Citation
Agricultural Policy, Crop Failure and the ’Ruriganiza’ Famine
(1989) in Southern Rwanda: a Prelude to Genocide ?
by
Philip VERWIMP
Development Economics
Center for Economic Studies
Discussions Paper Series (DPS) 02.07
http://www.econ.kuleuven.be/ces/discussionpapers/default.htm
June 2002
Agricultural Policy, Crop Failure and the ‘Ruriganiza’
Famine (1989) in Southern Rwanda :
a Prelude to Genocide ?
Philip Verwimp
Economics Department, Catholic University of Leuven
Naamsestraat 69, 3000 Leuven, Belgium
Research Fellow, Genocide Studies Program, Yale University
Email : philip.verwimp@econ.kuleuven.ac.be
Abstract of the paper
The paper analyses the agricultural policy of the Habyarimana regime, which ruled Rwanda from 1973 to 1994.
Econometric analysis of rural household survey data is used to investigate the effects of the 1989 crop failure in
southern Rwanda on children’s health status. The paper shows that children in southern Rwanda are chronically
malnourished, more then in other prefectures of Rwanda. It is shown that the 1989 crop failure developed into
famine and the causes of this development are investigated. It turns out that the Habyarimana regime did not
respond to early warnings of famine conditions and pretend it did not know what was going on. The relationship
between this non-response to famine, agricultural policy in general and the 1994 genocide is demonstrated.
J.E.L : I12, Q18
Keywords : agriculture, famine, survey research, Rwanda
This research was funded by the Fund for Scientific Research (Belgium, Flanders). The author thanks Daniel Clay
for the permission to use the Michigan University data file and Lode Berlage and Stefan Dercon for many helpful
comments during the research process. The author also thanks numerous Rwandans and expatriates who were
interviewed on food policy in Rwanda or who helped to trace documents that could inform the author on the topic.
The paper was presented at the Frontiers of Development Economics conference, Leuven, May 30th , 2002. All
responsibility remains with the author.
1
Agricultural Policy, Crop Failure and the ‘Ruriganiza’
Famine (1989) in Southern Rwanda :
a Prelude to Genocide ?
“Rwanda is clearly at a crossroads, in that the old strategy is no longer viable : the vision of a nation of selfsufficient peasants, meeting through their labor alone their needs for food and shelter , leading tranquil and
meaningful lives centered around the local community, unbeholden to the world without, that vision is no longer
sustainable.”
World Bank, Rwanda Agricultural Strategy Review, 1991, p.1
1. INTRODUCTION
In the second half of the eigthies, Rwanda entered a period of economic decline. Food
production per capita decline, low coffee and tea prices , unfavourable weather conditions,
unsolved refugee questions, demographic growth, mounting corruption and political unrest
characterise this period. With the fall of the Berlin Wall, western donors began to demand
democratization, also in Rwanda. On top of a drought and a local crop failure in 1989 came a
rebel invasion in late 1990. These rebels were the offspring of the 1959 Tutsi refugees who had
not been allowed to return to Rwanda. The government argued that « Rwanda did not have
enough land to settle the former refugees » (MRND decision, 1986). In the subsequent years,
Rwanda experienced a growing demand for democratisation both from oppostion parties within
the country, from the donor community and from the rebels. From 1990 to 1993 a series of
battles occurred between the rebels and the government forces, the latter winning these battles
with the support of French paratroopers. A peace agreement was signed in Arusha in 1993. Hard
2
core MRND supporters however did not accept power sharing with the rebels and had already
begun preparing large scale massacres of the civilian Tutsi population from the beginning of the
civil war.
It is in this context of growing political and economic unrest that a nation-wide rural household
survey was conducted by the Rwandan Department of Agricultural Statistics (DSA) and
Michigan State University under the direction of Dan Clay. The first round of this survey took
place from October 1988 till March 1989 (the 1989a season), the second round took place from
April 1989 till September 1989 (1989b). This was repeated in 1990 and 1991. The survey thus
contains data from a sesaon before the 1989 local crop failure (which took place in 1989b), from
the crop failure season itself and from two seasons following the local crop failure (1990a and
1990b). Since the civil war broke out in October 1990, the survey also has data from two seasons
following this outbreak (1991a and 1991b). The 1989b crop failure was ‘local’, i.e. limited to the
southern prefectures (especially Gikongoro). The civil war could also be labelled ‘ local ‘ , at least
in the time frame covered by the survey (the 1991b season ran till September 1991) because the
battles were restricted to the northern prefectures (especially Byumba and Ruhengeri).
In fact, the above mentioned survey consisted of several parts : an agricultural survey (6 rounds
1989-1991), a market transaction survey (4 rounds, 1990-1991), a coffee survey (1992, covering
the three previous years) and a health survey (3 rounds, one in 1991 and two in 1992). The
sample size is 1248 households, representing 78 clusters (one cluster in one sector of one
commune) of 16 households per cluster. This method yields a nationally representative sample.
In the present study, these surveys will be linked together to obtain one large dataset with
detailed information on household members, health, agricultural production, land, livestock, and
market transactions. The outcomes of these surveys can be compared to the 1984 national
household survey (which used another sample) to get a picture of changes over time.
The main objective of the paper is the study of the 1989 crop failure. For this purpose, we will
use agricultural data on production and consumption and data on the nutritional status of
children in Rwanda. It is generally accepted that height for age is a good indicator of long-run
child nutritional status. Low height for age, compared to an international standard is an indicator
of chronic malnutrition. We will study the determinants of the nutritional status in Rwanda and
relate it to the 1989 crop failure, the ruling agricultural policies, the civil war and the subsequent
3
genocide. Recently, genocide scholars have argued that more attention should be given to study
the links between famine and genocide (Jonassohn, 1999). It is the purpose of this paper to relate
the agricultural policy of the Habyarimana regime and the 1989 famine with the genocide. In
doing so, we want to combine the tools of economic analysis with the passion for human rights.
The content of the text is as ordered as follows : part two of the text discusses the characteristics
of the rwandan peasant economy, part three describes the 1989 food crisis and entitlement
failure. . In part four, we discuss the causes of the famine and relate these to agricultural policies.
Part five presents the econometric analysis of children’s nutritional status and part six links the
famine with the genocide.
From the data, and this should be stressed, it wil become that we are not dealing with a famine
that killed several thousand people in the course of a few months. In terms of mortality, we
cannot talk of a real famine. If however we consider famine as a prolonged period in which a
serious food crisis breaks down the fabric of society, the word ‘famine‘ applies to southern
Rwanda in 1989. We are dealing with a creeping agricultural crisis, the result of several years of
hardship in which households have depleted resources and cannot cope anymore with another
crop failure.
2. AGRICULTURE IN RWANDA
2.1. The income of Rwandan households
Fully 93% of Rwanda’s population live in rural areas and nearly all rural households are engaged
in farming. On average, households cultivate 0.89 hectare of land, the vast majority of these
landholdings being owner operated. Beans, sorghum, sweet potatoes and casssava are the main
food staples and bananas, potatoes and coffee the main cash crops. Farming is labor intensive,
women’s labor is particularly important in food crop production and men’s labor is crucial in
cash crop production and animal husbandry. Hoes and machetes are the basic farm implements ;
animal traction is nonexistent. Marginal lands once set aside for pasture or left in long fallow are
4
now taken into cultivation. Rural informal and formal credit markets are severely
underdeveloped.1
According to Bart (1993), the Rwandan peasantry has set up an inventive and efficient
agricultural system.2 This system, that has managed to avoid famine most of the time, is a
complex mixture of food and cash crops, cultivated on slopes and in valleys, combined with
some small livestock (and cattle for the richer peasants) and some income from beer brewing,
other business activities or an off-farm income. The chosen mix being the result from personal,
household, local and climatic characteristics. Between the altitude of 1500m and 1700m, we find
the ideal conditions for many crops, explaining the high population densities found on this
altitude level. Clay et al. (1995) have calculated that two thirds of the cultivable land in Rwanda is
cultivated. The rest consists of fallow land, pasture and woodlot. Half of Rwanda’s cultivated
fields are intercropped and 56% of these grow bananas in association with food crops such as
beans, sorghum and sweet potatoes.3 Intercropping appears to be a response to land scarcity, as
it is practised more often by households with relatively few land per person. Less intercropping
occures in high-altitude areas where few or no bananas are grown. In terms of shares of land, the
main crops are bananas (26%), beans (17%), sweet potatoes (11%), cassava (9%), sorghum (9%)
and mais (7%).
Two-thirds of bananas grown are beer bananas, making beer brewing an important source of
cash income for Rwandan households. Bart (1993) reports that this income varies between 500
and 3000 RWF per month, representing on average 50 liters per household per month. For
owners of large banana fields, this amount can be much higher. In my own calculations with the
DSA data set , I found that beer sales represent on average 10% of gross household income, the
equivalent of 5800 RWF per year and totalling 267 liters sold. If we only take the beer selling
households (78%), average gross beer income is 7500 RWF, representing 13% of gross income.
1
2
3
Clay, D., Reardon, T., Kangasniemi, J, Sustainable Intensification in the Highland Tropics : Rwandan
Farmers’ Investments in Land Conservation and Soil Fertility, Economic Development and Cultural Change,
1998, p. 363.
Bart, F., Montagnes d’ Afrique, terres paysannes, le cas du Rwanda, 1993.
Clay, D., Promoting Food Security in Rwanda through sustainable agricultural productivity : meeting the
challenges of Population pressure, Land Degradation and Poverty, International Development Paper no. 17,
Michigan State University, Departements of Agricultural Economics and Economics, chapter 3, p.16.
5
Table 1: Average gross income and the sources of income (1990), n=1182
Gross income
per household
Sources of income at the household level (in %)
RWF
US$
(1)
subsistence
cons. of crops
55.945
700
54.0
(1)
(2)
(3)
(4)
(5)
(6)
(2)
crop sales
(3)
beer
8.6
10.4
(4)
livestock
cons.
sales
3.4
5.1
(5)
off-farm
(6)
transfers
14.7
3.8
includes crop consumption from own production
includes sales of all crops (food, domestic cash crops and export crops)
includes the sales of artisanal brewed beer (banana and sorghum been)
includes livestock and livestock products consumed from own production and sold
includes income from skilled&unskilled off-farm work and from business activities other than beer sales
includes all kinds of gifts of food, beer and livestock received
Source: Author’s calculations from DSA data set.
Next to income from beer sales, the average Rwandan household has income from crop
production for own consumption (54% of gross income), from crop sales (8.6% of gross
income), from livestock and livestock products 8.5%, from off-farm activities (14.7%) and from
transfers (3.7%). On average, monetary income - the sum of beer sales, crops sales, livestock
sales and off-farm income - per household is about 38.9% (8.6+10.4+5.1+14.7) of total gross
income, some 21.768 RWF per year.
2.2. Land scarcity and population pressure
It is well-known that the average size of the Rwandan farm is very small. In 1984 it was 1.2 ha
per household and decreased to 0.89 ha per household in 1990. This is a reduction of 25% over
just 6 years. Since these are averages, it means that many peasants have less then 0.7 ha, the
generally accepted minimum size needed to feed the average household. Population density in
Rwanda in 1990 was 300 persons per square kilometer, the highest in Africa and one of the
highest in the world.
Kangasniema (1998) states that economic theory does not have a clearcut answer on the
relationship between agriculture and commercialization under population pressure. On the one
hand, declining farm size pushes farmers to cultivate and sell crops that have high returns per
6
acre of land and to buy those crops that have only low returns to the scarce resource.4 On the
other hand, the existence of transaction and transportation costs and the associated failures in
the markets for food and other products may make households subsistence-oriented. Only
empirical research can find out what farmers actually do.
Farms are not only small in Rwanda, but also highly fragmented. Blarel a.o (1992), using a 19871988 World Bank Survey data set of 232 households, found that 40% of farms have 8 or more
parcels. In their regressions however, neither farm size nor farm fragmentation seem to have a
negative effect on land productivity.5 Place and Hazell (1993), using the same World Bank data
set, found that Rwandan peasants invest – defined as short term improvements (continued
mulching or manuring) or long term improvements (planting trees, trenching, destumping or
green fencing) – more in land when they have secure rights on their land.6 But this does not
mean that yields on these fields are higher. Place and Hazell at least did not find a productivity
increasing effect of land rights. They point out that other constraints can be more binding such
as technology , credit and fertilizer.
The findings of both Blarel a.o and Place and Hazell are not confirmed in a study by Byiringiro
and Reardon (1996) who used a much larger (nationally representative) 1990-1991 data set of
1248 rural households. They find a strong inverse relationship between farm size and land
productivity and a positive relationship between farm size and labor productivity.7 Small farms
invest twice as much per hectare in soil conservation compared to large farms. Byiringiro and
Reardon show that erosion severely reduces farm yields in Rwanda.
Several reports published by a research team from Michigan State University show a clear and
dramatic decline in yields of all major crops between 1984 and 1991. As a result of this decline
and of population growth, per capita food production dropped by 25%. Half of the farmers in
the DSA data set report declining productivity. The fact that 1984 was considered to be a modest
4
5
6
7
Kangasniemi, J., People and Bananas on steep slopes : Agricultural Intensification and Food Security under
Demographic Pressure and Environmental Degradation in Rwanda, Ph.D Disseration, Department of
Agricultural Economics, Michigan State University, 1998, p.127.
Blarel, B.,Hazell, P., Place, F., Quiggin, J., The Economics of Farm Fragmentation : Evidence from Ghana
and Rwanda, The World Bank Economic Review, Vol. 6, no. 2, p. 252.
Place, F and Hazell, P, Productivity effects of Indigenous Land Tenure Systems in Sub-Saharan Africa,
American Journal of Agricultural Economics, February 1993, p.14-15.
Byiringiro, F. and Reardon, T., Farm Productivity in Rwanda : the effects of farm size, erosion, and soil
conservation investments, Agricultural Economics 15, 1996, p. 132-135.
7
drought year suggests that the observed decline in production between 1984 and 1989-1991 was
real.8 The report also states that the decline for tubers, the main providers of calories for the
poor, is particularly strong. Half of Rwanda’s farmland suffers from moderate to severe erosion.
Regional differences are outspoken : the Northwest (Gisenyi prefecture) agroecological zone
produces twice as much output per unit of land as the Southwest (Gikongoro and Cyangugu
prefecture).
