Short-run effects of grid electricity access on rural non-farm entrepreneurship and employment in Ethiopia and Nigeria
Grid access alone did not shift rural non-farm entrepreneurship in Ethiopia or Nigeria
Read the published paper ↗doi.org/10.1016/j.wdp.2022.100473
Does grid access shift rural livelihoods
Nigeria and Ethiopia are major hotspots of energy poverty, with 58 and 90 million people respectively living without electricity. Together they account for almost 20% of the global electricity access gap. In 2015, 36% of the rural population in Nigeria reported access to the national grid, against just 8% in rural Ethiopia.
Rural electrification is often expected to bring a structural shift: connected households move away from farming towards non-farm enterprise and employment. We ask whether gaining grid access produces that shift in the short run, meaning the two to four years after connection. Prior evidence is mixed and depends heavily on the identification strategy, that is, the assumptions used to separate the effect of electricity from everything else that differs between connected and unconnected households.
Doubly robust difference-in-differences on two panels
We use the World Bank's Living Standards Measurement Study (LSMS) panel surveys, collected with national statistics agencies in three waves: 2010, 2012 and 2015 in Nigeria, and 2011, 2013 and 2015 in Ethiopia. We keep only rural households present in all three waves that did not move house. We drop households that were always electrified or that gained and then lost electricity. This leaves 2,594 of the 3,466 rural households originally surveyed in Ethiopia. In Nigeria it leaves 1,920 (household treatment) or 1,650 (community treatment) of 3,356.
Treatment is household connection to the national grid. In Nigeria we also estimate the effect of community connection, which we consider arguably more defensible because it removes household self-selection into a connection. We apply a difference-in-differences (DiD) design, which compares how outcomes change over time in newly connected households against comparable households that stayed unconnected. Specifically, we use the doubly robust estimator of Sant'Anna and Zhao, extended to staggered treatment timing by Callaway and Sant'Anna. It combines inverse probability weighting with outcome regression, so the estimate holds if either stage is correctly specified. Identification requires parallel trends conditional on baseline covariates, including distance to road, market and administrative centre, household size and composition, education, and an asset index.
A precisely framed null in both countries
We find that rural electrification considered alone was insufficient to trigger shifts in non-farm entrepreneurship or non-farm employment in the two to four years after grid connection, in either country. Estimated changes in the number of non-farm enterprises per household are generally not statistically significant at the 95% level. That is, we are unable to identify a treatment effect distinct from what we would expect under normal sampling variation. This holds for household and community treatment in Nigeria and for household treatment in Ethiopia. We see no evident differences across enterprise types, or for enterprises co-owned by women in the household.
On employment, we find some evidence of a rise in overall employment among both male and female adults in Nigerian households. The Nigerian data appear to suggest a shift away from non-farm employment (not statistically significant) and towards farm employment (statistically significant). None of this is evident in Ethiopia, where error bars are much wider, likely because farm-based self-employment is so common. Our treatment groups are small relative to the control groups, which may leave too little statistical power to detect smaller effects.
Justify electrification by the capabilities it provides
For rural Nigeria, our results run contrary to recent work on the same dataset, which found a positive association between electrification and non-farm employment using instrumental variables and matching. Our work is not a replication and was conceived independently. The divergence suggests how strongly the identification strategy can shape the result. It aligns with a growing literature urging caution about inference from identification strategies that may be biased.
We speculate that the structural economic effects of electrification should be evaluated over longer periods. Complementary infrastructure and capacity can take substantial time to develop, and short survey panels may miss slower shifts. We close with a caution on interpretation. A strict cost-benefit analysis informed by studies such as ours might conclude that rural electrification is not worthwhile, especially in remote areas. We consider that view too narrow and propose that rural electrification is justified by the capabilities and freedoms it provides households and businesses. Designing electrification policy around expected short-term economic shifts will likely disappoint policy makers and may increase inequity. We argue that infrastructure provision should instead be embedded in frameworks of human well-being, such as Decent Living Standards.
@article{pelz_short-run_2023,
title = {Short-run effects of grid electricity access on rural non-farm entrepreneurship and employment in {Ethiopia} and {Nigeria}},
volume = {29},
copyright = {All rights reserved},
issn = {24522929},
url = {https://linkinghub.elsevier.com/retrieve/pii/S2452292922000819},
doi = {10.1016/j.wdp.2022.100473},
language = {en},
urldate = {2022-12-09},
journal = {World Development Perspectives},
author = {Pelz, Setu and Pachauri, Shonali and Falchetta, Giacomo},
month = mar,
year = {2023},
pages = {100473},
}