UC Davis

This dissertation adds to our understanding of how mobile phones play a role in improving development opportunities. They not only provide a channel to communicate with others, but generate troves of data about the economic lives of subscribers. I contribute to the literature by examining three aspects of the potential of cell phones as development tools.

In chapter one, I study the most widespread form of digital credit: airtime loans. This service allows prepaid customers to borrow small airtime advances for a fee instead of purchasing recharges that must be paid upfront. Relying on rich administrative data from a mobile network operator in Haiti, I show access to loans increases total communication expenditure by 16% but with distinctly heterogeneous impacts. When airtime loans become available, poorer customers more than double their mobile communication spending, while access to loans leaves expenditure of the highest tercile unchanged. These differences exist despite relatively uniform patterns of loan usages between the poor and non-poor. I argue these differences are driven by distinct motivations for requesting airtime loans, with poorer customers using the loans to relax short-term liquidity constraints at critical communication times whereas non-poor customers primarily use these loans for convenience, as it gives them more discretion in when to visit airtime vendors.

In chapter two, I build on the evidence that demonstrates mobile phone metadata, in conjunction with machine learning algorithms, can be used to estimate the wealth of individual subscribers, and to target resources to poor segments of society. This paper uses survey data from an emergency cash transfer program in Haiti, in combination with mobile phone data from potential beneficiaries, to explore whether similar methods can be used for impact evaluation. A conventional regression discontinuity-based impact evaluation using survey data shows positive impacts of cash transfers on household food security and dietary diversity. However, machine learning predictions of food security derived from mobile phone data do not show statistically significant effects; nor do the predictions accurately differentiate beneficiaries from non-beneficiaries at baseline. Our analysis suggests that the poor performance is likely due to the homogeneity of the study population; when the same algorithms are applied to a more diverse Haitian population, performance improves markedly. We conclude with a discussion of the implications and limitations for predicting welfare outcomes using big data in poor countries.

In chapter three, I provide evidence on the determinants of the adoption of mobile money services. In contrast with previous research that centers on adoption after a service is launched, I study a mature platform that experiences a stagnating user base. I combine a survey with a randomized component with mobile money transaction logs to test if informational videos induce people to open an account and try new products. My results show that awareness of mobile money services is high and, even if having an account is free, many people use the service indirectly by asking others to make transactions for them. The intervention increased adoption by 5.4\%. However, a large share of new users came from the group that declined the opportunity to watch the videos, indicating more than simply information drove their decision to adopt. I do not find the videos increased the usage of additional services by people with an account at the time of the survey. Taken together, my results show further growth of the mobile money platform requires increasing the number of services available to attract additional customers and incentivize the daily usage of mobile money for economic transactions.