UC Berkeley

Electricity is widely considered to be critical for economic growth, but recent literature provides mixed evidence on the welfare implications of rural electrification in developing countries. Understanding the welfare impacts of electrification requires detailed knowledge about the determinants of consumer demand for electricity in rural, low-income settings. My dissertation uses a combination of field experiments and natural experiments to contribute to current knowledge of consumer demand for electricity and explore the associated implications for consumer welfare. I employ a unique product to address current limitations in the literature: pay as you go (PAYGo) solar. PAYGo solar is explicitly designed to lower barriers to electrification for rural, low-income households. It does so my requiring down payments which are small relative to the cost of purchasing a solar home system outright or being connected to the grid, and by allowing for flexibility in ongoing payments to use the system. In each chapter of my dissertation, I partner with a PAYGo solar company operating throughout sub-Saharan African to study a distinct aspect of consumer demand for solar electricity. The first chapter of my dissertation studies consumer responses to randomly offered bulk discounts and monthly rewards for electricity purchases in Rwanda and Kenya. Both types of incentives potentially alter the average and marginal price that consumers pay for solar electricity. I find that most consumers do not respond to either type of incentive. While this could suggest that demand for electricity is inelastic on the intensive margin, I provide suggestive evidence that uncertainty over the future marginal utility of solar as well as liquidity constraints may also play important roles in explaining consumer non-responsiveness to price. In the second chapter of my dissertation, I use a second field experiment with solar customers in Rwanda to formally explore the role of liquidity constraints and transaction costs in shaping demand for solar electricity. The key observation underlying this chapter is that PAYGo solar contracts render electricity access a strictly perishable good, or a good that cannot be stored. I show that transaction costs for perishable goods create welfare loss. The loss comes from the trade-off between transaction costs and the waste that occurs when perishable goods expire, a trade-off that is compounded by liquidity constraints. I explore the trade-off between waste and transaction costs through a field experiment where I randomly offer 2,000 current solar customers a line of credit for solar access time. The line of credit both alleviates liquidity constraints and lowers transaction costs. Consumers who previously bought in bulk respond by eliminating wasteful consumption, reducing demand by up to 6.4%. Those who are the most likely to be liquidity constrained increase demand by 88%. My results illustrate that transaction costs for perishable goods distort willingness to pay in opposite directions for different subsets of consumers. I estimate consumer surplus from electricity under the less distorted conditions of my experiment. My estimates indicate a stronger cost-benefit proposition for universal electrification than others in the literature, but indicate that marginal households' willingness to pay still falls below current cost-covering levels. In the final chapter, I turn to the aspect of demand for electricity that I could not address with either of the field experiments in the first two chapters: the initial decision to adopt electricity. I specifically ask how much electrification choices depend on the expected price of electricity on the intensive margin. I use the phased rollout of a subsidy for PAYGo solar electricity in Togo to measure impacts on electrification. The subsidy reduces the intensive margin price of electricity by 18--42%, with the largest discounts going to consumers choosing the smallest systems. The subsidy increases the adoption of small systems by 85%--135% under different specifications, implying relatively high elasticities of 2.02--3.25. Increases in adoption of medium and large systems range from 23% to 63%, implying elasticities between 1.33 and 3.54. I compare welfare estimates from the subsidy in Togo to my work on the intensive margin of PAYGo in Rwanda in chapter two. Discrepancies between the two studies point to limitations of revealed preference measures for durable goods: both intensive and extensive margin demand may be shaped by a variety of non-price factors that fail to accurately measure welfare. I conclude by offering directions for future research on the economics of rural electrification.