Gender and policy effects: Fertilizer use, crop diversification, and sampling weights in informative sampling

The Food and Agriculture Organization of the United Nations (FAO) has indicated that closing the gender gap in agriculture can reduce food emergencies. The FAO reported that 44% of Zambia\u27s population were undernourished between 2006 and 2008. In Zambia, over 70% of the population and 78% of females derive their income from agricultural production. Designing policies that target females can reduce food security emergencies. This thesis examines statistical issues related to complex survey design, and the influence of input subsidies on input usage and crop diversification among small-scale farm households for the nation of Zambia. The 1999/2000 Post-Harvest Survey and subsequent supplemental waves were used to study the effect of gender and policy gender effects on input usage and crop diversification. The data were collected in a stratified multi-stage sampling design which could result in informative sampling issues. In the presence of informative sampling, sampling weights are related to the response variable resulting in inconsistent estimators. Thus, the estimation of gender and policy gender effects could result in incorrect policy recommendations. Generalized Method of Moments (GMM) optimal instrument theory was used to obtain a consistent fixed-effects estimator in Generalized Linear Mixed Models (GLMM) in the presence of informative sampling. Estimation of a Double-Hurdle model indicated that the subsidy is key factor in the farmer\u27s decision of whether to adopt new agricultural technology and how intensely to adopt it. Estimation of two crop diversification indices indicated that the subsidy affects the number of crops planted by the household. Results indicated that in terms of proportional distribution of crops in the farm, subsidized female headed households were more responsive to the subsidy compared to male headed households. Estimation of crop shares in a multivariate regression model indicated that the subsidy encourages greater farmland allocation to maize and groundnuts, and less farmland allocation to cassava and millet. Crop shares estimation results also indicated that unsubsidized female headed households allocate less farmland to maize and cotton, and more farmland to groundnuts. Finally, simulation results showed that the proposed fixed-effects GMM estimator for GLMM models is consistent under informative sampling