The Effect of Initial Land Distribution on Income Growth and Distribution in Kenya

Abstract

Poverty is a major challenge worldwide and poverty reduction remains a critical aspect of developmental efforts. The rate and extent of poverty reduction associated with agricultural growth in primarily agrarian societies has commonly been assumed to depend importantly on the distribution of productive assets in general and land assets in particular. However, recent evidence by Jayne, et al., (2003) and Jayne et al., (2014) has shown that landholdings are becoming more concentrated and operated land is declining in some African countries like Ghana, Zambia, and Kenya. This trend means that efforts by governments and donor agencies to reduce chronic poverty will partly depend on understanding the joint effect of the distribution of productive assets on growth and income distribution. Without this understanding, agricultural productivity growth may contribute less to rural poverty reduction than usually assumed. Understanding the potential adverse effects of the rising land concentration on agricultural growth’s poverty reduction potential, also offers an opportunity to limit the wastage of government and donor resources, as interventions aimed at poverty reduction can be targeted at other growth enhancing investments that yield higher returns than investments aimed at improving agricultural productivity. Against this background, this study explains the relationships between initial land distribution at the village level, growth of household income and the distribution of income at the village level in Kenya. Although a number of studies have looked at the relationship between inequality and growth, the difficulties associated with such a process remain central to the debates in the growth literature. Finding high quality panel data that allows for cross-country estimation using panel econometric techniques remains problematic. Even with high quality data, questions remain on the comparability of institutional structures across countries. The problem of endogeneity and how to address it also remains at the centre of growth empirics and past studies have revealed several weaknesses. In combination, these challenges have been cited as a major internal threat to validity of past growth empirical works. To overcome the econometric challenges in the majority past studies, this study uses the system generalized method of moments to deal with the dynamic nature of the growth model and the unobserved individual- and time-specific effects. It also uses the first differences estimator for the econometric analysis of the effect of land distribution on income distribution since its construction is not dynamic. To address the endogeneity problem in both models, I use the “Jackknife” procedure and/or lagged values o of all endogenous variables. For instance, the endogeneity arising from the fact that growth and distribution affect each other, the village land gini coefficient corresponding to a household is computed using observations of operated land for other households in a village while excluding its own. This study also uses a 13 year panel micro-level dataset for Kenyan rural farming households to overcome the data-related challenges. By using a nationally representative panel dataset, the study also takes care of the unobserved heterogeneity that gets lumped up in the error term in the discredited single cross-sectional analyses. By adopting a joint determination of the effects of various covariates on growth and inequality, the study recognizes the fact that the two outcomes are generated by the same underlying processes and thus should not be analyzed in isolation. This is advantageous as it yields complete messages for the policy maker whose interest is in promoting growth while reducing inequality. While it seems impossible to fully solve for the endogeneity problem and thus establish causal effects, the study identifies the effects of several variables on growth and inequality while arguing that it is possible under certain qualifications, which are explained in the literature review section. The results reveal that operational landholdings at the village level are becoming fairly concentrated over time in Kenya and that rising land concentration has impeded growth. The distribution of both income and operational landholdings spatially is such that the two reduce as the distance from the district/headquarters rises.