Heckit,

provides an important adjustment for analyses like the ones contained in this thesis.

4.3.1 Heckman’s two-stage procedure
Heckman corrects for a censored sample by including in the regression series an
Inverse Mill’s Ratio, defined as:
7». = [f((0t + BXi/O)]/[F((a + [KO/0)]-
Described well by Wohl (1992), and therefore not restated in full here, Heckman’s
estimation procedure entails running a normal PROBIT analysis using the same
dichotomous variables described in the following section for BWS participation and non-
participation. The independent variables also remain the same. The results of the
PROBIT analysis are used to calculate (or + BXQ/O’, which is then used to compute 1i.
Ultimately, by including the Inverse Mill’s ratio as an independent variable in the
successive OLS series, which comprises the second stage of Heckman’s two-stage
procedure, consistent and asymptotically normal estimators of a and [3 are obtained

because the Inverse Mill’s Ratio corrects for the missing variable bias.

92
4.3.2 The PROBIT models

Binary choice, PROBIT models are used to estimate participation rates in each
BWS division. PROBIT is selected because it is useful in estimating the effects of one
or more independent variable on a dichotomous dependent variable, for which the
outcome is yes or no. In this case, the outcome depends on whether the household
participates in the production and/or sale of bananas and/or wine.

The PROBIT models use cumulative normal functions and rely on maximum
likelihood in their estimation (see Maddala, 1992). By dividing the sample into
participants and non-participants (dummy variable Y equals 0 for non-participation or 1
for participation below), PROBIT allows me to examine data in the context of what
determines actual participation (as opposed to level of participation). In keeping with the
production and marketed surplus functions cited above, these models are as follow:

(1) PRODBVij = [3,le“ + 1312qu + 1513an1 + U“,-

where YI = 0 if PRODBVij S 0
1 if PRODBVij > 0

(2) SELLBVij = BZIjPZi + BZZIZZI + B23jK2i + U2ij

where Y2 = 0 if SELLBVij S 0
1 if SELLBVij > O

(3) PRODWVTJ' = 1331,1331 + B32jz3i + BBBjKBi + U30

where Y3 = 0 if PRODWVij S 0
1 if PRODWVij > O

(4) SELLWVij = 134111341 + 1342;241 + B43jK4i + U41)

where Y, = 0 if SELLWV, s 0
1 if SELLWVij > 0

 

93

As shown above, PROBIT enables Y,, Y2, Y3, and Y4 to serve as the model’s dependent
variables.

PRODBVij is the indirect utility function of household i, based on its decision j
to grow beer bananas. Similarly, SELLBVij is the indirect utility function of banana-
growing household i, associated with its decision j to sell bananas. PRODWVij is the
indirect utility ftmction of banana grower i, based on its decision j to produce wine. Last,
SELLWVD- is the indirect utility function of wine-producing household i, associated with
its decision j to sell some quantity of wine.

The right-hand variables (P, Z, and K) represent the price, farm, and factor of
production characteristics of each household. They are described in greater detail, along
with a series of hypotheses, in the following paragraphs. Uij is a stochastic component
that captures the unmeasured determinants of choice (Glick and Sahn, 1993, p. 4). In all

cases, as defined by the PROBIT model, Uij is normally distributed (N ~ 0,1).

4.3.3 The OLS models
Four ordinary least squares models are also used in this analysis. This is because
OLS, after selecting out participating households, is a better tool than PROBIT for
examining continuous dependent variables. Although the PROBIT estimations are
expected to yield interesting insights into the influences acting on a household’s decision
to participate in each of the BWS’s four subsectoral components, the OLS models are

anticipated to shed light on the magpitudes of participation. As such, they better describe

 

94

the determinants of different levels of production and sales - information which will allow

me a different angle through which to explore the ME potential of BWS activities.

The four OLS models take the following form:

(1)

(2)

(3)

(4)

PRODBVI) = Blljpli + 131sz11 + BlBjKli + BMjkli + U

lij

where Y = PRODBVij if PRODBVij > 0
0 if PRODBVij s 0

SELLBVij = BZUPZ, + Bzszz. + 1323,19. + Bra-4r + U21)

where Y = SELLBVij if SELLBVij > 0
o if SELLBVij s 0

PRODWVg = 1531,1331 + I332)Z3i + B33jK31 + 1334901 + U30

where Y = PRODWVij if PRODWVij > 0

0 if PRODWV-- < 0

I] -
SELLWVI)‘ = B4le4i + [342,241 + B43jK4i + I344j7‘v4i + U40

where Y = SELLWVij if SELLWVij > 0
0 if SELLWV, s 0

In this OLS selection process, unlike in the PROBIT model, only households who

participated in the specified BWS activity are retained for observation. The continuous

dependent variables stand for banana production, banana sales, wine production, and wine

sales. All dependent variables are measured in FRW. The description of the first three

independent and U variables are the same as for the PROBIT models. Lambda,

representing the Inverse Mill’s Ratio, represents the fourth independent variable.

95

4.3.4 Variable justification and hypotheses
In the following justification of variables, I intersperse a number of variable-
oriented hypotheses which I plan to test through the regressions. The first pertains to
prices, the next three to farm characteristics, and the last two to factors of production.

All hypotheses refer to ceteris paribus conditions.

4.3.4.1 PRICES

My price vector, p, which is used in both the production and marketed surplus
regressions, includes the mean monthly producer prices (by each of Rwanda’s ten
prefectures) for four of Rwanda’s seven major cropsz. Identified by the core data as
those crops with the highest production values nationally, prices are included for sweet
potatoes, coffee, sorghum, and potatoes. I also include the price of beer bananas in all
regressions, although the price of wine is omitted because of its non-inclusion in the
original data set. Also, any attempt to estimate and include wine price results in a
collinear situation. Therefore, since bananas and wine are highly correlated, I am also
using the price of bananas as a proxy for the price of wine.

Prices are included for the following reason. In Chapter 2, the supplemental
survey pointed to a preference for other crops as constraining banana production. As
traditional microeconomic theory states, price changes frequently lead to a series of

substitution effects which result in new preferences (Nicholson, 1989). Since household

 

2Beans, maize, and manioc are omitted from the series because of high price
correlations.

96

profits rise or fall in response to price changes, these changes impact the household’s
budget constraint and farm decisions associated with this constraint.

For example, if the price of a frequently sold but never consumed commodity
increases (like com), the household may decide to produce more corn at the expense of
another commodity (like bananas). Thus, in this example, the price of corn directly
influences the production and sale of bananas. Indirectly, it influences wine production

and wine sales, leading me to propose the following:

HYPOTHESIS 1: As the selling prices of other crops increase, banana production and

sales decrease.

4.3.4.2 FARM CHARACTERISTICS

My vector of farm characteristics, 2, includes total household income per AB in
1990, cultivated land per AE, distance to market, child to adult ratio, and education of the
household head. These particular variables are included for the following reasons.

Income is included as a relevant farm characteristic because household income per
adult equivalent is shown in Chapters 2 and 3 to be significantly higher among large wine
sellers than small wine sellers. Even after subtracting the value of banana and wine sales,
this distinction is evident. Likewise, banana sellers earn significantly lower incomes than
non-banana sellers. In other words, it is probable that lower incomes result in a more

frequent need to sell bananas for emergency funds (see Chapter 2). Thus, higher income

97

households are in a better position to manufacture wine, leading to the following

hypothesis:

HYPOTHESIS 2: Higher income households are less inclined to sell bananas for cash
and more inclined to participate in wine production and, hence,

wine sales.

In terms of landholdings, Clay and Reardon (1994) claim that farmers use virtually
all arable land for agriculture in Rwanda and that they tend towards crop diversification.
In conjunction with the fact that 90% of all households in Rwanda grow bananas (see
Chapter 2), and that the Spearman correlation between cultivated landholdings and
participation in three of the four BWS activities is significant, it is justifiable to assume
that more land implies more land planted in a highly-valued crop - bananas. In turn,
banana production is significantly correlated with banana sales, wine production, and wine
sales, thus implying that participation in these activities will also increase as landholdings

increase. I therefore propose the following:

HYPOTHESIS 3: Households with larger landholdings are more inclined to
participate in all stages of the BWS. Among households who
participate, those with larger landholdings (per AE) exhibit higher

levels of participation.

 

98

Nearby markets imply improved trading opportunities. Since farm families in
developing countries are generally described as ’efficient’ and ’profit-maximizing’ (Ellis,
1988; Schultz, 1964), it can be assumed that those households located nearer to markets
will be in a better position to take advantage of the BWS opportunities offered by the
markets. Spurred on by larger profit margins than households located far from markets
(whose transport costs to market decrease profits), such households are indeed better able

to maximize BWS profits. Thus, I propose:

HYPOTHESIS 4: Households located closer to markets are more likely to participate
in banana and wine sales. Among households who sell bananas and

wine, those located closer to markets sell larger quantities of each.

