(iPoster) Infrastructure Inequality and Household Solar Adoption: Evidence from Pakistan

• In Event: Poster Session: Science, Technology, and Environmental Politics

Abstract

With sharply declining global solar panel prices, electricity generation through distributed (rooftop) solar panel has increased rapidly in the Global South in recent years. Yet subnational patterns of solar adoption in these countries vary widely. In developed countries, existing research consistently associates household solar uptake with wealth and education, framing solar as adopted primarily by affluent households. In the Global South, however, this relationship breaks down: solar adoption is often observed in poorer areas that lack reliable public energy services. This raises an important question: why are poorer areas in the Global South adopting solar at higher rates than richer areas?
This paper argues that solar uptake is caused not by wealth but by existing public infrastructure provision. Where the state has invested in infrastructure, households face weaker incentives to adopt distributed alternatives. Conversely, in areas characterized by infrastructural neglect, solar technologies function as substitutes for unreliable or absent public provision, increasing adoption even at lower income levels.
Pakistan provides a compelling case to examine this relationship. In recent years, it has emerged as one of the world’s largest importers of solar panels, and solar adoption extends well beyond affluent urban households. This paper uses Pakistan to study how variation in public infrastructure provision shapes contemporary solar uptake in a largely privatized solar market.
The analysis draws on tehsil-level (administrative level 3, roughly equivalent to county level in the US context) data from the 2023 population census on household lighting sources. The dependent variable is the percentage of households using solar panels. The primary explanatory variable is natural gas access, measured as the share of households connected to the gas network, which serves as a proxy for long-run public infrastructure provision. The models include controls for wealth, male literacy rates, female literacy rates, the share of owner-occupied housing, and the share of employment in agriculture.
As a first step, I estimate correlational models using Ordinary Least Squares (OLS) and a Beta-Binomial specification. These analyses produce results consistent with the paper’s theoretical expectations: tehsils with lower levels of public infrastructure provision exhibit significantly higher rates of solar adoption, even after controlling for socioeconomic characteristics. The Beta-Binomial model, in particular, accounts for the right-skewed and zero-heavy distribution of solar adoption (i.e., a large number of tehsils have no or very low solar adoption) and serves as a robustness check on the OLS results.
Building on these findings, the next stage of the analysis focuses on causal identification. Because gas infrastructure is not randomly allocated, I implement an instrumental variables strategy that exploits exogenous variation in geography, specifically altitude, as a cost shifter for gas network expansion. Altitude affects the feasibility of extending gas infrastructure through engineering constraints while being plausibly exogenous to household solar adoption once photovoltaic potential, socioeconomic characteristics, and baseline development are controlled for. To capture baseline economic development while avoiding post-treatment bias, I also control for night-time lights data measured before the onset of large-scale solar diffusion in Pakistan, which qualitative evidence places around 2015. All models include district fixed effects, restricting identification to within-district variation.

Author

• Unregistered Participant