PLACE IN REI'URN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

’13

 

 

 

DATE DUE DATE DUE DATE DUE
651:9 60370 l ' OED 4 01333

 

RI30409

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1M cJCIRC/DubDuost-n“

 

PEOPLE AND BANANAS ON STEEP SLOPES: AGRICULTURAL
INTENSIFICATION AND FOOD SECURITY UNDER DEMOGRAPHIC
PRESSURE AND ENVIRONMENTAL DEGRADATION IN RWANDA
By

J aakko Kangasniemi

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

Department of Agricultural Economics

1998

ABSTRACT
PEOPLE AND BANANAS ON STEEP SLOPES: AGRICULTURAL
INTENSIFICATION AND FOOD SECURITY UNDER DEMOGRAPHIC PRESSURE
AND ENVIRONMENTAL DEGRADATION IN RWANDA
By

Jaakko Kangasniemi

This research uses national farm survey data and other information to investigate

4/ X ;,1 agricultural intensification in Rwanda. The main focus is on how farmers are changing _

 

 

'0.

their land use in response to population growth.

 

The study shows that under demographic pressure, Rwandan households are
expanding cultivation at the expense of pasture and fallow and relying increasingly on
four crops: bananas, beans, sweet potatoes, and maize. While beans, sweet potatoes and
maize are staples grown mostly for subsistence consumption, the expansion of bananas is
largely a market-driven process. In 1990, bananas brought more cash to Rwandan farm
households than coffee. The exchange of banana beer for food helped many poor
households achieve food security. In part, this strategy is based on high taxes on

industrially produced beer and low barriers for the informal imports of beans and cereals.

The results also Show that land scarcity pushes Rwanda’s smallholders to grow
bananas in dense multi-layered associations with other crops rather than on separate
fields. In much of Rwanda, banana intercropping already accounts for more than halfof

the erosion-prone hillsides. Soil losses on fields intercropped with bananas vary greatly

depending on crop management practices, but are generally well below those recorded on

fields without bananas.

Population growth itself need not push Rwandan farmers to replace their beer
bananas with food crops, but restrictions on food imports or declining banana yields due
to pests or diseases could do so. However, if Rwandais economy recovers and if policies
target food security rather than food self-sufiiciency, market-oriented intensification with

bananas, perennial export crops, agroforestry, and mixed farming is likely to continue.

The large and expanding role of banana intercropping suggests that research and
extension to make banana associations more productive and sustainable deserve high
priority. Especially increases in the productivity of food bananas would contribute
greatly to food security. Both environmental and food security concerns argue for policies
that invest in rural infrastructure and tax consumption rather than trade, thereby allowing
farmers to specialize increasingly in high-value cash crops that in Rwanda happen to be

also less degrading than the staple food crops.

ACKNOWLEDGEMENTS

This dissertation builds on the data collection activities of the Division of
Agricultural Statistics (DSA) of the Ministry of Agriculture in Rwanda. Financial support
for these efl‘orts as well as for Michigan State University’s food security research,
including much of my work, came from the Rwanda Mission of the United States Agency
for International Development. Ultimately, the credit should go to American taxpayers
and, above all, to over a thousand Rwandan families who donated their time in answering

weekly survey questions, and to nearly one hundred enumerators who recorded their

responses.

I also would like to express my gratitude to other sources of support. The first
year of my graduate studies at Michigan State University was supported by the ASLA
Fulbright I Helsinki School of Economics scholarship; subsequent support came from the
' Yrjo Jahnsson Foundation, the Finnish Cultural Foundation, and the Wainstein
Foundation. During much of the dissertation write-up I was able to use the facilities of
the Institute of Development Studies, Helsinki University.

In the course of my graduate program, I have benefitted greatly from the support
and encouragement of a wide range of people. John Staatz, my major professor, has
provided valuable support and guidance ever since my first contact with Michigan State
University in 1990. His insights into economic development, his skills as a teacher and an
advisor, and his commitment to “making a difference” have been a constant source of

inspiration for me.

Thomas Reardon, my thesis advisor, was always full of enthusiasm and provided
valuable comments on various aspects of the dissertation. I have also benefitted from
other members of my thesis and guidance committees. Michael Weber was instrumental

iv

in arranging the resources necessary for carrying out my research. Along with James
Shaffer he refocused my attention on the policy relevance of the research questions and
on the importance of communicating the results to policymakers. Daniel Clay was my
principal guide to agricultural survey work and data analysis at DSA. Carl Eicher, A.
Allan Schmid, and Richard Bernsten supported my studies in various ways over the

years.

I would also like to thank colleagues floor the Division of Agricultural Statistics.
My special thanks go to Anastase Murekezi who lead the Division during a very difficult
period, and to Gedeon Mudacumura whose hard work was admirable and who taught me
a lot about Rwandan agriculture. David Tardif-Douglin had many insights into the data
and the issues. He and his family, like Daniel Clay and his family, helped make our stay
in Rwanda enjoyable.

In Michigan, I have also benefitted from the dedicated work of Janet Munn,
Sheryl Rich and others in the staff of the Department of Agricultural Economics.

A number of fellow students and fiiends helped and encouraged me in various
ways. I am especially grateful to Julie Howard, Scott Grosse, and KyOsti and Liisa
Pietola.

Most of all, I would like to express my gratitude to my wife, Teija Kallio, and our

children Farm and Paula. Without their support and sacrifices this work would not have
been possible.

TABLE OF CONTENTS

 

 

 

 

 

LIST OF TABLES .............................................................................................................. x
LIST OF FIGURES ......................................................................................................... xii
LIST OF ABBREVIATIONS ......................................................................................... xiii
Chapter 1
INTRODUCTION .............................................................................................................. l
1.1 Intensification and Commercialization ...................................................................... l
1.2 Organization of the Study ........................................................................................... 6
Chapter 2
BACKGROUND ON RWANDA ....................................................................................... 8
2.1 Geography and Climate ............................................................................................. 8
2.2 People and History ................................................................................................... 12
2.3 Agricultural Developments ...................................................................................... 16
2.3.1 Pre-Colonial Practices: Extensive Grazing and Small Plots ............................. 16
2.3.2 Colonial Changes: New Crops and More Cattle ............... ....18
2.3.3 Revolution and Independence: More People and More Crops ..... 21
2.3.4 Recent Stagnation: More People and More Trouble -- - ..... .-22
Chapter 3
REVIEW OF AGRICULTURAL INTENSIFICATTON LITERATURE ........................ 25
3.1 Malthusian Fears and Boserupian Hopes ........................ 25
3 .2 Conditioning Factors ............................................................................................... 28
3. 2. 1 Introduction ......... ..... - ..... - .................. 28
3 ..2 2 Environmental Conditions ................................................................................ 31
3 .2 3 Market Incentives -- - - .... - -37
3. 2. 4 Availability of Intensification Techniques ........................................................ 41
3.2.5 Land Tenure ...................................................................................................... 43
3.2.6 Capacity to Invest .............................................................................................. 46
3.3 Chapter Summary .................................................................................................... 47

Chapter 4

 

 

 

 

 

 

 

DATA AND THEIR LIMITATIONS .............................................................................. 48
4.1 Introduction ............................................................................................................. 48
4.2 Sampling Design ...................................................................................................... 48
4.3 Surveys Conducted .................................................................................................. 50
4.4 Quality Control and Data Cleaning ......................................................................... 53
4.5 Data Transformations and Resulting Variables ..... - - _ -54

4.5.1 Overview ............................... - ................ -- 54
4.5.2 Production, Land Use and Yield Data ............................................................... 54
4.53MarketTransactionsandRuralExports. -5..- -57
4.5.4 Income and Its Components .............................................................................. 58
4.5.5 Food Availability .............................................................................................. 63
4.6 Limitations and Their Implications for Analysis ..................................................... 64

Chapter 5

INTENSIFICATION OF LAND USE .............................................................................. 66
5.1 Main Land Use Options ........................................................................................... 66
5.2 Land Use Trends in Recent Past .............................................................................. 72
5.3 Patterns of Land Use by Farm Size ......................................................................... 75
5.4 Multivariate Analysis of Land Use .......................................................................... 80

5.4.1 General Model ................................................................................................... 81

5.4.2 Explanatory Variables and Hypotheses ............................................................. 82

5.4.3 Regression Results on Land Use ....................................................................... 87

5.5 Environmental Impacts of Intensification ............................................................... 92

5. Conclusions ............................................................................................................. 97
Chapter 6 \

PARTICIPATION IN AGRICULTURAL PRODUCT MARKETS 99

% 6.] Income Patterns . - - - - 99

6. 2 Patterns of Agricultural Transactions -- - -- - - - - --106

6.2.1 National Patterns - - - --.....106

6. 2. 2 Agricultural Transactions In Banana Zone ...................................................... 112

6. 2. 3 Crop Transactions by Income In Banana Zone ............................................... 115

6.2.4 Crop Transactions by Farm Size in the Banana Zone -- -- - - 119

vii

6.2.5 Some Concluding Remarks ............................................................................. 125

 

 

 

 

 

 

 

 

6.3 Multivariate Analysis of Crop Transactions .......................................................... 126
6.3.1 General Model, Explanatory Variables, and Hypotheses ................................ 126
6.3.2 Regression Results on Crop Transactions ....................................................... 129

6.4 Conclusions ........................................................................................................... 132

Chapter 7
BANANAS AND FOOD SECURITY ........................................................................... 134

7.1 Introduction - - -- -- ..................................................................................... 134

7.2 Background Information on Bananas .................................................................... 134
7.2.1 General Characteristics ofBananas - _ _ - - - 134
7.2.2 Bananas on East Afiican Highlands ................................................................ 138
7.2.3 Role of Bananas in Rwanda’s Farming System ................................................ 141

7.3 Bananas for Food ................................................................................................... 142

7.4 Bananas for Cash ................................................................................................... 150

7.5 Cash for Food ........................................................................................................ 154

7.6 Future Role of Bananas in Food Security .............................................................. 157
7.6.1 Declining Terms of Trade for Banana Beer? .................................................. 157
7.6.2 Declining Yield Advantage of Beer Bananas? - ............ - . - .....160
7.6.3 Fertilizer Subsidies? ........................................................................................ 162
7.6.4 Transition to Edible Bananas? ......................................................................... 166

7.7 Chapter Summary .................................................................................................. 168

Chapter 8
OTHER RESPONSES TO DEMOGRAPHIC PRESSURE .......................................... 171

8.1 Introduction ........................................................................................................... 171

8.2 Other Agricultural Responses - _ -- - _- -- -- - - ..171
8.2.1 Investments in Erosion Control -_ - -- - - - 171
8.2.2 Agricultural Inputs - - -- - - ......... - -- - ......... _ 177
8.2.3 Livestock Practices -- -- - -- - - - 179

8.3 Income Diversification .................................. -- - 185

8.4 Demographic Responses ........................................................................................ 187

8.5 Conclusions ........................................................................................................... 190

\/

Chapter9

 

 

 

 

 

 

CONCLUSIONS AND POLICY IMPLICATIONS ...................................................... 191
9.1 Summary of Results ................................... 191
9.1.1 Land Use ......................................................................................................... 191
9.1.2 Market Participation - - - ............. . - - - -- - - - ---.l92
9.1.3 Bananas and Food Security -- -- -- - - -193
9.2 Update on Recent Developments ........................................................................... 195
9.3 Implications for Policy .......................................................................................... 197
9.3.1 Agricultural Research and Extension - .................... - .......... 197
9.3.2 Soil Conservation - - - -- - - . ...................... - -- -- -203
9.3.3 Tax and Trade Policy -- - -------- -- ----206
9.3.4 Infiastructure ................................................... - - - - - - 209
9.3.5 Off-Farm Employment .................................................................................... 210
9.4 Suggestions for Future Research ........................................................................... 213
Appendix A
REGIONAL PATTERNS OF RURAL EXPORTS .............................................. 223
REFERENCES ............................................................................................................... 228

LIST OF TABLES

Table 5-1: Characteristics of Main Crops (National Averages) ......................................... 68
Table 5-2: Land Use by Farm Size .................................................................................... 76
Table 5-3: Crop Shares by Farm Size ................................................................................ 78
Table 54: Comparison of Cropping Zones ....................................................................... 79
Table 5-5: Land Use by Farm Size in Banana Zone (% of cultivable land) ...................... 80
Table 5-6: Intercropping by Farm Size in Banana Zone .................................................... 81
Table 5-7: Land Use Model I: Shares of Cultivable Land under Each Use ....................... 88
Table 5-8: Land Use Model II: Shares of Cultivated Land under Each Crop .................... 89
Table 6-1: Gross Household Income Sources by Income Quartiles ................................ 100
Table 6-2: Net Household Income Sources by Income Quartiles .................................... 101
Table 6-3: Net Household Income Shares by Income Quartiles ...................................... 103
Table 6-4: Net Household Income Shares in Banana Zone ............................................. 104
Table 6-5: Net Household Income Sources by Farm Size in Banana Zone ..................... 106
Table 6-6: Rural Transaction Quantities .......................................................................... 108
Table 6-7: Rural Transaction Values ............................................................................... 110
Table 6-8: Net HOusehold Sales by Cropping Zone ........................................................ 113
Table 6-9: Net Household Sales in Frw by Cropping Zone ............................................. 114
Table 6-10: Net Household Sales Quantities by Income Quartile in Banana Zone ......... 116
Table 6-11: Net Household Sales Values by Income Quartile in Banana Zone .............. 118
Table 6-12: Net Household Sales Quantities by Farm Size in Banana Zone .................. 120
Table 6—13: Net Household Sales Values by Farm Size in Banana Zone ........................ 121

Table 6-14: Net Sales per Household by Farm Size in Banana Zone .............................. 123

Table 6—15: Net Sales per Hectare by Farm Size in Banana Zone ................................... 125
Table 6-16: Net Household Sales Model ......................................................................... 130
Table 7-1: Caloric Availability by Crop and Income Quartile ........................................ 145
Table 7-2: Shares of Caloric Availability by Crop and Income Quartile ........................ 147
Table 7-3: Shares of Caloric Availability in Banana Zone .............................................. 149
Table 7-4: Household Cash Incomes by Source and Income Quartile ............................ 151
Table 7-5: Household Cash Income Shares by Income Quartile ..................................... 152
Table 7-6: Households Getting >1/3 of Cash fi'om a Source ........................................... 154

Table 7-7: Household Cash Expenditures on Food Crops and Beer by Income Quartile156

Table 8-1: Household Livestock Holdings by Farm Size ................................................ 182
Table A-1: Net Household Sales and Gifis by Geographic Region ................................. 224
Table A-2: Net Household Sales and Gifts by Agroclimatic Zone .................................. 225
Table A-3: Net Household Sales and Gifts by Prefecture ............................................... 226

LIST OF FIGURES

Figure 2-1: Map of Rwanda ................................................................................................ 9

C-value

DAI

DSA

FAO

FRw

MINAGRI

MINIPLAN

MSU

RPF

TLU

USAID

LIST OF ABBREVIATIONS

Adult equivalent

National Agricultural Commission (of
Rwanda)

Index of crop cover
Development Alternatives Inc.

Division of Agricultural Statistics (of
Rwanda’s Ministry of Agriculture)

Food and Agriculture Organization of
United Nations

Rwandan francs

(Rwanda’s) Ministry of Agriculture and
Livestock)

(Rwanda’s) Ministry of Planning
Michigan State University
Rwandan Patriotic Front
Tropical livestock unit

United States Agency for International
Development

xiii

 

Chapter 1

INTRODUCTION

1.1 Intensification and Commercialization
During the next quarter of a century, the population of sub-Saharan Afi'ical is

projected to double (World Bank 1995). As a result, hundreds of millions of rural
Afiicans will not be able to survive by expanding cultivation as their parents and
grandparents had done. Instead, they will have to increase production by substituting
other inputs for land. If food security is to improve under demographic pressure, farming
systems need to undergo major changes. This requires public policies and investments
that facilitate sustainable intensification.

Africa already has several pockets of high population density, where land is scarce
and where farmers are intensifying by inventing and adopting new agricultural practices,
by adjusting their crop mix, and by purchasing new inputs (Kates, Hyden, and Turner 11
1994). In addition to being a home to tens of millions of people, these intensively farmed
areas are pioneers, struggling now with many of the problems that other areas will face

soon as their population grows.

As the most densely populated country in continental Afiica, Rwanda is in the
forefiont of agricultural intensification. Like most other densely inhabited areas in Afiica,
Rwanda has found that the fast transition fiom extensive growth to intensification is
difficult. After being one of the few Afiican countries where food production increased
faster than population in 1960-1980 (Harrison 1987), Rwanda’s agriculture apparently
failed to keep pace with population growth after the mid-1980s, when the land frontier

 

‘ Hereafter, “Africa” is used to denote sub-Saharan Afiica.

became virtually exhausted (MINAGRI 1992). The resulting impoverishment and
declines in food access also contributed to the genocide of 1994 (Grosse 1994; Olson
1994; SellstrOm and Wohlgemuth 1996).

Four years later, Rwanda’s agriculture has largely recovered from the events of
1994 (Economist Intelligence Unit 1996a, 1996b, 1997), but the challenge of
intensification remains unsolved. To feed the expanding population without destroying
their resource base, Rwandan farmers need support for their intensification efforts from
V appropriate public investments and policies. The design of appropriate interventions
requires information on how the farming system is changing under population pressure.
Policymakers need to know how farmers are using their increasingly scarce land and what
are the factors that determine whether they resort to unsustainable practices to survive on

their tiny farms. Since! many of the public policies affect farmers via markets, the

 

relationship between intensification and commercialization is of particular interest.

1 f 2....

General literature on semi-subsistence agriculture provides important insights and

 

’7“ hypotheses, but cannot give specific answers. For instance, theory has no unequivocal
conclusion on whether agriculture will become more or less commercialized as it
intensifies during demographic pressure. On the one hand, population growth can push
farmers to specialize in cash crops that give highest returns to scarce land (von Braun, dc
Haen and Blanken 1991; I-Iillman 1981). Moreover, since very small farms do not have
enough land to restore fertility via fallowing, they may need to produce cash crops to be
able to buy fertilizer and other inputs (Ruthenberg 1976; von Braun and Kennedy 1994).

\lVOn the other hand, many researchers have pointed out that due to high transaction costs
and unreliable markets for food and for cash crops, poor rural households may have good
reasons to prioritize subsistence production and produce for the market only after most Of
their subsistence food needs have been met (de Janvry, Fafchamps and Sadoulet 1991;
Fafchamps, de Janvry, and Sadoulet 1995; von Braun and Kennedy 1994). Where the

 

rational for “food first-strategy” prevails, land scarcity brought about by population
growth can reduce cash crop production and push households back towards subsistence
production. On the other hand, population growth could also undermine the rational by {f L J

helping to reduce transaction costs and make markets more reliable.

Empirical studies on Rwanda’s agricultural intensification also have considerable
gaps. Many have focused on small, unrepresentative areas. In most of Rwanda, farming
systems are dominated by bananas, mostly grown for beer’, but in attempting to select
measwherethepressmeonresomcesisgreatesgmanyofthebeststudiesofthe
intensification process have focused entirely or mostly on high-altitude and/or marginal
areas where few bananas are grown (von Braun, de Haen and Blanken 1991; Olson 1994;
Ford 1994; Charlery de la Masseliere 1989, 1992, 1993a). While there are good reasons
to study the most food insecure or the most densely populated areas, the past studies can
cast little light on the ongoing intensification process in the banana-growing areas that
cover most of Rwanda. The focus of the present study is on these areas and specifically
on the role of bananas in Rwanda’s agricultural intensification. I

When empirical studies have looked nationally at the questions discussed above,
they have often done so on the basis of quite limited evidence and come up with
contradictory answers. The ongoing and expected crop mix changes are a case in point.
The conventional wisdom is that the transition to roots and tubers that has been going on
for decades will continue in the future. However, in one of the most detailed studies of

~ . rural Rwanda, Bart (1993) predicts that the current peri-urban cooking banana gardens

 

 

2 Banana beer is a rapidly fermented, poorly storable drink with a low (and varying)
alcohol content. In Rwanda, it is typrcallycfiroduced artisanally out 0 beer bananas
(plus small amounts of saw) that are 'stinct from cookin bananas. Because it is
mostly fi'uit-based, banana rs occasionally called “wine,” ut this risks creating
confusion with another product, a lon -fermentation more storable beverage known as
wine (“Vin”) that is also made artisanaglly out of bananas in Rwanda, but is not nearl as
important as the ubiquitous beer (Miklavcic 1995). Also distilled hard liqueurs can
made out of bananas, both industrially and artisanally, but they play only a minimal
role in rural Rwanda.

\[I

give a preview of tomorrow’s agriculture in Rwanda’. Several researchers have argued
that if population growth reduces per capita food production, farmers will substitute food
crops far their beer bananas (Jones and Egli 1984; Cambrezy 1984). However, the land
use statistics for 1984-1990 indicate that while per capita food production declined,
farmers planted more beer bananas and coffee (MINAGRI 1992).

Regarding commercialization, there is a wide consensus among researchers (e. g.,
Charlery de la Masseliere 1992; Jones and Egli 1984; Cambrezy 1984) and policymakers
in Rwanda and in donor agencies (Ntezilyayo 1986; CNA 1992a; World Bank 1991) that
under demographic pressure Rwandan farmers are reverting back to subsistence.
However, the land use statistics for 1984-1990 cited above do not Show any such trend.
Moreover, many of those warning about declining cash crop production in fact focus on
export crops, ignoring bananas and other non-export crops that are fi'equently grown for
sale.

The purpose of this study is to inform agricultural policies in Rwanda by
looking specifically at ongoing changes in land use and at the participation of rural
households in agricultural product markets in the context of demographic change
and environmental degradation. More specifically, we ask what explains households’
land use and marketing decisions, how households are responding to demographic
change, and why some households are more successful in finding sustainable solutions to
the increasing land scarcity than others. In less detail, we also review literature and some
data on other agricultmal and non-farm responses, seeking to put the land use and
marketing decision in a wider perspective on what are households’ strategies to cope with

the increasing demographic pressure on resources.

 

3 Although Bart’s book was ublished in 1993, his statistics on the “current” situation
were mostly fiom the 19 national agricultural survey.

Although the focus is on Rwanda, the study may also have wider relevance for the
highlands in eastern and central Afiica. Together these areas cover close to 100 million
hectares and are home to perhaps 100 million people (Lewis and Berry 1988). Most of
these highlands have much higher population densities than the surrounding lowlands.
Like Rwanda, they typically have sufficient rainfall for cultivation and steep slopes, some
of which are fertile and volcanic, but most relatively poor and prone to erosion. Many of
the highlands have sedentary farming systems that resemble much more Rwanda’s
farming systems than those of their surrounding lowlands. With population growth, many
of these areas are rapidly approaching the problems that Rwanda is struggling with today.
Just as this study repeatedly refers to the experiences documented in the highlands of
Kenya, Tanzania, Burundi, Uganda, Democratic Republic of Congo‘, and Ethiopia, those
working on these highlands may find the observations and conclusions of this study
useful. The findings should be of particular interest for those focusing on Burundi, which
is very similar to Rwanda in many respects.

In Rwanda, intensification is a key food secrnity question. In a predominantly
agricultural, landlocked country with almost no unused cultivable land and limited hope
for rapid increases in non-farm employment, it is only a slight exaggeration to say that
food security succeeds or fails with agricultural intensification. But intensification is also
an environmental issue, particularly in the less densely populated countries like Tanzania.
Failed intensification may not lead to starvation there as easily as in Rwanda, but may
instead force farmers to expand to new areas.5 Whether the remaining forests and
wetlands on the East Afiican highlands and many of the species endemic to these areas
can be saved depends largely on how well agricultural intensification succeeds in the

 

‘ Former Zaire, hereafter referred to as “Congo.”

’ Examples of outrnigration in the context of failed intensification abound in the
literature. Ruthenberg (1968 and 1976) discusses several cases and Kajumulo-Tibaijuka
(1984mrtbsélof one case that has much relevance for Rwanda and is discussed in
some ow.

currently farmed areas. By informing policies on agricultural intensification, this study

also hopes to contribute to environmentally sound development.

1.2 Organization of the Study
This study comprises nine chapters. Chapter 2 provides a brief overview of the

geography and history of Rwanda as well as a somewhat more detailed review of
agricultural developments in this century. Since the main issue in Rwanda’s agriculture in
this century is the expansion of population fiom 1-2 million to about 8 million in 1994,
much of the discussion looks at the long-term trends in agricultural intensification.

In chapter 3, the literature of agricultural intensification is reviewed. Although the
theoretical starting point is on the writings of Thomas Malthus, Esther Boserup and their

followers, the main focus is on empirically based studies on the African continent.

Chapter 4 describes the surveys which form the empirical basis of this study and
discusses the methods used in transforming raw survey data into analytically meaningful
numbers. I

The results of the study are presented in chapters 5, 6, and 7. Chapter 5 examines
the patterns and determinants of land use, focussing on how the most land-scarce farmers
differ in their land use fi'orn that of their neighbors with more land per person. Chapter 6
studies the patterns and determinants of market participation. The main focus is on how
and to what extent Rwanda’s poor rural households use markets to survive with their
limited land resources. In chapter 7, we look in some detail at the contribution of bananas
for food security.

Chapter 8 draws from other literature and somewhat also fi'om DSA data to look
on other aspects of agricultural intensification, including soil conservation investments,
livestock practices, and the use of agricultural inputs. The purpose is to put the results of
this study into a wider perspective.

Finally, in chapter 9, we summarize the results of the study and discuss their

implications for agricultural policy in Rwanda and similar countries.

Chapter 2

BACKGROUND ON RWANDA

2.1 Geography and Climate
Rwanda is a small, landlocked country located close to the equator in the highland

region of east-central Afiica (Figure 2-1). It is bordered by Tanzania to the east, Uganda
to the north, Congo to the west, and Burundi to the south. Most international trade
normally flows through Uganda and Kenya, whose port town Mombassa lies about 1700
kilometers to the east by road (Sirven, Gotanegre, and Prioul 1974).

With a land area of 26,338 square kilometers (2.6 million hectares), Rwanda is
slightly smaller than Belgium, its former colonial ruler. Almost the entire country is over
1000 meters above sea level and about 1.5 million hectares are over 1500 meters, the
altitude Often used to define highlands (Lewis and Berry 1988). Due to the high altitude,
Rwanda enjoys a mild temperature, averaging about 19 degrees Celsius. While
temperature shows little seasonal variation, rainfall follows a bimodal pattern. The longer
rainy (and hence growing) season ranges from February to May, and the shorter one from
October to December (N duwayezu 1990). Summer months are dry and little agriculttual
work is done in August.

 

 

Source: The World Bank

 

Figure 2-1: Map of Rwanda

Despite its small size, Rwanda contains substantial climatic variation, largely
attributable to altitude. The eastern third of the country is relatively low (1000-1500
meters) , warm (above 21 degrees Celsius), and, except for the swamps, dry (800-1000
mm). The natural vegetation consists mostly of woody savannah, not closed forests as

 

elsewhere. The landscape is hilly, but not as mountainous as elsewhere. By Rwandan
standards, the East is a drought-prone region, much of which has only recently become
settled, and still is relatively sparsely inhabited.

The central plateau lies between 1500 and 1900 meters and has milder
temperatures (17-20 degrees Celsius) and higher rainfall (1000-1300 mm) than the
eastern lowlands. This is the historic core of Rwanda, “the land of thousand hills” , and
has a long history of dense settlement.

To the west of the central plateau lies the Congo-Nile Divide, which consists of a
mountainous chain running north-south and separates the Nile watershead to the east and
Lake Kivu to the west. Much of the mountain chain is between two and three kilometers
of altitude, cool (15-17 degrees Celsius) and rainy (more than 1200 mm rainfall). The
slopes are generally steep and highly prone to erosion. In northwestern Rwanda, the
divide consists of a chain of volcanic mountains, several of which peak above 3500 1
meters. The volcanic soils are much more fertile and less prone to erosion than the soils

elsewhere on the Divide.

The shores of Lake Kivu resemble the Congo-Nile Divide in rainfall but
central plateau in temperature.

Administratively rural Rwanda is divided into ten prefectures, named for their
respective capitals (Sirven, Gotanegre, and Prioul 1974; Grosse 1994). In addition, there
is one urban prefecture consisting of Kigali, the capital city. Much Of the rural Kigali, as
well as Byumba in the north-east and Kibungo in the south-east cover eastern lowlands
andhavelargefarmsbythestandardsowaanda. GitaramaandButareonthehigh
plateau are the historic center of the country. Also Gikongoro and Kibuye on the Congo-
Nile Divide belonged to the core but had a lower agricultural potential and were partly
cleared for cultivation only in this century. Cyangugu in the south-west is separated hour

10

the rest of Rwanda by a large national park, Nyungwe forest, and relies much on cross-
border trade with Congo and Burundi. The northwestern prefectures of Ruhengeri and
Gisenyi are very densely populated, but were historically somewhat separate from the

core.

For analytic and policy purposes, several agricultural classification schemes have
been devised for Rwanda. The most widely used is the division of Rwanda into 12
agricultural regions (Delepierre 1982). Producing representative survey data for 12
regions, some of which are quite small, would either require very large national samples
or result in unreliable statistics for some of the regions. Therefore, the Division of
Agricultural Statistics (DSA) of Rwanda’s Ministry of Agriculture used in its survey
work another classification, which divides the county into five agroclimatic zones:
North- West, South- West, Center-North, Center-South, and East. The problem with this
scheme is the great variation within the zones. North-West, for instance, includes both
some of Rwanda’s most productive banana groves by Lake Kivu, temperate volcanic
highlands that are too cool for bananas but highly productive under potatoes, marginal
farming systems on the non-volcanic and erosion-prone parts of the cool highlands, and
warmer mid-altitude areas typical for much of the rest of the country. Agroclimatic
differences between Center-North and Center-South are probably much smaller than
those within North-West. When referring to these five zones below, we call them

geographic regions.

A third scheme, used in some form by Jones and Egli (1984), Ange (1987) and
Nduwayezu (1990), separates three large zones mostly based on altitude: Central (or
High) Plateau consists of the historic “land of thousand hills” between 1400 and 2000
meters, East includes the warmer and drier regions at lower altitudes to the east of the
High Plateau, and Congo-Nile Divide covers the high-altitude watershed lands on the

mountain chain to' the west of the Plateau. In addition, there are two smaller zones:

11

Volcanic Highlands on the northern edge of the Congo-Nile Divide and Kivu, a narrow
strip between the Divide and Lake Kivu. This is the classification used implicitly in the
beginning of this section. However, nationally representative farm surveys have not been
designed to provide reliable information for each of these five zones. Therefore, statistics

computed for each of the two small zones have large margins of error.

2.2 People and History
Rwanda is the most densely populated country on the African continent. In early

1994, its population was about 7.9 million, which translates into a density of roughly 300
inhabitants per square kilometer or three per hectare (Grosse 1994). The density relative
to cultivable land (outside national parks) was about double this, and the most densely
inhabited regions of the country had rural population densities close to 1000 per square
kilometer. Although not exceptionally high by the standards of many rice-growing areas
of Asia, such densities are rare in Africa (Kates, Hyden, and Turner II 1993).

In part, Rwanda’s high population density is related to its altitude. Also in
Tanzania, Burtmdi, Congo, Uganda, Kenya, and Ethiopia, highlands tend to have much
higher population densities than the surrounding lowlands. Highlands are generally more
suitable for agriculture due to higher rainfall and more fertile soils and they also provide
some protection from tropical vector-bom diseases such as malaria and sleeping sickness
(Lewis and Berry 1988). Moreover, the isolation promoted both by geography and by the
precolonial kingdoms protected the populations of Rwanda and Burundi fi'om the great
epidemics (cholera and smallpox) of the nineteenth century (Gourou 1953).

At the turn of the century, colonial administrators estimated Rwanda’s population
to be about two million (Guichaoua 1989), but some later estimates for that period are
barely above one million (Nzisabira 1989a, 1989b). Since the estimate for 1949 is below
two million, interpretations of what happened in between vary widely. Grosse (1994)

12

 

considers the higher initial figures to be more credible and sees no population growth,
whereas Olson (1994) states that population doubled between 1920 and 1950 due to
advances in health care and food production. In any case, there were considerable
fluctuations. A series of famines occurred in the late 19208, probably decreasing
population through death and outrnigration. Population increased in the 19303, perhaps to
almost two million, only to be reduced to some 1.5 million by a major famine during
World War II. The period ofrapid growth started after the war. Between 1950 and 1994
the population increased fourfold, starting from about two million, reaching three million
in 1964, four million less than ten years later, five million before the end of 1970s, and
climbing to almost eight million by April 1994 (Hamand 1982; Lalou 1993; Grosse
1994).

According to the 1991 census, 95 percent of the population is rural. Even allowing
for some urban undercounting and misclassification of both individual people and
locations as rural when they in fact were urban (Grosse 1996), it remains true that
Rwanda is one of the most rural countries in the world, with more than 90 percent of its
population living in rural areas.

The only sizable city is the capital, Kigali, with about 300,000 residents in 1988
(Nduwayezu 1990). The other 11 oflicially recognized towns were estimated to have a
total population of about 200,000. Beyond that, there are few hamlets or even large
villages, since Rwandan farmers usually have their dwellings in the middle of their fields.

Unlike most African counties, Rwanda has a long history as a “nation” that
shares a single territory, language, and culture. Yet, there are three ethnic groups, Hutu
(88-90% of the population), Tutsi (9-11%), and Twa (less than 1%). The Twa presumably
descend tom the original hunter-gatherers, the Hutus are descendants of the Bantu
farmers that migrated to Rwanda perhaps 1000 years ago, and the Tutsis are believed to

13

descend tom the pastoralists who came some 500 years ago establishing a feudalistic
contol over the others (Vidal 1985). By 1900, the kingdom of Rwanda encompassed
much of what is today Rwanda. The kingdom was a highly statified society, where
Tutsis were rulers and herders, while Hutus were primarily cultivators and the Twa
hunters and artisans. However, some fluidity existed in ethnic identity. Some Hutus
accumulated cattle and became regarded as Tutsis, while some Tutsis who lost their herds
aswmed Hutu identity. '

Rwanda and neighboring Burundi, which resembles it in many respects, were
occupied by German colonial troops at the turn of the century, but the presence of the
Germans was minimal. During the World War 1, Belgium seized the territory and started
to administer it through the existing Tutsi monarchy. This “indirect rule” favored the
Tutsis and made the feudalistic relationships between Tutsi chiefs and Hutu peasants
more exploitative and the ethnic identities more rigid (N ewbury 1988). At the same time,
colonialists promoted Christianity and other new ideas that undermined the legitimacy of
the Tutsi rule. In the 19503, educated Hutus started to mobilize and articulate the .
discontent. A violent rebellion ensued in 1959, and tens of thousands of Tutsis left as
refugees. Belgian administrators decided to promote majority rule, organizing elections in
1960-61, abolishing monarchy, and transferring power to a Hutu-dominated government
in 1962. In the southern part of the colony, and later in independent Burtmdi, Tutsis
remained in power.

Rwanda’s first government under president Gregoire Kayibanda was explicitly
pro-Hutu. Inter-ethnic violence escalated and almost 200,000 more Tutsis left for
neighboring counties (Grosse 1994). The remarnrn' ' g Tutsis lost their privileges, including

exclusive rights to grazing lands. Increasingly, Hutus and Tutsis alike were farmers.

14

In 1973, Major General Juvenal Habyarimana replaced Kayibanda in a bloodless
coup. President Habyarimana preached reconciliation between ethnic groups and
allocated government positions proportionally to them. Intermarriage became common
and Habyarimana was widely credited for creating ethnic peace in Rwanda. This view
was not, however, shared by most of the exiled Tutsis, few of whom returned. Moreover,
a regional split developed, as president Habyarimana, a northener, channelled government
jobs, educational opportunities, and development funds disproportionately to the
northwestern prefectures of Ruhengeri and Gisenyi.

In late 1990 and again in early 1993, the Rwandan Patriotic Front (RPF),
dominated by the exiled Tutsis, attacked Rwanda fi'om Uganda. The war greatly
increased ethnic tensions and further weakened the economy, already sufi‘ering fiom low
coffee prices and agricultural stagnation. Under pressure fi'om donors, Habyarimana was
forced to legalize opposition parties in 1991. Groups close to Habyarimana started to
mobilize support among the Hutus by vilifying the Tutsis and by labeling the more
conciliatory Opposition parties as RPF collaborators (Prunier 1995).

In August 1993, President Habyarimana, opposition parties, and RPF reached an
agreement on power sharing, tansitional government, and multi-party elections. The
implementation of this “Arusha Accord” was repeatedly delayed. In April 1994, the
assassination of President Habyarimana unleashed simultaneously a genocide and a civil
war. In the genocide, between half a million and one million Tutsis and mostly southern
Hutus were killed by extremist Hutus associated with the regime. In the war, the RPF
invaded the county, which ended the killings of the Tutsis by August 1994, but also
caused a mass exodus of almost two million Rwandans (mostly Hutus) to neighboring

counties.

15

For two and a half years most of the (new) refugees stayed in the camps in Congo,
Tanzania, and Burundi. Meanwhile, hundreds of thousands of earlier “old caseload”
refugees (mostly Tutsis who had left in and after 1959) returned to Rwanda. Some of the
“old caseload” refugees returned fearing the genocidal campaign that was continuing in
and around the refugee camps, especially in Congo. The situation remained tense also in
Rwanda, where dozens of killings and revenge killings were reported monthly and where
tens of thousands of suspected genociders were kept without trial in overcrowded prisons.

In November 1996, a (largely Tutsi) rebellion in Eastern Congo tiggered a
sudden return of more than half-a-million (Hutu) refugees, followed by fitther
repatriations of most of the remaining refugees in 1997. In December 1996, hundreds of
thousands of refugees were repatiated fiom Tanzania. After these returns, the total
number of people inside Rwanda’s borders exceeds that of 1990, when most of the
statistics used in this study were collected, and is probably close to the pre-genocide level
of 7.8 million (World Bank 1998). Within few years, Rwanda’s population density will
climb to unprecedented levels. Despite the genocide of close to one million people,
adaptation to high population pressure remains a cental issue in rural Rwanda.

2.3 Agricultural Developments

2.3.1 Pro-Colonial Practices: Extensive Grazing and Small Plots
When the Europeans arrived at the turn of the century, much of the “country of

thousand hills” was covered by almost treeless grasslands. Except for small areas in the
dry eastern savannas, the grasslands were created by centuries of grazing and fiequent
burnings. While the burnings favored grasses that the cattle could utilize, they also
reduced the amount of organic matter in the soil, thereby decreasing its capacity to hold
water and nutrients. Over time, the grasslands became degraded, producing much less
biomass than the vigorous forests that once had covered the hills. The forests still covered
higher mountains and steeper slopes, especially those on the Congo-Nile Divide. These

16

slopes and watersheds had less fertile soils, were partly too temperate for the crops, and
degraded fast if cleared for grazing or cultivation.

Cultivation was limited to small plots on the hilltops and upper slopes that were
easiest to cultivate and clear with fire (Meschy 1985, 1989). Wet valleys were mostly
reserved for grazing during the dry season. Some of the plots were permanently cropped;
others were periodically left fallow. The fertility of the cultivated plots was based on
fallowing, household refuse, and especially on cattle manure. Livestock played a key role
in fertility management by tansporting nutrients and organic matter from the grasslands
to the cultivated plots. Although most livestock and all land was owned by the Tutsi and
most cultivation done by the Hutu, the two groups were integrated through the ubulralce
contactual system. In the system, Hutu clients guarded the cattle of Tutsi patons and
gave them a portion of their crops, receiving in rettun male calves, milk, blood, manure

and protection against outsiders (Olson 1994).

The main crops were sorghum, finger millet (eleusine), taro (colocase) and
traditional legumes (peas, cowpeas). Beer bananas were also grown widely, although
most beer may have been made out of sorghum. Farmers had also started to grow beans,
sweet potatoes, and maize, but none of these was yet a major staple (Jones and Egli 1984;
Martin 1987).

This picture of pre-colonial agriculture holds for the historical core of Rwanda,
the cental plateau, where bulk of the population lived. Most of the savannas of the East
had very low population densities and almost no cultivation. The large marshes in
southeastern Rwanda and the montane forests on the Congo-Nile watershed were largely
uninhabited.

l7

2.3.2 Colonial Changes: New Crops and More Cattle
To fight the famines that had periodically devasted the county, Rwanda’s Belgian

administors enacted several policies that were supposed to increase or stabilize the
availability of food (Guichaoua 1989; Vis 1975; Olson 1994). In 1924, they passed a law
requiring all farmers to plant and maintain 15 ares offamine crops and 35 ares ofother
food crops per adult. The required fields were much larger than the plots cultivated
before. The implementation of the law would have increased dramatically the work load
of the farmers and the amount of land exposed to erosion by cultivation, while decreasing
soil fertility due to shorter fallow periods and reduced availability of manure. Moreover,
the rule would have reduced the availability of grazing lands. Consequently, the law was
widely resisted, and it is unclear whether the total acreage requirement had much direct
impact.

In contast, the requirement to grow famine crops was enforced rigorously by the
colonial administation, especially after the famines in the late 19203 (Guichaoua 1989;
Leurquin 1963). The famine crOps were non-seasonal roots (sweet potato and casSava)
and tubers (white potato), that were supposed to feed the families if cereals and other
seasonal crops failed. Cassava was promoted at low altitudes, particularly in the East,
white potatoes at the high altitudes of the Northwest, and sweet potatoes throughout the
county. Farmers’ responses were mixed. White potatoes were well received on the
volcanic highlands of the Northwest, where farmers had been lacking crops that do well
in temperate climates. Sweet potatoes provided high yields in terms of calories, had been
gaining grormd already before, and continued to expand. The situation was entirely
difi'erent for cassava, which was actively opposed by farmers. In addition to the new taste
they did not like (yet), farmers may have resented the crop because it did not do
particularly well in much of Rwanda. Cassava is a lowland crop and its yields are much
lower in the cool highlands. Its superiority on badly degaded fields was not needed yet

18

and its resistance to drought was valuable mostly in those parts of the county that were
still under pasture.

Another influential colonial policy was to give farmers rights to cultivate land in
valleys (Olson 1994). The Tutsi herders who depended on the valleys for dry-season
pasture, especially in drought years, resisted the conversion to crops, but nevertheless
some valleys were cleared for cultivation. Belgians also promoted cultivation in
uninhabited areas at high elevations and in the east through the paysannat resettlement
scheme, which emphasized heavily export crops. Despite these policies, the Tutsi by and
large were successful in opposing the conversion of grazing lands into cropping lands.
Consequently, the new crops had to increase mostly at the expense of old crops and
fallow, particularly in the 19503, when the population exceeded the pre-famine levels and
continued to grow fast. In addition to the actively promoted roots and tubers, farmers
planted more bananas and switched away from low-yielding peas and finger millet ’
(Leurquin 1963). Thanks to the new crops and improved varieties introduced by the
Belgians, living conditions remained stable or improved despite the population pressure
(Leurquin 1960). Especially the volcanic highlands in the Northwest, which were fertile
but too cool for Rwanda’s old crops, benefitted tom the temperate crops (white potato
and wheat) intoduced and promoted during the colonial period.

Two other influential land use policies were enacted and enforced rigorously by
the Belgians. In 1931, Rwandans were obligated to plant communal forests, and in 1933,
each farmer was required to plant a certain number of coffee tees (Olson 1994). The
former policy was intended to produce constuction wood and fire], reduce erosion, and
fight famines by increasing rainfall, while the purpose of the cofi‘ee requirement was to

provide revenue for the colonial administation.

19

The colonial rulers failed to realize the key role of cattle manure in Rwandan
agriculture or the role of milk and blood in diets. They felt that cattle contributed little to
human nutition, caused erosion, and were owned for little else than prestige. They
wanted to reduce livestock numbers and rationalize livestock management by
slaughtering more animals and by improving veterinary services. The net impact of the
measures was the Opposite of what was intended. Citing Jean-Paul Harroy (1954), who
became the Resident Governor of Ruanda-Urundi in 1955, Olson (1994) reports that
cattle population increased tenfold between 1920 and 1950 due to veterinary services. As
a leading advocate of the colonial policies to reduce cattle numbers, Harroy had an
interest in creating alarming numbers, and his “cattle explosion” appears grossly
exaggerated Nevertheless, it is clear that Belgians not only failed to reduce cattle

numbers but even contributed to an increase.

After the famines of the late 1920s, Belgians identified erosion as one of the main
causes leading to famine and wrote alarming reports on vast ravines and gullies
everywhere on the densely populated cental plateau (Scaetta 1932). Although many of
the grass lands were quite degraded already before the Belgians, their policies that
expanded cropping at the expense of fallow and increased livestock populations
apparently made things much worse (Olson 1994). In 1934, colonial authorities
intoduced anti-erosive measures coercing farmers to dig infiltration ditches on their
fields (Gatera 1980). After the severe famines of the 1940s, the policy was greatly
expanded. Throughout the colony, farmers were mandated to participate in umuganda,
forced communal labor to dig erosion ditches, plant grass strips on contour lines
(typically on the upper banks of the ditches), and to plant trees on erosion-prone lands.
Dinning of pastme was outlawed (Olson 1994). By 1960, 40 percent of cultivated lands
and 8 percent of pastures had been protected (Nduwayezu 1990; Sirven, Gotanegre, and
Prioul 1974). The laborous works were bitterly resented by farmers.

20

2.3.3 Revolution and Independence: More People and More Crops
As the authority of the Belgians started to wane in the end of the 19503, the

enforcement of the unpopular agricultural policies weakened and finally stopped. Dming
the first decade of independence, almost no anti-erosion stuctures were built, some of the
old stuctures were actively destroyed and most of the rest gradually disappeared due to

lack of maintenance (Brondeau 1991).

In rural Rwanda, the principal consequence of the “social revolution” was to make
lands previously reserved for pasture available for cultivation. Together with the
accelerating population growth, this led to a fast increase in cultivation both locally in the
densely populated heartlands and through the settlement of the uninhabited areas,
particularly in the East. Thus, the process of agricultural intensification under population
pressure that characterized the 19503 was ended by the social revolution and replaced by
extensification for almost two decades (Olson 1994; Bezy 1990). Also the exodus or
death of large numbers of Tutsi and even larger numbers of their cattle made room for
those that remained. It was only after mid-19703 that Rwanda’s agriculture again-faced

population pressure.

During the period of extensive growth, farmers eagerly planted more bananas and
beans (Delepierre 1970; Bart 1993). This form of expansion can be interpreted to be a
sign of relative prosperity, since banana beer, cooking bananas, and beans are relatively
expensive foods, whereas root crops are poor man’s food. Bananas could also be grown
extensively on the relatively large farms of the East, and both bananas and beans grew
much better on the newly cleared fields than on the old overcultivated lands. Sorghum
was favored in the drought-prone eastern savannas but elsewhere lost ground to maize,
which typically yields better in well-watered areas. Cofi‘ee was promoted heavily and at
times coercively throughout the county and tea was promoted in small areas around the
factories (Chapuis 1986). Potato production in the Northwest benefited greatly tom

21

externally ftmded agricultural research and extension (Haverkort 1986a, 1986b).
Although the potato success story was an exception to the general pattern of technical
stagnation and modest investments in agriculture, economic policies in Rwanda were not
nearly as anti-agricultural than they were in many other Afi'ican counties (Azam 1989).
In the 19603 and 19703 Rwanda was one of the few countries in Africa where food
production grew faster than population (Harrison 1987).

As cultivation expanded and soil conservation was neglected, erosion rates
increased. In the mid-19703, Habyarimana’s government re-launched the colonial anti-
erosion policies, with the same soil conservation techniques and similar or even sticter
enforcement (Brondeau 1991; Olson 1994). By the end of the decade, 17 percent of the
cultivated lands were “protected against erosion” (Musema-Uwimana 1979), and in the
early 19803, Habyarimana’s government started to push even harder to achieve 100
percent by 1988. Communal authorities were required to report quarterly on their
activities, with progress linked to development activities (Guichaoua 1991). In 1988, 38
of 133 rural communes were over 90 percent protected and even Kibungo with its gentle
slopes and lower rainfall was estimated to be 74 percent covered (Olson 1994). As in the
colonial period, farmers resented the forced labor and the inflexible design of the soil

conservation works.

2.3.4 Recent Stagnation: More People and More Trouble
By the end of the 19703, almost all cultivable land outside the national parks had

been opened up for cultivation. Communal pastures were mostly gone, private pastures
were largely marginal lands on steep slopes and on leached hilltops, and the remaining
valleys were mostly large wetlands that were difiicult to drain on farmers’ own initiative
and ofien had an important role in the water system (Bart 1993). During the 19803,
farmers could expand cultivation only on marginal lands and at the expense of fallow.
Without ofl’setting measures, both options were likely to increase degradation and

22

erosion. In a 1991 national survey, 47 percent of field blocks were reported by their
cultivators as degraded (Clay 1994).

To cope with the rapid population growth, farmers planted more sweet potatoes
and bananas, which gave high caloric yields, or, in the case of beer bananas, superior
monetary returns to land. In the late 19803, these and other adjustments were merely
enough to offset the impacts of degradation and pest attacks that may have been partly
caused by it. Food production stagnated and per capita food production declined
considerably. The degraded western highlands were particularly hard hit. The droughts in
1989 and 1993 lead to starvation in the poorest areas, especially in the prefectures of

Gikongoro and Kibuye.

In the early 19903, the authoritarian contol of Habyarimana’s government on
farmers started to wane. Forced communal labor came to an end, and the constuction of
anti-erosion works stopped, just as it had done 35 years earlier in the last days of the
colonial government. While some farmers continued to maintain the stuctures, others
actively destoyed theirs. Centalized contol remained longer on coffee, which provided
the government with desperately needed export earnings. By 1994, many farmers had
started to neglect their cofl‘ee tees and some had altogether abandoned theirs, but few had

dared to uproot their coffee tees.6
We-.-m_ ~_ _ -—--. .-. 2 L _-.,_._..,

After the genocide and war in 1994, large areas of Rwanda, particularly in the
East were almost empty of people. Since then, much of the East has been inhabited by the
“old caseload refugees”, that is, the Tutsi who left the county more than 30 years ago,
and their descendants. Along with the people, large herds of cattle have arrived, implying

 

’ In Jan 1994, we asked a farmer in the refecture of Byumba what she would do, if
she coul uproot her coffee trees. She rep red that she would umt and Sow sorghum
and beans instead. Although the prohibition ainst uprooting Ofiici 1y been

ed the local agricultural extensromst w o was with us rapidly reminded her that
she would not be allowed to do so. Some other farmers, however, told us that they were
happy with their cofl‘ee tees and might even plant more.

23

that parts of the East were again reserved for pasture. The return of hundreds of thousands
of refugees in late 1996 brought man-land ratios back to the pre-genocide level in most

parts of Rwanda

24

Chapter 3

REVIEW OF AGRICULTURAL INTENSIFICATION LITERATURE

3.1 Malthusian Fears and Boserupian Hopes
Much of the literature on how rural societies respond to population growth is

colored either by pessimistic fears or by optimistic hopes. The gloomy view sees
population pressure leading to declining living standards and ultimately to poverty and
famine. As desperate people resort to unsustainable responses, they mine their lands,
thereby destoying their environment. In contast, the optimists perceive additional
people as the “ultimate resource” (Simon 1981) that not only is sufficient to maintain the
standard of living but also may be needed for economic progress and environmental
recovery (Kates, Hyden, and Turner II 1993).

The pessimistic view is usually labelled “mo-Malthusian”, afier Thomas Malthus,
the classical economist who popularized the idea that human populations constantly tend
to grow beyond their means of subsistence (Blaug 1985). To avoid misery and starvation
caused by population growth, people should apply “preventive checksz” moral restraint
and contraception.’

The key assumption of the Malthusian View was the idea of diminishing returns.
Along with David Ricardo and other classical economists, Malthus believed that best
lands generally were opened up for cultivation first. If population grew, farmers would
have to increase production either by cultivating poor lands (extensive margin) or by
adopting practices previously considered too laborous (intensive margin). Both options

would reduce returns to labor and, hence, wages and access to food.

 

7 Malthus, unlike most present-day neo-Malthusians, advocated moral restraint and
condemned contaception as “Improper.”

25

Today, the neo-Malthusian view often goes with the term “human carrying
capacity.” The idea is borrowed from ecology, and increasingly neo-Malthusians worry
that population growth undermines ecological sustainability, not just that it reduces
human welfare (Lambert 1982). Angelsen and Fjelstad (1995), for instance, present a
simple model Of a farm household, in which farming becomes more intensive and soil

erosion increases if marginal returns to labor decline due to population growth.’

Some pessimists would reject both the neo-Malthusian label and the prescriptions
that emphasize birth contol (Kates, Hyden, and Turner 11 1993). Whether calling
themselves neo-Marxist or something else, they typically see either local population
growth itself or the associated negative results (human misery, environmental
degradation) as resulting from external exploitation (Blaikie 1985; Blaikie and Brookfield
1987). For instance, the capture of good agricultural lands for export crop production may
force the displaced smallholders to farm unsustainably on marginal lands. The suggested
solutions include land reform, local empowerment, self-reliance and variously defined

forms of “alternative development”.

