The Future of Smallholder Farming in Eastern Africa - Uganda ...

densely populated Nyeri and Kakamega districts (Migot-Adholla et al 1994b). Although
the study looked into the relationships between land tenure, agricultural investment and
farm productivity, it never established land tenure effects on the environment. It is more
than a decade ago since the study was carried out and it would be interesting to see
whether things have changed with time. Is it possible that the findings of the study and
others carried out in other countries still hold We have deliberately chosen agroecological
zones that totally differ from those in the 1988 study so as to establish whether
the relationships established in that study can hold in a totally different agro-ecological
zone. This would form a good basis for the necessary policy formulation.
The agricultural sector in Kenya has registered poor growth over the last decade with
growth falling to negative 2.4 percent in the year 2000 (Economic Surveys 1997 – 2001).
This has been blamed on among others, increasing frequency of drought and falling
agricultural productivity. These two may be an indication that Kenya is unable to achieve
sustainable growth. Empirical evidence shows that environmental degradation, more so
clearing of the bush, including those in major water catchment areas, can contribute to
climate change and even desertification. Moreover, the usefulness of vegetation cover to
formation of rainfall cannot be underestimated (Heyer and Waweru, 2000).
The declining land productivity, according to reports available at the Ministry of
Agriculture and Livestock Development, has necessitated increased use of farm inputs
such as fertilizer and farmyard manure. Increases in cost of these inputs has raised the cost
of production and reduced the competitiveness of Kenyan agricultural products in local
and international markets. In neighboring Uganda and Tanzania, the land is still relatively
richer and requires almost no fertilizer or manure thereby making the cost of production of
commodities like maize and beans lower than in Kenya. 27
Various methodologies have been used in the past studies, varying based on data and
researcher preferences. Migot-Adhola et al 1994a adopted single equation econometric
models and relied on Logit regression models. This was critical for their data in which the
dependent variables were binomial. The Ordinary Least Squares (OLS) may be used in
cases where the dependent variable is continuous. Roth et al (1994) used such a method
where land values formed the dependent variable. Migot-Adhola et al 1994a also used a
continuous dependent variable, crop yield. Each model has to be carefully applied due to
the possibility of mis-specification, multicollinearity, heteroscedasticity and
autocorrelation. Some researchers use two- or three-stage least squares to deal with these
problems. With discrete dependent variables or those requiring censoring, tobit regression
analysis is more commonly used. An example is Otsuka et al (1997), in which the
dependent variable was the proportion of area under different land tenure regimes.
To determine the nature of variables used, two variable analyses are often used, including
Pearson Correlation Coefficient. In this study, various regression techniques are used, with
the yield model being based on the OLS that is substantially reduced to minimize
occurrence of mis-specification or simultaneity bias. A Tobit Model is developed for
determinants of environmental degradation since the dependent variable requires
censuring and is discrete. A Logit model is also used due to the dummy nature of the
dependent variables.
27 It should be noted that these countries still have expanse of land to be cleared for cultivation and the soils
have not been exhausted as in Kenya.