Accounting for self-selection bias in assessing impacts of agricultural innovations : the case of farmer support programs in the Eastern Cape

Abstract

The study uses cross-sectional data to test whether FSPs improve technical efficiencies (TEs) of smallholder maize farmers. Using plot level data collected from 30 FSP and 66 non-FSP farmers drawn from Mgojweni, Mabetshe, Lujecweni, Bantingville, Canzibe and Dumasi villages of the Eastern Cape, this study estimates a Cobb-Douglas stochastic production frontier and uses it to compare the efficiency scores of the two farmer types. The results show that FSP adopters had relatively higher TE scores, with over 50% having scores of above 70%. However, a t-test for equality of mean TE scores revealed no statistically significant differences between them (t=-1.3969, p=0.1662), suggesting that the FSPs cannot explain the TE variances. Given that participation in FSPs was not random, a propensity score matching techniques was used to account for self-selection bias. After accounting for self-selection bias the results revealed that FSP adopters were on average 205% more efficient relative to non-adopters. These results underscore the importance of accounting for self-selection bias in demonstrating the impact of agricultural innovations.