The optimistic view is usually associated with Esther Boserup, a contemporary

Danish economist, although it also has deep roots in classical political economy. Malthus
himself talked about “the optimal population” and starting from Adam Smith classical
economists had emphasized the progressive role of specialization and its dependence on
the sizeof ESE-@159? (Blaug 1985). Classical economists were well aware that among

{parse populations “rude devices of roduction” could be more profitable than “the most
efiW’Eoopurauon densities grew, people would innovate and
adopt more productive methods. .

 

' This arises from a formulation that does not allow the costs of soil conservation to be a
function of wage rates. Neither is there any induced technical change in the model. If
either of these complications were introduced, the model would no more yield simple
Malthusian conclusrons.

26

Yet, classical economists saw technical progress mostly in industy and
rtation, underestimating its scope in agriculture. Above all, they failed to

appreciate how technical progress could help to increase the fiequency of cropping

(Boserup 1989). Drawing from the economic history of Europe and somewhat also fiom

developments elsewhere, Boserup (1965, 1981) concluded that while population pressure
first leads to diminishing returns, inducing people to adopt more labor-intensive practices
that they could avoid before, it also enables and encourages them to develop and adopt
innovations that mitigate or even reverse the decline in labor productivity that would

otherwise follow.

The key variable In Boserup’ s virtuous cycle 13 the fiequency of cultivation. ‘
Under low population pressure, people usually practice shifting cultivation and use long
(bush) fallows to restore fertility and contol weeds. With population growth, fallow
periods get shorter (grass fallows) and eventually give way to permanent cultivation 'or
even to carefully contolled irrigated farming with two or more cropping cycles per year.
As fallows disappear, households start to maintain fertility by manuring their fields and
adopt better hand tools or animal taction to contol the weeds. Investments in land
become worthwhile. Increasing population density reduces the per capita costs of roads,
irrigation systems, and other forms of infiastucture, expands markets, and promotes
specialization. Further innovation follows and purchased inputs become ubiquitous in
agriculture.

For Boserup, the beneficial upward spiral of population growth and technological
improvements is a possibility that many societies have been able to achieve, not
something that is guaranteed to happen. In particular, if people fail to invent or adopt
other measures of fertility restoration when they shorten the period of fallow, they may
mine their lands and be left with “the choice between starvation and migration” (Boserup
1965). ’

27

In the historical cases referred to by Boserup, population growth rates had been
low, often below 0.1 percent per annum (Boserup 1985). Consequently, the populations
involved had had centuries to invent and adopt appropriate intensification responses.
Boserup herself stated (1965: 118) that the intensification process may not work if
population densities are very high or grow very fast. More recently several authors have
claimed that sustainable intensification is failing to take place in much of Africa, where
population growth rates often exceed three percent and where population densities double
from one generation to the next (Ho 1985; Lele and Stone 1989; Binswanger 1988; Ehui,
Williams, and Swallow 1995).

On the other hand, cases of sustainable intensification do exist, even in places
where both population densities and growth rates have been high. Based on their
comparative study often densely populated areas in western and eastern Africa, Hyden,
Kates, and Turner 11 (1993) had some doubts on how much further these areas can adapt
to population growth, but they were convinced that outside the pockets of high population
density much scope for intensification exists in Africa. Moreover, in four of the ten cases,
intensification had been accompanied by “increases in food availability or improvements
in general material well-being.” Tiffen and Mortimore (1992) and Tifi‘en, Mortimore, and
Gichuki (1994) report on a highly successful intensification in Machakos, Kenya.
Although population increased five-fold between 1930 and 1990, erosion rates declined,
me landscape became greener, and the per capita value of agricultural production
increased considerably.

3.2 Conditioning Factors

3.2.1 Introduction
Basically Boserup studied how one variable, population density, had afl'ected

agriculture in human history. She demonstated how changes in population density in a

wide variety of environments had pushed agricultural societies fi'om one farming system

28

to another. While not denying the role of other factors, she considered population density
, pivotal and warned against assuming that low production in sparsely inhabited areas
indicates low potential (Boserup 1989). In her view, Afiica produced less food than India
because it had fewer people, not because it had less fertile land. She also emphasized that
while labor-saving tools and techniques promote intensification, lack of them has not
prevented some societies (like the Mayas) from intensifying. Moreover, Boserup
emphasized that many of the changes in institutions and policies that promoted
agricultural intensification were in fact responses induced by population growth. Private
land tenure, for instance, generally emerged where investments in land were needed to
feed the growing population. Similarly, the development of high-yield rice and wheat

varieties was a political response to population pressure in Asia.

Other researchers have focused on how factors other than population affect
intensification. Kates, Hyden, and Turner II (1993) difi'erentiate between intensification
theories that emphasize (subsistence) consumption needs and those that emphasize
market demand. Chayanov and Boserup supposedly exemplify the former, whereas many
economists from Schultz (1964) to Hayami and Ruttan (1985) have been proponents of
the latter, arguing that small farmers respond to market incentives. Kates, Hyden, and
Turner 11 lament that each side has largely ignored the other, although evidence is
accumulating that neither approach alone can tell the whole story. In their own study,
Kates, Hyden and Turner 11 perceive the relationship between population and agricultural
intensification as being modified by two sets of variables: market access and
environmental conditions. Thus, they hypothesize that the success of agricultural
intensification in responding to demographic change depends crucially on the

environmental and market conditions in which it is taking place.

That intensifying farmers are responding both to market incentives and
subsistence needs is undoubtedly tue but in implying that most others had missed this

29

elementary fact, Kates, Hyden and Turner 11 are largely creating a stawman that is easy
to knock down. Far from ignoring markets, Boserup (1989) wrote that commercial
farmers respond primarily to markets and warned that if food prices are kept artificially
low by food aid or subsidized imports, intensification may not be forthcoming. Similarly,
the economists emphasizing responses to prices typically did so in order to attack the
myth of an “irrational tadition-bound peasant”, not because they saw market prices as the
only things that matter. Moreover, much of what Schultz (1964) and other economists
wrote emphasized economic incentives that were more than just market incentives and

clearly included on-farm endowments and subsistence nwds.

Along with market incentives and environmental conditions, several researchers
have emphasized the availability of tools, methods, varieties and inputs needed for
intensification, farmers’ knowledge about them, land temue and other institutions
governing the use of land and water, government policies affecting all these, and cultural
and social factors including the norms regarding the sexual division of labor. Bilsborrow
(1987) casts the net even wider, suggesting that the scope for migration and that for
fertility decline should also be considered, since people may fail to intensify if they have
easy access to lucrative jobs or uncultivated fertile land elsewhere. This, however, is a
fiamework for a different question, namely how households choose between the three
options: coping on less land per person, moving out, and restricting their fertility. Our
focus here is on the scope and stategies of the first option.

Before discussing in detail how each of the above factors afl’ects intensification,
we ttun briefly into a more general question on how these factors should be stuctured in
the fiamework of analysis. Most researchers have explicitly or implicitly conceptualized
them as factors that either together with demographic change affect agriculture or
condition how farmers respond to demographic change. In contrast, Clay, Guizlo, and
Wallace (1993) argue that the impacts of population growth depend crucially on the set of

30

intermediate variables they label as the structure oflarufltolding and define to include
four key variables: size, fiagmentation, fiagility, and land tenure. In their fi'amework,
population growth changes the stucture of landholdings, which in turn changes farmers’
behavior regarding land use and investment decisions and, through them, may cause land
degradation. Population growth not merely increases the pressure on resources but also
may create obstacles for sustainable intensification. If, for instance, population growth
contibutes to the concentation of land holdings, lands may increasingly be cultivated by
tenant farmers, who may not have as good incentives as owner-operators to invest in land.
And if population growth pushes farmers to cultivate fragile lands, this may also cause
degradation.

The argument that demographic changes affects behavior through tenure and other
intervening factors is undoubtedly tue, but the “stucture of landholdings” ~fiamework is
problematic in not showing the influence of outside factors. For instance, although the
authors discuss in detail how different crops have very different environmental
consequences, the market incentives that induce households to grow protective rather
than erosive crops or vice versa do not fit easily in their fiamework. The framework of
Clay, Guizlo and Wallace (1993) appears to be designed for the purpose of tracing the
various direct and indirect impacts of population growth in a given context, not for
analyzing how the impacts differ as the context varies. Since the latter is the main focus
here, we proceed by discussing how each of the factors conditions the impact of
demographic growth on agricultural change.

3.2.2 Environmental Conditions
For Boserup, one of the key questions in intensification is how to maintain or

even improve land fertility without unacceptably high labor inputs. Another main
question is how to make water available for progressively higher frequencies of cropping.
While the questions were much the same everywhere, the answers naturally varied with

31

environments. Ruthenberg’s classical study on topical farming systems (1976) provides
many of the answers. It shows that while population growth is pushing nnal areas fi'om

one farming system to another, environment limits the options.

Farmers in temperate climates have typically responded to population growth by
moving fi'om fallow systems to permanent upland (non-irrigated) cultivation, but in the
humid topics dense populations have more often resorted to either irrigation (mostly wet
rice) or perennial crops (Ruthenberg 1976, chapter 6). Without mineral or organic
fertilizer, most topical soils rapidly lose much of their fertility under permanent upland
cultivation. The warmer and more humid the climate, the greater the problem: high
temperature promotes rapid decomposition and high rainfall contibutes to leaching and
erosion. Permanent production of annual crops typically provides neither much additional
organic matter nor much protection from the sun to slow down decomposition nor much
uptake of nutients from the subsoil. The loss of organic matter increases the leaching of
nutients to the subsoil, while reducing water infiltation. Land degradation caused by
leaching, the loss or organic matter, and periodic water scarcity is often augmenbd by
erosion. Soils poor in organic matter and water infiltation capacity, loosened periodically
by cultivation, and not protected by crops against the erosive impact of raindrops are
vulnerable to large soil losses, particularly if they are located on steep slopes and on
impermeable soils.

The fertility loss is generally less severe on the cool, topical highlands, in semi-
arid climates, and on fertile volcanic soils. If excessive erosion is prevented, continuous
cultivation of annual crops can often be sustainable in these environments even without
fertilizer. In contast, for permanent cultivation in humid and semi-humid climates to be
sustainable without either irrigation or large quantities of external inputs, it typically must
“mimic rainforest” (Altieri, Letoumeau, and Davis 1984). Examples include tee and
bush crops, such as oil palm and coffee, home-gardens such as the banana-based farming

32

systems in much of East African highlands, and various agroforesty systems. Another
possibility is to bring either large quantities of manure and other organic matter fi'om
elsewhere, or to use mineral fertilizer to grow the crops and the organic fertilizer (green
manure) needed to maintain the organic matter content of the soil. Where modern inputs
are unavailable or expensive, intensification typically reduces returns to labor, which is
why it only occurs when suflicient land for more extensive forms of production is no

longer available.

The island of Ukara on Lake Victoria in Tanzania is a well-known example of the
very low productivity of permanent farming on poor soils in semi-humid topics (Ludwig
1968; Ruthenberg 1964, 1976). For more than a century, the pOpulation density on Ukara
has exceeded 500 persons per hectare, which is too high to allow the restoration of
fertility through fallowing. Farmers practice careful soil husbandry with manuring,
leguminous green manure/cover crops, and labor-intensive methods of erosion contol.
Yet, yields are modest and returns to labor are extremely low. While quite sustainable
environmentally, Ukara has been a case of poverty agriculture. Whenever possible, young
people have left the island to practice more extensive agriculture on the mainland.
According to Ruthenberg (1964: 31), Ukara shows that “only an exteme emergency
situation can induce the inhabitants to take up pre-technical measures to preserve the
fertility of the soil.”

Eastern Nigeria is an example of intensification toward permanent farming in
humid climates. Population densities are now higher in Eastern Nigeria than on Ukara,
but two or three generations ago they were still low enough for fallow systems.
Consequently, farmers have not yet adopted as careful and laborous soil husbandry
practices as in Ukara. Lageman’s (1977) study of three villages suggests that land fertility
and yields are drastically lower in the village with the highest population density (nearly

thousand persons per square kilometer). Despite much stonger efforts to maintain soil

33

fertility, farmers in the densest village are mining their lands, albeit not in the sense of
taking nutients fiom the soil with the crops and selling them in the market. Increasing
frequency of cultivation lowers yields primarily by reducing the contents of organic
matter and nitogen in the soil. Traditionally fertility management has focused on
importing nutients and organic matter from outer fields and pastures to the compounds
(in-fields) that produce most of the output, but with population growth there are too many

compormds and too few out-fields for this to work.

The book edited by Turner II, Hyden, and Kates (1993) includes three case studies
from Eastern Nigeria. Unfortunately, they shed little light on whether workable responses
to continued population growth have been found in the pockets with the extremely high
rural population densities, like the one studied by Lageman in 1974-75. In one case
(Awka-Nnewi), total production is decreasing and in another (Imo State), yields are
declining (Hyden, Kates, and Turner 11 1993). The third case (N gwa-lgbo) has a much
lower population density and has thus far intensified successfully. In all three cases, some
farmers are responding to the fertility crisis by using mineral fertilizer, financed in part by
market gardening, which is becoming more important. Farmers have also planted more
oil palms and other tees and shifted to less demanding and higher-yielding crops,
particularly fi'om yams to cassava. Land husbandry is improving and irrigation is
expanding.

In sum, permanent upland cultivation of annual crops is so problematic in the
topics that farmers tend to opt either for irrigation or for perennial crops. In general, the
tansition to irrigation farming is a major intensification option in humid climates and
also in semi-humid and semi-arid areas provided that water sources can be found
(Ruthenberg 1976; Bray 1986; Ishikawa 1967). Irrigation allows higher frequencies of
cultivation and can proVide much productive employment. Typically farmers start to
regulate water flows first in the valleys where irregular flooding is a problem and

34

irrigation is easy. In fallow systems such valleys have often been marginal lands, and
among the last areas to be cleared for cultivation. With increasingly skillful irrigation
they become prime rice fields. The widely credited fertility of rice fields (e.g., Pearce,
Barbier, and Markandya 1990) actually is often created by decades of careful husbandry
on soils that may originally have been quite poor (Bray 1986). Where population growth
continues, farmers start the laborous conversion of lower slopes into rice terraces.
Improvements in water control, fertility management, and crop husbandry can typically
increase output greatly with modest or no declines in labor productivity (Geertz 1963;
Bray 1986). Fertilizer, pesticides, and improved varieties tend to have higher payofl‘s than
in most other farming systems (Ruthenberg 1976; Byerlee 1990). In much of Asia, wet
rice systems support very high rural population densities, often exceeding 3000 persons

per square kilometer.

In contrast to humid and semi-humid climates, where the transition to permanent
upland cultivation is difficult but irrigation has great potential, farmers in semi-arid
climates can more easily intensify from fallow systems to continuous cultivation, but tend
to have great difficulties in intensifying further, except in the vicinity of irrigable water
sources. 0n the other hand, the Machakos success story mentioned above (T ifi‘en,
Mortimore, and Gichuki 1994) comes from a semi-arid area, showing that considerable
intensification within the system of permanent cultivation is possible. Soil and water
comervation works can reduce erosion and nutrient losses and increase the availability of
water for plants. The development of animal traction is often the bottleneck. Physical
responses to mineral fertilizer are often stronger than in more humid regions, but the
profitability of fertilizer naturally depends on relative prices. The semi-arid cases
included in Turner 11, Hyden, and Kates (1993) were not particularly successful,

experiencing both environmental degradation and declining per capita production levels.

35

Although cooler climates make tropical highlands more conducive to permanent
cultivation than the lowlands, suggesting better prospects for intensification, many of the
highlands already have high population densities. Moreover, highlands often have steep
slopes and poor, shallow soils, which make them vulnerable to erosion. In Rwanda and
Burundi, for instance, the chain of mountains that form the divide between the basins of
the Congo and the Nile is a particularly poor and degraded region. Verhaegen and
Degand (1993) report how intensification is failing on Congo-Nile Divide in northern
Burundi. Soil fertility is low and declining, and pests are attacking beans and sweet
potatoes, the main staples. Marginal returns to labor are collapsing with population
growth. No innovations capable of bringing about sustainable intensification have been
widely adopted. Uwizeyiarna (1991) describes similar problems in Rwanda, concluding
that agricultural intensification there is reaching a dead end.

Some highlands have much better prospects due to their volcanic soils that tend to
be highly fertile, impermeable and hence not very prone to erosion (Ruthenberg 1976;
Pearce, Barbier, and Markandya 1990). For instance, in contrast to the generally ’
impoverished Congo-Nile Divide highlands, the volcanic part of the divide, including
much of the Ruhengeri prefecture, is Rwanda’s potato basket with high yields, reasonable
incomes and moderate levels of soil losses. Ford (1993) concludes that Ruhengeri is
“marginally coping” with population growth, and Hyden, Kates, and Turner 11 (1993)
label it a qualified success story.

Outside the highest altitudes that are too cool for bananas, tropical highlands often
can intensify sustainably with banana groves. A well-documented case is that of Bukoba,
Tanzania. On relatively poor soils, “0.3 hectares of bananas mixed with beans are
sufficient to cover the basic food requirements of one family” (Ruthenberg 1976: 239).
Although the high productivity of banana gardens was in part achieved by importing
nutrients from degraded pastures, the system appeared capable cf supporting very high

36

population densities (Friedrich 1968). Some of the banana gardens had existed for a
century or more, had virtually no erosion, and showed stable or increasing yields. In the
19705, however, the banana groves were attacked severely by new pests, which
rmdermined the viability of the system (Kajumulo-Tibaijuka 1984).

In sum, the scope for agricultural intensification in the tropics depends greatly on
environmental conditions such as rainfall, soil type, steepness of slope, and temperature.
Prospects are good on volcanic highlands that enjoy fertile soils, abundant rainfall and
modest rates of decomposition. Also farmers in semi-arid tropics can often intensify
relatively sustaimbly fi'om fallow systems to permanent cultivation. In most of humid
and semi-humid tropics and on the steep slopes of tropical highlands, sustainable
intensification from fallow systems to permanent upland cultivation of annual crops is
extraordinarily difficult. Technically, soil fertility can be maintained either with
industrially produced inputs or with an organic approach based on very large labor inputs,
but returns on these inputs tend to be low. In these conditions, sustainable intensification

has typically been based either on irrigation or on perennial crops.

3.23 Market Incentives
Market conditions can influence intensification in multiple ways. First, market

production can make intensification profitable and provide farmers with the resources to
invest in land. Second, access to markets and relative market prices influence crop choice.

Third, input markets may promote or discourage specific forms of intensification.

In the case of perennial crops, the crucial role of market access is well rmderstood
(Ruthenberg 1976). Most perennial crops represent intensification: they provide high
returns to land and much employment per hectare, thereby supporting high population
densities. Most provide better crop cover against erosion than annual crops. Most are
produced entirely or mostly for sale, often to overseas markets. Many need local

37

processing, and for some, production is feasible only in the neighborhood of the
mocessing facilities. This description applies to tree crops such as oil palm, rubber,
cocoa, coconuts, and cloves, to shrub crops such as coffee and tea, and to perennial field
crops such as sugar cane, pineapples and sisal. Bananas are a partial exception in being

also a major subsistence crop.

When farmers facing increasing land scarcities voluntarily choose perennial cash
crops over subsistence food crops, the choice typically helps themin coping with the
scareities. Examples include smallholder tea in Meru, Kenya (Hyden, Kates, and Turner
II 1993), coffee in Bukoba, Tanzania (Friedrich 1968), and beer bananas in Rwanda (Bart
1993). No such positive presumption can be made if farmers have been forced to grow
perennials. Laure (1986) and Braun, Haen, and Blanken (1991) found that tea produced

lower returns to land and labor in Gisenyi, Rwanda than the most lucrative alternative

9

crops.

Similarly, in Clay et al. (1995) we found that at the prices that prevailed in 1990,
cofi‘ee was not an attractive crop for many of the Rwandan smallholders who had cofi'ee
bushes, which they were not allowed to uproot.

In several of the cases included in Turner 11, Hyden, and Kates (1993), market
gardening for nearby urban areas was instrumental in justifying and financing purchased
inputs and labor-intensive investments in land. The Machakos success story reported by
Tifi‘en, Mortimore, and Gichuki (1994) was largely based on terracing and other
investments and inputs paid for and made lucrative by the production of non-food and
horticultural products for the Nairobi market. Both Hyden, Kates, and Turner 11 (1993)

 

’ This conclusion probably only holds for tea grown on high-quality fields, es ially in
the wet valleys suitable for sweet or white potatoes. The opportunity cost 0 tea is much
lower on the thin soils of acidic hills, where bananas do not grow at all and where
annual food cro 8 would grow only poorly and lead to massrve erosion. Moreover, the
privateprofitab 'ty ofteaappearstohavebeenpoorianandaatleastinpaltbecause

38

and Pearce, Barbier, and Markandya (1990) conclude that further intensification in
densely inhabited highland areas is likely to involve more fruits, vegetables and other
special products. Increased imports of starchy staples from elsewhere are typically

essential for intensification in the highlands.

In contrast to perennials that typically provide good crop cover against erosion,
vegetables and other annual crops produced for urban markets are often highly erosive.
Pearce, Barbier, and Markandya (1990) report how the lucrative market for temperate
vegetables has encouraged upland farmers in Java to cultivate steep volcanic slopes high
on the mountains. Since volcanic soils are inherently fertile and not very prone to erosion,
the losses of topsoil appear to have little impact on yields. Although erosion imposes
significant off-site costs, its real on-farm costs are modest and perceived costs even
smaller. Hence, farmers see no reason to invest in soil conservation. The authors,
however, see this as an exception to the normal rule that good farm prices are a key to

soil conservation and sustainable intensification.

The argument for good overall incentives does not apply for areas where farmers
would respond primarily by expanding agriculture to new areas. Neither does it carry to
individual crops. Improved incentives to grow erosive crops can clearly make agriculture
less sustainable. In Java, Pearce, Barbier, and Markandya (1990) were worried about the
substantial price increases for cassava, a highly erosive cash crop.'0 Similarly, Verhaegen
and Degand (1993) lament that the most lucrative cash crops in Burundi (tobacco and

potato) are highly erosive.

When important quantities of purchased inputs are used in agriculture, input

prices play a key role in determining incentives. When discussing the development

 

of the indirect taxation of farmers through overvalued exchange rate, distorted pricing
structure, and inefficient parastatals (World Bank 1991).

39

options of various topical farming systems, Ruthenberg (1976) repeatedly refers to
relative prices, especially to the price ratios between fertilizer and farm products.

Although almost all input use can contibute to intensification, some inputs may
be unsustainable, bringing short-term gains at the expense of long-term production losses.
Fertilizer is often regarded such an input. Unlike in the industrialized countries and in
parts of Asia, excessive nutients are seldom the problem in Africa, Instead, the concern
is about the loss of organic matter, which may ensue if cheap fertilizer encourages
farmers to neglect soil conservation or the organic fertilizers that would provide both the
nutrients and the humus. With this reasoning, Pearce, Barbier, and Markandya (1990)
suggest that fertilizer subsidies in Indonesia should be abolished. McIntire, Bourzat and
Pingali (1992) argue that impediments for the use of mineral fertilizer are encouraging
the use of manure and other organic fertilizers in sub-Saharan Afiica.

On the other hand, mineral fertilizer increases crop growth, which means better
crop cover against erosion and provides more organic matter that can be used to improve
soil quality. Ruthenberg (1976) believes that fertilizer generally increases the stability of
topical agriculture. McCown et al. (1992) started their research in semi-arid Kenya by
focussing on biological fixation of nitogen by legumes but ended up advocating the
fertilizer-augmented soil enrichment (FASE) strategy that emphasizes the importance of
using both organic and mineral fertilizer. Similarly, a German project (PAP) that
sponsored research on “ecofarming” in Rwanda and originally had been quite critical
towards “artificial fertilizer,” found that the leguminous plants that were cental on their
agenda were diflicult to establish on degraded lands without mineral fertilizer (nger and
Rottach 1983; Adelhelm and Kotschi 1985; Egger and Martens 1987; Kotschi, Pfeiffer,
and Grosser 1983). Kotschi (1991) concluded that improved fallow complements but

 

'° Unlike in Afiica, where cassava is usually a subsistence food crop, it is grown in Java
primarily for export to Europe as animal feed.

40

cannot substitute other forms of fertilization. This conclusion is also shared by
Balasubramanian and Blaise (1993).

Comparable arguments can be presented for and against other agricultural
chemicals. While the standard argument against “environmentally unsustainable”
pesticides that supposedly destoy the ecological balance (Buchholtz 1984) has merit in
many cases, certain forms of sustainable intensification can be highly dependent on
industially produced chemicals. For instance, zero-tillage, which is increasingly used in
mechanized agriculture to save fuel and/or to conserve water and which might have
substantial potential for soil conservation in the topics is highly dependent on herbicides
and insecticides (Ruthenberg 1976: 159; Byerlee 1996).

In sum, sustainable intensification is most likely in market conditions that reward
farmers for investing in their land, for using inputs and methods that increase and
maintain productivity, and for choosing crops that provide high returns without mining
the soil. In general, good access to well-functioning markets allows farmers to specialize
based on their comparative advantage. With markets, land-scarce farmers can deviate
from their consumption needs towards products that give higher returns to land, and
farmers on steep slopes can focus on crops that provide good protection against erosion.
Lucrative agricultural terms of trade promote input use and investnents in soil

conservation, but may also encourage expansion to new areas.

3.2.4 Availability of Intensificatlon Techniques
Boserup (1965) recognized that failure to invent or adopt appropriate methods of

intensification may lead to severe problems, but nevertheless assumed that most farmers
have several options and can respond to population growth by choosing more labor-
intensive practices. However, Ruthenberg (1976) described in detail how some

pepulations had a variety of attractive intensification techniques to choose from, whereas

41

others had no viable techniques available and were either getting poorer or migrating
elsewhere as population grew. Although migration may indicate superior opportimities
elsewhere rather than dearth of intensification options in the current location (Bilsborrow
1987), it is clear that the availability of intensification techniques varies considerably in
time and place. Lipton (1990) argues forcefully that the principal constaint of
agricultural growth in much of Afiica is the poor supply of improved agricultural
technologies.

While some agricultural intensification techniques can be invented by farmers
themselves or borrowed from elsewhere, many are results of organized agricultural
research. Although such research can be seen as just another form of the Boserupian
population-induced intensification process (Boserup 1989; Hayami and Ruttan 1985),
time lags in agricultural research are long and the results available to farmers depend on a
large number of factors, only one of which is the population growth that may have pushed
governments, donors, or private companies to invest in agricultural research. Hence,
when analyzing the responses of rural households to population pressure, the availability
of intensification techniques is often most usefully teated as an independent factor that
conditions the responses. A further argument for this approach is the fact that agricultural
research is one of the key policy levers that donors and governments can pull in order to

make intensification more successful or sustainable.

In nine of the ten densely populated districts examined by Hyden, Kates, and
Turner II (1993), including all the successful cases of intensification, new crops or
improved varieties had played an important role. In Ruhengeri, Rwanda, for instance,
progress had been partly based on the development of improved potato varieties, the
provision of fungicides they needed, and the constuction of a paved road from the region
to the capital, the main market for potatoes (Ford 1993). When the temperate crops
(potato and wheat) had originally been intoduced druing the colonial period, the

42

intensification prospects at the high altitudes had greatly improved. Previously
intensification even on the fertile volcanic lands had been hampered by lack of suitable
crops, because the crops that yielded well at lower altitudes (bananas, beans, sweet

. potato) did not grow well on the cool highlands (Nduwayezu 1990).

In Machakos, Kenya the “inflows of ideas from external sources” were crucial for
successful intensification (T ifi‘en, Mortimore and Gichuki 1994). The sources included
not only extension agents and development projects but also many other sources. Indian
customers in Nairobi, for instance, provided farmers with the seeds of the Indian
vegetables that they wanted to buy.

In sum, sustainable intensification is made more likely by the availability of
appropriate technologies. Over time, the development of new technologies is a key issue,
whereas in cross-sectional analysis differences in information about the available

technologies can explain differences in intensification.

3.2.5 Land Tenure
For Boserup (1965; 1989), the emergence of individualized land tenure was an

integral part of the intensification process induced by population growth. In long-fallow
systems individual tenure of land during the fallow period is neither necessary nor easy;
enforcing such property rights is seldom worth the touble. Since virtually all farm work
is oriented towards short-term benefits during the short cultivation period, communities
generally recognize and enforce only use rights during cultivation. As population grows
and fallow periods get shorter and finally disappear, investments in land improvements
become economically attactive for the community. They are also privately profitable for
households if those who invested can count on getting the benefits. To ensure this,
farmers generally claim and communities recognize stonger property rights on land,
which gradually lead to individualization.

43

As already mentioned, Boserup studied slow changes that gave societies hundreds
of years to adapt to new situations. She fully recognized that where populations double in
20-30 years institutions such as land tenure may fail to change fast enough, and
inadequate security of tenure may constain intensification. Not surprisingly, the
population explosion of the past decades has created a lively debate on whether or to
what extent lacking security of tenure is causing inappropriate incentives, and what, if
anything, governments should do to provide security of tenure. At one exteme are those
who assume that the lack of individual property rights leads to the “tagedy of commons”
(Hardin 1968) and advocate rapid privatimtion and formal registation. At the other
exteme are those who attibute problems primarily to externally caused disruptions in the
local governance of the commons and advocate measures to stengthen it (Blaikie and
Brookfield 1987).

Empirical evidence fi'om Afiica suggests that while land tenure in many parts of
Afiica is gradually becoming more individual and less communal, an abrupt tansition to
fieehold tenure has only limited benefits and often contibutes to landlessness (Bruce
1986). Moreover, a wholesale tansfer of the land tenure administration from taditional
institutions to the government is very costly and may decrease rather than increase the
security of tenure. Therefore, the emerging consensus on land tenure calls for policies that
prevent outsiders (including government oflicials) from grabbing land, establish some
ground rules for the fair functioning of local taditional institutions of land tenure
administation without tying to take over the entire operation, and promote the gradual
individualization, recognizing that communities will differ in what they privatize. For
example, commrmities may move at their own pace from a situation where cultivation
rights are private but grazing rights during fallows are communal into a situation where

those planting tees or leguminous fodder crops on their fallows are allowed to exclude

others’ livestock by constucting fences, and ultimately into a situation where livestock

ownersarerequiredtokeeptheiranimalsout.

Another incentive problem associated with tenure arises when owners do not farm
their lands themselves. According to the usual argument, the tenant has no security of
tenure and therefore no incentive to invest in rented land. However, this does not explain
why the landlord would not find it profitable to pay for the investment and then increase
the rent to reflect the higher productivity of the field. Alternatively, landlords and tenants
could agree that in the end of the lease period landlords would compensate the tenants for
any improvements made. That these options are not entirely unrealistic can be seen in that
the Asian lowlands with their extaordinarily high investments in the creation and the
maintenance of irrigation systems are often operated by tenants and owned by absentee
landlords. Thus, the usual argument must be replaced by a more complex one that
emphasizes the costs of designing and enforcing contacts. Two related approaches .
known as the information-theoretic approach (Stiglitz 1986) and the tansaction costs
approach (Williamson 1985) offer different terms and techniques, but both see the -
problem as one in which costs related to uncertain or asymmetic information prevent

landlords and tenants from having contacts that would give tenants the same incentives

 

45

to invest as an owner-operator would have." For instance, the costs of supervising how
much tenants use manure on rented fields may be so high that landlords and tenants may
find it advantageous to specify nothing even if the result were that no manure will be used
on the field.

In sum, investnents in soil conservation, water contol, and perennial crops are
typically crucial for sustainable intensification and are more likely to happen when land
tenure arrangements and other institutions are such that those who invest in land can also

reap the benefits. To some extent, evolution towards such institutions is induced by the

pressures to intensify.

3.2.6 Capacity to Invest
Even when physical and economic incentives to invest in land and to adopt the

appropriate intensification techniques are in place, households may not have the resources
to invest (Reardon and Islam 1989; Clay and Reardon 1994). Fundamentally, this may
also be seen to result fi'om high tansaction and information costs that lead to failing
credit markets. The practical implication is that wealth and incomes influence '
intensification.

As intensification measures themselves can increase incomes or arrest declines
that would otherwise occur, the intensification process is characterrzed' by “multiple
equilibria.” In other words, relatively small differences in initial conditions may put
households and communities on diverging paths towards stikingly dissimilar outcomes.
For instance, a household fortunate enough to own a head of cattle may reach higher
incomes with the manure, may therefore be able to invest in water contol and, with the

additional income, enter a virtuous cycle. Meanwhile, a neighbor without the animal may

 

" Also in the privatization debate above the root cause of the problem is tansaction costs.
In their absence, the degradation of and underinvestnent in the commons could be
avoided by appropriate contacts between communities and their members.

46

be forced to “mine the soil” in an increasingly vicious cycle. Similarly, ofi-farm incomes
may be the source of liquidity that allows households to invest in intenfication (Reardon
1997).

3.3 Chapter Summary
Although much of the literature is vague or careless about the multitude of factors

affecting the responses of rural populations on demographic change, and although many
scholars make sweeping generalizations without spelling out the specific conditions
under which the generalizations were observed or would be expected to hold, there
nevertheless seems to be a wide consensus on the factors that matter. The prospects of
intensification depend on environmental conditions, market incentives, availability of
intensification techniques, on institutional arrangements such as land tenure, and on the

capacity of the households to invest in intensification.

47

Chapter 4

DATA AND THEIR LIMITATIONS

4.1 Introduction
The empirical analysis in this study is based on national farm survey data mostly

fi'om agricultural year 1990”. The farm surveys were carried out by the Division of
Agricultural Statistics (DSA) of Rwanda’s Ministy of Agriculture and Livestock
(MINAGRI) with assistance from the United States Agency for International
Development (U SAID) through a cooperative agreement with Development Alternatives
Inc. (DAI) and Michigan State University (MSU).

4.2 Sampling Design
The sampling procedures of the DSA were designed primarily to provide

nationally representative statistics on Rwanda’s agricultural economy at reasonable costs.
A secondary objective was to provide information on regional variation. To achieve these
objectives, DSA used statified random sampling with 21 stata. Each statum
corresponds to a unique combination of agro-ecological zone and prefecture. The idea
was to design a sample that would give reasonably accurate estimates for each or
Rwanda’s five main agro-ecological zones", and hence, for Rwanda as a whole, while
also allowing for disaggregation of national aggregates by prefectlte, which was
inevitably less reliable but nevertheless wanted by policymakers.

 

‘2 The definition of “agricultural year” used by DSA has varied over time; in 1990, the
agricultural year ran from September 1989 to August 1990.

‘3 These were called geographic regions in section 2.1., where also the limitations of this
scheme of regionahzatron are discussed.

48

Depending on the size of the stata, 2 to 8 administative sectors were randomly
selected within each.“ Altogether, 78 sectors out of the 1,490 in the county were chosen.
Next, to reduce further the costs of tavelling by enumerators, primary sampling units
were selected by choosing randomly one census distict from within each sector. Finally,
samplers resorted to cells, the smallest administative units, corresponding roughly to 100
families. Four cells were randomly selected from within each distict. Only afier these
steps were households selected randomly from within each cell.

The selection was made out of a list of all farm households. Since the intent of the
survey was to cover all agricultural production, the definition of farm households was
broad. All rural families that engaged in agricultural activities on their own account were
considered farm households. In rural Rwanda, this definition includes most but not all
established households and excludes mostly young wage laborers and salaried employees
who no more live with their parents. The sample included a handful of households that
had household members employed in skilled (school teachers, masons) or unskilled

(laborers) occupations and earned most of their income from ofl‘-farm activities.

Within each cell, four households were randomly selected for the intensive
sample. In addition, four households were selected for the extensive sample and four
more for the reserve sample. The purpose of the extensive sample was to test whether less
frequent interviews and hence less costly data collection could yield sufficiently reliable
results. No data on the extensive sample have been used in this study.

The sample included 78 primary sampling units (‘Villages”"), each covered by

one full-time enumerator. Each enumerator visited four households every week in each of

 

“ More precisely, the random selection of sectors had been made for the 1986-1988
sam 1e and was largely kept in place for the 1989-1992 sample so that the enumerators
wo d not have to relocate.

‘5 “Village” is a somewhat roblematic term since rural Rwandans live scattered on the
hills, not in clusters clear y fiom one another. This, of course, is also tue for
many “villagers” of the worl .

49

I
l

K
* \

‘.\1

the four cells (“sub-villages”) chosen fi'om his sector. Whenever households dropped out
of the intensive sample for whatever reasons, they were immediately replaced by
households taken tom the reserve sample. While the replacement in data collection was
immediate (so as to avoid distorting the incentives of enumerators), both the new and the
replaced households were typically dropped out of the analysis for the season when the
replacement happened.

4“th

\The sample rs not self-weighting. 3Although the number of sectors chosen from

\_———--~- ”4‘“. (##me

each stata were roughly proportional with population, this adjustment was not perfect.

 

More importantly, within stata, households from sectors with smaller populations were
more likely to be selected than households from elsewhere. Sample weights to adjust for
these biases were computed based on the population estimates tom the 1978 census.
Whenever a household was dropped from the analysis because of missing or poor data,
the weights were adjusted accordingly

A...“

.. "‘\-...._ “..d

The sample used in this study was selected rn 1988, and data on it were collected
for the crop years of 1989, 1990, 1991, and 1992. Since late 1990 (crop year 1991), data

 

collection in parts of northern Rwanda has been disrupted by the civil war. Hence, the

most complete data is for crop year 1990.

Despite the disruptions and the intensive nature of the survey, most households
remained in the sample throughout the period. Four out of five households included in the
beginning of crop year 1989 were still interviewed weekly three years later.

4.3 Surveys Conducted
The basic task of agricultural sample surveys in most developing counties is to

collect data on agricultural production. In Rwanda, the first nationally representative farm
survey was carried out in 1983-1984, and since crop year 1986 DSA had collected

50

production data annually. Surveys on land use were added in 1989 and those on

agricultural transactions a year later.

Despite the expansion, production surveys remained a core activity and the only
survey requiring each enumerator to visit and interview each household every week. This

was deemed necessary in an environment where crops are harvested throughout the year
/and not just once or twice a year. Farmers were provided with standardized buckets and

‘

 

 

 

asked to report harvested output in terms of buckets.

V Data on revenues and expenditures (and gifts) were collected monthly. Despite its
1

 

name, the revenue and expenditure survey was not a comprehensive household budget

survey. We side covered also wages and salaries and not just agricultural

sales, the expenditure side focused exclusrvely on the purchases of agricultural products,
W, Wm“... .. .. .

labor and inputs. Even with this omission, the survey was very laborous, and included

.____..——-————--"""

more numbers to key in than all the other surveys combined. For most tansactions, both

physical quantities tansacted and amounts of money paid or received (if any)w-ere
, recorded.

Land use data were collected twice a year (once per season). Unlike production
and tansaction data, the land use data were based on direct observation. Enumerators
measured the dimensions and angles of each field. If the field was intercropped,

- enumerators estimated the planting densities of all the principal crops, using the local
(1 densities of pure stands as a yardsticks. The focus was on operational holdings, including
both owned and rented fields, but excluding fields that were rented out for others. Only
one-half of the fields were included in the 1989A season and three-quarters during the
next two seasons. Complete land use data for the entire intensive sample is available for

seasons 1990B, 1991A and 1991B.
\_

/

51

 

Two types of livestock surveys were fielded regularly by DSA. Livestock
inventory data were collected once a year, while animal production surveys were
conducted quartely.

Demographic information on household members was collected once in 1989 and
again during the 1991A cropping season. The demographic survey included questions on
age, sex, literacy, education, and occupation.

In addition to the regular surveys, DSA conducted some special surveys. For our
pmposes, the most important is the survey of agroforesty fielded in 1991. Despite its
name, the agroforesty survey had a broad focus, including questions on farm
productivity, soil conservation, and farming practices. The agroforesty survey also asked
whether each parcel was owned or rented.

No data were collected on the use of family labor on the farms. The demographic
data sets tells how many members were primarily farming in each household, but there is
no information on who did what and how much labor was devoted for agriculture, not to

‘ speak of how much was devoted to each crop.

Another significant omission concerns wealth. While data were collected on the

i ' t
two most important asset classes in rural Rwanda, land and livestock, the focus was on

M

 

operational holdings. No information was gathered on land rented out or livestock lent to

 

52

others. “ Further, no listings were made of possessions such as bicycles or radios or of
other potential proxies of wealth such as the type of roof on or the number of windows in
the house.

No commrmity-level data on agroclimatic conditions, health facilities, market
places or roads were collected as part of DSA’s regular survey program. However, on the
basis of detailed maps and with the assistance of Jean Bosco Sibomana, the former
national supervisor of DSA, we were able to add to the data set sector-level estimates of
altitude, rainfall, and distance to paved road.

4.4 Quality Control and Data Cleaning
To ensure the quality of data collected, DSA relied primarily on continual taining

and close supervision of its 78 enumerators. It had full-time supervisors in each
prefecture, so that the performance of each enumerator could be monitored weekly. The
supervisors visited enumerators weekly, collected filled questionnaires fi'om them,
screened for inconsistencies and anomalies, and went back to the farms with the
enumerators if needed. Supervisors themselves were monitored by a national supervisor,
and also DSA’s analysts frequently made field visits to see how their questionnaires were
understood and interpreted and how enumerators adhered to the standard operating

procedures.

In the DSA’s oflice, questionnaires were again inspected visually before data 3
enty. However, most data cleaning occurred after data enty. Complicated data cleaning "
programs were used to screen for anomalies and to print out listings of suspicious cases
(Lassiter, 1992; Kangasniemi, Riley, and Peters, 1993). With the list in hand, data

cleaners rechecked the questionnaires and reentered data. Decision rules for anomalies

 

" It is common in Rwanda for r households to borrow livestock from their better-off
neighbors. In exchan e for ' g care of the animals, the poor get manure, which helps
them to maintain so' fertility.

53

not attributable to data enty errors were unclear and little or no documentation exists for

how or whether anomalies were corrected.

All data cleaning in the DSA office was done survey by survey. No formal
procedures were in place to ensure that information gathered in one survey was consistent
with that gathered in another. Although supervisors and enmnerators informally could

weed out exteme inconsistencies, many anomalies remained.

As many of the computations made for this study were based on many surveys,
they inevitably revealed inconsistencies such as sales of a given crop by households that
had neither produced nor bought the product. Although it was not possible to address all
or even most of the anomalies, some cleaning was made and a handful of households with

very anomalous data was dropped out of analysis.

4.5 Data Transformations and Resulting Variables

4.5.1 Overview
The creation of analytically interesting variables out of the raw data collected by

enumerators requires complex tansformations based on numerous definitions and
assumptions. While the tansformations can be understood fully only by studying what

the hundreds of lines of tansformation algorithms actually do, this section seeks to
describe some of the key definitions and assumptions needed to assess the potential and
the limits of the data. Most of the tansformation algorithms used by DSA were created ‘2‘
by Greg Lassiter.

4.5.2 Production, Land use and Yield Data
The raw production data were in terms of buckets by crop and form of the

product. Using standard conversion factors, quantities were converted into kilograms and

non-typical forms were converted into the equivalents of the preferred form (e.g., green
beans were converted into dry-bean equivalents). Note that all production data were

54

collected at farm level; there is no way of telling how much beans or bananas came fi'om

any one field or whether fields close to the compound were more productive than more
distant fields.

Land use data, in contast, was originally at the field level. The raw data on land
area consisted of measurements of angles and lengths of each side of each field, which
were quite staightforward to convert into ares (100 ares = 1 hectare). Aggregated over
the fields of each household, this gave an estimate of operational holdings.

Estimating land use by crop was more complex because of the prevalence of
intercropping. The raw data included planting densities of the associated crops for each
field, expressed as a percentage of the typical planting densities of these crops on purely
cropped fields. These “lmstandardized densities” typically added up to more than 100
percent. To allocate the estimated area of the field between the associated crops, the
densities were scaled down (divided by their sum) to create “standardized densities” or
shares that added up to 100 percent. If, for instance, the unstandardized densities of a 10
are plot were 60 percent for beans, 60 percent for sweet potatoes and 30 percent for
maize, the standardized densities (shares) were 40 percent, 40 percent and 20 percent,
respectively, and the household was considered to have 4 ares of this field lmder beans, 4

ares under sweet potatoes, and 2 ares under maize.

With farm-level estimates of output and area by each crop at hand, yields can in
principle be computed. In practice, there are several reasons why household-level yield
estimates are highly unreliable. First, although the use of most fields remains constant
over the season, some crops are harvested throughout the year, implying that a significant
minority of the fields change fiom one use to another during the season. For instance, a
sweet potato field may be harvested and replanted with cassava just before the

enumerator arrives to observe land use. Thus, the recorded sweet potato output for the

55

season can be hundreds of kilograms, even though the recorded area under sweet potato is
zero. Ifthe farmer in this case had sweet potato on no other fields, the computed yield
would be infinite. And since the cassava that replaced sweet potato would not be
harvested during the season, its estimated yield would be zero.

Another explanation for the extemely noisy farm-level yield data has to do with
minor crops on associated fields. For simplicity, enumerators were advised to ignore
crops with lmstandardized densities below five percent. Because of this simplification,
areas under minor crops are understated on many farms and, if calculated, household-
level yields would be overstated (and infinite in some cases).

While these problems cause anomalies at the household level, there is little reason
to think that they systematically bias either land use or production data. Therefore, these
problems do not distort national or other aggregated yield estimates computed by
aggregating hectares and kilograms first and then dividing."

Thus, the frequent allegation that the prevalence of intercropping leads to a
significant acmss-the-board underestimation of yields and land productivity, is not tue
for DSA’s data. It is an argument that holds for farm surveys that altogether ignore
significant minor crops. In DSA’s system, the output of minor crops is included even
when these crops were left out from land use statistics. At most, the errors in the

measurement of land use caused by intercropping can increase the estimated areas and

 

‘7 Note that the aggregation of household-level yieldestimates leads to aggregate
estimates that are biased downward, since it effectlvely excludes the “anomalous”
kilo (infinite yields) but includes the analogously “anomalous” hectares (zero
Eel ). All yield estimates computed for this stu 3 were made by aggegating first

'lograms and hectares and.then dividin . Some SA publications, ough, lnclude
estimates com uted by; the incorrect me od, underestimating yields slightly (0-8%,
on e crop . ~

56

hence depress the estimated yields of some major crops by reducing the area estimates
and inflating the yield estimates of some minor crops."

It is most likely to be tue that DSA’s numbers underestimate agricultural
production, yields and food availability, but the reason is not because the data collection
procedures failed to account for intercropping. Instead, the principal culprit is likely to be
underreporting, especially of those products that are harvested in small quantities
throughout the season.

4.5.3 Market Transactions and Rural Exports
The raw data on market tansactions included numerous units and several forms of

many products. As in the case of production data, the first step was to convert non-
standard units into kilograms and several product forms into one per each crop.

One problem with the rural tansaction data is the fact that some farm households
are also part-time taders, buying products in order to sell. At the same time, many ,
households buy and sell the same crop without being taders, for instance because they
run out of their own supplies before the harvest. To study the degree of rural
commercialization, the tansactions of the latter should be included, whereas those of
taders should be excluded in order to avoid doublecounting. In practice, we sought to
separate the two by netting purchases and sales monthly. For instance, a household that
bought 300 kilograms of potatoes in March and sold 500 kilograms during the same
month was assumed to be tading 300 kilograms and selling 200 kilograms from his own
production. In contast, a household which bought 40 kilograms of beans in November
and sold 50 kilograms of beans in January was assumed to be selling the full amount.

 

" At the national level, this dippears to be a theoretical curiosity with little impact on the
estimates. At sub-national evels, maize yields could be somewhat afl‘ected, because
maize is fi'equently grown as a minor component in intercropped fields.

57

The monthly netting adjusts gross sales only modestly to eliminate trading. If
netting is done over the year and over many or all surveyed households, the results are
called “net sales and gifts” or “rural exports?”9 These no more indicate the importance of
markets for producers or consumers but instead estimate the quantities of farm products
moved from rural areas.

4.5.4 Income and Its Components
Since total household income is such a comprehensive concept, using farm survey

data to compute estimates for total household income in semi-subsistence agriculture
requires many conceptual decisions on the definition of income, a large number of
computations and tansformations, and numerous shortcuts and assumptions to deal with
data limitations and computational constaints. Subsistence consumption must be valued,
and adjustments may be needed for seeds and other inputs and for cross-household

differences in the purchasing power of money.

The complex computations used to derive income estimates from the DSA survey
data were originally created by Scott Loveridge (1992), and later substantially modified
by the present author. Despite many changes that appeared quite significant, our revised

income estimates were highly correlated with those of Loveridge.

Definition of income. Income is defined as a sum of the value of home
consumption of main crops net of purchased inputs, sales of agricultural products net of
purchased inputs, net gifis received, wages earned, and other incomes such as land rents
received.” Among the items that conceptually should be part of the definition but are not

included in the estimates due to unavailability of data are (1) home consumption of minor

 

‘9 Thus, our terminolo difi‘ers from that used by Loveridge (1992). His “net sales” is
close to our “adjust gross sales”, although his adjustnents moved more tansactions
fi'om sales and purchases to tading than d ours.

3° Home consumption is defined as production minus sales and gifts given (both adjusted
monthly to account for tadlng and recrprocal exchanges). Thus, the income estimates
exclude in-kind gifis given and include in-kind gifts received.

58

crops, (2) home consumption of livestock products, and (3) the value of the meals
provided by employers and relatives.

Production of minor crops such as cabbages and tomatoes was not asked in the
DSA surveys, although all sales, including those of minor crops were recorded. Hence,
home consumption cannot be estimated for minor crops. Collecting information on the
minor crops would have increased costs, and might even have reduced the quality of A
income estimates by adding to the complexity of the farm survey. An earlier national i3
household budget survey (ENBC Rural) found that the 15 main crops (or 17, ifthe three ‘
types of bananas were considered separate crops) included in the DSA survey accounted l"
in the mid-19805 for more than 97 percent of all calories consumed with much of the rest 1‘.

sq

coming from industial products like sugar and edible oils (MINIPLAN 1988). \}

Livestock products are like minor crops in that only tansactions were recorded,
implying that production for home consumption was not included in income estimates.
The omission is not as important as it appears, since while most rural households practice
mixed farming, livestock is kept largely for manure and for cash income, not for home ;\
consumption. Most rural Rwandans are too poor to consume sizable amounts of meat, / \
eggs or milk. According to the ENBC rural budget survey the 15 main crops accounted i
for more than 96 percent of all proteins consumed (MINIPLAN 1988). \J

More significant is the omission of the meals provided by employers and

relatives. These are clearly important, since especially the poorest households are known
to send their children to eat with the relatives or to work for richer households, ofien

primarily for wages in kind. In some parts of the county, particularly in the Northwest, it
is customary for the employer to feed the employees during the working day. The
omission of these meals presumably is one reason why the food availability estimates for

59

the very poorest households are often impossibly low (<1000 kcal/day/adult equivalent
on average). Unfortunately, little is known of the magnitude of this omission.

Valuation. A common Wm use national producer price
estimates to value production and home consumption, whereas household budget surveys

Maw-

frequently value home consumption by using national consumer price estimates. In other

words, farm surveys effectively assume that the alternative for home consumption is to

 

sell at thejah—tional producer price, whereas consumption surveys assume that the
f.“

alternative is to buy at the national consumer price. Both approaches are simplifications

 

in that they overestimate the relative importance of home consumption of a given product

in areas where its price is low and underestimate it in areas where its price is high. For

instance, if we used the national mean potato price to value the home consumption of

potatoes in Rwanda’s northwestern potato zone, where potato prices are low, we could be

claiming that a farmer who sold half of his potatoes got much more income from the

other balfthat was consumed at home and valued higher. Even more anomalously, a i. I) 2

subsistence-oriented potato farmer who consumed all his potatoes at home could appear 7 3 it i a.
i

more prosperous than a market-oriented potato farmer who sold all his potatoes and then

bought potatoes for food at the low local prices. -

In theory, computations should use prices that are as close as possible to the
opportunity costs that households face and then adjust for price differences to ensure
comparability. Since the DSA surveys have collected price and quantity information on
tens of thousands of tansactions in difl‘erent locations at difi‘erent times, we have tied to
get a little closer to this ideal than is common in farm household surveys. Given the fact
that most rural households in Rwanda both buy and sell farm products, and many even
buy the same crop during some months they sell at other times, we valued home \
consumption by using the average of local purchase and sales prices. We estimated

 

monthly producer and consumer prices for each of Rwanda’s ten rural prefectures by first

60

aggregating values and quantities and then dividing. Where only few tansactions were
made, annual prefectural or even annual national prices were computed instead."