The driving force behind declining land productivity, soil erosion and declining per capita
incomes is population growth. Peasants are forced to cultivate marginal unfertile lands, often on
steep hill sides. Boserup (1965) has argued that agricultural innovation is spurred by decreasing
productivity. She argues that as land becomes scarcer, peasants will adopt shorter fallow periods,
apply more fertilizer and work the land more intensively. This process will allow the peasantry to
avoid the malthusian trap over population growth and hunger.
From a survey of 14 case studies, Wiggins (1995) has concluded that that the Boserupean model
of population-induced agricultural intensification is a correct representation of evolving realities
in the African countryside.9 He argues a.o. that farmers have seized opportunities to grow new
crops as marketing becomes more attractive as a result of better roads. Platteau (2000) however
considers this picture too optimistic and points at several constraining factors, especially at the
technological level.10 He argues that intensification has proceeded at a very slow pace relative to
population growth, in spite of improved access to markets. Interestingly, Platteau considers
Rwanda a perfect illustration of this distressing possibility :
« Population growth has reached extemely high levels without giving rise to major technical progress
susceptible of providing a decent livelihood to the growing number of peope living on the land. »
2.3. Food crops, coffee and banana beer
The effect of population pressure on crop choices and land use in Rwanda has been analysed by
Kangasniemi (1998). Using the 1989-1991 nationally representative farm household survey I will
also use, he finds that Rwandan farmers prefer to keep and even expand their banana holdings.
8
9
Clay, D. et al., Promoting Food Security in Rwanda, ibidem, Michigan State University, chapter 3, p. 34.
Wiggins, S.Changes in African Farming Systems between the mid-1970s and the mid-1980s, Journal of
International Development, vol.7, n°6 ,1995, pp.807-848
8
In fact, under population pressure, farmers prefer to grow bananas, brew and sell banana beer.
With the money, the farmers buy food crops and finance other expenditures. Bananas are
popular because they are not only a source of monetary income, they also provide income the
whole year round, a significant difference with e.g. coffee. Bart (1993) writes that banana
cultivation is not labour-intensive, is socially very important and provides protective cover
against erosion. All these factors combine to make bananas the most important and most
preferred crop for Rwandan peasants. Clay et al. (1995) add that the farming practises needed
for higher yields require more labor then other crops. The brewing activity of course is also
labor-intensive.
Kangasiemi (1998) argues that the exchange strategy (production of banana beer for
consumption of beans) is the preffered strategy by the Rwandan farmer to cope with population
pressure :
« 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 bananas per hectare. In this respect
bananas are different from any other major crop, including coffee. This suggests that Rwandan farmers are
not substituting food crops for cash crops to cope with land scarcity caused by population growth. While
households with less land per person are not making any dramatic transition to cash crops either, they
clearly prefer to keep their bananas and to rely on the exchange of banana beer for food. ‘ Food first’ is not
their strategy of achieving food security. » (p.133)
The main argument in support of the economic rationale behind the exchange strategy is simple :
it can be computed that the cash revenue from a hectare of beer bananas sold as beer would on
average have bought 1446 kilogram of beans (Kangasniema, p.157). This is 72% 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 improves the food entitlements of those engaged in the exchange strategy.
The importance of the brewing of banana beer is such that 80% of Rwanda’s rural households
get more cash from bananas than from coffee or any other sigle crop, and that, for 41% of the
households, bananas are a more important source of cash than all other crops combined.
Clay et al. (1995) report that sweet potatoes are the single most important source of calories for
Rwandans. Sweet potatoes have more calories per-kilogram than potatoes or cooking bananas.
10
Platteau, J.Ph., Institutions, Social Norms and Economic Development, Harwood, 2000, ch. 2, p.26.
9
Except for high-altitude areas, where potatoes grow best, only bananas produce more calories
per-hectare than sweet potatoes. Sweet potato production is not labor-intensive, compared to
bananas beer production or potatoes and it is less demanding of soil quality and moisture.
2.4. Regional differences in 1990
Table 2
Prefecture
average
land size
per hh
average
gross y
per hh
monetary*
income as %
of gross y
Ruhengeri
Butare
Buymba
Cyangugu
Gikongoro
Gisenyi
Gitarama
Kibungo
Kibuye
Kigali
Rwanda
0.79
0.81
1.01
0.66
0.89
0.45
0.92
1.35
1.16
0.95
0.89
65.443
39.134
63.702
33.878
30.376
51.175
53.828
113.170
32.757
70.513
55.954
40
37
41
41
39
45
38
37
33
38
39
Altitude
(m)
Rain
(mm)
Distance
to paved
road (km)
2115
1660
1888
1917
1917
1946
1630
1469
2100
1581
1802
1124
1120
919
1567
1556
1313
1070
733
1275
822
1126
11
14
17
30
28
9
15
18
66
31
22
N=1146
* Monetary income is the sum of income from crop sales, beer sales, livestock sales and off-farm income.
In terms of gross household income, the southern prefectures of Butare, Gikongoro and the
eastern prefectures Kibuye and Cyangugu are poorer than the northern, central and western
prefectures. Compared to Ruhengeri and Kigali prefectures for example, average gross
household income in the southern prefectures is only half. This has not much to do with the size
of the farm : farms in Rwanda are very small everywhere (with the possible exception of
Kibungo prefecture). Regional differences in household income are better explained by soil
fertility and productivity, suitability of the soil for high yielding crops (such as sweet potatoes in
the north), access to public infrastructure (such as roads and markets), availability of off-farm
jobs, the price of coffee and tea for regions growing these crops. We come back to regional
differences in the next section, where we discuss the food crisis that struck southern Rwanda in
1989.
10
3. THE FOOD CRISES AND ENTITLEMENT FAILURE OF 1989
3.1. The entitlement to food
Farmers will use part of their production of food crops for own consumption. A. Sen calls this
the direct entitlement of food the farmer has. When the farmer also earns a wage, produces cash
crops, receives transfers or has another source of income which he can use to buy food, the
farmer has an exchange entitlement in Sen’s terms. Most farmers have a combination of direct and
exchange entitlements to food. A farmer producing beans has a direct entitlement to beans. An
increase in the price of beans will not endanger the direct entitlement of the bean producer.
Farmers who are not producing beans will see their exchange entitlement to beans decreased
when,ceteris paribus, the price of beans increases. This allows us to stress the importance of
relative prices in the conservation of a farmer’s exchange entitlements. As 80% of Rwandan
peasant households are net buyers of beans, the price of beans is very important.
Taking a look at direct entitlements first, we note that beans, bananas, potatoes and sweet
potatoes are the major food crops for Rwandans. Bananas are often used to brew beer and thus
give acces to exchange entitlements. Coffee is an export crop and also gives acces to exchange
entitlements. In 1989, production of beans and sweet potatoes declined sharply in Gikongoro.
Since sweet potatoes are the most important crop in Gikongoro (in contrast to Rwanda as a
whole were bananas are the most important crop), this decline caused a severe loss of direct
entitlement to food. Together with the production of sweet potatoes, the production of beans
declined as well with 50 % in both seasons (see figure 2). Importantly, both figures 1 and 2 show
that the production decline in 1989 was a local and not a country-wide phenomenon. The value
of production in Rwanda declined gradually over the years (as did the quantity of bananas
produced), but there was no sudden decline in 1989. That yeat even shows a upsurge in the
national value of production.
In contrast to direct entitlment failures, which can occur as a result of bad weather, rain, crop
disease, drought or other calamnities, exchange entitlement failures depend on relative prices and
thus on market forces. The entitlement approach focusses on starvation and not on famine
11
mortality. People starve when they cannot command enough food. As Sen indicates, many
famine deaths are the result of epidemics, but this is not of our interest here. People will more
easily become victim of sickness or desease when they lack adequate food consumption. The
command over food a peasant has, by means of his direct and his exchange entitlements over
food is thus at the centre of our attention. The production decline in seasons 1989a and especially 1989b
in Gikongoro (and parts of Butare) prefecture (see figure 1) are evidence of direct entitlement failures. This
affects consumption if households are not able to compensate for this trough purchases of food
on the market. In the next sections we will see whether exchange entitlement failures were also
part of the peasant situation.
Figure 1: Value of production over ten seasons in RWF (1987 a season prices)
32000
28000
24000
20000
16000
12000
Gikongoro
8000
1987a
Rwanda
1988a
1987b
1989a
1988b
1990a
1989b
1991a
1990b
1991b
source : author’s calculations from DSA dataset
Figure 2 : Production of major crops in Rwanda and Gikongoro
in average kilograms per household per year/season
Production of beans
140
140
120
120
100
100
80
80
60
60
40
40
20
20
0
1987a
0
1988a
1989a
1990a
1991a
1987b
1988b
1989b
1990b
1991b
12
Production of bananas
1400
1400
1200
1200
1000
1000
800
800
600
600
400
400
200
200
0
0
1987a
1988a
1989a
1990a
1987b
1991a
1988b
1989b
1990b
1991b
Production of sweet potatoes
800
800
700
700
600
600
500
500
400
400
300
300
200
200
100
100
0
1987a
0
1988a
1989a
1990a
1991a
1987b
1988b
1989b
1990b
1991b
Production of coffee
30
30
20
20
10
10
0
0
1989a
1990a
1991a
1989b
1990b
1991b
Legend : full line ____ for Rwanda, dotted line ------ for Gikongoro
Left-hand graphs are for the first season (October-March), right-hand graphs are
for the second season (April-September)
In order to detail the relative importance of different food items consumed by Rwandan
households, we look at table 3. It is shown that, in terms of value, beans , bananas and sweet
potatoes were the most important crops in the Rwandan diet in 1990. Together they constitued
67% of the average crop consumption per month in the 1990a season. Very few meat is
13
consumed, wheras large quantities of banana beer were consumed. The later took 20% of total
consumption in Rwanda, 15% in Gikongoro and 7% in Ruhengeri. Table 3 also shows other
regional differences, in particular the importance of potatoes in the Ruhengeri diet and sweet
potatoes in Gikongoro.
Table 3: Consumption data per household per month (season 1990a)*
Crops
Beans
Potatoes
Peas
Mais
Manioc
Bananas
Sweet potatoes
Sorghum
kg
Rwanda
value
%
21.96
16.85
0.99
12.69
16.43
42.14
46.70
6.37
869.62
282.74
49.50
288.82
281.45
715.54
683.22
190.54
26
8
1
9
8
21
20
6
Total of 8 crops
Adult equivalents
Consumption of 8 crops,
meat and beer per ae
per month
8.54
8.89
1.33
4.07
3.70
2.17
51.27
4.84
3360.9
Meat
Banana beer (liter)
kg
889.05
(15.6)***
4.90
904.33
354.59
155.71
66.56
92.67
67.32
39.87
726.33
141.40
22
9
4
6
4
2
44
9
kg
24.36
59.55
1.58
18.52
3.30
15.16
46.08
19.05
1644.4
181.3
(48.7)**
28.22
Gikongoro
value
%
311.38
(20.9)***
4.59
441.29
946.04
833.82
79.11
421.69
71.14
290.73
629.80
549.55
25
22
2
11
2
8
16
14
3821.8
69.76
(67.7)**
9.73
Ruhengeri
value
%
207.16
(42.1)**
10.63
340.41
(30.1)***
5.09
858.42
*
quantities are the sum of production in season 1990a minus kg sold and given to others plus kg bought and
received from others. Prices are average market prices in RWF per kg in season 1990a for Rwanda and in
the prefectures of Gikongoro and Ruhengeri. Percentages is the percent of the total value of 8 crops.
** % households with no meat consumption
*** % households with no consumption of banana beer
14
3.2 Some Exchange entitlements11
3.2.1 Coffee versus beans
Export crops are another way to secure one’s living in a peasant economy. In Rwanda, peasants
had three main ways to earn some monetary income : selling crops on the market , brewing beer
or performing off-farm labor. Now, we concentrate on the export crop, coffee. Table 4 shows
the evolution of coffee cultivation over several years. For the whole of Rwanda, about half of the
peasants were cultivating coffee in 1989. On average, they held about 150 coffee trees. Over
three years, this number stayed about the same, but the income from coffee dropped by one
third. The regional differences (not in the table) are sharp : households in Gikongoro doubled
their number of trees between 1989 and 1991 only to end up in 1991 with half of the coffee
income compared to 1989. Wether this investment strategy was the right one, remains to be
discussed. Altough peasant investment behaviour is not the main topic of this chapter, we need
to look at the value of coffee in relation to other crops in section (5.1.2) of this chapter. The
hopes of ‘a good coffee harvest ’ expressed by farmers in discussions with NGOs on the crisis,
however have not come out. In Butare, coffee income also dropped sharply over three years
time, but peasants did not invest in new coffee trees.
Table 4 : Coffee production over several years, household level averages
% households doing coffee
number of coffee trees
kg sold
coffee income (RWF)
1989
46.7
149
48.3
6038
1990
155
48.4
4846
1991
154
34.1
3932
89-91%
+3.3
-29.3
-34.8
Coffee was Rwanda’s most important export crop and was responsible for high state revenue in
the late seventies (see chapter 2). For the farmer, coffee is a source of monetary income. As long
as the producer price of coffee is high, the farmer has an economic incentive to grow coffee.
11
We deal with coffee and livestock in this section. For Rwanda as a whole, the exchange entitlement of banana
beer is important as well, but coffee and livestock are more important than bananas in Gikongoro. In addition,
famine researchers also study the exchange entitlement of the agricultural wage. However, I only have data on
wages from the 1990 and 1991 seasons and thus not during the crop failure season. Wage data from 1990 and
1991 do not show entitlement failures for wage earners.
15
When that price decreases, he may want to replace coffee with other crops. Bananas and beans
are the major competitors of coffee for land allocation in the peasant household. By decree,
coffee trees may not be intercropped, neglected, abandoned or destroyed. As can be observed
from figure 3, the price of coffee was fixed by the government, first at 122.5, then 125, then 100
and later at 115. The relative price of coffee versus beans (indexed) however has schown a severe
decline between 1986 and 1993. Where 1kg of coffee would buy the farmer 1kg of beans in
1986, the farmer needed 3kg of coffee for 1kg of beans at the end of 1993. It is not surprising
then, that towards the end of the eighties, farmers started ripping out coffee trees, an act
forbidden by penal law. Table 4 on coffee production already showed that the average coffee
income in Rwanda decreased by 1/3 in the course of 3 years (1989-1991). This general decline
was due to decreased production per tree, but masked big differences between prefectures.
