Econometric analysis of smallholder agricultural households for Zimbabwe

Thumbnail Image
Zindi, Christopher
Journal Title
Journal ISSN
Volume Title
The objectives of this study were to test for separation of agricultural household models and to estimate separable and nonseparable econometric models of maize output supply, yield response to price, acreage demand, consumption demand and marketed surplus supply. Regression models run include OLS, 2SLS, SUR and 3SLS. The study uses a cross-sectional data set of peasant households spread across five natural regions of Zimbabwe. The sample size used in this study is 215. Three main methods are used to test for separation of household production and consumption decisions. Results of the tests support separation thus implying that production and consumption decisions of households are nonjoint. Despite the lack of support for nonseparation, nonseparable models are estimated as well since they are relevant to the environments in which peasant households operate. Results of hypothesis tests suggest that maize and fertiliser prices generally have insignificant effects on separable and nonseparable models of the functions investigated. However these variables have some significant effects on some amily farm labour demand functions. Dropping the numeraire price in the maize consumption function improves the statistical significance of maize price in this function. Significance of the numeraire price in the maize output supply, marketed surplus and consumption function suggests that dependent variables of these functions increase with an equiproportional increase in all prices. Results have to be interpreted with caution since insignificance of prices may be due to errors in the model such as difficulties in measuring expected prices and constraints on production. Although separation is supported, there are no marked differences in results obtained using separable or nonseparable agricultural household models. Results imply that household production and consumption decisions are not conditioned by the nature of the household model used.