While the use of local prices ensures that the estimated relative roles of home
consumption and sales are not very biased, they distort regional income comparisons by
\ exaggerating incomes in the areas of high pficeSWerstafing incomes in the areas of
low prices. To correct for this/all income components/not just home consumption, need
to be adjusted up in low-price m andtrown/m high-price areas. Our approach was to
compute household-level food price indices based on what was the estimated value of
food availability (whether produced at home or purchased) at local prices and at national
prices. Consider, for instance, a household that lives in a low-price area and was
estimated to have earned 60,000 Frw in cash from sales and 40,000 Frw in kind fi'om
home consumption (at the low prefectural prices). Suppose that 20,000 Frw of the cash
earnings were used to buy food at the low local prices so that the estimated value of food
a availability was 60,000 F rw. If the value of these quantities at national prices would have
a ’ been 70,000 F rw instead, our method would adjust for the low local prices by adjusting
.\) ./ incomes up by 10,000 Frw (10 percent). To avoid distorting numbers on market-
5 orientation, we would adjust both sales and home consumption up by 10 percent. Since
the DSA data base bad no information on non-agricultural prices, we only corrected for

   
  

' erences in (the most important) agricultural prices.

As important as the use of local prices and the corrections for differences in
purchasing power may appear in a context where regional price variation is substantial,
their quantitative significance turned out to be quite limited. The overall picture painted

 

2' Annual prefecttnal rather than monthly national rices were used, since prices varied
more across prefectures than across months. Pre ectural rather than sectoral prices were
estimated, since a vast majority of sector-month-product combinations bad no or only
few tansactions on which to base the computations.

61

by our estimates is essentially the same as that presented by the numbers computed by
Scott Loveridge, who did not adjust for differences in the purchasing power of money.

Conversion factors. Most rural households grow either bananas or sorghum or
both and sell at least some taditional beer made out of these ingredients. The amount of
bananas and sorghum estimated to be left for home consumption depends crucially on
how much ingredients are estimated to be needed for the beer that was sold. Very little is
known on the actual conversion factors. Relying on the mid-point estimates for small
beer brewers by Hagblade (1987) we assume that 3.41 kilograms ofbananas and 0.11
kilograms of sorghum are needed to produce one liter of banana beer and that 0.36
kilograms of sorghum are needed to brew one liter of sorghum beer.

Unfortunately, neither the beers nor the ingredients are of standard quality. More
bananas are probably needed at higher altitudes where bananas have lower sugar contents.
Anecdotal evidence suggests that beer produced for sale is more diluted and hence
requires less ingredients than beer produced for own consumption. Consequently, the
conversions associated with beer brewing intoduce considerable uncertainty to income
estimates.

No change in assets. Our simplified assumption that sales plus consumption
minus inputs equal production implies that changes in assets were ignored. Implicitly we
assume that there were no changes in inventories and that labor was hired to plant and
wad, not to invest in long-term biological capital or land improvements. Both
assumptions are probably reasonable simplifications: Rwanda’s main farm products are
poorly storable and therefore crop inventories and changes in them are minimal
(Loveridge 1988). Alhough investments in soil conservation, tees, and perennials are

sizable, most labor is hired to participate in current production.

62,

One important part of the omission of capital accounts is the simplified teatment
of livestock sales. For simplicity, all livestock sales revenues were considered current
income, whereas the weight gains of the animals not sold during the year were ignored
altogether. Moreover, livestock purchases which conceptually are inputs, were not
deducted fiom the sales. In the absence of capital accounts deducting livestock purchases
would artificially reduce the incomes of some of the wealthier households. On the
aggregate, the omission of livestock purchases biases livestock incomes upwards, but
since home consumption of livestock products is ignored altogether, the overall bias is
probably downward. The real problem is that income estimates for any year are highly
dependent on whether households happened to sell livestock during the year. Income
statistics overestimate the incomes and the role of livestock for some households, while
underestimating them for most others. As a result, the statistics exaggerate the

concentation of livestock incomes.22

4.5.5 Food Availability
Food availability was defined as production minus sales minus gifts given plus

pitchases plus gifis received. As was the case with income computations, the definition
efi‘ectively assumes that there were no change in inventories. As discussed above,
estimates of bananas and sorghum left for home consumption were based on the

somewhat arbitary assumptions on how much bananas and sorghum were needed for the
recorded beer sales.

Kilograms were converted into kilocalories by applying the conversion factors
from MINAGRI (1988). The availability of calories from bananas (that is, bananas not
sold either as such or as beer) was computed as such, despite the fact that most bananas
were brewed into beer. Thus, the estimated availability of kilocalories overstates the

 

22 An alternative mach not used here would be to ignore livestock sales and instead
simply assume the income from hvestock is proportional to livestock holdings.

63

likely consumption of calories. A more detailed discussion of the implications is provided
in Chapter Seven.

\ Availability was computed for the 16 main crops. According to a nationwide
) consumption survey in the mid-1980s, these crops accounted for more than 97 percent of
calories and 96 percent of all proteins consumed in rural Rwanda. More important

than the omission of minor crops, milk and other livestock products and industially
produced food products such as sugar and vegetable oil is probably the omission of meals
provided by employers. Especially for the very poorest households, who sometimes even
send their children to work for food wages, this omission can greatly understate food
availability.

4.6 Limitations and Their Implications for Analysis
Detailed inspection of small samples of households by Scott Grosse and the

present author have showed that the quality of the data set is quite uneven. Despite the
various efl‘orts to contol data quality, numerous data enty and other errors exist in the
data.

For many variables, the tails of distributions are not very meaningful. For
instance, the estimates of food availability per adult equivalent are impossibly low for a
significant minority of households. In some of these cases households have probably
understated their production or purchases or overstated sales, in other cases a data enty
error may have been the problem, in still others household members may have been fed
by their employers or may have been consuming their inventories. At the same time some
households have impossibly high availability estimates of perishable products. Income
estimates suffer from similar problems.

Excluding outliers tom analysis often helps but seldom solves the problem, since

the cut-ofi‘ points for outliers are arbitary and can easily bias results. In some cases,

64

aggegation over time, products, or households is necessary. Yield estimates discussed
above are one example; food availability estimates are another. In principle, DSA’s data
are detailed enough that food availability could be estimated month by month, so as to
study how different household groups change their diets over time and which are the
crops that carry the food insecure households over e lean n. In practice, however,
such estimates would be noisy and not particularly informative. Above all, the absence of
inventory data means that monthly disaggregation could not shed much light on whether
the food insecure households survive without storable products (beans and cereals) during
the lean period. Therefore, our analysis of food availability is based on aggregation over
12 months.

One possibility is to use methods that give less weight to exteme observations.
Throughout this study we rank households to quartiles based on income or farm size and
base discussion largely on these rankings. No reference is made to the “poorest five
percent,” or “the worst nourished one percent” since many households in these categories
would be there only because of data errors.

Finally, analyses have been repeated for several subsets, although only one or a
few have been reported here. Throuhgout, the aim has been to focus on robust findings

that do not change due to minor variations in definitions, specifications or samples.

65

 

Chapter 5

INTENSIFICATION OF LAND USE

5.1 Main Land Use Options
In this section we compare the main agricultural land use options available to

Rwandan smallholders. We review national statistics on yields (returns to land), whether
measured in terms of subsistence (nutitive) or market value. We also discuss labor
requirements of the main crops and the degree of protective crop cover that each crop

provides against erosion.

Rwandan smallholders cultivate on average two-thirds of their cultivable lands.
The rest consists of fallows, pastures, and woodlots. Fallows are the taditional way of
restoring land fertility. Along with pastures they yield benefits through grazing. Woodlots
provide increasingly valuable forest products. In general, uncultivated land uses give low
returns to land, which explains why they over time have lost ground to crops. Some
recent studies suggest, however, that tees already are highly attactive cash crops in
certain parts of the county (Cook and Grut 1990). Unfortunately, no data are available on
the typical returns to pasture and woodlots.

The main crops in Rwanda are, in the order of the land share, bananas, beans,
sweet potato, cassava, sorghum, maize, coffee, white potato, peas and taro (colocase).
Together, the main crops cover over 90 percent of the cultivated land area. Minor crops
include soybeans, peanuts, wheat, finger millet (eleusine), rice, cocoyams, tea, pyretbrum,
as well as various fruits such as papaya and avocado and many vegetables, including

cabbages and tomatoes.

66

Rwanda has two main agricultural seasons per year, although bananas, roots and
tubers, and to some extent also other crops are harvested throughout the year. Table 5-1
shows the national mean yield estimates of major crops during the six seasons of 1989-
1991. For cross-crop comparability, all yields are expressed per six-month seasons.23
Since beer bananas and cooking bananas are not separated in land use statistics, only the

average yield estimate for total bananas can be computed.

In terms of market value, white potatoes have highest yields, followed by
bananas, sweet potato, and beans. However, the high national figure for potatoes is due to
the volcanic highlands in northwestern Rwanda, where yields are more than twice as high
as elsewhere. For most Rwandan farmers, bananas are by far the most remunerative cash
crop, whereas cofl‘ee was not particularly attactive at the prices that prevailed in the late
19808 and early 19905.

 

2’ Note that this differs from the FAO practice of expressing yields per growing period for
annuals and per year for perennials.

67

Table 5-1: Characteristics of Main Crops (National Averages)

 

Yield per hectare per season Index Sales Caloric
of Crop price cost
Cover (Paw/kg) (PRw/
(C-value) 1000
(r) kcal)
Height value Energy Proteins
(kg) (‘000 (mill. (kg)
PRw) kcal)

Beans 838 29 2.50 164 19 31 13
Peas 272 16 .90 56 15 62 19
Sorghum 1016 23 3.10 73 40 19 9
Maize 1010 23 3 30 86 35 19 9
Sweet Potato 4527 40 4 20 65 23 8 9
Cassava 2185 17 2 20 11 26 9 7
White Potato 6102 78 3.50 73 22 10 26
Taro 1580 20 1 30 22 35 12 15
Cooking Banana 6788 51 S 40 48 . 4 7 10
Beer Banana 6788 31 5 40 48 4 4 6
Banana Beer 6788 47 1 30 9 4 27 41
Coffee 256 21 N.A, N.A. 2 83 N.A

Note: Mean banana yield used for all bananas.

Sources: Yields: 1989—1991 means computed from DSA/MINAGRI farm survey data;
C-valuea: Lewis (1986); Prices: 1990 means computed from DSA/MINAGRI data.

 

In terms of food energy, sweet potatoes and cooking bananas are far ahead of the
others, followed by white potatoes, maize, sorghum, and beans. Beans are superior
providers of proteins. Cassava, taro, and peas are low-yielding by every measure.
Bananas are high-yielding subsistence crops if consumed directly, but if they are turned

into beer, their yields in terms of calories or proteins are low.

Data on labor use are notoriously laborous and costly to collect. They were not
collected by the Division of Agricultural Statistics (DSA) of Rwanda’s Ministy of
Agricultlue and Livestock (MINAGRI), whose data we used above to compute returns to
land. Hence, no comparable figures on returns to labor can be presented. Experts on
Rwandan agriculture generally believe that coffee is the most laborous crop (Bart 1993;

68

Graaff 1986). Pure banana groves do not require much work if grown extensively, but
intensive production that often involves intercropping with annual food crops can be
laborous. Beans require more labor per hectare than cereals, and cereals more than root
crops. Cassava, especially, requires little labor (Holden 1993). Thus, it appears that cash
returns to labor are high for potatoes in the northwestern highlands and for extensively
cropped bananas in much of the rest of the country. Cofl'ee is not nearly as attractive (at
the low prices that prevailed in late 19803 and early 19905). The makings of the main

food crops—sweet potato, beans, sorghum, maize, and cassava—in terms of their returns
to labor apparently vary greatly depending on local agroclimatic conditions and prices.

Crop-specific returns on land and labor vary considerably depending on
agroclimatic conditions. The higher the altitude, the cooler the climate and the lower the
yields of cassava, which grows best on the warm lowlands found in Rwanda only on the
soutwestern and southeastern edges. Bananas yield well at the mid-altitudes (around 1500
meters) that cover much of the country. At higher altitudes they grow more slowly and
have a lower sugar content, which reduces their value for beer-brewing. Few. bananas are
grown above 2000 meters altitude. The range of sweet potatoes, beans, and maize reaches
somewhat higher, but at the temperate highlands (around 2500 meters), sweet potatoes
are typically outyielded by potatoes, beans by peas, and maize by wheat. Of export crops,
cofl’ee grows better at lower and tea at higher altitudes.

Rainfall increases in Rwanda from east to west and the difference matters mostly
for cereals. Maize is the dominant cereal in the well-watered western provinces, whereas
sorghum thrives in the dryer eastern savannas. Although bananas in general are
vulnerable to droughts, only a very small area in eastern Rwanda is too dry for bananas.
Most of the East is very suitable for bananas. Cassava, sweet potato, and tea are known to
grow relatively well on degraded lands, whereas cereals, beans, and bananas suffer

69

considerably from degradation (Barampama 1993). Cooking bananas are more
demanding than beer bananas.

Producer prices per kilogram can be interpreted as proxies for transportability. At
the 1990‘ prices, coffee was by far the most transportable crop, followed by peas, beans,
banana beer, and cereals. Root crops and bananas have low values per weight, which
implies that their production for non-local use depends crucially on transportation costs.
The relative prices of beer bananas and banana beer shows how brewing increases
transportability. Ifbrewing isconsideredpartofbananaproduction,beerbananasarea
labor-intensive crop with high market returns per hectare. The fact that cooking bananas
arealmosttwiceasexpensiveasbeerbananas suggeststhatonsimilarlandcooking
bananas yield considerably less than beer bananas. Thus, our second column cannot be
interpreted to mean that farmers could earn higher income by switching from beer
bananas to cooking bananas.

The last column of Table 5-1 reports consumer prices relative to caloric content.
Of the products normally eaten, cassava provides cheapest calories. Beer bananas would
be even cheaper, but are considered famine food and are usually eaten only after serious
crop failures. Brewing increases price and destroys many calories, making banana beer a
very expensive source of food energy. Cooking bananas are relatively expensive food,
which suggests that in consumers’ view, they are a very different product from beer
bananas. Cereals provide as cheap calories as sweet potatoes despite much lower returns
to land. White potatoes are quite expensive food, although this does not hold for the

prime potato production areas.

Since Rwanda is a hilly country, with an average field slope of 16.7 degrees,
erosion is a major concern. Erosion depends on a number of factors, including the degree

of protective cover provided by the vegetation. Crop cover, in turn, depends on the crop

70

species and varieties grown, on the timing and density of planting, on the crop
management practices followed, etc. In general, perennials that are kept mulched
throughout the year provide best protection against erosion, whereas annuals that
establish slowly leaving much of the ground bare during the early rainy season, are most

erosive.

Table 5-1 also presents the crop cover indices (C-values) estimated for Rwandan
conditions by Lewis et al. using the Universal Soil Loss Equation (U SLE). These
estimates are specific to the prevailing agronomic practices and are based on empirical
measurements in Rwanda and Kenya. Coffee fields that are usually mulched carefully in
Rwanda are least vulnerable to erosion. Also tea, Rwanda’s second most important export
mop, provides good crop cover. Pure banana groves are typically covered by mulch
(banana leaves and trunks), which protects them well against erosion. However, farmers
are ofi'rcially urged to thin their banana groves to get more productive banana trees, which
means less mulch and more erosion. Moreover, when bananas are intercropped with
annuals, most of the crop residues are typically removed or turned under the soil, making
land much more vulnerable to erosion. Of the annual crops, beans and roots and tubers
typically provide more crop cover against erosion than maize and sorghum, the main
cereals.

In sum, outside the northwestern potato zone, market incentives are typically
tilted towards bananas, subsistence incentives towards sweet potatoes, beans and bananas,
followed by cereals. On degraded lands, cassava and sweet potatoes often produce most
food; on better lands, beans, cereals, and especially bananas produce more food or market
value. 0n steep slopes, concern for soil conservation might induce households to prefer
bananas, coffee, or woodlots over annual food crops.

71

5.2 Land Use Trends in Recent Past
Between 1944 and 1994, Rwanda’s population grew fi'om less than two million to

almost eight million. Changes in land use and cropping patterns have been the central
mechanism through which agriculture has been able to respond to population growth.

One hundred years ago most cultivation took place on the gently sloping hillsides
that were easiest to clear with fire and cultivate with hoes (Bart 1993). Over the past
century, cultivation has gradually expanded to the wet valleys, steep slopes, dry savannas,
and temperate highlands previously reserved for grazing or left under natural forests.
Frequency of cultivation has increased; fallowing has declined.

In the beginning of this century, sorghum, which was grown both for food and for
beer, was the dominant staple and other important crops included beer bananas, finger
millet (eleusine), and traditional legumes (Jones and Egli 1984, Martin 1987). New world
crops, including beans, sweet potatoes, and maize had been introduced but had not yet
challenged the reliance on traditional cereals and legumes.

A series of famines in the late 19203 provoked the Belgian authorities to promote
non-seasonal crops: cassava in low-altitude areas, sweet potatoes in medium-altitude
areas, and potatoes in high-altitude areas (V is 1975, Guichaoua 1989). Cultivation of
coffee was made compulsory in 1927 (Guichaoua 1989) and in the 19308 also the
cultivation of the non-seasonal famine crops became mandatory (Leurquin 1963). Sweet
potato gained ground rapidly, but was soon attacked by a disease, which greatly
contributed to the 1943-44 famine (V is 1975). Colonial administrators responded by
emphasizing more cassava, which was originally bitterly resented by farmers, and by
introducing more resistant sweet potato varieties (Leurquin 1960). Farmers also
responded to famines by planting more beer bananas, which can be eaten during times of
distress (Bart 1993; Leurquin 1963; Close 1955).

72

By 1967, cereals were still believed to take nearly one-third of cultivated land
(Bart 1993). Roots and tubers had increased to 14 percents and provided more calories
than cereals. Leguminous plants with 37 percent of cultivated land provided enough
proteins. One-fifth of cultivated lands was believed to be under bananas.

In the 19603 and 1970s, cultivated area grew approximately as fast as population.
The area under and the production of roots and tubers and cofl‘ee may have increased by
more than three-fold, while cereals and beans grew more slowly and lost ground relative
to roots and tubers (Cambrezy 1984; Bart 1993; World Bank 1991; CNA 1991b). Banana
production expanded fast in the 19603, as most of the colonial restrictions on bananas
were discontinued after independence (Delepierre 1970). Thereafter, banana groves
continued to expand slightly faster than cultivated area, but since many of the new groves
were planted on marginal lands, production increased more slowly, failing to match
increases in population (Bart 1993; CNA 1991b).

The above numbers were based on various expert opinions. Agricultural surveys
with nationally representative sampling started in 1984, and only in 1989 a systematic
approach was introduced to deal with intercropping, which is a pervasive feature of
Rwandan agriculture (DSA 1991). This suggests caution when discussing the long-term

trends in land use.

According to the national agricultural survey for agricultural year 1991, cereals
have lost their dominance and only covered 14 percent of cultivated land. Their role in
food consumption, whether measured in terms of food energy or proteins, had dropped to
below 20 percent. Finger millet had almost disappeared and sorghum had lost ground
nationally, despite its suitability for and popularity in the dry eastern lowlands, where
cultivation had expanded most (Bart 1993). In beer brewing, sorghum had been
increasingly replaced by beer bananas (Jones and Egli 1984; Bart 1993). Rice had been

73

expensively promoted for the drained marshes (Rushemeza 1991), but nationally it was

still a minor crop.

Roots and tubers that cover 29 percent of the cultivated area and bananas with 24
percent provide almost two-thirds of the food energy, while legumes (primarily beans)
with 24 percent provide halfof the proteins. Sweet potato is by far the most important
root crop and is grown almost throughout Rwanda. Cassava is grown mostly in areas that
are too dry or degraded for sweet potato. White potatoes are a minor, low-yielding
subsistence crop in most of Rwanda, but a major, high-yielding cash crop in the cool
northwestern highlands. Their prominence is partly due to climatic conditions, and partly
dependent on effective research, extension, and input supply. Bananas are mostly beer
bananas, but especially in eastern Rwanda and around cities also cooking bananas are a
major crop. Legumes are mostly beans, and increasingly climbing beans rather than bush
beans (Sperling et al. 1994). Except for some high-altitude areas, peas have almost
disappeared. Soybeans are starting to gain ground, primarily on degraded lands and in
areas where beans have suffered greatly from diseases (Olson 1992; World Bank 1991).

Compared to the first agricultural survey in 1984, the 1990 survey suggested that
theareaunderbananashadincreasedby27percentandthatundersweetpotatoesby 33
percent. Also coffee, cassava, potato and soybean acreages had increased, while beans,
peas, and cereals had declined.

Insum,threemaintrendsinlanduseoverthiscenturyianandacanbe
identified. First, annual and perennial crops have expanded at the expense of pasture,
fallow, and forest. Second, the shares of cultivated fields under roots and tubers, bananas,
and coffee have increased at the expense of cereals, and, more recently, legumes. Third,
traditional cereals (finger millet and sorghum) have been replaced by maize and
traditional legumes (cowpeas and peas) by bush beans and these in turn by climbing

74

beans and soybeans. Despite quite heavy-handed promotion of some crops at the expense
of others at times, the trends are probably mostly voluntary choices by farmers. Sweet
potatoes and cassava continued to expand after colonial promotion ended, and ofiicial
discouragement did not prevent the expansion of bananas, although it may have slowed it
somewhat during the late colonial period.

Above we have simplified by averaging over the two seasons. Most land uses in
Rwanda are quite stable, with little difference between the seasons. The main exception is
sorghum, which in much of Rwanda is grown exclusively during the second (longer)
season. The ofi'setting change concerns mostly beans and maize, which nationally cover
about twice as much land during the first season than during the second season. Seasonal
differences are least important in the North-West, where the rainfall is higher and more
evenly distributed than elsewhere in Rwanda (N duwayezu 1990).

5.3 Patterns of Land Use by Farm Size

The main difl‘erence in land use between small and large farmers is the much
larger share of small farms that is under the crops. Using data from the national
agricultural survey of 1991 and averaging over the two seasons, Table 5-2 reports that
the quartile of households with least cultivable land relative to the size of the family
(hereafter “small farmers”) cultivated 90 percent of their cultivable land, while the
quartile with most land per adult equivalent (hereafter “large farmers") had crops on only
69 percent of their agricultural land. Large farmers had three times as large a share under
fallow than small farmers, and also had much more pasture.

75

Table 5-2: Land Use by Farm Size

0 OP CULTIVABLE LAND FARM SIZE QWTILE (Ares/AB) All Farms
<9 9-14 14-24 >24
Bananas 19 20 18 19 19
Roots and Tubers 28 26 23 18 24
Legumes 22 19 19 16 19
Cereals 13 13 12 11 12
Coffee . 6 5 5 4 5
Other 2 2 3 2 2
TOTAL CULTIVATED 90 85 80 69 81
Fallow 10 14 18 25 17
Pasture 1 1 2 5 2
Woodlot 8 10 13 8 10

Source: DSA/MINAGRI, Agricultural Survey 1991; 1095 households

 

To a large extent, the disappearance of fallow and pasture reflects the Boserupian
intensification process in which farmers respond to population pressure by increasing the
frequency of cultivation. Notably, woodlots do not show the same pattern. Unlike pasture
and fallow, which in Rwanda are often “luxuries” of large farmers, the woodlots that
remain on farms are largely located on steep slopes and on other marginal lands.
Regionally disaggregated tabulations not reproduced here indicate that the relatively rich
and large farmers in the East have almost no woodlots, whereas the impoverished farmers
on the Congo-Nile Divide allocate much land for trees. Our definition of cultivable land
used in grouping farmers to farm size quartiles excludes woodlots but includes pasture
and fallow. Even the inclusion of pasture and fallow is debatable, since while most of
these lands undoubtedly can be and periodically are cultivated, some farmers resort to

76

fallowing not because they have much land but because their fields are so poor that they
yield little without fallowing.“

Regarding the crop mix, the table is less clear, showing mainly that small farmers
have higher shares of most crops. To better identify patterns in the crop mix, we focus on
the allocation of cultivated land in Table 5-3. As the historical developments would
suggest small farmers have higher shares of white potatoes and root crops. But the
historic expansion of bananas at the expense of other crops is nowhere to be seen, and the
decline of bean cultivation in the 19803 under strong demographic pressure clearly does
not mean that small farmers would shun beans. Overall, the crop choices of the small
farmers are not strikingly different from those of the large farmers. Some observers even

refer to the “considerable stability” of the crop mix (MINAGRI 1992).

 

1‘ In regional comparisons, cultivated-rather than cultivable land would probably be a
better basis for assessing land scarcrty, since much of the regional variation in fallowing
and pasture a pears to be associated with land quality. Withrn each region, however,
larger and ric er households generally have much more fallow and pasture than their
less fortunate neighbors. Thrs sg‘ggest that in household-level analysis like ours,
cultivable land should be defin to mclude pasture and fallow.

77

Table 5-3: Crop Shares by Farm Size

8 OP CULTIVATBD LAND EARN SIZE QUARTILS (Ares/AS) All Farms
<9 9-14 l4-24 >24
Banana 21 23 22 27 23
Sweet Potato 15 14 13 11 13
Cassava 8 9 10 10 9
White Potato 6 4 3 2 4
Total Roots and Tubers 32 31 29 26 29
Beans 18 16 17 16 17
Peas 3 3 3 4 3
Total Legumes 24 22 23 22 23
Sorghum 5 8 9 9 8
Maize 9 7 5 6 6
Total Cereals 15 16 16 16 16
Coffee 6 6 6 6 6
Other Crops 2 2 3 3 3

Source: DSA/MINAGRI, Agricultural Survey 1991; 1095 households

 

The differences between the longitudinal and cross-sectional results suggest that
the associations between land scarcity and land use may be lost because small and large
farmers difi‘er also by agroclimatic or other factors. A key agroclimatic distinction in
Rwanda is between highlands, where bananas do not grow well, and the rest of the
country. Part of the non-banana highland area enjoys fertile volcanic soils and specializes
in white potatoes. Outside the potato zone, the highlands are mostly marginal and
degraded. In our sample, the distinctions between the zones are clear. Ten of the 78
enumerators worked outside the banana zone, four of them in the potato zone. Within the
banana zone, there were some households with almost no bananas, but in all the clusters
of four households bananas covered at least 10 percent of cultivable land. In the potato
zone, bananas were almost non-existent. In the marginal zone, both bananas and white

potatoes were rare.

78

Table 5-4: Comparison of Cropping Zones

Banana Zone Potato Zone Marginal Zone
Altitude (ml 1710 2214 2184
Rainfall (mm) 1083 1161 1574
Farm Size (ares/AB) 19 11 22
b Cultivated 82 90 60
8 under Banana 21 0 2
3 under Sweet Potato 11 1 13
3 under lhite Potato 1 38 3
Households in sample N-960 N-57 N-78

Source:DSA/MINASRI, Agricultural Survey 1991; 1095 households

 

Table 5-4 shows that the cropping zones are closely associated with farm size:
farms in the potato zone are tiny, whereas those in the marginal zone are relatively large.
This suggests that despite the strong national cross-sectional correlation between land
scarcity and potato production, potatoes are a significant response to population pressure
only for a small minority of Rwandan farmers. More generally, the table hints that the
national patterns on land use may give little guidance on how Rwandan farmers are

intensifying. Moreover, intensification strategies are likely to differ by zone.

Hereafter, we focus on the banana zone, which covers some 85 percent of our
sample and presumably roughly the same share of Rwanda. Table 5-5 shows that the
strong association between farm size and white potatoes seen above indeed does not hold
for the banana zone. Otherwise, the results do not change much. Land scarcity appears to
be associated primarily with beans and sweet potatoes, but not with bananas, which cover
only slightly mallet shares of cultivable land on smaller farms and, relative to the

cultivated land area, are no more important on small farms than on larger farms.

To the extent that intercropping is a way to increase returns to land, it can be

hypothesized to be one intensification strategy. Table 5-6 shows that intercropping indeed

79

Table 5-5: Land Use by Farm Size in Banana Zone (% of cultivable land)

EARN SIZE QUARTILE (Ares/AB) All Farms
<9 9-14 14-24 >24
Banana 22 23 20 20 21
Boots and Tubers 27 26 22 18 23
Sweet Potato 14 13 10 7 11
Cassava 8 9 9 7 8
white Potato 1 1 1 1 1
Legumes 23 19 19 15 19
Beans 19 14 14 12 15
Peas 2 2 2 2 2
Cereals 11 11 12 10 11
Maize 5 4 3 3 4
Sorghum 5 7 7 7 6
Coffee 7 5 5 5 6
Other 2 2 3 2 2
TOTAL cumrva'ran ' 91 as 91 71 32
Fallow 8 13 16 24 16
Pasture 0 1 3 5 2

Source: DSA/MINAGRI, Agricultural Survey 1991; 960 households

 

is associated with land scarcity, with smaller farmers having much higher shares of their
land under various crop associations. Half of the difference arises because small farmers
have less fallow, woodlots and pasture, but even if we look at only the cultivated fields,

small farmers clearly practice more intercropping.

/' Moreover, small farmers intercrop differently. Their associations include more
ll crops, and have more often bananas, beans and sweet potatoes among the associated
"M. crops. Thus, it appears that land scarcity is pushing households towards multistorey

\ banana gardens where food crops are grown under banana trees (Conway 1983).

5.4 Multivariate Analysis of Land Use
The cropping patterns discussed above show how the land uses of land-scarce

farmers difl'er from those of their larger neighbors, but they do not go very far in

80

Table 5-6: Intercropping by Farm Size in Banana Zone

am 3123 ounarrna (Ares/AB) All farms
<9 9—14 14—24 >24

3 of cultivable
intercropped 63 54 48 41 51

t of cultivated
intercropped 68 62 58 58 62
with bananas 45 38 33 30 36
with beans 40 32 29 27 32
with sweet potatoes 23 22 17 14 19
with cassava 19 18 16 15 17
with sorghum. 9 10 11 11 10
with maize 24 19 18 18 20
with coffee 2 2 1 2 2
Crops/association 2.4 2.3 2.2 2.1 2.2

Source: DSA/MINAGRI, Agricultural Survey 1991; 957 households

 

explaining whether the differences arise because land scarcity pushes small farmers to
behave differently. We next study the association between land scarcity and land use by
controlling for a number of other factors. Since the patterns were found to be highly
different on the highlands, it is likely that also the relationships between land use and
other factors vary by cropping zone. Access to markets, for instance, is likely to provide
very different incentives for land use depending on which are the most lucrative cash
crops in the area. We therefore restrict our study to Rwanda’s banana zone, which covers
roughly 85 percent of the rural households.

5.4.1 General Model
Semi-subsistence smallholder agriculture like that in Rwanda is characterized by

very high tansportation and transaction costs (Binswanger and Rosenzweig 1986). As
explained by Coase (1937) and elaborated by Williamson (1985), North (1990), and
Fafchamps, de Janvry, and Sadoulet (1995), the presence of such costs makes resom'ce
allocation dependent on who owns the assets. Land, labor, and credit markets cannot be

81

assumed to equate factor costs across farm households. Consequently, resource allocation
decisions, including those on land use, cannot be modelled merely as a firnction of
relative prices and the physical input-output relations. Instead, the opportunity costs
(shadow prices) of using land tend to be higher and those of using labor tend to be lower
on the farms that have less land per person. Instead, models should also include relative

factor endowments as well as household subsistence needs and consumption preferences.

In practice, many of the theoretical determinants are not observable and various
proxies need to be used. Given our focus on the impacts of increasing land scarcity

associated with demographic pressure we sketch a model with four sets of variables:

Land use = f (market incentives, physical conditions, person/land ratio, other
household characteristics).

Market incentives refer to prices or to other measures of how lucrative it is to
produce certain crops for sale while relying on markets for other crops. Physical
conditions are proxies for land productivity, that is, for the yields farmers can expect on
their fields from each crop. Physical conditions are likely to have strong correlations both
with land use and with other determinants, which makes them crucial control variables
for the model. Person/land ratio measures both subsistence needs and the relative
endowments of land and labor. Other household characteristics are proxies for

preferences and non-land endowments.

5.4.2 Explanatory Variables and Hypotheses
Land scarcity. We expect that land scarcity (high person-to-land ratio in terms of

adult equivalents per cultivable hectare) is associated with more land being allocated to

82

uses that give high returns to land at the expense of uses that give low returns to land.
Based on the characteristics of the major crops in Rwanda (Table 5-1), we expect land
scarcity to be associated with high shares for bananas and sweet potato, which nationally
have superior productivity both in terms of market value and in terms of food energy. We
expect low shares for sorghum, peas, taro, and cassava, which are nationally quite low-
yielding by every measure.

No unambiguous hypothesis can be made for beans, since while their land
productivity is considerably below that of sweet potato and bananas in terms of market
value and food energy, they are superior to all other crops in terms of proteins. In a
commercialized setting, we could expect the market value to matter most and small farms
to specialize in bananas and sweet potato, and purchase rather than grow beans. But in
Rwanda’s still quite subsistence-oriented rural economy where diets are very deficient in
proteins, subsistence production of proteins clearly must be very important. Also the
value of the nitrogen fixed by leguminous plants such as beans may increase as land

becomes scarcer and fallowing no longer can be relied on to restore fertility.

Also the expectation for maize is ambiguous, since while maize does not provide
very high returns to land in Rwanda, it produces more food per hectare than sorghum,
especially in the well-watered western regions. To the extent that cereals are subsistence
foods and substitutes for each other, maize can be hypothesized to gain ground from

sorghum as land becomes more scarce.

Overall, the share of cultivated land can be expected to increase with higher land
scarcity and that of pasture and fallow to decline. Little can be said about woodlots, since
no data is available on returns to trees. While cultivation has historically expanded at the

expense of forests, some results suggest that in many land-scarce areas trees could be
highly remunerative cash crops (Cook and Grut 1990; Current, Lutz, and Scherr 1995).

83

 

Market incentives. In theory, market incentives could be expressed in terms of
(the expected levels and variances of) prices faced by farm households at the farm gate,
but in practice we observe only market prices. Since Rwanda’s main agricultural products
are bulky (Table 5-1) and local infrastructure is often poor, the unobserved farmgate
shadow prices are likely to be quite different fiorn the observed prices”. Moreover, local
market prices reflect both local demand and local supply. Instead of indicating good
economic incentives (strong demand), high prices may be a sign that the crop in question
just does not grow well in the area (weak supply). When agroclimatic and other factors
affecting supply are imperfectly controlled for, regression analyses often seem to suggest
that farmers react to high prices by producing less. Since data on local prices are also

quite noisy, we do not attempt to measure market incentives by prices.

Instead, we use distance to paved road as a proxy for the incentives to produce for
the market and to rely on it for food. The point, of course, is not that only roads matter for
commercialization. While proximity to road may be the key determinant of whether a
given farmer has access to wider markets, the overall commercialization of Rwandan
agriculture may depend more on the growth of urban economy and on trade policies than
on road construction. Thus, what our analysis finds about the “impacts” of good access
may also suggest what are some of the rural consequences of growing urban demand for
food and other agricultural products.

We expect that good market access encourages'households to specialize in cash
crops that yield well in terms of market value. We expect this to happen at the expense of
less remunerative land uses, including food crops with low monetary returns. We expect

 

2’ In Kagera, on Tanzanian side of Rwanda’s eastern border, large quantities of cooking
bananas are thrown away or fed to livestock in good harvest years (Kajumulo-Tibaijuka
1984). This suggests that transportation costs exceeded market prices and the perceived
farmgate producer prices were close to zero. DSA data suggest that some households in
eastern Rwanda may do the same.

84

also changes towards bulky cash crops at the expense of crops with lower yields but
higher value per unit of weight.

In Rwanda’s banana zone, this implies above all that bananas should benefit from
good access to paved roads. Whether sold for food or as beer, their tansportability as
measured by value-to-weight ratio is far below that of coffee, the other main cash crop in
Rwanda. The land uses that can be hypothesized to lose ground are pasture, fallow,
cereals, cassava, taro, peas, and coffee. For beans the expectation is again ambiguous,
since, as discussed above, their role in fertility management can offset the consequences
of modest yields in terms of market value. While the latter is an argument for expecting
that households close to roads might take advantage of markets and buy rather than
produce beans, the latter suggests that market access which increases the value of the
increased soil fertility could promote bean production.“ In other words, in a farming
systems perspective beans may not be a low-yielding crop that farmers are tempted to
giveupiftheycanbuybeansinsteadu

Physical conditions. Rwanda is a mountainous country, where agroclimatic
conditions vary primarily with altitude. Most of the county is between one and two
kilometers above sea level, but roughly a quarter of agricultural land is located above
2000 meters and some areas in eastern and southwestern parts of the county are below
1000 meters. Highlands are cool and well-watered, lowlands much dryer, especially on
the eastern side of the Congo-Nile Divide. Soils also vary considerably, but no estimates
of the soil types or of degradation on the surveyed farms were available in the data set.
Estimates obtained from soil maps were not helpful, presumably because they were much
too crude. Instead, we have computed an index of land quality, based on the estimated

 

2‘ In many counties a reasonable assumption would be that good road access implies
improved access to fertilizer and reduced incentive to grow beans. As discussed below,
fertilizer consumption in Rwanda in 1990 was so minimal that variation in fertilizer
access hardly played any significant role in crop choice. Neither farmers close to roads
nor those in more remote areas used notable quantities of fertrhzer for beans.

85

yields of major crops (weighted by national prices) in each sector. Since not all
differences in physical conditions are captured by rainfall, altitude, and land quality, we
also have included dummy variables for agroclimatic zones. To the extent that the
impacts of rainfall, altitude, and land quality are non-linear, such impacts also can get
captured by zone dummies.

Multicollinearity among the indicators of physical conditions creates problems for
the estimation and interpretation of regression coeficients. Fortunately, these coefficient
estimates as such are of minor interest. Since our main focus is on demographic change
and on factors that are changing or could be changed by policy, the role of rainfall,
altitude, land quality, and zone dummies is primarily to contol for physical conditions.
This role is not compromised by high correlations among the indicators of physical

conditions.

The general hypothesis, of course, is that each crop claims larger shares in those
conditions that are more suitable for it. Of Rwanda’s major crops, bananas and cassava
are known to thrive at lower altitudes and sorghum does better in dry conditions than
maize. White potatoes are known to grow best at the very high altitudes, whereas sweet
potatoes and beans grow better at lower altitudes. Cassava and sweet potato are known to
sufl‘er relatively modestly from degradation, whereas cereals are considered quite
demanding.

Household characteristics.Two key characteristics that can be hypothesized to be
associated with land use are the age of the household head and ofi-farm incomes. As
population growth has greatly increased the number of agricultural households during the
past decades, young people have often settled either on previously uncultivated lands or
on the outer fields of their parents’ farms. We expect that households with older heads
devote higher shares of their land to perennial crops such as bananas and cofi‘ee.

86

Ofi-farrn incomes are a proxy for liquidity and can be hypothesized to be
associated with lucrative crops that require large cash outlays or long payback periods. In
Rwanda’s volcanic highlands, white potatoes might be such a crop, but in the banana
zone studied here none of the main crops requires many purchased inputs other than
labor. Since both cofl‘ee and (intercropped) bananas are labor-intensive and take years to
reach their full production, we hypothesize that these are associated with high off-farm

incomes. .

Education is highly correlated with off-farm incomes, either because high
incomes are needed to pay for education or because education raises earnings. Only off-

farm incomes were used in our specification.

5.4.3 Regression Results on Land Use
Two sets of Ordinary Least Squares regressions on land use are estimated using

the variables described in the previous section. In the first set of OLS regressions, the
dependent variables are percentages of cultivable land under each land use and in the
second set, they are percentages of cultivated land under each crop. The first set focusses
on how farmers respond to demographic change by expanding cultivation at the expense
of fallow and pasture, while the second set looks specifically at variation within the crop
mix. Table 5-7 reports the results for the first set, and Table 5-8 for the second one.

87

Table 5- 7: Land Use Model 1: Shares of Cultivable Land under Each Use

Banana Sw.Pot Cassava Beans Sorghum Maize Coffee Fallow Pasture

.---------------------------------------------------------------------------

Man/land,farm .18** .02 -.06 .17** -.05 .11** -.11** -.13** -.13**
Han/land,vill -.04 24** - 02 12** .00 27** .10* -.29‘* .04
Dist. to road -.17** 02 06 02 .09* 08* .02 .04 .07
Altitude -.11* 33 - 49** 11* .21** 12** -.26** .19** -.07
Rainfall -.06 04 20** - 37** -.14** 04 .06 -.01 -.13**
Land quality -.01 -.22** - 04 - 07* .15** 06 .00 .09* -.13**
Age Of head .12** -.04 - 13** 00 .07* 06 -.07* -.02 .03
Off-farm inc. .04 -.02 - 07* 04 -.04 04 -.08** -.02 .02
Bast .14** - 02 - 24** 14** .16** 24** -.12** -.10* -.04
Congo-Nile .03 -.06 - 13* 00 .07 15** -.09* -.04 02
volcanic -.03 -.14** - 01 26** .01 12** -.03 -.10** 00
Lake Kivu -.05 -.06 - 12** 26** -.12** 08* .28** -.09* - 04
ADJ R2 11 28 22 28 14 17 23 18 04

Source: DSA/MINAGRI, Agricultural Survey 1991; 960 households
* - significant at 5% or better, ** - significant at It or better

 

Population pressure. Table 5-7 confirms that small farmers indeed cultivate a
higher share of their cultivable land by allocating less land for fallow and pasture. In
other words, the pattern observed in the cross-tabulations above, appears to hold when
rainfall, altitude and the other factors in the regression specification are contolled for.

Both Table 5-7 and Table 5-8 show that land-scarce households allocate more
land to bananas, beans and maize than others on the average. The relationship between
sweet potatoes and population pressure is more complex. The results show that while
more sweet potatoes are grown in secteurs with higher population pressure, no significant
association was found within secteurs between sweet potato and land scarcity. If the
sectoral association arose only because some unmeasured factors (such as soil type) are
linked to both population pressure and land use, we could not conclude that population
pressure is pushing land-scarce farmers to grow more sweet potatoes. Yet, the post-World
War II expansion of sweet potatoes during the period when Rwanda’s population grew

88

Table 5—8: Land Use Model 11: Shares of Cultivated Land under Each Crop

Banana Sw.Pot. Cassava Wh.Pot. Beans Sorhum Maize Coffee

Han/land,farm .12** -.02 -.09“ -.05 .13** -.07* .07* -.12**
Han/land,vill - 15** .20** - 06 18** .08 -.05 23** .05
Dist. to road - 17" .04 06* 05 .00 .09* 12** .02
Altitude - 07 .37** - 49*” 04 .l7** .2S** l7** - .27‘”
Rainfall -.08 .02 19** 00 -.43** -.17** 01 .05
Land quality .03 -.22Mr - 05 - 03 -.O7 .18** 07 -.02
Age of head .12** -.05 - 12** 01 .00 .06 06 -.07*
Off-farm inc. .04 -.01 - 08 00 .04 -.04 06 -.07*
Bast .11* -.07 - 28** 01 14** .16** 24** -.14**
Congo-Nile .03 -.06 - 14** 03 00 .09* 16** -.10**
volcanic -.06 -.1S** - 01 14" 24** -.03 10** -.03
Lake Kivu -.07 - 09** - 14** - 04 25** - 10* 07 .25**
ADJ R2 12 32 .22 04 2‘7 18 15 19

Source: DSAYMINAGRI, Agricultural Survey 1991; 960 households
* - significant at St or better, ** . significant at 1* or better

 

four-fold suggests a link between the two trends. One possible explanation is that
population pressure pushes small farmers to consume sweet potatoes (which provide
cheap calories) but not to produce them (because other crops provide higher returns to
land). This would explain why the sectoral association exists while the farm-level
association does not and this would also be compatible with the trend towards sweet
potatoes over time. We will examine this question below using data on sales and
purchases for the same set of households.

Although small farmers clearly allocate a higher share of their land to bananas
than their neighbors with more land, sectoral land scarcity is associated with less bananas
in the crop mix. Just as above in the case of sweet potatoes, there are two possible
explanations for the paradox. First, there may be omitted or poorly measured factors that
explain why densely inhabited areas have less bananas than their other characteristics

would suggest. Potential candidates include soil type, prevalence of banana diseases, or

89

market demand for bananas. Second, land scarcity may push farmers to produce more
bananas (which provide high returns to land) but consume less (because bananas and
banana beer provide relatively expensive calories), in which case the small size of the
neighbors would be an argument to grow food rather than bananas. Again, this hypothesis
is something we will study more thoroughly in the next chapter.

Table 5-7 shows that none of the main food crops significantly loses ground as
pressure on land increases. In other words, the expansion of bananas, beans, and maize
occurs primarily at the expense of pasture and fallow. However, as Table 5-8 shows, the
relative importance of sorghum and cassava declines significantly in the crop mix. Cofi'ee
is favored on the larger farms of the more densely populated areas. One possible
explanation is that the official persuasion to grow coffee has been strongest in these areas,
but either has not been enforced equally strictly on the smaller farms, or has been resisted
more actively by the small farmers desperate to grow more food instead.

The results are broadly compatible with the hypotheses stated above. Sorghum
and cassava are less labor intensive and lower-yielding than the other major crops and all
crops require more labor than fallow and pasture.

Market access. Banana production is significantly associated with short distance
to paved roads. In contrast, cereals and cassava are more important in more remote areas
than in areas close to paved roads. This suggests that improvements in roads and
presumably also in other forms of marketing infrastructure would encourage farmers to
substitute bananas for cereals and cassava. These results are compau'ble with the
hypothesis that good market access encourages specialization in bulky but otherwise

profitable cash crops and reduces reliance on low-yielding subsistence crops.

90

 

Land quality and agroclimatic factors. As discussed above, our proxies for the
physical environment, especially for soil quality, are quite imperfect and highly collinear.
Consequently, the coefiicients for them are difficult to interpret and often not significant.

Nevertheless, the hypothesis that low land quality would be associated with
higher shares of land devoted to less demanding crops got strong support in the case of
sweet potato. The hypothesis also got some indirect support in the case of cassava, since
the dummy variables show that cassava was associated with location on the central
plateau, which is known to be quite degraded. Since land productivity is affected by labor
use, inputs, skills, and other factors in addition to land quality, the dummies for zones

may capture some land quality and degradation impacts.

The other side of the hypothesis is that demanding crops should be favored on
highly productive fields. This was supported for sorghum but not for bananas, maize, or
beans. In the case of bananas, the explanation may be historical, since both the high share
of bananas and low yields may result from a long history of fiequent cultivation in the
area. The explanation for beans may be that while they grow poorly on poor fields, their
ability to fix nitrogen is most needed there.

The associations between land use and agroclimatic conditions are generally
strong, although the high correlation between altitude and rainfall create some surprises.
For instance, the association of banana production with the warm and relatively dry
conditions outside the high-altitude areas was captured by rainfall rather than altitude in
our model, whereas the opposite was true for the association of sweet potato and higher-

altitude areas.

Household characteristics. The age of the household head is positively associated
with banana production. This is what we expected for a perennial that takes years to reach

its full productive potential. The association with the other main perennial, cofi‘ee, is to

91

the opposite direction. This shows that coffee trees are not a similar investment of
embodied labor as banana trees and suggests that other considerations, including
government policies to promote coffee, have been more important for the decision to
grow cofi‘ee. Of the annual crops, cassava is grown more frequently by younger
households. This was not hypothesized, and might be related to unmeasured difierences
in land quality. Many young households occupy poor lands that until recently were
considered marginal for cultivation and reserved for pasture. Cassava may be the only
food crop that gives decent yields on such fields.

The hypothesis that perennials are investments that households with high ofl-farm
cash incomes are more capable of doing was not supported. Apparently, the fact that
‘ banana groves can be intercropped throughout the establishment period makes the
investment so small that cash is a minor constraint. Coffee, in turn, is related to low off-
farrn incomes. This may be related to farmer interest (or official persuasion) to plant
coffee in places where few alternative sources of cash have been available. In principle,
coffee is also a source of liquidity that could be hypothesized to reduce farmers’ interest
in off-farm employment, but given how poor cash source cofi‘ee was around 1990
compared to most other land uses, this explanation does not appear to be very credible.

5.5 Environmental Impacts of Intensificstion
As shown above, increasing land-scarcity forces Rwandan farmers to cultivate a

larger share of their land at the expense of forest, pasture, and fallow. Since non-cropping
uses of land provide more vegetative cover against erosion (have lower C-values) than
most food crops (Table 5-1), the expansion of cultivation increases erosion in Rwanda.
On the other hand, demographic pressure also pushes farmers to grow more perennials,
which protect soils better than annual food crops and can even exceed the level of crop
cover provided by degraded pastures and woodlots planted with non-protective trees such

92

as eucalyptus. Moreover, farmers also respond to increasing land scarcity by relying more

on dense crop associations, which may increase vegetative cover on cultivated fields.

Accurate assessment of the net result of all these offsetting impacts would require
detailed field experiments on soil losses and statistical analysis on their determinants,
with land scarcity as one of the explanatory factors. Since data on soil losses tend to be
too costly to collect in regular farm surveys, researchers are left with the option of
combining farm survey data on land use with empirical information on the average
degree of protective crop cover provided by each land use. Ifthere exists a relatively
small number of relatively uniform land use types, this approach should provide

reasonably accurate approximations at low cost.

Unfortunately, this assumption is questionable in Rwanda. Although a handful of
crops dominate Rwanda’s agriculture, more than one-half of cultivated fields are covered
by crop associations, with numerous crop combinations and infinite variation with respect
to the relative importance of each crop in the association. Few empirically-based crop
cover estimates are available for crop associations, and there is little empirical basis for
computing estimates for them on the basis of crop cover estimates of the constituent

crops and on the densities of each crop.

Despite these caveats, we computed two sets of field-level crop cover estimates in
Kangasniemi and Reardon (1995). Our unadjusted C-value was an average of crop-
specific C-values weighted by the estimated land shares of each crop in a given
association. There were three reasons to think that this typically represents the lower
bound for the true C-value. First, DSA’s land use data show that planting densities on
intercropped fields are higher than on purely cropped fields, which suggest that there is
more vegetation, and thus more crop cover against erosion. Second, the few empirical

measurements available indicate that the average C-value of the crops grown in the

93

associations overstates the C-value (understates protective crop cover) of the association
(Lewis 1986). Third, in the literature, reduced erosion is frequently mentioned as one of
the reasons why farmers choose to intercrop (Fusscl and Serafini 1985; Koponen 1988;
Ruthenberg 1976; Lewis and Berry 1988; Richards 1983, 1985).

Our second set of estimates therefore was computed assuming that the higher the
planting densities on intercropped fields, the better the crop cover and the lower the C-
value that measures the exposure to erosion. The exact form of the adjustment was
arbitrary: we divided the unadjusted C-value by the sum of the densities of the crops

grown in the association.

With these assumptions, our data analysis gave one clear conclusion. At high
altitudes (above 1900 meters), small farms were significantly more exposed to erosion
than large ones. Elsewhere, results depended on how the measure of vegetative cover was
adjusted to account for high cropping densities. Without any adjustment, small farms
appeared to have less vegetative cover; with our adjustment, there appeared to be no

association between farm size and exposm'e to erosion.

The key difference between high and low altitudes was the dominance of bananas
in the latter areas. Essentially, the result shows that if intercropped banana groves provide
as good protection against soil losses as our adjustment implies, the expansion of banana
intercropping offsets the increases in erosion that otherwise would arise from the land use
changes brought about by agricultural intensification.

There are two reasons to think that this result is too optimistic, even ifthe
expansion of bananas would continue. First, the idea of comparing planting densities in
crop associations with those in pure stands, for which we have C-value estimates, fails for
bananas. Traditional pure banana groves are typically dense, at least if they are old. The
oficially recommended density, which is used as the benchmark pure density in the data

94

collection, is much lower than the traditional density (CNA 1990). While thinning
supposedly increases banana production, at least in the short term, it reduces the
production of crop residues and increases erosion. In the collection of land use data, the
benchmark density for bananas refers to the thinned groves, but the only available C-
value estimates refer to “typical” pure groves which presumably are quite dense.

To illustrate, consider a bean/banana field with unstandardized densities of 80
percent for beans (C=19°/o) and 80 percent for bananas (C=4%). Our unadjusted C-value
estimate for this field was simply the average (C=l 1.5%), while the adjusted C-value was
much lower (C=l l.5/1.6=7%). However, notwithstanding the recorded high density, this
field probably has less than half of the number of bananas trees of the typical pure banana

grove.

Second, the key explanation for minimal erosion (low C-values) in the taditional
banana groves is mulch: typically the soil is fully covered by crop residues. The mulch
protects the soil from the direct erosive impact of raindrops and, more importantly, slows
down the movement of water, thereby increasing infiltration and taming the downhill
flow into a non-erosive trickling that does not take the topsoil with it. To grow annual
food crops with bananas, farmers typically remove the crop residues from much of the
soil. While densities clearly influence crop cover by affecting the amount of vegetative
material available for mulching, this impact is probably dwarfed by variations in how the
mulch is used. Thus, the 80/80 bean/banana grove discussed above could easily provide
crop cover that is closer to that of a bean field than that of a dense, fully mulched banana
grove. Where-this is the case, both the adjusted and the unadjusted C-value estimates

could exaggerate the degree of protective cover and understate the exposure to erosion.

