This week I would like to explore more something I saw during my recent visit to Ghana. As I explained in a previous post , a conversation with a rural bank manager made me realize that in Ghana, just like in the United States, people take payday loans.
For those who are unfamiliar with this type of credit, this is basically a short-term, high interest loan secured against an upcoming paycheck. The bank manager in Ghana explained that while this was quite pricey credit, clients regarded this as a boon since it was cheaper than any other option they had for short term credit. At the same time, the conversation drifted to a discussion of why people tend to take these loans one after another in what might appear to be either irrational (it would be much better financially to save) or desperate behavior.
A colleague alerted me to an interesting forthcoming paper  by Marianne Bertrand and Adair Morse where they tackle this issue by running a set of experiments to tease out possible behavioral remedies for persistent payday borrowing.
So why is this important, at least in the US context? According to Bertrand and Morse, in 2007 Americans paid an estimated $8 billion to borrow $50 billion. Just to hammer home how expensive this source of credit is: the way the loans work is that borrowers have to pay $15-17 for every $100 they borrow by the time they get their next check, which for most is every two weeks (interesting side note: apparently the use of fixed fees instead of interest rate means that usury laws don’t bite). Thus, the implied APR on this is over 400%. The price for this type of loans I saw in Ghana was significantly less – but only because the pay period in Ghana is monthly – the interest rate there was around 15-18% due by the next month.
So starting from the point that if individuals are making a fully informed, utility-maximizing choice, providing information shouldn’t change things, Bertrand and Morse test this with four different interventions:
1. The APR treatment. Now the borrowers are already informed about the APR on the loan forms they get. But in this intervention, that information is provided again (443%) and compared to other sources of credit like credit cards, car loans and subprime mortgages (which, in case you were wondering, clock in at 10%). The idea here is to help borrowers “anchor” the APR for payday loans.
2. The Dollar Adding-up treatment. This puts the costs into dollar terms, comparing the cost of using a credit card to finance $300 from 2 weeks ($2.50) to 3 months ($15) relative to that for a payday loan ($45 for 2 weeks, and $270 for 3 months). This treatment hits on a number of ideas, including the aptly named “peanuts effect” which argues that that we don’t do well thinking about small numbers that could add up a lot over time.
3. The Refinancing treatment. This gives a graphical representation of the average number of times a payday loan is refinanced before it is paid back (on average only 25% pay back these loans without refinancing). The idea here is help borrowers think realistically about the future (and unlike options 1 and 2 it’s giving them information they didn’t already have).
4. The Savings Planner treatment which gives them a chart they can fill out with expenses they might be able to cut in order to save. In addition to being administered to folks who get none of the above, the savings planner is also combined with the above interventions since it’s potentially a complement.
Bertrand and Morse manage to get a large national payday lender to implement these treatments in its stores across 11 states. Given the complexity of the logistics involved, they randomized at the store-day level from May-September 2008. Among borrowers approached, they get a 21% take-up.
In addition to a couple of quick questions at the point of enrollment in the study, they also conducted a follow-up, more in-depth phone survey. The take-up rate for this is dismal – despite hiring a firm that had experience with this demographic, they only get a 15% response rate. This doesn’t allow them to combine this survey data with the experimental data, but some of the survey results help motivate this experiment further. First, when asked about 40% of the respondents are close to the real APR (>400%), but there is another mass of respondents who say it is around 17%. Second, when asked about the dollar fee for the loans, about 30% get it right, but most of the folks limit the dollar fees to what they would pay in one cycle (recall that on average they will refinance multiple times). Third, respondents are asked their expectations on how long it takes people to pay back these loans. The mean is about right (5-6 weeks based on historical evidence), but there is a lot of dispersion.
OK, so what do they find when these treatments are administered? In short: the information matters (and the savings planner does not – either alone or in combination). The Dollar Adding-up treatment results in an 11% decline relative to the control group in the likelihood that individuals take a loan. In addition, individuals in all three groups lower the amount that they borrow. The Dollar treatment results in a 23% decline, the APR in a 16% decline and the Refinancing treatment results in a 12% decline (all relative to the controls). They sum it up nicely: “our results suggest that information disclosure that is inspired by, and tries to respond to, the specific cognitive biases and limitations that may surround the payday borrowing decision has a non-trivial impact on some individuals’ decisions of whether or not to take on a payday loan.”
There is also some interesting heterogeneity in the results. The Dollar information treatment is significantly less effective in reducing borrowing among more highly educated borrowers. In addition, the Dollar treatment seems to have more of an effect on folks with higher (self-reported) self control, but the APR treatment seems to have more impact on those with lower self-control…Pretty interesting results and definitely fodder for further work.
Finally, since they also observe these borrowers over a number of cycles, Bertrand and Morse also look at the dynamics of the impacts (although they are limited to a couple of months on the post-intervention time-frame). They find that the effects aren’t immediate – it seems to take folks at least one pay cycle to get some savings together to move into a better equilibrium. And once there, they don’t revert, even though the information was a once-off intervention.
In the end, these results give us some food for thought on how policy can target specific behaviors to help “nudge” folks to a better financial equilibrium. Other examples and thoughts are most welcome.