Gikongoro prefecture is a case in point : average coffee income decreased by 50% , but the
number of trees had more then doubled in the 1989-1991 period. This means that Gikongoro
farmers had invested heavily in new trees, which only bear fruit after three years.12 From the
scarce but relevant literature on hunger in Gikongoro in 1989, we retain that peasants tell
investigators that they will starve next season when the next coffee harvest fails :
«Everybody’s eyes are focussed on the coming coffee season. When it is good, they can survive, when not, they
will die.» 13
As for the famine months of the 1989b season, coffee farmers in Gikongoro and in the whole of
Rwanda experienced a sharp exchange entitlement failure versus beans. The 1989b price shock
of beans and the (subsequent) loss of exchange entitlement however was only the beginning of a
more serious decline of exchange entitlements in the 1991a season and thereafter The price of
coffee remained at the 115 level till the beginning of the genocide (April 1994).
12
13
These investment figures are analysed in detail in the chapter 2, where I discuss the data on the coffee
economy.
Les Retombées de la Famine dans les Préfectures de Butare et de Gikongoro, Bureau Social Urbain, Caritas,
Kigali, Février 1990, p. 17.
16
Figure 3: The price and the exchange entitlement of coffee
Rwanda
Gikongoro
350
300
250
200
150
100
beans
50
coffee
0
exchange rate
Index of bean and coffee prices
index of bean and coffee prices
350
300
250
200
150
100
50
0
93
-19
CT 3
-O 99
01 R-1 2
P
-A 99
01 T-1 2
C
-O 99
01 R-1 1
P
-A 99
01 T-1 1
C
-O 99
01 R-1 0
P
-A 99
01 T-1 0
C
-O 99
01 R-1
P 9
-A 98
01 T-1
C 9
-O 98
01 R-1
P
-A 88
01 T-19
C 8
-O 98
01 R-1
P
-A 87
01 T-19
C 7
-O 98
01 R-1
P
-A 86
01 T-19
C 6
-O 8
01 R-19
-AP
01
93
-19
CT 93
-O 19
01 PR- 2
-A 99
01 CT-1 92
-O 19
01 PR- 91
-A 19
01 CT- 91
-O 19
01 PR- 90
-A 19
01 CT- 0
-O 99
01 R-1 89
P
-A 19
01 CT- 9
-O 98
01 R-1 8
-AP 98
01 CT-1 8
-O 98
01 R-1 7
P 8
-A 9
01 CT-1 87
-O 19
01 PR- 6
-A 98
01 CT-1 86
-O 19
01 PR-A
01
DATE
DATE
3.2.2 Livestock versus beans
Table 5: Average livestock holdings over several years, in TLU ’s (*)
Ruhengeri
Butare
Byumba
Cyangugu
Gikongoro
Gisenyi
Gitarama
Kibungo
Kibuye
Kigali
1989
0.82
1.19
1.04
0.31
0.84
0.83
0.97
0.77
1.24
0.93
1990
0.84
0.88
0.95
0.32
0.78
0.87
0.94
0.79
1.21
0.94
1991
0.79
1.06
0.93
0.20
0.65
0.76
0.97
0.78
1.29
1.02
% change
89-91
-3.6
-10.9
-10.5
-35.4
-22.6
-8.4
0.0
+1.3
+4.0
+9.6
Rwanda
0.91
0.86
0.86
-5.4
%households without livestock
89
90
91
20.3
17.1
26.0
36.6
44.8
49.0
35.5
39.1
47.3
42.7
53.9
58.4
24.7
39.3
43.8
32.8
35.3
35.3
40.8
44.9
46.9
30.4
28.3
21.7
22.1
25.6
27.9
30.1
23.1
26.6
32.0
35.1
38.4
(*) Tropical livestock units for Rwanda are calculated as cattle 1; pig 0.25; goat 0.17; sheep 0.17;
1140 other hh were surveyed in three rounds 1989-1991.
Table 5 shows that Rwandan households, on average, have decreased their livestock holdings
during the three years of the survey. During this period, household livestock holdings (in tlu’s)
declined on average by 5.4%. The decline in the southern prefectures, Gikongoro and Cyangugu
was much larger than in other prefectures. Gikongoro prefecture has the highest percentage of
households who lost all livestock between 1989 and 1991, namely an increase of 77% of
17
households without livestock (the Rwandan average being 20% increase). Kibungo, Kibuye and
Kigali prefectures on the other hand have seen an increase in the number of households holding
livestock over this period..
With incomplete risk markets - entailing interdependence of production and consumption
decisions – a production input such as livestock can be an important asset which helps protect
the consumption of poor people (Ravalllion, 1997, p.1223). It is known from other countries
that the value of livestock decreases in periods of famines. Sen for example, writing on the
Ethiopian famine (1972-1974) says that the economic distress of the pastoralists was not
confined only to the loss of animals (by drought or displacement from traditional grazing land) ,
«The exchange entitlement associated with any given stock of animals vis-à-vis grains also fell sharply. It is
sometimes overlooked that a pastoralist does not live only by eating animals or consuming animal products
like milk. He also exchanges animals for other means of sustenance, chiefly grains.» 14
We only have observations on the price of sheep and goats from October 1989 to September
1991, thus missing the high point of the famine in Gikongoro. From figure 4 however, we retain
that holders of goats and sheep in Gikongoro got less value for their livestock in terms of beans
compared to the Rwandan average. As we have stated several times, the 1989-1991 period is a
time of prolonged agricultural crisis in Gikongoro and the livestock data may well be an
indication of this. As crops fail because of drougth, lack of inputs or plant diseases, peasants are
forced to eat or sell (parts of) their livestock in order to survive. When crop failure is a covariate
shock (as opposed to an ideosyncratic one), many households sell livestock at the same time,
thereby causing a sharp decline in the market price of livestock (Ravaillion, 1997, p.1223). From
the livestock data, we cannot say that livestock owners experienced a sharp exchange entitlement
loss verus beans.
14
Sen, A., Poverty and Famines, 1981, p. 105.
18
Figure 4: The price and the exchange entitlement of sheep and goats
300
200
100
beans
livestock
0
exchange rate
index of bean price and livestock
index of bean price and livestock
180
160
140
120
100
80
60
40
1
99
P-1
-SE 1
01
99
L-1
-JU 91
01
-19
AY
1
-M
01 -199
AR
-M
01
91
19
N0
-JA
01 -199
CT
-O
90
01
-19
UG
-A
01
90
19
N-JU
01 990
R-1
-AP 90
01
-19
EB
-F
01
89
-19
EC 9
-D
01 -198
CT
-O
01
91
-19
UG 1
-A
01 -199
N
-JU 1
01 -199
R
-AP 1
01 -199
B
0
-FE
01 -199
EC
-D 90
01 -19
CT
-O 90
01 -19
UG 0
-A
9
01 -19
N
-JU 90
01 -19
PR 0
-A
01 -199
B
-FE 89
01 -19
EC
-D 89
01 -19
CT
-O
01
DATE
DATE
We conclude from this section that coffee producers experienced an exchange entitlement
failure relative to beans in the period of the food crisis. This came on top of a direct entitlement
failure. The price of coffee continued to decline after 1989, making the exchange rate of beans
versus coffee not particularly low in 1989 compared to other years. In general, what we observe
is in fact very close to Sen's observations on the Ethiopian famine (1972-1973). Sen observed
that the peasant population in Wollo starved to death without substantial rise in food prices. 15
The peasants experienced a direct entitlement failure and could not command food at the market
by lack of purchasing power. As Amartya Sen (p. 101) writes
«The biggest group of destitutes in the Wollo famine came from the agricultural background, and indeed
were farmers – both tenants and small land-owning cultivators. The entitlement decline here took the form
of direct entitlement failure without involving the market in the immediate context. The output – typically
of foodgrains – was severely reduced, and this led to starvation in a direct way. In so far as the Ethiopian
farmer eats the food grown by the family without becoming involved in exchange to acquire food, the
immediate influence affecting starvation is the decline of the food grown and owned by the family, rather
than the fall in the total food output in the region as a whole. The distinction is important, since the FoodAvailability-Decline-approach would focus on the latter variable. The hunger of the Wollo peasant had a
more direct origin.
But, of course, once his own crop had failed, the Wollo peasant would have tried to get hold of food trough
the market in so far as he could have exercised market command. But since the agricultural failure also
amounts to a collapse of his source of market command (namely his income), he was not in a position to
supplement his reduced food output by market purchase. The Wollo agriculturalist could not provide much
effective demand for food in the market, and despite widespread starvation the food prices in Dessie and
elsewhere recorded very little increase.»
15
Sen, p. 69.
19
As for peasants in Gikongoro, crop production of the household farm is indeed the major
source of the household income (up to 70%) and household consumption. Only off-farm labour
(which is in short supply) and the sale of livestock products are sources of purchasing power that
are independent of agricultural output. The sale of beer (which depends on the supply of
bananas and sorghum) and crops on the market is not independent of agricultural output. High
bean prices in the beginning of the 1989b season (April-June 1989), can therefore be an
indication that households still had some cash to command food in the market. The absence of
(very) high bean prices in subsequent seasons (1990-1994) therefore may then be an indication
that households on the one hand produced more beans themselves and on the other hand had
no resources anymore to command beans in the market.
4. THE CAUSES OF FAMINE IN SOUTHERN RWANDA
4.1. Food import and export policy
The central political idea concerning agriculture and indeed the central issue of Habyarimana’s
regime was food self-sufficiency. In numerous speeches the president stressed the importance of
food self-sufficiency. This had to be reached by increased agricultural production. In a 1984
speech before heads of state of Africa and France, in Bujumbura, Habyarimana expressed this as
follows :
“In Rwanda, food self-sufficiency has always been the primary objective of our development effort. Given the
periodic famines which regurlarly afflicted us in the past, our population density considered the highest in
Africa, and our extreme shortage of arable land, we have had no option but to intensify food production.
Living constantly at the very edge of subsistance, the Rwandan peasants has had no other choice, and
desired no other choice than to dedicate his efforts to increasing the production of food…” 16
We remark that the vision of Habyarimana concerning the choices and the preferences of
peasants as far as crop production is concerned, is completely opposite to empirical research
findings by Kangasniemi (1998) as we reported in section (2.3) of this chapter.
16
World Bank, Rwanda, Agricultural Strategy Review, 1991, p.12.
20
Every four years, a general development plan was presented in which nation-wide production
targets for the main crops were outlined and where the problems to increase production were
addressed. For Habyarimana, the main method to increase production was to work more and to
work harder. In economist’s terms, to apply more labour to the land. In a 1987 speech
Habyarimana said that his policy of food self-sufficiency should not be equated with autarky. He
said he realized that Rwanda could not produce all the food it needed.
17
But, in stead of food
imports, it was better to rely on food aid, Habyarimana said. The latter namely does not bring the
balance of payment out of equilibrium. Food aid, he added, should be sold above the price of
locally produced crops, in order not to reduce the production incentives of the peasants.
In a 1991 report (p.12) on Rwanda’s agricultural strategy, the World Bank writes that
“The Second Republic has been characterized by the vision of an inward-looking, non-importing,
autonomous approach to food self-sufficiency, at both the national and the farm level. In the case of the
latter, subsistence rather than market production has usually been encouraged. “
Although the Habyarimana regime embraced the vision of a self-sufficient agrarian nation,
Rwanda was far from self-sufficient. In a June 1989 report, an official of the Ministry of
Agriculture published the results of a small survey of informal border trade conducted in April
and May 1989. He shows that Rwanda appears to export substantial manufacturing goods,
potatoes small-stock, poultry and vegetables to the surrounding countries, and that it is at the
same time an important importer of cooking oil, beans, sorghum, bananas, sugar and maize. 18
This survey can be complemented with official trade statistics. Table 6 shows that Rwanda
imported food for a average annual value of 40 million US dollar during the eighties. Rwanda
exported food for 90 million dollar yearly.
17
18
Habyarimana, J, Discourse at the first encounter with public servants, January 30, 1987.
Ngirumwami, J.L., Ministry of Agriculture, Résultats de l‘enquête Commerce Frontalier au Rwanda, June
1989. No case of re-export was found by the author, indicating that all import was meant for domestic
consumption.
21
Table 6 : Value of Food Imports and Exports in million US dollar*
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
Imports
37,3
42,8
37,4
48,5
63,3
53,4
38,5
36,8
41,0
46,7
Exports
67,4
69,3
60,6
62,9
68,7
160,1
121,5
91,2
104,7
88,6
* Source : UCTAD Trade Statistics, 1992 yearbook.
These results indicate two things : first, the vision of the Rwandan leadership did not correspond
with reality : Rwandan agriculture was not self-sufficient, there was a large volume of cross
border trade. Second, trade continued at the time of the crop failure, meaning that there was no
nationwide shortfall in agricultural production, as documented by the production statistics of the
rural household survey (see earlier sections). Table 6 shows us even that the value of food export
doubled in the second half of the eighties, compared to the first half. This remarkable upswing in
food exports is for the moment unexplained, but coincides with a sharp decline in export
earnings from coffee. Food exports could have been used to partly offset the decrease of these
coffee export earnings. At the same time, imports of food in the second half of the eighties
remain at the same level as in the first half. These import figures show 1988, the year before the
crop failure to be the year with the lowest value of food imports in the decade. This has to do
with the 'name ' that was given to the years 1987 and 1988 by the Habyarimana regime. 1987 was
called the year of food self-sufficiency and 1988 the year of the protection of the income of the
peasants. Gatete (1996) writes that the Rwandan government had banned all food imports in
1988.19
According to the Agricultural Strategy Review (World Bank, 1991), it is this informal trade (read
illegal cross-border trade) that is believed to be responsible for the import of beans that
19
Gatete, C., Food Security and Food Aid in Rwanda, Wihogora, June 1996, p. 2.
22
eliminated the food shortage in Southern Rwanda.
20
“Informal trade” means that under normal
conditions the imports of crops such as beans, sorghum and bananas are officially banned. A
division in the Ministry of Commerce was responsible for assessing the needs for imports based
on a projection of national production and consumption. This division only used data on official
imports which became available two weeks after arrival in the country. The division does not
make any projections on consumption or production and does not have information on
unofficial (informal) imports from neighbouring countries.21 Nevertheless it is estimated that in
1986, a fairly good crop year, unofficial imports accounted for 50% of the beans and sorghum
sold on Rwanda’s markets. If the division anticipates food shortages, a complex process of
import licensing was set in motion. During the 1989 crop failure, the World Bank writes (1991,
p.25), the Commerce Ministry decided to authorize ten importers to import beans from Uganda
when the price had reached 60 RWF. This was a too late to prevent starvation and emigration.