The dramatic increase in erosion is not inevitable, though. Preliminary research
results from Burundi suggest that if crop residues are put into small contour lines 2-3

95

meters apart they provide better protection in thinned banana groves against soil erosion
than the traditional practice of simply leaving crop residues cover the soil more or less
haphazardly (Rishirumuhirwa 1992, 1993a and 1993b). Ifconfirmed by fmther studies,
this suggests that intercropped bananas can be a highly sustainable form of agricultural
intensification on the steep hillsides of the Great Lakes Region.

In sum, except for the high altitudes, where population growth is increasing
erosion, it is not clear whether demographic pressure is making land uses more erosive in
Rwanda. The expansion of bananas is oflsetfing some of the damage done by the
expansion of annual crops, but the net impact depends on the details of how farmers
intercrop.

From the sustainability point of view, agricultural land uses difl‘er from each other
not only in how well they protect the soil against erosion but also in how much organic
matter they produce. In the tropics where decomposition rates are high, low production of
organic matter can easily reduce the organic matter content of the top soil and thereby
compromise its capacity to hold water and nutrients, especially nitrogen (Ruthenberg
1976). Most annual crops produce so little organic matter that lands degrade unless
fallowed periodically. Also coffee produces only modest amounts of organic matter, but
soil fertility is typically maintained by bringing in mulch from elsewhere. Among the
principal crops of Rwanda, bananas are superior producers of organic matter, which,
along with the low rates of erosion, explains how banana groves can remain highly
productive for several decades without fallowing (Kajumulo-Tibaijuka 1984). Trees,
bushes. and anti-erosion grass lines also produce organic matter which can be used to

improve soils.

Thus, even if we consider degradation associated with the loss of organic matter

rather than merely that associated with erosion, the bottom line is broadly similar: The

96

expansion of annual crops at the expense of fallow and pasture is making the situation
worse, whereas the expansion of bananas is making it better.

5.6 Conclusions

In this chapter we have presented household-level survey data showing how land-
scaree Rwandan farmers allocate a higher share of their land to cultivation in general and
to roots and tubers in particular than do farmers with relatively large amounts of
cultivable land per adult equivalent. We have further shown that the national pattern of
small farmers relying heavily on white potatoes only arises because farms are small on
the volcanic soils of the cool northwestern highlands, where potatoes thrive. In most of
Rwanda, farming systems are banana-based and the key difference between farm size

categories is the heavy reliance of small farmers on sweet potatoes.

Controlling for a set of other factors, we have shown evidence suggesting that
small farmers respond to land scarcity by growing more bananas, beans, and maize.
Notably, bananas are among these crops but sweet potatoes are not. In light of the
longitudinal trend towards roots and tubers, we hypothesize that smallholders may be
responding to population pressure by growing more bananas, and exchanging them for
sweet potatoes. Thus, population growth would be increasing the importance of sweet
potatoes in the diets and the role of bananas on the fields of small farmers. We will study
these hypotheses below.

The results also show that in areas close to paved roads, farmers grow more
bananas and less cereals and cassava. Since bananas protect the soils much better against
erosion than cereals and cassava do, the specialization made possible by infrastructure
development, urban demand and food imports appears to increase the environmental

sustainability of agricultural intensification in Rwanda.

97

We have found no evidence that lack of liquidity would be a constraint that
prevents poor households from choosing profitable or sustainable land uses. For most
Rwandan households, the principal intensification option, and the one that also happens
to be a clear step towards environmental sustainability, is increased reliance on bananas,
and this option is characterized neither by large start-up costs nor by long payback
periods.

Environmentally, the land use changes associated with population growth are a
mixed bag. The expansion of annual crops at the expense of fallow and pasture increases
erosion and reduces the production of organic matter. Both trends contribute to
degradation. On the other hand, the expansion of bananas and, to a lesser extent, of coffee
and tea, makes agriculture somewhat more sustainable. The net impacts are unclear,
except for the high-altitude areas, where land use is becoming less sustainable due to

demographic pressure.

98

Chapter 6

PARTICIPATION IN AGRICULTURAL PRODUCT MARKETS

6.1 Income Patterns
In this section we briefly look at the overall patterns of incomes. The purpose is to

provide the background for the more detailed analysis of market participation.

Using farm survey data to compute estimates for total household income in semi-
subsistence agriculture requires many conceptual decisions on the definition of income, a
large number of computations and transformations, and numerous shortcuts and
assumptions to deal with data limitations and computational constraints. Subsistence
consumption must be valued, and adjustments may be needed for seeds and other inputs
and for cross-household differences in the purchasing power of money. In Rwanda, the
importance of traditional beers means that the estimates on how much ingredients are
needed for beers are among the most important parameters in income computations. Our

#7 . “- ..—---'

if- Ekefinition of incownlurinciple includes implicit income from subsistence food

J

production (that ' , food production for home consumption), sales of agricultm'al products

 

and beer, off-farm wages and salaries, gifis received and some minor items. The main
issues in how the definition was operationalized and what auxiliary assumptions and
shortcuts were made were discussed above in chapter four. The details are available in the
transformation programs, which consist of thousands of lines of SPSS code.

99

Table 6-1: Gross Household Income Sources by Income Quartiles

 

Rwandan Francs Income Quartile (FRw/AE) .All Farms
<S370 5370- 8040- >13120
8040 13120

Home Consumption 11,161 19,323 28,044 45,396 25,757

farm Sales 4,125 9,568 16,250 26,135 13,881
Crops 1,541 3,643 7,360 13,271 6,372
Beer 1,834 4,004 5,805 8,272 4,943
Livestock 750 1,921 3,085 4,592 2,565

Purchased Inputs -707 -1,486 -2,600 -7,653 -3,054
Ag. Labor -342 -909 -1,749 -6,944 -2,429
Other Ag. -19 -45 -131 -l41 -84
Beer Inputs -346 -532 -719 -567 ~542

Labor Sales 3,097 4,609 6,551 31,118 11,088
Agricultural 1,659 1,891 1,794 1,308 1,669
Other unskilled 602 873 1,453 1,815 1,179
Skilled 836 1,845 3,303 27,996 8,241

Gifts Received 771 830 1,273 2,279 1,275

Other 401 222 705 2,696 983

Net Income 18,847 33,066 50,223 99,971 49,930

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

 

Table 6—1 shows the estimated gross household incomes by source, costs of
purchased inputs, and total net incomes both for rural Rwandan households on average
and for income quartiles computed by making households in terms of income per adult
equivalent. The “pm-industrial” nature of Rwanda’s agriculture can be seen in that the
use of fertilizer and other external inputs is minimal. The only inputs purchased in
significant quantities are agricultural labor and ingredients (sorghum and bananas) for
traditiorml beer brewing.” The fact that households prefer to use their cash for labor
rather than for chemical inputs, along with the fact that households with highest incomes

 

’7 In addition, some farmers purchase seeds and capital inputs ' 'ly livestock) and
pay cash rents, but due to data problems, these are not mclu ed in income
computations. Other DSA data show that lessthan 10% of 'cultural land is rented,
often at no or low charge (perhaps because it rs ofien rented m relatives).

100

Table 6-2: Net Household Income Sources by Income Quartiles

Francs/household Income Quartile (FRw/AE) All Farms
<5370 5370- 8040- >13120
8040 13120
Home Consumption 10,904 18,755 26,911 40,872 24,174
Farm Sales 3,675 8,651 14,783 23,006 12,410
Crops 1,501 3,505 7,007 11,911 5,911
Beer 1,438 3,307 4,841 6,904 4,093
Livestock 736 1,840 2,934 4,191 2,406
Labor Sales 3,097 4,609 6,551 31,118 11,088
Agricultural 1,659 1,891 1,794 1,308 1,669
Other unskilled 602 873 1,453 1,815 1,179
Skilled 836 1,845 3,303 27,996 8,241
Gifts Received 771 830 1,273 2,279 1,275
Other 401 222 705 2,696 983
Net Income 18,847 33,066 50,223 99,971 49,930

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

use only a fraction of a percent of their cash outlays for purchased inputs, suggests that

liquidity is not the primary contraint of input use. Instead, inputs are either unavailable or

considered unprofitable by most farmers.”

The inequality of income distribution in rural Rwanda can be seen in that the

estimated mean net income is five times as high in the richest quartile of households
(hereafter “the rich”) as it is in the poorest quartile (hereafter “the poor”). Although the
tails of the estimated income distribution probably are strongly afi‘ected by data errors
and although the rich have somewhat larger families than the poor, rural difi'erentiation is

rmdoubtedly a major issue in Rwanda.

 

2' A notable exception to this is the fungicide used to control mildew in potatoes and

“MDEKKS.

lOl

In Table 6-2 the purchases of beer inputs are deducted from beer sales and the
costs of labor and other inputs are allocated proportionally to home consumption and
farm sales. Note that the beer sales revenue is net of purchased inputs, and not net of the
imputed values of home-grown ingredients. In other words, it includes returns to both
farming of bananas and sorghum for beer, and brewing for sale.29 As discussed below, the
market value of ingredients accounts for most of the market value of beer. Rather than
being a separate income-generating activity, traditional brewing (at least that of banana
beer) is more correctly seen as the last phase of agricultural production needed to process
products for transportation and sale. Its role is close to that of peeling and drying cassava
for sale. Table 6-3 shows the shares of net income sources.”

 

2’ To avoid double-counting the estimated quantities of bananas and sorghum used for
beer sales were deducted from home consumption.

3° The ntages in table 6-3 were computed at household-level and then aggregated,
not e other way round. Instead of reportm the shares of total income commglfrom
each source, they show how much househo ds on average 1get income from eac source
without givm more weight to the more prosperous house olds. Thus, skilled labor is
reported m Ta 1e 6-3 to account for 7% of average rural incomes, although its share of ‘
aggregate rural incomes, implicit in Table 6-2, 15 16%. The difference arises since
incomes from skilled labor are highly concentrated among the high-income households.
Both percentages would obviously be much lower if the sample excluded teachers,
masons and o ers employed in off-farm activities and instead covered only households
earning a bulk of their income from agriculture.

102

Table 6-3: Net Household Income Shares by Income Quartiles

 

Percent Income Quartile (FRw/AE) A11 Farms
<5370 5370- 8040- >13l20
8040 13120
Home Consumption 58 58 54 49 55
Farm Sales 18 24 28 27 24
Crops 8 10 13 14 11
Beer 7 9 9 8 8
Livestock 3 5 6 5 5
Labor Sales 16 14 13 19 16
Agricultural 9 7 5 2 6
Other unskilled 3 2 3 2 3
Skilled 4 5 5 15 7
Gifts Received 5 3 3 3 3
Other 2 1 1 3 2
Net Income 100 100 100 100 100

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

 

Tables 6-1 to 6-3 show that households in rural Rwanda are still quite subsistence-
oriented, earning on average just above half of their income in the form of food crops
grown for home consumption. Sales of agricultural products and labor account for most
of the other half. Poor households are typically somewhat more subsistence-oriented than
those in the other end of the income distribution. Sales of crops and livestock are more
important for better-off households, whereas beer incomes on average account for almost
one-tenth of the estimated incomes in all income quartiles. Wage labor supplements on-
farm incomes in all quartiles, but while the better-0E households sell primarily skilled
labor at high wages, the poorest households work primarily in agriculture at low wages.
Gifts, land rents and other off-farm sources of income are relatively unimportant in rural
Rwanda.

103

Table 64: Net Household Income Shares in Banana Zone

Percent Income Quartile (FRw/AB) All Farms
<537O 5370- 8040- >13120
8040 13120
Home Consumption 60 58 55 48 55
Farm Sales 18 25 29 27 25
Crops a 10 13 14 11
Beer 8 10 11 8 9
Livestock 2 5 5 5 4
Labor Sales 15 14 12 20 15
Agricultural 9 5 4 1 5
Other unskilled 3 3 3 2 3
Skilled 4 5 5 16 8
Gifts Received 4 3 3 3 3
Other 2 1 1 2 2
Net Income 100 100 100 100 100

Source:DSA/MINAGRI, Agricultural Survey 1990; 955 households

 

In the preceding chapter we showed that on the temperate highlands (“potato
zone” and “marginal zone”), where roughly 15 percent of Rwanda’s farmers reside,
farming systems are quite different from the rest of Rwanda, which we called “banana
zone.” In accordance with our focus on banana zone in chapter five, we tabulate also
income shares for just this area. Table 6-4 shows that the income patterns do not change
much if one leaves out the temperate highlands. The explanation lies in that the two zones
of the temperate highlands diverge from the national patterns into opposite directions.
The volcanic potato zone is relatively highly commercialized and has few livestock,
while the marginal zone relies more on subsistence food production and has more

livestock.

Finally, we take a look at the relationship between farm size and household

income within the banana zone. Table 6-5 shows that only on-farm incomes are closely
\MR

104

\ associated with farm size. On average, household‘sinflthevhgnom farm size quartile get

 

 

more income from labor sales than those in the top quartile. However, labor incomes are
dominated by earnings from skilled employment, which, as discussed above, are highly
concentrated among a handful of households with members who work as teachers,
masons and other skilled employees. While skilled jobs have allowed a small number of
land-scarce households to escape poverty, most small farmers are poor.

Nevertheless, the mean on-farm income in the bottom farm size quartile is nearly
twice as high as that in the bottom income quartile (Table 6-2). This suggests that many
of the smallest farmers have relatively high yields, whereas many of the poorest
households are poor not just because of land scarcity, but also because of the low

productivity of their land. The simple correlation coefficient between farm size and

‘ income, both scaled by family size expressed in adult (consumer) equivalents, is .43,

implying that variation in farm size “explains” only one-fifth of the variation in income.
The association between farm size and income would obviously be tighter, if the survey

had excluded those earning most of their income fi'om outside agriculture.

105

Table 6-5: Net Household Income Sources by Farm Size in Banana Zone

Francs/household RARM SIZE QUARTILE (Ares/AB) All Farms

Home Consumption 16,432 21,715 29,281 34,104 25,549

Farm Sales 7,913 10,954 13,937 18,721 12,990
Crops 4,117 4,497 5,779 9,449 6,012
Beer 2,564 4,096 5,661 5,895 4,585
Livestock 1,233 2,362 2,497 3,377 2,392

Labor Sales 16,032 7,017 14,746 9,143 11,606
Agricultural 2,301 1,867 1,097 507 1,426
Other unskilled 2,389 1,127 726 498 1,164
Skilled 11,342 4,024 12,923 8,137 9,016

Gifts Received 882 1,196 1,232 1,677 1,256

Other 703 550 1,359 1,085 924

Net Income 41,962 41,433 60,555 64,730 52,325

SourcerDSA/MINAGRI, Agricultural Survey 1990; 960 households

6.2 Patterns of Agricultural Transactions

6.2.1 National Patterns
Despite the importance of subsistence production, Rwanda’s rural households sell

and buy large quantities of agricultural products. Much of the trade occurs within the
rural sector, allowing households to exchange their surpluses of some products for those
that they do not produce in sufficient quantities. But rural areas also export significant
quantities of some products to urban areas and neighboring countries, while relying on
imports of other products.

Table 6-6 reports the estimated sales, gifts given, purchases, gifls received, and
net outflows in (exports from) the rural sector in agricultural year 1990. Since some
households are also trading (buying in order to sell), the first four measures are adjusted
by netting monthly and recording only the sums of non-negative monthly net figures for

106

each household." The adjustment has no impact on the fifth column, which we primarily
focus on below. Negative net sales and gifts indicate that rural areas were importing
products from elsewhere. The estimates are based on DSA’s nationally representative
rural household survey sample. Statistics on beer bananas and sorghum include only
transactions as such, not those in the form of traditional beers.

 

3’ The month] netting admittedly is an imperfect way of eliminating tradin . It fails to
exclude ing when purchases and sales do not occur during the same endar month
and it incorrectly excludes some purchases for consumption and sales from own
production, especially in the case of penshable products such as beer.

107

Table 6-6: Rural Transaction Quantities

Millions Sales Gifts Purchases Gifts Net Sales
of units Given Received and Gifts
Beans 16 11 5.47 75 91 5.39 -59 71
Peas .63 .21 2.35 .30 -1.81
Peanuts .86 .11 1.03 .07 -.13
Soybeans 1.07 .25 1.50 .25 -.44
Sorghum 25.42 1.93 50.57 3.31 -26.53
Maize 4.84 3.62 7.71 2.82 -2.07
Wheat 2.11 .19 .98 .15 1.16
Finger Millet .11 .09 .05 .08 .07
Rice 2.23 .24 3.54 .22 -1.29
Cassava 42.71 4.57 68.25 2.80 -23.77
White Potato 77.95 5.99 52.85 7 ° 3.23 27.86
Sweet Potato 58.04 17.22 48.15 8.60 18.51
Taro 4.96 1.48 6.06 1.11 -.74
Cocoyam .05 .02 .11 .04 -.08
Cooking Banana 47.02 7.63 16.99 4.87 32.80
Beer Banana 59.01 10.42 61.71 5.93 1.79
Fruit Banana 7.26 .54 2.53 .41 4.86
Coffee 41.13 .00 .00 .00 41.13
Banana Beer 207.08 39.97 6.99 12.71 227.35
Sorghum Beer 67.62 15.32 8.75 11.05 63.13
Milk ' .06 .00 .00 .00 .05
Cattle .13 .02 .05 .03 .07
Pigs .19 .01 .07 .01 .12
Sheep .28 .02 .09 .02 .20
Goats .65 .05 .20 .07 .43
Chicken .64 .10 .38 .14 .22
Ag. Labor 23.68 N R 34.30 N.R —10.63
Non-Ag. Unsk. L. 15.02 N.R. 8.20 N.R 6.82
Skilled Labor 41.43 N R 4.12 N R 37 31

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households
Units are kilograms, except for beer (litres), milk (litres), livestock
(heads), and labor (mandays). N.R.- Data not recorded.

108

According to Table 6-6, Rwanda’s rural sector sells very large quantities of
traditional beers, especially banana beer. In part, this reflects underreporting of beer
purchases by the respondents of the rural sample.32 Nevertheless, rural households sell
very large quantities of beer to urban areas. Other crops exported from rural areas in large
quantifies include cofl‘ee, cooking bananas, white potato, and sweet potato. Large rural
trade occurs inbeerbananas,buttherearealmostnorural exports. Beerbrewingis
characteristically batch production and requires some rural assembly, but long—distance
transportation typically occurs after brewing, which greatly increases value to weight
Also root crops are largely traded within the rural sector.

 

3’ If we assume that Rwanda’s urban pulation (or more 993,111.: non-farm
po ulation not covered by DSA’s armhousehold survey) m l was 700, 000 and that
,000 o fthis scpopulation was in the us that consumed beer, taking the rural
eXpo rts reportem Table 6-6 at face v negro an assuming no net exports would imply
that each urban beer drinker on average consumes nearly two liters of beer per day.
This rs clearly not credible and suggests that most rural purchases were not reported by
the respondents.

109

Table 6-7: Rural Transaction Values

FRv/household Sales Gifts Purchases Gifts Net Sales
Given Received and Gifts
Beans 387 154 2,362 178 ~1,998
Peas 31 12 100 19 ~76
Peanuts 56 7 74 6 ~17
Soybeans 32 8 47 8 ~15
Sorghum 382 30 1,038 67 ~692
Maize 72 69 168 65 ~91
Wheat 61 5 24 4 37
Finger Millet 4 3 3 4 ~0
Rice 67 7 224 12 ~162
Cassava 292 29 375 19 ~73
White Potato 634 49 618 40 25
Sweet Potato 384 113 354 63 81
Taro 47 14 61 10 ~10
Cocoyam 1 0 1 0 ~1
Cooking Banana 264 41 103 30 173
Beer Banana 191 34 244 22 ~41
Fruit Banana 33 2 8 1 27
Coffee 2,684 0 0 0 2,684
Other Ind.Crops 413 0 36 0 377
Beef' 14 0 620 0 ~606
Other Meat 6 0 224 0 ~218
Fish 15 0 212 0 ~197
Banana Beer 4,271 821 163 323 4,607
Sorghum Beer 754 169 100 123 701
Milk . 147 0 80 0 68
vegetables 117 0 172 0 ~55
Fruits 25 0 15 0 10
Fertilizer 0 0 7 0 ~7
Pesticides 0 0 68 0 ~68
Forest Prod. 232 0 50 0 182
Fodder Crops 6 0 7 0 ~1
Cattle 1, 037‘ 151 351 223 615
Pigs 297 15 61 19 232
Sheep 220 17 70 18 148
Goats 733 57 203 78 509
Chicken 104 16 57 22 40
Ag. Labor 1,702 N.R. 2,428 N.R ~725
Nun-Ag. Uhsk. L. 1,228 N.R. 327 N.R. 901
Skilled Labor 8.285 N.R. 498 N.R. 7,787
Total 25,287 1,825 11,571 1,353 14,188

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households
N.R.- Data not recorded

110

Rmal Rwanda imports large quantities of beans, sorghum, and cassava, although
much of the sorghum is used for beer, and sold. further. It is particularly noteworthy that
Rwandan farmers not only fail to supply urban areas with beans, the main source of
proteins in Rwandan diets, but also supply only a fraction of the rural market demand for
beans.

In contrast to beans, rural Rwanda does supply cities with meat. Table 6~7 reports
the estimated transactions in money terms and also includes products such as meat, for
which data was collected on the cash values of the transactions but not-on the quantities
transacted. A comparison of meat purchases and livestock sales indicates that rural
Rwanda exports large numbers of livestock to urban areas and neighboring countries.

Table 6-7 shows that none of the main food crops brings significant amounts of
cash to rural Rwanda. Although white potatoes, sweet potatoes, cooking bananas, beans,
cassava, and sorghum are important sources of cash for some farmers, most of the trade is
intra-rural. Moreover, the three food crops exported from rural Rwanda in notable
quantities (cooking bananas, white potatoes, and sweet potatoes) are much cheaper
products than the beans imported from neighboring countries.

To finance the imports of beans and all the industrial products, rural Rwanda sells
labor, beer, coffee, and livestock, in this order. The role of unskilled labor resembles that
of food crops in that both are important sources of cash for some households but neither
brings significant amounts of cash to rural areas.33 However, as discussed in detail below,
the roles differ in that the poor are net sellers of unskilled labor but net buyers of all food

crops.

 

’3 That the farm households on the aggre‘ gate appear to be net importers of agricultural
labor should not be interpreted toomdicate migrato labor fiom the cities or
neighboring countries. Indeed, m1 grato labor pro ably flowed from rural areas to
cities and not vice versa (Olson, l 942. instead, the numbers probabl reflect the
samplin design that sought to cover arm households and excluded arm workers that
did not ve eir own farms.

lll

Interestingly, the single most important flow of cash to agricultural households
consists of the wages of skilled workers, such as teachers, soldiers, and employees of

development projects. However, as we will see below, labor incomes, especially those
from skilled employment, are highly concentrated.

6.2.2 Agricultural Transactions in Banana Zone
In this section we tabulate the estimates of the net sales and gifts given (rural -

exports) by cropping zone. Much of the discussion deals with the banana zone, which
covers 85 percent of the observations and is the primary focus of this study. Table 6~8
presents the quantities of crop and labor sales per household and Table 6-9 shows the

summaries for all the recorded transactions in monetary terms.

112

Table 6-8: Net Household Sales by Cropping Zone

units/household Cropping Zone Rwanda
Banana Potato Marginal
Zone Zone Zone
Beans ~44 ~97 ~57 ~48
Peas ~1 ~S ~2 ~1
Peanuts ~0 ~0 ~0 ~0
Soybeans ~0 0 ~0 -0
Sorghum ~16 ~67 ~65 ~22
Maize ~2 ~21 7 ~2
Wheat 1 ~0 1 1
Finger Millet 0 0 0 0
Rice ~1 ~2 ~1 ~1
Cassava ~17 ~52 ~23 ~19
White Potato ~37 1,158 ~22 23
Sweet Potato 23 ~53 ~54 14
Taro ~1 0 0 ~1
Cocoyam ~0 0 0 ~0
Cooking Banana 30 ~5 ~1 26
Beer Banana 5 0 ~37 2
Fruit Banana 5 ~6 ~1 4
Coffee 37 0 3 33
Banana Beer 199 ~3 40 178
Sorghum Beer 48 62 60 50
Ag. Labor ~12 20 17 ~8
Non-Ag. Uhsk. L. 5 7 11 5
Skilled Labor 30 22 22 29

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households
units are kilograms, except for beer (litres)and labor(mandays).

 

Although the three zones
high-altitude zones are so small that the na onal patterns largely hold for the banana
zone. Like Rwandan households on average, households in the banana zone exchange
skilled labor, banana beer, cofi‘ee, and livestock for beans and cereals. Some agricultural
laborers fi'om the high-altitude zones are employed in the banana zone and some white

 

potatoes are imported from the potato zone.

113

Table 6~9: Net Household Sales in Frw by Cropping Zone

FRw/household Cropping Zone Rwanda
Banana Potato Marginal
Zone Zone Zone
Beans ~1,885 ~3,913 ~2,305 ~2,015
Peas ~56 ~266 ~152 ~73
Peanuts ~21 ~3 ~18 ~20
Soybeans ~15 0 ~21 ~15
Sorghum ~550 ~1,984 ~1,842 ~715
Maize ~69 ~825 84 ~95
Wheat 42 ~16 39 38
Finger Millet ~1 ~0 15 ~0
Rice ~169 ~237 ~43 ~163
Cassava ~52 ~294 ~199 ~75
White Potato ~568 11,214 ~239 38
Sweet Potato 160 ~561 ~530 74
Taro ~11 0 ~4 ~10
Cocoyam -1 0 0 ~1
Cooking Banana 200 ~46 ~8 173
Beer Banana ~23 0 ~265 ~39
Fruit Banana 33 ~22 -3 27
Coffee 3,062 0 215 2,704
Other Ind.Crops 416 ~112 296 381
Beef ~654 ~528 ~99 ~608
Other Meat ~242 ~17 ~46 ~216
Fish ~226 ~12 ~8 ~199
Eggs 32 4 22 30
Money 10 0 16 10
Banana Beer 5,142 ~104 982 4,581
Sorghum Beer 680 916 824 702
Milk 77 2S 0 68
vegetables ~59 ~213 67 ~57
Fruits 11 ~5 12 11
Fertilizer ~5 ~24 ~26 ~7
Pesticides ~17 -1,035 ~41 ~69
Forest Prod. 164 ~99 614 184
Fodder Crops ~1 ~10 1 ~1
Cattle 597 931 717 622
Pigs 244 0 215 230
Sheep 130 167 388 151
Goats 494 699 641 515
Chicken 42 ~4 14 38
Ag. Labor ~1,101 2,081 1,888 ~727
Non-Ag. Dusk. L. 860 1,259 1,296 911
Skilled Labor 8,319 4,671 4,179 7,838
Total 14,989 11,636 6,676 14,219

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

114

\_ 5 True to its name, the potato zone is highly dependent on white potato exports. It
a if} also relies heavily on pesticides, primarily fimgicides for potatoes, whereas most farmers
. in other zones do not use industrially produced agricultural chemicals. Because of their
profitable potato sales, potato farmers can both augment the output of their tiny farms by
purchasing large quantities of beans and sorghum and have almost as much cash left for

their non-staple needs as farmers in the banana zone, where farms are much larger.

 

Also households in the marginal zone buy significant amounts of beans and
sorghum, but because they lack profitable cash crops, their incomes from sales are small
and hence they have very little left for their other needs.

6.2.3 Crop Transactions by Income in Banana Zone
This section focusses on how the patterns of crop transactions vary by household

income within the banana zone. Table 6-10 shows the estimates for crop and labor sales
quantities per household by income quartile, and Table 6-11 presents the summaries for

all the recorded transactions in monetary terms.

The tables show that food transactions are strongly associated with poverty and
that the nature of this association varies by crop. On average, households in all but the
richest quartile buy significant amounts of beans and sorghum. The rich are roughly self-
suficient, imports being the source of supply for the poor. In contrast, cassava, the third
important food purchased by poor households, is a minor cash crop for the rich. The

‘ regional analysis in the Appendix A suggests that the poor buyers are nevertheless largely
dependent on imports, since the rich sellers are located mostly in eastern Rwanda and
supply mostly urban consumers in Kigali, whereas the poor buyers are found mostly in
western Rwanda, especially in the south-western prefecture of Cyangugu, where most

cassava in the market is presumably supplied by imports from Congo.

115

Table 6-10: Net Household Sales Quantities by Income Quartile in Banana Zone

 

units/household Income quartile (FRw/AB) .All Farms
<5350 5350- 8020~ >13230
8020 13230
Beans ~53 ~60 ~57 ~5 ~44
Peas ~1 ~1 ~1 ~2 ~1
Peanuts ~0 ~0 ~1 1 -0
Soybeans ~2 -0 ~0 0 ~0
Sorghum ~26 ~24 ~23 8 ~16
Maize ~5 ~S ~3 7 ~2
Wheat ~0 1 3 ~0 1
Finger Millet ~0 ~0 0 ~0 0
Rice ~1 0 ~1 ~2 ~1
Cassava ~44 ~52 ~17 46 ~17
White Potato ~17 ~29 ~37 ~64 ~37
Sweet Potato ~43 7 35 89 23
Taro ~6 ~7 1 10 ~1
Cocoyam ~0 ~0 ~0 ~0 ~0
Cooking Banana ~5 ~1 ~6 132 30
Beer Banana 19 4 ~33 32 5
Fruit Banana 1 4 7 6 5
Coffee 10 22 49 66 37
Banana Beer 64 148 224 355 199
Sorghum Beer 47 49 64 32 48
Ag. Labor 14 10 ~4 ~69 ~12
Non-Ag. Uhsk. L. 9 8 13 ~11 5
Skilled Labor 7 16 19 77 30

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households
Units are kilograms, except for beer(litres) and labor(mandays).

 

Sweet potato, Rwanda’s single most important staple food, is another minor cash
crop for richer households and another food crop purchased by the poor. Since bean,
sorghum, and cassava supplies come primarily fi'om neighboring countries, sweet potato
is the only major crop bought primarily by the poor Rwandans fi'om their better-ofl‘
neighbors. This is in clear contrast to white potatoes, which are relatively expensive and
are purchased primarily by better-off households.

Cooking banana is another relatively expensive food crop, but unlike white
potato, it is sold rather than purchased by the better-off rural households. As shown in

116

 

Appendix A, much of the explanation is regional: the main production area and the only
region where cooking bananas are an important cash crop is in eastern Rwanda, where
i farms are large and households relatively prosperous.

In the banana zone, two items, beer and cofl‘ee, dominate agricultural sales
revenues in all income quartiles. Although the net sales are concentrated strongly among
the rich households, beer and coffee sales are relatively more important for the poor,
since while the rich also sell modest amounts of food crops and cattle, the poor rely more
exclusively on beer and coffee. While the income tabulations above showed that even the
poorest get some income fiom food crop sales, the net flow tabulations here indicate that
as a group the poor are net buyers and would be hurt rather than helped by higher prices.
This naturally is only the direct impact and indirect impacts, especially those transmitted
through labor markets, could alleviate or aggravate the situation, depending on what
brought about the higher food prices.

117

Table 6-11: Net Household Sales Values by Income Quartile in Banana Zone

 

FRw/household Income quartile (FRw/AE) .All Farms
<5350 5350~ 8020~ >13230
8020 13230
Beans ~2,127 ~2,467 ~2,306 ~610 ~1,885
Peas ~34 ~54 ~49 , ~86 ~56
Peanuts ~17 ~35 ~65 36 ~21
Soybeans ~59 ~15 2 7 ~15
Sorghum ~732 ~735 ~715 ~15 ~550
Maize ~154 ~151 ~88 ' 118 ~69
Wheat 4 38 119 0 42
Finger Millet ~3 ~3 2 ~2 ~1
Rice ~82 ~35 ~167 ~395 ~169
Cassava ~317 ~294 ~64 465 ~52
White Potato ~273 ~450 ~564 ~973 ~568
Sweet Potato ~396 19 247 740 160
Taro ~93 ~87 19 113 ~11
Cocoyam ~1 ~2 ~0 ~0 ~1
Cooking Banana ~28 1 ~42 875 200
Beer Banana 54 ~52 ~177 96 ~23
Fruit Banana 14 29 . 49 37 33
Coffee 796 1,830 4,035 5,487 3,062
Other Ind.Crops 50 408 376 811 416
Beef ~131 ~301 ~663 ~1,509 ~654
Other Meat ~55 ~119 ~254 ~534 ~242
Fish ~131 ~207 ~228 ~333 ~226
Eggs 17 54 52 3 32
Money 0 41 ~1 ~3 10
Banana Beer 1,716 3,950 5,914 8,826 5,142
Sorghum Beer 628 672 921 488 680
Milk 21 87 111 82 77
vegetables ~44 ~53 42 ~182 ~59
Fruits 5 8 48 ~18 11
Fertilizer ~6 0 ~11 ~2 ~S
Pesticides ~0 ~18 ~45 ~3 ~17
Forest Prod. 119 44 96 404 164
Fodder Crops 0 8 12 ~23 ~1
Cattle ~106 390 622 1,449 597
Pigs 34 347 316 258 244
Sheep 56 137 143 179 130
Goats 176 237 597 957 494
Chicken 30 49 107 ~20 42
Ag. Labor 1,221 814 ~272 ~6,158 ~1,101
Non-Ag. Unsk. L. 614 891 1,282 625 860
Skilled Labor 616 1,803 3,130 27,784 8,319
Total 1,381 6,778 12,532 38,978 14,989

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

118

The poorest households sell very little skilled labor (otherwise they would not be
poor) but augment their farm product sales revenues by working for richer farmers. .
Important as these incomes are, they do not match the sales revenues from beer. The
poorest use most of their sales revenues to buy beans and other staple foods. The bottom
line in Table 6-11 shows that the richest quartile accounts for almost two-thirds of the
rural demand for non-agricultm'al products.

6.2.4 Crop Transactions by Farm Size in the Banana Zone
, As discussed above, poverty in rural Rwanda is only loosely associated with farm

' size, primarily for two reasons ' ofi‘~farm incomes are important also in rural areas
and make income distribution less dependent on land endowments. land quality
varies and while some of the large farms, especially those in eastern Rwanda, are
relatively prosperous, others, especially those on the Congo-Nile Divide, are among the
poorest due to the low productivity of their fields.

  
 

 

While transactio -‘ u - . by income ~ tegories are valuable for understanding

. 3 would be distributed@ize

are the first approximation of how population growth in rural areas would affect rural

how the benefits and costs of prr -

supply and demand. Table 6-12 tabulates net sales quantities by farm size and Table 6-13

presents the transactions in monetary terms.

119

Table 6-12: Net Household Sales Quantities by Farm Size in Banana Zone

 

Obits/household Farm size quartile (ares/AB) .All Farms
<9 9~14 14-24 >24
Beans ~65 ~58 ~52 ~13 ~46
Peas ~1 ~2 ~1 1 ~1
Peanuts ~1 ~0 ~1 2 ~0
Soybeans ~0 ~1 0 ~0 ~0
Sorghum ~16 ~29 ~17 4 ~15
Maize ~8 ~4 l 6 ~l
Wheat 0 l 2 1 1
Finger Millet ~0 ~0 0 ~0 0
Rice -6 ~2 ~2 5 ~1
Cassava ~101 ~11 ~20 64 ~15
White Potato ~66 ~25 ~31 ~29 ~37
Sweet Potato ~37 5 58 67 24
Taro ~8 ~3 ~l 9 -1
Cocoyam ~0 ~0 0 ~0 ~0
Cooking Banana ~10 1 15 113 31
Beer Banana ~10 ~20 ~7 53 4
Fruit Banana ~1 7 4 12 6
Coffee 38 31 29 50 37
Banana Beer 113 171 248 275 203
Sorghum Beer 27 62 63 48 50
Ag. Labor 13 5 ~18 ~53 ~14
Non-Ag. Uhsk. L. 15 10 ~4 0 5
Skilled Labor 40 18 40 26 31

Source:DSA/MINAGRI, Agricultural Survey 1990; 1095 households
units are kilograms, except for beer (litres) and labor (mandays).

 

Regarding food crops, the big picture is still that the top quartile purchases less or
sells more farm products than the bottom quartile. Small farms, just as the poor ones, are
net buyers of beans, sorghum, white potato, cassava and sweet potato, whereas large
farms, as the rich ones, are net sellers of cassava and sweet potato. The large difference
between the two rankings comes in the main cash crops and in labor incomes. Whereas
the coffee and beer sales of the bottom income quartile were only a small fraction of
those of the top income quartile, the differences between the farm size quartiles are less
extreme. Labor sales are even higher in the bottom farm size quartile than they are in the
top quartile.

120

Table 6-13: Net Household Sales Values by Farm Size in Banana Zone

FRw/household Farm size quartile (ares/AB) All Farms
<9 9~14 14-24 >24
Beans ~2,628 ~2,355 ~2,157 ~753 ~1,953
Peas ~67 ~98 ~56 36 ~45
Peanuts ~73 ~32 ~81 95 ~21
Soybeans ~21 ~35 14 ~7 ~13
Sorghum ~545 ~848 ~628 ~107 ~529
Maize ~251 ~112 10 83 ~64
Wheat ~0 61 78 25 41
Finger Millet ~1 ~3 4 ~5 ~1
Rice ~460 ~202 ~160 85 ~178
Cassava ~677 ~7 ~113 572 ~41
White Potato ~959 -415 -512 ~435 ~573
Sweet Potato ~380 41 460 527 171
Taro ~106 ~48 ~13 118 ~10
Cocoyam ~1 ~3 1 0 ~1
Cooking Banana ~64 4 117 743 208
Beer Banana ~153 ~144 ~20 198 ~26
Fruit Banana ~0 46 20 78 37
Coffee 3,139 2,583 2,428 4,196 3,097
Other Ind.Crops 15 88 654 867 413
Beef ~748 ~389 ~681 ~852 ~666
Other Meat ~188 ~150 ~313 ~364 ~255
Fish ~561 ~40 ~211 ~104 ~223
Eggs 61 23 16 39 34
Honey 41 ~11 3 ~8 5
Banana Beer 3,022 4,514 6,381 6,862 5,230
Sorghum Beer 377 857 846 728 707
Milk ~123 116 ~27 321 78
vegetables ~161 30 ~139 37 ~55
Fruits ~1 18 23 9 12
Fertilizer 0 ~6 ~4 ~8 ~5
Pesticides -3 ~10 ~40 ~21 ~18
Forest Prod. 37 92 439 125 173
Fodder Crops ~13 16 ~0 ~3 0
Cattle 32 310 899 844 528
Pigs 150 365 178 302 251
Sheep 125 109 146 138 130
Goats 338 362 455 856 508
Chicken 1 86 65 37 48
Ag. Labor 1,186 556 ~1,646 ~4,915 ~1,261
Nun-Ag. unsk. L. 2,106 1,030 164 185 852
Skilled Labor 11,366 3,810 12,176 7,095 8,511
Total 13,812 10,209 18,775 17,620 15,098

Source:DSA/MINAGRI, Agricultural Survey 1990; 1095 households

121

While the above tables show clearly that small farmers rely more on markets for
their food, the message regarding cash cr0ps is less clear. Both farm size and cash crop
sales are much smaller in the household quartiles with less land per person. To find out
whether two small farms typically sell more cash crops than one large farm operating the
same amount of farm land, we next tabulate farm transactions per hectare, again
restricting the analysis to Rwanda’s banana zone. Moreover, to understand the role of
sorghum, which is both a staple food product and an ingredient of sorghum and banana
beer, we convert beer transactions into the estimated ingredient needs included in those

transactions.“

 

3‘ Beer and fiuit bananas are combined, since fiuit bananas are frequently used in beer
brewing along with beer bananas.

122

Table 644: Net Sales per Household by Farm Size in Banana Zone

Kg/household Quartiles of C-ABLB/AB All Farms
<9 9-14 14-24 >24
ares/AB ares/AB
Beans ~65 -58 -52 -13 -46
Peas -1 -2 -l l -1
Peanuts -1 O -1 2 0
Soybeans O -1 0 0 0
Sorghum 6 12 33 52 26
Maize -8 -4 1 6 -1
Wheat 0 1 2 1 1
Finger Millet O 0 O 0 0
Rice -6 -2 -2 S -1
Cassava ~101 -11 -20 64 -15
White Potato -66 -25 -31 -29 -37
Sweet Potato - -37 5 58 67 24
Taro -8 -3 -1 9 -1
Cocoyam 0 O O O 0
Cooking Banana -1o 1 15 113 31
Other Banana 374 570 841 1001 702
Coffee 38 31 29 50 37

Source:DSA/MINAGRI, Agricultural Survey 1990; 960 households
Banana and sorghum sales also include sales as beer.

 

Table 6-14 indicates that when transactions in the form of banana and sorghum
beer are included, all farm size quartiles in Rwanda’s banana zone are net sellers of
sorghum. The sizable sorghum imports discussed above are not for food but for
transformation into beer. Primarily, sorghum purchases are needed to process beer
bananas for sale, a transformation that increases value to weight six-fold and thereby

greatly reduces transportation costs. Accounting for sales as beer, sorghum is one of the
W'

In Table 6-15, the transaction quantities are scaled by the hectares of cultivable

 

 

 

land. The purpose of the table is to shed light on what happens to the aggregate rural

123

exports (net sales) when population growth makes farms increasingly small. Even if small
farmers sold less of everything, their sales per hectare could be higher, suggesting that as
farms get subdivided under demographic pressure, marketed supply will increase rather
than decrease. According to Table 6-15, this holds for bananas and for coffee, but not for
sorghum. In other words, unless these associations are spurious results caused by
agroclimatic or other variation, demographic change should increase the marketed \
supplies of bananas and coffee, but reduce those of sorghum. In the next section we wi\

study this further, controlling for agroclimatic and other variation.

124

Table 6-15: Net Sales per Hectare by Farm Size in Banana Zone

Kg/hectar Quartiles of C-ABLB/AE All Farms
<9 9-14 14-24 >24
ares/AB ares/AB
Beans -297 -102 ~58 -19 -116
Peas -8 -5 -1 1 -3
Peanuts -4 -1 -1 1 -1
Soybeans -4 -2 O 0 -1
Sorghum 10 25 35 3O 25
Maize —35 -7 -1 2 -10
Wheat 0 2 3 1 1
Finger Millet 0 0 0 O 0
Rice -31 -3 -2 4 -8
Cassava -498 -6 -27 28 -119
White Potato -317 -49 -42 -22 -104
Sweet Potato -214 10 64 44 -21
Taro -42 -9 -2 7 -11
Cocoyam 0 0 0 0 0
Cooking Banana -65 6 6 64 4
Other Banana 1218 932 937 678 935
Coffee 140 51 33 46 66

Source:DSA/MINAGRI, Agricultural Survey 1990; 960 households
Banana and sorghum sales also include sales as beer.

 

6.2.5 Some Concluding Remarks
The transaction patterns presented above show that the partially but not fully

overlapping categories of small (land-scarce) and poor households are net buyers of food.
The national average pattern—the exchange of labor, beer and coffee for imported beans
and sorghum—holds for the mid-quartiles, but large and rich farmers also sell bulky food
crops to their neighbors (sweet potatoes) and to urban consumers (cooking bananas).

Finally, as discussed above, it should be noted that the survey covered not just
full-time farmers but also households with members employed in ofl-farm occupations. If

the definition of farm households were more exclusive, the role of labor sales would

125

appear much smaller than what the tables above suggest. Above all, the average numbers
tabulated for the smallest households would change. In our tables, the inclusion of the
relatively high-income semi-farmers, who can afford to eat better than their neighbors and
drink their banana beer instead ofselling it, increases the net purchases offood and
decreases the net sales of banana beer, especially in the bottom farm size category.

6.3 Multivariate Analysis of Crop Transactions
The transactions patterns discussed above show that land-scarce households on

average are simultaneously net buyers of food and net sellers of beer bamnas (mostly in
the form of beer) and coffee. Relative to farm size, small farmers produce more beer and
cofl‘ee for sale than their neighbors with more land. The patterns do not, however, go very
far in explaining whether land scarcity actually pushes households to buy more food (per
hectare) and sell more cash crops (per hectare). Thus, based on the patterns only, man
bersaid on whether increasing land scarcity brought about by demographic change tends

 

 

 

to increase the supply of cash crops to the market. To be able to Wfi'fisqfisfion,
we next study the association between land scarcity and net sales by controlling for a
number of other factors. Like above, we argue that relationships are likely to be quite

difi‘e‘rent in the high-altitude areas, and therefore focus on Rwanda’s banana zone.

6.3.1 General Model, Explanatory Variables, and Hypotheses
The approach used here is analogous to the one used above to model land use.

Decisions on market participation and those on land use are obviously closely associated.
In a cross-sectional analysis they are determined simultaneously. Both market conditions
that afl‘ect households through market tansactions and on-farm conditions that affect
households through the agricultural input-output relationships influence land use and
market participation.

126

Thus, we see market participation as a function of market incentives, physical
conditions, land scarcity, and other household characteristics. Market participation is
defined as net sales in kilograms per hectare.

Land scarcity or “man/land ratio” is measured in terms of adult equivalents per
cultivable hectares. As discussed above, no unequivocal hypoheses regarding the impact
of land scarcity on market participation can be made on the basis of economic reasoning.
Instead, theory suggests that the association between net sales and land scarcity is
determined by the interplay of two objectives. On the one hand, land scarcity pushes
households to grow and sell crops that give high returns to land and to purchase those that
give only low returns to the scarce resource. Based on the characteristics of the major
crops in Rwanda (Table 5-1), this would imply more specialization in (beer) bananas and
sweet potatoes and less in coffee. 0n the other hand, the existence of transactions and
transportation costs and the associated failures in the markets for food and other products
may make households subsistence-oriented. Where this prevails, cash crop sales should
be closely associated with farm size, since only those with some land left after
subsistence needs are met, would produce cash crops for sale. However, also food crop
sales would decline with farm size, since large farmers would be able to have a margin of
safety and would produce marketable surpluses in most years, whereas small farmers
could have deficits despite their single-minded focus on food crops.

In Rwanda, both arguments suggest that cofi'ee sales per unit of land should
decline with farm size. Cofl‘ee is neither a subsistence food crop nor provides high returns
' to land (at the prices that prevailed in 1990). Therefore, we expect that coffee sales per
1 unit of land decline with farm size. This hypothesis implies that the apparent association
1‘“ between land scarcity and coffee sales per hectare that was seen in the cross-tabulations
\ above should not hold under multivariate analysis.

l
“.1

127

 

_..———
_ __..4u..._--—-~—

For bananas the implications are less clear. If the costs and uncertainties of relying
on the market are indeed high enough to justify a strategy that gives priority to
subsistence food crops, bananas could decline as land gets scarce. On the other hand, if
market transactions are not prohibitively expensive, one could expect that land-scarce
farmers would rely more on bananas, which, when turned into beer, provide high returns
on their scarce land.

Sincenone ofthe foodcropscanmatchthehighretmnstolandprovidedby
bananas, our hypothesis is that the net sales ofall food crops per unit ofland should
decline as land gets more scarce. The decline can be expected to be stronger for cassava,
cereals, and beans, which are more transportable and provide lower returns to land than

sweet potatoes.

Market incentives are again problematic, since, as discussed in the previous
chapter, our measures of agroclimatic conditions are so imperfect that if prices were used
to measure market incentives, they would actually capture much of the variation
associated with agroclimatic variation. In other words, high prices for a given crop would
be associated with lower sales—not because farmers do not behave economically but
because high prices are associated with poor growing conditions. Thus, we only use
distance to paved road as a proxy for market access, hypothesizing that households with
good market access rely more on the sales of profitable but bulky cash crops, particularly
bananas, and also rely more on the purchases of food crops that are not very lucrative in
terms of returns to land and labor. For cofl‘ee, the hypothesis is that proximity to roads
should reduce farmers’ reliance on coffee, which is less profitable and more transportable
than banana.

Physical conditions are again measured by rainfall, altitude, and land quality
index, and are used primarily to control for other factors. The proxies, especially rainfall

128

and altitude, are highly collinear, and the coefficients themselves have little relevance.
The general hypothesis is that the net sales of each crop are higher in those conditions
that are more suitable for the crop. For instance, net sales of bananas are expected to be

lower at higher altitudes.

Household characteristics used in the model are the age of household head and
ofi-farm incomes (which are highly correlated with education). In accordance with the
hypotheses of the previous chapter, we could hypothesize that households with older
heads as well as those with higher off-farm incomes produce and sell more bananas and
cofi‘ee. However, households with large off-farm incomes can afl'ord to consume more,
which should reduce the net sales of most products, especially those of beer, which has a
high income elasticity of demand (Ansoanuur 1991).

6.3.2 Regression Results on Crop Transactions
The results of the OLS regressions on net sales per hectare are presented in Table

6—16. The key results concerning the association between land scarcity and net sales
controlled for other factors are found on the first line of the table.

The numbers confirm our hypothesis concerning cash crops. As expected, the
coefficient for beer bananas is positive and that for cofi‘ee is negative, suggesting that,
other things being equal, increasing land scarcity brought about by population growth is
associated with expanding market supply of banana beer but declining market supply of

coffee.

129

Table 6-16: Net Household Sales Model

Beer Cooking Sweet Cassava Beans Sorghum Maize Coffee
Banana Banana Potato

Man/land,farm .17** -.26** -.48** -.50** -.73** -.10** -.18** -.15**
Man/1and,vill .08* -.08* .09* -.12** -.04 .00 -.07* .07
Dist. to road -.09* -.04 -.03 .05 .03 -.01 .03 .02
Altitude -.14** .03 .04 .07 -.12** -.OS -.16** -.16**
Rainfall -.15** .01 -.02 .00 .14** -.14** .10* -.04
Land quality .03 .08* .02 .10"r .03 .04 .14" .09*
Age of head .04 -.01 -.05 .01 .03 -.06 .06 .00
Off-farm inc. -.09* -.03 .07 -.13** .19** -.06 .05 -.02
Bast .02 .01 - 04 06 .12** 14** -.05 -.06
Congo-Nile .03 .05 -.05 .01 -.01 OS - 08* -.02
Volcanic .14** -.01 -.10* .00 .03 .12 -.05 -.04
Lake Kivu -.05 .11 - 03 06 -.03 05 - 30** 31.1.

ADJ. R2 .12 .07 .17 .40 .56 .10 .17 .17

Source: DSA/MINAGRI, Agricultural Survey 1991; 960 households
* - significant at St or better, ** - significant at 1% or better

 

The result for coffee suggests that the positive correlation between land scarcity
and coffee sales shown in Table 6-15 is spurious, reflecting the impact of intervening
variables and not specialization in coffee as a response to population pressure. Table 6-16
suggests that agroclimatic conditions might be the explanation. Not only are cofi‘ee sales
associated with land quality but they are also much more important on the shores of lake
Kivu, which is characterized by small farms and good soils. Moreover, coffee sales have
a weak positive association with the man-land ratio of the village, which also correlates
with high land quality. But within villages, small farmers sell less cofi‘ee per hectare than
their neighbors with more land.

The result that land scarcity is positively correlated with beer banana sales (as
such or as beer) but negatively with cooking banana sales, is consistent with the fact that
beer bananas require much more work per unit of land than cooking bananas when
brewing labor is also considered. While the agronomic suitability of Eastern Rwanda for

130

cooking bananas is clearly one reason why farmers there rely so much on cooking
bananas, our findings suggest that large farm size in the East is another explanation.
Within villages large farmers rely more on cooking banana sales and less on beer banana
sales than do their smaller neighbors.

The net sales of all the main food crops decline with farm size. The longitudinal
implication of this cross-sectional result is that as population grows and farms get
smaller, the market demand for all food crops will increase.

Unlike the coefficient for the man-land ratio of the farm itself, the coefiicient for
the man-land ratio of the village is significantly positive in the case of sweet potato.
Together these two numbers suggest that in densely inhabited villages many large farmers
are net sellers of sweet potatoes. As demographic pressure increases, the market demand
for sweet potatoes, this crop is gaining ground as a cash crop. None of the other major
food crops is doing so, which is consistent with the fact that sweet potatoes provide

higher returns to land than any other food crops.”

Market access, represented in the model by distance to paved road, had a
significant coefiicient only in the case of beer bananas. As expected, farmers sold more
beer bananas in areas close to paved roads. The coefficients for cassava, beans, maize,
coffee, and cooking bananas had the hypothesized signs, but were not statistically
significant at the 5 percent level.

As expected, high off-farm incomes were associated with small sales of beer
bananas. The same was true for cassava. The net sales of beans, in contrast, were found to
increase with off-farm incomes. Since the net sales of beans are typically negative, this
means that those with above-average ofi-farm incomes bought below-average quantities

 

3’ In the potato zone on Rwanda’s temperate. highlands, white potatoes already are a
major cash crop. With few exceptions, white potatoes are a rumor, low-yield
subsistence crop in the banana zone.

131

of beans. This is an anomalous result with no clear explanation. The fact that employees
are sometimes fed by their employers may be a partial explanation. Another partial
explanation is the fact that some of the households with large off-farm incomes bought
meat, eggs or fish instead of beans.

In sum, the regression results presented in this section show that in Rwanda’s
banana zone households with less land per adult equivalent sell more bananas per hectare
and buy more food per hectare. This suggests that as population grows and farms get
smaller, the practice of growing bananas and exchanging banana beer for food is
becoming an increasingly important part of the food security strategies of the poor.

