Improved Personal Identification Flashcards
Author of the paper?
Giné et al (2012).
The experiment:
RURAL MALAWI.
Randomised fingerprinting of loan applicants to test the impact of improved personal identification. Rural Malawi characterised by an imperfect identification system and limited access to credit.
How does fingerprinting lower adverse selection?
Riskier individuals that would other default may now take out smaller loans (or not borrow altogether) to ensure access to the credit market in the future.
How does fingerprinting lower moral hazard?
Borrowers may have a greater incentive to ensure that agricultural production is successful either by exerting more effort or diverting fewer resources away from production.
Effects of fingerprinting:
Substantially higher repayment rates for the subgroup of borrowers with the highest ex ante default risk.
No impact on the repayment rates for the group with the lowest default risk.
Farmers chose smaller loan sizes - lower adverse selection.
Higher-default risk farmers divert fewer inputs away from the contracted crop (paprika) - lower moral hazard.
Actions of loan officers:
Approval decisions and monitoring of lending clubs by loan officers did not differ across treated and control groups.
Borrowers with relatively low probability of success:
Most affected by the intro of dynamic incentives.
Borrow the lower amount and invest it all in cash crop production (complete opposite without dynamic incentives).
On average, how much lower were the loan sizes of fingerprinted borrowers?
MK693 lower.
How much lower were the loan sizes of the lowest quintile of expected repayment?
MK2,657 lower.
Consistent with the theoretical model’s prediction that the ‘worst’ borrowers will respond to the fingerprinting by voluntarily reducing their loan sizes.
Overall loan pool therefore less weighted toward low-quality borrowers.
Lowest quintile borrowers: the increase in eventual fraction paid?
32.7% point increase.
Lowest quintile borrowers: the increase in the likelihood of eventually being repaid?
40.8% point increase.
What is the increase in paid inputs by the lowest quintile borrowers?
MK6.566 more paid inputs - almost 90% more than the equivalent unfingerprinted group.
Why can’t we make concrete predictions about the effect fingerprinting has on revenues and profits?
The imprecision of the estimates prevents us.
Are there problems of selection bias in the borrowing pool in response to fingerprinting?
Perhaps, BUT… full-sample regression results are very similar to those form the borrower-only regressions. Therefore, selection into borrowing is not driven by treatment effect estimates.
Could there be idiosyncratic differences in the repayment rates between treatment and control groups?
Perhaps, BUT… robustment checks show that the main results are not mechanically driven by idiosyncratically low repayment rates among some control farmers in certain localities.