The ratio of children to adults is included as a relevant farm characteristic because
an increased child to adult ratio implies additional responsibilities in the household which
are associated with child care. Intuitively, adults have less time available to manufacture
and sell banana wine - a fact reported by a number of supplemental survey households.
Rather than have their bananas go bad, households with high child to adult ratios sell
them for a small profit.

Education is frequently cited as one of the most important determinants of
economic grth (Parish and Willis, 1993; Boissiere, Knight, and Sabot, 1985), and is
now a frequent component of household-level research (see Kimhi, 1994; Reardon,

Delgado, and Matlon, 1992). Although Chuta and Liedholm (1979) did not find

99

education to be a good measure of entrepreneurial capacity (they actually found a negative
relationship), and Chapters 2 and 3 did not show the relationship between education and
participation in the BWS to be statistically significant, McPherson (1992) found it a
worthwhile predictor of microenterprise growth in Zimbabwe.

On one hand, it can be said that education encourages farmers away from farm-
oriented activities, perhaps towards more remunerative occupations. On the other hand,
barring accessible off-farm opportunities, education may enhance a household’s ability to

maximize profits (i.e., through participation in the BWS).

4.3.4.3 FACTORS OF PRODUCTION

The factors of production vector, k, contains total inputs purchased by the
household (including labor purchases) and climatic factors. Climatic conditions, of which
rainfall was particularly noted in Chapters 2 and 3 as influencing banana production, are
aggregatedly accounted for in the agroclimatic zone variables. These variables also take
into account soil quality and susceptibility to erosion, both factors of which are thought
to influence banana production (Birasa et al., 1992) (see Appendix 2).

Purchased inputs may influence banana production in two ways. First, if the
inputs (including labor) are targeted towards banana production, then higher levels of
inputs probably mean a higher yield of bananas. In turn, wine production and possibly
wine sales presumably increase. On the other hand, if the inputs are targeted away from
bananas, then fewer bananas are likely produced by the household - either due to

receiving fewer inputs to enhance grth or to smaller quantities of cropland planted in

 

100

bananas (i.e., more land is planted with input-receiving crops.) Since aggregate inputs are
included in the data set but their destination unknown, it is not possible to predict their
influence on either banana production or sales. They are nevertheless included in the
regressions.

Zonally, the East zone is endowed with milder temperatures and less rainfall,
conditions better-suited to the production of bananas than anywhere else in Rwanda
(UNR, 1989). As the descriptive chapters point out, this agroclimatic distinction favors

participation in all BWS activities. Since Chapters 2 and 3 did not examine ceteris

 

paribus conditions, the real impact of climatic influences on banana and wine production

and sales are so far unknown. I therefore propose:

HYPOTHESIS 5: Households outside the East zone are less likely to participate in
any stage of the BWS. On average, those who participate outside

the East zone do so to a lesser degree.

4.4 Descriptive Statistics

Excluding the selection bias variable included in the OLS regressions, 16
independent variables are used in each analysis (2 PROBIT regressions and 2 OLS
regressions). Each variable is thought to influence, in some way, a household’s decision
to produce and/or sell bananas and/or wine. Descriptive statistics are provided for all
dependent and independent variables in Table 4.1. Spearman correlations are cited in

Table 4.2 for the dependent variables.

 

.5

 

101

Note that Tables 4.1 and 4.2 also include descriptive information for wine sales
as a percentage of non-banana income earned by the household, wine sales as a percentage
of wine production, and wine purchases. I use these three dependent variables in six
additional analyses (again one PROBIT and one OLS regression per dependent variable,
containing all the same independent variables) to supplement my core findings.

The task now is to determine the degree of individual effect of each variable on

participatory decisions, holding all other variables constant.

4.5 Regression Results
Table 4.2 shows the regression results of the PROBIT analyses. The second-stage
OLS results are shown in Table 4.3. An overview of my findings is presented in Sections

4.5.1 through 4.5.4.

4.5.1 Participation in banana production
Seven variables are found to significantly influence the decision to participate in
banana production - the prices of sweet potatoes, sorghum, and potatoes, level of income,
distance to market, quantity of inputs purchased, and location in Rwanda’s western zones.
As sorghum and potato prices increase, households shift away from banana
production. They shift towards banana production as sweet potato prices go up. The first
two findings support Hypothesis 1, that increased selling prices of other crops cause

banana production to decline, although the latter one does not. This may be explained

102

Table 4.1: Descriptive statistics (household-level, N=853)

 

 

VARIABLE MEAN S.D. RANGE UNITS
PRODUCTION:
Banana production 6446.90 7801.50 0-89830 FRW
Wine production 12732.00 15172.00 0-137300 FRW
SALES:
Banana sales 190.60 754.72 0-10150 FRW
Wine sales 4538.70 7763.20 0-99800 FRW
Wine/non-banana income .12 0.34 -7-3 RATIO
Wine sales/wine production .30 0.72 0-19 RATIO
PURCHASES:
Wine
358.69 2999.40 0—82450 FRW
PRICE:
Sweet potatoes
Coffee 8.79 1.53 6.66-12.87 FRW/KG
Sorghum 83 .33 0.94 81 .26-84.56 FRW/KG
Potatoes 21.48 3.36 17.81-31.32 FRW/KG
Beer bananas 14.49 3.65 9.34-19.41 FRW/KG
4.54 0.90 3.45-6.13 FRW/KG
FARM
CHARACTERISTIC: 11308.00 10858.00 66460-109000 FRW
Non-banana income/AE 15.41 13.06 0.62-178.4 HA
Cultivated land/AE 4.48 1.46 2-8 KMS
Distance to market 1.02 0.81 0-5 RATIO
Child/adult ratio 1.64 0.89 1-6 ca
Education
FACTORS OF
PRODUCTION: 2698.90 7392.40 0-88400 FRW
Inputs .16 -- 0-1 dummy
Northwest zone .14 -- 0-1 dummy
Southwest zone .24 -- 0-1 dummy
North Central zone .20 -- 0-1 dummy
South Central zone .26 —- 0-1 dummy

East zone

Source: DSA Household Survey, 1990, Kigali, Rwanda

 

3Education categories: 1 = no formal education, 2 = some primary school, 3 =
finished primary, 4 = post-primary school, 5 = some secondary education, 6 = finished

secondary school.

 

 

 

 

 

 

 

 

82: m 5 mm. 2N5 .- onmv. :8 m _ 5. name.
83:55
2:3
55595
2mm. coco; mhoo. owmm. muco. 85. 2.3. 2:3
Roam DE?
0885
£55.52
5mm? moo. 82: 55. $18. 38... ommm. B23 2:3
She. 33. mem. coco; 2 K. .35.- meow. mus—am DE?
55505
wnmc. mmco. 3mm. 2 2.. coco; goo. $3. 2:3
mofim
2 3. 85. 58.. 58.- goo. coco; «2:. «555
55:55
38. :mo. ommm. 342.. Nmnm. M33. 82: «55m
55505 0885
83555 255 $55.52 55255 55505
05? B25 053 E28 2:? mofim 55? 05>» mofim «55m «555

 

 

8355, 55:55 .5 50:38.50 ”me 035.

m3

 

104
Table 4.3: Decision to participate: PROBIT results4

 

 

 

VARIABLE BANANA BANANA WINE WINE
PRODUCTION SALES PRODUCTION SALES
Intercept 7.1165 -3.S115 7.4269 12.3580
(1.12) (-0.53) (1.17) (2.22)**
Sweet potatoes 0.2001 -0.2327 0.1901 0.0970
(2.22)** (~4.22)** (2.12)** (1.89)
Coffee -0.0428 0.0617 -0.0453 -0.1385
(-0.54) (0.80) (-0.57) (-2.11)H
Sorghum -0.0951 -0.0412 -0.0968 -0.0167
(-3.48)** (-1.92)* (-3.S4)** (-0.84)
Potatoes -0.1049 0.0153 -0.9365 -0.0233
(-2.14)** (0.64) (-1.92)* (-0.92)
Beer bananas -0.0254 0.0258 -0.0761 -0.0736
(-0.14) (0.26) (-0.40) (-0.71)
Non-BWS income/AB 0.0000 0.0000 0.0000 0.0000
(2.32)** (1.24) (2.29)** (0.22)
Cultivated land/AB 0.0051 -0.0091 0.0050 0.0202
(0.66) (~2.20)n (0.65) (3.86)*'
Distance to market 0.2471 0.0062 0.2506 0.1088
(3.34)** (0.17) (3.3S)** (2.73)**
Child/adult ratio -0 . 1376 -0 . 0207 -0 . 1391 -0 .0772
(-1.44) (-0.31) (-1.4S) (-1.18)
Education -0.0914 0.0678 -0.0653 0.0819
(-0.98) (1.04) (-0.69) (1.21)
Inputs -0.0000 -0.0000 -0.0000 -0.0000
(~2.77)** (-2.27)** (-2.90)** (-2.14)**
Northwest zone -2.5779 -0.1702 -2.4910 -1.7665
(-5.68)** (~0.74) (~5.46)** (~7.50)**
Southwest zone -1.2839 0.7264 -1.1781 -1.7506
(-2.60)** (2.75)** (-2.36)** (-6.09)**
North Central zone -0.6769 -0.3916 -0.6514 -0.5086
(-1.67)* (-2.32)** (-1.60) (-2.81)**
South Central zone -0.3648 0.4196 -0.2470 -0.4619
(-0.89) (2.50)H (-0.59) (-2.38)"
% Correct Predictions 90.27 80.30 90.27 79.13

 

Source: DSA Household Survey, 1990, Kigali, Rwanda

** = significant at .05 level * = significant at .10 level

 

4t-ratio in parentheses.