6.4 Conclusions
The tabulations presented in this chapter show that while bananas and cofi‘ee are

the main cash crops for all farm size categories in Rwanda, food crops are sold in
significant quantities only by relatively large farmers. Throughout the country large
farmers achieve and exceed self-sufficiency in sweet potatoes, maize, cassava, and
cooking bananas, and in the sparsely inhabited eastern lowlands they also succeed in
producing marketable surpluses of beans. Small farmers, in contrast, often need to
purchase food to supplement their own production. In the case of beans their own
production only accounts for halfof the consumption. Many small farmers could produce
more food by giving up cash crops but have instead opted for a partial specialization in
cash crops.

Poorest farmers are more subsistence-oriented than smallest farmers. They sell
little banana beer and coffee, produce primarily cheap root crops, purchase only minimal
amounts of beans and have extremely low consumption levels. While this may in part

show the dead end of the subsistence farming under demographic pressure, a more
N— --—

 

important explanation is agroclimatic. Subsistence-orientation and poverty are linked
W I

132

largely because soil degradation contributes to both. This is associated with the fact that
in Rwanda, the principal cash crops (bananas and coffee) are more demanding than some

of the main food crops (sweet potatoes and cassava).

Multivariate analysis that controls for agroclimatic and other factors shows that in
Rwanda’s banana zone, households with less land per adult equivalent sell more beer

M'

W
bananasperh .Inthisrespectb are 'fferent anyoW“

including cofi‘ee. This suggests that Rwandan farmers are not subsfitutinngQd cropsform ,

W

cash ficrops to cope with land scarcity caused by population growth. While households

w,.....—~ NH"

with less and less land per person are not making any dramatic transition to cash crops
either, they clearly prefer to keep their bananas and to rely w of banana

beer for food. “Food first” is not their strate of achi ' food securi .

 

133

Chapter 7

BANANAS AND Poor) SECURITY

7.1 Introduction
In the previous two chapters we have seen that bananas are central in Rwanda’s

agriculture. In most of the country it is not an exaggeration to call the farming system
banana-based. The relative importance of bananas has increased during the past decades
and the cross-sectional evidence we have presented above shows that, controlling for
other factors, bananas seem to be associated with land scarcity. This suggests that the
relative role of bananas should increase firrther with population growth. However, we
have also shown that for many households bananas are largely a cash crop, implying that
the expansion in supply associated with population growth is contingent on changes in

market demand.

In this chapter we take a more detailed look at the contribution of bananas for
food security in rural Rwanda. We examine why so many food insecure households grow
bananas for beer rather than food for themselves and discuss the factors that afl‘ect the
future role of bananas in Rwanda’s food security equation. To put the survey results in
perspective, we start by reviewing literature on bananas and banana-based farming
systems in the highlands ofcenfial and eastern Afiica.

7.2 Background Information on Bananas

7.2.1 General Characteristics of Bananas
Banana is one of the most important food crops of the world, and it is grown

throughout the tropics (Stover and Simmonds 1987). Botanically, banana (genus Musa) is
not a tree, but a gigantic herb. The plant springs fi'om an underground stem, forming a

134

false trunk. Each trunk produces only one bunch of fi'uit, and is then cut down. The stem
can live for decades.

Most bananas are starchy when green and sweet at matmity (Moan 1985). Most
bananas grown for export and eaten uncooked when ripe belong to the AAA triploid
group, whereas most bananas that are cooked when green and that are typically known as
plantains belong to the AAB triploid group (Perrault 1978). Nevertheless, the distinction
between sweet bananas and plantains does not always correspond to botanical
classification.

In Africa, plantains belonging to the AAB triploid group dominate West Afiican
lowlands and the Congo drainage basin in central Afiica. Plantains are mostly cooked for
food, and provide very low yields (about 2 tons per hectare per year), but do not require
much labor either (Perrault 197 8). In the highlands of Central and East Afiica, varieties of
the AAA triploid group predominate, most of them belonging to a subgroup that is found
nowhere else (Perrault 1978). In Rwanda, 90 percent of bananas were estimated to be
AAA triploids (Delepierre 1970). Yields are considerably higher than for plantains in
Zairian lowlands, usually exceeding ten tons per hectare per year. In Rwanda and
Burundi, bananas are mostly grown for beer; in Uganda and Tanzania mostly for food
(Perrault 197 8). '

Bananas grow best on deep soils with abundant rainfall and temperature that
varies between 25 and 30 centigrades. However, hundreds of varieties exist, and as a
group, bananas can grow satisfactorily on a wide variety of soils, as long as the climate is
suficiently warm and not subject to prolonged dry spells (Champion 1963, Stover and
Simmonds 1987). In much of Central and East Afiica, mid-altitude highlands provide the
best conditions, since the sun'ounding lowlands are often too dry for bananas, whereas the
high mountains are too cold (Stover and Simmonds 1987). Although the absolute altitude

135

limit due to low temperature is quite high, bananas grow so slowly in cool areas and have

so low sugar contents that few bananas are grown above 2000 or 2100 meters.

Bananas thrive on soils that are high in organic matter, but at least on the more
fertile soils banana groves can efi'ectively mulch themselves and gradually build up the
contents of organic matter (Perrault 1978, Kajumulo-Tibaijuka 1984). Neither fallowing
nor nutrients imported fiom elsewhere in the form of chemical or organic fertilizer are
needed to maintain fertility in subsistence—oriented banana groves, although both can be
used to increase yields on poor soils, or to replace nutrients exported in the form of
bananas or banana leaves.

Nutrient uptakes by bananas are modest and little leaching of nutrients or
volatilization of nitrogen occurs in well-mulched banana groves (Champion 1963).
Compared to cereals, bananas have particularly low uptakes of phosphorus. Fertilizers
designed to be used on cereals are usually not suitable for bananas.

On small farms, pests such as weevils and nematodes and diseases such as black
Sigatoka (black leaf streak) and F usarium vilt (Panama disease) typically keep average
banana yields at a fraction of the genetic potential. Until the 19505, this was true also on
most export-oriented banana plantations, but during the subsequent 30 years average
yields on plantations almost tripled to 40-60 tons per hectare per year (Stover and
Simmonds 1987). Improved pest and disease control along with breeding for higher
yields were the keys to the spectacular yield increase. The impact of these advances on
semi-subsistence production has been relatively modest, both because pest control has
relied heavily on fi'equent applications of costly pesticides and because banana research
and extension has mostly focused on high-input production on commercial plantations.
Among smallholders the adoption of improved varieties and external inputs is sporadic.
Combined with the rapid spread of banana diseases and pests over the past decades, this

136

means that average yields in smallholder conditions have stagnated or decreased in many
parts of the world. This is in great contrast to the broadly shared benefits of technical
progress in wheat, rice, and maize over the past decades in most of the developing world.

Bananas are often intercropped with coffee, timber trees, and food crops such as
beans, maize, or sweet potatoes. The relative densities of the crops vary widely and
cultural practices range from deep hoeing to zero-tillage. Typically none of the associated
crops yields as well as it would on pure stands, but the total production per hectare is
higher (and so are the ”standardized" yields computed by allocating the intercropped
fields among the associated crops). Occasionally intercropping is a temporary phase
during the establishment period of perennials. In Papua New Guinea highlands, for
instance, mixed vegetable gardens are gradually converted into cofl'ee/banana gardens
and eventually into coffee/casuarina stands, and the point of intercropping is to minimize
the need for weeding (Bourke 1989). In eastern Kivu, close to Rwanda's western border,
intercropped banana groves became pure as they matured (Perrault 1978). More often,
however, intercropping is a permanent feature of banana groves. In much of eastern and
central Afiica, albeit not in Rwanda, coffee is the main crop associated with bananas. At
least at lower altitudes and warmer temperatures, cofl‘ee bushes benefit from the shade
and the mulch provided by banana trees.

The uses of bananas are not limited to beer brewing and direct consumption as
such, although these typically dominate. Numerous processed food products (flour,
flakes, chips, puree, juice, vinegar), as well as non-foods ranging from glue to cloth, can
be made out of various parts of banana plants (Stover and Simmonds 1987). Some of the
processing techniques are cheap and simple and are used by small-scale operators,
particularly in India, Sri Lanka, and the Philippines. Also in Tanzania the practice of
drying bananas and pounding them to flour has a long history (Koponen 1988).

137

As such, bananas are bulky and perishable products, which greatly reduces their
value for food security and makes their potential as cash crops crucially dependent on
access to roads and markets (Perrault 1978). Processing can often solve one or both of
these problems. Drying increases both transportability (vaiue-to-weight ratio) and
storability; brewing only the former, since traditional banana beer is highly perishable.

7.2.2 Bananas on East African Highlands
Ineasternandcentral Afiicabananashavebeenoneofthemainstaples sincethe

precoloniai period. In precoloniai Tanzania, for instance, ”grain cultivation with a fallow
period was practiced in the lowlands and plateaus with moderate rainfall whereas banana
was predominantly the crop of well-watered volcanic highlands and other pockets of
exceptional soil fertility”5 (Koponen 1988). Some of the banana-based systems supported
very high population densities, maintained fertility through manuring and mulch, and
practiced irrigation and soil conservation. Bananas were used in many forms (fi'esh,
cooked, dried and pounded to flour, mashed for brewing, etc.).

Quantitatively, the historical role of bananas is difficult to estimate. One German
settler claimed early this century that three-quarters of Afiican crops consisted of bananas
(Koponen 1988), but this may refer to quite a limited geographic area in Tanzania and it
may also underestimate crops intercropped with bananas.

In the mountains of Usambara, banana lost its role as the chief staple after 1900,
primarily to maize and secondarily to cassava (Fleuret and Fleuret 1980; Koponen 1988).

The reasons for the shift are unclear, as is the extent to which such a transition occm'red

 

3‘ It is likely that many of these fertile “pockets” were results of f ' stems that
concentrated nutrients and organrc matter in the banana ves aroun e com unds
(Perrault 1978, Kajumulo-Tibaijuka 1984, Olson 19‘96‘2. e contrast between e
banana growing “in-fields” and the annually cultiva or grazed “out-fields” was often
great both because nutrients and organic matter were brou t from out-fields for
consumption and recycling in and around compounds, an because out-fields were
typically more exposed to sun, fire, and water, and hence had higher levels of

138

elsewhere. Nevertheless, banana continues to be an important crop, often the main staple,
in the well-watered highlands of the region.

The Chagga (or Wachagga) homegardens on the slopes of Mount Kilimanjaro are
one of the present-day banana-based systems (Ruthenberg 1964; Femandes, O'Kting'ati
and Maghembe 1989). They support a population densitythat exceeds 500 persons per
square kilometer, and are regarded as a relatively sustainable system. The homegardens
are dominated by bananas and intercropped with coffee and beans. Most households also
have a lowland plot, mostly allocated to maize and beans. Most farmers keep
permanently stabled cattle and goats, mostly fed with leaves and stalks from bananas,
supplemented with hay and grass brought from the drier lowlands. In the early 19608,
Ruthenberg (1964) saw the system as improving; yields were increasing as farmers were
starting to use fertilizer and pesticides and adopting "rational" milk production practices.
By the late 1980s, little progress had occurred. Femandes, O'Kting'ati and Maghembe
(1989) reported that little fertilizer was used and included fertilizer and intensified milk
production among their recommended improvement options. They pointed out that the
system had a low productivity and was failing to deal with population growth, which had
resulted in considerable outrnigration. Nevertheless, they suggested that the gardens were
relatively sustainable and productive, and therefore worth extending elsewhere, including
the uplands of South-West Rwanda. ,

Another well-documented banana-based system is that of the Bahaya in Bukoba,
Kagera, the region of Tanzania that borders Rwanda. The system relies on intensive
manuring and mulching and is credited not only for arresting erosion but also for building
up fertility on soils that initially were badly leached (Friedrich 1968; Ruthenberg 1976).
However, the poor grass lands that each farm has in addition to its fertile banana groves

 

decomposition, volatilization of nitrogen, leaching of nutrients to sub-soil, and soil
erosron. .

. 139

may not offer a valid comparison, since the system is partly based on the transfer of

nutrients and organic matter from the poor pastures to the intensively cropped infields.

In the early 1960s, farmers produced primarily cooking bananas for home
consumption, and secondarily robusta coffee and cooking bananas for sale. Perennials
were intercropped with beans and maize, which were grown with zero-tillage (Friedrich
1968). Returns to land and labor were relatively high compared to many other African
farming systems (Ruthenberg 1976).

Twenty years later, the farming systems around Bukoba were reported to be under
severe stress (Kajumulo-Tibaijuka 1984). In the 19703, cooking bananas had been
attacked by weevils and nematodes, causing yield declines of 20-95 percent. Also beans,
the main source of proteins, had been under severe pest attacks. Farmers had responded
by increasing root crop production in banana groves, or by replacing cooking bananas
altogether by cassava and sweet potatoes. They also had planted more beer bananas that
are higher-yielding and less susceptible to the pests. Beer bananas were replacing both
cooking bananas and coffee as sources of cash, and beer had also become a much more
important part of subsistence consumption. Some farmers even had migrated to more
remote parts of the region to avoid the banana pests. Kajumulo-Tibaijuka (1984)
concluded that coffee production will not recover unless the pest problem of cooking
bananas can be solved and suggested that bananas should be ”number one priority crop"

inthearea.

The problems faced by banana growers in Bukoba apparently reflect the situation
in many other parts of the East Afiican highlands as well: Due to pests and diseases, and,
in some cases, due to the impoverishment of the land, yields of bananas, in particular
those of cooking bananas, are stagnating or declining. In the most severely affected areas

farmers are substituting roots and cereals for bananas. At the same, prices relative to other

140

crops are increasing. In response to the high prices and to pOpulation growth, farmers in
less afl‘ected (and typically less fertile) areas are planting more bananas. In Uganda,
agricultural statistics suggest that the area under cooking bananas (matoke) increased by
some 50 percent fiom the early 19708 to the early 19903, whereas production barely
increased (World Bank 1993: Uganda: Agriculture; FAO/WF P 1997). Although the
shorter-term fluctuations in yields are dominated by the collapse due to civil strife and
recovery thereafter, the long-term picture is one of declining productivity and yields.

7.2.3 Role of Bananas in Rwanda’s Farming System
Bananas have a long history in Rwanda, but it is not clear how important they

were in the past. Jones and Egli (1984) report that before 1900 most beer was made out of
sorghum, not out of bananas. Bart (1993) claims that bananas dominated farming already
at the turn of the century, albeit only in the central region. As permanent cultivation later
expanded to valley bottoms and swamps, towards the Congo-Nile Divide, and towards
the dryer Eastern savannas formerly used for pasture, bananas conquered new territory.

In 1966-67, bananas were estimated to cover some 130,000 hectares (Nyabyenda
1988; Bart 1993). Delepierre (1970) reports that during the 1960s the area planted to
bananas increased by 7.6 percent per year, more than three times as fast as the population.
He attributed the increase to the superior profitability of beer bananas compared to other

craps.

During the 19703, the area under bananas continued to expand (Gotanegre 1983).
Bart reports an estimate of 249,000 hectares for 1982, and Nyabuenda an estimate of
253,000 hectares for 1985. Charlery de la Masseliere (1993) sees the expansion of

bananas to as high as 2,200 meters altitude in places as a sign of strong pressure on
traditionally settled lands at lower elevations.

141

On most Rwandan farms, bananas are the main crop of the intensively cropped in-
fields around the compound. Virtually all manure and household waste are used on
bananas, which respond well to organic matter (Raquet 1991; Kotschi et al. 1991). In
recent years, also the vulnerability of banana groves to thefts has encouraged farmers to
plant bananas on their closest fields that are easiest to supervise (Bart 1993).

Jones and Egli (1984) claim that "banana groves tend to nest in small folds in
hillsides down to which soil nutrients have been washed fiom the annual-cropped soils
above. Thus banana groves tend to occupy the more fertile and less steep hillside land."
The claim is not supported by DSA data, except for the prefecture of Gisenyi, where
many of the hilltops and upper slopes are too high and cool for bananas. Elsewhere,
bananas play their biggest role on hilltops, followed by upper slopes. In the valleys, the
share of land under bananas is much smaller.

7.3 Bananas for Food
Since most bananas produced in Rwanda are beer bananas and are typically

consumed as beer, the role of bananas in human nutrition depends crucially on how much
nutrients are lost in the conversion to beer. Unfortunately, the estimates found in the
literature vary substantially. Perrault (1978: 336) gives the highest estimate, claiming that
90 percent of the calories are lost in the brewing process. According to Loveridge (1989:
226), who refers to MINAGRI (1985) “a kilo of beer bananas provide only 17 percent of
the calories that a human obtains from a kilo of cooking bananas because calories are lost
in the brewing process.” Jones and Egli (1984: 67) suggest that the alleged high losses are
exaggerations associated with the oflicial anti-beer policies that consider beer drinking
anti-social and beer brewing wasteful. Referring to FAO statistics, they claim that “every
kilogram of bananas consumed directly furnishes 751 calories,” whereas “every kilo of
banana fermented into beer furnishes 438 calories.” This implies that only 42 percent of
calories would be lost in the conversion to beer. In its annual statistical report, DSA uses

142

the estimates of 802 kcal/kg, 196 kcal/kg, and 523 kcal/kg for cooking bananas, beer
bananas (after brewing) and fruit bananas, respectively. The conversion factors are taken
from MINAGRI (1988) and suggest that ifthe beer bananas consumed as such were
comparable to cooking bananas, the implied loss would be 75 percent, and if they were
more like fi'uit bananas, it would be 62 percent.”

While Jones and Egli may be right in that the highest estimates of the
fermentation losses are biased for political reasons, much of the wide variation, from 42
m90pementinesfimawdlossespmbablynsulmfiomthefaathatesfimatesambased
on small samples of heterogenous goods. Neither beer nor bananas are products of
standard quality. Recipes for making beer are known to vary, and the highest loss
estimates probably come from computations in which the caloric content of “more
diluted” beer is divided by the amount of bananas used for “less diluted” beer. Moreover,
also small amounts of cereals (mostly sorghum) are used for making banana beer. The
studies may have differed in how they divided the caloric losses between cereals and
bananas.

In the tables below, we present estimates on the (potential) availability of calories.
For each crop, availability is defined as own production plus purchases plus gifis received
minus sales (either as such or as beer) minus gifts given. Beer bananas consumed as such
are assumed to be comparable to fruit bananas in their caloric content. No adjustment is
made for calories lost in brewing for home consumption.” One argument for this
approach is that the destruction of calories by brewing is the producers’ choice and if
foodisunusually scarce, farmers may eatbananas insteadofbrewingthem forhome

 

3’ While some of the calories are indeed lost in the brewing proce others are included in
the by-products of brewing. A patently none of the estimates es any attempts to
account for how these by-p ucts are used. To the extent that the byproducts are used
for fodder and contribute to the avarlabrlrty of animal products, the estimates overstate

asses.

143

consumption.39 Thus, while the numbers below do not show the best estimate for calories
actually consumed, they might tell more about food security. Moreover, given the huge
variation in the estimated brewing losses, it may be more informative to give the numbers
on availability and then establish what the various estimates On brewing losses would
suggest for consumption. A third argument is that the numbers for all the other crops are
estimates of availability rather than consumption. Sorghum, finger millet, and wheat are
used partly for beer, too. In remote areas where transportation costs make bulky products
non-tradable, occasional surpluses of cooking bananas and sweet potatoes are probably
fed to animals.‘0 No quantitative information on the caloric losses caused by these factors
is available. Thus, it can be argued, the only way to be consistent is to focus on

availability and not on consumption.

 

3' Calories lost in brewin for sale is accounted for by the conversion of beer sales into
banana and sorghum es.

3’ Brewing for sale is a different story, since, depending on prices, brewin and
exchanging beer for food may be exact] what food scarcrty calls for. erefore,
lclaloriirls 13st in brewing for sale cannot assumed to be potentially available for the

ouse o .

‘° In Kagera, on Tanzania’s side of Rwanda’s eastern border, farmers were found to seek
food security for soor years b producing more bananas than they needed in normal
years (Kajumulo- ibarjuka l 84). Consequently, bananas had almost no market value
m§ood years and in remote areas farmers at trmes threw awag'alarge quantities. Also
D A’s numbers show that many farmers in Eastern Rwanda ve much more coo '
bananas available than they possibly can eat. In some cases this robably results from
data errors (underreported sales or overreported production or om unrecorded
consurzrcption by relatives or employees, in others it probab y indicates that surpluses
were f to animals.

144

Table 7-1: Caloric Availability by Crop and Income Quartile

 

1000 kcal/hh- Income quartile (Paw/AB) All Farms
<5350 5350- 8020- >13230
8020 13230
Beans 325 481 675 924 598
Peas 17 25 30 46 29
Peanuts 2 6 13 41 15
Soybeans 30 57 54 96 59
Sorghum 148 194 320 645 323
Maize 115 176 254 363 225
Wheat 8 7 18 15 12
Bleusine 3 3 3 3 3
Rice 3 9 13 21 11
Cassava 130 210 277 320 233
lhite Potato 32 50 94 153 81
Sweet Potato 539 735 739 714 682
Taro 28 48 60 . 65 50
Cocoyam 1 2 3 4 2
Cooking Banana 100 199 359 799 359
Other Banana 193 372 629 1144 578
Total 1672 2574 3542 5352 3261

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

Table 7-1 shows mean food energy availabilities fiom each crop by income
quartile in kilocalories and Table 7-2 indicates what is the average share of each crop of
the total amount of calories available from the 16 main crops. According to a nationwide
consumption survey in the mid-19803, these crops accounted for more than 97 percent of
all calories consumed in rural Rwanda. Fruit bananas, which account for about one-tenth
of the volume of all bananas and are used both for food and for beer, are combined in the
tables-with beer bananas.

145

Table 7-2 shows that if bananas (and other crops) available were consumed as
such, they could provide almost one-quarter of all the calories for an average household."
In practice, most beer and some fruit bananas are brewed, which probably reduces the
average share of bananas in caloric consumption to somewhere between 17 and 20
percent.‘2 Although traditional brewing probably destroys more than one-tenth of all
calories available to rural Rwandans", which is considerable waste in a country where
many people go hungry, bananas unequivocally are one of the three comer-stones of
Rwanda’s food supply. The old assumption that since most bananas are turned into beer,
bananas do not provide many calories, is clearly unfounded.

 

" Note that the numbers in Table 7-2 are means of household-levelagercentafi, not
percentages computed from household-level means reported in T 1e 5-1. ey tell
about average households without giving more weight to households with higher
availability. Since households with higher availabilrtg get a higher share of it from
bananas, e centfies (r9445) reported in Table 7- or bananas are smaller than

those (11+18 im ii in able 7-1. An ave e household ets 24 percent ofits caloric

availability from but 29 percent of ories availab e for Rwandan households
come fiom bananas.

“Assuming that 62 percent of calories were lost in brewing and that 85 percent of “other
bananas’ were brewed, the production of banana beer on average destroyed 305 Meal
(thousands of kilocalories) per household, which reduces total availability by 9.3
percent. Guessing that the brewin of sor um beer destroyed one-third of calories
available fiom sor um, the total osses m traditional brewin would reach 412 Mcal
or 12.6 percent 0 total availability. The reduced estimate of ories available fiom
bananas would account for 22 percent of the reduced total. As explained in the previous
footnote, this computation gives more weight to households with higher caloric
availability. The average computed without such weighting is 18.4 percent.

‘3 See the com utations in previous footnotes: Note that this only applies to calories that
were availab e to rural households, and avarlabrlrty rs defined to excludes products that
were sold either as such or as beer. If we add the caloric losses that appltiutlo marketed
beer and use the assumptions of the previous footnote, we can estimate t brewing
destro 5 about 15 percent of the calorres produced in Rwanda. In contrast, Jones and
Egli ( 984 estimated the total losses from banana brewing to be only 2.75 percent on
the Great es Highlands as a whole and onl slightly more in Rwanda. In addition to
using a very low estimate for the losses per ki ogram, they grossly underestimated the
relatrve importance of bananas in the farming systems, especially in the East (“Kagera
Piedmont” rn therr termrnology). .

146

Table 7-2: Shares of Caloric Availability by Crop and Income Quartile

Percent Income quartile (FRw/AB) All

Finger Millet

Rice

Cassava

White Potato

Sweet Potato 3
Taro

Cocoyam

Cooking Banana 5 7 10 14 9
Other Banana 11 14 17 21 15
Total 100 100 100 100 100

0
1
0
2
9
7
Wheat 0
0
0
8
2
3
2
0

O

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

 

Nevertheless, for the poorest households which typically are among the most
food-insecure, the role of bananas as direct suppliers of calories is much smaller than it is
for the largest households. In the poorest quartile, only 16 percent of caloric availability
comes fi'om bananas. After adjusting for brewing losses, perhaps only 12 percent of the
actual consumption of dietary energy comes from bananas. Sweet potatoes provide
Most three times as many calories for the poorest quartile as bananas and beans almost
twice as many. In the diets of the very poorest households in rural Rwanda, bananas are

clearly secondary to sweet potatoes and beans.

147

As Table 7-3 shows, leaving out the areas that are too cool or otherwise unsuitable
for bananas does not change the picture very much. Even in Rwanda’s banana zone,
bananas are twice as important in the diets of the richest quartile as in the diets of the
poorest quartile.

As the multivariate analyses in the previous two chapters showed, the negative
association between poverty and bananas does not mean that poverty forces households to
grow and eat something else instead. Instead, the negative association results largely from
agroclimatic and historical factors. Rwanda’s banana basket in the East is relatively
sparsely inhabited and prosperous whereas, on the watersheds of Congo-Nile Divide,
most crops grow poorly and bananas almost not at all, leaving people in extreme poverty.
In other words, the poor consume relatively few bananas in part because many of them

live in areas where bananas do not grow well.

148

Table 7-3: Shares of Caloric Availability in Banana Zone

Percent Income quartile (FRw/AE) All

0

1
Peanuts 0
Soybeans 2
Sorghum 9
Maize 4
Wheat 0
Finger Millet 0
Rice 0
Cassava 8
White Potato 1
Sweet Potato 33
Taro 2
Cocoyam 0
Cooking Banana 6
Other Banana 13 15 19 23 18
Total 100 100 100 100 100

Source:DSA/MINAGRI, Agricultural Survey 1990; 955 households

There are, however, three reasons to believe that the direct consumption of
bananas contributes to food security more than the estimated shares in consumption
suggest. First, food security depends not only on average contributions but also on their
variation. In Rwanda, bananas tend to produce something even when other crops fail
completely.“ For instance, several farmers we met in January 1994 told us that their
beans had died because of the drought, but their beer bananas kept on producing, albeit
less than in normal years. Moreover, bananas, along with roots, are nonseasonal and

 

“ Although bananas are vulnerable to droughts in the sense that they cannot tolerate long
dry seasons, they are not very sensitive to the lesser irregularities that uently cause
havoc to annual crops in countries like Rwanda (Stover and Simmonds 1 84).

149

produce also during the nutritionally critical preharvest season. Second, the poorest
households probably eat more of their available beer bananas than their more fortunate
neighbors and, more importantly, can in poor years convert what they normally drink into
food by brewing and exchanging, or, if beer prices collapse and food prices rise, by eating
the beer bananas." Third, as an extreme coping mechanism during famines, farmers can
uproot their bananas, dry and pound the roots and eat them the same way as dried cassava
is eaten. Some 60 years ago, starving Rwandans even preferred dried banana roots to
dried cassava imported by Belgian colonial administrators fiom Congo to prevent
starvation (Bart 1993). Although cassava is now preferred, banana roots were eaten
during the 1989 drought in southern Rwanda (Pottier 1993).

7.4 Bananas for Cash
In addition to supplying food for direct consumption, bananas contribute to food

security as a cash crop that can be exchanged for food. In this section, we focus on the
cash incomes rural Rwandans get fiom the banana subsector and fiom other sources.

 

" As discussed below, in normal years eating beer bananas instead ofbrewing them into
beer and exchanging for food is unlikely to improve food entitlements.

150

Table 74: Household Cash Incomes by Source and Income Quartile

 

Rwandan Francs Income quartile (FRw/AB) .All Farms
<5370 5370- 8040- >13120
8040 13120

net farm sales 3,675 8,651 14,783 23,006 12,410
Coffee 622 1,527 3,302 4,513 2,468
Banana (+ beer) 1,170 2,958 4,383 7,005 3,845
Cooking ban. 25 65 142 796 250
Sorghum (+ beer) 515 728 1,192 1,852 1,062
Other crops 632 1,599 2,971 5,446 2,629
Livestock 736 1,840 2,934 4,191 2,406
Labor sales 3,097 4,609 6,551 31,118 11,088
Agricultural 1,659 1,891 1,794 1,308 1,669
Other unskilled 602 873 1,453 1,815 1,179
Skilled 836 1,845 3,303 27,996 8,241
Cash Gifts 179 207 360 784 377
Other cash income 401 222 705 2,696 983
NET CASH INCOME 7,352 13,689 22,399 57,604 24,859
Total net income 18,847 33,066 50,223 99,971 49,930

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

Table 7-4 shows the mean cash income from each source by income quartile and
Table 7-5 shows the share of net cash income that households on average get from each
source. In all income quartiles, bananas and banana beer bring households more cash than
coffee, which has typically been considered the “predominant source of cash” in rural
Rwanda (Jones and Egli 1984: 33). Moreover, the numbers are net of input costs and the
imputed value ofcereals sold in the form ofbanana beer is reported in the table with
sorghum and sorghum beer. Rural Rwanda as a whole earns approximately as much cash
from traditional beers and their ingredients as from all the other crops combined. Overall,

151

Table 7-5: Household Cash Income Shares by Income Quartile

 

Percent Income quartile (FRw/AE) All Farms
<5370 5370- 8040- >13120
8040 13120
was farm sales 55 64 70 64 63
Coffee 10 10 13 12 11
Banana (beer) 18 24 23 20 21
Cooking ban. 0 1 1 2 1
Sorghum (beer) 7 6 6 5 6
Other crops 11 12 15 15 13
Livestock 9 13 13 10 11
Labor sales 36 31 25 29 30
Agricultural 20 15 10 4 12
Other unskilled 7 5 6 4 5
Skilled 9 11 10 21 13
Cash Gifts 4 2 2 3 3
Other cash income 5 2 3 5 4
NET CASH INCOME 100 100 100 100 100
CASH % OF TOTAL 38 40 44 49 43

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

 

one-half of rural incomes are in the form of cash; the other half consists of the imputed
value of subsistence food production.“5

High average incomes from an activity do not always mean that the activity
contributes greatly to food security, since incomes that are highly concentrated may be of
marginal benefit to the food-insecure households. The examination of distributional
issues can be started by comparing income quartiles. Table 7-4 shows that the quartile

with highest incomes earns six times as much income fiorn bananas than the poorest

 

“ Due to lack of data and computational limitations, several minor income components
are excluded fiom our income estimates. An important omission is perhafis the value of
the meals provided by employers. These can be important for some poor ouseholds,
especially those with many teenage children. Also non-marketed non-agricultural
activities such as brewing are exc uded, as are livestock products produced for home
consumption.

152

quartile, but the same kind of inequalities hold for most other sources of income as well.
Table 7-5 confirms that, on average, households in the most prosperous quartile do not
earn any higher share of their cash fi'om bananas than households in the other quartiles.
The real difference between the quartiles can be found in labor sales. The poor depend
crucially on farm work, while the rich get most of their wages from skilled jobs.

To see whether incomes from a given source are highly concentrated within each
quartile among those with highest incomes, we can compare the shares implicit in Table
7-4, where the means are disproportionately afl‘ected by higher-income households and
Table 7-5, where household-level shares are averaged without giving more weight to
households with higher incomes. The numbers in Table 7-4 imply that 16 percent of the
aggregate cash incomes of the poorest quartile of households come from bananas,
whereas Table 7-5 reports that on average, households in the poorest quartile get 18
percent of their cash incomes from bananas. The difiemnce indicates that within this
quartile, bananas are relatively more important for lower-income households. Similar
reasoning can be used to show that both overall and within each quartile, cash incomes in
general and labor incomes in particular are unequally distributed, contributing to income
inequalitY-

While these averages show that banana sales on average are approximately as
important for the poor as for the rich, they still leave the possibility that only a small
number of people earn cash by selling bananas and banana beer. To find out whether
banana sales are important for many households, we next tabulate the share of households
in each quartile earning at least one-third of their cash income from each source. Table 7-
6 shows that for all income quartiles, more than one in four households gets at least one-
third of their cash income from bananas, whereas cofl‘ee contributes as much for only half
tint many households. Production of food crops for sale plays an important role for only
one in ten poor households. Labor sales are important for a much larger share of

153

Table 7-6: Households Getting >1/3 of Cash fiorn a Source

Percent Income quartile (FRw/AE) .All Farms
<537O 5370- 8040- >13120
8040 13120
Coffee 11 10 14 13 12
Banana (beer) 24 3O 27 25 26
Sorghum (beer) 6 4 4 3 4
Other crops 10 10 14 16 13
Livestock 8 14 14 7 11
Labor sales 46 42 31 37 39
Cash gifts 4 2 1 3 2
Other cash 6 2 2 4 3

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

households, including one-half of the households in the poorest quartile. Further
computations, not shown in the tables, indicate that more than 80 percent of Rwanda’s
rural households get more cash from bananas than from coffee or any other single crop,
and that, for 41 percent of the households, bananas are a more important source of cash
than all the other crops combined. One in two households gets more cash fiom bananas
than fiom wages and 15 percent of households get more than 50 percent of their cash
revenues from bananas sold as such or as beer.

7.5 Cash for Food
Much of the policy-oriented literature on Rwanda’s agriculture has argued that

beer bananas are not a proper cash crop, since traditional beers are sold primarily in the
rural areas and do not bring in cash fiom outside (Jones and Egli 1984). A related
argument claims that cash from banana beer contributes little to food security because it
is used largely to buy more beer (Miklavcic 1995). According to this argument, cash fi-om

154

beer bananas represents illusory earnings and should more properly be seen as a lubricant
in the process in which households take turns in producing batches of beer and share the
poorly storable product with their neighbors.

It is very difficult to examine the merits of these arguments. As we will see below,
interviewed households did not admit to buying much beer, but numerous studies fi'om
many parts of the world have found that alcohol consumption and purchases in general
are grossly tmderreported in household surveys (Deaton 1997). While there is no doubt
that rural Rwanda exports large quantities of banana beer to Kigali and other urban
centers (Miklavcic 1995), and probably also sells some banana beer to neighboring
cormtries (Loveridge 1988, 1992), it is also clear that DSA’s rural transaction data
significantly ‘tmderestimate rural purchases and overestimate rural exports.

With these caveats we report the cash expenditures on main crops and beer in
Table 7-7. The differences between the income quartiles are striking. While the poorest
quartile uses one-half of its net cash incomes for food crops, the richest quartile uses only
one-tenth for the same purpose. In other words, although only a small fi'action of
aggregate cash income is used for food, more than one-half of the cash in the hands of the
food-insecure buys food. Reported beer purchases are minimal in all quartiles, but much
more money is used for sorghum and beer bananas that can be used to make beer".
Unfortunately, it is not clear whether the purchases of beer ingredients were resold in the
form of beer. Although adjustments for beer and sorghum that only “pass through” the
household were made, they were not perfect and some misclassification was bound to

happen, especially for the poorest households, who rely most on purchased ingredients.

 

‘7 In an earlier study based on the same data set, Loveridge 1992) finned banana beer
purchases that were roughly 30 times as large as those m able 7- . This was incorrect,
resulting fiom a typing error in the SPSS program that efi‘ectively added sales to .
Blusrchases. The error was not carried over to numbers on rural ex rts (net sales) and in

' text Loveridge drew no conclusions from the erroneous num rs.

155

Table 7- 7: Household Cash Expenditures on Food Crops and Beer by Income Quartile

 

Francs Income quartile (FRw/AE) All Farms
<5370 5370- 8040- >13120
8040 13120
Food Crops 3,558 4,243 5,037 4,818 4,414
Beans 1,830 2,327 2,565 2,055 2,194
Sweet Potato 514 333 303 194 336
Sorg.&Beer Ban. 527 552 844 776 675
Banana Beer 58 96 135 331 155
Sorghum Beer 70 as 80 156 99

Total Cash Income 7,214 13,449 21,806 53,919 24,097

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households

 

Although most of the aggregate cash incomes remain unaccounted for in the table,
and although a significant fraction of the unaccounted share undoubtedly goes to beer,
this primarily applies to those who are far from starving. Half of the cash of the poor is
accounted for by recorded food crop purchases, leaving very little for industrially
produced foods (e.g., vegetable oil) and for non-food items.

The higher-income households probably buy much more beer than they admit.
Although these transactions bring no cash to the rural areas, they redistribute food access
sothatthepoorestgetmoretoeat.

If we compare net sales and production, we can see that the poorest quartile of
households increases its food availability by 15 percent through market transactions and
this does not yet include the meals provided by employers. Moreover, cash incomes from
beer sales and other sources contribute to nutritional status and other well-being not only
by financing food purchases. Indeed, the nutritional benefits of additional food decline
fast and above the very low levels money ofien buys more nutritional benefits if it is used

156

for medicines, soap, sanitation, health care or education instead of food (Blanken 1988;
von Braun, de Haen and Blanken 1991).

7.6 Future Role of Bananas in Food Security
We have now seen that many poor households in Rwanda grow beer bananas, sell

banana beer, and buy food, especially beans. Both the widespread adoption by farmers
and the numbers discussed above suggest that the exchange strategy is in the best interest
of the farmers, which at the very low income levels is not far from saying that it improves
their food security. Using the numbers in Table 5-1, it can be computed that the cash
revenue from a hectare of beer bananas sold as beer would on average have bought 1446
kilograms of beans“. This is 72 percent more than the estimated average national bean
yield. This shows that at the relative prices and yields that prevailed in 1990, the strategy
of exchanging banana beer for beans succeeds in substituting labor for land and in
improving the food entitlements of those engaged in the exchange strategy. In this
section, the focus is on the factors that determine the future role of bananas in Rwanda’s
food security. We start by asking what the prospects are that relative prices or yields will
change enough to undermine the basis of the banana beer exchange strategy. .

7.6.1 Declining Terms of Trade for Banana Beer? ,
As discussed above, the exchange strategy is associated with land scarcity so that

as population grows the market supply of banana beer can be expected to increase, while
that of food should decrease. Other things being equal, banana beer prices should decline
whereas food prices could increase. In principle, these changes could undermine the
economic basis of the beer exchange strategy. More generally, one needs to ask what are
me prospects that prices collapse if banana production increases.

 

" This uses producer price (sales price) for beer and consumer urchase) price for beans
and hence shows the advantage of exchange net of the costs c ed by the traders.

157

Consumption studies suggest that in Rwanda banana beer is a luxury consumed
primarily by households with relatively high incomes". This implies that the demand for
banana beer depends greatly on real incomes. If economic growth increases the
purchasing power of large segments of the population, demand for banana beer increases
much faster than population. Indeed, it is unlikely that banana beer supply could increase
as fast as demand, if the economy grows satisfactorily. If, for instance, both population
and per capita incomes grow by 2.5 percent per year, and if we, probably conservatively,
use 1.0 as the expenditure elasticity of demand for banana beer, supply should increase by
5 percent per year to prevent prices from rising. .

If, however, real incomes decline, demand stagnates or even declines. Market
prices can react dramatically, since stagnation in incomes not only reduces demand but
also increases marketed supply as some of the impoverished producers exchange a higher
share of their output for food. In part, the profitability of beer bananas in 1990 reflected
Rwanda’s economic growth that until 19803 had been sufficiently fast to allow a large
share of the rapidly growing population to maintain the traditional diet that includes large

quantities of beer.

Thus, the future of Rwandan smallholders’ exchange strategy depends largely on
how the overall economy develops. If real incomes increase, additional net sales coming
from the increasing number of small farms specializing in beer need not depress prices,
since demand will also grow rapidly. But if the stagnation of the late 19808 and the early
19903 continues in the nearby future, prices could decline and thereby force some farmers
to revert from banana-based specialization towards subsistence food production.

 

' ‘9 Although industriallgproduced beer is preferred to traditional beers, it is too expensive

for most Rwandans. 0th in urban and rural areas, the demand for banana beer rs
estimated to increase fast with incomes, in rural areas more than proportionally (that is,
it is estimated to be a luxury) (Ansoanuur 1991).

158

Apart from economic stagnation, changes in taxes on industrially produced beer
could depress banana beer prices. In the past, industrially produced beer has been taxed
heavily in Rwanda, whereas traditional beer has mostly escaped taxes (I-Iaggblade 1987).
The choice has largely been dictated by fiscal needs and administrative feasibility. Since
nearly all households manufacture some beer, taxes on traditional beers have been
thought impossible to administer. With the increased commercialization and rural-to-
urban transportation of traditional beers, the administrative feasibility of taxing them has
improved and the fiscal gain from doing so has increased. Thus, although radical cuts of
taxes on industrially produced beer may be unlikely for fiscal reasons, the same fiscal

considerations make new taxes on banana beer a distinct possibility.

The scenario where higher bean and cereal prices would choke off the banana-
based specialization on Rwanda’s land-scarce farms would be most likely in a situation
where Rwanda’s borders were more effectively closed to informal imports of beans and
cereals. In the early 19905, most of the beans bought by Rwanda’s land-scarce farmers
were imported fiom neighboring countries, mostly from Congo (Loveridge 1992).
Without higher trade barriers major increases in long-term average bean and cereal prices
are unlikely, since the relatively high transportability and storability of beans and cereals
suggests that supply is quite elastic. In other words, small increases in relative bean and
cereal prices would be suficient to expand the areas in Congo and Uganda that supply
beans to Rwanda

The opposite scenario is that of lower food prices brought about by more open
borders, improved infrastructure, lower transportation costs, higher agricultural
productivity in the supplying areas, or food aid. These obviously would encourage more
specialization in banana beer at the expense of food crops. One of the unintended
consequences of the massive food aid Rwanda has received since 1994 has been to
encourage Rwandan farmers to produce beer rather than food.

159

7.6.2 Declining Yield Advantage of Beer Bananas?
Just as the decline in relative banana beer prices could make specialization in beer

bananas unattractive, lower yields for bananas or higher yields for food crops could
destroy the economic basis of the exchange strategy. Whether this will happen depends
largely on policies affecting agricultural research, extension, and input supply.

All of Rwanda’s main food crops have average yields that are only a fraction of
those achieved in agronomic trials in Rwanda and are also far behind the yields harvested
by smallholders in the well-watered areas of Zimbabwe and Kenya (CNA 1990). This
suggests that with concerted efi‘orts in research, extension and input supply, the yields of
any of the main crops could be increased significantly, with major implications on what
farmers would choose to grow.

The potential of agricultural research in changing cropping patterns is illustrated
by Rwanda’s experience with white potatoes. By introducing local varieties resistant to
late blight (the main problem limiting potato production in Rwanda) and by supplying
pesticides to control mildew, the national potato improvement program (PNAP) has
greatly increased yields and production. Between mid-19705 and mid-1980s the area area
under potatoes more than doubled (Bart 1993). The initial success was limited to high
altitudes but the great interest of farmers elsewhere suggests that if improved planting
materials, technical advice and pesticides were made more widely available, the area
under potatoes could expand further (Little and Horowitz 1988).

Another recent change in cropping patterns has been a shift from bush beans to
climbing beans (CN A 1990). The latter require much more labor per unit of land, but
have much higher yield potential in favorable conditions, are less susceptible for the
damage caused by excessive rain and thrive at higher altitudes. More importantly, some
new climbing bean varieties are resistant to root rots, which had become a major
constraint in the production of the traditional bush bean varieties. Where climbing beans

160

grow well, farmers can get more beans by growing climbing beans than by growing
bananas and exchanging beer for beans. By the late 19803 climbing beans had become
popular in some fertile, well-watered densely inhabited areas, particularly in Gisenyi, and
in the 19905 they have spread throughout much of the country (Sperling and Munyaneza
1995). In the areas most affected by root rots, many farmers had stopped producing beans

before the introduction of the new varieties resistant to root rots.

However, these two success stories are exceptions to the general picture which is
one of minimal technical progress associated with underinvestment in agricultural
research. Relative to the size of agricultural output, few countries in the world have
invested as little in agricultural research as Rwanda. In terms of funding, the neglect has
not always been visible in international statistical comparisons, since agricultural research
in Rwanda, like in many other African countries, is fimded mostly by donors. As a result,
the level of fimding has been highly volatile. Moreover, a large share of the recorded
expenditures has been spent on a handful of expatriate researchers.

Even a quite limited effort could be effective, if it is focused tightly on a
suficiently small number of key constraints and concentrated on selecting, adapting and
transferring technologies developed elsewhere, provided that appropriate technologies
exist. The success of the national potato improvement program was based on a tight focus
on two problems (late blight and mildew) and two solutions (varieties resistant to late
blight and pesticide to control mildew) (Amani and Haverkort 1986; Haverkort 1986a,
1986b, Wilcock and Ndoreyaho 1986). Moreover, the program targeted an inherently
fertile volcanic small area where potato was the cash crop (Durr 1983). Hence, land
fertility was much less of a problem than in most other parts of Rwanda, and getting
inputs and advice into the hands of the farmers was relatively easy. Much of the push for
creating a critical mass behind the program came hour the International Potato Center

(CIP), which was eager to demonstrate impact by capitalizing on its success with late

161

blight resistant varieties. In the case of climbing beans, the keys to success were
improved varieties and strong support fiom CIAT (Sperling and Munyaneza 1995).
However, outside these two programs the situation is quite different. In general, the
limited resources of Rwanda’s national agricultural research institute (ISAR) have been
spread extremely thinly on tens of commodities and hundreds of production constraints
(ISAR 1989; Eicher, 1990).

In light of the limited resources and lack of focus, it appears tmlikely that research
alone could increase food crop yields in the nearby future so dramatically that
specialization in bananas would lose its appeal. More likely challenges for the banana-
based intensification include declining banana yields due to pests and diseases, and
higher food crop yields brought about by large fertilizer subsidies.

7.6.3 Fertilizer Subsidies?
One policy change that could change the role of bananas in Rwanda is the

introduction of major subsidies on mineral fertilizer. While fertilizer subsidies have not
played a large role in Rwanda’s past, they may soon become a major issue in Rwanda’s
food policy.

In pre-genocide Rwanda, fertilizer use was minimal and mostly limited to export
crops. Several projects distributed fiee or heavily subsidized fertilizer, but quantities
remained small. The main deterrant was the extremely high international transport cost of
“more than $270 per ton from Kigali to Europe,” which alone was enough to more than
double the cost of fertilizer in Rwanda (World Bank 1991)?0

Yet, it is precisely the costly international transportation that may push Rwanda or
its donors to subsidize fertilizer. Even if Rwanda’s farmers fell short of the “rule-of-

 

’° It is unclear whether this refers to the shortest, and the only fully halted route to the
sea Via Uganda to Mombasa. In the 19805 a litical blockade by ganda often forced
Rwanda to use more costly routes (Bart 1987f0

162

thumb” response ratio of ten kilograms of cereals per one kilogram of fertilizer (Pinstrup—
Andersen 1982), costly transportation means that it is less expensive to increase food
supply in Rwanda by importing fertilizer from international markets than by importing
cereals from there.

Since the genocide, the focus in and around Rwanda has been on emergency
relief, and hundreds of thousands of tons of food aid have been trucked or flown in
without much concern for costs. However, the return of hundreds of thousands of
refugees in late 1996 made it clear that food aid would be needed for years to come and
increased interest in alternatives that would be financially more sustainable in the long
term. In 1997 and 1998, the Government of Rwanda and USAID oflicials were
discussing whether donors could “jump-start Rwanda’s rural economy” by shipping in

large quanties of fertilizer.

In pre-genocide Rwanda, the idea of increasing food production by promoting the
I use of fertilizer was already part of the conventional wisdom. In the early 19805,
increased use of fertilizer and other external inputs was declared one of the corner stones
of the oflicial food security strategy (MINIPLAN 1983; Bart 1987). Ten years later,
Rwanda’s National Agricultural Commission went further, stating that increased use of
fertilizer was “a question of life and death” (CNA 1992a). The commission computed that
to make fertilizer use “financially viable,” farmgate prices should be subsidized by 40
percent. For fertilizer use to achieve the level deemed necessary by the commission, the
total subsidy should be more than ten million dollars per year, which is far more than
what has been used for agricultural research. Fertilizer needs were computed by adding
the estimated uptakes of nutrients by crops to the estimated losses of nutrients caused by

soil erosion.’1

 

5' This, of course, leads to a great overestimation of nutrient outflows since in Rwanda
most crops are used and recycled on the farms and most of the topsoil lost due to

163

Also many donor documents emphasized the importance of fertilizer (Jones and
Egli 1984; Wilcock and Ndoreyaho 1986; World Bank 1991). Instead of advocating
costly subsidies, donor experts argued that fertilizer either is already profitable or could
be made so by appropriate investments in agricultural research and extension. Hence,
they called for research on more fertilizer-responsive crop varieties and for the
development of easy-to—extend packages that would incorporate fertilizer.

To substantiate their claims that fertilizer pays or at least could increase
production if promoted with subsidies and supported by relevant research, some of the
above documents summarize the results of agronomic trials on fertilizer response. The
computations carried out for the National Agricultural Commission suggest that bush
beans, bananas and sorghum are so unresponsive that fertilizer use would remain
unprofitable in all reasonable price scenarios (CNA 1992a). In contrast, climbing beans,
soybeans, potatoes, and wheat, along with Rwanda’s main export crops, cofl‘ee and tea,
have given high returns to fertilizer in the trials (CNA 1992a; World Bank 1991). Results
for sweet potatoes, cassava, and maize were also somewhat encouraging (CNA 1992a;
Wilcock and Ndoreyaho 1986). More importantly, combining fertilizer with varieties that
are highly responsive to it, could increase yields dramatically. In particular, the
combination of fertilizer and hybrid maize is assumed to have much potential, as it has in
parts of Kenya and Uganda. In the World Bank studies one of the key arguments was that
research; on fertilizer response deserves high priority (Jones and Egli 1984; World Bank
1991).

 

erosion is deposited on the fields lower on the hillside. On the other hand, main?! fields
in Rwanda are so degraded that merely replacing current nutrient outflows co d not
make them productive (K6nig 1992).

164

In Rwanda, fertilizer subsidies would probably encourage the production of high-
response crops at the expense of bush beans, sorghum, and bananas.’2 The relative
importance of these three traditional crops would decline, while the role of newer crops
such as soybeans and potatoes would increase. Since sorghum has been losing its relative
role for the past forty years or more and since there already has been a clear trend fi'om
bush beans to climbing beans and soybeans, fertilizer subsidies would only hasten the
trends. In the case of bananas, however, a major reversal could happen. Bananas have
expanded secularly for several decades, but could decline if bypassed by the production
increases brought about by subsidized fertilizer.”

Much depends on the results of the associated research agenda. The combination
of major fertilizer subsidies and locally adapted highly responsive maize, cassava, and
sweet potato varieties could stop the expansion of bananas and gradually transform much
of Rwanda fi'om a system based on banana intercropping into a system dominated by
monocmpped cereals or roots and tubers.

 

5’ In theory, the acreage under the responsive crops could decrease if higher yields and
production brought about by cheap fertilizer dramatically reduce prices. However, as
discussed above, beans cereals, and cassava are traded regionally, which reduces the
impact of national fertilizer subsidies on firjices. The disincentive efi‘ect would be felt
mostl in the lus areas of Congo and anda, where slightly lower food prices
woul not be 0 set by significantly lower ertilizer prices.

’3 Two major caveats for these conclusions should be noted. First, most of the fertilizer
trials noted above have examined responses to common] available compound fertilizer
types containing nitrogen, phosphorus and potassium. 'le this may be a reasonable
starting point and the most cost-effective fertilizer application for some crops and some
soils, more profitable fertilizer applications could certaian be found by better tailoring
the application of chemical ferti rzer to soil conditions an to the nutrient uptakes of

each crop. For instance, banana groves that failed to give attractive returns to

compound fertilizer might have economically interesting responses to potassium.

Second, fertilizer re use trials typrcall focus on short-term drrect impacts of the

nutrients and ignore ong-term impacts ugh changes in soil structure. Given that

bananas thrive on deep soils with much or anic matter and that increased biomass
production brought about by fertilizer can used to build up the organic matter
content of the 5011, it is ssible that certain fertilizer-augmented soi ennchment
strategies could be pro table for bananas, although the straightforward fertilizer trials
fail to show sufficient impact.

165

7.6.4 Transition to Edible Bananas?
The numbers in Table 5-1 show that in addition to being highly profitable cash

crops, bananas produce more calories per hectare than any other main crop in Rwanda.
Moreover, cooking bananas supposedly yield even higher cash returns to land than beer
bananas. Yet, as we have seen, cooking bananas are not a particularly popular food for
the poorest farmers.

The solution to the paradox lies in the yield gap between beer and cooking
bananas. Since beer, cooking and fiuit bananas were not separated in land use statistics,
Table 5-1 includes only the aggregate banana yield. Although our data set cannot show
the gap directly, we can indirectly conclude that cooking banana yields must be
considerably higher than beer banana yields. Only a large yield gap can explain the fact
thatcookingbananasaremuchmore expensivethanbeerbananasandyetsmall farmers
prefer beer bananas.

