Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article

Degree

Doctor of Philosophy

Program

Economics

Supervisor

Timothy Conley

2nd Supervisor

Lisa Tarquinio

Abstract

This thesis consists of three chapters on economic development in sub-Saharan Africa. The first two chapters investigate financial inclusion in the region, specifically analyzing the impact of access to savings and credit through a digital financial tool called Mobile Money. The third chapter examines methods for making inferences using the Demographic Health Surveys (DHS) for rural West Africa.
For the past two decades, mobile money, a cellphone-based payment infrastructure, has been the key player in bringing financial services to the unbanked in developing economies. It provides a means for saving and has been widely used to make peer-to-peer (P2P) transfers, shown to be important in helping households deal with bad shocks. Recently, lenders in developing countries have used mobile money to extend digital credit loans to a wider population including the unbanked.
In Chapter 2, I use unique administrative data on mobile money transactions, to examine the impact of mobile credit on mobile money use. Using an event-study difference-in-differences approach, I observe statistically significant declines in P2P transfers in the first, 3 months following the initial take-up of mobile money loans. The most substantial impact occurs in the third month, with a 95% confidence interval indicating a 14% to 31% decrease in the number of transfers sent and a 16% to 34% decrease in the number of transfers received. This decline in the volume of P2P transfers made is associated with a decline of similar magnitude in the number of unique accounts with which transfers are made. This effect is solely driven by borrowers who become delinquent in repaying their loans. I argue that the results are driven by the repayment enforcement mechanism which allows garnishment of mobile money wallets, causing borrowers to avoid using mobile money until their
debt is fully repaid.
In Chapter 3, I examine the role of mobile money as a savings tool in both formal and
informal savings practices in Kenya. Using ownership of multiple SIM cards as an instrument for M-Pesa savings, I find that M-Pesa savings have a statistically significant impact on ROSCA participation and bank savings. However, I find no evidence of an impact of M-Pesa saving on SACCO participation nor on savings done at home. Under the assumption of the usual exact IV exclusion restriction, I obtain a 95% confidence interval of 22% to 99% increase in ROSCA participation and a 20% to 75% increase in bank savings. The impact of M-Pesa savings on ROSCA participation is still statistically significant even when allowing for a violation of the exclusion restriction up to a direct effect of 6% (approximately 50% of the effect of a rural-urban dummy differential on ROSCA participation). Additionally, the impact of M-Pesa savings on bank savings remains statistically significant under violation of the exclusion restriction up to a direct effect of 5% (approximately 60% of the effect of a rural-urban dummy differential on bank savings).

Chapter 4, written in co-authorship with Aldo Sandoval Hernandez, uses data from Northern Nigeria and a linear model of mother’s education predicting a child’s vaccination index to demonstrate the presence of spatial dependence in the Demographic Health Surveys and illustrate its importance for inference. We then review the performance of different methods of inference that account for spatial dependence within a Monte Carlo simulation experiment where outcomes are simulated to have the same spatial correlation structure as identified in the DHS data. We find heteroskedastic-robust standard error estimators as well as clustered standard errors with small clusters over reject the true null. Conley’s (1999) estimator with a uniform kernel performs well at low-distance cutoffs but breaks down at longer-distance cutoffs. Bester Conley and Hansen (2011) clustered standard errors with large clusters defined by latitude and longitude coordinates perform well but can be conservative with 3 groups. Finally, clustered standard errors using Nigerian states perform surprisingly well and are less conservative than BCH with 3 groups.



Summary for Lay Audience

My thesis is comprised of three chapters on economic development in sub-Saharan Africa. The first two chapters investigate financial inclusion in the region, specifically analyzing the impact of access to savings and credit through a digital financial tool called Mobile Money. The third chapter examines methods for making inferences using the Demographic Health Surveys (DHS) for rural West Africa.

Mobile money, a cellphone-based payment infrastructure, has played a crucial role in bringing financial services to the unbanked in developing economies over the past two decades. Initially introduced as a means of storing money electronically and providing a fast, cheap, and easy way to transfer money across distances, mobile money quickly gained traction. In regions where most individuals had limited access to financial services and relied heavily on informal transfers, it soon became a popular means for saving and making peer-to-peer (P2P) transfers. Building on its success, mobile money providers expanded their range of services beyond savings and P2P transfers to include a diverse set of other financial services. One of the most common of these new services was the digital credit product. Mobile money credit has since grown in popularity and is considered transformative for the credit market in developing countries.

In Chapter 2, I use unique administrative data on mobile money transactions from Ghana, to study the impact of mobile money credit on mobile money use. Several studies have shown that mobile money use, particularly for P2P transfers, has a positive impact on households. However, I observe statistically significant declines in P2P transfers in the first, 3 months following the initial take-up of mobile money loans. The most substantial impact occurs in the third month, with a 95% confidence interval indicating a 14% to 31% decrease in the number of transfers sent and a 16% to 34% decrease in the number of transfers received. This decline in the volume of P2P transfers made is associated with a decline of similar magnitude in the number of unique accounts with which transfers are made. This effect is solely driven by borrowers who become delinquent in repaying their loans. I argue that the results are driven by the repayment enforcement mechanism which allows garnishment of mobile money wallets, causing borrowers to avoid using mobile money until their debt is fully repaid.

In Chapter 3, I examine the role of mobile money as a savings tool in both formal and informal savings practices in Kenya. Using ownership of multiple SIM cards as an instrument for M-Pesa savings, I find that M-Pesa savings have a statistically significant impact on ROSCA participation and bank savings. However, I find no evidence of an impact of M-Pesa saving on SACCO participation nor on savings done at home. Under the assumption of the usual exact IV exclusion restriction, I obtain a 95% confidence interval of 22% to 99% increase in ROSCA participation and a 20% to 75% increase in bank savings.

In Chapter 4, I focus on examining various methods of inference with the DHS data from rural West Africa. The DHS is an important source of data for empirical research across several countries. And just like in any study, there are likely to be unobservable factors that are correlated across observations, at least for those sufficiently close to each other. Researchers using DHS employ a variety of methods of inference to account for the possible spatial dependence, however, different methods of inference can produce substantially different results making it difficult for the researcher to choose an inference method. This chapter shows the results of a Monte Carlo simulation conducted using a model that is calibrated to match the DHS dataset from northern Nigeria to evaluate the performance of various methods of inference that allow for spatial dependence across observations.

Download

Share

COinS