105

 

 

 

 

5t-ratio in parentheses.

Table 4.3: Decision to participate: PROBIT resultsS (continued)
VARIABLE RATIO OF RATIO OF
WINE SALES WINE SALES WINE
TO NON-BWS TO WINE PURCHASES
INCOME PRODUCTION
Intercept 12.1580 12.3580 11.1830
(2.18)** (2.22)** (1.86)*
Sweet potatoes 0.1031 0.0970 0.0672
(1.99)** (1.89)* (1.29)
Coffee -0.1368 -O.1385 -0.1342
(-2.08)** (-2.11)** (-1.95)*
Sorghum -0.0161 - —0.0167 -0.0701
(-0.81) (-0.84) (-3.29)**
Potatoes -0.0245 -0.0233 -0.0220
(-0.96) (-0.92) (-1.03)
Beer bananas -0.0739 -0.0736 0.1577
(-o.71) (-o.71) (1.76)
Non-BWS income/AE 0.0000 0.0000 0.0000
(0.23) (0.22) (1.22)
Cultivated land/AB 0.0207 0.0202 0.0050
(3.94)** (3.86)** (1.25)
Distance to market 0.1106 0.1088 -0.0285
(2.77)“r (2.73)** (-0.89)
Child/adult ratio -0.0862 -0.0772 -0.0001
(-1.31) (-1.18) (-0.00)
Education 0.0844 0.0819 0.1263
(1.24) (1.21) (2.19)**
Inputs -0.0000 -0.0000 0.0000
(-2.20)** (-2.14)** (0.60)
Northwest zone -1.7679 -1.7665 -0.3052
(~7.50)** (-7.50)** (-1.46)
Southwest zone -1.8134 -1.7506 —0.5183
(-6.27)** (-6.09)** (-1.96)**
North Central zone -0.5091 -0.5086 -0.2176
(-2.81)** (-2.81)** (-1.55)
South Central zone -0.4638 -0.4619 0.1297
(-2.39)** (-2.38)** (0.86)
% Correct 79.48 79.13 67.17
Predictions
Source: DSA HouseholdTSurvey, 1990,4Rigali, Rwanda
** = significant at .05 level = significant at .10 level

106

Table 4.4: Participation level: OLS results"

 

 

VARIABLE BANANA BANANA WINE WINE
PRODUCTION SALES PRODUCTION SALES
(N = 762) (N = 176) (N = 763) (N = 614)
Intercept 50449.00 -70923.00 119240.00 50017.00
(1.57) (-0.35) (1.90)* (0.89)
Sweet potatoes 464.49 -1903.70 905.38 -542.96
(1.69)* (-0.26) (1.71)* {-0.90)
Coffee -452.34 933.68 -1097.10 -421.48
(-1.22) (0.36) (-1.51) (-0.62)
Sorghum -203.43 -459.70 -410.68 -117.78
(-1.76)* (-0.32) (-l.81)* (-0.77)
Potatoes -157.23 51.87 -344.54 82.33
(-1.38) (0.10) (-1.57) (0.45)
Beer bananas -1442.30 162.45 -3062.80 -1073.30
(-3.12)** (0.10) (-3.41)** (-1.71)*
Non-banana income/AB 0.1677 0.0595 0.3500 0.1600
(5.57)** (0.23) (5.99)** (3.73)**
Cultivated land/AE 112.83 -65.43 211.20 34.95
(5.48)** (-0.22) (5.27)** (0.67)
Distance to market 167.32 216.45 356.16 -305.77
(0.95) (0.40) (1.04) (-0.96)
Child/adult ratio 111.91 -163.53 137.81 933.05
(0.34) (-0.16) (0.22) (1.93)*
Education -397.37 613.33 -870.28 -651.27
(-1.31) (0.26) {-1.47) (-1.47)
Inputs 0.2327 -0.1699 0.3676 0.2596
(5.32)** (-0.20) (4.29)** (3.81)**
Northwest zone 609.42 746.10 -6266.30 3578.20
(0.25) (0.13) (-1.31) (0.60)
Southwest zone -2200.80 6619.30 -6165.20 4528.60
(-1.54) (0.29) (-2.24)** (0.75)
North Central zone -2129.90 -3224.80 -5829.00 -714.23
(~3.01)** (-0.24) (-4.24)** (-0.51)
South Central zone -1508.10 4122.70 -2077.90 215.72
(-1.95)* (0.29) (-1.38) (0.16)
Lambda ~1622.00 11834.00 1539.10 ~742.62
(-0.51) (0.27) (0.25) (-o.11)
Adjusted R2 .3231 .2787 .3167 .1281

 

 

Source: DSA Household Survey, 1990, Kigali, Rwanda

** = significant at .05 level * = significant at .10 level

 

6t-ratios in parentheses.

107

Table 4.4: Participation level: OLS results7 (continued)

 

 

VARIABLE RATIO OF RATIO OF
WINE SALES WINE SALES WINE PURCHASES
TO NON-BWS TO WINE
INCOME PRODUCTION
Intercept -0.0947 16.7380 64876.00
(-0.05) (1.33) (0.34)
Sweet potatoes 0.0036 -0.0417 -1624.00
(0.18) (-o 31) (-1.48)
Coffee 0.0070 -0.1974 -716.71
(0.32) (-1.30) (-0.29)
Sorghum -0.0056 -0.0190 -140.12
(-1.10) (-0.51) (-0.12)
Potatoes -0.0025 0.0098 387.96
(-0.40) (0.22) (1.05)
Beer bananas -0.0066 -0.0888 -104.48
(-0.30) (-0.55) (-0.04)
Non—BWS income/AB -0.0000 -0.0000 0.0385
(-2.78)** (-0.15) (0.40)
Cultivated land/AE -0.0004 0.0147 19.95
(-0.20) (1.31) (0.24)
Distance to market -0.0241 0.0753 -339.41
(-2.25)** (1.05) (-o.71)
Child/adult ratio 0.0178 0.0021 609.69
(1.06) (0.02) (1.39)
Education -0.0137 0.0604 732.26
(-0.88) (0.55) (0.37)
Inputs 0.0000 -0.0000 0.0264
(2.46)** (-0.81) (0.38)
Northwest zone 0.3893 -2.1670 -541.18
(2.11)** (-1.78)* (-0.11)
Southwest zone 0.2422 -1.3546 7292.00
(1.25) (-1.09) (0.86)
North Central zone 0.0828 -0.4073 -7.30
(1.77) (-l.28) (-0 00)
South Central zone 0.0910 -0.3824 1281.30
(1.97)** (-1.19) (0.56) ‘
Lambda -O.2717 2.8232 5214.70
{-1.35) (2.12) (0.24)
Adjusted R2 .0324 .0861 .0771

 

Source:

** = significant at

.05 level

 

7t-ratios in parentheses.

DSA Household Survey, 1990, Rigali, Rwanda

significant at .10 level

108

by the possibility that sweet potatoes and beer bananas serve as better complements than
substitutes. In other words, it may be the case that higher sweet potato prices cause
farmers to grow more sweet potatoes Q91 beer bananas. This point warrants further
exploration.

The fact that banana growers with higher household incomes are more likely to
grow bananas than households with lower incomes may be attributable to richer
households needing fewer remunerative crops. They may simply be in a better financial
position to produce bananas and manufacture wine.

Contrary to Hypothesis 4, participation in banana production increases as
households distance themselves from markets. In retrospect, this is probably because there
is less opportunity to sell more remunerative crops; thus farmers grow more bananas and
manufacture wine for home consumption.

As for the finding that higher levels of purchased inputs significantly influence
households to 1_1pt_ grow bananas, this is probably due to the household preferring to grow
crops which require more intensive inputs (including land). Farmers are thereby
discouraged away from banana production, as they are equally discouraged from it by
living in the less climatically suitable westerly zones (in support of Hypothesis 5).

Once households decide to participate, however, several other variables become
relevant as influences determining leLeI of participation. Although sweet potato and
sorghum prices influence banana production levels as well as the decision to participate,

banana prices signal a decline in the quantity of bananas grown. This is not surprising

 

'A

 

 

109

since climatic influences which cause decreased production ultimately result in increased
prices.

