Credit Lines as Insurance: Evidence from Bangladesh
Lane, Gregory. 2020. "Credit Lines as Insurance: Evidence from Bangladesh." Working paper. | Link
The author estimates that the availability of guaranteed post-disaster credit induced agricultural households to undertake more productive investments ex-ante and may have helped reduce the impact of a flood ex-post.
The experiment was conducted in 2016 and 2017 in Bangladesh through a partnership with BRAC, a micro-finance organization. Emergency loans for flood-affected agricultural households were guaranteed in half of 200 BRAC branches in flood-prone parts of Bangladesh (areas bordering major rivers that had experienced flooding historically). Micro-finance clients in treated areas were contacted one month before planting and informed about guaranteed loan availability, allowing them to change cultivation decisions even before any flooding occurred.
Clients were eligible for an Emergency Loan if they had a sufficiently high credit score, created by a weighted index of past repayment behavior that was found to predict loan default. About 40% of clients had credit scores that made them eligible for an Emergency Loan. Eligible clients were slightly older and had slightly lower levels of annual income, savings, and livestock than ineligible clients. The maximum amount eligible clients could borrow was 50 percent of their most recent loan.
A flood event was defined as (a) a government-maintained river gauge showing water levels above a certain level for at least one day and (b) a BRAC employee determining that at least 20 percent of the area served by the branch had experienced flooding. Once a flood event has occurred, all eligible borrowers in the service area could request the Emergency Loan, regardless of whether the borrower experienced flooding herself.
The Emergency Loan was issued at a 25 percent interest rate and had to be repaid within a year. These terms are similar to the most common loan issued by BRAC. In 2016, 40 control and 51 treatment areas experienced a flood event as defined above, although in most cases the flooding did not cause substantial damage. In 2017, 73 control and 63 treatment areas experienced such a flood event, and the flooding was in general more severe.
Administrative data and a survey of 4000 borrowers and 800 BRAC staff was used to measure outcomes. Both eligible and ineligible borrowers were surveyed. Takeup of the Emergency Loan was 2.9 percent in 2016 and 5.4 percent in 2017.
Eligible borrowers to whom the availability of the Emergency Loan was randomly guaranteed (“treated households”) increased the amount of land they rented by 30 percent. Total land devoted to agricultural production increased by 15 percent. Loan availability increased the probability of cultivating any crops by 10 percent. Non-agricultural business investment increased by 30 percent, but the increase was only marginally significant. Crop production among treated households increased by 17 percent, and per-capita consumption increased by 8 percent. Value of business stock increased by 23 percent. Input use (fertilizer and pesticides) and input expenditure per acre did not change.
Comparing treated and control households in areas that did not experience flooding, Lane shows that the availability of the emergency loan increased crop production by 33 percent; there was no significant difference in consumption. This finding shows that more efficient ex-ante investment is an important channel through which the guaranteed Emergency Loan improved outcomes.
Comparing treated and control households in areas that experienced flooding, Lane shows that treated households lose 90 percent of the production gains they experience from the availability of the loan, but have consumption that is 10 percent higher than flooded control households.
Finally, Lane estimates that offering the Emergency Loan was overall profitable for the lender.