Syndicate content

It's Hard To Save at Home

Markus Goldstein's picture

So say 87% of the respondents in a survey used by Dupas and Robinson in an interesting forthcoming paper on what happens when you help people get set up with bank accounts in Kenya. And, as we will see, this problem seems to be particularly acute for women.  

Their study takes place in western Kenya, in the rural market town of Bamala, population 3,500 (they point out that this is the 189th largest town in Kenya).   There’s one bank in town, and the next nearest bank is 25km away. As you might imagine, not a lot of folks have bank accounts. In fact, at baseline only 0.5% of the daily income earners they surveyed had a bank account with the bank in town (2.2% had a bank account elsewhere). Why not?   Reasons ranged from inability to pay the account opening fees to a lack of information about the bank and its services. 

Into this setting comes an intervention designed to get folks into the banking system.   Dupas and Robinson and their team went out and found a bunch of small scale entrepreneurs.   They end up with two main groups: female vendors (selling foodstuffs or charcoal for example) and male bicycle taxi drivers (known in Kenya as boda drivers). They did have a small sample of male vendors, but this is too small for inclusion in the results that follow.   Stratifying by gender and occupation (given the high correlation, this basically reduces to stratification by gender), they randomly offered half of the respondents the intervention.  

Individuals selected for treatment were offered the option to set up a bank account at the village bank with no cost to themselves.   This entailed the research team paying for the account opening fees, as well as making the deposit of the minimum balance ($1.43), which individuals cannot withdraw.   Now, the thing to keep in mind is that this bank account had a negative effective interest rate.   There was no interest paid on deposits (and inflation is well above zero) and there was a sliding scale of fees for withdrawal. In addition, the bank is somewhat inaccessible: it’s open from 9-3 Monday to Friday and there are no ATM cards issued. 

The data for this study are kind of interesting and bear a bit of discussion.    First, they have a baseline survey, which gives them a bunch of covariates. Second, with the respondents’ consent, they pull data on each deposit and withdrawal from the bank’s administrative record.   Third, they collected a bunch of time and risk preference data and a measure of cognitive ability (the varied timing on when this was administered limits what they can do with this). 

Finally, starting 2-3 months after the intervention, they collected detailed data on income, expenditure, business activities, labor supply, and transfer through daily logbooks.   For those who were literate, enumerators checked in with the respondents twice a week.   For those who were illiterate (33% of women and 9% of men) enumerators visited every day to help fill out the logbooks. 

As you might imagine, this is pretty intensive work for the respondents.   To sweeten the incentive to provide data, Dupas and Robinson paid folks 71 cents for each week the logbook was filled out correctly. But even with this, some folks didn’t keep the logbooks for the whole three months of this part of the data collection exercise (more on this in a minute). In the end, these data will give them daily average measures of a number of business metrics.   However, on one metric they end up having trouble: profits or, more specifically, revenues.   Their hypothesis, which makes sense, is that the respondents did not keep good records of each sale during the course of the day (investments tended to be more lumpy and hence better recorded). Hence average profits tend to be negative – and so they end up not using this. 

Now, as you can imagine, if someone came and asked you to do a daily logbook you might tell them to get lost.   And a non-trivial number of folks did this to the research team.   Of those they could track down after the baseline survey, 17% refused to fill out the logbooks – and here again gender differences show up – 7% of women declined but for men this figure was 21%.   It turns out that this attrition, at least among men, may be associated with unobservable characteristics, and this will temper their results for men (although, as you will see, there remains a good reason to believe their results on men).  

So that’s the data – now what happened?   Given the offer, most people (87%) elected to open a bank account.   So there is a demand for bank accounts out there.   Now 40% of those offered an account opened one but never made a deposit.   The remaining folks used the accounts a fair bit. Women used them more than men – their mean level of deposits was twice a (around $40 versus $18). Indeed, the average man used them so little that when they estimate the effect of treatment on men actively using the bank account, the coefficient is pretty close to zero, and not significant (while for women it is 0.40 and significant). Remember this part is based on administrative data on not subject to the logbook attrition issue.   So their conclusion that this intervention seems to have had no impact on men seems warranted. 

Overall deposits for women weren’t that frequent, but they were fairly substantial: the median woman deposited an average of 1.6 times daily (household) expenditures.   And an interesting result is that this does not seem to have displaced savings through other vehicles.  

This increased savings by women seems to have translated into increased investment - Dupas and Robinson find that the bank account led to a 180 shilling increase in daily investment over a base of 300 shillings. However, this is significant at 10%, and when they trim the top 5% of the data, significance drops to 14% (and it looks like they are not powered here for the gender disaggregated effect). On the expenditure side, the results are stronger : food expenditure for women shows a significant increase – going up by over 16%. In addition, women’s private expenditure (i.e. on items that can be attributed to consumption clearly by them), also goes up, by more than 36%.  

So what could be going on?   It’s hard to tell.   Women clearly aren’t using this as overnight storage for their business capital.   But they are saving up significant chunks of cash and putting them in the bank.   Given that there is a negative real interest rate, this is clearly about protecting the money from something.   Is it themselves or is it pressure to share?