On the other hand, land and inputs now join income as significant factors
influencing quantities of bananas grown. For households who have made the decision to
grow bananas, higher incomes imply a lesser need to rely on subsistence crops (and thus
an improved capacity to participate in banana production). Relatedly, increased land
makes it physically possible to grow more bananas, and input purchases (including labor
for banana tree maintenance) also increase production levels. As anticipated, habitation
in the central zones significantly decreases production quantities, although habitation in
the western zones does not result in a significant finding.

In response to Hypothesis 3, although the yes-no decision to grow bananas is
shown by the model to be independent of landholdings, magnitude of participation is
significantly influenced by this variable. Since virtually all households (90%) grow some
bananas (yet all do not have large landholdings), it makes sense that landholdings affect
how many bananas a household can grow but not necessarily its desire to grow bananas.

As predicted in Hypothesis 5, all households outside the East zone are less inclined
to grow bananas and produce significantly fewer bananas when they do. This is again
related to the more suitable agroclimatic conditions (altitude, rainfall, and soil quality)
discussed in Chapters 2 and 3. In other words, agroclimatic circumstances influence
production decisions, even ceteris paribus.

Finally, neither child to adult ratio nor level of education influences banana

production in any way. This most likely implies that children above a certain age are as

 

110

capable as adults of caring for banana trees. Banana production may also not require

exceptional skills provided only by additional education.

4.5.2 Participation in banana sales

No variables are found to significantly influence 1.6195 of participation in banana
sales. This is probably because the volume of sales was so small in 1990. In other
words, those households who participated in banana sales sold relatively few bananas and
did not exhibit sufficient variation among themselves to enable the identification of
significant sales influences. This section therefore concentrates exclusiver on the
variables which influence the decision to sell bananas.

To start with, two price variables are identified as significant influences. In
support of Hypothesis 1, increased sweet potato and sorghum prices decrease the
likelihood that households will sell bananas. Since income is increased by selling more
of these other crops at higher prices, households are better able to meet emergency
expenses without resorting to banana sales. This means that households are able to keep
their bananas for wine production, hopefully generating even higher revenues by
manufacturing and selling wine.

Unlike banana production, landholdings negatively influence the decision to sell
bananas. The more land per AE farmed by a household, the less inclined that household
is to sell any bananas. Although discordant with Hypothesis 3 (thereby supporting earlier
assertions that banana sales are truly an insignificant subsectoral component), this finding

makes sense in that more land implies more income from other crops. In turn, more

 

 

 

111

income implies a lesser need for the kinds of emergency funds described in Chapter 2,
funds which are commonly obtained by selling bananas. In other words, larger
landholdings indirectly enable households to participate more frequently in wine
production (by not having to sell bananas), as the next section shows.

In relation to Hypothesis 4, market proximity is also shown to have no influence
on banana sales. Although my hypothesis was that households closer to markets would
be more inclined to sell bananas, it seems that banana sales are on average so small that
markets play no role. Despite the banana market observed near Kigali, most banana sales
are apparently conducted in an informal setting, probably between neighbors. Banana
sales are again marked as the least interesting component of the banana wine subsector.

The data offer no support to the banana sales aspects of Hypotheses 2 and 5.
First, income exerts neither a significant positive nor negative influence on a household’s
decision to sell bananas. Second, habitation outside the East zone (particularly in southern
Rwanda) actually increases the likelihood that households will participate in banana sales.
In retrospect, this is not surprising since average households produce fewer bananas in
these zones. Not having adequate numbers of bananas with which to manufacture wine
(i.e., quantities which justify extensive labor use), it is possible that households opt to sell
bananas with greater frequency.

In addition, education and age of the household head have no impact on banana
sales, nor does the household’s child-to-adult ratio. These findings may be due to the low
volume and random nature of banana sales, which make it difficult to assign variable

cause.

 

112

4.5.3 Participation in wine manufacturing

Both the PROBIT and OLS results for wine production are nearlv identical_tc_>
those for banana production. This is not surprising in light of the fact that the Spearman
correlation between banana production and wine production is .98 (see Table 4.2). Since
only a small percentage of Rwanda’s bananas are ever sold and virtually all beer bananas
are used to manufacture wine, these regressions are expected to be very similar.
Nevertheless, I will briefly review the findings - although their interpretation is the same
as that found in the description of banana production results (see Section 4.5.1).

First, the same crop prices influence wine production as banana production, and
in the same magnitudinal directions. Second, higher incomes influence households to
manufacture wine. Higher incomes also influence households to produce wine in greater
quantities (supports Hypothesis 2). Third, landholdings do not influence the decision to
manufacture wine. Rather, larger landholdings influence households to produce greater
quantities of wine, undoubtedly in response to the quantities of bananas produced by these
households (supports Hypothesis 3). Fourth, households outside the East zone
(particularly in the Southwest and North Central zones) are inclined to produce less wine
than households in the East zone (supports Hypothesis 5). Again, neither education nor

child to adult ratio impacts wine production.

 

113

4.5.4 Participation in wine sales

In addition to agroclimatic zone, the decision to sell wine is significantly
influenced by four variables. Coffee price and purchased inputs exert a negative effect;
landholdings per AE and distance to market a positive effect.

The first finding, related to the price of coffee, is not surprising. High prices
induce households to plant more coffee as a cash crop. Thus, coffee inhibits banana
production, with its revenues precluding the necessity of selling wine for cash.

Purchased inputs, on the other hand, are used more readily to promote crops other
than bananas. Therefore, purchased inputs imply at least one other crop which yields
revenues for household use. As a result, wine sales become a less important source of
cash.

As for landholdings, it is not surprising that more land implies a higher rate of
participation in wine sales. After all, if banana production is high, wine production is also
(recall the .97 correlation between banana and wine production), thus implying an
available surplus of wine for sale.

An anomolous finding is noted regarding proximity to the nearest market. In
Hypothesis 4, it was predicted that households located closer to markets would be more
inclined to participate in wine sales. Among those who participate, it was predicted that
the further away the market, the smaller would be the m of wine sales. Although
the latter expectation is significantly borne out by the data, distance to market is found
to positively influence the decision to participate. In other words, the further away the

market, the more likely the household is to sell wine. I expect this is because the local

114

market for banana wine is more reliable, requiring fewer transaction costs, than a distant
market. Also, as it becomes more and more difficult to transport other kinds of
commodities to a distant market, reliance on wine sales as a source of income grows.
Although most households are self-sufficient in a number of subsistence crops, they are
generally not self-sufficient in wine“. Located far from markets, households need to earn
cash somehow. Participation in wine sales, to whatever degree, alleviates some of this
problem.

The last two influences concerning the decision to participate in wine sales involve
purchased inputs and agroclimatic location. Each of these variables induces a household
to sell less wine. Decreased willingness to participate as a function of purchased inputs
is consistent with all other BWS participation decisions reported so far. The zonal
finding, on the other hand, supports Hypothesis 5.

In terms of variables which influence levels of wine sales’, no price variables,

 

with the exception of beer banana price, exert a statistically significant influence. As
explained before, the higher banana price is likely attributable to a poor harvest of
bananas. A poor harvest implies less wine production and less wine available for sale -
thus pushing up the wine price (in this case, proxied by the banana price).

Among those who sell wine, the finding that increased income results in decreased

participation levels in wine-selling activities may be explained by the fact that those with

 

8This is mostly because of its short shelf-life, extreme labor intensity, and popularity
throughout Rwanda. What this implies is a lot of households buying and selling a lot of
wine.

9Among those who participate in wine-selling activities.

 

5‘1

115

more money, shown to produce more wine as income rises, are less financially strapped
than lower income households. Thus, preferring to consume more wine or give it away
as gifts, they are less in need of the revenues which wine sales generate. This finding
negates Hypothesis 2.

The wine sale regressions yielded no information concerning Hypotheses 3, 4, and
5. Neither landholdings, distance to market, nor agroclimatic location exerted a
significant influence on participatory decisions relating to wine sales. Likewise, education
exerted no influence, although the child to adult ratio positively influenced wine sales -

probably the result of seeking additional revenues for child support.

4.5.5 Supplementary findings

To supplement the above wine sales findings, for which the OLS regression
structure only accounts for 13% of the model’s variability (Adjusted R2 = .1281), I am
also running three additional pairs of regressions. The first includes ratio of wine sales
to non-BWS income as the dependent variable; the second includes ratio of wine sales to
wine production. The third examines wine purchases as the dependent variable, to serve
as a complement to wine sales. Regression results are shown in Tables 4.3 and 4.4.

The first two pairs of regressions, which respectively control for wine sales with
regard to income and production levels, yield virtually identical results regarding the
decision to participate in wine sales. In addition, they differ only minimally from the
original wine sales regression, in that sweet potato prices join coffee prices in exerting an

influence on wine sales participation. What this tells me is that my original wine sales

 

116

PROBIT regression is fairly reliable, even without controlling for income or production
levels. The number of correct predictions is 79% in all three cases.

