Burundi's economy is predominantly based on agriculture which is not only a source of food
for most households but also a source of revenue both in food sales and in the creation of
labour opportunities. The agricultural sector of Burundi employs 90 percent of the workforce
through small-scale, mainly subsistence-oriented family farming units. The sector
contributes about 95 percent to the food supply and represents more than 50 percent of
GDP. It is the first earner of foreign exchanges accounting for 80 percent of the country’s
exports, mainly coffee and tea. The country is recognized as potentially self-sufficient in food
production thanks to its bimodal rainfall pattern allowing three cropping seasons.
However, the performance of this sector is very poor and 70 percent of the population lives
under the national poverty line. The high prevalences of poverty and hunger are mainly
attributed to the rapid population growth which leads to land scarcity and degradation. The
increase in agricultural production remains generally lower compared to the rate of
population growth. The main objective of this study was therefore to find out what really
matters among small scale farms and suggest a model for agricultural production in
response to the problem.
As an attempt to answer the above problem, a research survey was carried out in Ngozi, one
of the northern province of Burundi with a heavy population density (462 inhabitants per
square kilometre). Surveys were held in august 2010. The sampling frame was based on a list
of 360 households selected at random on 90 “collines” or hills (institutional demarcation) of
Ngozi in 2007 by the project UA-FACAGRO (Université d’Anvers-Faculté d’Agronomie de
l’Université du Burundi). A sample of 60 farms was purposely selected using a cluster
analysis based on resource endowments and livelihood strategies. This stepwise approach
was used in order to avoid sampling errors and hence to ensure a more reliable and
representative sample.
Results of statistical analysis showed that smallholder farmers are disadvantaged in both
asset endowments and labour remunerations. The limited access to land leads to livelihood
diversification through both on-farm and non-farm activities. Poor farmers have to diversify
in order to get the amount of money necessary to buy food and to cover additional family
expenses. However, the limited non-farm job opportunities constrain them to engage in
agricultural labour where the levels of payment are very low. The low wages provided to
agricultural labourers are mainly attributed to the low agricultural income which results
from risk attitude of farmers and market failure. Despite the inverse relationship between
farm size and labour availability in the region, the level of complementarities between large
and small farms is very poor. The rate of own consumption among farmers (both rich and
poor) is very high which adversely affects productivity and hence rural wages.
Nevertheless, despite the rampant land scarcity, the model developed in this Master’s
Dissertation shows that there are possibilities for households of Ngozi to optimally produce
crops that meet their food security requirements while generating a certain level of income.
iv
By growing a limited number of crops, farmers can benefit from an optimum land use which
further contributes in improving farm output and increasing agricultural wages offered to
landless households. However, the model highlighted in infrastructure especially storage
system a necessary condition for the success of the agricultural specialization in fighting
rural poverty.
Finally, despite the great effort made in producing this Master’s Dissertation, no claim can
be made to have fully covered the issue of agricultural production in Burundi. Amongst other
limitations encountered, the complexity of agricultural system of Burundi and the use of
Linear Programming model which relies on average farms drive up the author to formulate
suggestions towards more research. Further researches could assess the scope of exchanges
between farms both in input and outputs using Mathematical Programming based MultiAgents
Systems in order to capture the real world heterogeneity and simulate sustainable
resource use.