The World Bank’s Financial Access Database , which has quite recent representative data for most countries in the world, indicates that in Africa, for instance, only about a quarter of the population has a bank account. And less of them use it.
So why don’t more people use banks? Since folks occasionally ask me if I am a teller at the World Bank, this is a question I wonder about, particularly in Africa where I work. A nice new paper  by Simone Shaner gives us some insight into this – and some thought provoking results across gender lines.
So the obvious reason why people don’t use banks is that they are a pain – a pain to get to, costly to use, and generally inconvenient. And, when interest rates are low, there are a lot of decent informal alternatives. So making the account free (as I discussed in my post  on a paper by Dupas and Robinson) and low on transaction costs, ought to make things easier.
But Shaner lays out some reasons why this might not automatically be the case. First, if you have self-control problems, like I do, then making your money hard to get to might be something you actually are willing to pay for. Second, if you face demands from your kin for loans or gifts, putting cash in a hard to reach place might make it a bit easier to say no. Finally, if you are trying to shield your money from grabbing or other kinds of interference from your spouse, having it off-site might be a good thing (for you, anyhow). So, ultimately, this is an empirical question, of the simple economic theory versus the messy real world constraints variety.
Shaner worked in western Kenya (the bubbling cauldron of new experiments in economics) with a commercial bank to test this. Basically, her experiment is to offer couples (with ID cards and interested in opening a bank account) the option to open any and all of three types of account: single accounts for the husband and wife, and a joint account. These accounts are free (she covers the $1.25 non-withdrawable starting balance). After making the account opening decision, accounts were then randomly selected to receive a free ATM card. This is not a trivial subsidy – ATM cards from this bank usually cost $3.75 – not a small amount for the study population (and indeed, only 7% of the control group buy these cards on their own). One interesting side note to keep in mind is that by sequencing the randomization in this way, Shaner is isolating the impact of the ATM card, removing any impact the offer of the card might have had on the incentive to open the account in the first place – indeed this effect is the focus of her analysis. Finally, Shaner also randomly varies the interest rate that folks receive on their account
To look at the impacts of this, Shaner uses a combination of a baseline survey (which includes some discount rate elicitation with actual payoffs) and administrative data. No endline survey – so this is clearly on the cheaper side of evaluations and means she is going to be constrained to bank account outcomes.
Before turning to what she finds, it’s interesting to look at how people are saving pre-intervention. The most common tool (85% of husbands and 90% of wives) is savings at home, which averages around one week worth of earnings. This is interesting to me – all of the times I’ve asked about under the mattress cash, I haven’t got numbers this high (including in Kenya) – so clearly she is doing something to gain people’s trust. The other most common tool for savings is ROSCAs. In terms of bank based savings account, 32% of men and 12% of women report owning a bank account (much lower than the Kenyan averages in the national data in the World Bank’s financial access database).
So what does she find? Two thirds of couples opened a joint account and 45-47% of couples opened up single person accounts – with no significant difference across men and women. Use of these accounts isn’t high: 45% of couples saved in at least one account, and this number is only 27% when she excludes the folks who won at least one cash prize (in the elicitation part of the survey – many of whom chose to deposit the newly won cash in the bank).
On average, ATMs cards, which make it easier to access your cash, do increase the use of bank accounts by 0.19 standard deviations. When she unpacks this, cards give a 0.302 standard deviation boost to joint account use, a 0.265 standard deviation boost to men’s account use but no significant boost to women’s account use. One important caveat here (and one where she points out that the absence of an endline is an issue) is that she cannot tell if this increased bank account use came at the expense of other savings mechanisms.
It is important to note that while we see this kind of heterogeneous response to the ATM card treatment, there is no difference between savings rates or average balances in the control group (remember, they too got bank accounts). So what’s going on with the women who got ATM cards?
It could be a lot of things, but Shaner zeroes in on bargaining power. The idea here is that having an ATM card might make it easier for men to get at their wives’ money. After they get ahold of the PIN, and maybe the card, they can help themselves. The card itself isn’t concealable, because the lottery for the card was held while the couple was sitting at a table together. Shaner tests the role bargaining power might play by bringing in four potential measures of bargaining power: relative income, education, literacy, and age – with the idea that having more of these things (at baseline) gives you more clout in the household.
Using this measure, she unpacks the heterogeneity further. First of all, men’s positive response to the free ATM cards appears to be driven by men who are above the median in terms of bargaining power. If the women have more bargaining power, then the effect for men having an ATM card isn’t significantly different from zero. When the woman is offered an ATM card, the effect if she has below median bargaining power is actually negative, with a positive but not significant response if her bargaining power is above the median. Bottom line: giving an ATM card to women with lower bargaining power reduces their use of bank accounts.
Shaner goes through some robustness checks to convince us we’re seeing this effect. These include bringing in the interest rate experiment and showing that while folks are more likely to use bank accounts if the interest rate is high, this response doesn’t change with bargaining power (and hence the ATM treatment isn’t just picking up a response to better account terms).
This is a thought-provoking study – making it clear that there are non-trivial gender dimensions in extending access to finance. And it’s not just price which matters.