As for participation levels, although ratio of wine sales to non—BWS income
appears to yield more interesting results than wine sales to wine production (in which case
only one variable appeared minimally significant), its structure accounts for only 3% of
the sample’s variability (R2 = .0324). This compares to 8.6% of variability in the second
modification. Again, these results point to the finding that neither ratio is any better than
the definition used in the original wine sales OLS equation.

Last, although the wine purchase regressions yield no statistically significant
influences for participation levels, the decision to participate in wine purchases is affected
by a number of variables. Even though this PROBIT model predicts wine buyers with
only 67% accuracy, coffee and sorghum prices, habitation in the Northwest zone, and
level of education appear to influence this decision to buy.

The negative relationship between coffee and sorghum prices may be explained
by the fact that these are two frequently sold crops. In other words, as their prices go up,
income from their sales also rise, possibly enabling households to purchase higher-grade
alcoholic beverages. If so, wine purchases may decline. Habitation in the Northwest zone
may cause decreased numbers of households to purchase wine because of a preference for
sorghum wine.

Note how different these wine purchase findings are from those related to the
decision to sell wine. In particular, income, landholdings, and market proximity no longer

influence a household’s decision to participate, although education does. The income and

 

117

landholding findings imply that wine buyers cannot be differentiated in terms of
household assets or financial status. As for education, although wine purchases is the only
category in which education plays a role, I am at a loss to explain this finding. Since
education in no way impacts production or sales of bananas or wine, I can only surmise
that it influences purchases through an increased ’ability’ to participate in market
transactions. Recalling that education rarely extends beyond some primary school, and
most transactions are informally conducted between neighbors (remember that market
proximity plays no role in the decision to purchase wine), this explanation is not very

satisfactory.

4.6 Conclusion

Having constructed models which predict participation in banana and wine
production with greater than 90% accuracy, I am justified in saying that income, distance
to market, purchased inputs, and zonal location are the most significant determinants of
participation in banana and wine production. Sweet potato, sorghum, and potato prices
also play a role. Although income and inputs maintain their effectiveness in the decision
about what quantity of bananas to grow or wine to produce (among those who decide to
participate), landholdings unsurprisingly replace market proximity as a primary factor in
the decision about how much to participate.

My banana and wine sales PROBIT models are able to predict participating
households with 80% accuracy. A. number of statistically significant variables are

identified as influences. Sweet potato and sorghum prices affect banana sales decisions,

 

b.

118

as do landholdings, purchased inputs, and zonal location. Only coffee prices influence
wine sales, along with landholdings, distance to market, inputs, and zonal location. Note
that, excluding prices, distance to market is the only factor which alters the list of
variables influencing these two kinds of sales decisions.

Last, my models for banana and wine sales participation levels leave much room
for improvement. No statistically significant variables are identifiable in the former case;
income and inputs are the only robustly significant factors in the latter case. In reality,
banana sales are most influenced by banana production levels, and wine sales by wine
production levels. Although possible to create sales models in which production serves
as an independent variable, the hazards of endogeneity and potential biasing increase
significantly. Since such modeling has the tendency to become very complicated, such
an attempt is best left to a Doctoral rather than Masters thesis. Nevertheless, I feel that
the content of the models used in this thesis are adequate for the purposes of this work.

In the following chapter, I use the above findings to establish a set of policy

recommendations regarding the banana wine subsector.

CHAPTER FIVE

CONCLUSIONS AND POLICY IMPLICATIONS

5.1 Policy Implications

The question to address now is whether or not the above research findings justify
the implementation of policies to promote Rwanda’s BWS. Assuming that the intention
of such policies is to stimulate the economy by augmenting incomes for the greatest
number of rural households, it is crucial to understand that not all households have the
capacity to participate equally.

Interest in the BWS was first generated by the informal observation that so many
households in Rwanda grow bananas and manufacture wine. Upon examining the data,
this figure was shown to be 90%. Nevertheless, as presented in Chapters 2 and 3,
although 80% of all wine producers (rich and poor alike) sell some wine, relatively rich
households produce and sell the largest quantities. Thus, in its present condition, the
BWS is most remunerative for those who are already better off financially. These are
also the households with the most land.

Assuming it were possible to target exclusively poor households with some form
of assistance, two things are important to remember. First, most poor households are

located away from the banana-abundant East zone (mostly in southern and western

119

 

120

Rwanda). Thus, they are situated in agroclimatic zones inherently less favorable to the
BWS. Second, assuming equal banana-growing potential, poor households have generally
smaller landholdings and thus decreased capacity to expand banana production. In other
words, even if it were possible to drastically improve Rwanda’s transportation
infrastructure - to enable households easier, less expensive transportation of wine to
nearby markets - it is unlikely that Rwanda’s poorest households would benefit.

There is also the issue of wine preservation. As learned from the supplemental
survey, the shelf-life of wine is presently two to three days. Purchases of large quantities
of wine by cabaret owners are therefore risky in rural areas (where demand is generally
smaller), and transportation of the product to distant locales (i.e., urban areas) even
riskier. If the product spoils ahead of time, either the producer, wholesaler, or cabaret
owner loses money.

Although it may be possible to correct this problem through alternative
preservation means, large wine sellers (i.e., richer households) are more likely than small
wine sellers to benefit'. Sales would increase, especially in urban areas, and sellers
would be in a position to withhold production in favor of more competitive wine prices.

It is unquestionable that the urban market has room for expansion. As migrant

workers (young men in particular) move from rural to urban areas seeking new

 

1During the supplemental survey, we encountered one large wine producer,
Zigirinshuti Vincent, who operates his own bottling plant and actually exports banana
wine to other parts of Africa and Europe. Having devised a simple method for
preserving banana wine over several years, Mr. Zigirinshuti would not reveal his secret
to us, but stated (with a smile) that even the poorest households in Rwanda have the
capacity to preserve wine as he does.

 

121

opportunities, their taste for banana wine does not change (attested to by the rapidly
growing wine market outside Kigali). However, their ability to manufacture it does.
Lacking inputs and time, urban residents rely more and more heavily on the wine supplied
by city cabarets.
Therefore, a solid campaign to increase the availability of wine to urban areas has
potential. This campaign would require:
1) village outreach to teach people about the income merits of
selling wine to urban areas (as opposed to consuming it at
home),
2) extensive research into (and dissemination of knowledge
aboUt) improving the shelf-life of wine and regulating its
quality.
3) improved infrastructural services to transport bananas and/or
wine (including the possible development of wine-
manufacturing cooperatives, like that observed at Gisenyi),
and
4) marketing services to establish regular buyers in Kigali and
other urban areas.
The costs associated with such an effort are undoubtedly high and, as indicated
above, the outcome questionable. Rich households with more land are likely to benefit
most, and poor households to fall further behind financially. Nevertheless, it is possible

that upstream and downstream opportunities (associated with urban-oriented growth of the

 

122

BWS) would be made available to poor households (i.e., as input suppliers, laborers, or
wholesalers).

A second alternative is to seek new uses for bananas and banana products. Since
beer bananas and banana wine are such an important part of Rwandan culture, it is
unrealistic (anytime soon) to consider replacing their majority with crops or activities
which may be better-suited to microenterprise development and household-level economic
growth. Also, since bananas are cited as one of Rwanda’s foremost anti-erosive crops
(Kangasniemi and Reardon, 1994) and the residue from wine production is regularly used
as mulch to preserve soil quality (and as animal fodder), the environmental role of beer
bananas (and wine) is clear. Thus, alternative uses for bananas - like manufacturing juice
instead of wine or using the banana leaves to make paper-like products - would continue
to protect farm environments, stimulate household incomes through other kinds of
microenterprise development, and inhibit wine consumption. Over time, such new uses
may even encourage farmers to replace beer bananas with fruit or cooking bananas.

These topics will be resumed when I discuss future research possibilities.

5.2 Methodological Conclusions

The joint implementation of subsectoral and household-level findings has
contributed significantly to this research. For example, although the household-level data
show that the volume of banana sales throughout Rwanda was small in 1990, the
subsectoral findings point to a few key banana trading routes (primarily towards Kigali

and Butare from the East zone) which would have been missed by examining the

123

household-level data alone. Thus, even though banana sales are considered the least
interesting component of the BWS for r_n_opt_ of Rwanda, banana trading is quite important
in a few key locations.

In addition, the 1990 household-level data focused exclusively on rural
landholders. Had it not been for the 1993 subsectoral component of this project, the
urban demand for wine would have gone unnoticed and an entire policy option omitted
from the discussion.

Although my recommendation is to research and promote banana-based activities
beyond the scope of wine, the Rwandan government may opt to do otherwise. If so, the
subsectoral findings will inform decision-makers about the locus of demand for wine in
Rwanda’s urban locations. Analogously, the household-level findings will realistically
identify the likely winners and losers of a policy aimed at strengthening the banana wine

subsector. In any case, knowledge is enhanced.

