Public investments play an important role in transforming poverty cycles for rural households. Since public-investment portfolios are often diversified to more effectively allocate scarce resources to meet a variety of needs, it is perhaps more critical for policymakers to know which investments yield the greatest returns towards welfare objectives. This study replicates research on a set of rural Ethiopian households conducted with the objective of quantifying the impact of public investments on the livelihoods of the rural poor. The pure replication supports the original authors’ findings under the central assumption that household consumption levels, capital stock and access to technology change slowly over time. This replication also presents an alternative interpretation of the original qualitative results, with new quantitative evidence on the impact of investments in road infrastructure and the provision of agricultural extension service. Road development is shown to have a direct impact on short-term consumption and crop income growth, though there is weaker evidence of a long-run effect. However, extension services are shown to yield no impact in the short term or in the long run. The approach to the replication is supported by validation tests for the key underlying assumptions made in the original study and includes an application of an improved dynamic panel Generalized Method of Moments (GMM) estimator, which corrects for the bias in standard error estimates generated by the efficient two-step GMM estimator used to draw inferences in the original work. The replication findings suggest high economic returns to further investments in road infrastructure but indicate serious shortcomings in the effectiveness of extension services to support consumption and crop income growth for the rural households studied over the sample period. These results highlight public investments in Ethiopia that are likely to be highly effective when scaled nationally versus investments that appear too inefficient to successfully scale under the current capacity.

The paper is part of the International Initiative for Impact Evaluation (3ie) Replication Paper Series.