Abstrato

A Production Function Analysis for Private Peasant Holdings Crop Farms in Ethiopia: An Application of Robust Regression

Taddesse Kassahun and Fentaw Abegaz*

Four major crop producing regions in Ethiopia, i.e., Tigray, Amhara, Oromia and SNNP were included in the study. Three models for production function: linear, exponential and Cobb Douglas were considered and assessed for statistical model diagnostics. The statistical model diagnostics suggested that crop production function was found to be represented by the Cobb-Douglas function based on data from 2007/2008 agricultural sample survey. The Cobb-Douglas production function was first fitted using ordinary least squares (OLS). As expected, the parameter estimates using OLS were misleading due to occurrence of outliers; hence robust regression was taken as an alternative. Then many of the parameter estimates received the expected signs, R2 values increased and standard errors of parameter estimates decreased. In general, farm size, fertilizer, seed, oxen power and human labor were important to maximize crop yield. The great contribution was found to be due to farm size in each of the regions except in SNNP where it was due to human labor. Education variable was found to be statistically insignificant and received negative sign in Tigray and Amhara. Production elasticities for each of the inputs except farm size in Tigray, Amhara and Oromia suggested that the relation between inputs and output was inelastic.