The reduction and surely the banning of imports of course is detrimental to a farming
population that, under the pressure of population growth, concentrates efforts on high value
crops. The exchange strategy (banana beer for beans) that we have discussed earlier on, is only
viable when beans are in effect imported into Rwanda. In the absence of imports, the exchange
strategy is useless and farmers have to grow subsistence crops themselves.
As with licensing in general and under famine conditions in particular, large profits and
speculation from the side of the traders and from the side of the license giving authority are
possible. Import policy in Rwanda under the Habyarimana regime was managed by S.
Rwabukumba, member of the Akazu and brother of Agathe Kanziga, the wife of the president.
He headed the powerful company ‘ La Centrale ‘ which was in charge of official import into
Rwanda. At the same time this person, who had not even finished his secondary school, was in
charge of the foreign currency division of the National Bank. Since import cannot occur without
foreign currency, Rwabukumba sat in a key position. ' La Centrale ' did not import basic foods
needed by the Rwandan population, but imported high quality products for the Rwandan elite.
Well-informed people knew that ‘ La Centrale ‘ did not pay import or export taxes.22 Both the
20
21
22
World Bank, Ibidem, p. 26.
World Bank, Ibidem, p. 25.
Reyntjens, F., Testimony, Cour d’ Assisen, Brussels, April 20, 2001.
23
non-food and food imports for the elite took up a large part of the foreign currency. As Bézy
argues, the food that was imported was destined for:23
-
the expatriates from industrialized countries, to allow them to continue their food habits
the growing number of tourists
interns in schools who benefited from non-traditional food
the national elite who lived a european-style of live
the whole of the urban population (5%) who had introduced milk powder, tomato
concentrate, imported cooking oil, far , sugar and rice (whose domestic production remained
too low) into its eating habits.
Reading the 1987 speech of Habyarimana in which he pleaded for import restriction, one gets
the impression that the Rwandan peasantry had to produce more food in order for the Rwandan
elite to continue its import of luxury products, both food and non-food. Indeed, each dollar that
is saved on import of basic foods can be used for something else. Additional information, found
in numerous articles of the journal Kinyamateka, seems to validate this statement : between June
1989 and December 1989, Kinyamateka published several articles in which they revealed that
government authorities embezzled public funds on a large scale. A.Sibomana, priest and editor of
Kinyamateka put it this way :
« Rwanda was a country which still had the reputation of being well run, ruled by a sort of ‘enlightened
despot’, Juvénal Habyarimana. But you don’t become an honest man just by knowing how to quote French
poetry to President Mitterrand. Juvénal Habyarimana and his people were plundering the country while the
peasants were starving. We had evidence that he or his wife were diverting funds allocated to buying food for
the population to import luxury items instead, for example televisions which were sold at vastly inflated
prices. We also had information on drug trafficking. »24
4.2. A malfunctioning food marketing system
Accroding to Pottier (1993), OPROVIA, the leading Rwandan agency in food marketing,
intervened only to a limited scale in food markets.25 It was OPROVIA’s task to assure price
stability, especially for beans and sorghum. The Rwandan government however expected that
OPROVIA could be run as a profitable business, which was not possible when OPROVIA had
to buy surpluses in the post-harvest period above the market price and sell from its stocks in any
23
24
Bézy, F., Bilan d'un régime, 1962-1989, Louvain-La-Neuve, 1990, p. 13. The author also refers to J.
Nzisabira, dans Bulletin des Séances, ARSOM, XXXIV, n° 4, 1987, p. 641.
Sibomana, A., Hope for Rwanda, p. 25.
24
hungry period below the market price. According to Pottier, OPROVIA’s struggle became acute
following the bad harvests from 1988 and the drought in 1989 :
“The impact of the 1988 harvest failures, caused by excessive rainfall, was unusually harsh since the
authorities had banned all food imports that year; 1988 being Rwanda’s Année de l’ Autosuffisance !”26
In fact, as we earlier saw, 1987 was decreed to be the year of food self-sufficiency, not 1988. But
this does not make a big difference. Food imports were low in either year, since the president
wanted to keep the balance of payments under control. One could say that Habyarimana was a
conservative and parsimonious accountant. The 1988 slogan was «the year of the defense of the
peasant income». For 1989, the slogan was «The year of rural self-organisation». The latter
meaning that the peasants should not wait until the administration comes to help them, but take
matters into their own hands.
According to Pottier, it is exactly the policy of food import minimization that made it impossible
for Rwanda to re-supply its food stocks at the time of the 1988 bad harvest. This is one reason
why, one year later in 1989, food stocks where almost empty. Official inquiries revealed that the
Rwandan Government was setting OPROVIA an impossible task. OPROVIA’s impasse, Pottier
writes, was proof that no real progress had been made since the drawing up of agreements with
the EC in 1982. The second report of the Interservice Commission warned, with regard to
Rwanda’s internal food trade :
«I has been noted that even when food is abundant, the shops and silos belonging to the cooperatives and
communes remain poorly stocked, even though these structures are theoretically the most appropriate for
supplying all of the country’s communes. It is therefor a matter of urgency that all the serivices concerned
(ministeries , OPROVIA and agricultural projects) to decide how best the decentralized stockage system
could be activated in the interest of a nationwide distribution of foodstuffs »27
The government of Rwanda nevertheless continued to believe that OPROVIA will take the
initiative for inter-regional food commerce. (Rapport, p.23) The expectation, Pottier argues, that
OPROVIA could raise the funds required for effective inter-regional distribution of surpluses in
times of crisis is unrealistic. In fact, the contrary was true, there were signs of gross neglect
towards the very apparatus for state-controlled marketing :
25
26
27
Pottier, J., Taking stock, Food Marketing reform in Rwanda 1982-1989, African Affairs, p. 6 and p. 15.
Pottier, J., ibidem, p. 15. During the Habyarimana regime, every year was given a slogan. 1988 was called
the year of food self-sufficiency.
Republic of Rwanda, Rapport sur la Situation Alimentaire de notre Pays, Ministry of Agriculture, Octobre
1989, translation by J.Pottier, ibidem, 1993, p. 20.
25
«The Rwandan governement still needs to reimburse OPROVIA the promised 28.000.000 RwF it lost
in 1988 after selling at artificially low prices the sorghum it had bought too dearly in 1986.» 28 And ,
«The Rwandan government should reajust the agricultural prices at which it bought from OPROVIA in
1989 to ensure that this organisation in future can reduce the losses it must incure.» 29
There is also an information problem in the game : one can argue that the population, through
the speeches of Habayrimana (restriction of import), can start believing that future public stocks
of grains or other crops will be low and that the government will not be able to intervene in the
case of a crop failure. This information problem may lead to high expected prices.30
Relying on this discussion of OPROVIA, we argue that the Rwandan elite never really cared to
set up a functioning food marketing system. Pottier (1993, p. 27) is right on the mark when he
writes that
«It may even be fair to suggest that the slow progress – in food marketing – was yet another illustration of
how policy makers tend to regard improved nutritional status as the by-product of agricultural strategies
rather than a goal in its own right.» 31
This is exactly what we have been saying when discussing the peasant ideology of the
Habyarimana regime in another paper of the author : the regime considered the peasants as
producers (of food and coffee), not as consumers.32 The leadership did not equip OPROVIA
with the means to succesfully secure access to food for the destitute; it did not pay back debts; it
banned food imports and increased food exports; it professed to secure the income of the
peasants and considered the application of more labor to agriculture as the solution.
4.3. Agricultural policy under Habyarimana
In 1991, the World Bank writes that
“A localized crop failure like that of 1989 in Butare and Gikongoro brings in train forced emigation
abroad of several thousand, a large number of children dropping out of school , and deaths from starvation“
33
.
28
29
30
31
32
33
Republic of Rwanda, Compte-rendu de la Réunion tenue au Minagri en date du 02/05/1989 sur la situation
alimentaire du Rwanda en Avril 1989, Ministry of Agriculture, p. 4.
Republic of Rwanda, Compte-rendu, p. 4.
I refer to Ravallion, M., Markets and Famines, 1987 for an exploration of the price effects of information.
Pottier, ibidem, p. 27.
Verwimp, P., Development ideology and Genocide, Journal of Genocide Research, November 2000.
World Bank, Rwanda Agricultural Strategy Review , 1991, p. 1.
26
The direct cause of this crop failure was a severe drought in Southern Rwanda. A prolonged
period of drought had started in 1984 and reached its climax in 1989. Rwandan peasants have
developed capacities to deal with food crisises. The Gikongoro population however was already
very poor and vulnerable to adverse schocks. In 1989, they were at the limits of their resources.
The incapacity of the peasants to cope with the crop failure were due to the absence of credit, of
fertilizer, the anti-erosion practices, tax policy and the general neglect of the prefecture. Again,
these elements are closely linked to explicit policy choices. We discuss them one by one.
Although peasants contributed vasts amounts of savings to the well-known Banques Populaires,
very few loans were given to peasants. Half of the loans were given to traders and businessmen
whereas this group only made up 14% of the people with deposits at these banks. 34
Under Habyarimana, Rwanda was the country in the world that used the lowest amount of
fertlizer per capita.35 Again, this has to do with the import restrictions but also with the
distribution of the available fertilizer. The regime wanted to restrict its use for industrial crops
only (tea and coffee). The absence of fertilizer is cited by a peasant in the IWACU film as one
reason for the incapacity to cope with the crop failure.
The Habyarimana regime undertook a massive and nation-wide anti-erosion effort. The regime
obliged all peasants to dig ditches on their plots. Ditches were considered the best means of
combating erosion. Peasants who refused to do this were fined, just as in the case of neglect of
the coffee trees. Guichaoua (1991) who interviewed farmers in 1989, reports that farmers
resented the harsh labour requested in the nationwide ant-erosion policy of digging ditches. No
doubt erosion was one of the main problems of the country given that peasants have to cultivate
the slopes of hills. However, as Kangasniemi (1998, p.174) writes, Rwanda’s anti-erosion
campaign was rigid and simplistic, leaving no room for local physical conditions (rainfall, soil
type, slope) indigenous practices, and household specific economic circumstances (crops,
livestock).
A Belgian brother in the north of Rwanda, specializing in agriculture had developed an
alternative way of combating erosion. He taught the peasants to construct radical terraces. When
34
I refer to Uvin and to IFAD for this.
27
the author visited the brother in his home in Byumba in August 2000, a very large area was
covered with radical terraces. During the interview the brother related his story :
«From the beginning of his arrival in Rwanda (1972) I believed that intensive agriculture was thé
solution to the countries problem. Government policy under the Habyarimana regime however was build
on extensive agriculture. The government believed that production would increase if all available land
would be taken under cultivation. In order to prevent erosion on the hills, the regime obliged the
population to dig ditches, which was very unpopular among the peasants and which did not work. From
1977 onwards I practised radical terracing on 10 ha of land. Donors and engineers showed some
enthusiasm but the Ministry of Agriculture remained sceptic. In 1985 Habyarimana visited the project in
Byumba and was delighted by the results. However, he remained silent upon my requests to implement
my techniques on a larger scale. My fields were namely more productive then the fields without radical
terracing. In 1987, I wrote a very critical report on the functioning of the communal agricultural
officials ( they were doing nothing but commanding and bully the peasants). I believes that Habyarimana
saw the report because from 1988 onwards the president himself encouraged peasants to adopt radical
36
terracing in his speeches.«
The brother did not have to oblige the peasants to adopt radical terracing, they came to him to
learn the technique by themselves.37 The contrast with the mandatory erosion techniques decreed
by the regime could not be greater : upon the beginning of the democratisation process (19891990) and the decrease in power of the MRND, peasants neglected the maintenance of the antierosion ditches or destroyed the ditches.38 This also resembles the treatment of the coffee trees.
Peasants started to rip them out, even before 1991, although this was prohibited by law.
While the absence of credit and fertilizer, the mandatory coffee cultivation, the umuganda policy
and the anti-erosion measures where not confined to the prefecture of Gikongoro (indeed they
were nation-wide policies), their cumulated effect may exhaust a population that is on average
poorer than in other prefectures. While peasants all over Rwanda suffer from mismanagement
and bad policies, it are always the poor who are hit most. Under normal weather conditions, a
poor population may just be able to survive despite bad government policies. The occurance of a
drought may unveil their vulnerability to crop failure and famine.
In line with Habyarimana’s ideology of a hard working and moral peasant population (discussed
in the first part of the disseration) , Habyarimana tried to reduce the production of beer bananas
and the consumption of banana beer. It is worth noting that Habyarimana discouraged the
35
36
37
World Bank, Rwanda, Agricultural Strategy Review, 1991, p.
Interview in Byumba, September 6, 2000. This text is a summary of the part of the interview that discussed
agriculture.
This means that, according to the Belgian Brother, farmers in Byumba adopted radical terracing. Nduwayezu
(1990) in contrast, writes that no Rwandan farmer adopted radical terracing. For several reasons, the
brothers’ testimony is much more credible than Nduwayezu’s affirmation.
28
cultivation of bananas, even though bananas produce a very high yield per acre, and despite the
fact that a large domestic banana market existed in Rwanda.39 In a 1979 speech, he is very explicit
about that:
“Despite the opposition and the misundersta nding that I have seen in this question, I remain convinced that
the extension of the wine banana and, in certain regions, the appropriation of land for its cultivation, are a
great handicap for development and for the food equilibrium of the population. The “myth of the banana
tree” must disappear as well as the myth of the “nice corners” and that of the burning of bushes so-called
regenerating effect”40.
The regime considered beer drinking anti-social and beer brewing wasteful (Kangasniemi, 1998,
p.142). For the farmer household however, as we have seen in section two, beer brewing and
selling is very important for food security. Pottier (1992) writes that Habyarimana wanted to
reduce the cultivation of bananas because he believed occult rituals were being performed in the
banana plantations. A more economic and down-to-earth approach however, suggests that
bananas are a direct competitor to coffee, a cash crop that was strongly promoted by the regime.
Habyarimana’s main argument for high food prices was the protection of peasant revenue. While
that argument made sense from a producers' point of view, it is nonsense once you acknowledge
that the majority of the population were net-buyers (as opposed to net-sellers) of beans and
other staple food. Because the Habyarimana regime only looked at the peasantry as producers of
food and of cash crops, it never acknowledged that many peasants were net-buyers of beans (and
of food in general). Net-buyers depend on the price of food for their survival and it does not
need to be stressed that they benefit from low instead of high bean prices. The policy of import
restriction was continued in 1988 "to preserve the income of the peasant ".
