Who even needs a pound of flesh?
Improving loan repayment rates using behavioral science
“Why, I am sure, if he forfeit, thou wilt not take his flesh — what’s that good for?”
-Salarino, a bystander in the play
The antagonist in Shakespeare’s Merchant of Venice is a Jewish moneylender called Shylock. In the story Antonio — the merchant of Venice — takes a loan from Shylock for his business. The collateral for the loan is a pound of his own flesh from whichever part of his body pleases Shylock (all in good sport, goes the quote). Antonio’s business doesn’t do well and he doesn’t pay up. This sets the background for what develops into a complex and rich discussion on greed and the nature of villainy that is the play.
Credit is like salt. Too little in your economy and it’s bland, without much action and uninspiring. Too much and you could ruin everything like we saw in the 2008 meltdown. The thing is, credit is only as good as the creditor — which is to say, credit only has value if paid, something. It’s not uncommon for loan recipients to struggle with understanding and living up to the terms and conditions of their loans, as Shylock must have known. This is why financial institutions perform comprehensive investigations of potential clients’ anticipated credit repayment behavior.
Several studies on credit repayment behavior have found that credit behavior generally revolves around personal and environmental factors that might influence borrowers’ attitudes on repayments. This was no different for Nigeria’s Government Enterprise Empowerment Program. This social investment program was designed to ensure beneficiaries adopt digital financial services through low interest, short term microloans — and it was experiencing challenges with repayment.
As part of the first Policy Innovation Unit in Nigeria — which had a clear mandate to support the delivery of social protection programs — Busara designed a set of behavioral interventions to improve loan repayment, which we termed the color categorization intervention.
This intervention sought to test the effect of social comparison and repercussive SMS prompts on repayment among GEEP beneficiaries. The SMSes were sent to beneficiaries based on their assignment to different color categories (which was based on their repayment behavior).
For the intervention, we sampled 60,000 beneficiaries from a segment pool which is referred to as the ‘antagonist group’ by GEEP. These are customers who have not made a single repayment since they received their loans. As such, the color allocation criteria was primarily hinged on how far along the loan tenor the beneficiaries were, since all customers were in default.
We split each color group into three with one group serving as the control group. The other two groups were sent two different behaviorally framed SMS notifications. The first purely communicated the color categories of beneficiaries, while the other incorporated a ‘consequence’ element. This was done to enable us assess if any observed response to the SMSes was simply as a result of the colors in themselves and what they connote socially, or if it was bolstered by the fear of a negative consequence.
The SMSes were sent to beneficiaries, based on their treatment assignment, every week for a period of 6 weeks. During this period, beneficiaries’ repayment records were closely monitored and their color categories were regularly updated based on their repayment performance.
What happened?
While both SMS treatments saw an improvement in repayment, the pure color SMS treatment did better.
Overall, the pure color SMS treatment yielded the highest repayment over the 6 weeks testing period. All individuals in this treatment arm made a combined payment that was 60% higher than the combined payment from the control arm. This was primarily driven by the orange color category, where we observed a 225% increase in repayment in pure color SMS treatment relative to the control. Interestingly, most beneficiaries who made a payment in this category made a partial payment, not necessarily the full amount owed.
However, despite the improvement in repayment in each SMS treatment, the majority of color movements were negative
Most beneficiaries regressed during the course of intervention into worse colors: Red and Black. This suggests that while in relative terms (to the control group), the intervention was found to have positive associative effects, the majority of beneficiaries in each treatment arm were unresponsive to the social/relative ranking color SMS.
These results suggested two main courses of action. First, expand the Orange color category to accommodate more beneficiaries: Beneficiaries were found to be more receptive to the color categorization messages — particularly the pure color SMS for the Orange category. We believe the relatively strong effect of the Orange color was primarily as a result of the positioning of the Color — close to Red (danger) but just as close to (safety) — which induced some degree of risk compensation among beneficiaries. As such, it may be worthwhile to leverage this underlying mechanism by expanding the criteria for the Orange color to enable it to accommodate more beneficiaries, including those outside the antagonist group.
Second, explore the use of commitment devices in making salient the need for repayment: Despite the noticeable improvements in payments in the SMS treatment arms, the overall repayment rate within the entire sample was surprisingly low (0.2%). Given this, it may be necessary to explore the use of commitment devices, such as simple, personalized repayment calendars which beneficiaries are required to sign upon loan disbursement and over the course of repayment. Devices that require personalized action, such as these, tend to be effective in driving salience.