5.3 Further Research

Strategically, the goal of this thesis has been to evaluate the potential income-
enhancing attributes of participation in Rwanda’s banana wine subsector. In light of my
findings that the sales volume of bananas is small and wine sales concentrated, it is
difficult to imagine that the encouragement of wine production could significantly shift
income into the hands of those most food insecure.

I therefore suggest that future research examine alternative uses for beer bananas -

uses which could be promoted by the government, thus encouraging the initiation of a

 

124

number of different microenterprise opportunities throughout rural Rwanda. As
environmentally necessary as bananas are towards combatting soil erosion, and as
imperative as it is for Rwandan households to seek non-farm sources of income, research

of this kind is likely to yield very satisfactory results.

5.4 Summary

The banana wine subsector is undoubtedly an important component of everyday
life in Rwanda. Nevertheless, its potential for augmenting household incomes and
redistributing wealth among rural communities is dim. Although banana and wine
production are common, banana and wine sales are rarely used as opportunities to enhance
household wealth. Households which are in a position to sell large quantities of wine are
generally well-off and land-rich. Policies geared towards enhancing the BWS would thus
help this segment of the population rather than those truly in need of food security.

Although Rwanda’s urban demand for wine is presently on the rise, the financial
and social costs associated with supplying this market are potentially high (especially in
light of Rwanda’s recent past). Thus, Rwanda’s best option may be to seek alternative
uses for beer bananas and other banana tree products (i.e., leaves). The identification of
such alternatives would encourage the continued planting of banana trees - one of
Rwanda’s best erosion control measures. It would also identify more suitable ME
activities for promotion in rural areas and perhaps inhibit the over-consumption of banana

wine.

 

125

There is no doubt that banana wine will continue to thrive in Rwanda, even in the
aftermath of its horrific war. In a recent letter from my Rwandan counterpart, Vincent
Habimfura, presently living in a refugee camp near Bukavu, Zaire, "... les Rwandais
continuent a survivre de la biére de banana: achat de bananes, fabrication et vente de la
biere de banane entre eux, dans differents camps. Les problemes restent presque le méme,

le plus grand étant toujours la conservation"2 (letter from Zaire, 1995).

 

2In English: "Rwandans continue to survive with the help of banana wine: banana
sales continue, as does the fabrication and sale of wine among refugees, even between the
different refugee camps. The problems with wine are almost the same in the camps as
at home in Rwanda, the greatest problem being conservation".

 

APPENDIX A

 

APPENDIX A
1993 Survey on Participation in the Banana Wine Subsector
by

Vincent Habimfura
Pamela Riley

DECISION TO PRODUCE AND SELL BANANA WINE
1. Depuis combien de temps produisez-vous la biére (b.b.) pour vendre?
2. Comment étes-vous devenus producteur de la b.b.?

(1) Heritage d’un terrain

(2) Heritage d’un bananeraie
(3) Achat d’un terrain

(4) Achat d’une bananeraie
(5) Autres

3. En cas d’achat, comment avez-vous pu trouver l’argent?

(l) Crédit
(2) Salaire
(3) Agriculture
(4) Commerce
(5) Autres

4. La production de la b.b. semble une activité relativement refissi?
(l) Oui (2) Non
Si oui, quelles sont les raisons de cette réussite?
(1) Grande bananeraie
(2) Bonne Bananeraie
(3) Un bon entretien

(4) Main d’oeuvre bon marché
(5) Autres

126

 

127

Exercez-vous d’autres activités rémunératrices en déhors de la production de la
b.b.?

(1) Autres cultures vivrieres
(2) Cultures industrielles
(3) Commerce

(4) Artisanat

(5) Autres

Y aurait-il des activités génére'es par la production de la b.b.?
(l) Oui (2) Non

Si oui, lesquelles?

Sont-elles rentables par rapport a la b.b.?

(1) Oui
(2) Non
(3) Egalement rentables

BANANA WINE PRODUCTION

8.

Y a-t-il une variation de la production de la b.b. au cours de l’anne'e?
(1) Oui (2) Non

Si oui, combien de mois dure (1) la bonne saison:
(2) la mauvaise saison:

Qu’est-ce qui explique cette difference?
(1) Changements clirnatiques saisonniers:

(a) Quand il y a abondance de pluies et que la bananeraie croit (de
a ...), i1 y a une petite production, mais quand vient la saison séche
(de a ...) il y a un murissement rapide des regimes de banane, la
production devient abondant.

(b) Autres.

 

128

9. Combien de fois fabriquez-vous la biere de banane par mois?

(1) Pendant la mauvaise saison:
(2) Pendant la bonne saison:

10. Combien de régimes de banane utilisez-vous chaque fois?
(l) Pendant la mauvaise saison:
(l) Gros: (2) Moyens: (3) Petits: (4) Total:
(2) Pendant la bonne saison:
(l) Gros: (2) Moyens: (3) Petits: (4) Total:
11. Quel le prix du regime de banane:
(1)Gros: (2) Moyen: (3) Petit: (4) Kilo:
12. Quelle est la production annuelle moyenne des bananes a biere?

Formule: [(8.8.l)*(9.1)*(10.1.4)] + [(8.12)*(9.2)*(10.2.4)]

13. Quelle est la tendance de la production moyenne des bananes a biere les 6 demiers
mois par rapport a la production normale?

( 1) Hausse
(2) Baisse
(3) Stable

14. Quelle est la tendance de la production des bananes a biere par rapport a l’année
passée:

A 1992? (1) Hausse (2) Baisse (3) Stable
A 1991?

A 1990?

 

15.

16.

17.

18.

19.

20.

21.

22.

129

Quelles sont les contraintes a la production des bananes a biere?
(1) Manque de pluies
(2) Maladies
(3) Autres

Combien de jerricans (litres) produisez-vous chaque que vous faites la b.b.?

(1) Pendant la mauvaise saison?
(2) Pendant la bonne saison?

Quelle est la production annuelle moyenne de la b.b.?
Forrnule: [(14.1)*(9.1)*(8.1)] + [(14.2)*(9.2)*(8.2)]
Quand est-cc que vous avez produit la b.b. la derniére fois?
(1) une semaine
(2) deux semaines
(3) trois semaines
(4) un mois
Combien de jerricans (litres) avez-vous produit la derniere fois?
Combine (quel pourcentage) de jerricans (bouteilles) avez-vous:
Vendus?

Autoconsomme’s?

Quelle est la tendance de la production de la biere de banane les 6 derniers mois
par rapport a la production normale?

(1) Hausse
(2) Baisse
(3) Stable

Quelle est la tendance par rapport a l’année:
1992? (1) Hausse (2) Baisse
1991?

1990?

 

130

23. Y a-t-il des contraintes a la production de la b.b.?
(1) Bananes a biere
(2) Marché de la b.b.

(3) Temps
(4) Autres

BANANA BEER PRICE
24. Y a-t-il une variation saisonniere des prix de la b.b.?
(1) Oui (2) Non
Si oui, raisons?
(1) Variation de la production, grand prix pendant la mauvaise saison
et petit prix pendant la bonne saison
(2) Demande
(3) Autres
25. Quelle est le prix d’un jerrican (une bouteille) pendant
(1) la bonne saison?

(2) pendant la mauvaise saison?

26. Quelle était le prix moyen d’une bouteille de b.b. pendant les 6 derniers mois?

27. Quelle est la tendance du prix de la b.b. par rapport a l’anne'e
1992? (l) Hausse (2) Baisse (3) Stable
1991?

1990?

28.

29.

30.

131

A qui vendez-vous régulierement votre biere de banane?

(1) cabaretiers ambulants a domicile

(2) votre cabaret

(3) un cabaretier voisin permanent du centre
(4) n’importe quel cabaretier du centre

(5) marché voisin

(6) acheteurs a domicile

(7) autres

Quelle est la tendance des revenues de la vente de la b.b. ces 6 derniers mois par
rapport a la normale?

(1) H (2) B (3) S
Quelle est la tendance des revenus de la b.b. par rapport a l’anne'e:
1992? (1) H (2) B (3) S
1991?

1990?

QUALITY OF BANANA WINE

31.

32.

33.

Combien de qualités de b.b. produisez-vous?
( 1) une (2) deux (3) plus
De quoi depend la qualité de votre biere?

(1) Dilution

(2) Qualité de la terre

(3) Abondance de la bananeraie
(4) Prix de chaque qualité

(5) Buts de la production

(6) Autres

Quelle qualité pensez-vous rentable a produire?

 

132

INPUTS REQUIRED FOR BANANA WINE PRODUCTION

34.

35.

36.

37.

38.

39.

40.

41.

Quelles activités nécessaires pour produire la b.b?
(1) Entretien de la bananeraie
(2) Fabrication de la b.b.
(3) Puiser de l’eau
(4) Autres

Combien de fois entretenez-vous votre bananeraie par an?