In that respect, it needs to be said that an agricultural policy of food-selfsufficiency at the
national, regional, local and household level is, among other things, an anti-pastoral policy. As
Amartya Sen has argued before, pastoralists do not live of their cattle alone. They engage in
trade, they exchange cattle products for food crops. If you want every family to grow all the food
it needs for its own consumption at its own farm, a pastoral way of life is impossible. Pastoralists
namely do not cultivate. To repeat an expression by G.Kayibanda : « The Tutsi must also cultivate « .
38
39
40
Ntezilyayo, A., L'agriculture, une priorité dans la reconstruction nationale in A. Guichaoua, ibidem, p. 324.
Little and Horowitz, ibidem.
Habyarimana, J, Speech at the opening of the 1980 new year, 1979, p.135.
29
5. CHILD MALNUTRITION : AN ECONOMETRIC ANALYSIS
5.1 Data on child malnutrition in Rwanda
We use data from the 1992 Demographic and Health Survey. 1933 children between 0 and 60
months were surveyed three times between December 1991 and August 1992. The da ta include
antropometrics of child and mother, disease and vaccination information, community health
infrastructure and use of health care. 858 of these children came from households that were
included in a agricultural research project undertaken by Michigan State University (the rest of
the 1933 children came from households who were not part of the agricultural survey). The
demographic data are panel data (first round in 1991, second and third round in 1992), but the
collection of economic and agricultural data did not take place in 1992. For that resaon, we are
only using the 1991 cross-section with demographic, agricultural and economic data.
Table 7 : prefectural averages of antropometric indicators
Prefecures
Ruhengeri
Butare
Byumba
Cyangugu
Gikongoro
Gisenyi
Gitarama
Kibungo
Kibuye
Kigali
Rwanda
n
207
211
197
153
122
223
228
182
154
256
1933
n = 1933
haz
-2.18
-2.04
-2.21
-2.17
-2.40
-1.97
-2.14
-1.68
-1.77
-1.68
-2.00
whz
0.64
-0.49
-0.06
0.02
0.17
0.06
-0.05
-0.36
-0.02
-0.17
n
100
100
80
61
50
99
104
89
71
104
n = 858
haz
-2.24
-2.38
-2.40
-2.56
-2.62
-1.96
-2.00
-1.55
-1.62
-1.82
-0.037
858
-2.07
whz
0.73
-0.40
0.15
-0.05
0.14
0.03
-0.13
-0.40
0.04
-0.18
-0.01
haz : height for age z-score, whz : weight for height z-score
Table 8: Malnutrition, in percentage, n=1933
z-scores
above -2 (not malnourished)
-2>x>-3 (moderately malnourished)
below -3 (severely malnourished)
heigt for age
47.7
27.6
24.7
weigth for height
94.4
4.4
1.2
30
Height for age and weight for height are generally accepted as good indicators of the nutritional
status of children.41 The first is a measure of long-run nutritional status. Children with low height
compared to their age are stunted which is considered as an indication of chronic malnutrition.
Growth retardation generally occurs between 6 months and 2 years of age.
Though a stunted child may have some cath up growth, for the most part, a child whose growth has faltered
in the first two years of life will be on a different growth trajectory during the rest of his/her life.
(Alderman, 1993)
52% of the children in our sample are smaller then the ‘ standard ‘ child of the same age. It is
then no surprise that wasting (low weigth-over-age z-score) is less of a problem in Rwanda. More
then half of the children are indeed relatively small, meaning that their body has adapted to low
levels of food intake. Since the body is relatively small, it is easier to have a ‘ normal ‘ weight over
height observation. The z-score measures the deviation of each antropometric outcome from a
standard series of antropometric outcomes.
Figure 5: Height-for-age for 1933 children in Rwanda compared to Gikongoro
0.0
mean z-score
-.5
-1.0
-1.5
-2.0
-2.5
Rwanda
-3.0
0-5
Gikongoro
6 -11
12-23
24-35
36-47
48-60
AGEGROUP
Figure 5 suggests a negative correlation between age and the height-for-age z-score in Rwanda.
The older the young child gets, the worse its z-score. For Gikongoro, such a correlation seems not to
exist.
41
Thomas, D., Public Policy and anthropometric outcomes in the Côte-d’Ivoire, Journal of Public Economics,
1996, p. 156.
31
Another way to look at this negative relationship is the following. In Gikongoro Prefecture,
babies younger than 23 months show a very low z-score, meaning they face a severe degree of
chronic malnutrition. The price of beans in Gikongoro reached its highest point during the
month of June in 1989. We consider this month to be the high point of the famine in
Gikongoro. The first round of the nutrition survey was run in the month of December 1991,
meaning that 30 months separate the period of the survey from the high point of the famine. We
add to that the information that babies start to be exposed to food other that mother’s milk from
the age of six months onwards. From this age onwards they are no longer shielded from price
variations in the staple food. We split the 1933 children in the sample in two groups, the first
born before and the second born after the price shock of beans. We then investigate whether the
observed difference in z-scores for both groups of children between Gikongoro and the other
prefectures is significant or not. The group that experienced the price shock was at the time of
the survey 36 months or older. We use three categories of malnutrition, the first having a z-score
higher then -2, which contains children that are not malnourished. The second category is for
moderately malnourished children with z-scores between –2 and –3 and the last category is for
severely malnourished children with z-scores below –3. In table 9, we test whether the
distribution of children over these 3 categories differs significantly between Gikongoro and the
other prefectures.
Table 9: Distribution of malnutrition
Age
ha-z-score
Rwanda
< 36 m above -2
-2>x>-3
below -3
#
607
268
216
%
55.5
24.5
20.0
Gikongoro
#
%
26
38.8
18
26.7
23
34.5
> 36 m above -2
-2>x>-3
below -3
307
260
254
37.4
31.7
30.9
21
13
17
41.2
25.5
33.3
32
Table 9 (con.): Chi-Square tests for Chronic Malnutrition for children <60 months of age
Age
Value
df
< 36 months Pearson Chi-Square
Number of Valid Cases
11.043
1094
2
Asymp.Sign.
(2-sided)
0.004
> 36 months Pearson Chi-Square
Number of Valid Cases
0.964
821
2
0.617
The Chi-Square test in table 9 suggests that the distribution of the z-scores of children above the
age of 36 months does not differ between Gikongoro and the rest of Rwanda. For children
below the age of 36 months however, this difference is significant, Gikongoro has much more
children in category 3 (severely malnourished) than Rwanda as a whole. We did not find a
significant result for any of the other prefectures. This is new evidence for the observation that
the 1989 crop failure was not a single-peak famine, but was rather an episode in an enduring
period of hunger, a period that continued in the early nineties. For many peasant families in
Gikongoro, the 1989 crop failure was not the end point of a period of drought and hunger, but
rather the beginning of an even worse period. Given the anthropometric evidence, it appears that
older children have withstood the crop failure whereas children born after June 1989 in
Gikongoro suffered the most. The latter had mothers who consumed less food than needed
during the pregnancies of these children, resulting in lower weight at birth and started of their
lives under bad food conditions. From figures 1 and 2, we observe that food production in
Gikongoro after 1989 never reaches pre-1989 levels.
33
5.2. Explaining children’s nutritional status
Table 10a: Regressions of antropometric indicators for children
0 to 60 months of age, Height-for-age score is dependent variable, n=820
Child characteristics
age of child (in months)
age squared
sex
characteristics of mother
age
body mass index
read and write
married
Regression 1
OLS
Regression 2
2SLS
Regression 3
2SLS
Regression 4
OLS
-0.0800 ***
(0.000)
0.0008 ***
(0.000)
-0.0009
(0.993)
-0.0768 ***
(0.000)
0.0008 ***
(0.0001)
-0.0225
(0.8441)
-0.0749 ***
(0.0000)
0.00076 ***
(0.0003)
-0.0072
(0.9512)
-0.0719 ***
(0.000)
0.0007 ***
(0.000)
0.0043
(0.970)
-0.0011
(0.889)
0.0381 *
(0.056)
0.2240 *
(0.061)
0.547 ***
(0.002)
-0.0051
(0.5523)
0.04155 **
(0.0525)
0.1388
(0.3082)
0.45721 **
(0.0139)
-0.0057
(0.5180)
0.04375 **
(0.0444)
0.1137
(0.4027)
0.4452 **
(0.0190)
0.0017
(0.845)
0.03310
(0.121)
0.315 **
(0.012)
0.508 ***
(0.007)
-0.0018
(0.537)
-0.0538
(0.1541)
-0.0608 *
(0.0861)
-0.0108
(0.725)
IV
IV
0.5841 ***
(0.0033)
-0.00126
(0.3957)
0.6303 ***
(0.0000)
-0.00128
(0.3989)
Household characteristics
Household size
livestock (tlu)
farm size (acres)
log farm output (FRW)
distance to health center
regional dummy variables
crop failure affected region
civil war affected region
0.261 ***
(0.002)
-0.00136
(0.332)
-0.300 *
(0.094)
-0.297 *
(0.034)
IV
IV
IV
prices
beans
IV
bananas
IV
potatoes
IV
sweet potatoes
IV
Constant
Adjusted R²
N=820
-4.391 ***
(0.000)
0.147
-7.411 ***
(0.0002)
0.123
-7.831 ***
(0.0000)
0.128
-0.0068
(0.908)
0.00239 **
(0.017)
-0.0028 *
(0.079)
-0.305
(0.137)
-0.836 ***
(0.006)
0.0280
(0.281)
-0.0673 *
(0.060)
-0.132 **
(0.011)
-0.0543
(0.214)
0.789
(0.583)
0.140
34
significant at the 10% level ; ** significant at the 5% level ; *** significant at the 1% level
log farm output is instrumented for in regressions 2 and 3
Variables used in the regressions :
Age of the child in months and its square;
Sex 0 for male, 1 for female ; age of the mother in years ;
Body mass index of the mother is the weight in kilograms divided by the squared height in meters ;
The read/write and the marriage variables are dummies, with 1 for yes anf 0 for no ;
The size of the household is the number of household members ;
Tropical livestock units (tlu’s) 1 for cattle, .25 for pigs, .17 for sheep and goats, .10 for chicken and rabits ;
The farm size variable is the area of land cultivated by the household ;
The farm output variable in the regressions for Table10a is the log of the average value of agricultural production
over six seasons, from the 1989a sesaon to the 1991b season ; farm output was measured with constant 1987 prices ;
70% of all households with children<5y have reported farm output for 6 seasons. Households from the remaining
30% were included when at least 3 seasons were reported.
The regional dummies are 1 for the region affected by the crop failure (Gikongoro and parts of Butare) and for the
region affected by the civil war (Ruhengeri and Byumba).
The price variables are average, deflated prices over six seasons for beans, bananas and potatoes in Table10a and
average deflated prices in the child’s crucial growth period in Table10b.
Discussion of the results
In previous sections of this chapter, I have documented two facts :
(1) Gikongoro experienced a sharp decline in farm production in 1989, a decline from which the
households in that prefecture never fully recovered in the subsequent years. The data show
that farm production remained very low in the years 1990 and 1991. Rwanda on the whole
however did not experience a sharp decline of farm output in 1989.
(2) The health status of children in Rwanda in December 1991, measured by the height-for-age
antropometric indicator was such that ¼ of the children was moderately undernourished and
¼ was severly undernourished. For very young children (under 36 months of age in
December 1991) Gikongoro was doing significantly worse than the other prefectures.
We will first discuss the regression results presented in Table 10a. The first regression
(Regression 1) adds to the above picture : it shows that the health status of children can be
explained by a number of variables of which farm output is one. The regression shows that
children from households with a higher farm output (in value terms) have a better health status.
We will discuss econometric problems with this regression hereafter, but first look at the results.
In Regression 1, the shape of the height-for-age curve (in figure 4) is confirmed by the
significance of both the age variable and the age squared variable. These results on age are a
35
confirmation of findings in other studies ; this effect is among the best documented effects on
childrens nutritional status in developing countries. We do not observe a sex-effect in the heightfor-age-z-score.
As far as the characteristics of the mother are concerned, the mother’s age does not have an
effect on the child’s nutritional status. Her body mass (BMI), on the other hand does. As for
body mass, a well-fed and robust mother can invest more time and effort in the well-being of
her child. The BMI (BMI = weight in kg/(height in m)2) effect seems reasonable since most
activities in child care require a lot of physical effort, especially in developing countries. A strong
mother, ceteris paribus, has less difficulties in lifting her child, carrying her child, feeding her
child and performing other actvities in the household that benefit the child’s health indirectly. At
this point it is interesting to look at the results in a paper by A.Bhargava (1997) on the
determinants of adult well-being in Rwanda. Bhargava found that low incomes and high food
prices reduce the households’ energy intakes, thereby forcing the adults to spend additional time
resting and sleeping.42 He also found that mothers from households with a large number of
children spent more time working in agriculture. They namely allocate household tasks to
children. Bhargava concludes that, since poor nutritional status diminshes adult productivity,
larger families might be viewed favorably in Rwanda. Since we do not have time allocation data
in our data set, we cannot test for Bhargava’s findings directly. The variable on household size
proves insignificant in the regression and does not change the effects of any of the other
variables.
In order to illustrate the importance of mother’s BMI, we can look a little bit closer at the
antropometric characteristics of the mothers in our sample. Figure 6 schows the height and the
weigth of 520 Rwandan mothers. the average mother in Kibungo prefecture is 6 cm taller than
her counterpart in Gikongoro. Since mother’s height is not determined by the farm output or
consumption of the last couple of years, we are observing here a long term effect of (under)
nutrition. Figure 5 shows very well that Gikongoro is in general a poor area, not just in the late
eighties and early nineties, but across decades. These mothers were undernourished in the period
of their own youth. In terms of heigth, Gikongoro’s mothers score exceptionally poor. The
42
Bhargava, A., Nutritional Status and the allocation of time in Rwandese Households, Journal of
Econometrics, 77, 1997.
36
figure also shows that mothers in Ruhengeri and Gisenyi are doing very well compared to the
rest of Rwanda, especially in terms of weight.