Cooking bananas are grown mostly in the East, where banana yields are much
higher, and nationally their mean yield may well be close to that of beer bananas. Even
on many farms, cooking bananas may yield as well as beer bananas, since they are more
demanding and grown usually on the most fertile fields just around the compound“. But
on similar land, cooking bananas must yield considerably less than beer bananas, so that
farmers could not increase their cash incomes simply by substituting cooking bananas for
beer bananas.

Across Rwanda’s eastern border, in Kagera, Tanzania, the superior yields of beer
bananas are well documented. According to Kajumulo-Tibaijuka (1984) farmers in this
area had substituted beer bananas for cooking bananas because the former are higher-
yielding and less susceptible to pests, which had become a major problem in the 19705.

 

5‘ Jean-Bosco Sibomana, personal communication.

166

____ ___._.—'——fi

Pest problems were particularly severe in areas which had been covered by banana groves
for several decades. This gives the hint of an alarming prospect that the important role of
of cooking bananas in eastern Rwanda may be only temporary. Until independence, much
of Kibungo was reserved for pasture and most of the banana groves in the area have been
planted quite recently. Thus, the pests which have devastated cooking banana groves in

Kagera may become a much more severe problem in eastern Rwanda in the near future.

Technically, beer bananas are just as edible as cooking and fruit bananas, but the
taste is different and, as Table 5-1 shows, consumers are willing to pay nearly twice as
much for cooking bananas as for beer bananas. Indeed, while beer bananas are a famine
food eaten usually only under extreme deprivation, cooking bananas are a highly
appreciated food crop. A more expensive source of calories than sweet potato, maize,
sorghum, or cassava, cooking bananas are consumed primarily by the better-ofl'
households. Eastern Rwanda, especially Kibungo prefecture where cooking bananas grow
best and are cheap, is a partial exception in that cooking bananas are the main staple there
for poor and rich alike.

In addition to parts of Kibungo, cooking bananas increasingly dominate in
Rwanda’s pen-urban areas. Based on this tendency, Bart (1993) predicts that peri-urban
cooking banana gardens may show the path for the rest of the country. However, the
evidence presented above on prices and consumption patterns suggests another
explanation. Instead of being a response to land scarcity, the peri-urban cooking banana
gardens probably reflect proximity to high-income consumers. Urban consumers demand
both cooking bananas and banana beer. Since the latter is much more transportable (has
four times more value per weight) than the former, pen-urban and other well-connected
ueasspecializeincookingbanamswhflethemommmoteareasspecializembmma
beer.

167

The prices and consumption patterns of bananas suggest that depending on
economy-wide trends, two entirely different developments for edible bananas are
possible. If Rwanda prospers, the demand for cooking bananas increases fast, inducing
increased production, especially in peri-urban areas, along roads and in Kibungo. But if
Rwanda stagnates, and ifboth the demand for banana beer and beer banana prices
collapse, millions of poor Rwandans may be forced to eat beer bananas more regularly,
and not just during famines, as in the past. Depending on how much and how persistently
they dislike beer bananas, producers may either start substituting root crops for bananas
orlearntoacceptbeerbananasasafoodcrop.

Obviously, the above scenarios hold only if the price and yield gaps between beer
and edible bananas persist or become even wider due to pest attacks. If agricultural
research succeeds in narrowing the gap, by improving edible banana varieties or their
management practices or by developing new varieties which yield as much as beer
bananas but are appreciated as food, the entire picture changes. Based on such
achievements, edible bananas could expand independent of what happens to incomes,

improving considerably food security and environmental sustainability in Rwanda.

7.7 Chapter Summary ‘
In Rwanda, the grass-roots view on bananas is strikingly different from the

aggregate view. From the aggregate point of view, the brewing of bananas into beer
represents an important waste of food. In theory, a reduction in the waste and a
redistribution of the saved food to the needy could significantly reduce undernutrition in
Rwanda, if not eliminate it altogether. But from the point of view of hundreds of
thousands of poor households, bananas are crucial for their food security, not least via the
exchange of banana beer for cheap staples.

168

The evidence presented in this chapter indicates that (1) despite the nutritional
losses associated with brewing, bananas provide more than one-fifth of the calories
consumed in rural Rwanda; (2) bananas are the single most important cash crop and
second only to labor sales as a source of cash; (3) cash incomes fi'om bananas are widely
distributed in rural Rwanda and no more concentrated among the better-ofl' households
than other sources of income; (4) the poorest households use halfof their cash income to
buy staple foods, especially beans imported informally from neighboring countries; (5)
food purchases are essential for food security, since subsistence production is seriously
deficient in proteins and lipids; and (6) beer banana yields are so high that land-scarce
households typically can get more beans by growing bananas and exchanging beer for
beans than by growing beans themselves.

The key role of bananas in the food security equation of rural Rwandans is based
primarily on two ratios: the yields of bananas are relatively high compared to those of
other major crops and the prices farmers can get for banana beer they sell are relatively
high compared to the prices they need to pay for the beans and cereals they buy. In
several plausible scenarios these ratios could change sufi'rciently to undermine the basis
ofbanana-based specializationfion Rwanda’s hillsides. A tax cut on industrially produced
beer that would reduce banana beer prices, import restrictions that would increase food

 

prices, fertilizer subsidies that would increase food crop yields, and pests that would
reduce banana yields are among the possibilities. Obviously, reverse scenarios that could
increase the contribution of bananas for food security also exist. Technical progress that
would increase the yields of cooking bananas close to the current yield levels of beer

bananas is one such scenario.

The purpose of exploring the above scenarios is not to predict the future of
Rwanda’s banana-based intensification process. Instead, the scenarios can illuminate

169

some of the consequences of selected policy choices and thereby provide the background
for the policy discussion in the concluding chapter.

170

Chapter 8

OTHER RESPONSES TO DEMOGRAPHIC PRESSURE

8.1 Introduction
In the preceding chapters we have focused on how farmers with difi‘erent amounts

of land per person differ in their land use and participation in agricultural product
markets. We have found that in most of Rwanda, land scarcity is strongly associated with
more intensive cropping patterns, dominated by the intercropping of bananas and food
crops. This suggests that banana-based intensification of cropping patterns is a major
response to demographic pressure in Rwanda

The purpose of this chapter is to put the banana-based intensification into a wider
perspective by looking at the other forms of agricultural intensification and the bigger
picture of how agricultural households respond to population pressure on and off their
farms. Drawing from existing literature and hem DSA data, we look at agricultural and

non-farm income strategies. This is followed by a very brief review of the demographic

responses.

8.2 Other Agricultural Responses

8.2.1 Investments in Erosion Control
As mentioned in section 2.3., strictly enforced policies to fight soil erosion were

introduced in Rwanda dming the colonial era, discontinued at independence, re-launched
in the 19705, and intensified during the 19805. Most often farmers were advised and
forced to dig infiltration ditches and plant grass strips, both on the contours. On steeper
slopes, contour hedgerows rather than grass strips were usually the recommended
method. Since the result of well-maintained grass lines or hedgerows is a gradual build-

171

up of terraces, contour lines with vegetative barriers are often called (progressive)
terraces. By 1990, most sloping fields were “protec ” by at least some anti-erosion
stuctures. According to the DSA data set discussed in the preceding chapters, Rwandan
fields had on average 205 meters of grass strip per hectare, 161 meters of anti-erosion
ditches, and 56 meters of hedgerows (Clay, Reardon, and Kangasniemi 1995).”

Many observers have praised Rwanda’s beautifully “protected” hillsides and
Harrison (1987) even argues that diligent soil conservation is one of the key reasons why
Rwanda was among the few Afiican countries that until mid-19805 could increased food
production fast enough to match the rate of population growth. However, both the poor
performance of Rwanda’s agriculture since the mid-19805 and detailed studies on how the
soil conservation investnents work on Rwanda’s steep hillsides cast doubt on these

conclusions.

The anti-erosive measures promoted during the colonial period and again after
1973 were developed in the 19405 to reduce soil erosion on slopes of less than ten
degrees (Moeyersons 1989). By Rwandan standards these are gentle slopes. One-half of
Rwanda’s fields have slopes that exceed 11 degrees, the mean slope of fields is estimated
to be 16.7 degrees, and two-thirds of the soil losses occur on slopes that exceed 21
degrees (SESA 1986).

Several researchers argue that on steeper slopes ofi'rcial measures can be
ineffective or even harmful. Bidou (1989) argues that the anti-erosion ditches promoted
in Burundi and Rwanda by the Belgians were not able to deal with heavy rains, and if not
properly maintained, could lead to gullying. Nduwayezu (1990) shares these allegations

 

’5 Hedgerows were not necessarily woody perennials. In the surv , pennisetum ureum
lines were classified as hedgerows, unless they were k short?y cutting reguralrl'lg for
fodder (Jean Bosco Sibomana, personal commrmication .

172

and also criticizes infiltation ditches for being costly to make, requiring lots of

maintenance, not increasing production, and not reducing erosion between the ditches.

Nduwayezu (1990) advocates grass lines, claiming that if made according to the
official recommendation, they reduce soil losses to one-tenth of what they would
otherwise be. However, on steep slopes this apparently is neither quite tue nor good
enough. Byers (1990), Brondeau (1991), and Lewis (1992) report average annual soil
losses of 54 t/ha, 83 t/ha, and 62 t/ha, respectively, even with vegetated barriers. While
much lower than those achieved on similar conditions without barriers, they are too high
to allow continuous cropping in the long term. Moreover, as pointed out by Ndayizigiye
(1989, 1992), among others, grasses and hedges take up nutients which, if not replaced
with fertilizers, reduces crop yields. Unless the grasses and hedges themselves can be
converted into valuable products by feeding them to livestock, farmers may not be able to

reap any benefits.

The key problem is that just as infiltation ditches fail to arrest erosion between
ditches, grass lines do little to reduce erosion between the lines. Instead, they lead to the
accumulation of topsoil from the top of the interline area to the bottom, as the terrace
progressively is formed. In much of Rwanda, the loss of topsoil fiom the upper sides of
the terraces is a serious problem. Rwandan soils are fertile but the fertility resides only in
the topmost layers, which are fairly thin (Moyersons 1989). Subsoils are often acid, poor
in main nutients, and have potentially toxic levels of aluminium (Zaag et al. 1984). At
least on fragile highlands, terracing has reduced fertility and increased soil acidity (Lewis
1992)

Farmers often cultivate too close or even under the progressive terrace to make it
collapse and then spread the accumulated fertile topsoil more widely (Lewis 1992; Bart

1993). But even without such man-made soil losses the formation of progressive terraces

173

can destabilize steep hillsides and lead to gullying (Webster and Wilson 1980). Where
soils are deep, even landslides can follow (Ndayirukiye 1989), as many Rwandan farmers
experienced during the unusually heavy rains in May 1988 (Byers 1988).

To some extent, the problems experienced on steeper slopes could be addressed
by reducing the distance between grass lines and increasing the width of the vegetative
barrier, as is the official recommendation (Nduwayezu 1990). However, this would leave
less space for crops and reduce current incomes, unless the plants grown on the barriers

have considerable value as fodder, fuel, or food.

The key message from the literature is that Rwanda’s anti-erosion campaign was
rigid and simplistic. It has advised and at times forced farmers to adopt practices with
little regard to their physical environment (rainfall, soil type, slope) and almost no regard
to the economic environment and to what farmers were growing on their fields and
whether they owned any livestock. Little attention was paid to indigenous practices
already used by farmers in many parts of the county (Derenne 1989). It stands to reason
that the interests of and benefits for farmers varied widely, ranging from those who were
clearly hurt by the measures to those whose needs were so well met by the measures that
they did even more than was required. In between the extemes must be some who were
initially reluctant but then were persuaded by the benefits they saw and others whose
efforts would have been better spent on other activities but who nevertheless decided to
maintain the stuctures when the most laborous task of putting them in place was done.
Not much information exists on where in this range most farmers were and what they
would have done in the absence of the compulsion. Guichaoua (1991) who interviewed
farmers in 1989, reports that the anti-erosion campaign was bitterly resented and mist'usted,
with measures generally seen as either useless or even harmful. Yet, many farmers agreed
that soil conservation in some form was needed. What they resented were the specific things
they were asked to do and the degree of compulsion that was used.

174

Even in a context of a government campaign to promote soil conservation, one
could argue that the large variations between households reflect the economic, social, and
environmental incentives faced by farmers and therefore could be used to get some
understanding on what is going to happen now that the enforcement of the soil conservation
policies has (temporarily?) ceased. This is the approach taken by Clay and Reardon (1994)
and Clay, Reardon, and Kangasniemi (1995), who used the DSA data set to examine the

determinants of soil conservation investnents.

The dependent variable in the analysis is the amount (meters) of three types of
conservation investments (grass stips, anti-erosion ditches, and hedgerows) per hectare
on the farm. Independent variables include measures of monetary and physical incentives
to invest, risk of investnent, wealth and liquidity and other household characteristics. No
proxies of the degree to which the policies were actually enforced in the area were used.

Some of the results supported the hypotheses: off-farm employment, small farm
size, and large household labor force appeared to promote investnents, whereas
insecurity of land tenure appeared to discourage them. However, also statistically
significant anomalous results were found: The index of agricultural profitability had a
statistically significant positive association with grass stips and anti-erosion ditches, but
a statistically significant negative association with hedgerows. Also price variation and
knowledge of conservation techniques had coefficients that were positive for some forms
of soil conservation investnents and negative for others. High agricultural wages, which

make investnents more costly and could be hypothesized to discomge conservation,
appeared to encourage it.

The regression equations explained only a small fiaction (6-1 3%) of the variation
in the investnents. This is not surprising in light of the claims that much of the variation

may have been related to difi‘erences in how rigorously the policies were enforced

175

(Guichaoua 1991). Neither does it necessarily invalidate the results, if the unmeasured
degree of enforcement is not correlated with the explanatory variables used. Thus, the key
question is whether the implicit assumption that enforcement is uncorrelated with the

other determinants is warranted.

If, for instance, all adults were mandated to constuct a certain quantity ofterraces
on their fields, this would imply that enforcement was correlated positively with family
size and negatively with farm size. Thus, the key result suggesting that households with
much labor and little land indeed intensify by investing in soil conservation could be a
result of enforcement rather than of voluntary responses.

In smn, although the result that land scarcity is associated with investnents in soil
conservation appears plausible, the evidence to support this assertion is not very stung.
Further, although the degree to which farmers invested in soil conservation because they
were forced to do so is unclear, the evidence is strong that compulsion was an important
factor in the 19805. The additional incentives that population growth may create for the
constuction of progressive terraces in the nearby future are likely to be more than offset
by the fact that the enforcement of the rules requiring farmers to conserve their soils
effectively ended in the early 19905.

More importantly fi'om the point of view of this study, even if population pressure
brought about significant increases (or decreases) in how farmers invest in and maintain
their progressive terraces, it is not clear how such changes would afl‘ect their incentives to
intensify through banana-based intercropping. To the extent that progressive terraces
reduce soil erosion, they may reduce the need to plant bananas for soil conservation
purposes. On the other hand, the profitability of bananas along with their sensitivity to
variations in soil fertility suggests that if progressive terraces work as intended, the

payoffs to planting bananas could increase.

176

In conclusion, while investnents in soil conservation in principle can change
farming systems and cropping patterns, we have found little evidence to suggest that soil
conservation in Rwanda is fundamentally changing the basis of the banana-based
intensification patterns discussed in the preceding chapters. Neither the prospects of soil

conservation nor its impacts on cropping patterns are clear.

8.2.2 Agricultural Inputs
Increased use of external and on-farm inputs is another way of coping with

increasing land pressure that potentially could change cropping patterns and modify
Rwanda’s ongoing banana-based intensification. This section briefly reviews the
evidence on input use in Rwanda to explore whether ongoing changes in input use could
be modifying the basis on which Rwanda now is heading towards labor-intensive banana-
based intensification.

In the early 19905, Rwandan farmers used only minimal quantities of external
inputs. The only significant exceptions were the use of pesticides to cont'ol mildew on
potatoes in the northeastern part of the county and the use of fertilizer on tea plantations.
Apart from these cases and occasional distibutions of free or heavily subsidized fertilizer
by externally funded development projects, farmers largely ignored the fertilizer
recommendations of the research and extension systems. According to DSA’s national
survey, less than five percent of parcels received any purchased inputs in 1991 (Clay,
Reardon, and Kangasniemi 1995).

As shown above in Table 6-1, according to the 1990 tansaction survey, Rwanda’s
rural households spent on average only 84 fiancs on purchased inputs (other than labor).
Roughly 30 times more were spent on the purchases of agricultural labor, suggesting that
liquidity is not be the principal constaint on low input use. Instead, farmers choose to
allocate their cash to labor rather than to fertilizer and pesticides, either because chemical

177

inputs are not available, because returns on them are low, or because farmers fail to
appreciate the returns. The fertilizer tials discussed above found that, on most of
Rwanda’s major crops, fertilizer use was not particularly profitable at the prices that
prevailed in the late 19805.

In contrast to purchased inputs, on-farm organic inputs are widely used in
Rwanda. According to DSA’s national survey, almost 70 percent of parcels received
manure, compost, and/or mulch in 1991 (Clay, Reardon, and Kangasniemi 1995). As
discussed below in more detail, most livestock are kept in stalls at least during the nights
and the resulting manure is applied to nearby fields. Mulching is common on fields
covered by bananas and cofl'ee, and much of household waste along with some crop
residues is composted and spread on the fields.’6 The recycling of the human “nightsoil”
is less rigorous, but banana tees are often planted on top of former latine pits.

Given the central role of organic inputs in fertility management on Rwanda’s
farms, understanding how that role is changing would be valuable. General literature on
intensification provides no clear guidance. As discussed in Chapter 3, the use of organic
inputs can be highly labor-intensive and therefore increasingly careful mulching and
manuring is typically a key part of labor-based intensification associated with land
scarcity (Goodland, Watson, and Ledec 1984). On the other hand, land scarcity can
reduce the availability of organic inputs. Noting that much of the mulch used on cofl‘ee
fields comes from uncultivated (or fallowed) valley bottoms that are being cleared for
cultivation, Jones and Egli (1984) predicted that mulching will decrease in Rwanda.

 

" There is an important difference in how coffee and banana fields are mulched. The
former involves hard work by farmers to bring in the mulch, whereas the latter occurs
naturally on established banana groves, unless farmers remove the crop residues either
to cultivate other crops with bananas or to use the residues for other purposes such as
the (mandatory) mulching of coffee.

178

Empirical evidence provides little support for this prediction. Using DSA’s
national survey data, Clay, Reardon, and Kangasniemi (1995) examined the determinants
of organic input use on the parcels of Rwandan households. The results on this topic are
subject to similar caveats as those on soil conservation investnents discussed above.
Most notably, both the production of coffee and the nearly universal mulching of it were
stongly influenced by government policies and their enforcement. Variation of these was
not measured and little is known about how enforcement was correlated with the

socioeconomic factors studied.

With these caveats we can note that the use of organic inputs was significantly
and negatively associated with farm size. In other words, large farmers were less likely to
use organic inputs on any given parcel than their neighbors with less land. Further,
organic input use favored fields that were close to the residence, owned rather than
rented, flat rather than steep, and located on the hill rather than in the valley.”

The result does not support the claim that organic input use is declining under
demographic pressure due to the reduced availability of organic inputs. Rather, it appears
that declines in fallows and pastures have been offset by the increased planting of
bananas, tees, and hedgerows. In conclusion, there is little evidence to suggest that the
direction of Rwanda’s intensification is fundamentally changing as a result of ongoing

changes in input use.

8.2.3 Livestock Practices
Although it is fiequently assumed that the role of livestock must inevitably decline

when demographic pressure forces households to grow food for themselves rather than

 

’7 The results also included a statistically significant positive association between the size
of the parcel and the likelihood of input use, but thrs cannot be interpreted to suggest
that a plant growing on a large arce was be more likely to enjoy organic inputs than a
plant owing on a small parce . Input use was not quantified and it rs rather obvious
that a arge parcel (that in Rwanda typically consists of several difl‘erently cropped

179

grass for their animals, empirically based literature paints a quite different picture
(Ruthenberg 1976, Jahnke 1982). While large herds are not compatible with high
population density, mixed farming often is. In low-income and low-input conditions,
moderately dense livestock populations are often used to convert crop residues and weeds
into animal products, draft power, and manure, which plays a key role in fertility
management (McIntire, Bourzat, and Pingali 1992; Decuyepere et a1 1982). Where local
incomesarehigherorlargermarketseasilyaccessiblesothatfarmersfaceastrongdemand
for animal products, intensification with purchased inputs and specialized fodder crop
production can lead to very high livestock densities. On parts of Kenyan highlands, for
instance, a shift to intensive dairy production has allowed households to earn higher
incomes on increasingly small farms (Jahnke 1982).

In Rwanda, total livestock numbers increased during the post-World War 11 period
but much more slowly than population. As a result, the relative importance of livestock,
especially that of cattle, in Rwanda’s rural economy declined. Large herds disappeared
and increasingly farmers have substituted small ruminants for cattle. While Rwanda had
.29 cattle, .12 sheep and .40 goats perperson in 1953, the numbers in 1989 were .10, .10,
and .26, respectively (Rwamasirabo 1990).

Expert opinions on the current and firture role of livestock in Rwanda’s rural
economy vary widely. Before the events of 1994 fundamentally changed the situation,
most observers believed that Rwanda was overstocked with animals. For instance,
Rwamasirabo (1990:66) estimated that Rwandan farmers have "roughly three times the
amount of livestock recommended for adequate animal nutition and ecological
sustainability". He referred to the "unmatched expansion" and asked whether livestock

in Rwanda has been "tansformed from a resource that generates income and helps

 

fields) 15 more likely to receive inputs and be therefore classified as organically
fertilized than a small parcel.

180

improve nutrition to a destoyer of natural resources". At the same time some other
observers saw room for expansion and argued that soil conservation with vegetative
barriers can only succeed if farmers can turn the grasses and hedges into valuable

products with livestock.

In Table 8-1, we use the DSA 1990 data set to take a cross-sectional look at
lifestock holdings. The last column gives the national averages: While only 31 percent
households have cattle, 50 percent have goats and 24 percent have sheep. The holdings
overlap so that 28 percent of households have neither ruminants nor pigs (but many of them
have chicken or rabbits). To get a measure of livestock pressme on land, we compute
stocking rates (livestock densities) by converting each species into topical livestock units
(T'LU) and dividing by farm size. Ifwe aggregate over all farms, the average stocking rate is
.89 TLU/ha; excluding households without livestock gives an average rate of 1.23
T'LU/ha."

Compafingfmmsizequarfilesweseethatasthepopulatondensitymaeasesand
farmsizedecreases,the shareofhouseholdswithcattledecreasesasdoestheaverageherd
size. However, among those who do have cattle, farm size declines more than herd size so
that average stocking rate increases. The same applies also for goats and sheep. More
importantly, the overall stocking rate is highest on the smallest farms even if we average
over all households, not just over those with animals.

 

5' These numbers are considerably lower than those com uted by Rwamasirabo (1990)
using the essentially comparable data fiom 1989 Agri tural Surve . The difl'erence
arises primarily from how livestock numbers are converted into TL 5. Rwamasirabo
assumed that all cattle on average were 250 kg, goats one-sixth of that, etc, whereas we
assume that these appl onlyhto adult animals, and that the conversion factors for yormg
animals should be smaller. e question is empirical and not tivial since our conversion
factors lead to stocking rates that are only two-thirds of those used by Rwamasirabo. In
the absence of measurements fiom Rwanda we chose to use conversron factors that are
gompatiblirggyose of the International Livestock Center for Africa (ILCA) as reported

y ahnke .

181

Table 8—1 : Household Livestock Holdings by Farm Size

FARM SIZE QUARTILE (Ares/AB) All Farms
<9.S 9.5-1S.0 15.0-25.5 >25.5
CATTLE
% with cattle 13% 24% 37% 48% 31%
average herd 1.88 1.92 2.40 2.91 2.44
density ('rLU‘lha) 4.61 2.31 1.44 1.11 1.83
% keep in corral 70% 64% 62% 53% 60%
% permanently in’ 25% 19% 28% 14% 20%
a cutscarry’ 78% 64% 61% 55% 61%
COATS
% with goats 42% 45% 54% 59% 50%
average herd 2.51 2.50 3.02 3.27 2.86
density (nati/ha) .ss .40 .29 .21 .41
% keep in corral 55% 52% 38% 37% 44%
a permanently in’ 35% 22% 21% 21a 24%
% cutacarry’ 58% 46% 40% 37% 45%
SHEEP
% with sheep 18% 22% 28% 31% 24%
average herd 2.15 2.02 2.61 2.87 2.47
density (TLtJf/ha) .66 .33 .27 .17 .32
% keep in corral 62% 60% 45% 47% 52%
% permanently in2 44% 24% 27% 16% 26%
% cutacarry’ 61% 49% 49a 45a 50a
TOTAL LIVESTOCK
% with livestock 60% 67% 78% 83% 72%
average herd (TLUU .55 .75 1.13 1.55 1.04
density (TLU1/ha) 1.94 1.28 1.03 .89 1.23
% with corral 57% 55% 51% 52% 54%
% keep all in” 39% 26% 24% 20% 27%
% mostly cutacarry’ 57s 49% 47% 49% 50%
TLU/ha (ALL FARMS) 1.16 .86 .80 .73 .89

1TIM-Tropical Livestock units, '% keep animals permanently in corral,
3% rely primarily on stall-feeding (cutacarry), not grazing
Source:DSA/MINACRI, Agricultural Survey 1990; 1035 households

Someofthestocldngratesonsmallfirmsareveryhigh, suggestingthatthefodder
for these animals may have come fiom outside the firms. Another possible explanation is
that small farmers have adapted livestock management practices that substitute labor for
landResultsfiomelsewhere showthatagivenareacansupportmoreanimals,ifitisnot
grazed, but instead firmers keep their livestock in corrals (stables, stalls), cut grass, leaves,

182

and other fodder, and carry it to the animals. This cut and carry system or stall-feeding is
more labor-intensive than grazing, but more compatible with crop production, especially if
the crops are perennials. Moreover, it helps conserve nutients. Much of the efi‘ectiveness of
manure is based on nitrogen which is easily lost under exposure to sun (McIntire, Bourzat,
and Pingali, 1992). While stall-feeding does not guarantee effective use of manrue, it makes

it much easier.

Table 8-1pre5ents three indicators of stall-feeding: percentage of households who
reportthatthey keeptheiranimals fullyorpartly in stables, percentage reportingthatthey
have adopted permanent stabulation, and percentage reporting that they rely more on fodder
thatthey cutandcarrytotheiranimalsthanonnaturalpastme. Eachofthese indicators
shows more stabulation on small firms. But even on the largest firms one-half stabulate
partly, and one-fifth keep all their animals permanently in stable.

Longitudinal evidence about the trend from grazing to stall-feeding is also clear.
During the colonial period cattle were nearly always grazed and no tadition of fodder crop
production exists (Bart 1993). In the 1984 survey, one-fourth of cattle-owners reported that
they kept their cattle in stable (corral), which suggests that stall-feeding has been increasing
also during the 19805. Bart (1993) reports that stables, especially those with tin roofs, have
become signsofmodernity,inpartbecauseoftheincreasingriskthatcattlethatarenotkept
confined damage crops.

While small farmers on average have more livestock per unit of land even when
averaging over all households and not just those with animals, two out of five households in
the smallest firm size category did not have any livestock in 1990. Clearly, lack of fodder is
not the reason for not having even one goat. Literature provides two other explanations.
First, due to poor veterinary services, livestock mortality in Rwanda is exceptionally high,
making animals quite risky investnent for very poor households (Rwamasirabo 1990).

183

Second, many households in Rwanda are probably too poor to buy livestock and have no
access to credit, except possibly at excessively high rates (Bart 1993). That lack of liquidity
rather than lack of fodder or poor profitability of livestock plays the key role is illustrated
by the common practice in which poor households agree to take care of the animals of their
neighbors with no other compensation than the manure.

The tend towards stall-feeding is compatible with the banana-based intensification
pattern discussed in the preceding chapters. Banana groves produce large quantities of
leaves that can be used for fodder. Moreover, the supply of banana by-products is not nearly
assessonalasthatofmostotherfodderresomees. Thismatters, sincetypically itisnotthe
total annual fodder availability but instead the dry-season minimum that limits livestock

production.

On the other hand, the use of banana leaves for fodder competes with their use as
mulch, which is the key to the environmental sustainability of banana groves. The t'adeofi‘
is obviously less severe, if banana leaves are fed to animals only during the dry season,
when grass availability is at the minimum.

Another link between stall-fed animals and bananas is through manure, which in
Rwanda is most often used for bananas and crops intercropped with them (Bart 1993;
Kotschi etal. 1991).

In sum, intensification based on labor and not on purchased inputs characterized
both what smallholders in pre-genocide Rwanda did in their livestock practices and what
they did in their cropping practices. The tends were compatible, even mutually
reinforcing. The much-lamented overstocking appears to have had been oversold, in part
because both the intensification of cropping patterns (banana gardens, tee planting) and
the establishment of vegetative anti-erosion stuctures were increasing the availability of
fodder.

184

8.3 Income Diversification
In the two preceding chapters we have already presented some data on the role of

rural off-farm incomes. In this section we revisit these findings to ask how continued
population growth will affect households’ opportunities to find ofi-firm employment, and
what, in turn, are the implications for the agricultual intensification process discussed in
the preceding chapters.

Several previous studies that have looked at some regions of Rwanda have
emphasized that off-farm employment already is an important source of income for rural
households (von Braun, dc Haen and Blanken 1991). The numbers presented above show
that this is also We for rural Rwanda as a whole. According to Table 6-5, labor sales
accounted in 1990 for 22 percent of aggregate net household incomes in Rwanda’s
banana zone (22 percent in rural Rwanda on average) and for almost 40 percent of the net
household incomes of the quartile with least land per adult equivalent. However, as
discussed above, these numbers are effectively weighted by income, and therefore are
stongly influenced by a small number of high-income households with members
employed as teachers or other civil servants.” Most households rely much less on labor
sales, and, as shown in Table 6-3, the mean share of income coming fiom labor sales was
only 16 percent in 1990.

For all but the richest quartile of households, unskilled labor was a larger source
of income than skilled labor, and in the poorest quartile agricultual wage labor was
typically more important than all skilled and unskilled non-agricultural jobs combined.

Thus, for the poorest households, access to agricultural wage employment is crucial for
food security.

 

’9 By the standards of rural households, civil servants were well paid in Rwanda in 1990.
According to the DSA data 5 the mean daily wage for civil servants was
approximately five times that or unskilled farm and non-farm workers. Also masons
and carpenters earned 2-4 tines as much as unskilled laborers.

185

Unfortunately, this access is likely to decline under demographic change. As
shown in Table 6-12, sales and purchases of agricultural labor are stongly associated
with firm size. The bulk of the demand for agricultural labor comes fiom the top firm
size quartile, whereas households with the least land per person are typically net sellers of
rmskilled labor. The longitudinal implication of this cross-sectional observation is that,
other things being equal, the subdivision of firms under population pressure reduces the
demand for and increases the supply of unskilled labor. Without either major land-saving
and labor-using innovations in agriculture or rapid increases in non-agricultural
employment to ofi‘set these tends, either real wages for unskilled workers will decline or
rural rm(der)employment will increase.

Moreover, many of the non-agricultural jobs directly or indirectly face the same
predicament. To the extent that subdivision of farms makes households poorer and less
self-suficient in food, it reduces the rural demand for the services of masons, carpenters,
and tailors, among others. And to the extent that smaller firms pay less direct and indirect
taxes, the capacity of the state to employ teachers and other civil servants may decline. In
other words, without technical progress in agriculture, population growth threatens to
impoverish poor households not only by allocating them less land but also by turning the
process of agricultural growth multipliers in reverse (Haggblade, Hazell, and Brown
1989; Mellor 1986; Mellor and Desai 1985).

Reduced opportunities for wage labor increase further the pressure for labor-based
intensification in agriculture. The cross-sectional intensification patterns discussed above
may well understate the upcoming changes. To the extent that tomorrow’s land-scarce
firmers have less access to off-firm incomes than today’s small farmers, they will be
forced to practice more intensive agriculture than is found today on tiny farms. Thus, the
second-round impacts of demographic change through labor markets make the story of
labor-based intensification towards banana intercropping even more credible. However,

186

as discussed above, this is not the only second-round impact; another one is felt through
prices. To the extent that demographic change reduces incomes not just among the rural
poor but also among the urban consumers, thereby forcing them to cut back on their
consumption of banana beer and cooking bananas, it may reduce banana prices and
undermine the scope for banana-based intensification.

8.4 Demographic Responses
Apart fiom agricultual intensification and the expansion of the non-firm

economy, rural populations can also respond to increasing demographic pressure by
migrating elsewhere or by adopting measures to limit population growth. These
demographic responses interact with the economic responses in different ways.
Therefore, we review here very briefly the nature of demographic responses in Rwanda,
drawing primarily on Olson (1994), and discuss the implications for agricultural

intensification and for the development of the non-farm economy.

Migration fiom densely inhabited highland areas to surrounding lowlands has
been a cental response throughout the region. Rwanda has a long history of sending
permanent and temporary migrants to less densely populated areas in neighboring
counties. During the early colonial period such emigration was frequently encouraged by
the colonial administators, who considered Rwanda overpopulated. After independence
Rwandan politicians continued to talk about the need to find more land for Rwandans in
neighboring counties, but as the neighbors were reluctant to host immigrants, little
international migration occurred.‘so

Most migration during the post-World War 11 period has been internal and, until
the 19805, mostly rural-to-rmal. The big picture was that young men and women moved

 

‘° Althoth some 200,000 tutsis left the county in the 19605, most of them considered
tlhge’rr‘rse ves rgfirgees, declared their intention to return, and eventually did so after the
genocr e.

187

eastward from the densely inhabited high plateau. In 1945-61 the large movements were
from Gikongoro to Butare and horn Ruhengeri to Byumba. After independence, most
migrants went to the rural areas of Kigali prefecture, often to areas that had previously
been reserved to the tutsi pastoralists. Since the end of the 19705, the eastern province of
Kibungo became the main destination for the rural-to-rural migrants. Rural-to-urban
migration was limited to the capital city, and between 1978 and 1991, the estimated
population of Kigali more than doubled to 230,000.

Also temporary migration plays an important role in Rwanda. Using the men-to-
women ratios ofthe 1991 census, Olson (1994) estimates that the main destinations of
short-term migration are the city of Kigali and Bugesera in the south of Kigali (nrral)
prefecture. Interestingly, Kibungo, the most sparsely inhabited region, did not appear to
have as many temporary migrants as the earlier migration destinations. In part, this may
be associated with the stong reliance of Kibungo on cooking bananas, which require less
labor per hectare than most other crops.

Fertility rates remained high until until the late 19705, but then started to fill
rapidly. According to the 1991 census, Rwanda’s total fertility rate had declined to 6.2,
which is about 30 percent below the rate computed based on the 1978 census results
(Grosse 1994).

While the spread of modern methods of contaception undoubtedly contibuted to
Rwanda’s fist fertility decline, the main explanation was the increasing scarcity of land
and other Opportunities, which postponed the marriages of the young and raised the
interest in fimily planning. The steepest fertility declines were recorded in the western
prefectures (Gisenye and Kibuye), where the pressure on resources is strongest. In
contast, fertility rates had barely declined in the relatively sparsely inhabited eastern
prefectures (Byumba and Kibungo). News fiom the post-genocide Rwanda reports that

188

birth rates are up again, presumably because many households are responding to the
losses of family members by having more children (The Economist 1997). Together with
the return of hundreds of thousands of “old-caseload” refugees, this suggests that
demographic pressure in most of Rwanda is about as high as it was before the genocide

and increasing faster.

In sum, demographic responses have clearly reduced the pressure on resources in
the most densely inhabited parts of Rwanda and given households more time to intensify
their firming and diversify outside agriculture. In many areas of Rwanda’s historic core,
migration has fully compensated for natural population growth, keeping local pressure on
resources at bay. However, now that the unoccupied frontiers have been settled, the
pressures to intensify and diversify are getting stonger. Even if fertility rates resumed
soon their rapid downward tend, Rwanda’s young population would continue to grow

fast.

The key agricultural implication of the past demographic responses has been a
shift towards banana-based farming systems. Hundreds of thousands of people have
moved to areas where beer bananas are the main cash crop and, less universally, cooking
bananas the main food crop.

Although the eastern frontiers are now settled, the tend towards the areas
dominated by bananas is likely to continue. Farms in these areas are still larger and more
fertile than those in most other parts of Rwanda. As a result of above-average birth rates,
some immigration from more densely populated areas, and less outnigration to cities
than from other rural areas, the share of Rwanda’s rural p0pulation living in these areas

will increase.

189

8.5 Conclusions
In this chapter, we have seen that although there are number of other responses to

demographic change in addition to the land use and marketing responses discussed in the
previous chapters, they do not appear to change the situation fundamentally. The big
picture is still that of labor-based intensification and increased reliance on banana
intercropping and on the market for food. Despite all the gradual changes, firmers are
losing their battle to survive on the hills of Rwanda. Without significant technical
progress in agricultue and/or major increases in non-agricultural income earning

opportunities, food secmity in rural Rwanda is set to deteriorate.

190

Chapter 9

POLICY IMPLICATIONS AND SUGGESTIONS FOR FUTURE RESEARCH

9.1 Summary of Results

None of the findings summarized below is entirely new. All of them have been
stated, guessed or documented somewhere in the literature on Rwanda’s agricultural
development. However, the main findings are contary to what some of the most 10 (a W l: :1 .1
knowledgeable observers of rural Rwanda have recently stated or what is the A +36%“ x9” “1)“:

V) 1
conventional wisdom based on which policies have been or are formulated. \ U. .\ '
\r 51’ ‘

9.1.1 Land Use
The key finding of this study is the association of beer bananas and banana

intercropping with land scarcity. National cross-tabulations fail to show this link because
firms for various reasons happen to be large in the areas most suitable for bananas. But
within regions, small farmers rely more on beer bananas than large farmers. The
conventional wisdom that land scarcity pushes farmers to switch from beer bananas to

weet potatoes and other food crops gets no support from cross-sectional data. Instead,

pulation growth is increasingly converting Rwanda’s hills into intercropped banana

gardens, where food production occurs under bananas. While some types of intercropping
are labor-saving methods used in land-abundant conditions, Rwanda’s intercropped
banana groves are not. Instead, they are a form of agricultural intensification, helping
people to substitute labor and biological capital for increasingly scarce land.

In a complete contest to bananas, sweet potatoes are favored in densely inhabited
areas, but within the areas small farmers allocate no more land to sweet potatoes than do

eir neighbors with more land. This occurs despite the fact that sweet potatoes have high

191

l caloric yields and are the main staple for the poor. Clearly, small firmers do not seek Q)
.H’ food self-suficiency at all cost but instead respond to the superior returns that beer
‘1 _W

X' bananas can provide §0Jtheir increasingly scarce resource.
34v C t. \l H0 A R
Apart from bananas and sweet potatoes, our results confirm the conventional f1 k); __ ,/

I. ~
g wisdom. Land scarcity pushes households to expand cultivation at the expense of fillow ‘1“
OJ)” and pasture. leans and maize, along with W are the crops that expand most. J

-\

One policy lever that appears to have a significant impact on land use is physical (11:5)
{I infiastucture. In areas close to paved roads, firmers grow more bananas and less cereal:

 

 

71% (6 Cf“; and cassava.

/ \

 

{0” Although the expansion of cultivation at the expense of fallow and pasture
increases the exposure of Rwandan hillsides to water erosion, the expansion of bananas
offsets the negative impact at least in part. It is not clear whether population growth is
making land uses more or less erosive. Dense pure banana groves protect the soil well
against erosion, but the protective cover provided by banana intercropping varies greatly

depending on management practices. The officially promoted practices have clearly been

 

 

 

more erosive than the taditional practices followed by farmers. my; ,1 \(‘c

\V"

The transition of Rwandan smallholders to banana intercropping only applies to
the agroclimatic areas that are suitable for bananas. It does not hold for the highest
@ altitudes (above 2000-2100 meters) that are too cool for banan£_ On temperate #

 

 

 

highlands, population growth is unequivocally making agriculture more erosive and less
sustainable.

9.1.2 Market Participation
The findings on market participation are corollary to those on land use. Most

significantly, when agroclimatic variation is contolled for, the net sales of bananas (as
suchorasbeer)perunitoflandarehigheronsmall farmsthanon large firms. This

192

t
‘I
1

@ suggests that as population growth increases land ’scarcity, the market supply of banana 7B I!

 

beer will increase, anommonly assumed. Coffee sales per unit of land, it

 

contast, tend to decline with firm size when agroclimatic variation is contolled for. This 7‘.
impliesthat; tiger things being equal, demographic pressure would tend to reduce y
Rwanda’s coffee exports.

 

Dim/‘1 gift: The sales and purchases ofall food cropsare strongly associated with land

7%, ”A scarcity. On average, small farmers are net buyers and large firmers net sellers. Other
things being equal, population growth reduces market supply and increases market 7 ‘1
demand for all food crops.

t)

In other words, the cross-sectional patterns show that firmers are Wag
to land scarcity by giving priority to food. They are Mintensifying production by

Hi... _~_..-—

g
intercropping bananas and food crops. As a result of this intensification, both subsistence X I
W

 

 

 

 

 

was .r. «twat-1W

food production and beer production for sale increase per unit of land, while both decline
per person and per household.

As expected, households with better access to markets were found to rely more on 1 l 0
the exchange of banana beer for food.

9.1.3 Bananas and Food Security
Despite the waste of calories associated with the brewing of bananas into beer,

bananas provide more than one-fifth of the calories consumed in rural Rwanda. Although
the most food-insecure rural households rely more on sweet potatoes and other cheaper
sources of calories, the direct consumption of cooking bananas and banana beer is not
unimportant for them. On average, bananas account for more than one in ten calories
7’} consumed by the poorest households. In the eastern parts of the county cooking bananas
"/ are the principal staple food for most households.

193

6’9

Much of the contibution of bananas to food security comes through the exchange
of banana beer for food. Bananas are the single most important cash crop and second only
to labor sales as a source of cash. The cash incomes from bananas are widely distibuted
and no more concentrated among the better-off households than incomes in general. The
poorest households use much of their cash incomes to buy food, especially beans, which

are a valuable supplement to their protein-deficient diets.

Our computations suggest, given existing prices and yields, that many land-scarce
households can get more beans by growing bananas, brewing beer and exchanging beer
for beans than by growing beans themselves. Moreover, bananas are less vulnerable to
droughts than annual food crops and protect the soils better against erosion. In other
words, while the use of scarce land for beer bananas reduces food availability at the
national level, it improves the food security of many households by increasing their

incomes and “exchange entitlements” to food.

The argument that demographic pressure is pushing households to substitute
cooking bananas for beer bananas gets no support from our data. Unlike beer bananas
which have high yields and tade at low prices, cooking bananas are a much more
expensive staple, presumably because they are more demanding and/or yield less.
Effective demand rather than land scarcity appears to be the explanation for the expansion
of cooking bananas in peri-urban areas.

Since independence, population has grown much faster in eastern Rwanda where
bananasarethemainstaple,thanintherestofthecounty.Aseasteranandastillhas
much larger farms and higher household incomes and therefore can cope with additional
people better than other parts of the county, the tend is likely to continue. This will tend
to increase the national importance of bananas. Apart fiom that, the future role of bananas
in Rwanda’s food security equation will depend crucially on what Rwandans can afford

194

to consume. If Rwanda’s economy grows and if the benefits of the growth are shared
broadly, cooking bananas and banana beer will in the medium term become more
important than they are now. If, however, population growth is accompanied by
economic stagnation, the demand for traditional banana products will decline, eroding the
economic basis of the strategy of exchanging banana beer for food. Depending on how
much and how persistently poor Rwandans dislike beer bananas, they would either
substitute root crops for bananas or learn to accept beer bananas as a food crop.

9.2 Update on Recent Developments
The results summarized above were obtained primarily from the analysis of a

large nation-wide survey data set collected in 1989-1991. Although no comparable data
set exists for later years, Rwanda has undergone so dramatic events in the 19905 that we
cannot go from 1990 data to 1998 policy implications without examining how the
situation may have changed since 1990. Drawing primarily from World Bank (1998), we

“‘—
list the following developments:

1. Although close to one million lives were lost during the events of 1994,
Rwanda’s 1998 population is back to the pre-genocide levels. Consequently, rural
population densities are now considerably higher than they were in 1990, when the data
set used above was collected. Hence, the question of how households respond to

increasing land scarcity remains relevant

2. In cont-ast to the population, agriculture has not recovered fully. A large part of
the physical and organizational infrastructure for agricultural services was destroyed in
1994 (Kumar et al. 1996). In early 1998, production was estimated to be at some 70
percent of the pre-genocide level. Large quantities of food aid are brought in to fill the
gap and distributed primarily through food-for-work projects. Although many relief and

19S

development projects have distributed farm tools and seeds, the shortage of sweet potato
vines and other planting materials is still acute in many areas.

3. Agriculture suffers from a severe labor shortage, primarily because hundreds of
thousands of adults were killed in 1994, and because some 100,000 Rwandans are still in
prison, most of them suspected genociders awaiting trial. Restrictions on labor
movements limit the ability of labor-deficient areas from getting workers from labor-
surplus areas. Further, food-for—work projects draw labor from agriculture. Much land lies
fallow because households do not have enough labor to cultivate it or because they
participate in food-for-work projects. However, the entry of young people into the labor
force can be expected to turn the labor shortage into a labor surplus in a few years.

4. The security of land tenure remains uncertain, and few households have started

to make long-term investments such as the planting of perennial crops (coffee and
bananas).

5. Cropping systems are largely similar to what they were in 1990. The most
significant difference may be the expansion of root-rot-resistant bean varieties. Relative
to most other crops bean prices have somewhat declined. Sweet potatoes are still the
dominant cheap staple for the poor, and cooking bananas are still a more expensive
source of calories than cereals. Information on beer bananas or on banana beer has not
been collected.

6. World cofl‘ee prices have increased, but higher prices have not been passed on
to farmers. Like in 1990, coffee does not appear to be a lucrative cash crop, except for
those who could not get bulkier products to the market. Many vehicles used to transport
agricultural products were destroyed in 1994 and have not been replaced.

I96

7. Large cattle heards were brought by returning “old-caseload” refugees from
Uganda to Eastern Rwanda, but elsewhere livestock were decimated by the genocide and
the ensuing war and dislocations, and have not recovered.

9.3 Implications for Policy
In a predominantly rural country, almost all areas of public policy have some

connections to how farming systems are coping with rapid demographic change. The
focus here is on implications for agricultural policies, broadly understood.

9.3.1 Agricultural Research and Extension
The main policy implication of this study concerns the very high contribution that

technical progress on bananas, especially an edible bananas, would make to food
security and environmental sustainability in rural Rwanda. As summarized above,
binanas are the single most important crop in terms of land use and income and a key to
soil conservation on the steep hillsides of Rwanda. Since bananas can produce more
undome and food per hectare than any other crop in most part: of Rwanda, their role has
expanded with population growth and could potentially increase much further. However,
banana production suffers from severe pest and disease problems and without new
solutions to these problems, banana production in general and cooking banana production
in particular could decline dramatically, as it has done in some nearby areas in
neighboring countries. This would have dire consequences to food security and

environmental sustainability.

The substantial benefits of technical progress are a strong argument for giving
bananas a high priority in agricultural research and extension, provided that agricultural
research and extension on bananas are neither much more costly nor less likely to
produce results than research and extension on other crops. Experiences fiom other

countries suggest that research and extension on bananas have a high likelihood of

197

\

producing significant results at reasonable costs. Technical progress in the export-
oriented banana plantations in Central America and elsewhere has been spectacular. The
stock of methods and results (including germplasm) that researchers focussing on the
semi-subsistence banana groves of Afi'ican highlands can apply and adapt is substantial.
Researchers in the region already have some interesting results in the pipeline. Moreover,
technical progress in smallholder banana production does not rely on industrially
produced fertilizer to the extent that productivity increases in cereals typically do. Given
the high costs of importing fertilizer to Rwanda, this is a significant advantage and one

argument for prioritizing bananas in technology development and transfer.
W

Apart from bananas, Rwandan agriculture is increasingly dominated by two crops:
sweet potatoes are the most important source of calories while beans are the main source

of proteins. Together, these three crops provide more than two-thirds of calories, proteins

‘3 and cropping incomes in rural Rwanda, and for the poorest households, their shares are

l

\ even higher. To a large extent, the stagnation of Rwandan agriculture and the

deterioration of the food security situation in the 1980s and early 19903 resulted fi'om the
LR of technical progress in these three crops. Although detailed appraisals of research
projects are needed before definitive conclusions can be drawn, it is a good first
approximation that returns to research investments are likely to be large when they focus
on widely grown crops so that innovations can potentially have much impact. Moreover,
the gap between current average yields and on-station experimental yields is huge in
Rwanda, suggesting that much potential for technical progress may exist (Ndamake
1984a, 1984b). Thus, our findings suggest that Rwandan policymakers could increase the
contribution of agricultural research to food security by ensuring that it has a strong
focus on the three main food crops.

Note that the argument here is based on expected payoffs, not on any anti-market
“food first” approach. Both sweet potatoes and cooking bananas are bulky and poorly

198

transportable products and therefore essentially non-tradable wage goods. If their
productivity stagnates in the context of rapid population growth, the ensuing higher prices
will not only hurt the net buyers, including most poor rural households, but can also
choke of? growth in other parts of the economy, especially in the coffee sector. Major
increases in productivity, in turn, could release resources for other industries, including
export crop production. To a lesser extent, the argument also applies to beans, since
although they are more tradable, transportation of imported beans to rural areas is
expensive and a shift fi'om locally importable to locally exportable entails substantially
lower local prices. '

The other side of the coin is that none of the three main crops can pull Rwanda
out of poverty. All three are essentially local wage goods with little direct potential for -
ego: earnings. Theflbetter the researfic‘hfand extension” efforts targeted at them succeed,
the mommempfiws of the bulky food products decline, and the stronger becomes the
case for promoting productivity increases not only in the main food crops but also in
export-oriented activities that could take advantage of the resources released from food
crop production (Pearce 1990). Above all, this reasoning puts coffee on the agenda of
agricultural research and extension. Although not nearly as important for rural households
as the three main food crops, cofi'ee is Rwanda’s main source of export earnings and the
only widely grown export crop in the country. In the conditions that prevailed in 1990 (as
well as in those that prevailed in 1998) coffee was not a particularly profitable crop,
which raises questions on whether farmer interest in replanting coffee could be
sufficiently high to bring improved technologies to the fields and to justify investments in
research and extension. Obviously, the payoffs depend crucially on how good the
improved coffee technologies are, but positive returns are also more likely if the efforts
advocated above are already bringing major productivity increases (and price decreases)
to the poorly tradable bulky food crops (e.g., sweet potato and bananas), if open borders

199

"‘_——v

 

help to lower the prices of the more tradable food crops (e.g., cereals and cassava flour),
and if policies do not discriminate against the coffee sector.

Apart from coffee, similar arguments hold for many other agricultural activities
that currently play little role compared to the main food crops, but eventually will become
local or even national engines of rural growth. These range fi'om tea and horticultural
products to dairy operations and fish ponds. Each of these sub-sectors is more likely to
emerge and respond positively to investments if productivity in the food crops that now
employ and feed bulk of the population increases.

The suggestion to focus research more tightly on the three main food crops and on
coffee is not an argument for cutting funding on other lines of research. As the success of
white potato research in Rwanda shows, research on crops that are nationally of minor
importance can have high returns and contribute significantly to food security (Amani
and Haverkort 1986; Haverkort 1986a, 1986b; Blanken, von Braun, and de Haen 1994;
Monares 1984, 1987). In a predominantly agricultural country with poor prospects for
non-agricultural exports to pay for massive food imports, investments in the generation
and dissemination of agricultural innovations should probably be much larger and the
research agenda should even in the future include more than just the 34 main crops.
Nevertheless, the risk for Ereading research resources too thinly is real.

In most of Rwanda, the main competitors to the three main crops on the research
agenda are probably cereals (maize and sorghum) and cassava." Any of these crops
could probably become much more productive and gain ground if concerted efforts were

targeted at it.“2 Unfortunately, the expansion of maize and cassava could increase soil

 

" Also white potatoes and wheat com for the same funds, but they mostly target the
high-altitude areas, where none of e three main crops grows well.

‘2 The ensuing price reduction obviousl would offset part of the increased incentive to
$2“! these crops, but as discussed be ow, the extensrve informal cross-border trade in
se crops limits pnce changes.

