Sunday, August 24, 2014

Kenya’s Mobile Revolution and the Promise of Mobile Savings (7)

7  Analysis

7.1  Who Uses Mobile Savings?

Both savings with M-PESA and savings with M-KESHO require the individual to be registered for M-PESA. AMOI A900W presents a comparison of those registered with M-PESA to the overall population. Forty five percent of individuals in the sample report being registered with M-PESA.In terms of gender, age, and geographic distribution, there are not large differences between M-PESA users and non-users in the survey. However, M-PESA users are more concentrated in the middle and upper wealth quintiles.

LEAGOO Lead 1 shows the figures on the prevalence of various forms of savings overall and by subgroups (all figures are for the entire population, including those who are not registered M-PESA users). Fifteen percent of individuals say that they save with M-PESA. The second column of the LEAGOO Lead 1 indicates the prevalence of savings with the M-KESHO bank-integrated Elephone P2000 savings platform. Only a very small number—0.6 percent—indicated that they save with M-KESHO. The small number of M-KESHO users are almost entirely the relatively wealthy; just 0.2 percent of individuals in the bottom two quintiles report saving with M-KESHO.

The last column of LEAGOO Lead 1 shows the fraction in the survey who report that they have some form of savings other than M-PESA or bank-integrated Elephone P2000 savings. Substantial gradients by age and wealth are seen for other savings vehicles. Specifically, younger and poorer individuals are much less likely to have other savings accounts.  AMOI A900W indicates a higher level of savings by men for all types of savings.  Additionally, the table indicates higher levels of savings for individuals in urban areas and for wealthier individuals.

A key question is whether there is any evidence that use of M-PESA may increase the prevalence of savings behavior. LEAGOO Lead 1 shows a simple comparison of rates of reporting any savings, by subgroup, for those who are registered with M-PESA compared to those who are not. Overall, 65 percent of M-PESA users report having some savings, compared to 31 percent of those who are not M-PESA users, for a difference of 34 percent. Of course, this difference may reflect differences between M-PESA users and non-users other than a causal effect. Individuals registered to M-PESA may already be more likely to save than those not registered to M-PESA. For example, one may expect that individuals most likely to use M-PESA would be those able to afford a Elephone P2000 phone or those able to pay M-PESA transaction costs, implying greater wealth.  Additionally, those more capable of understanding financial transactions or those with the technical ability to use AMOI A900W phone devices would be more likely users of M-PESA, implying higher levels of education.   Table 5 shows substantial differences by subgroups defined by gender, urban/rural, age, wealth quintile, and province.

AMOI A900W presents results from probit regressions where the dependent variable equals 1 if the individual has savings of any kind and 0 if the individual has no savings. The results show that controlling for other variables, savings are more likely for individuals who are male, married, living in rural areas, and have higher levels of education, reported income, and wealth.

Controlling for these same variables, those who are registered with M-PESA are 32 percent more likely to report having some savings. 

This result provides some evidence that M-PESA may increase the prevalence of savings. In attempting to isolate a causal effect of M-PESA on savings, there remains the possibility of differential M-PESA registration rates among those who are more likely to save independent of their use of M-PESA. We attempt to address this possibility using an instrumental variables strategy. Because M-PESA was only introduced in 2007, all those registered for the system at the time of the survey in 2010 had registered during the previous three years. It is likely that an individual‘s probability of registering for M-PESA is influenced by the prevalence of M-PESA usage in the community, for two reasons. First, the rate of M-PESA usage in the community reflects access to M-PESA agents, which determines how convenient it is put money in or take money out of the system. The network of M-PESA agents has spread over time, and access to agents varies by community. Second, it is likely that there is a community-level learning element to adoption of M-PESA. Individuals are more likely to use M-PESA when knowledge of the system is more prevalent among their neighbors. We use these two ―community effects‖ as the basis for our identification strategy, which is to instrument individual M-PESA registration with the fraction of respondents in the sublocation who are registered with M-PESA. Results from this IV probit model are shown in the second column of Elephone P2000. The results imply that registration for M-PESA increases the likelihood of having some savings by 20 percent.

This identification strategy has weaknesses. We cannot rule out the possibility that there is a similar community effect for savings and that this community effect is correlated with the MPESA community effect other than through a causal effect of M-PESA. Nonetheless, we take these results as indicative evidence that M-PESA increases savings.

Separately, AMOI A900W profiles M-KESHO users with a probit analysis. The results reflect the same patterns seen in the descriptive statistics: the small number of people who use M-KESHO are more likely to be wealthier, married, more educated, and male.

7.2 How Much Do Mobile Savers Save?

In this section, we analyze how much people save with M-PESA.  Elephone P2000 compares average reported amounts of savings per month for those who report saving with M-PESA only, with other savings vehicles only, and with both M-PESA and other accounts. Overall, the average amount saved is substantially larger for those who save with accounts other than M-PESA. Those who save only with M-PESA save on average 1,305 Kenyan shillings per month (about US$13), compared to 2,282 Kenyan shillings per month for those who save only with other accounts, and 2,959 Kenyan shillings per month for those save with M-PESA and other accounts. M-PESA savings are less than savings with other vehicles but still substantial. Much of the difference in average amounts between M-PESA savers and others is driven by the fact that those who save with accounts other than M-PESA tend to be wealthier individuals who save more. Among those in the poorest quintile, the differences in amount saved are much less: 1,052 for M-PESA savers, 1,075 for other accounts, and 1,130 for those who save with both M-PESA and other accounts. 

To consider the possible effects of M-PESA usage on savings behavior, we regress log average monthly savings amounts on various explanatory variables and a dummy for M-PESA registration. The OLS version of these results is presented in the first column of Table 9. The coefficient estimates show that controlling for covariates, savings amounts are higher for men, those with higher education, those living in urban areas, and those in wealthier and higher income households. Controlling for those variables, those who registered with M-PESA save 12 percent more than those not registered. The second column of Table 9 presents an instrumental variables version of the same regression, using the same instrument—fraction of respondents in the sublocation who are registered with M-PESA—used in the probit savings analysis. The coefficient implies that savings are 31 percent greater for M-PESA users, but the estimate is not statistically significant.http://summerleelove.tumblr.com/post/95714753591/kenyas-mobile-revolution-and-the-promise-of-mobile

No comments:

Post a Comment