The Impact of Agricultural Extension and Roads on Poverty and Consumption Growth in Fifteen Ethiopian Villages
Original Researchers: Stefan Dercon, Daniel O. Gilligan, John Hoddinott, and Tassew Woldehanna
Original Publication: American Journal of Agricultural Economics
Replication Plan: Bowser’s Replication Plan
Current Status: Completed Replication Study
The Original Study
This study assesses the impact of improvements in agricultural extension and road quality on consumption growth and poverty in Ethiopia from 1994 to 2004.The Ethiopian government has continued to encourage agriculture extension through large-scale investment with a belief that increased agriculture extension investment; government can better equip farmers with new technologies and schemes and can better prepare farmers for adverse events. The same is believed to be true for improved road quality: Increased public investment in roads brings better access to farms and reduces operational costs. This in turn reduces the costs of collecting inputs, helps farmers to obtain high sales prices and, with increased profits, leads to investment in diverse farming activities. To capture the growth process as a result of improved technologies, the authors used an empirical growth model method. Correlations were addressed using an instrumental variable estimator. A fixed-effects estimator was used to control for time-invariant household characteristics. This approach addressed attrition, and the instrumental variable was used to reduce measurement bias in the regressor.
In this replication study, the original authors employ a GMM-IV with fixed effects to control for selection bias. Many targeted development programs, especially those related to public investments, suffer from selection biases due to initial conditions (Jalan and Ravallion, 2002; Khandker et al., 2009). This replication aims to further investigate the original findings from (Dercon et al., 2009) by first replicating the original results and validating the assumptions which drive the primary empirical model. Following this, the replication will examine the sensitivity of the results to evenly spaced time intervals using the authors' GMM-IV fixed effects model, with alternative considerations for the set of instruments used to correct for possible endogeneity. Then, first-differenced and system GMM-IV estimators will be deployed on the evenly spaced data, with the inclusion of the 2009 survey round data. And finally, the replication will consider intermediary outcomes along the causal chain to clarify the theory of change.