200

losses on the steep hillsides of Rwanda, since neither crop provides much protection
against erosion. Although hedgerow intercropping systems, terraces or herbicide-
dependent minimum-tillage methods may solve the problem in the future, it is
questionable whether any of these will have much impact for many years to come. In the
nearby future, bananas will continue to play a central role in soil conservation, while
cereals and cassava will remain highly erosive. It should also be noted that cereals and
WWle than sweet potatoes and cooking bananas. Thissgggests

 

 

'W r-r— ,

that productivity gains in cereals and cassava would tend to have less impact on prices

 

..-__._._—a

WWW.“ Wit-MM“ ”yr—WW

fin."

and more impact on cropping patterns than those in sweet potatoes and cooking bananas.
Consequently, reflh on tradable food crops could hurt coffee production, wherease

research on non-tradable food ' likely to help it.

 

 

Another, closely related policy implication of our findings is that agricultural
research and extension should emphasize intercropping. As summarized above,
intercropping, especially with bananas, is widespread, is expanding with population
growth, and apparently is making the farming system somewhat more sustainable. Yet,
agricultural research in the past has done relatively little to improve intercropping
systems and agricultural extension has even discouraged some forms of intercropping.
Moreover, when intercropping has been studied, the focus has often been on non-banana
associations, none of which have not been adopted as widely as associations with

bananas.

Emphasis on intercropping implies focusing on the performance of the system
rather than on that of individual components in artificial isolation. This requires more
than adjustments to the “downstream” on-farm trials. Breeders, for instance, need to seek
varieties that make good components for the associations, rather than individual “star
players” intended to replace entire “teams”. If farmers typically grow their beans under

201

bananas, the screening trials of beans should happen in similar conditions, so that

tolerance to shading and other interactions get included among the the selection criteria.

One important crop association that has almost completely been neglected in
research and discouraged in extension, is the intercropping of cofl‘ee and bananas.
Belgian colonial administrators banned cofi‘ee intercropping in Ruanda-Urundi, arguing I
that coffee does not need shading at high altitudes. However, farmer practices in the
highlands of neighboring countries suggest that declines in coffee yields are more than
ofl‘set by the value of bananas and by the labor savings achieved when mulching
materials are produced right next to the coffee bushes. Farmers clearly are convinced
cofi‘ee and bananas perform better in combination than on separate fields. Both in Kag
in Tanzania, in Kivu in Congo, and in southern Uganda farmers intercrop most of their
cofi’ee with bananas, some of it at altitudes comparable to Rwanda. Indeed, to the extent
that the policies forcing farmers to monocrop coffee have now come to an end,
intercropping may well be already returning and, depending on prices, may in the long
term be the only form of coffee production attractive to farmers.63 Thus, research on and
extension of coffee-banana intercropping may be crucially important for Rwanda’s export

' gs.“ Or, if the government of Rwanda is to rehabilitate cofl'ee production in one
way or another, intercropping may be an option that provides the needed export earnings

at lower opportunity costs than forced monocropping of cofl‘ee. ,..——--—t

 

‘3 In early 1994, we observed how farmers in Butare and Byumba were starting to
intercrop food crops with coffee. We did not see any bananas planted between cofi‘ee
trees, but this may reflect mostly the fear of farmers that the ban on intercropping might
soon be reintroduced and the bananas planted between coffee trees destroyed.

“ Given that (1) coffee trees suffers greatly from excessive shading but can actually
benefit from a modest de of shading, and that (2) the amount of work need to
transfer mulching materi s fi'om banana groves to coffee fields depends largely on
distance, an option that 3 worth exploring is a compromise in which bananas and
coffee are not fully mix ut spaced in alleys.

202

9.3.2 Soil Conservation
Our results suggest that enhancing the contribution of bananas for soil

conservation deserves high priority in Rwanda ’s soil conservation policy. Although the
shortcomings of the old engineering-based soil conservation approach as well as the key
role of bananas in protecting Rwandan soils are widely acknowledged in literature and
although several researchers have advocated greater reliance on vegetative measures,
there has been surprisingly little interest in how the environmental contributions of
bananas could be increased. In part, bananas may have been neglected because of the
incorrect assumption that they would (or, in the case of beer bananas, should) inevitably
lose ground under demographic pressure. Another reason may simply be the failure to
realize that although the traditional dense banana groves conserve soils well, the mulch

produced by bananas ofien contributes little to erosiontcontrol on intercropped banana

groves.

This study has, noted many developments that argue for focussing on bananas in
soil conservafiofifihe old approach that relied on ditches, grass lines, and hedges
leading to the formation of‘ “ progressive terraces” was unpopular, inefl‘ective on steeper
slopes, and largely ended in the early 19903 as the government lost its ability to enforce
it. econ in general and intercropped bananas in particular are gaining ground
under demographic pressureO, cultivation is expanding on steep slopes, where
perennial crops may be the only way of protecting the soils@ declining land
fertility rather than erosion is the main soil conservation issue in Rwanda, and bananas
can alleviate both problems. This is in great contrast to the engineering solutions that at

best do little to maintain fertility and at worst make the problem worse.

(.1! In principle, the contribution of bananas for soil conservation can be increased in
' three difl'erent way First, the expansion of bananas can be promoted by research and
extension as discussed above, or by favorable prices as discussed in the next section.

1

203

Secon better use could be made of the mulch produced by bananas, especially in the
intercropped banana groves. Although research on this has been minimal and only
preliminary results (from Burundi) are at hand, it appears that relatively small changes in
banana management practices could greatly reduce soil erosion in intercropped banana
groves (Rishirumuhirwa 1992; 1993). The proposed changes may not be ready for .
nationwide extension, but they are sufliciently promising to warrant further research .
efi'orts. .

® banana production can be targeted at areas where the good vegetative cover
it provides is most valuable. This has implications for priority setting in research,
extension, and rural infrastructure. Improved banana varieties would contribute most to
soil conservation if they were targeted at the erosion-prone hillsides, where soils are often
degraded and acidic and temperatures often too low for current banana varieties to do
well. Soil conservation is an argument for focusing banana research and extension on the
production constraints that prevail on the hills. The same argument suggests that cereal
and root crop research should focus on the constraints that matter most in the valleys and
flatter fields.

Naturally, soil conservation is only one argument among many others. Since
cooking bananas are more demanding than beer bananas and since Rwanda’s prime
cooking banana areas (Kibungo) are relatively flat and not very vulnerable to erosion,
there may be a trade-off between soil conservation and food production objectives.
Banana research that targets erosion-prone bills, for instance by screening for edible
banana varieties that are less demanding or more tolerant to acidity, may not be the best
strategy to boost banana production, but it may be the most cost-efl'ective method of

conserving soils on steep hillsides.“

 

‘5 Comparable tradeoffs may also exist for other croafs. For instance, preliminary results
suggest that researchers could break through the titude (and temperature) barriers that

204

The suggestion to make a better use of bananas in soil conservation efl’orts can be
seen as a component of a broader approach of basing soil conservation more on
vegetative cover and less on labor-intensive structures. Under a variety of names this
approach is being studied and extended in many parts of the world (F ujisaka and Garrity
1988, 1989; Garrity 1989; Getahun 1991; KOnig 1992a, 1992b; Nair 1989b, 1989c;
Shepherd et al. 1996; Osiname and Tonye 1996; Palmer 1992; Paningbatan 1994; Roose
et al. 1988; Roose et al. 1992; Stark 1996; Tacio 1993; Young 1989, 1995). While the
economic viability of alley cropping and other forms of agroforestry may often be
questionable on level land, the potential of vegetative soil conservation measures on
sloping lands is considerable (Young 1989; Garrity 1989; Hoekstra 1994; Kessler and
Wiersum 1995; Jabbar, Larbi, and Reynolds 1994). And on steep slopes, perennial crops
or multi-storey “home gardens” rather than annual crops accompanied by vegetative

barriers are increasingly seen as the best option (Young 1995).

The principal alternative for soil conservation based on bananas and other
perennials is a return to the engineering approach, probably with a strong focus on proper
(“radical”) terraces. In the early 19903, Rwanda’s National Agricultural Commission saw
terraces, along with fertilizer subsidies, as the key to soil conservation and higher
agricultural productivity (CNA 1992a).

In contrast to intercropped bananas that have expanded spontaneously on
hundreds of thousands of farms, no Rwandan farmers have voluntarily adopted radical
terraces, which require large inputs of labor to build and maintain (Nduwayezu 1990).
Except perhaps for the volcanic highlands where bananas do not grow and where potatoes

 

now constraining cassava production on the mountains of Rwanda éMulindangabo and

Rollinl989). The immediate gains for food security could be signi cant, since many of
the hi -altitude areas are degraded watersheds where other crops tEzrf‘orm very poorly

and w ere many households are extremely poor. However, given t cassava provides

much poorer vegetative cover than sweet potato, the gains might not be sustainable.

205

are a lucrative cash crop, it is highly questionable whether radical terraces could spread

significantly without massive subsidies or strong coercion.

9.3.3 Tax and Trade Policy
A second set of policy implications arises fiom the findings that Rwandan

smallholders, especially those with least land per person, are highly dependent on food l

m'hfi-Auwafi- .- -

purchases financed by the sales of labor, banana beer, and cofl‘ee. These food security 3* V
(a... _ , ”I _-

.mn‘
E...
M.”h*-..‘ ‘aw_ w“...-_._.,

 

strategies are corEJgent on certain relative prices, and adverse changes in the policies

that afi‘ect those prices would tmdermine the access of some food insecure households to \
food. Wd, appropriate policy changes informed by these strategies could be
of considerable help to the poor.

The most straightforward implication concerns the direct and indirect taxes on
industrially produced beer. Although the primary justifications for high beer taxes in
developing countries are typically fiscal and administrative, in Rwanda their
distributional and allocative impacts have much merit compared to most other sources of

public finance. Beer taxes are paid by the relatively afiluent drinkers of industrial beer,

. but by increasing the demand for and the prices of traditional beers, they also transfer

incomes fi'om the (mostly urban) drinkers of traditional beer to the (mostly rural) brewers,
banana growers, and their employees, thereby. improving the food access of the poorest
rural households“ They subsidize environmentally benign forms of land use and

1.67

discourage the use of alcoho

 

“ Note that our results are specific to Rwanda’s banana zone, where most traditional beer
is made out of bananas that are a lucratrve cash crop for most rural households,
. including the r. The distributional impacts of taxes on industrial beer are less
equitable in g aces where traditional beers are made mostl out of cereals and where the
r are net uyers of cereals. Thus, the conclusions are 1' ely to hold also for much of
urundi and for the neighboring mid-altitude regions in Tanzania, Congo, and Uganda,
but not to areas that are either too cool (high altitudes) or too dry for bananas.

‘7 It is not clear, however, whether taxes on industrially produced beer promote public
health. Traditional beers sometimes include alarming uantities of hazardous
substances. The advocates of development assistance or the construction of public

206

On the other hand, high taxes on industrially produced beer discourage cofi‘ee
production by making beer bananas a more profitable crop. High beer bananaiprices have
been an unintended consequence of fiscally motivated taxation of industrial beer. The
heavy-handed promotion of coffee in Rwanda during the past decades can be seen in part
as a way of ofi‘setting this disincentive. As Rwandan policymakers seek to liberalize
agricultural policies, they are likely to pay more attention to economic incentives and
question the indirect subsidies of beer bananas created by high beer taxes.

However, lower taxes on industrially produced beer or new taxes on banana beer
would be not just a distributionally and environmentally questionable policy change but
also a rather blunt tool to promote cofl‘ee. At the relative prices that prevailed in 1990 as \
well as at those that prevailed in 1998, many other crops besides bananas appeared to )k

\x‘ provide higher returns to land and labor than were available in coffee production.

No similar case exists for the taxation of most other industrially produced or
'f imported goods. Sugar and vegetable oil, for instance, rather than competing with
( environmentally benign crops sold by most poor farmers are themselves products that

   
    
  

most poor households buy. Taxes on gasoline and vehicles are also partly paid by the

rural poor, since they increase marketing margins, thereby reducing the prices farmers
W m
receive for their products and increasing those they pay for the non-local products they

 

buy. Border taxes in general change the terms of trade against export producers, which
ests that many of the poor coffee growers are adversely affected. All these taxes may
bejustified on fiscal grounds, but beer taxes are unique in that the majority ofthe rural
poor who are net sellers of banana beer gain fiom them. It is true, though, that the biggest
gainers fiom the higher banana beer prices brought about by taxes on industrially

 

breweries in Afiica and elsewhere have often depicted traditional brewing as a public
health hand, but this may be enggerated (Kortteinen 1989a, 1989b . Nevertheless,
the health benefits of lower total cohol consumption brought about y hi beer taxes
are t1partially offset by the shift in consumption fiom industrial brews to 'tional ones
wi more variable quality.

207

%D

’\°\

produced beer are relatively afiluent rural households with large farms that are suitable
for bananas and well connected to wider markets.

Being net buyers of beans, cereals, and cassava flour, the poorest segments of
rural Rwanda depended in the early 1990s crucially on trade policies. Since the cross-
border food imports were largely informal, trade policies influenced prices not so much
through the formal tarifi‘ rates but through the decisions that allowed the borders to be
relatively porous so that the traders could import food from the neighboring countries,
primarily Congo. During the 1990s, eastern Congo may have lost much of its ability to
produce surpluses for export, but at the same time maize and cassava production in
southern Uganda has increased dramatically. Open borders to the imports of basic staples
and to the exports of the goods that pay for them contribute significantly to food security
'n rural Rwanda, and also promote environmentally sustainable land uses on the steep

illsides by lowering the prices of erosive crops (cereals and cassava).

As discussed above, fertilizer subsidies would probably trigger environmentally
unfavorable land use changes, since highly erosive crops, such as maize, happen to be
more responsive to fertilizer than less responsive crops, such as bananas and sweet
potato. This is not to say that chemical fertilizer is environmentally harmful in Rwanda.
On the contrary, chemical fertilizer can increase vegetative cover on impoverished
hillsides and thereby reduce erosion. The success of introducing agroforestry on degraded
lands may depend on access to fertilizer.“ Moreover, yield increases brought about by
fertilizer may reduce the pressure to bring new land under cultivation. Nevertheless, the

 

“ More precisely, the yoffs to trees and cover crops should increase with improved
access to or redu rice of nutrients other than mtrogen. With respect to mtro en, the
picture is more comp ex, since .if the trees and. cover crops are leguminous and r the
abili to fix nitro en plays an important role in their benefits, payoffs to agroforestry
coul actually dec ine with 1m roved access to (reduced price of) nitrogen ertilizer
(Akyeampong and Hitimana 1 6).

208

prospect of adverse environmental impacts through unfavorable land use changes

weakens the case for fertilizer subsidies.

9.3.4 Infrastructure
The findings of this study on farm-level land use patterns support the argument

that could be made a priori based on economic reasoning and information on rural and
mban consumption patterns: In Rwanda, investments in rural infiastructure can promote

environmentally sustainable agricultural intensification

The key to this conclusion is the association of banana production with roads.
Given the bulkyness of bananas and the perishability of banana beer as well as the role of
bananas as the main cash crop, it is not surprising that farmers with good access to roads

 

 

and hence to urban markets are more likely to produce bananas for sale andrely‘rflngrehon

-5" --H,m.__n_-. mud-cur...- n.-

_,._..--"- “a...

 

 

specialization increases farmers’ incomes and reduces erosion on their sloping lands.

Investments in rural roads and bridges would not merely allocate banana fields
from previously connected but more distant areas to the newly connected areas. Also food
consumption patterns would change. Urban diets in Rwanda are now strikingly different
from rural ones. Urban consumers eat much less sweet potatoes and bananas than their
rural counterparts, and much more cereals and white potatoes. Even controlling for
income, a large gap exists, presumably because bananas and sweet potatoes are more

/expensive relative to cereals and white potatoes in urban areas than in the rural areas.
" \ t This, in turn, depends largely on costly transportation, which is reflected in wide
9, marketing (gross) margins (T ardif-Douglin 1991). Investments in rural roads would

'.\ !

uce the prices of the rural staples in urban areas, but would have little impact on the
pfiWfied fiom abroad. As a result, there would be
some substitution of locally produced sweet potatoes and bananas for imported cereals in

209

Ar
\1

mbanareas,andsmallshiftsfi'omthebulky staplestocerealsandbeansinmralareas.
Since sweet potatoes and bananas have much higher yields than cereals and beans, the
reallocation of land from the latter to the former would increase total food production and

reduce food imports.

Since Rwanda already has good paved roads connecting prefectural capitals to
Kigali, the highest priority is likely to lie in feeder roads. By neglecting feeder roads y

 

while investing heavily in highways, policymakers in Rwanda and in donor agencies have
favored food imports (via the good paved roads fi'om Congo and Uganda) as well as

W

market-oriented food production in relatively distant areas alongthe paved highways

MN“. “W ~g‘ am I‘m ‘1' n

 

 

 

 

 

 

(bananas in the southwestern part of the county (Kibungo) and white potatoes in EC,

 

 

 

northwestern Rwanda).
9.3.5 Off-Farm Employment
Literature on rural Rwanda has repeatedly emphasized the need to promote off-

farm employment, not just to provide additional income but also to encourage farmers to
invest, commercialize and take risks in their agricultural activities (Ezemenari, 1993).

Our findings confirm that for the households with smallest farms and/or most

impoverished fields, labor sales play a crucial role in helping them to bridge the gap

between what they produce and what they need to survive. Unfortunately, the poorest sell ‘
mostly agricultural labor, and as farms get smaller with populationg growth, the demand 3
for it isdeclining. Finding ways to reverse that trend and/or to increase the demand for
unskilled labor outside agriculture is vital for the welfare of the rural poor.

Most of the discussion on how various policies and programs can best create non-
farm jobs is beyond our scope. Here we only focus on four issues suggested by our

results.

210

_ 0’9. the viability of banana-based processing activities in rural Rwanda should
be assessed. The processing of bananas and banana byproducts in Rwanda is almost
exclusively limited to brewing. The other small-scale processing activities that are
widespread in parts of Asia and also used in Afiica, are almost non-existent in Rwanda.
Bananas are seldom dried into flakes or chips or pounded into flower, although drying
could greatly improve storability and transportability. Yet, survey results suggest that at
least in Eastern Rwanda, surplus cooking bananas are sold at very low prices, and
sometimes even thrown away or fed to hvestock As Rwanda’s urban population grows,
the potential for processing activities that improve storability and transportability is
clearly improving. Although it may be premature to advocate the promotion of processing

activities, the opportunities appear worth exploring.

It might also be possible to reduce the use of scarce resources for banana beer by
improving the transformation process (Haggblade and Minot 1987). If combined with
taxes on banana beer, such an intervention, if successful, could improve the performance

of the sub-sector without increasing beer consumption.

Second, e dependence of the poorest quartile of rural households on the sales of
unskilled la r together with the disappearence of employment opportunities outside the
main agricultural seasons suggests a useful role for labor-intensive public works during
the summer months. At sufficiently low wages such works can be self-targeting and cost-
effective means of improving food security among the working poor (Braun, Teklu, and

Webb 1991). They cannot, however, target assistance to those who cannot work.

If properly planned and implemented, public works are not only social safety nets
but also productive investments. The list of possible activities range fiom the construction
of roads and fish ponds to soil conservation and tree-planting (Engle, Brewster, and

Hitayezu 1993). A particularly interesting target for the public works consists of various

211

investments in water control. In addition to employment during construction, such
investments could permanently increase the demand for labor during the season when
additional employment is most needed. Large-scale irrigation schemes in Rwanda and
elsewhere in Africa have been expensive and performed poorly, but the experiences fi'om
smaller schemes are often encouraging. Public works that use manual labor and simple
tools rather than heavy machinery to level land and to build dams, canals and reservoirs
are worth studying.

Third, he government of Rwanda should avoid high and rigid agricultural
minimum wages. Outside the busy seasons, Rwanda’s rural minimum wage was in 1990
far above the opportunity cost of most people’s time and also above the value of the work
for the employer. Although not universally enforced, minimum wage appears to have had
some impact, often limiting rural employment to those activities and periods where labor
productivity exceeded the minimum wage. This slowed down labor-based intensification,
which typically involvs low-return investment activities during periods when the
opportunity cost of labor is low.

As population growth reduces the average farm size, an increasing share of the
more productive work will be done by owner-operators, leaving less for the laborers. If
intensification activities combine labor inputs and (other) cash outlays, they may appeal
neither to the poor households who face a credit constraint nor to their wealthier
neighbors who face an artificially high price of labor. More generally, a minimum wage
and the limits it implies on what is worth doing send a fundamentally wrong signal in a

situation where the combination of demographic pressure, environmental degradation,
W v—A— --— ..

 

 

W‘~M// -. “’w‘" .--,,.u.- ‘

 

M M...

212

Undoubtedly some employees were earning higher daily wages because of the
minimum wage laws, but their total earnings may have been lower. F ull-time year-round
employment at minimum wage is a rare anomaly in Rwandan agriculture; instead of
having regular jobs, farm laborers get temporary assignments that often take only days
and seldom more than weeks. Thus, during less productive periods employers can
flexibly cut days worked if the law prevents them from cutting wages. It follows that
lower agricultural minimum wages, at least outside the busy months, would probably
increase employment, encourage intensification and promote food security in rural
Rwanda

,

'ums or/skills are high in rural Rwanda Much ofthe premium is due to the

\\

relatively high salaries paid to teachers, soldiers, and other civil servants, but also rural
households themselves pay often high wages for employees with specific skills. Masons

@ore investment in the development of skills should be considered. Wage

 

and carpenters, for example, typically earn two to four times as much as unskilled
laborers. The high private returns to formal education and on-the-job training implied by
rvthe wage diflerentials also suggest that social returns are high. The creation of rural non-
\ farm employment depends crucially on the viability of enterprises that combine the

t} efforts of skilled and unskilled employees and the high skill premiums suggest that in
/

t

Rwanda the availability of skilled employees is a serious constraint.

9.4 Suggestions for Future Research
Our discussion above on policy implications already covered the implications for

agricultural research and extension, emphasizing the great social value that technical
progress on the three major crops would have and arguing that also the case for coffee
research may be strong, especially ifsubstantial technical progress in food crops releases
resources fi'om the production of bananas, sweet potatoes, and beans. The suggestions in

this section focus on policy-relevant socioeconomic research. In other words, we ask

213

what are the researchable knowledge gaps that should be filled in order to inform policies
related to agricultural intensification in Rwanda. Some of these gaps have been
mentioned above as limitations and caveats of the results of the present study, while
others stem fiom anecdotal evidence encountered in the field, fi'om major changes that
have taken place since the data used in this study was collected, or from unsuccessful

attempts to use existing data to illuminate important issues.

Timely information on production and prices. This study has repeatedly
highlighted the dependence of rural households on the exchange of labor, banana beer,
and coffee for food. This implies that the ability of households to command food depends
not just on their food production but also on the prices they face. An illustrative case is
the one we observed in early 1994 in southern Rwanda, where the influx of refugees fiom
Burundi had reduced rural unskilled wages at the same time as a drought had increased
bean prices. Neither trend alone looked particularly alarming, but their combined impact
was that the amount of beans food deficit households could get in exchange of their labor
had declined to less than halfthe normal level. This shows that efforts to alert
policymakers to dangerous declines in household food entitlements must include timely

collection, processing, analysis and dissemination of data on production, prices, and
M

 

wages. Although the focus on timeliness that is vital for early warning monitoring is rare
F—m

in nationally representative farm surveys, combining the two efforts could have
significant benefits, not least because timely data processing can improve quality in data

collection.

Cooking bananas versus beer bananas. A major question raised by this study is
whether pest problems are forcing Rwandan farmers to replace their cooking bananas by
beer bananas. Pests have devastated many cooking banana-based communities in Kagera,
Tanzania, on the other side of Rwanda’s eastern border, but we have not found evidence

of any major transition fiorn cooking bananas to beer bananas in Rwanda. Perhaps the

214

 

best guess is that whatever shifts to beer bananas may have taken place in some areas
have been ofi‘set at the national level by the expansion of cultivation in Bastem Rwanda,
where cooking bananas grow well. As these relatively recently established cooking
banana groves are maturing, the pest problems are probably increasing, but just how
severe the problem is and whether it is forcing farmers to replace their cooking bananas
by beer bananas or other crops is not clear. One constraint for analysis is that the land use
statistics used in this study do not differentiate between cooking bananas and beer
bananas. Given the key role of bananas in Rwanda, especially in the eastern parts of the
country, improved understanding ofthwtude and the nature of the banana pest
prob@ed. This calls for collaboration between social scientists
engaged in nationally representative surveys and agricultural scientists working on the

pest problems in Rwanda.

A related issue concerns the direct use of beer bananas for food. As discussed
above, anecdotal evidence suggests that, in some parts of Rwanda, beer bananas have
becomeanacceptedpartofthe regulardiet. Inpart, this may beareactionto extreme
poverty, but is this the whole story? Are there other explanations such as innovations in
preparation? Are there policy interventions that could promote the use of beer bananas as
food. Given that beer bananas are considerably cheaper than cooking bananas and, if
considered edible, would provide cheaper calories than any other staple and more food
per unit of land than any other food crop, the answers to these questions might have
significant implications for food security. A case study on “the emerging custom of
Wg beer bananas” is cm

W " ""- “ Wm/x—HM

 

Rural income patterns. Numerous externally funded projects are seeking to
alleviate poverty in post-genocide Rwanda, and some of these have been designed with
an explicit assumption that efforts to target projects at poorest areas or at the poorest
households are not warranted, because rural poverty is rather universal. The income and

215

food availability computations used in this study as well cast considerable doubt on this
assumption as do many earlier studies (McAllister 1994). While most rural Rwandans

admittedly are poor by the standards of most outsiders, households with small farms, poor
/\
soils, many dependents and little access to ofi-farm incomes are far more likely to suffer

A.“

 

 

It..-“

from deprivation than others. In 1990, most households on the degraded hillsw 1n the

 

prefectures of Gikongoro and Kibuye were much more food msecure than most

m.“

households in the prefecture of" Kibungo. While the main patterns are likely to hold still,
much has changed during and after the genocide and new We
distribution and income/Enema is warranted. Given the great complexity and cost

involved in collecting data on household mcome, it might be advisable to develop
i, “relatively quick and relatively clean” methods that are sufliciently inexpensive so that

 

their use for the targeting of poverty alleviation efforts is feasible.

 

Impact of AIDS on agriculture. Population growth, increasing land scarcity,
declining farm size, and the increasing need to substitute labor for land were the starting
points of this study. While the expansion of AIDS in rural Rwanda is likely to somewhat
slow down rural population growth, it is not expected to stop it. AIDS is, however,
creating an increasing number of very poor households with both scarce land and
inadequate supply of labor (Hunter, Bulirwa and Kisseka 1993). Gillespie (1989)
suggests that in Rwanda increased mortality due to AIDS “might provoke a gradual shift
in cropping patterns away from relatively labor-intensive legumes, to tubers such as
potato, cassava and sweet potato” and is concerned about the ensuing declines in soil
fertility. Thus, land and labor poor households are of special interest not only fi'om the
point of view of food security and the ineffectiveness of labor-intensive public works in
alleviating this kind of poverty, but also matter from the point of view of environmental
sustainability and land use. Given that AIDS, which in pre-genocide Rwanda was still

216

largely an urban phenomenon has now greatly spread in rural areas, a special research
focus on the affected households may be warranted.

Labor markets and labor use in agriculture. Although most Rwandans are still
predominantly farmers, many rural households earn a significant share of their income
from ofl-farm employment. In pre-genocide Rwanda, skilled employment was a major
source of income for a small minority of households and the sales of unskilled labor
helped many of the most land-scarce households to achieve food security. Since 1994, the
situation has changed greatly. Two key policies in the current (1998) situation are severe
restrictions on labor mobility and major investments in food-for-work projects. The
Why probably hurt greatly the food sectuity of some pooflfluseholds,
whereas the food-for-work projmamed for drawing labor fi'om agriculture.
Improved understanding of how the rural labor market functions, what are the impacts of

 

 

 

 

 

the mobility restrictions and the employment projects, and what are the seasonal patterns

of labor use in agriculture could help in designing better interventions.

Intrahousehold allocation of beer income. Although we have documented that
banana beer sales are a major component of household cash incomes and that the most
food insecure households spend a large share of their cash income on food, it is not clear
who inside the household controls beer income, whether it is spent differently from other
types of income, and how the goods and services paid for by beer income are distributed
among household members. The suspicion of many is that beer income is largely
controlled by men and spent on beer and other non-necessities, and therefore contributes
less to household food security and children’s well-being than other types of income. The
assumption that incomes controlled by men tend to be less well spent than incomes
controlled by women has been supported by many studies and may now be considered to
be a “stylized fact” (Braun and Kennedy 1994). In contrast, the assumption that men
control banana beer incomes is more questionable. The knowledge gap lies in who

217

controls which incomes inside Rwandan households. Literature provides some guidance,
suggesting that women often brew and sell beer, that they probably have more control
over beer income than over cofi'ee income, and that men genérally control cash incomes
much more than the imputed incomes fiom subsistence food crops (Bucyedusenge et al.
1990; Miklavcic 1995; Martin 1987). While somewhat helpful, this does not tell whether
beer brewing mostly tranfers incomes fi'om male beer drinkers to female beer makers or
whether it more often moves resources from the female-controlled subsistence domain to
the male-controlled exchange economy where children’s food security gets less emphasis.
Since the food security implications of policy decisions such as whether to impose taxes
on banana beer sales depend on the answers, efl‘orts to examine these issues further may
be justified. Moreover, DSA might want to take another look at the consumption of
banana beer. Given that consumers typically report only a fiaction of their beer purchases
and that also the DSA data set used in this study clearly understates rural consumption, it
might be advisable to study beer transactions also from another angle by surveying
traders and the keepers ofrural bars.

The role of trees in agricultural intensification Although this study has focused
on changing land use, it has paid little attention to trees that increasingly are not grown on
separate woodlots but on the boundaries of fields or on contour hedges and hence do not
get normally accormted for in land use statistics.69 Yet, numerous studies argue that
increased planting of trees is one of the most dynamic features of the ongoing labor-based
intensification in densely inhabited areas such as Rwanda, and is crucial for
environmental sustainability and household food security (den Biggelaar 1994;
Gahamanyi I989; Nair 1989a, 1989b, 1989c; Shepherd 1996; Tacio 1993; Arnold 1995;
Young 1989; Chambers and Leach 1989; Egli 1985). Most socioeconomic and farm

 

‘9 DSA’s Agrofo Survey did collect information on trees but the data had many
problems that grea y reduced their usefulness.

218

management research on agroforestry consists of case studies. Few attempts to use
representative national farm household data to understand the role and the constraints of
trees in agricultural intensification exist. As the capacity for collecting and analysing
farm survey data is rebuilt in post-genocide Rwanda, it is probably advisable to pay more
attention to trees.

Paths of intensification Several students of Rwanda’s rural economy have argued
that interactions between crop choice, livestock, input use and soil conservation decisions
play key roles in agricultural intensification. According to one argument, households may
not be interested in investing in anti-erosion hedges or grass lines unless they have
livestock that can convert grass and leaves into valuable animal products. Similarly, the
profitability of input use may depend on soil conservation investments. Understanding
such interactions is crucial for finding most effective interventions. For instance, if
livestock really is a key to making hedges and grass lines pay, improved veterinary

services might be a cost-effective way of promoting soil conservation.

Despite the strong intuitive appeal of these arguments, attempts to study such
interactions with DSA’s cross-sectional data set have not been very successful. As
discussed above, Rwanda’s strictly enforced agricultural policies provide a partial
explanation. If, for instance, households are forced to plant hedges and grass lines
whether or not they have animals, the observed link between livestock and soil
conservation may be weak or non-existent. Another explanation can be soil quality,
which obviously greatly afl‘ects land use, soil conservation, and input use, but is poorly

controlled for in DSA’s cross-sectional data set.

The question to ask is whether a difi'erent research approach would allow DSA to
look at these interactions and the resulting intensification paths. Perhaps a detailed look at

219

 

 

the case histories of a sample of households could cast light on the interactions. Or

perhaps participatory group discussions with farmers could be useful.

Emerging success stories. While DSA’s nationwide sample survey is an excellent
tool for studying the state of Rwanda’s rural economy and major ongoing changes such as
the increasing reliance on banana intercropping examined in this study, it cannot tell
much about new innovations and emerging tends that touch only few respondents in the
sample. Yet, where the farming system is technically quite stagnant, policymakers are
interested in and deserve to be informed about the exceptional cases where technical
progress and/or investnents are taking place. Apart from the well-documented success in

M

white potatoes, some Rwandan farmers have interesting experiences with climbing beans,

 

 

 

mixed farming, agroforesty, horticultural crops, and the development of valleyfibkottoms.
Understanding the ingredients of these limited successes and examining the preconditions
for wider adoption requires that DSA’s work on national surveys is supplemented by case
studies of the emerging (or potential) success stories. Again, collaboration between social
scientists and agricultural scientists is needed to identify the cases that are worth studying
in detail.

Processing of sweet potato and banana into dried products. Two of Rwanda’s
three main food crops are bulky and perishable. As discussed above, processing them into
more storable and tansportable dried products is not common, presumably because it has
not been of much interest in the past, when sweet potatoes and bananas played a smaller
role in food supply and when most consumers were rural. With urbanization, the potential
contibution of processing is increasing. The potential for processing is probably greatest
in areas where current or potential production exceeds local demand and where prices are
considerably below national levels, because costly tansportation isolates the location
tom national markets. Promising opportunities may also exist in areas where production

and prices have unusually large seasonal variation. Where are these areas and what is the

220

magnitude of the gaps that processing could bridge? While expertise in processing is
needed to evaluate and intoduce innovations and investments to respond to the
opportunities, national farm survey data may be useful in understanding the nature and
size of the opportunities. Policies on processing could also be informed with demand-side
research, including studies on the potential demand for processed products m urban areas

w‘ ‘l mwfimm— " '“1 IMH‘

(Boughton, Reardon, and Wooldridge 1997; Kennedy and Reardon I994) "" " “W“

 

Liberalization of agricultural policy. The liberalization of Rwanda’s coercive
agricultural policy started before the genocide and presumably has continued thereafter.
Relatively little is known on how farmer behavior has changed as a result. Information on

the extent to which farmers have stopped mulching their coffee, are intercropping coffee

 

 

 

 

with other crops or are even abandoning coffee would be relevant for the design of coffee

 

 

policy. If, for instance, farmers were found to be widelywabandoning coffee in some areas
but not in others and if this was shown to be a voluntary response and not a reflection of
differences in the extent to which coercive cofi‘ee policies continue, implications for
where efforts to promote policy are likely to yield best results would be clear. Similarly,
the extent to which farmers still invest in soil conservation, maintain the terraces they
were once forced to build, or tear down the stuctures for short-term gains is not known
and has major implications for policy. The higher the degree to which the old
engineering approach to soil conservation has been abandoned by farmers, the more there

is need to develop and promote new measures such as the planting of perennial crops on
steep slopes.

Land tenure and land markets. During the field work in 1994 we repeatedly heard
about land tansactions, both sales and rentals. Some observers were convinced that land
sales were putting (or would put) land into the hands of more effective operators and
should therefore be encouraged, while others assumed, based on anecdotal evidence or
literature on what had happened elsewhere (Hill 1986), that most land tansactions were

221

distress sales by the poorest rural households and were creating a class of rural landless
workers. DSA’s 1990 data set has some information on land tenure, but major changes
have taken place since 1994 and substantial concerns exist on whether farmers now have
sufficient security of tenure to invest in their land. Given that such investments are
important both for household food security (bananas), export earnings (cofi‘ee) and
environmental sustainability (soil conservation), improved understanding of the situation
is needed to guide policy. Initial focus of data collection should probably be in examining
how insecure tenure is constaining investnents.7o In the long term, efl‘orts probably
should be targeted at understanding land tansactions. Policy implications obviously
differ considerably depending on whether land sales are predominantly made by those
with better non-farm opportunities and declining interest in farming, or by the poorest
households that sell to survive when crops fail or to pay for taxes, school fees, or

medicines.

 

7° Farmer perceptions of the situation are crucial since lack of secruity may even
encourage land-based investments if farmers believe that investments stengthen their
claim on the land they operate. Conseguently, policy details matter greatly and research
to mform policies can have high payo s.

222

APPENDIX A

REGIONAL PATTERNS OF RURAL EXPORTS

To disaggregate the national patterns of the net sales and gifts (rural exports)
presented in Chapter 6, the regional patterns are preser'rted in the next three tables. Three
regional classifications are used. The patterns for the five geographic regions are shown
in Table A-1 , those for the five agroclimatic zones are shown in Table A-2, and those for
Rwanda’s ten rural prefectures in Table A-3. Note that the numbers of observations in

some of the areas in the last two tables are quite small."

 

7' More 1precisely, the number of clusters alsecteurs) fi'om which the interviewed
house olds were selected was small in ese categories.

223

Table A71 : Net Household Sales and Gifts by Geographic Region

Hill. of units North- South- Center- Center- East Rwanda
West West North South
Beans -13 -10 -26 -1O -2 ~61
Peas -1 -0 -0 -0 -O -2
Peanuts -0 -0 -0 -0 0 -O
Soybeans -0 -O -O O -O -O
Sorghum -9 -9 -10 -6 7 -27
'Maize -2 -0 -1 -O 0 -3
Wheat -0 O 1 -0 -0 1
Finger Millet 0 0 0 -O O 0
Rice -0 -1 -1 -O 1 -1
Cassava -7 -15 -12 -2 13 -23
White Potato 58 -1 -16 -5 -6 31
Sweet Potato -3 -3 13 7 4 18
Taro -0 -1 0 1 -1 -1
Cocoyam 0 -o -0 -0 -0 -0
Cooking Banana -0 3 0 0 29 32
Beer Banana -11 -8 1 2 16 -1
Fruit Banana -0 0 0 2 3 5
Coffee 7 6 9 5 12 39
Banana Beer 28 16 40 38 94 216
Sorghum Beer 11 11 15 12 11 61
Ag. Labor 0 2 -2 -1 -10 -11
Non-Ag. Uhsk. L. 2 3 1 1 -1 7
Skilled Labor 5 3 16 3 10 36

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households
Units are kilograms, except for beer (litres) and labor (mandaye).

 

Table A-1 and Table A-2 confirm that the relatively sparsely inhabited eastern
lowlands, that are more drought-prone and were taditionally considered less productive,
are now Rwanda’s food basket. They produce surpluses in most crops, exporting not only
banana beer and cooking bananas, but also cassava and sorghum. As Table A-3 shows,
most of the surpluses come tom the prefecture of Kibungo. Unlike any other prefecture,
Kibungo even manages to be self-sufficient in beans. Farmers in eastern Rwanda are net
buyers of labor, providing employment to temporary migrants. Nevertheless, the numbers
are small, showing that even in the East most households operate family farms and hire
workers only to supplement family labor.

224

Table A-2: Net Household Sales and Gzfls by Agroclimatic Zone

Mill. of units Agroclimatic Zone
East High Congo-Nile Kivu Vblcanic
Plateau Divide Lakeshore Highlands

Beans 1 -32 -17 -7 -6
Peas -O -o -1 -0 -O
Peanuts 0 -0 -0 -0 -0
Soybeans o -o -o -o o
Sorghum 9 ~18 -14 -2 -3
Maize. 1 1 -1 -2 -1
Wheat -0 -0 1 0 .0
Finger Millet o -0 0 -0 0
Rice 1 -1 -O -1 -0
Cassava 8 -12 O -16 -4
White Potato -7 -16 -5 -9 67
Sweet Potato 3 18 -1 2 -4
Taro 1 -1 -O -1 0
Cocoyam -0 -O —O O 0
Cooking Banana 27 1 3 2 -0
Beer Banana 17 -5 0 -19 7
Fruit Banana 3 2 1 O -0
Coffee 11 14 1 12 0
Banana Beer 88 79 22 25 3
Sorghum Beer 8 27 21 -1 6
Ag. Labor -10 -3 2 -0 1
Non-Ag. Uhsk. L. -1 2 1 3 0
Skilled Labor 9 15 6 3 2

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households
units are kilograms, except for beer (litres) and labor (mandays).

 

The historic land of a thousand hills in central Rwanda exports banana beer,
coffee and some sorghum beer and sweet potatoes, and imports beans and sorghum, along
with some cassava and white potatoes. Center-South is considerably poorer than Center-
North, and although farmers there produce less beans and sorghum than their colleagues

in Center-North, they also import much less.

Farmers cultivating the watershed lands of the Congo-Nile Divide buy large
quantities of beans and cereals although their earnings from selling beer are low. None of

225

Table A-3: Net Household Sales and Gifis by Prefecture

Mill. of units BUT BYU CYA GIK GIS GIT EGO KBY RIG RUB
Beans -6 -10 -4 -6 -9 -6 2 -3 -8 -11
Peas 0 -0 -0 o -O -0 -0 -O -0 -1
Peanuts -0 -0 -0 -0 -0 -0 O O 0 -0
Soybeans -0 -0 -0 0 -0 O -0 -0 -O -O
Sorghum -5 —5 -1 -7 -4 -1 3 -5 1 -3
Maize -O -0 -1 -0 -2 -0 O 1 O -1
Wheat 0 O -0 0 0 -O 0 0 -O 0
Finger Millet -O 0 0 -O O 0 0 0 -0 0
Rice 1 -o -O —0 -O -0 -0 -o -1 -0
Cassava -1 3 -13 -2 -4 -6 5 -O -2 -2
white Potato -3 —3 -0 -O 13 -5 -1 -0 -9 39
Sweet Potato 2 2 -O -1 3 6 2 -2 -1 6
Taro -2 0 -1 -0 -0 1 O 0 O O
Cocoyam 0 0 -0 -O 0 -0 -O O -0 0
Cooking Banana -0 2 3 0 O -1 25 0 1 1
Beer Banana -1 -0 -2 -6 -17 5 12 -1 4 5
Fruit Banana 1 1 0 -o -0 1 1 0 1 -0
Coffee 3 2 5 1 8 7 5 O 7 0
Banana Beer 27 20 9 8 27 28 41 5 38 15
Sorghum Beer 14 177 -0 12 2 1 5 6 1 4
Ag. Labor -1 1 1 2 1 -3 -6 -0 -3 -3
Hen-Ag. Unsk. L. 0 0 3 1 2 1 -2 0 1 -0
Skilled Labor 1 4 1 2 4 4 5 1 8 6

Source:DSA/MINAGRI, Agricultural Survey 1990; 1184 households
Units are kilograms, except for beer (litres) and labor (mandays).
BUT-Butare, BYU-Byumba, CIA-Cyangugu, SIX-Gikongoro, SIS-Gisenyi,
GIT-Gitarama, KGO-Kibungo, KEY-Kibuye, RIG-Kigali, RUB-Ruhengeri

 

these crops grow well on the marginal lands, and farmers are forced to use most of their

cash to buy staple foods, even if that leaves very little for everything else.

Western Rwanda imports large quantities of beans, sorghum, and cassava. Table
A-3 shows that the southwestern prefecture of Cyangugu is particularly dependent on
cassava imports, presumably fi'om Congo.

The fertile volcanic highlands of the Northwest send very large quantities of white
potatoes to Kigali and to lower-altitude rural areas. In exchange, potato farmers buy small

amounts of crops that do not grow well at high altitudes.

226

REFERENCES

227

References

Adelhelm, R, and J. Kotschi. 1985. Development and introduction of self-sustaining
practices in tropical smallholder farms. Entwicklung und Landicher Raum 19: 17-

Akyeampong, E., and L. Hitimana 1996. Agronomic and economic appraisal of alley
cropping with Leucaena diversifolia on an acid soil in the highlands of Burundi.
Agroforestry systems 33: 1-1 1.

Altieri, M. A., D. K. Letoumeau, and J. R. Davis. 1984. The requirements of sustainable
agroecosystems. In Agricultural sustainability in a changing world order. , ed.
Gordon K. Douglass. Boulder, CO: Westview Press.

Amani, M., and A. J. Haverkort. 1986. La multiplication de semences de pommes de
terre. In Developpement de la culture de la pomme de terre. , ed. T. Gatarasi and
A. J. Haverkort. Ruhengeri, Rwanda: Institute des Sciences Agronomiques du
Rwanda (ISAR).

Ange, A. 1987. Caractérisation et zonage des ressources naturelles au Nord et aut Centre
du Rwanda Bases d'interprétation des systémes de culture paysans et des lans
d'expérimentation agronomique. IRAT/CIRAD, Monpellier, France.

Angelsen, Arild, and Odd-Helge Fjelstad. 1995. Land reforms and land degradation in
Tanzania: Alternative economic approaches. Chr. Michelsen Institute, Bergen,
Norway.

Ansoanuur, James S. 1991. Price elasticity of supply and expenditure elasticity of
demand of fprmc1pal agricultural commodities in Rwanda. Kigali, Rwanda:

Ministry 0 Planning.

.... 1992. Food consumption parameters for agricultural households in Rwanda and their
application in analysing the impact of the structural adjustment program. Kigali,
Rwanda: MINAGRI, Division des Statistiques Agricoles.

Arnold, J. E. M. 1995. Retrospect and prospect. In Tree management in farmer
strategies: Responses to agricultural intensification , ed. J. E. M. Arnold and
Peter A. Dewees. Oxford: Oxford University Press.

Azam, J. P. 1989. The supply of manufactured goods and agricultural development.
Paris, France: OECD Development Centre.

Balasubramanian, V., and N. K. A. Blaise. 1993. Short season fallow management for
sustainable production in Africa. In Technologies for sustainable agriculture in
the tropics, ed. J. Ragland and R. Lal. Proceedings of two international symposia
held in San Antonio, Texas, and Denver, Colorado, 1990 and 1991, respectively.
Madison, Wisconsin: American Society of Agronomy.

Balasubramanian, V., and A. Egli. 1986. The role of agroforestry in the farming systems

in Rwanda with special reference to the Bugesera-Gisaka—Mikongo region.
Agroforestry systems 4:271-89.

228

Balasubramanian, V., and L. Sekayange. 1991. Effect of tree legumes in hedgerows on
soil fertility changes and crop performance in the semi-arid highlands of Rwanda
Biological Agriculture and Horticulture 8:17-32.

Barampama, Angelo. 1993. Le manioc en Afrique de l'Est: Role et perspectives dans le
développement agricole. Paris: IUBD - Karthala.

Bart, Francois. La strategic alimentaire du Rwanda. In Crise agricole et crise alimentaire
darts les pays tropicaux. , ed. CNRS. Bordeaux, France: CNRS.

--. 1993. Montagnes d'Afiique, terres paysannes: Le cas du Rwanda. Bordeaux, France:
Presses Universitaires de Bordeaux.

Bergen, D. W. 1988. Influences de la densité de la population sur la taille de l'exploitation
agricole et sur l'intensification de l'agriculture traditionelle. Bujumbura, Burundi:
Institut des Sciences Agronomiques du Burundi (ISABU).

Bezy, F. 1990. L'evolution economique et sociale du Rwanda depuis l'independance
(1962-1989). Mondes en Developpement 18: 1 5-38.

Bidou, Jean-Etienne. 1989. Erosion des sols au Burundi. In Géographie et aménagement
dans l'Afrique des Grands Lacs. Bordeaux: CRET, Institut de Geographic,
Université de Bordeaux III, pp. 3-20.

Bilsborrow, Richard E. 1987. Population pressures and agricultural development in
developing countries: A conceptual framework and recent evidence. World
Development 15: 1 83-203.

Binswanger, Hans. 1988. Technology, investment and institutional priorities for rapid
agricultural growth in sub-Saharan Africa In Food policies in Africa. , ed. Jo Ann
Paulson and Charles Ben Pike. Papers presented at a Conference Sponsored by the
University of Minnesota, November 6, 1987. Minneapolis, MN: University of
Minnesota, African Studies Council.

Binswanger, Hans P., and Prabhu L. Pingali. 1984. The evolution of farming systems and
agricultural technology in sub-Saharan Africa. Washington, DC: World Bank.

Binswanger, Hans P., and Mark R Rosenzweig. 1986. Behavioral and material
determinants of production relations in agriculture. Journal of Development
Studies 22:503-39.

Blaikie, P. 1985. The political economy of soil erosion in developing countries. London
and New York: Longman.

Blaikie, P., and H. Brookfield, eds. 1987. Land degradation and society. London:
Methuen.

Blanken, J. 1988. Ernahrungssichenmg bei knappen Bodenressourcen: Das Beispiel
Ruanda. Entwicklung und Landlicher Raum 4:28-31.

229

Blanken, Jttrgen, Joachim von Braun, and Hartwig de Haen. 1994. The triple role of
potatoes as a source of cash, food, and employment: Effects on nutritional
improvement in Rwanda. In Agricultural commercialization, economic
development, and nutrition , ed. Joachim von Braun and Eileen Kennedy.
Baltimore and London: The Johns Hopkins University Press for the International
Food Policy Research Institute.

Blaug, Mark. 1985. Economic theory in retrospect. 4th ed. Cambridge: Cambridge
University Press.

Boserup, E. 1989. Agricultural growth and population change. In The new Palgrave:
Economic development. , ed. John Eatwell, Murray Milgate, and Peter Newman.
London: W.W.Norton. First published in "The new Palgrave: A dictionary of
economics" by the same editors (1987, Macmillan).

Boserup, Esther. 1965. The conditions of agricultural growth Chicago: Aldine.

-—-. 1981. Population and technological change: A study of long-term trends. Chicago:
University of Chicago Press.

--. 1985a. The impact of scarcity and plenty on development. In Hunger and history. , ed.
Robert I. Rotberg and Theodore K. Rabb. Cambridge: Cambridge University
Press.

--. 1985b. Economic and demographic interrelationships in sub-Saharan Africa.
Population and Development Review 1 1:383 -96.

Boughton, Duncan, Thomas Reardon and Jeffrey Wooldridge. 1997. Determinants of
Diversification of Urban Sahel Diets into Maize: A Contingent Valuation Study
of Processed Maize Demand in Mali. Contributed paper presented at the August
1997 Meetings of the International Association of Agricultural Economists,
Sacramento, California.

Bourke, R. Michael. 1989. Food, coffee and casuarina: An agroforestry system from the
Papua New Guinea highlands. In Agroforestry systems in the tropics. , ed. P. K. R.
Nair. Dordrecht: Kluwer.

Braun, Joachim von. 1994. Production, employment, and income effects of
commercialization of agriculture. In Agricultural commercialization, economic
development, and nutrition , ed. Joachim -von Braun and Eileen Kennedy.
Baltimore and London: The Johns Hopkins University Press for the International
Food Policy Research Institute.

Braun, Joachim von, Howarth Bouis, and Eileen Kennedy. 1994. Conceptual framework.
In Agricultural commercialization, economic development, and nutrition , ed.
Joachim von Braun and Eileen Kennedy. Baltimore and London: The Johns
Hopkins University Press for the International Food Policy Research Institute.

Braun, Joachim von, I-Iartwig de Haen, and Juergen Blanken. 1991. Commercialization of
agriculture under population pressure: Efl‘ects on production, consumption, and

230

firs'trition in Rwanda. Washington, D.C.: International Food Policy Research
'tute.

Braun, Joachim von, and Eileen Kennedy. 1994. Conclusions for agricultural
commercialization policy. In Agricultural commercialization, economic
development, and nutrition , ed. Joachim von Braun and Eileen Kennedy.
Baltimore and London: The Johns Hopkins University Press for the International
Food Policy Research Institute.

--, eds. 1994. Agricultural commercialization economic development, and nutrition
Baltimore and London: The Johns Hopkins University Press for the International
Food Policy Research Institute.

Bray, Francesca. 1986. The rice economies: Technology and development in Asian
Societies. Oxford: Basil Blackwell.

Brondeau, F. 1991. Anthropisation et dynamique morpho-pédologique des Bassins-
Versants: Essais de classification des geosystemes rwandais. DEA Thesis,
Université de Bordeaux III, Talence, France.

Buchholz, J. 1984. Grime Revolution oder 6kologischer Landbau: Beobachtungen in
Bangladesh und Rwanda. In Landwirtschafl = Hungerschafi? Umbruche und
Krisen in den Agrarsystemen der Dritten Welt, ed. K. Fliege and L. Ramalho.
Saarbrficken, Germany: Verlag Breitenbach.

Bucyedusenge, Germaine, Hyacinthe Fabiola, Anita Regmi, and Laurence Uwamariya.
1990. Le role de la femme dans l'agriculture rwandaise. Kigali, Rwanda:
Ministére de l'Agriculture et de l'Elevage, Division des Statistiques Agricoles.

Byers, A. 1988. Averses catastrophiques, glissements de terrain et inondations dans les
communes de Nyakinama et Nyamutera en prefecture de Ruhengeri en Mai 1988.
Kigali, Rwanda: Ruhengeri Resource Analysis and Management (RRAM)
Project.

--. 1990. Preliminary results of the RAM project soil loss and erosion control trials,
Rwanda, 1987-1988. In Research needs and applications to reduce erosion and
sedimentation in tropical steeplands. , ed. R. R. Ziemar, C. L. O'Loughlin, and L.
S. Hamilton. IAHS Publication 192. Gainesville, FL: IAHS.

Cambrezy, Luc. 1984. Le surpeuplement en question: Organization spatiale et ecologie
des migrations au Rwanda Paris: Ed. ORSTOM.

--. 1986. Environnement et densités dc population: Le recours a l'histoire. Cahiers
d'Etudes Afiicaines 26:63-73.