Figure 6: Average heigth (in cm) and weight (in kg) of mothers per prefecture
158
59
Weight of mothers (kg)
Height of mothers (cm)
58
156
154
152
57
56
55
54
53
52
150
51
Ruhengeri
Byumba
Butare
Gikongoro
Cyangugu
Gitarama
Gisenyi
Kibuye
Kibungo
Ruhengeri
Kigali
prefecture
Byumba
Butare
Gikongoro
Cyangugu
Gitarama
Gisenyi
Kibuye
Kibungo
Kigali
prefecture
Having a married mother is better for the child’s long term health compared to having a nonmarried mother. Ideally, one would have data on intra-household distribution of food, to test for
this directly, but in the absence of this, we can only observe the effect of the marriage variable
on children’s nutritional status.
Another characteristic of the mother benefits the child’s long-term nutritional status, i.e. literacy.
A mother who can read and write is generally better informed about the practices of child care.
She can read the users application of medicin, she may have an informed knowledge of child
feeding practices, she may be more alert towards early signs of child illness, seek early assistance.
The literacy variable is significant in Regression 1.
Two regional dummies introduced in the model, namely one dummy for the region affected by
the crop failure and one dummy for the region affected by the civil war both prove significant.
The civil war in the north of Rwanda (Ruhengeri and Byumba prefectures) has a negative effect
on the health status of children living in that region. This is also true for children living in the
region of the crop failure (Gikongoro and parts of Butare prefecture). Since we already observed
37
the effect of farm output on child health (through the farm output variable), it means that in
both regions child health is influenced by a region-specific characteristic. In the north, civil war
negatively influences child health, whereas for Gikongoro, the region specific effect is also
negative. The latter effect, combined with the linear effect of farm output on child height-overage, suggests that the total effect of the crop failure in Gikongoro was non-linear. This supports
the thesis that the 1989 season was particularly bad for child health and thus supports our view
of a crop failure that developed into famine. Caution should be paid to these findings, however.
The latter effect may be suffering from multicollinearity, a problem we now turn too.
There is problem of endogeneity in Regression 1. This occurs when it is not possible to treat one
or more explanatory variables as given or exogenous. In such cases, it can be argued that some of
these explanatory variables are correlated with the equation’s error term. As a result the OLS
estimator is biased and inconsistent.43 The problem in fact is that a number of observable and
non-observable explanatory variables that are correlated with farm output and with health status
are missing in Regression 1. When one does not correct for this, one misinterpretes the effect of
farm output on health. The observed effect of farm output is not reflecting the cause of a good
or a bad health status, but merely the consequence of differences in observed and non-observed
assets and endownments (land, cattle, household size,…). The solution to this requires an
alternative estimation method. To derive a consistent estimator, it is necessary that our model is
identified. For this identification, we use so called instruments or instrumental variables. An
instrumental variable is a variable that can be assumed to be uncorrelated with the models error
term, but correlated with farm output. When we allow for an arbitrary number of instruments,
the estimator is sometimes referred to as the generalized instrumental variables estimator
(GIVE), also know as the two-stage least squares estimator (2SLS).
The 2SLS procedure allows one to obtain the same estimator in two steps, both of which can be
estimated by least squares. In the first step, the endogenous variable, farm output is regressed on
all instruments. In the second step, the original model (with height-for-age as dependent
variable) is estimated while replacing the endogenous variable (farm output) with its predicted
value from the first step.
38
This procedure is performed in Regressions 2 and 3. In Regression 2, farm output is predicted
with one instrumental variable (farm size). The results prove to be similar as those of Regression
1 with age, age squared, mother’s BMI, the marriage dummy and the predicted farm output
showing a statistically significant relation with the child’s health status. The literacy variable
looses its significance. Because the two-stage-least-squares estimation method was used, the
estimators in regression 2 are consistent.
We can use the estimated coefficients in Regression 2 to calculate the effect of farm output on
child health in Gikongoro, keeping all other variables on the country mean. This means we
calculate
Hˆ
= α + β 1 lnO R + β 2 X
A
and
Hˆ
= α + β1 ln OG + β 2 X
A
with height-for-age at the left hand side, O for farm output, R for Rwanda and G for
Gikongoro ;
since
Ĥ
= −2. 07; β 1 = 0 .5841; ln OR = 9. 9775 and ln OG = 9 .2452
A
This means that β 1 ln OG − β 1 ln OR = −0. 42
The result indicates that the average haz-score (child chronic malnutrition or stunting) for
Rwanda (–2.07) goes down by 20% (-0.42/-2.07) when all variables except farm output are set
on the country mean and farm output in Rwanda is replaced by mean output in Gikongoro. This
result thus gives the pure effect of farm output in Gikongoro (relative to Rwanda) on child
nutritional status.
In regression 3, the 2SLS method was used again, but this time 8 instrumental variables were
used to predict farm output. In the first step, farm output is predicted by area of land cultivated,
43
We refer to Verbeek, M., A Guide to Modern Econometrics, John Wiley and Sons, 2000, Chapter 5 for an
elaborate discussion of the endogeneity problem and the econometric ways to deal with it.
39
livestock, two regional dummies and four prices. In the second step, the predicted value of farm
output from the first step is used as regressor to estimate height-for-age. Again, the results of
the estimation are similar to regressions 1 and 2, but the significance of the predicted value of
farm output has increased a lot compared to regression 2, thereby confirming the validity of our
results. The negative effect of household size, observed in regression 2 is now significant (at the
10% level).
An additional regression was performed using all instrumental variables as explanatory variables
in an OLS-regression thereby omitting farm output as a regressor. This is regression 4. In this
regression, mother’s BMI looses its significance. The literacy variable and the distance to the
health centre become significant, effects we did not observe in regressions 2 and 3. This means
that there is some kind of multicollinearity going on between farm output, literacy and distance
to a health centre. Farm output is picking up some of the effects of literacy and distance. Since I
have tried to deal with multicollinearity in the course of the four regressions discussed in this
section, we should not overestimate the problem caused by these effects. The civil war dummy is
now very significant and the significance of the other regional dummy has decreased. New is the
effect of prices. Increased prices of bananas and potatoes have a significantly negative effect on
the child’s nutritional status. Both these crops are essential in the Rwandan household economy
and the Rwandan diet.
The livestock variable, measured as the 3-year average holding of livestock (in tropical units),
does not have a significant effect on the nutritional status of children. This result is somewhat
surprising as we have seen that in Rwanda, households saw their livestock holdings decrease in
the 1989-1991 period. This decrease in livestock is either the result of sales or of direct
consumption of livestock.44 Sales can be used to purchase food. This means that household
consumption, in the short run, may have benefitted from a decrease in livestock holdings.
44
Neglecting loss or death of livestock.
40
Table 10b: Regressions of antropometric indicators for children
6 to 60 months of age, Height-for-age score is dependent variable , n=627
Child characteristics
age of child
age squared
sexe
characteristics of mother
age
body mass index
read and write
married
Household characteristics
Household size
Regression 5
OLS
Regression 6
2SLS
Regression 7
2SLS
-0.0422 **
(0.034)
0.0003
(0.249)
-0.0695
(0.577)
-0.0534 **
(0.0128)
0.0004
(0.1346)
0.0986
(0.4493)
-0.0497 **
(0.0216)
0.0004
(0.1818)
-0.0803
(0.5425)
-0.0039
(0.133)
0.0003
(0.271)
0.102
(0.439)
0.00126
(0.893)
0.0326
(0.143)
0.198
(0.142)
0.471 **
(0.016)
-0.0017
(0.8586)
0.03723
(0.1184)
0.1439
(0.3386)
0.40513 **
(0.0470)
-0.00139
(0.8878)
0.03363
(0.1574)
0.1873
(0.2097)
0.38466 ***
(0.0609)
-0.0045
(0.648)
0.0313
(0.189)
0.319 **
(0.025)
0.430 **
(0.036)
-0.01916
(0.568)
-0.0534
(0.2129)
-0.04808
(0.2303)
IV
IV
IV
-0.0153
(0.673)
-0.0837
(0.210)
0.0025 **
(0.027)
0.4451 **
(0.0346)
-0.0018
(0.2565)
0.37148 ***
(0.0263)
-0.00159
(0.3371)
livestock (cruc) (tlu)
farm size (acres)
log farm output (cruc) (FRW)
distance to health center
regional dummy variables
crop failure affected region
civil war affected region
prices
beans (cruc)
0.199 **
(0.023)
-0.00187
(0.235)
-0.134
(0.498)
-0.327 **
(0.041)
IV
IV
IV
bananas (cruc)
IV
potatoes (cruc)
IV
sweet potatoes
IV
Constant
Adjusted R²
N=627
-4.272 ***
(0.000)
0.054
-6.333 ***
(0.0019)
0.047
-5.6052 ***
(0.0009)
0.047
Regression 8
OLS
-0.0024
(0.169)
-0.125
(0.570)
-0.511 **
(0.049)
0.019
(0.341)
-0.044
(0.207)
-0.076 *
(0.070)
0.012
(0.737)
-1.689 *
(0.064)
0.052
41
significant at the 10% level ; ** significant at the 5% level ; *** significant at the 1% level
log farm output is instrumented for in regressions 2 and 3
Variables used in the regressions :
Age of the child in months and its square;
Sexe 0 for male, 1 for female ; age of the mother in years ;
Body mass index of the mother is the weight in kilograms divided by the squared height in meters ;
The read/write and the marriage variables are dummies, with 1 for yes anf 0 for no ;
The size of the household is the number of household members ;
Tropical livestock units (tlu’s) 1 for cattle, .25 for pigs, .17 for sheep and goats, .10 for chicken and rabits ;
The farm size variable is the area of land cultivated by the household ;
In the regressions for Table 10b, the farm output variable is the average value of agricultural production in the
seasons that are crucial in the child’s growth path, meaning between the 6th and the 23rd month. If, for example, a
child was 40 months old at the time of the survey (December 1991), he are she was born in september 1988. This
means that the period March 1989 through September 1990 was the crucial period, which corresponds with the
agricultural seasons 1989b, 1990a and 1990b. The same reasoning applies to the number of livestock (in tlu’s) and
the price variables for the regressions in Table 10b.
The regional dummies are 1 for the region affected by the crop failure (Gikongoro and parts of Butare) and for the
region affected by the civil war (Ruhengeri and Byumba).
The price variables are avergage deflated prices in the child’s crucial growth period in table10b.
In table 10b, we have tried to dig a little bit more in the farm output data. From the literature,
learn that from a nutritional point of view, the child goes through its most crucial months
between the 6 th and the 23 rd month after birth. We then asked whether we could find a
significant effect of farm output on the child’s health status during this period. Ideally we should
put the output variable in this crucial period in a regression with the average output over the six
seasons (as was used in table 10a regressions). The multicollinearity problems however were
evident, and instrumenting did not help, since we could not find good instruments for farm
output in the child’s crucial period. That is why we used only farm output in the crucial period in
the regressions reported in table 10b. Since children below the age of 6 months are protected
from the nutritional environment by their mothers (they only get brestfeeding), these children
had to be left out from the regressions in table 10b. The results of the regressions 5 to 8 are
roughly the same as those of regressions 1 to 4, this being not surprising, since we used the same
estimation methods and only slightly different variables. Prices were taken during the child’s
crucial nutritional period, just as the number of livestock. We observe that farm output - in the
crucial period - has a very significant effect on the health status of the child. A number of other
variables (mother’s BMI for example) are not significant anymore at the 10% level. Observe also
the significance of the regional dummy in the region with civil war (norhtern Rwanda). This
42
variable, as in table 10a seems to be picking up effects that are not picked up by the farm output
variable. The crop failure region dummy does not show an extra effect for that region, probably
due to the low number of observations of children from Gikongoro between 6 and 23 months
of age during the crop failure. As for the prices of staple foods in the crucial period is concerned,
only the price of potatoes has a negative and significant effect on the child’s health status.
6. ON FAMINE AND GENOCIDE
6.1. Political economy background
According to the peasants interviewed in a documentary film (details in appendix 2), the 1989
famine was worse than others because one could not even leave one’s children with the
neighbours. During earlier famines, peasant families showed more solidarity towards each other’s
misery. « Rich families would take in children of poor families « a woman said. The film also mentions 4
cases in which peasants killed thieves in 1989 without intervention of the judicial authorities.
Taking policing in their own hands was a precursor of much worse violence 5 years later. The
reasons for the breakdown of solidarity at the end of the eighties may be found, we believe, in
the duration of the agricultural crisis. If one family has a food problem during a few weeks in a
specific year, it is not difficult for the neigbours to help this family, but when an entire region
experiences a crisis for several years in a row, it is not surprising that after a few years people
have no surplus left to share with others. And the rich lived in Kigali.
Moreover, the mentioned documentary (see appendix 2) also shows how peasants complained
that fertile land had been occupied by "the rich". The peasants say that the communal authorities
gave fertile land that previously belonged to the peasants, to a group of wealthy people in order
to establish a large scale tea-plantation.45 Peasants were forcibly removed from their land without
compensation. The wages earned on the tea plantation only covered part of the nutritional needs
of peasants families. Removal of peasants from their land with little or no compensation also
occurred in other places in Rwanda, e.g. in Gisovu commune, Kibuye prefecture, also for a tea45
IWACU, ibidem.
43
plantation and again in Gikongoro for a large scale "development project". The latter project
earmarked land for wealthy people from Kigali as well as for some foreigners.
Interestingly, Pottier argues that the elite in the Habyarimana regime, consisting of Hutu from
the North of Rwanda, value land in a special way. When they took power, in 1973, Pottier writes,
their ambition was to restore their own pre-Tutsi culture – a culture dominated by powerful
landowners (abakonde) who attracted clients (abagererwa) through land.46 The institution of
ubukonde, restricts access to land and creates political and economic ties between patrons and
clients. Ubukonde, Pottier argues, is important for understanding food and development policy,
because the young Hutu state intervened in 1961 to authenticate and legitimate its existence.
6.2. From crop failure to famine : the non-response of the Habyarimana regime
The crop failure that struck several communes in Butare and Gikongoro prefectures in 1989 was
the result of a five year period of drought (starting in 1984) from which the region never really
recovered. The drought was so severe in 1989 that the roots of crops turned into ash. Crop
failure however does not have to lead to famine. Indeed, there are numerous cases in which crop
failures do not cause hunger. Several circumstances can avert famine. If a population has other
sources of income outside farming, famine can be averted by importing food through usual
market mechanisms. The population affected by crop failure will buy the necessary food items
on the market. Policy measures averting famine are well-kwown. The government can
temporarily employ people in public works (India), the government can hand out food aid to
vulnerable people, the government can allow aid organisations to hand out food, to name just a
few policy measures. The literature on famines and real life cases also advise governments to
stop taxing poor people struck by crop failure. Governments are advised to stop levying taxes
during famines, requering mandatory labour dues, and school fees. These measures will not avert
46
Pottier, p. 29.