Qui entretien votre bananeraie?
(1) Les membres de la famille
(2) Main-d’oeuvre
(3) Parentes
(4) Autres
Combien de temps (homme-jours) la famille dépense a l’entretien de la
bananeraie?
Pendant combien de temps (homme-jours) utilisez-vous la main-d’oeuvre?

Combien payez-vous pour la main-d’oeuvre (homme-jours) par jour?

Comparer les frais de main-d’oeuvre des 6 derniers a la normale?
A l’année 1992? (1) H (2) B (3) S
A 1991?
A 1990?

Utilisez-vous les bananes a biére a d’autres fins que la production de la biére de
banane?

(1) Vendre
(2) Manger
(3) Cadeaux
(4) Autres

 

 

 

42.

43.

44.

45.

46.

47.

133

Utilisez-vous des re'gimes de banane pour la fabrication de la b.b. venant en dehors
de l’exploitation?

(l) Oui (2) Non
Si oui, d’ou viennent-ils?

(1) Achat (2) Emprunt (3) Cadeaux (4)_Autres
Combien?

(1) Gros:

(2) Moyens:

(3) Petits:

Valeur?
Quels sont d’autres intrants utilisez-vous pour la fabrication de la b.b.?

(1) Sorgho (2) Eau (3) Bois de chauffage (4) Autres
Comment obtenez-vous ces intrants?

(1) Produits (2) Achete's (3) Emprunt (4) Autres

Quelle est leur valeur?

Utilisez-vous la main d’oeuvre pour la fabrication de la b.b.?
(1) Oui (2) Non
Si oui, combien payez-vous pour la main d’oeuvre une fois?
Utilisez-vous la main d’Oeuvre pour la vente de la b.b.?
(1) Oui (2) Non

Si oui, pour combien chaque fois?

 

134

48. Quels sont les revenue nets de la b.b.?
Formule: Total outputs - total inputs
1. Total Inputs:
2. Total Outputs:
49. Comment utilisez-vous vos revenues de la b.b.?
(l) Réinvestis dans la production de la b.b.
(2) Achats des biens de premiere nécissité
(3) Investis dans d’autres activités productrices
(4) Faire l’épargne
(5) Autres
Si épargne, que comptez-vous faire de l’épargne?
(1) Augmenter la production de la b.b. dans l’avenir
(2) Pour la mauvaise récolte
(3) Pour améliorer les conditions de vie, d’hygiéne et sanitaires
(4) Pour l’avenir des enfants

(5) Pour d’autres activités plus productrices
(6) Autres

BANANA WINE PURCHASES
50. Achetez-vous souvent de la b.b.?
(1) Oui (2) Non
Si oui, combien de fois par mois?

Quelle quantite' achetez-vous en moyenne?

51. Achetez-vous de la b.b. pour revendre?
(1) Oui (2) Non

Si oui, combien de jerricans par mois?

52.

53.

54.

135

Oil achetez-vous régulierement la b.b.?

(1) cabaret voisin permanent
(2) n’importe quel cabaret voisin

Quel est votre mode de payement?
(1) cash (2) credit biére (3) autre
Quelle est la tendance de vos achats de la b.b.?

(1) H (2)13 (3) s

COFFEE PRODUCTION

55.

56.

57.

58.

Produisez-vous du cafe?
(1) Oui (2) Non

A combien estimez-vous les revenues annuels du café?

Entendez-vous:

(1) augmenter
(2) diminuer
(3) continuer

a produire 1e café dans les proportions actuelles?

Quelles sont les contraintes a la production du cafe?

(1) Soleil

(2) Manque de paillage
(3) Prix qui dirninue
(4) Difficile a produire
(5) Culture d’épargne
(6) Maladies

 

 

136

FUTURE

59.

60.

61.

62.

63.

64.

Voudriez-vous extendre votre bananeraie?
(1) Oui (2) Non
Y a-t-il des contraintes a l’extension?

(1) terrain
(2) temps
(3) maladies
(4) autres

Votre domaine agricole a-t-elle tendance a diminuer ces derniers temps?
(1) Oui (2) Non

A cause: (1) vente
(2) heritage des enfants
(3) éboullements
(4) autres

Si elle diminuait de moitié, quelles cultures pensez-vous continuer a produire?
(1) bananeraie
(2) autres cultures vivrieres
(3) cultures industrielles

Pensez—vous continuer et extendre la culture de la bananeraie?

(1) Oui (2) Non

Souhaiteriez—vous que vos enfants entinuent a exercer l’activité agricole et cultiver
la bananeraie?

(1) Oui (2) Non

Si non, quelles autres activités voudriez-vous que vos enfants exercent dans
l’avenir?

(l) Commerce
(2) Agents
(3) Autres activités rémunératrices

 

 

65.

137

Quel niveau d’études leur souhaiteriez-vous?
(1) Supérieur
(2) Moyen
(3) Qui leur permet de vivre

Et que faites-vous pour les y preparer?

(1) Epargne
(2) Investissement

DEMOGRAPHIC INFORMATION

66.

67.

68.

69.

70.

71.

La taille de votre ménage se serait-elle augmentée durant les trois dernieres
années?

(1) Oui (2) Non
Louez-vous la terre des autres?
(1) Oui (2) Non
Louez-vous la terre aux autres?
(1) Oui (2) Non
Avez—vous vendu la terre ces trois dernieres anne’es?
(l) Oui (2) Non
En avez-vous acheté?
(l) Oui (2) Non

Combien de gens vivent actuellement dans votre foyer? Leur genre, age, et
statut?

Nombre: Filles: Garcons: Age (du petit au grand):

APPENDIX B

APPENDIX B

Description of Rwanda’s Five Agroclimatic Zones

There are three regional classification schemes that are used for various purposes
by researchers and policy-makers in Rwanda. All three are based on differences in soils,
altitude and rainfall, and as such also show marked differences in cropping patterns, farm
size, livestock ownership, and other important household and regional characteristics. The
first was developed by Delepierre (1974), and divides the country into 12 agro-ecological
regions. More recently, the CNA has expanded the number to 18. This classification
scheme draws upon a more comprehensive data base, particularly in soil characteristics,
and has been useful for targeted, commune- and secteur-level development projects. A
third classification scheme (Clay and Dejaegher 1987) has been devised to capture the
major delineating characteristics of the first two, while summarizing these differences in
just five zones that can be used effectively for national level socioeconomic (rather than
purely agronomic) analysis. The five-zone classification is judged to be the most suitable

for our purposes; some of the defining characteristics of these zones are described below.

NORTHWEST ZONE
This zone comprises the prefecture of Gisenyi, part of Ruhengeri and part of
Kibuye. It has mostly volcanic, fertile soils that are highly susceptible to erosion. Its high

altitudes mean abundant rainfall and cooler temperatures. Major cash crops are coffee,

138

 

139

white potatoes, and pyrethrum. Bananas are not grown at elevations above 2,000 meters.
Staple food crops include maize, sorghum, and beans. The North-West includes both
temperate highlands with fertile and/or recently cleared volcanic soils and well-watered
lowlands by the lake Kivu. Much of the zone is very densely populated and the typical

agricultural working day is longer than it is elsewhere in Rwanda.

SOUTHWEST ZONE

The Southwest region comprises Cyangugu, the southern part of Kibuye and
western part of Gikongoro prefectures. It is characterized by high altitudes, steep slopes,
and high rainfall, with concomitant soil erosion and soil acidity problems. Soils have a
high proportion of clay and range from poor to moderately suitable for agriculture. A
substantial but diminishing part of the Southwest Zone is covered by a natural, "protected"
forest. Major cash crops are bananas and coffee. The most important food crops are
beans, sweet potatoes, "colocase", and cassava. Soilsare poor, and sometimes degraded
and acidic on the steep slopes of the Zaire-Nile divide; soils are fertile on the coast of
lake Kivu. Although not as densely populated as the North-West, the pressure on
resources is higher in the South-West, which is the poorest zone in Rwanda. The zone is

not self-sufficient in food and depends on imports from Zaire and Burundi.

NORTH CENTRAL ZONE
The region covers part of Ruhengeri, Byumba, and Kigali. Its a region well

known for its high mountains with very steep slopes, a region susceptible to erosion too.

140

Maj or cash crops are potatoes, wheat, and coffee in the southern part. Food staples include
beans, peas, sweet potatoes, maize, and sorghum. The zone is less densely populated than
the Center-South and some northern and eastern parts of it have been agricultural frontier

until recently. Agroclimatically it is quite similar to the Center-South.

