Abstract

This thesis contains four closely related essays which address the empirical issues pertaining to the causes, consequences, and households’ responses to food price shocks in Uganda. The first essay investigates the nature of volatilities in agricultural commodity prices in Uganda between 2000 and 2012 by focusing on six key food staples, namely matooke, cassava, maize, sweet potatoes, beans, and millet flour. It studies the behavior of monthly price volatilities of these commodities, examines the extent of their volatility spillovers, identifies their macroeconomic and environmental drivers, and uncover their differential impacts using respectively the General Autoregressive Conditional Heteroscedastic (GARCH), the Vector Autoregressive (VAR) and the Seemingly Unrelated Regression (SUR) models. I find evidence that both unconditional and conditional price volatilities have significantly increased since January 2008 for most commodities, period of turmoil in the global food markets. The GARCH (1, 1) estimates indicate a strong persistence in volatility for most commodities while results from the Exponential GARCH (1, 1) models suggest the presence of asymmetric and leverage effects of unexpected price shocks for half of the commodities. In addition, the VAR estimation results detect limited and mostly unidirectional spillover effects across food commodities. Finally, historical price volatilities of most commodities are found to be primarily affected by volatilities in consumer price indices, fuel prices, and rainfall, with less evidence of strong seasonality effects as previously reported. The second essay presents an empirical analysis of the welfare impacts of food price changes in Uganda using three waves of the Uganda National Panel Surveys (UNPS) spanning over the years 2005-2011. It theoretically investigates the implications of labor market imperfections and households’ heterogeneity in terms of their net positions in both food and labor markets and compares welfare estimates between separable and non-separable models. Through the estimations a panel stochastic production frontier function and a censored-Quadratic Almost Ideal Demand Systems (QUAIDS) with expenditure and shadow wage endogeneities, the results suggest that the welfare effects of price changes (measured in terms of compensating variations) were globally lower in the non-separable agricultural model, implying a high degree of labor market frictions. Furthermore, I find that the welfare effects were unevenly distributed both within and between household groups. Particularly, although agricultural households benefited from price increases as a group between 2005/6 and 2009/10, both significant and insignificant net buyers did suffer from price changes. Moreover, results from non-parametric estimations show that households at the extremes of the welfare distribution were more severely hit by food price instabilities than others. Finally, the essay suggests that the important dynamics in the net market positions observed during the sample period might be attributed to a cost-benefit analysis related to the potential welfare effects of food prices. The third essay explores the question of crop choices and land allocations in environments where farmers face uncertainties about end-of-season output prices and yield levels, weather variability, and formulate expectations about their future levels. Indeed, in the absence of credit and/or insurance markets, farmers are widely expected to adjust their land allocation decisions as a management tool against agricultural and market-related risks. However, little is actually known about the likely differential effects of each of these risk components on farmers’ decisions, particularly when current decisions are allowed to depend on previous choices. Using a nationally representative panel data set for agricultural households in Uganda spanning over the years 2005 – 2012, the paper proposes to investigate the role played by both price and yield risks on farmers’ crop choices and area allocations using a multivariate generalization of the Heckman-type two-step procedure: a multivariate crop selection and a conditional acreage share models. The crop selection problem is modeled as a dynamic multivariate probit regression and estimated through Simulated Maximum Likelihood and the Geweke Hajivassiliou Keane simulator, whereas the conditional acreage share model is estimated using a dynamic multivariate fractional logit model. In both the multivariate crop selection and acreage share models, the results reveal that, while own expected prices and yields are among the main drivers of farmers' crop choices and land share allocations, farmers are found to be more sensitive to changes in expected yield levels than in expected end-of-season output prices. In addition, yield risks, temperature and rainfall volatility appear to have more impact on acreage share decisions than market price risks. Finally, household characteristics are found to play a marginal role in explaining farmers’ crop selection and acreage allocation decisions. The fourth and last essay develops a modified standard Ramsey model to analyze households’ welfare growth and test the assumption that differential exposure to food price shocks leads to different welfare trajectories and to potentially increased risks of poverty traps. The essay focuses on two welfare indicators, namely consumption levels and asset indices, and employs a battery of econometric methods, ranging from parametric GMM fixed effects models to locally weighted scatterplot smoother (LOWESS), local polynomial regressions, and Ruppert et al’s (2003) semi-parametric penalized splines to address nonlinearities in welfare dynamics, identify and locate critical welfare thresholds, and test for the presence of single against multiple welfare equilibria. Using the full sample, I find nonlinearities in welfare dynamic paths and reduction in the growth rates of both consumption levels and assets holdings as a consequence of exposure to food price and asset shocks. However, there is no evidence of poverty traps caused by households’ exposure or vulnerability to food price shocks, but instead I identify only a single dynamic stable equilibrium, located slightly above the official poverty, towards which Ugandan households are converging in the long run. Finally, when disaggregating households into different sub-groups sharing similar characteristics, the empirical results reveal that Ugandan households are converging towards specific welfare equilibria, depending on their initial conditions, demographic characteristics, the extent of their vulnerability and differential exposure to food price shocks. Particularly, I found that, households exposed to food price shocks or above the vulnerability threshold index are expected to move in the long run to welfare equilibria located at lower levels than their unexposed or less vulnerable counterparts.