Accounting for technical efficiency differentials among smallholder tobacco farmers in Hurungwe Zimbabwe : impact of self-selection bias on contract participation

Abstract

This study used farm-level data to test whether contract farming can account for technical efficiency differentials amongst smallholder tobacco farmers from Hurungwe District in Zimbabwe independent of self-selection bias, in response to the need (i) to inform policy on contract farming in Zimbabwe as an agricultural finance model, and (ii) the literature on impact evaluation which hypothesizes that contract participation is not a random process. A sample of 240 smallholder tobacco farmers was split into a treatment (contract farmers) and a control group (non-contract farmers) to enable comparison. Using 2016/17 farm-level production data collected through face-to-face interviews by means of structured questionnaires, the study compared 75 contract and 165 non-contract farmers purposefully selected through stratified random sampling. A Cobb-Douglas Stochastic Frontier Production Function (SPF) model was used to estimate technical efficiency differentials across the sub-samples before and after accounting for self-selection bias using Propensity Score Matching (PSM) techniques. Without accounting for self-selection bias, the results show that contract farmers had a mean technical efficiency score of 83 percent (95 percent CI 0.799: 0.851) compared to 81 percent (95 percent CI 0.794: 0.819) for non-contract farmers. A t-test for equality of means showed no significant differences between the two groups (t=-1.4332, p=0.153), suggesting that participation in contract farming cannot account for the observed technical efficiency differentials. However, using PSM techniques to account for self-selection bias and the stratified matching algorithm, the results show that contract farmers were on average 4.8 percent (t=4.075, p=0.012) more technically efficient relative to their non-contract counterparts, suggesting that accounting for self-selection bias matters in evaluating the impact. In the second stage of the SPF, bio-physical, socio-economic and policy variables were used as covariates to investigate determinants of technical efficiency across the two groups. For this group of farmers, the results suggest that household size (t=2.34, p=0.020) education level (t=1.96, p=0.061), access to extension services (t=2.22, p=0.027) and tobacco farming experience (t=3.48, p=0.001), and membership to a farmers’ group (t=2.84, p=0.008) showed a positive effect on technical efficiency. Meanwhile area allocated to tobacco farming (t=-2.57, p=0.011) and off-farm income (t=-2.49, p=0.013) showed a negative effect on technical efficiency. These results suggest that, in addition to formulating policies that promote contract farming, policy makers should also work on policies that improve access to extension services, education, promote the formation of farmers’ groups and encourage farmers to join them if productivity in smallholder tobacco farming sector is to be increased. In a final model, the study established that membership to farmers’ groups (t=1.92, p=0.054), agricultural field day attendance (t=2.86, p=0.004), and farm size (t=4.65, p=0.000) increased the probability that farmers will choose to participate in contract farming. Thus, to promote contract farming, the government and policy makers play an important role by encouraging farmers to join farmers’ groups, attend agricultural field days in addition to making farmland accessible. In conclusion, there exist considerable productivity losses due to inefficiency among this group of smallholder tobacco farmers that could be addressed at policy level. The study recommends contract farming as one policy vehicle that could be used to address such inefficiencies in capital-constrained smallholder agriculture. Further the existence of self-selection bias in contract participation must be addressed before assessing the impact of contract farming on technical efficiency of smallholder tobacco farmers. While acknowledging the role of contract farming in addressing productivity losses in the smallholder tobacco farming sector, the study also noted that there are additional variables that could be targeted by policy if these efficiency losses are to be addressed.