This paper investigates an empirical puzzle in technology adoption for developing countries: the low adoption rates of technologies like hybrid maize that increase average farm profits dramatically. I offer a simple explanation for this: benefits and costs of technologies are heterogeneous, so that farmers with low net returns do not adopt the technology. I examine this hypothesis by estimating a correlated random coefficient model of yields and the corresponding distribution of returns to hybrid maize. This distribution indicates that the group of farmers with the highest estimated gross returns does not use hybrid, but their returns are correlated with high costs of acquiring the technology (due to poor infrastructure). Another group of farmers has lower returns and adopts, while the marginal farmers have zero returns and switch in and out of use over the sample period. Overall, adoption decisions appear to be rational and well explained by (observed and unobserved) variation in heterogeneous net benefits to the technology.

We explore the impact of reduced transaction costs on risk sharing by estimating the effects of a mobile money innovation on consumption. In our panel sample, adoption of the innovation increased from 43 to 70 percent. We find that, while shocks reduce consumption by 7 percent for nonusers, the consumption of user households is unaffected. The mechanisms underlying these consumption effects are increases in remittances received and the diversity of senders. We report robustness checks supporting these results and use the four-fold expansion of the mobile money agent network as a source of exogenous variation in access to the innovation.

Mobile money allows households in Kenya to spread risk more efficiently. In this paper we show that these efficiencies are achieved through deeper financial integration and expanded informal networks. Active networks are more geographically dispersed and support more reciprocal financial arrangements. Consistent with the reported reciprocity, mobile money users report a higher share of transactions as being for credit and insurance purposes.

We study the potential for asset collateralization to expand access to credit in rural Kenya. Increasing the share of a loan for a durable agricultural asset that is collateralized by the physical asset itself (from 0 to 96%) while reducing the share backed by financial assets increases loan take-up considerably, with only a very limited impact on repayment behavior and the lender's profitability. A Karlan–Zinman test finds evidence of small and marginally significant selection effects in some specifications but no evidence of moral hazard. We find no evidence that joint versus individual liability affects take-up or repayment. Loans had real impacts on investment, milk sales, and girls’ school enrollment. The lender, a savings and credit cooperative, responded to the study results by offering 80% asset-collateralized loans.