Chambers, Robert, and Melissa Leach. 1989. Trees as savings and security for the rural
poor. World Development 17:329-42.

Champion, Jean. 1963. Le bananier. Paris: G.-P. Maisonneuve & Larose.
Chapuis, O. 1986. La theiculture rwandaise. Cahiers d'Outre-Mer 154:117-42.

231

Charlery de la Masseliere, Bernard. 1989. Systemes de cultures et surfaces de densités
dans le Nord-Quest du Rwanda In "Geographic et aménagement dans l'Afrique
des Grands Lacs." Bordeaux: CRET, Institut de Geographic, Université de
Bordeaux III, pp. 33-46.

--. 1992. Le resserrement de l'espace agraire au Rwanda, les paysans dans la crise.
Etudes Rurales 125-126:99-1 15.

--. 1993a. Du "versant-terroir" aux territoires fi'agmentés: Organisation, dynamique et
crise de l'espace agraire au Rwanda. Cahiers Sciences Humaines 29:661-94.

4-. 1993b.2Versants du Rwanda: La gestion de l'incertitude. Réseau Erosion Bulletin
13: 4-30.

Clay, Daniel, F idele Byiringiro, Jaakko Kangasniemi, Thomas Reardon, Bosco
Sibomana, Laurence Uwarnariya, and David Tardif-Douglin. 1995. Promoting
food security in Rwanda through sustainable agricultural productivity: Meeting
the challenges of population pressure, land degradation, and poverty. International
Development Paper 17. East Lansing, MI: Michigan State University, Department
of Agricultural Economics.

Clay, Daniel C. 1996. Fighting an uphill battle: Population pressure and declining land
productivity in Rwanda International Development Working Paper 58. East
Lansing, MI: Michigan State University, Department of Agricultural Economics.

Clay, Daniel C., Mark Guizlo, and Sally Wallace. 1994. Population and land degradation.
East Lansing, MI: Michigan State University, Departments of Agricultural
Economics and Sociology, Geography and Resource Development.

Clay, Daniel C., and Laurence A. Lewis. 1990. Land use, soil loss and sustainable
agriculture in Rwanda Human Ecology 18:147-61.

Clay, Daniel C., and Thomas Reardon. 1994. Determinants of conservation investments
by Rwandan farm households. IAAE Conference Proceedings. IAAE Occasional
Paper 7. Contributed Paper for the 22nd Congress of the International Association
of Agricultural Economists, August 1994, Harare, Zimbabwe.

Clay, Daniel, Thomas Reardon, and Jaakko Kangasniemi. 1995. Sustainable
intensification in the highland tropics: Rwandan farmers’ investments in land
conservation and soil fertility. Economic Development and Cultural Change 46:
35 1 -77.

Close, J. 1955. Enquete alimentaire au Ruanda-Urundi. Mémoires 8 2:1-64. Academic
Royale des Sciences Coloniales.

CNA. 1990. Rapport preliminaire. Volume I. 1. Allocation et exploitation des terres de
marais. 2. Le terrassement pour la conservation des sols. 3. Les cultures associees.
Kigali, Rwanda: Commission National d'Agriculture.

232

--. 1991a. Rapport preliminaire, Volume H1: 1. Les productions agricoles du Rwanda, 2.
Zones agro-climatiques du Rwanda. Kigali, Rwanda: Commission Nationale
d'Agriculture.

--. 1991b. Rapport preliminaire, Volume IV: Di nibilité des terres, peuplement et
environnement au Rwanda Kigali, Rwan : Commission Nationale d'Agriculture.

--. 1992a. Rapport preliminaire. Rapport de synthese. Kigali, Rwanda: Commission
Nationale d'Agriculture.

. 1992b. Rapport preliminaire. Volume VII. 1. L'auto—organisation du monde rural. 2.
Analyse de la politique sectorielle et evolution de la legislation agricole. Kigali,
Rwanda: Commission Nationale d'Agriculture.

--. 1992c. Rapport preliminaire. Volume IX. La commercialization des produits agricoles
au Rwanda. Kigali, Rwanda: Commission Nationale d'Agriculture.

Coase, Ronald. 1937. The nature of the firm. Economica 4:233-61.

Conway, Gordon R., Ibrahim Manwan, and David S. McCauley. 1983. The development
of marginal lands in the tropics. Nature 304:392.

Cook, Cynthia C., and Mikael Grut. 1990. Agroforestry in sub-Saharan Afiica: A farmer's
perspective. Washington, DC: World Bank.

Current, Dean, Ernst Lutz, and Sara J. Scherr. 1995. The costs and benefits of
agroforestry to farmers. The World Bank Research Observer 10: 151-80.

Deaton, Angus. 1997. The analysis of household surveys: A microeconometric approach
to development policy. Baltimore: The Johns Hopkins University Press for the
World Bank.

Decuyefere, E., T. Behaeghe, H. Michels, and J. Van Isterdael. 1982. Integration de
'élevage dans des exploitations agricoles familiales du plateau central au Rwanda.

Modéle propos, et résultats sur le terrain. In Tropical animal production for the
benefit of man , ed. Anon. Antwerp, Belgium: Prince Leopold Institute of Tropical
Medicine.

Delepierre, G. 1970. Le bananier, 150,000 T de regimes par an. Bulletin Agricole du
Rwanda 3: 1 55-62.

--. 1982. Les regions agro-climatiques en relation avec l'intensité de l'érosion du sol.
Bulletin Agricole du Rwanda 15:87-95.

Delor-Vandueren, Annick, and 103 Degand. 1992. Burundi: Demographic, agriculture et
environnement. Brussels, Belgium: CIDEP.

den Biggelaar, Christofel Johannes Petrus Antonius Maria 1994. Farmer knowledge and

experimentation with trees and tree cultivation in agroforestry systems in Rwanda
Ph.D. Dissertation, Michigan State University, East Lansing, MI.

233

Dercnne, Benoet. 1989. De la chicotte aux billons, apercu des méthodes de lutte contre
l'érosion au Rwanda et en Burundi du XD(e siecle .. nos jours. Geneve Afiique
27:45-72.

Durr, G. 1983. Potato production and utilization in Rwanda Lima, Peru: International
Potato Center, Department of Social Sciences.

Economist Intelligence Unit. 1996a Country profile 1996-97: Rwanda and Burundi.
London: The Economist Intelligence Unit Limited.

-—-. 1996b. Country report: Uganda, Rwanda and Burundi. 2nd quarter 1996. London:
The Economist Intelligence Unit Limited. .

--. 1997. Country Report: Uganda, Rwanda and Burundi. 1st quarter 1997. London: The
Economist Intelligent Unit Limited.

Egger, K. 1986. L'intensification ecologique. Conservation (LAE) et amelioration des
sols tropicaux par les systemes agro-sylvo-pastoraux. Collection Documents
systemes Agraires 6: 129-35.

Egger, K., and B. Martens. 1987. Theory and methods of ecofarming and their realization
in Rwanda, East Africa In The Green Revolution revisited Critique and
alternatives. , ed. B. Glaeser. London, UK: Allen & Unwin.

Egger, K., and P. Rottach. 1983. Methoden des Ecofarming in Rwanda Tropenlandwirt
84:168-85.

Egli, A. 1985. La conservation des sols l'aide des méthodes agroforestieres: Le cas du
Rwanda Bulletin Recherches Agronomie Gembloux 20:561-87.

Ehui, Simeon, Timothy Williams, and Brent Swallow. 1995. Environmental factors and
policies encouraging environmentally detrimental land use practices in sub-
Saharan Africa. In Agricultural competitiveness: Market forces and policy
choice, ed. G. H. Peters and Douglas D. Hedley. Proceedings of the Twenty-
second International Conference of Agricultural Economists, held at Harare,
Zimbabwe, 22-29 August 1994. Aldershot, England: Dartmouth.

Eicher, Carl K. 1990. Afiica's food battles. In Agricultural development in the third
world 2nd ed., ed. Carl K. Eicher and John M. Staatz. Baltimore and London:
The Johns Hopkins University Press.

Ellis, Frank. 1988. Peasant economics: Farm households and agrarian development.
Cambridge: Cambridge University Press.

anle, C. R., M. Brewster, and F. Hitayezu. 1993. An economic analysis of fish
production in a subsistence agricultural economy: The case of Rwanda. Journal of
Aquaculture in the Tropics 8:151-65.

234

Eylands, V. J., and C. F. Yamoah. 1990. Sustaining soil fertility with alleycropping
systems in the highlands of Rwanda Journal of Farmrng systems Research and
Extension 1:31-36.

Ezemenari, Kene. 1993. Production risk and commercialization of agriculture 1n Rwanda:
Implications for household labour supply and food security. PhD. Dissertation,
University of Guelph, Canada.

Fafchamps, Marcel, Alain de Janvry, and Elisabeth Sadoulet. 1995. Transaction costs,
market failures, competitiveness and the state. In Agricultural competitiveness:
Market forces and policy choice, ed. G. H. Peters and Douglas D. Hedley.
Proceedings of the Twenty-second International Conference of Agn cultural
gconomimuj, held at Harare, Zimbabwe, 22-29 August 1994. Aldershot, England:

artrno .

Femandes, E. C. M., A. O'Kting'ati, and J. Maghembe. 1989. The Chagga homegardens:
A multistoreyed agroforestry cropping system on Mount Kilimanjaro (northern
Tanzania). In Agroforestry systems in the tropics. , ed. P. K R. Nair. Dordrecht,
The Netherlands: Kluwer Academic Publishers.

Fleuret. P., and A. Fleuret. 1980. Nutritional implications of staple food crop successions
in Usambara, Tanzania. Human Ecology 8:311-27.

Ford, R. E. 1990. The dynamics of human—environment interactions in the tropical
montane agrosystems of Rwanda: Implications for economic development and
environmental stability. Mountain Research and Development 10:43-63.

--. 1993. Marginal coping in extreme land pressures: Ruhengeri, Rwanda In Population
growth and agricultural change in Afi'ica. , ed. B. L. Turner, Goran Hyden, and
Robert W. Kates. Gainesville: University Press of Florida.

Friedrich, K. H. 1968. Coffee-banana holdings at Bukoba. In Smallholder farming and
smalllholder development in Tanzania , ed. H. Ruthenberg. Manchen: Weltforum
Ver ag.

Fujisaka, S., and D. P. Garrity. 1988. Developing sustainable food crop farming systems
for the sloping acid uplands: A farmer-participatory approach. Paper presented at
the 4th Southeast Asian Universities Agroecosystem Network (SUAN) Research
Symposium, Khon Kaen, Thailand, July 4-8, 1988. Los Bailos, Laguna,
Philippines: International Rice Research Institute.

--. 1989. Farmers and scientists: A joint efi‘ort in upland soil conservation research and
technology transfer. Paper presented at the International Workshop on
Conservation Farming on Hillslopes, in Taichung, Taiwan, 20-29 March, 1989.
Los Bailos, Laguna, Philippines: International Rice Research Institute.

Garrity, Dennis P. 1989. Hedgerow systems for sustainable food crop production on
sloping lands: Elements of a strategy. Paper presented at the Symposium
Workshop on the Development of Upland Hilly Areas 1n Region XII, Bansalan,
Davao del Sur, December 20-22,1989.Los Bafios, Laguna, Philippines:
International Rice Research Institute.

235

Garrity, Dennis P., and Percy E. Sajise. 1990. Sustainable land use systems research in
Southeast Asia: A regional assessment. Paper presented at the Workshop on
"Sustainable Land Use Systems Research and Development", New Delhi, India,
12-16 February, 1990. Manila, Philippines: International Rice Research Institute.

Gatera, Frederic. 1980. Accroissement démographique et deforestation au Rwanda
Bulletin Agricole du Rwanda 13:28-30.

Geertz, Clifford. 1963. Agricultural involution: The rocess of ecological change in
Indonesia Berkeley, CA: University of Cali ornia Press.

Getahun, Amare. 1991. Agricultural growth and sustainability: Conditions for their
compatibility in the tropical East Africa highlands. In Agricultural sustainability,
growth and poverty alleviation: Issues and policies. , ed. Stephen A. Vosti,
Thomas Reardon, and Winfried von Urff. Proceedings of the Conference held
from 23 to 27 September, 1991 in Felfading, Germany. Felfading: Deutsche
Stifiung filr Intematinale Entwicklung (DSE) and International Food Policy
Research Institute (IF PR1).

Gillespie, S. 1989. Potential impact of AIDS on farming systems. A case study from
Rwanda Land Use Policy 6:301-12.

Goodland, Robert J. A., Catharine Watson, and George Ledec. 1984. Environmental
management in tropical agriculture. Boulder, CO: Westview Press.

Gotanegre, J. F. 1983. La Banane au Rwanda. Cahiers d'Outre-Mer 36:311-42.

Graaff, J. de 1986. Coffee in Rwanda In The economics of cofl‘ee., ed. J. de Graafi‘.
Wageningen, Netherlands: OAE.

Greenland, D. J., and A. Wild. 1990. The case for chemicals. Ceres 22:15-19.

Grosse, Scott 1994. More people, more trouble: Population growth and agricultural
change in Rwanda. Report to USAID. Environmental and Natural Resources
Policy and Training (EPAT) project.

--. 1996. The economic and demographic determinants of child growth retardation in
rural Rwanda Ph.D. Dissertatron, University of Michigan, Ann Arbor, Michigan.

Guichaoua, A. 1989. Destins paysans et politiques agraires en Afiique Centrale. Tome I.
L 'ordre paysan des hautes terres centrales du Burundi et du Rwanda Paris:
Editions l'Harmattan.

--. 1991. Les ”travaux communautaires" en Afrique centrale. Revue Tiers Monde 32:551-
73.

Guichaoua, Andre. 1986. Budgets et strategies monétaires des paysans au Burundi et au
Rwanda. Revue Tiers Monde 27:375-94.

236

Haggblade, Steve, and N. Minot. 1987. Opportunities for enhancing performance in
Rwanda's alcoholic beverages subsector. Kigali, Rwanda: US. Agency for
International Development. ,

Haggblade, Steven, Peter Hazell, and James Brown. 1989. Farm-nonfarm linkages in
rural sub-Saharan Africa. World Development 17:1 173-1202.

Hamand, J. 1982. Pressure on the land. People 9:14-16.
Hardin, G. 1968. The tragedy of the commons. Science 163:1243-48.
Harrison, Paul. 1987. The greening of Afiica London: Palading Grafion Books.

Harroy, Jean-Paul. 1954. La lutte contre la dissipation des ressources naturelles au
Ruanda-Urundi. Civilisations 4:363-74.

Haverkort, A. J. 1986a. Yield levels of potato crops in central Afiica. Agricultural
Systems 21 :227-35.

--. 1986b. Recherche sur les semences botaniques de la pomme de terre. In
Developpement de la culture de la pomme de terre. , ed. T. Gatarasi and A. J.
(I-IIaverkort. Ruhengeri, Rwanda: Institute des Sciences Agronomiques du Rwanda

SAR).

Haverkort, A. J., and C. Rutayisire. 1986. Utilisation des engrais chimiques soud
conditions tropicales. l. Efl'et de l'application d'azote, phosphore et potasse sm' les
rendements de la pomme de terre en Afrique centrale. Potato Research 29:347-55.

Hill, Polly. 1986. Development economics on trial. Cambridge: Cambridge University
Press.

Hillman, Jimmye S. 1981. The role of e rt cropping in less developed countries.
American Journal of Agricultur Economics 62 (2):375-83.

Ho, T. J. 1985. Population wth and agricultural productivity in sub-Saharan Africa In
Proceedings of the flh agricultural sector symposium: Population and food , ed.
T. J. Davis. Washington, D.C.: World Bank.

Hoekstra, D. A. 1994. Factors influencing the economic viability of three agroforestry
systems in Eastern and Southern Africa Forest Ecology and Management 64:265-

Holden, S. T. 1993. Peasant household modelling: Farming systems evolution and
sustainability in northern Zambia. Agricultural Economics 9:241 -67.

Hunter, S. S., E. Bulirwa, and E. Kisseka 1993. AIDS and agricultural production:
Report of a land utilization survey, Masaka and Rakai districts of Uganda Land
Use Policy 10:241-58.

.237

Hyden, Goran, Robert W. Kates, and B. L. Turner. 1993. Beyond intensification. In
Population growth and agricultural change in Afiica , ed. B. L. Turner, Goran
Hyden, and Robert W. Kates. Gainesville: University Press of Florida

ISAR. 1989. Plan directeur national de la recherche agricole, 1990-2000. Document
provisoire. Kigali, Rwanda: Ministere de L'Agriculture de L'Elevage et des F orets
(MINAGRI), Institute des Sciences Agronornrques du Rwanda (ISAR).

Ishikawa, Shigeru. 1967. Economic development in Asian perspective. Tokyo:
Kinokuniya Bookstore.

Jabbar, M. A., A. Larbi, and L. Reynolds. 1994. Profitability of alley farming with and
without fallow in southwest Nigeria Experimental Agriculture 30:319-27.

Jahnke, Hans E. 1982. Livestock production systems and livestock development in
tropical Afiica. Kiel, Germany: Kieler Wissenschaftsverlag Vauk.

de Janvry, Alain, Marcel Fafcham; ps, and Elizabeth Sadoulet. 1991. Peasant household
behavior with missing markets: Some paradoxes explained. The Economic
Journal 101:1400-1417.

 

Jones, William 1., and Roberto Egli. 1984. Farming systems in Afiica: The Great Lakes
highlands of Zaire, Rwanda, and Burundi. Washington, DC: World Bank.

Kajumulo-Tibaijuka, Anna 1984. An economic analysis of smallholder banana-coffee
farms in the Kagera Region, Tanzania. Causes of decline in roductivity and
strategies for revitalization. Ph.D. Dissertation, Swedish Umversity of
Agricultural Sciences, Uppsala, Sweden.

Kangasniemi, Jaakko and Thomas Reardon. 1994. Demographic pressure and the
sustainability of land use in Rwanda IAAE Conference Proceedings. IAAE
Occasional Paper 7. Contributed Paper for the 22nd Congress of the International
Association of Agricultural Economists, August 1994, Harare, Zimbabwe.

'emi, Jaakko. 1996. Bananas in Rwanda: Wasteful Brewing or Food Security for
the Poor. Paper presented in “Local Econom ” seminar in May 19-20, 1996.
Helsinki, Finland: Helsinki University, Institute of Development Studies.

Kates, Robert W., Goran Hyden, and B. L. Turner. 1993. Theory, evidence, study design.
In Population growth and agricultural change in Afiica , ed. B. L. Turner, Goran
Hyden, and Robert W. Kates. Gainesville: Universrty Press of Florida

Kennedy, Eileen, and Thomas Reardon. 1994. Shifi to non-traditional grains in the diets
of East and West Afiica: Role of women's opportunity cost of time. Food Policy
19:45-56.

Kessler, J. J ., and K. F. Wiersum. 1995. Ecological sustainability of agroforestry in the
tropics. Agriculture and Rural Development 2:50-53.

238

KOnig, D. 1992a. L'agriculture écologique agro-forestiere. Une strategic integree de
conservation des sols au Rwanda Réseau Erosion Bulletin 12: 130-39.

--. 1992b. The potential of agroforestry methods for erosion control in Rwanda. Soil
Technology 5:167-76.

--. 1993. Degradation et erosion des sols au Rwanda Réseau Erosion Bulletin 13:31-40.

Koponen, Juhani. 1988. People and production in late precolonial Tanzania: History and
structures. Helsinki, Finland: Finnish Society for Development Studies, Finnish
fiaghropglagcal Society, Finnish Historical Society, and Scandinavian Institute of

can es.

Kortteinen, Timo. 1989a Agricultural alcohol and social change in the third world
Helsinki, Finland: The Finnish Foundation for Alcohol Studies.

--. 1989b. Potentials of state monopoly systems in preventing alcohol-related problems
in developing countries. In State monopolies and alcohol prevention , ed. Timo
gangeinen. Report 181. Helsinki: The Social Research Institute of Alcohol

es.

Kotschi, J ., J. Pfeiffer, and E. Grosser. 1983. A model of sustainable agriculture. Applied
Geography and Development 22: 108-27.

Kotschi, Johannes, Irmfried Neumann, Pieter Pietrowicz, and Jbrg Haas. 1991.
Standortgerechte Landwirtschaft in Ruanda: Zehn Jahre F orschung und
Entwicklung in Nyabisindu. Eschborn, Germany: Deutsche Gesellschafi filr
Technische Zusammenarbeit (GTZ).

Kumar, Krishna, David Tardif-Douglin, Kim Maynard, Peter Manikas, Annette Sheckler,
and Carolyn Knapp. 1996. Rebuilding post-war Rwanda Study 4 in David
Millwood, ed.: "The international response to conflict and genocide: Lessons from
the Rwanda experience." Joint evaluation of emergency assistance to Rwanda

Lagemann, Johannes. 1977. Traditional Afiican farming systems in Eastern Ni eria: An
analysis of reaction to increasing population pressure. Milnehen: Wel orum
Verlag.

Lalou, Richard. 1993. La situation demogz‘lafihique au Rwanda: Tendances actuelles et
futures. Kigali, Rwanda: World B

Lambert, A. 1982. Rwanda: Dynamique de la population et dynamique des resources
alimentaires. Université Catholique de Louvain, Departement de Demographic.

Larson, B. A. 1994. Changing the economics of environmental degradation in
M ascar: Lessons from the National Environmental Action Plan process.
Worl Development 22:671-89.

Laure, J. 1986. Des Vivres ou du the? L'alimentation et les conditions de vie defamilles
rwandaises. Paris: Editions de l'ORSTOM.

239

Lele, Uma, and Steven W. Stone. 1989. Population pressure, the environment and
agricultural intensification: Variations on the Boserup hypothesis. Washington
D. C.: World Bank.

Leurquin, Philippe P. 1963. Agricultural change in Ruanda-Urundi, 1945-1960. Food
Research Institute Studies 4:39-89.

Leurquin, Pierre. 1960. Le niveau de vie des populations rurales de Ruanda- Urundi.
Louvain and Paris: Ed. Nauwelars.

Levi, John, and Michael Havinden. 1982. Economics of Afi-ican agriculture. London:
Longman.

Lewis, Lewis A. 1986. Predicting soil 1033' m Rwanda Kigali, Rwanda: Ministry of
Agriculture and Livestock, Division of Agricultural Statistics.

--. 1988. Measurement and assessment of soil loss in Rwanda. Catena 12:151-65.

--. 1992. Terracing and accelerated soil loss on Rwandan steeplands: a preliminary
investigation of theim lications of human activities afl‘ecting soil movement.
Land Degradation Rehabilitation 3: 241-46.

Lewis, L. A., and L. Berry. 1988. Afi'ican environments and resources. London: Unwin
Hyman.

Lewis, Laurence A., Daniel C. Clay, and Yvan M. J. Dejaegher. 1988. Soil loss,
agriculture, and conservation in Rwanda: Toward sound strategies for soil
management. Journal of Soil and Water Conservation 43 :41 8-21.

Lipton, Michael. 1990. Responses to rural population growth: Malthus and the modems.
In Rural development and population: Institutions and policy. , ed. Geoffrey
McNicoll and Mead Cain. A supplement to Population and Development Review

15. New York: Oxford University Press and the Population Council.

 

Little, Peter D., and Michael M. Horowitz. 1987. Subsistence crops are cash crops: Some
comments with reference to eastern Africa. Human Organization 46:254-58.

--. 1988. Author's reply (to Robins and Ndoreyaho). Human Organization 47:271-73.

Loevinsohn, M. E., J. Mugarura, and A. Nkusi. 1994. Cooperation and innovation by
farmer groups: Scale 1n the development of Rwandan valley farming systems.
Agricultura Systems 46: 141-55.

Loveridge, Scott. 1988. Uses of food and market survey data to inform food security
policy in Rwanda Ph.D. Dissertation, Michigan State University, East Lansing,
MI.

---. 1991. Marketing in Rwanda: imports and infrastructure. Food Policy 16:95-104.

240

--. 1992. Sources of household revenue, farm exports, and their impact on food
availability among agricultural households 1n rural Rwanda Kigali, Rwanda:
Ministry of Agriculture and Livestock, Division of Agricultural Statistics.

Ludwig, H. D. 1968. Permanent farming on the Ukara. In Smallholder farming and
slmalllholder development in Tanzania , ed. Hans Ruthenberg. Munich: Weltforum
er ag.

Martin, Susan. 1987. Boserup revisited: Population and technology 1n tropical African
ggrrculture, 1900-1940. Journal of Impertal and Commonwealth History 16: 109-

McAllister, James Ernest. 1994. Differentiation among peasant households in Rwanda
Ph.D. Dissertation, Michigan State University, East Lansing, MI.

McCown, R. L., B. A. Keating, M. E. Probert, and R. K. Jones. 1992. Strategies for
sustainable crop production in semi-arid Africa Outlook on Agriculture 21 :21-3 1.

McIntire, John, Daniel Bourzat, and Prabhu Pingali. 1992. Crop-livestock interaction in
sub-Saharan Afiica Washington, D.C.: World Bank.

Mellor, John W. 1986. Agriculture 011 the road to industrialization. In Development
strategies reconsidered , ed. J. P. Lewis and V. Kallab. New Brunswick, NJ:
Transaction Books.

Mellor, John W., and Gunvant Desai. 1985. Agricultural change and rural poverty.
Baltimore, MD: Johns Hopkins University Press for the International Food Policy
Research Institute.

Meschy, L. 1989. La colline et le marais. La gestion des bassins versants au Burundi et au
Rwanda Etudes Rurales 115-1 16:129-51.

Meschy, Lidia 1985. Perception et modélisation des espaces densément peuplés. Cahiers
d 'Etudes Afiicaines 25:1 11- 17.

Miklavcic, Pamela Riley. 1995. Determinants of farm household participation in
Rwanda's banana wine subsector. Master of Science Thes1s, Michigan State
University, Department of Agricultural Economics, East Lansing, MI.

MINAGRI. 1988. Production agricole en 1986, Tome 1: Bilan d'autosuflisance
alimentaire par commune et par habitant. Kigali, Rwanda: Ministere de
l'Agriculture et de l'Elevage.

--. 1991. Enquéte Nationale Agricole 1989: Production, superficie, rendement, elevage et
leur evolution 1984-1989. Kigali, Rwanda: Ministere de l'Agricultme et de
l'Elevage, Division des Statistiques Agricoles.

--. 1992. Enquéte Nationale Agricole 1990: Production, superficie, rendement, elevage et
leur evolution 1984-1990. Publication DSA 26. Kigali, Rwanda: Ministere de
l'Agriculture et de l'Elevage, Division des Statistiques Agricoles.

241

--. 1993a. Enquéte Nationale Agricole 1991: Production, superficie, rendement, elevage
et leur evolution 1984-1990. Publication DSA 27. Kigali, Rwanda: Ministere de
l'Agriculture et de l'Elevage, Division des Statistiques Agricoles.

MINIPLAN. 1983. Strategic alimentaire au Rwanda: Objectifs chimes et programmes
d'actions. Kigali, Rwanda: Ministere de la Planification.

--. 1988. Enquéte Nationale sur le budget et la consommation des ménages (Milieu
rural). Volume 4: Consommation alimentaire en milieu rural. Kigali, Rwanda:
Ministere de la Planification.

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, Department of Agricultural Economics,
East Lansing, MI. '

Moeyersons, J. 1989. La nature de I 'érosion des versants au Rwanda Tervuren, Belgium:
Musee Royal de l'Afrique Centrale.

--. 1990. Soil loss by rainwash: A case study from Rwanda Z Geomorph 34:385-408.

--. 1991. Ravine formation on steep slopes: Forward versus regressive erosion. Some
case studies from Rwanda Catena 18:309-24.

--. 1993. Les essais récents dc lutte anti-erosive au Rwanda Réseau Erosion Bulletin
13:50-62.

Monares, A. 1984. Building an effective potato country programme: The case of Rwanda
Lima, Peru: International Potato Center, Department of Social Sciences.

--. 1987. Analytical framework for design and assessment of potato seed programs in
developing cormtries. In The Social Sciences at CIP. , ed. CIP. Report of the third
social science planning conference, held at CIP, Lima, Peru, September 7-10,
1987 . Lima, Peru: International Potato Center.

Montcel, Hugues Tezenas du. 1985. Le bananier plantain Paris: Editions Maisonneuve
et Larose.

Mulindangabo, J ., and D. Rollin. 1989. Resultats preliminairies de l'essai d'adaptation des
varietes de manioc aux conditions de haute altitude de la zone du projet Kigali
Nord au Rwanda. Bulletin Agricole du Rwanda 21 :244-48.

Musema-Uwimana, Alfred. 1983. La conservation des terres au Rwanda Bulletin
Agricole du Rwanda 16:86-94.

Nair, M. Achutan, and C. Sreedharan. 1989. Agroforestry farming systems in the
homesteads of Kerala, southern India In Agroforestry systems in the tropics. , ed.
P. K. R. Nair. Dordrecht: Kluwer Academic Publishers.

242

Nair, P. K. R. 1989a F ood-producing trees in agroforestry systems. In Agroforestry
systems in the tro ics. , ed. P. K. R. Nair. Dordrecht, The Netherlands: Kluwer
Academic Publi ers.

--. 1989b. The role of trees in soil productivity and protection. In Agroforestry systems in
gruebtlro ics. , ed. P. K. R. Narr. Dordrecht, The Netherlands: Kluwer Academic
1 ers.

--, ed. 1989c. Agroforestry systems in the tropics. Dordrecht, The Netherlands: Kluwer
Academic Publishers.

Ndamage, G. 1984a Sweet potato cultivation and research in Rwanda In Crop
improvement in eastern and southern Afiica: Research objectives and on- arm
testing. , ed. R. A. Kirkby. Ottawa, Canada: IDRC.

--. 1984b. Sweet potato production potential in Rwanda. In Tropical root crops:
Production and uses in Afiica, ed. E. R. Terry, E. V. Doku, O. B. Arene, and N.
M. Mahungu. Ottawa, Kanada: IDRC. '

Ndayirukiye, Silvestre. 1989. Milieu nature] et problemes d'aménagement agricole dans
les collines de l'Imbo: 1e cas de Kabezi au sud de Bujumbura In "Geographic et
aménagement dans l'Afiique des Grands Lacs." Bordeaux: CRET, Institut de
Geographic, Université de Bordeaux 111, pp. 33-46.

Ndayizigiye, F. 1992. Valorisation des haies arbustives (Calliandra et Leucaena) dans la
lutte contre l'erosion e11 zone de montagnes (Rwanda). Réseau Erosion Bulletin
12:120-29.

Ndayizigiye, Francois. 1993. Efl‘ets des haies arbustives (Calliandra et Leucaena) sur
l'érosion et les rendements des cultures en zone de montagne (Rwanda). Réseau
Erosion Bulletin 13:41-49.

Ndiaye, Serigne M., and Andrew J. Sofianko. 1994. Farmers' perceptions of resource
problems and adoption of conservation practices in a densely populated area.
Agriculture, Ecosystems and Environment 48:3 5-47.

Ndimira, P. F. 1991. Dynamique et problématique d'amélioration des'systemes
d'exploitation agricole au Burrmdi: Cas de la région de Remera Ph.D.
Dissertation, Université Catholique du Louvain, Faculté des Sciences
Agronomiques, Louvain-la-Neuve, Belgium.

Nduwayezu, Jean Damascene. 1990. Les fondements physiques, humains et economiques
du developpement du Rwanda. Ruhengeri, Rwanda: Editions Universitaires du
Rwanda

Newbury, C. 1988. The cohesion of oppression Clientship and ethnicity in Rwanda,
1860-1960. New York: Columbia University Press.

243

Ngenzi, Esdras. 1992. Application d'rme nouvelle méthode de gestion des sols dans une
zone montagneuse tropicale: Terrasses radicales au Rwanda Réseau Erosion
Bulletin 12194-102.

North, Douglass. 1990. Institutions, institutional change, and economic performance.
New York: Cambridge University Press.

Ntezilyayo, Anastase. 1986. L'agriculture—l'horizon 2000 cu comment doubler la
production vivriere au Rwanda. Revue Tiers Monde 27:395-417.

Nwafor, J. C. 1979. Agricultural land use and associated problems in Rwanda Journal of
Tropical Geography 48:58-65.

Nyabyenda, P. 1988. Importance socio-economique de la banane au Rwanda Bulletin
Agricole du Rwanda 21 :223-27.

Nyamulinda, V., and V. Ngiruwonsanga 1992. Lutte anti-erosive et strategic paysanne
dans les montagnes du Rwanda Réseau Erosion Bulletin 12:71-82.

Nzisabira, Jean. 1989a Accumulation du peuplement rural et ajustements structurels du
systeme d'utilisation du sol au Rwanda depuis 1945. Bulletin Agricole du Rwanda
21 :1 17-28.

--. 1989b. Accumulation du peuplement rural et ajustements agro-pastoraux au Rwanda.
Brussels, Belgium: CIDEP.

Okoth-Ogendo, H. W. O., and John O. Oucho. 1993. Population growth and agricultural
change in Kisii District, Kenya: A sustained symbiosis? In Population growth and
agricultural change in Africa , ed. B. L. Turner, Goran Hyden, and Robert W.
Kates. Gainesville: University Press of Florida

Olson, Jennifer M. 1992. Rapport de Mission. Kigali, Rwanda: Projet des enquétes
agricoles et analyse des politiques economiques du secteur rural (ASAP).

--. 1994a Land degradation in Gikongoro, Rwanda: Problems and possibilities in the
integration of household survey data and environmental data. Rwanda Society-
Environment Project, Working Paper 5. East Lansing, MI: Michigan State
University, Department of Geography and the Center for Advanced Study of
International Development.

--. 1994b. Farming systems of Rwanda: Echoes of historic divisions reflected in current
land use. Rwanda Society-Environment Project, Working Paper 2. East Lansing,
MI: Michigan State University, Department of Geography and the Center for
Advanced Study of International Development.

--. 1994c. Demographic responses to resource constraints in Rwanda Rwanda Society-
Environment Project, Working Paper 2. East Lansing, MI: Michigan State
University, Department of Geography and the Center for Advanced Study of
International Development.

244

--. 1994d. Farmer responses to degradation in Gikongoro, Rwanda Ph.D. Dissertation,
Michigan State University, Department of Geography, East Lansing, MI.

Osiname, Olu, and Jean Tonye. 1996. Alley farming research at the International Institute
of Tropical Agriculture (IITA). Agriculture and Rural Development 3:63-64.

Pagrri, Lucien. 1982. Rwanda de Malthus face l'espoir reel du progres economique.
Courrier Afiique-Caraibes-Paciflque 72:8-15.

Palmer, Jeff. 1992. The sloping agricultural land technology experience. In Let farmers
judge: Experience in assessing the sustainability of agriculture. , ed. Wim
Hiemstra, Coen Reijntjes, and Eric van der Werf. London: Intermediate
Technology Publications.

Paningbatan, E. P. 1994. Management of soil erosion for sustainable agriculture in
sloping lands. Philippine Agriculturist 77:1 17-3 8.

Pearce, David, Edward Barbier, and Anil Markandya 1990. Sustainable development:
Economics and environment in the third world London: Earthscan Publications.

Pearce, R. 1990. Traditional food crops in sub-Saharan Africa. Food Policy 15:374-393.

Perrault, Paul Thomas. 1978. Banana-maniac farming systems iof the tropical forest: A
case study in Zaire. Ph.D. Dissertation, Standford University, San Francisco, CA.

Pingali, Prabhu, and Hans P. Binswanger. 1988. Population density and farming systems:
The changing locus of innovations and technical change. In Population, food, and
rural development. , ed. Ronald D. Lee, W. Brian Arthur, Allen C. Kelley, Gerry
Rodgers, and T. N. Srinivasan. Oxford: Clarendon Press.

Pingali, Prabhu L., and Hans P. Binswanger. 1984. Population density and agricultural
intensification: A study of the evolution of technologies in tropical agriculture.
Washington, DC: World Bank.

Pinners, E., and V. Balasubramanian. 1991. Use of the iterative diagnosis and design
approach in the development of suitable agroforestry systems for a target area
Agroforestry systems 15:183-201.

Pinstrup-Andersen, Per. 1985. The impact of export crop production on human nutrition.
In Nutrition and Development. , ed. Margaret Biswas and Per Pinstrup-Andersen.
Oxford: United Nations University and Oxford University Press.

Platteau, Jean-Philippe. 1990. The food crisis in Africa: A comparative structural
analysis. In The political economy of hunger, Vol. 2: Famine prevention , ed. J.
Dreze and A. K. Sen. Oxford: Clarendon Press.

--. 1996. Physical infrastructure as a constraint of agricultural growth: The case of sub-
Saharan Africa Paper presented in the V111 General Conference of the European
Association of Development Research and Training Institutes, Vienna, 11-14
September, 1996. -

245

Pottier, J. 1993. Taking stock: Food marketing reform in Rwanda, 1982-89. Afi'ican
Aflairs 9215-30.

Pottier, J ., A. Nkundabashaka, and E. Croll. 1992. Intolerable environments: Towards a
cultural reading of agrarian practice and policy in Rwanda In Bush base, forest
farm: Culture, environment and development. , ed. E. Croll and D. Parkin. London
and New York: Routledge.

Pottier, Johan P. 1986. The politics of famine prevention: Ecology, regional production,
and food complementarity in western Rwanda Afiican Aflairs 88:207-37.

Prunier, Gerard. 1995. The Rwanda crisis I 959-1 994: History of a genocide. London:
Hurst & Company.

Raquet, Kurt. 1991. Production d'engrais verts a partir de jacheres arbustives a croissance
rapide en region montagneuse tropicale du Rwanda In Pratiques d'agriculture
ecologique pour petites exploitations tropicales. , ed. Johannes Kotschi.
Weikersheim, Germany: Verlag Josef Margraf for Le Centre Technique de
Cooperation Agricole et Rurale and Deutsche Gesellschaft fitr Techrnsche
Zusammenarbeit (GTZ).

 

Reardon, Thomas. 1997. Using evidence of household income diversification to inform
study of the rural nonfarm labor market in Africa World Development 25:73 5-47.

Reardon, T., V. Kelly, E. Crawford, K. Savadogo, and T. Jayne. 1994. Determinants of
agricultural productivity: Four case studies in Africa. East Lansing, MI: Michigan
State University, Department of Agricultural Economics.

Reardon, Thomas, Eric Crawford, and Valerie Kelly. 1994. Links between nonfarm
income and farm investment in African households: Adding the capital market
perspective. East Lansing, MI: Michigan State University, Department of
Agncultural Economics.

--. 1995. Promoting investment in sustainable intensification of African agriculture. East
Lansing, MI: Michigan State University, Department of Agricultural Economics.

Reardon, Thomas, and May Mercado-Peters. 1993. Self-financing of rural household cash
expenditures in Burkina Faso: The case of net cereal buyers. In Finance and Rural
Development in West Afiica. , ed. Cuevas and Benoit-Cartin. Proceedings of the
12th Rural Economy Seminar (CIRAD, Ohio State University, CN CA, CEDRES,
INERA) in Ouagadougou, October 21-25, 1991. Montpellier, 1993: CIRAD.

Reardon, Thomas, and Stephen A. Vosti. 1994. Links between rural poverty and
environment in developing countries: Asset categories and "investment poverty".
East Lansing, MI: Michigan State University, Department of Agricultural
Economics.

Richards, Paul. 1983. Ecological change and the politics of African land use. Afiican
Studies Review 26: 1-72.

246

--. 1985. Indigenous agricultural revolution. Ecology and food producnon in West
Africa London: Hutchinson.

Rishirumuhirwa, T. 1992. Ruissellement et erosion sous bananier au Burundi. Réseau
Erosion Bulletin 12:83-93.

--. 1993a. Facteurs anthropiques de l'érosion dans les regions des montagnes et hauts
plateaux de la CEPGL. Réseau Erosion Bulletin 13:53-62.

--. 1993b. Contribution des residues du bananier en conservation de l'eau et du sol.
Réseau Erosion Bulletin 13:53-62.

Robins, Edward, and Valens Ndoreyaho. 1988. Agricultural policy and practice in
Rwanda: A response to Little and Horowitz. Human Organization 47 :270-71 .

Roose, E., F. Ndayizigiye, L. Sekayange, and J. Nsengimana 1992. La gestion
conservatoire de l'eau ete de la fertilité des sols (G.CES): Une nouvelle strategic
pour l'intensification de la production et la restauration de I'environnement en
montagne. Réseau Erosion Bulletin 12:140-60.

Roose, E., V. Nyamulinda, F. Ndayizigiye, and E. Byiringiro. 1988. La gestion
conservatoire dc l'eau et de la fertilité des sols (G.C.E.S.): Une nouvelle strategic
de lutte anti-erosive pour le Rwanda Bulletin Agricole du Rwanda 21 :264-7 7.

Roose, Eric. 1985. Soil and water conservation lessons from steep-slope in
French-speaking countries of Africa In Conservationfarming on steep lands, ed.
W. C. Moldenhauer and N. W. Hudson. Ankeny, Iowa: Soil and Water
Conservation Society.

---. 1988. Evaluation duro rogramrne de recherche "Management dc l'eau et de sols" dc
l'ISAR ct dcs pro lemcs d'érosion au Rwanda Compte rendu de mission du 5 an
20 février 1988. Montpellier, France: ORSTOM.

Rosenzweig, Mark R., Hans P. Binswanger, and John McIntire. 1988. From land
abundance to land scarcity. The effects of population growth on production

relations 1n agrarian economies. In Population, food, and rural development, ed.
Ronald D. Lee, W. Brian Arthur, Allen C. Kelley, Gerry Rodgers, and T. N.
Srinivasan. Oxford. Clarendon Press.

Rossi, G. 1984. Evolution des versants et mise en valeur agricole au Rwanda Annales de
Géographie 515:23-43.

--. 1991. Croissance de la population, mise en valeu ct équilibre des versants: Quel
avenir pour le Rwanda? Cahiers d'Outre-Mer 44:29-47.

Rushemeza, Justin. 1991. The economics of rice production in Rwanda: An application of
the policy analysis matrix. Ph.D. Dissertation, Cornell University, Ithaca, NY.

Ruthenberg, Hans. 1964. Agricultural development' 1n Tanganyika. Berlin: Springer-
Verlag.

247

--, ed. 1968. Smallholder farm1ng and smallholder development 1n Tanzania: Ten case
studies. Munchen: Weltfonrm Verlag.

--. 1976. Farming systems in the tropics. 2nd ed. Oxford: Clarendon Press.

Ruttan, Vernon W. 1990. Sustainability is not enough. In Agricultural development in the
third world 2nd ed., ed. Carl K. Eicher and John M. Staatz. Baltimore and
London: The Johns Hopkins University Press.

--. 1991. Sustainable growth 1n agricultural production. Poetry, policy and science. In
icultural sustainability, growth and poverty alleviation: Issues and polrc1es ,
Step hen A. Vosti, Thomas Reardon, and Winfried von Urff. Proceedings of
the Conference held from 23 to 27 September, 1991 m Felfading, Germany.
Felfading: Deutsche Stifamg fiIr Intematinale Entwicklung (DSE) and
International Food Policy Research Institute (IF PR1).

Rwamasirabo, Serge. 1990. Goat production in Rwanda: Assessing the potential for
future development. Master's Thesis (Plan B), East Lansing, MI: Michigan State
University, Department of Agricultural Economics.

Scaetta, H. 1932. Les famines périodiques dans le Ruanda: Contribution a l'étude des
aspects biologiques du phenomena Institut Royal Colonial Belge, Section des
Sciences, Brussels.

Schultz, T. W. 1964. Transforming traditional agriculture. New Haven: Yale University
Press.

Scott, G. J. 1988. Potatoes in central Afiica: A study of Burundi, Rwanda and Zaire.
Lima, Peru: International Potato Center.

Sellstrdm, Tor, and Lennart Wohlgemuth. 1996. Historical perspective: Some
explanatory factors. In David Millwood, ed.: "The international response to
conflict and genocide: Lessons from the Rwanda experience." Joint Evaluation of
Emergency Assistance to Rwanda

Sen, A. K. 1981 . Poverty and famines: An essay on entitlement and deprivation Oxford:
Oxford University Press.

Sen, Amartya 1990. Food, economics, and entitlements. In Agricultural development in
the third world 2nd ed., ed. Carl K. Eicher and John M. Staatz. Baltimore and
London: The Johns Hopkins University Press. Reprinted from Lloyd's Bank
Review 160(1986):1-20.

Shepherd, K. D., E. Ohlsson, J. R. Okalebo, and J. K. Ndufa. 1996. Potential impact of
agroforestry on soil nutrient balances at the farm scale in the East Afiican
Highlands. Fertilizer Research 44:87-99.

Simon, Julian. 1981. The ultimate resource. Princeton, N.J.: Princeton University Press.

Sirven, P., J. F. Gotanegre, and C. Prioul. 1974. Géographie du Rwanda Brussels:
Editions A. De Boeck.

248

Smaling, Eric M. A. 1990a Two scenarios for the sub-Sahara: One leads to disaster.
Ceres 22:19-24.

--. 1990b. Unbalanced ledger: Africa's soils are being mined. Ceres 22:20-22.

Sperling, L., and S. Munyaneza. 1995. Intensifying production among smallholder
farmers: The impact of improved climbing beans in Rwanda Afi'ican Crop
Science Journal 3 :1 17-25.

Sperling, L., U. Scheide er, R. Buruchara, P. Nyabyenda, and S. Mrmyanesa. 1994.
Intensifying uction among smallholder farmers: The 11:31am of improved
climbing beans in Rwanda. Network of Bean Research in 'ca (CIAT).

Stark, M. 1996. Adoption and adaption of contour hedgerow farming in the Philippine
uplands: Results of an early case study. Ttopenlandwirt 97:3-16.

Stiglitzélsoseph E. 1986. The new development economics. World Development 14:257-

Stocking, Michael, and Nick Abel. 1992. Labour costs: A critical element in soil
conservation. In Let farmers judge: Experience in assessing the sustainability of
agriculture. , ed. Wim Hiemstra, Coen Reijntjes, and Eric van dcr Werf. London:
Intermediate Technology Publications.

Stoorvogel, J. J ., and Eric M. A. Smaling. 1990. Assessment of the soil nutrient depletion
in sub-Saharan Afiica, l983-2000. Report 28. Wagemngcn, The Netherlands:
Winand Staring Centre.

Stover, R. H., and N. V. Simmonds. 1987. Bananas. 3rd ed. Essex: Longman.

Tacio, H. D. 1993. 810 ing agricultural land technology (SALT): A sustainable
agroforestry sc eme or the uplands. Agroforestry systems 22: 145-52.

Tardif-Douglin, David. 1991. The role of sweet potato in Rwanda's food system: The
transition from subsistence orientation to market orientation. Report prepared for
the Agricultural Surveys and Policy Analysis Project in Rwanda Bethesda,
Maryland: Development Alternatives Incorporated.

Tardif-Douglin, David, Jean-Leonard Ngirumwami, Jim Shafl'er, Anastase Murekezi, and
Theobald Kampayana. 1996. Finding the Balance Between Agricultural and Trade
Policy: Rwanda Coffee Policy in Flux. International Development Working Paper
59. East Lansing, MI: Michigan State University, Department of Agricultural
Economics.

Tifi‘en, Mary. 1995. Population density, economic growth and societies in transition:
Esserup reconsidered in a Kenyan case-study. Development and Change 26:31-

249

Tiffen, Mary, and Michael Mortimore. 1992. Environment, population growth and
productivity in Kenya: A case study of Machakos Drstrict. Development Policy
Review 10:359-87.

Tifi‘en, Mary, Miclmel Mortimore, and Francis Gichuki. 1994. More people, less erosion:
Environmental recovery in Kenya. Chichester, UK: John Wiley & Sons.

Turner, B. L., Goran Hyden, and Robert W. Kates, eds. 1993. Population owth and
agricultural change in Afiica Gainesville, FL: University Press 0 Florida

Uwizeyimana, Lamien. 1991. Croissance démograhpique ct production agricole au
Rwanda: Impossible adéquation. Brussels, Belgium: CIDEP.

 

Verhaegen, Etienne, and 103 Degand. 1993. L'evolution des systemes agricoles au
Burundi: Une dynamique conflictuelle. Louvain-1a-Neuve, Belgium: CIDEP and
Academia-Erasme.

Vidal, C. 1985. Situations ethniques au Rwanda In Au coeur de I 'ethnie: Ethnies,
tribalisme et etat en Afiique. , ed. J.-L. Amselle and E. M'Boloko. Paris: Editions
La Decouverte.

Vis, H. L. 1975. Analyse de la situation nutritionelle dans la région des Grands Lacs
d'Afiiquc Centrale: L'impasse demographique. Revue Tiers Monde 16:567-94.

--. 1986. Avant propos. In Le Kivu montagneux: Surpopulation, sous-nutrition erosion
du sol., ed. W. Wils, M. Carael, and G. Tandem. Brussels, Belgium: Academic
Royale des Sciences d'Outre Mer, Classe des Sciences.

Vosti, Stephen A., Thomas Reardon, and Winfried von Urfl‘. 1991. Synthesis and
conclusions. In Agricultural sustainability, growth and poverty alleviation: Issues
and policies. , ed. Stephen A. Vosti, Thomas Reardon, and Winfried von Urfi‘.
Proceedings of the Conference held from 23 to 27 September, 1991 in Felfading,
Germany. F elfading: Deutsche Stiftung filr Intematinale Entwicklung (DSE) and
International Food Policy Research Institute (IF PR1).

Waller, David. 1993. Rwanda: Which way now? Oxford: Oxfam.

Wassmcr, P. 1984. L'érosion et ses variables régulatrices dans un boisement d'eucalyptus,
l'cxemple du boisement dc Josi. Bulletin Agricole du Rwanda 17:94-99.

--. 1991. La surpopulation et ses consequences dans les hautes terres de la zone
intertropicale, le cas du Rwanda. Geo-Eco-Tl'op 15:71-89.

Wassmcr, Patrick, and Michel Pierre. 1986. S pulation ct erosion. L'exemple du
Rwanda In Changements globaux en A ique durant le quaternaire: Passe,
present, firtur. , ed. Anon. INQUA-ASEQUA Symposium International, Dakar,
21-28 Avril 1986. Paris: Ed. de l'ORSTOM.

Weber, William. 1987 . Ruhengeri and its resources. An environmental profile of the

Ruhengeri Prefectme, Rwanda Kigali, Rwanda: Ruhengeri Resource Analysis
and Management Project.

250

Webster, C. C., and P. N. Wilson. 1980. Agriculture in the tropics. 2nd ed. London and
New York: Longman.

Wilcock, David, and Valens Ndoreyaho. 1986. The Rwanda social and institutional
profile. Volume two: Agriculture at a watershed. Washington, D.C.: Development
Alternatives, Inc. for USAID.

Williamson, Oliver E. 1985. The economic institutions of capitalism: Firms, markets,
relational contracting. New York: The Free Press.

Wils, W., M. Carael, and G. Tandem. 1986. Le Kivu montagneux: Surpopulation, sous-
nutrition, erosion du sol. Brussels, Belgium: Academic Royale des Sciences
d'Outre Mer, Classe des Sciences. Etude prospective par simulations
mathématiques.

Wischmeier, W. H., and D. D. Smith. 1978. Predicting rainfall erosion losses: A guide to
conservation planning. Washington, DC: US. Department of Agriculture.

World Bank. 1991. Rwanda agricultural sector review. Washington, D.C.: World Bank.

--. 1993a. Project completion report. Rwanda: Agricultural research project. Washington,
D.C.: World Bank, Africa Region.

--. 1993b. Uganda: Agriculture. A World Bank Country Study. Washington. D.C.:
World Bank.

--. 1995. World Development Report I 995. Oxford: Oxford University Press for the
World Bank.

--. 1996a Global economic prospects and the developing countries. Washington, D.C.:
World Bank.

--. 1996b. Implementation completion report. Rwanda: Second Integrated Forestry
Project. Washington, D.C.: World Bank, Africa Region.

--. 1998. Rebuilding an equitable society: Poverty and poverty reduction after the
gvenolciideéniwanda poverty update. Report No. 17792 RW. Washington, D.C.:
or B .

Yamoah, C. F. 1991. Choosing suitable intercrops prior to pruning Sesbania hedgerows
in an alley configuration. Agroforestry Systems 13:87-94.

Yamoah, C. F., J. R. Burleigh, and V. J. Eylands. 1992. Correction of acid infertility 1n
Rwandan oxisols with lime from an indigenous source for sustainable cropping.
Experimental Agriculture 28.417-24.

Yamoah, C. F., and M. Mayfield. 1990. Herbaceous legumes as nutrient sources and
gover crops in the Rwandan highlands. Biological Agriculture and Horticulture
:1-15.

251

 

Yormg, Anthony. 1989. Agroforestry for soil conservation Wallingford, UK and Nairobi:
CAB International and ICRAF.

--. 1995. Agroforestry as a viable alternative for soil conservation. Agriculture and Rural
Development 2:45-49.

Zaag, P. van der, R. S. Yost, B. B. Trangmar, K. Hayashi, and R. L. Fox. 1984. An
assessment of chemical properties for soils in Rwanda with the use of
geostatistical techniques. Geoderma 34:293-314.

252