44
famine, but will help to ease the burden of a crop failure. None of the aformentioned measures
were taken by the Habyarimana government in 1989.47
It is this non-response to famine that should be investigated. Devereux (1993) writes that nonresponse to famine (not giving food aid nor taking other measures to avoid famine) is to allow a
‘Malthusian final solution‘ to happen, namely to equilibrate population with resources. In the
introductory chapter, I demonstrated some Malthusian influences in the thinking of Rwanda’s
governing elites. Here however, I first address the motivation behind the non-response. For as
we have seen, man-made famine can be intentional or non-intentional. Whereas the direct cause
of the crop failure in Southern Rwanda was a long lasting drought, the relationship between crop
failure and famine is far more complex. The underlying reasons for a crop failure to develop into
a famine are identified by a political economy analysis (see sections 3.2, 4.1, 4.2,4.3 ,6.1). When
actual starvation occurs on a more or less grand scale, immediate measures can be taken to relief
the starving population. I will now discuss the reasons for non-intervention.
The Rwandan government did not react to the crop failure, it did not provide food or other aid
and thus allowed a crop failure to develop into famine. The height of the famine occurred in
June 1989, when the price of beans rose to 60 RWF per kilo. The regime did not mention the
famine, no government agency or ministry studied it or reacted to it. Only in March 1990, after
the publication of a documentary film about the famine, the government started providing food
aid. Although famine conditions became evident from March 1989 onwards, government
disbursed aid only one year later. The makers of the documentary film we mentioned earlier,
IWACU, a non-governmental cooperative organisation, made the film explicitly to prove that
famine existed in the country.
48
The government was indeed denying that famine existed. The
refusal to accept that famine struck Southern Rwanda had everything to do with the ruling
principle of food self-sufficiency. Acknowledging that famine existed would have been the same
as saying that Rwanda could not feed itself, the latter being the cornerstone of Habyarimana’s
regime. Famine was the ultimate proof that a policy of food-self-sufficiency at the household,
communal and national level did not protect households and regions from famine. The
47
The 1990 report of the Caritas Social Bureau states that (p.26) it would be useful to suppress several
mandatory taxes such as the personal tax, the tax on cattle, diverse contributions to the MRND, to the
communal projects,…and voluntary contributions to the Red Cross, the Church and so on. The fact that this
NGO is giving this advice, is a serious indication that these taxes continued to be levied during the period of
the famine.
45
government thus refused to acknowledge that its policy had failed.49 In order to prevent the
outside world from getting information about the famine, the government tried to prevent
journalists to enter the famine region and to write about it. A Belgian journalist was told by
security chief Augustin Nduwayezu not to write articles which irritated the highest authorities. 50
The government did not accept aid from foreign governments or international organisations. 51
This strategy may be wise if one wants to avoid disruption of domestic production, but then one
should provide relief from domestic sources, meaning organising food transport from food
surplus to food deficit regions. As we have shown, by the time of the famine in Southern
Rwanda, Northern and Eastern Rwanda were not experiencing a decline in food production.
Nevertheless, no transport of food was organised.
Besides not wanting to admit policy failures in the face of the stated self-sufficiency objectives,
the non-response of the Habyarimana government also had other reasons. A second reason for
the regime to passively allow the population to starve was that the people living in the famine
struck region had never been in favour of Habyarimana. The president was considered as a man
from the North who had shown no interest in the improvement of the living conditions of the
population in Southern Rwanda including Butare and Gikongoro. Moreover, in the presidential
elections held in December 1988, in which officially 99% of the population voted in favor of the
sole candidate, Habyarimana , the population of Gikongoro had voted en masse against the
president.52 According to one of our interviewees, a long term and well-informed resident of
Gikongoro, this result awoke the anger of the president. Nsengiyaremye (1992) writes that the
elections were falsified : in Gikongoro, the préfet did not even count the votes, but declared that
Habyarimana got 100% of the vote, which was the target set by Habyarimana's supporters. 53
48
49
50
51
52
53
Iwacu, personal conversation, Kigali, August 12, 2000.
During one interview, a respondent told the author that the Rwandan bureaucrats under Habyarimana were
used to tell their superiors only the things that pleased them, hiding real problems. While this attitude may
have retarded the reaction of the regime, one cannot conclude from this that the president was uninformed
about the famine.
Prunier, G., The Rwanda Crisis, History of a Genocide, 1995, p. 87-88.
Iwacu, personal conversation, Kigali, August 12, 2000.
Interview, Kigali, August 13, 2000. Unofficial results say that 78% to 90% of the population voted against
Habyarimana in December 1988. Another informant told us that the Gikongoro population already voted
against the president in the 1978 election.
Nsengiyaremye, D., La Transition Démocratique au Rwanda (1989-1993) in Guichaoua, A., Les Crises
politiques au Burundi et Rwanda (1993-1994), Université de Lille, 1995, p. 242.
46
In addition, it is known that during his regime people from the south (Hutu as well as Tutsi) did
not get promoted in the administration. All senior posts went to people from Habyarimana’s
own region. Fr. Nzamurambaho for example, a popular politician from Gikongoro and a former
minister of Agriculture wanted to present himself in the November 1988 legislative elections. He
was removed from the list of candidates without any explanation. The president, everybody
knew, only wanted people who depended on him and who were completely loyal to him. 54
Thirdly, the famine-struck prefectures, Gikongoro and parts of Butare, and especially the part of
both these prefectures that was used to be called Nyaruguru, had a large Tutsi population.
Several sources indicate that, by 1989 or even earlier, president Habyarimana was informed of
the probability of a military invasion of Rwanda by the FPR. This can be documented as follows
: in May 1989, at the summet of Nyagatare, the president of Uganda, Museveni, told
Habyarimana that he had better solve the refugee problem if he wanted to avoid a military attack.
55
All over Uganda people were talking about it and one can assume that Habyarimana’s secret
service also monitored the build up of the rebel movement. The year before, in 1988, Tutsi in the
diaspora had held a major conference in Washington DC. One of the conclusions of this
conference was the desire and the right to return. As early as 1987, Tutsi leaders in the diaspora
were fundraising for an invasion.56 The embassies of Rwanda in Kampala, Nairobi, Addis-Abeba
and Washington had warned the Rwandan government of the preparations for an invasion.57 In
1989, at the high point of the famine in Southern Rwanda, a fraction of the FPR made an
abortive attack in north Rwanda. Ogetunnu writes that Habyarimana's intelligence service had
several officers from the FPR on its payroll in order to sow division inside the FPR. This means
that even before the famine was under way, Habyarimana knew of the build up of FPR forces in
Uganda.
Therefore, the question is whether information on an upcoming civil war influenced the decision
not to intervene in Gikongoro in 1989. Had Habyarimana already made up his mind in 1989, to
consider the Tutsi inside Rwanda as accomplices of the FPR and thus as enemies of his regime ?
In the event of an FPR attack, Habyarimana may have anticipated that young Tutsi from Rwanda
54
55
56
57
Nsengiyaremye, D., ibidem, p. 242
Nsengiyaremye, D., ibidem p.247 and interview, Kigali, August 14, 2000
Adelman, H. and Suhrke, A., The Path of a Genocide: the Rwanda Crisis from Uganda to Zaire, Transaction
Publishers, 1999
Nsengiyaremye, D., ibidem, p. 247
47
would join the FPR forces (which eventually happened). The author interviewed several
survivors from the genocide in Gikongoro who confirm that a number of young men from
Gikongoro joined the FPR before the 1990 attack.58 Moreover, the regime’s leadership
considered Gikongoro prefecture quasi-lost to the FPR as it would be very easy to penetrate the
prefecture through Nyungwe forest from Burundian territory. There is an extra reason related to
this war strategy. If Habyarimana anticipated an upcoming civil war, he would preserve the
country’s food stocks (however small or large they may have been) for his own soldiers. Indeed,
why feed a population in the south that did not favor him in the first place and of which a
number were potential FPR-soldiers when at the same time food would be needed for the Hutu
army.
6.3. The 1989 famine as a prelude to the 1994 genocide
When one combines the different elements discussed in the previous sections, namely the
peasant and self-sufficiency ideology, the populist approach to population growth, and the
reaction to the crop failure , then the following picture emerges.
Severe drought causing crop failure in southern Rwanda in 1989, by itself did not lead to famine.
Famine only occurs when governments allow it to take place. It is my thesis in this chapter that
the Habyarimana government knowingly allowed a crop failure to develop into famine in
southern Rwanda in 1989.
We have seen that the peasantry can only escape the malthusian trap of high population growth
and declining agricultural productivity when governments promote technological innovations.
The Habyarimana government has never done so. The regime’s advice to the peasants was to
apply more labor to the already degraded soils, as if the Rwandan farmers where not numerous
enough to work the land.
58
Interview, Kigali, October 12, 2000.
48
As a result of the mismanagment of agriculture, peasants were oriented towards subsistence
production, a situation in which they were highly vulnerable to crop failure and direct
entitlement loss. We have seen that a policy of food-selfsufficiency is anti-pastoral. Pastoralists
do not cultivate crops to secure their living. And, empirical research has shown that cultivators
do not want to grow only food to secure their food-security. Apart from tea and coffee, the
government nevertheless relied on a policy of cultivation of all available land to achieve higher
food production. Once all the land was put into cultivation, peasants were unable to keep food
production and population growth in balance because no attention was paid to technological
innovation, fertilizer use, credit and market development. The forcefully promoted and
monitored cultivation of coffee was not beneficial anymore to the peasants from the end of the
eighties onwards ; they preferred to grow cash crops for the domestic market such as bananas
for the production of beer, a choice the Habyrimana regime opposed.
In contrast to the peasant-friendly rhetoric, we have seen that the agricultural policies benefitted
the commercial and political elite. Very few funds – expressed as a fraction of government
budget - were invested in agriculture. Altough 50% of the deposits in the Banques Populaires
came from peasants, they were only receving 10% of the loans. We have shown that the
Habyarimana regime never cared to set up a functioning food marketing system. The access to
food was considered a local and personal problem, to be settled by patron-client relationships.
This attitude, together with the officially sanctionned slogan of food-self-suffiency, led the
regime to deny and neglect the famine in southern Rwanda in 1989. As K.Jonassohn (1999) has
recently argued that
“In the present stage of the world’s development, the presence of serious starvation in a group is one of the
first sure signs that this group has been singled out for victimization “
The refusal to assist a starving population does not only hurt young men and potential soldiers,
but also women and children. A strategy of preventing food aid may then be genocidal. To be
sure, for lack of documents, I cannot prove this strategy or the genocidal intent of the refusal to
provide food aid. Some evidence, advanced by Kinyamateka, indeed points at self-enrichment and
embezzlement of government funds destined to buy food. Is this explanation incompatible with
a political, strategic or genocidal motivation ? I do not know. In any case, the story of the crop
failure in southern Rwanda gives ample evidence that the Habyarimana regime knowingly
allowed famine to occur. According to my research, Habyarimana’s agricultural policy, in
49
particular its ideological underpinnings, its political motivations and its economic consequences
for the Rwandan peasantry are an important part of a process that eventually led to genocide.
When, after a revealing documentary produced by IWACU in February 1990, president
Habyarimana visited the famine-prone region in March 1990, he told the press that his
administration had hidden this event for him, that he was not aware that famine was occuring in
Rwanda. Granting that the president’s aides did not tell him everything, we believe that he lied at
this point. From our interviews in Kigali with close collaborators of the former president, we
know that he received detailed intelligence reports twice a day about everything that happened in
the country. It is highly unlikely that the highest authority was not aware of a famine that struck
large areas in two prefectures and caused the emigration of 10.000 citizens to Burundi and
Tanzania.59
59
Discussing the credibility of Habyarimana, one should not forget that only 7 months after his denial of famine
conditions, in October 1990, the same president, Habyarimana, organised massacres in Mutura and Kibilira
(documented in a report of FIDH, an organisation of international human rights groups, 1993). The targets of these
local massacres were the local Tutsi, known for their pastoral way of living. We should also keep in mind that the
same president later on organised a fake attack on Kigali in order to draw in more French troops and that the same
president imprisoned 10.000 Kigali citizens, mainly Tutsi, on ch arges of conspiracy with the FPR. It is the same
president, Habyarimana, who denied to the press and to diplomats that massacres occurred in Rwanda in 1990, 1991
and 1992 in which about 2.000 Tutsi were killed.
50
APPENDIX 1 Figure A1 : Famine struck region in Butare en Gikongoro Prefectures
Paved road
Prefectural border
Communal border
Prefectural capital
50 km
M
M
R
K
K
K
M
G
N
R
B
R
M
N
K
Most struck communes are indicated with their first letter: in Gikongoro: Mubuga,
Rwamiko, Nyamagabe, Kinyamakara, Mudasomwa, Karama, Karambo, Rukondo,
Musango, Muko; in Butare : Nyakizu, Runyinya, Kigembe
51
APPENDIX 2
Published sources documenting famine in Southern Rwanda
One of the most interesting examples of the documentation of the food crisis in southern Rwanda is a film. In
February 1990, a group of researchers working with IWACU, a cooperative organisation working in the rural areas,
produced a film, “Haguma Amagara” which means “you only live once”. The film was a response of this grass-roots
organisation to the silence of the Habyarimana regime. Since the regime refused to acknowledge the famine, which
in February 1990 was already going on for more then eight months, IWACU decided to film the evidence. The film
shows poor peasants in southern Rwanda having nothing to eat, showing markets without products, peasants
deconstructing their houses to sell the parts for food, peasants migrating to other parts of Rwanda and to Burundi,
hungry children and mothers, adult males who are telling the film-maker that they are to weak to work. The film
indeed documents hunger in southern Rwanda in the 1989-1990 period. We come back on interviews with peasants
in this documentary when we discuss the causes of the famine.
Apart from the film, several Rwandan organisations have documented the food crisis and have published reports on
it. The author was able to trace three valuable reports. These reports were published over a period of more than two
and a half years. The first was written by the Social Bureau of Caritas in Kigali.60 The report cites administrative and
communal sources documenting hunger and starvation in several communes of Butare (Nyakizu, Runyinya) and
Gikongoro (Nyamagabe, Karama). It counts the number of death people due to starvation by commune : 5 in
Karama, 30 in Nyakizu and 107 in Nyamagabe. On top, 2.316 people had left Nyakizu to find food or work
elsewhere. (p.3) In Runyinya commune, about one hundred children had left school and 500 in Karama. Many
husbands had temporarily migrated. Crops were stolen at night and several cases of suicide had been reported.(p.11)
Peasants also report the absence of solidarity among neigbours. (p.19) In Gikongoro, peasants hoped that the next
coffee harvest would be good, if not they would starve to death (p.16-17). The Caritas report concluded that there
was nothing to eat anymore in the south, peasants were already eating the leaves and the roots of their plants.
