Continuing on the microfinance theme from last week, I came across a paper by Mahmoud El-Gamal, Mohamed El-Komi, Dean Karlan and Adam Osman which developed and tested a Grameen model and a "bank-insured RoSCA" in Egypt to see whether one was superior to the other in attracting clients and ensuring repayment. The bank-insured RoSCA (rotating savings and credit association) is a slightly modified form of the traditional RoSCA where each participation contributes regularly and each period, one member receives the contribution. In the modified form, the bank will collect a premium from the participants and if one of them does not pay the contribution, the bank will step in and make the payment and that participant will owe the bank. It is viewed by the authors as a simplified credit union.
Their test was focused on determining whether the RoSCA would lead to better, worse or the same take-up by participants, as well as whether it would lead to more frequent repayment. To do so, they used real money in an experimental setting with participants who shared the same demographics as a typical microfinance client (in a country with a large enough Muslim population that a portion would be expected to turn down the Grameen loan because it required interest payments). However, they did not include any factors that would affect the return on the investment that participants theoretically made (the investment was predetermined to have a certain return).
The reason for the higher frequency of repayment in the RoSCA model is that the loans between participants is interest-free. It will not benefit the participant to renege when the social cost of default is more than the amount of the loan (in the Grameen model, that tipping point is higher, at the amount of the loan plus interest). In addition, when the social cost of default (which applies to the single defaulter alone) is lower than the loan amount, the bank (which guarantees the RoSCA, and to whom the defaulter becomes a debtor) can raise the cost of the penalty (which applies to both participants) to a level where default by a participant no longer becomes attractive.
In practice, this was borne out in their field experiments. The RoSCA attracted greater levels of participation and also had lower probability of default, which is what the game theoretic model predicted. It creates one data point, albeit under simplified assumptions, for how a RoSCA could be used as an alternative to the Grameen model of microfinance, and perhaps even expanded in the future into larger, more formal financial institutions similar to credit unions. It will be interesting to see what research develops out of this paper--either from the co-authors or from other researchers.