SOUTH CENTRAL ZONE

This region encompasses much of the prefectures of Gitarama, Butare, and .
Gikongoro. It has sandy-loam soils, and serious degradation. Soil fertility ranges from
very poor to moderately suitable for agriculture. It is a region of well-watered marshes
which allow a third cropping season. Major cash crops are bananas and coffee while
favored staples are beans, sweet potatoes, cassava, and sorghum. The Center-South
includes the historical center of the country and much of it has struggled with high
population densities for a long time. Both emigration to other parts of Rwanda, farmers’
subjective assessments, and yield levels suggest that what has been prime agricultural land

has degraded during the past decades.

EAST ZONE
This zone corresponds to the entire prefecture of Kibungo and the eastern parts of
Kigali and Byumba. It is a region with gentle slopes and relatively low altitude. Rainfall
is less here than in the higher elevation zones to the west. Because it is drier, this eastern
plateau was traditionally reserved for pastoral uses. Though it is densely settled today,

farms are larger here than in the older western zones that became occupied several

 

 

141

generations earlier. Households in this region rely principally on coffee and bananas as
cash crops and sorghum, beans, and cassava as major staples. Drier and warmer than
the rest of Rwanda, the East has been the last agricultural frontier of the country. Outside
the Akagera national park little cultivable land remains unused, but farms are much larger

in this zone than elsewhere in Rwanda.

BIBLIOGRAPHY

BIBLIOGRAPHY

Birasa, E.C. et a1. (1992). Carte d’Aptitudes des Sols du Rwanda: Me'thodologie, Légende
et Unite's Cartoggaphigues. Republique Rwandaise - Ministere de l’Agriculture et

de l’Elevage, March.

Boissiere, M., J.B. Knight, and RH. Sabot. (1985). "Earnings, schooling, ability, and
cognitive skills". American Economic Review. Vol. 75, No. 5, pp. 1016-30,
December.

Boomgard, James 1., SP. Davies, S. Haggblade, and DC. Mead. (1989). Subsector
Analysis: Its Nature, Conduct, and Potential Contribution to Small Entepprise
Development. Michigan State University Department of Agricultural Economics,
International Development Working Paper No. 26.

Boughton, Duncan, J.M. Staatz, and JD. Shaffer. (1994). Analyzing the Impact of
Structurpl Adjustment on Commodity Subsectors: Currency Devaluation and the
Maize Subsector in Mali. Michigan State University Department of Agricultural
Economics, Staff Paper No. 94-71, October.

 

Chuta, Enyinna and C. Liedholm. (1979). Rural and Non-Farm Emploment: A Review
of the State of th_e Art. MSU Rural Development Paper No. 4.

Clay, Daniel and T. Reardon. (1994). Determinants of Farm-Level Conservation
Investments in Rwanda. Contributed Paper to the International Association of
Agricultural Economists for presentation at August 1994 meetings, Harare.

Division des Statistiques Agricoles. (1991). Enguete Nationale Agg'cole 1989:

Production, Superficie, Rendemeng, Elevage et Leur Evolution 1984-89.
Republique Rwandaise, DSA Publication No. 22, November.

Ellis, F. (1988). Peasant Economics: Farm Households and Agrarian Development.
Cambridge: Cambridge University Press.

Glick, Peter and DE. Sahn. (1993). Labor Force Participation, Sectoral Choice, and

Earnin s in Con Guinea. Working Paper 43, Cornell Food and Nutrition
Policy Program, June.

142

Goldberger, Arthur S. (1991). A Course in Econometrics. Cambridge, MA: Harvard
University Press.

Habimfura, Vincent and RR Miklavéic. (1995). "Le Vin de Banane dans l’Economie
Rurale au Rwanda: Contraintes et Strategies de Developpement" (unpublished).

Haggblade, Steven J. (1984). The Shebeeen gueen or sorghum beer in Botswana: the

 

impact of factory brews on gottage indps_‘tr_y. Ph.D Dissertation, Michigan State
University.

Haggblade, Steve. (1987). "Opportunities for enhancing performance in Rwanda’s
alcoholic beverage subsector". August.

Heckman, James. J. (1979). "Sample selection bias as a specification error".
Econometrica. Vol 47, pp. 153-61, January.

International Monetary Fund. (1993). Government Finance Statistics Yearbook.

Kangasniemi, Jaakko and T. Reardon. (1994). Demoggaphic Pressure and the

Sustainability of Land Use in Rwang, Contributed Paper to the International
Association of Agricultural Economists for presentation at August 1994 meetings,

Harare.

Kilby, Peter and Liedholm, C. (1986). "The role of nonfarm activities in the rural
' economy". World Congress of International Economics Association. New Delhi.
December.

Kimhi, Ayal. (1994). "Quasi Maximum Likelihood Estimation of Multivariate Probit
Models: Farm Couples’ Labor Participation". American Journal of Agg'cultural
Economics. Vol 26, November, pp. 828-835.

Liedholm, Carl. (1990). The Mamics of Small-Scale Indu_s_tgy in Africa and the Role
of Policy, Michigan State University, 1990.

Liedholm, Carl and DC. Mead. (1987). Small-Scale Industries in Developing Countries:
Empirical Evidence and Policy Implications, MSU International Development
Paper No. 9.

Liedholm, Carl and DC. Mead. (1990). Dypamics of Microenterprises: Research Issues
and Approaches, GEMINI/MSU, 1990.

Liedholm, Carl and DC. Mead. (1992). The Structure and Growth of Microentepprises

in Southern and Eastern Africa; Evidence from Recent Surveys, Michigan State
University.

143

144

Loveridge, Scott. (1990). Les Sources de Revenu des Ménages Agg'coles Rwandais, les

Exportations et Leur Impact sur la Disponibilité Alirnentaire en Milieu Rural,
Republique Rwandaise, Ministére de l’Agriculteur et de l’Elevage, Division des

Statistiques Agricoles, DSA Publication No. 24.

Maddala, GS. (1992). Introduction to Econometrics (2nd Edition). New York:
Macmillan Publishing Company.

McPherson, Michael A. (1991). The Hazards of Small firms in Southern Africa (draft),
Michigan State University, November.

Minot, Nicholas W. (1992). Distributional Effects of Currency Devaluation on

Households in Rwanda: An Application of Willingness-to-Pay Welfare Measures.
Ph.D Dissertation, Michigan State University.

Ngirumwami, Jean Leonard. (1992). Analysis of Food Crop Production and Marketing
Trends in Rwanda with Emphasis on m Beans. Plan B Paper, Michigan State

University.

Nicholson, Walter. (1989). Microeconomic Theog: Basic Principles and Extensions
(4th Edition). Chigago: The Dryden Press.

Nowak, Maria. (1989). "The role of microenterprises in rural industrialization in Africa".
In Microenterprises in Developing Countries (Ed: Jacob Levitsky). London:
Intermediate Technology Publications.

McPherson, Michael A. 1992. Growth and Survival of Small Southern African Firms
Ph.D Dissertation, Michigan State University.

 

Miklavéié, Pamela R and V. Habimfura. (1995). Rwanda’s Banana Wine Subsector:
Determinpnts of Farm Household Participation, Michigan State University
Department of Agricultural Economics, Staff Paper (to be published).

Parish, William L. and R.J. Willis. (1994). "Daughters, Education, and Family Budgets:
Taiwan Experiences", The Journal of Human Resources Vol. 28, No. 4., pp. 863-
898.

 

Parker, Joan C. (1995). Pattems of Business Growth: Micro and Small Enterprises in
Kenya. Ph.D Dissertation, Michigan State University.

Reardon, Thomas, C. Delgado, and P. Matlon. (1992). "Determinants and effects of
income diversification amongst farm households in Burkina Faso". Journal of

Development Studies, January.

145

Schnepf, Randall D. (1992). Nutritional Status of Rwandan Households: Survey

Evidence on the Role of Household Consumption Behavior. Working Paper 23.
Cornell Food and Nutrition Policy Program. March.

Schultz, T.W. (1964). Transforming Traditional Agg'culture. New Yaven: Yale
University Press.

Shaffer, James D. (1973). "On the concept of subsector studies". American Journal of
Agp'cultural Economics. Vol. 55, pp. 333-335.

Strauss, John. (1984a). "Marketed Surpluses of Agricultural Households in Sierra Leone",
American Journal of Agricultural Economics, Volume 66, no. 3, pp. 321-331.

Strauss, John. (1984b). "Joint determination of food consumption and production in rural
Sierra Leone: estimates of a household-firm model", Journal of Development
Economics Volume 14, pp. 77-104.

 

Université Nationale du Rwanda. (1989). Enguete - Diagpostic sur la Culture Bananiere,
Rapport Final -Institut des Sciences Agronomiques du Rwanda/Institut de
Recherches sur les Fruits et Agrumes, mars.

Wohl, Jennifer. (1992). A Regional Analysis of the Effects of Food Subsidy Policies on
Household Consumption Patterns in Morocco, MS Thesis, Michigan State

University.

World Bank. (1992). World Development Repprt 1992: Development and the
Environment. Oxford: Oxford University Press.

 

HICHIGAN STATE 0

l11111111111111)!!!lllllflllflljjflfifli“

4