Peasants had sold their belongings at low prices to secure some cash.
The second report was written by staff of the grass-roots NGO called CCOAIB, but the author could not get hold
of this report .61 The third was a report on a local agricultural survey of 300 households in several communes by the
Agricultural Development Project in Gikongoro. This project was run by the Ministry of Agriculture. 62 This report
stated that 25% of the surveyed households were indigent and that female headed households suffered more. Two
of the reports used the word ‘famine‘ to describe the food crisis in Southern Rwanda and the third spoke of
starvation.
60
61
62
Les Retombées de la Famine dans les Préfectures de Butare et de Gikongoro , Bureau Social Urbain-Caritas,
Kigali, Février 1990, 26 p.
Twizeyimana, P., and Uwimana, V., Portrait de la Pénurie alimentaire actuelle au Rwanda. Dévoilement
d‘une famine cachéé sous la verdure, CCOAIB, Kigali, Novembre 1989, 43 p. Several attempts to get a copy
of this report in Rwanda failed.
Gascon, J.-F., Pauvreté a Gikongoro, Résultats de l’enquête réalisée aupres des ménages indigents, Projet de
Development Agricole de Gikongoro, Document de travail, n.156, Juin 1992, 62 p.
52
Kinyamateka, the most important independent newspaper in Rwanda, also published information on the food crisis
in 1989. The newspaper’s articles infact made the famine known to the Rwandan public. André Sibomana, editor of
the newspaper in 1989, expressed it as follows :
« In 1989, a terrible famine struck the south of the country. There was a natural explanation for this phenomenon, but the
authorities did nothing to improve the situation. Worse still, I had evidence that part of the government’s assistance which was
intended for the population at risk had been diverted. It was a scandal. I decided to publish this information. We were
threatened and we were called liars, until I published photographs which were overwhelming. This had an immediate effect.
Readers wrote in to express their satisfaction : at last the truth was being told. »63
APPENDIX 3 : Figure A3 : Value of Production in 10 seasons
87a
87b
88a
88b
89a
Ruhengeri
Butare
Buymba
Cyangugu
Gikongoro
Gisenyi
Gitarama
Kibungo
Kibuye
Kigali
prod
26775
31285
29323
24833
19796
23710
33396
35617
25803
35419
%p
40
35.6
40
45.8
62.7
59.1
25.4
15.7
55.4
26.8
prod
23020
30707
35121
25240
19137
20137
38733
39722
15584
47539
%p
43.2
37.9
28.6
36.9
64.3
71.2
21.4
14.5
78.6
22.9
prod
29377
27111
28142
25447
18744
26934
33271
34305
29583
38964
%p
32.7
39.4
37.8
36.1
65.5
53.6
25.9
19.0
48.2
23.0
prod
21164
30168
30474
24550
19043
17799
34245
39361
14906
44777
%p
50.9
41.2
36.0
39.8
59.5
73.6
27.5
11.9
75.6
30.9
prod
27770
24627
26012
21699
13347
22771
33639
38661
20064
40727
%p
44.5
46.3
50.5
60.0
81.3
60.2
25.8
24.2
63.5
21.7
Rwanda
29171
39.3
30702
40.1
29684
37.1
28693
43.5
27589
45.8
N
1073
1089
1076
1075
1211
Production for these 10 seasons is the value of total production of 8 crops : bananas, potatoes, sweet potatoes,
beans, mais, peas, sorghum and manoic.
%p is the percentage of households that has a value of production lower then 2/3 of the mean value of production
in Rwanda in 1987a. This mean value was 19.447 RWF.
Ruhengeri
Butare
Buymba
Cyangugu
Gikongoro
Gisenyi
Gitarama
Kibungo
Kibuye
Kigali
prod
22030
27612
31073
22910
9853
14036
38299
52659
11123
52922
89b
%p
66.9
48.7
42.9
57.3
86.2
80.0
25.9
22.1
82.3
18.2
prod
26636
22760
31003
19492
10014
22855
32070
44268
17866
41386
%p
57.0
55.5
44.9
63.2
86.2
64.1
29.3
27.1
64.1
24.7
prod
24089
24484
33185
17604
12699
15839
32879
39231
11846
37629
%p
56.3
55.2
42.2
69.5
81.9
73.4
29.1
22.1
89.4
28.8
prod
28499
16399
19571
13848
17761
25845
34216
16855
30635
%p
48.8
66.9
71.4
83.5
69.8
45.1
34.1
73.3
38.5
91b
prod
%p
21937 59.7
20697 59.6
17299 68.9
12102 80.7
14215 76.9
31268 40.3
32380 32.3
11624 85.9
37186 30.7
Rwanda
29247
50.9
27691
49.5
25935
52.4
23072
57.1
23687
N
63
1202
90a
1230
90b
1230
91a
1038
56.3
1022
Sibomana, A., Hope for Rwanda, Conversations with Laure Guilbert and Hervé Deguine, Pluto Press,
English-language edition, 1999, p.22.
53
APPENDIX 4 : Insights from the model of the peasant household
The model of the peasant household provides a good framework to discuss the effects of changes in food prices.
This is important for the topic of this paper, as we have seen that the Habyarimana government believed that high
food prices offer an incentive for producers to produce more food. With the help of our household model, we can
judge the effects of such a policy of high food prices on the production and consumption decisions of peasants.
One possible application of the household model is found in the human resource development literature.
Households obtain consumption not only from marketed goods, but also from goods which are produced at home
using household labor. As Udry and Bardhan (1998) state, one’s utility may depend on a vector of consumption
goods c and on health, which depends on c and on time spent at home «producing» health. 64 The household
problem, then, can be written as
Max u (c a , c m , H )
w.r.t
c a , cm , H ≥ 0
subject to
p mc m + p x x = pa ( qa − ca ) + w( Lc + LF + LO − L)
L = Lc + LF + LO
H = H(c, Lc )
where Lc is household labor devoted to producing health. The separation property is maintained with respect to
production on the farm, but the production of health depends on preferences. The effect of prices on consumption
can be written as
∂c a ∂c a
∂c ∂y *
=
y*+ a
∂pa ∂pa
∂y * ∂pa
(1)
(1.1)
(2)
(1) is the direct effect on consumption for given income
(2) is the profit effect
Since we are especially interested in the health effects of (in)sufficient consumption, we are looking at
∂H ∂H ∂ca ∂H ∂Lc
=
+
∂p a ∂ca ∂p a ∂Lc ∂p a
(1.2)
54
Working out (1.1) yields
∂ca ∂c a
∂c
=
U * + (q a − ca ) a
∂p a ∂pa
∂y *
(3)
(4)
(1.3)
(5)
(3) is a pure substitution effect and is negative
(4) is the marketed surplus (+) or decifit (-)
(5) is the income effect and is positive (assuming ca is a normal good)
Furthermore,
∂Lc ∂Lc
∂Lc
=
U * + (q a − c a )
∂p a ∂pa
∂y *
(1.4)
And, substituting (1.3) and (1.4) in (1.2), yields
∂H ∂H
=
∂p a ∂ca
c
∂ca
∂c a ∂H ∂Lc
∂L
U * + (q a − ca )
U * + (q a − ca )
+
∂y * ∂Lc ∂pa
∂y *
∂ pa
A health demand function can be directly derived from the household model. The demand for health, as expressed
in several antropometric measures such as height for age, weight for height and weigth for age depends only on the
price of basic food crops, child characteristics and community factors.
Assuming that pa is the price of beans, the derivation makes clear that peasants who do not produce beans (e.g.
landless peasants) will see their consumption of beans decline as a result of the price increase. The effect on bean
producers is less clear, it can be either positive or negative, depending on the effect of the price increase on the
marketed surplus(qa – ca) :
ϑ (q a − c a ) ϑqa ϑc a
=
−
ϑp a
ϑp a ϑp a
(6)
64
(7)
− (q a − c a )
U*
(8)
ϑc a
ϑy *
(1.5)
(9)
Udry, C and Bardhan, Development microeconomics.
55
(6) is a production incentive and is positive
(7) is a pure substitution effect and negative (postive when we consider the –sign)
(8) is the marketed surplus
(9) is the income effect and is positive (negative when we consider the –sign)
the effect of p a on the marketed surplus depends thus on the sign and the size of (8) and (9) versus (6) and (7). If the
income effect is very large and stronger than the combined effects on production and substitution, then the overall
effect on the marketed surplus is negative.
In a land scarce environment, such as Rwanda, peasant’s capacity to grow their own food decreases and they need to
purchase food on the market. A price increase of beans lowers bean consumption of landless peasants and peasants
who do not produce enough beans for their own consumption.65 For landed and bean producing peasants, the
effect depends on the income elasticity of bean consumption. If this is very large, peasants may produce more
beans, not for the market but for their own consumption.
The model of the peasant household as outlined above shows that price policies to « preserve the income of the
peasant » or « to increase bean production » may effect (parts of) the peasantry in the opposite way. The policy of
high bean prices only benefits wealthy, landed elites who can afford to produce beans for the market. Landless
peasants will suffer from high bean prices.
65
Ravallion, M., Famines and Economics, Journal of Economic Literature, 1997, p. 1213.
56
BIBLIOGRAPHY
Adelman, H. and Suhrke, A., The Path of a Genocide: the Rwanda Crisis from Uganda to
Zaire, Transaction Publishers, 1999
Bart, F., Montagnes d’ Afrique, terres paysannes, le cas du Rwanda, 1993
Bhargava, A., Nutritional Status and the allocation of time in Rwandese Households, Journal of
Econometrics, 77, 1997
Blarel, B.,Hazell, P., Place, F., Quiggin, J., The Economics of Farm Fragmentation : Evidence
from Ghana and Rwanda, The World Bank Economic Review, Vol. 6, no. 2
Byiringiro, F. and Reardon, T., Farm Productivity in Rwanda : the effects of farm size,
erosion, and soil conservation investments, Agricultural Economics 15, 1996
Clay, D., Reardon, T., Kangasniemi, J, Sustainable Intensification in the Highland Tropics :
Rwandan Farmers’ Investments in Land Conservation and Soil Fertility, Economic Development
and Cultural Change, 1998
Clay, D., Promoting Food Security in Rwanda through sustainable agricultural productivity :
meeting the challenges of Population pressure, Land Degradation and Poverty, International
Development Paper no. 17, Michigan State University, Departements of Agricultural
Economics and Economics, chapter 3, p.16.
Gasana, E.,Butera, J.B., Byanafashe, D., Karekezi, A., Rwanda, chapter 8 in Comprehending
and Mastering African Conflicts, ed. Adedaji, A., 1999
Gascon, J.-F., Pauvreté a Gikongoro, Résultats de l’enquête réalisée aupres des ména ges
indigents, Projet de Development Agricole de Gikongoro, Document de travail, n.156, Juin
1992
Gatete, C., Food Security and Food Aid in Rwanda, Wihogora, June 1996
Habyarimana, J, Discourse at the first encounter with public servants
Habyarimana, J, Speech at the opening of the 1980 new year, 1979
IWACU, documentary film, February 1990
Kangasniemi, J., People and Bananas on steep slopes : Agricultural Intensification and Food
Security under Demographic Pressure and Environmental Degradation in Rwanda, Ph.D
Disseration, Department of Agricultural Economics, Michigan State University, 1998
Les Retombées de la Famine dans les Préfectures de Butare et de Gikongoro, Bureau Social UrbainCaritas, Kigali, Février 1990, 26 p.
57
Ngirumwami, J.L., Ministry of Agriculture, Résultats de l‘enquête Commerce Frontalier au
Rwanda, June 1989
Nsengiyaremye, D., La Transition Démocratique au Rwanda (1989-1993) in Guichaoua, A.,
Les Crises politiques au Burundi et Rwanda (1993-1994), Université de Lille
Ntezilyayo, A., L'agriculture, une priorité dans la reconstruction nationale in A. Guichaoua,
ibidem, p. 324
Place, F and Hazell, P, Productivity effects of Indigenous Land Tenure Systems in SubSaharan Africa, American Journal of Agricultural Economics, February 1993, p.14-15.
Platteau, J.Ph., Institutions, Social Norms and Economic Development, Harwood, 2000
Pottier, J., Taking stock, Food Marketing reform in Rwanda 1982-1989, African Affairs
Ravallion, M., Markets and Famines, 1987
Republic of Rwanda, Rapport sur la Situation Alimentaire de notre Pays, Ministry of
Agriculture, Octobre 1989, translation by J.Pottier, ibidem, 1993, p. 20.
Republic of Rwanda, Compte-rendu de la Réunion tenue au Minagri en date du 02/05/1989
sur la situation alimentaire du Rwanda en Avril 1989, Ministry of Agriculture, p. 4.
Sen, A., Poverty and Famines, 1981
Sibomana, A., Hope for Rwanda, Conversations with Laure Guilbert and Hervé Deguine,
Pluto Press, English-language edition, 1999, p.22.
Ravallion, M., Famines and Economics, Journal of Economic Literature, 1997
Thomas, D., Public Policy and anthropometric outcomes in the Côte-d’Ivoire, Journal of Public
Economics, 1996
Twizeyimana, P., and Uwimana, V., Portrait de la Pénurie alimentaire actuelle au Rwanda. Dévoilement
d‘une famine cachéé sous la verdure, CCOAIB, Kigali, Novembre 1989, 43 p. Several attempts to
get a copy of this report in Rwanda failed.
Udry, C and Bardhan, Development microeconomics.
Verbeek, M., A Guide to Modern Econometrics, John Wiley and Sons, 2000
Verwimp, P., Development ideology and Genocide, Journal of Genocide Research, November
2000.
Wiggins, S.Changes in African Farming Systems between the mid-1970s and the mid-1980s,
Journal of International Development, vol.7, n°6 ,1995, pp.807-848
World Bank, Rwanda, Agricultural Strategy Review, 1991
58
Copyright © 2002 @ the author(s). Discussion papers are in draft form. This discussion paper
is distributed for purposes of comment and discussion only. It may not be reproduced without
permission of the copyright holder. Copies of working papers are available from the author.
