Borrowing for a rainy day: emergency loans in Bangladesh

Blog post by a student on the job market.

Weather shocks are a constant and growing threat to much of the world’s rural population whose livelihoods depend on agriculture (Dercon, 2002). The cost of being exposed to these shocks is high: households sell productive assets or reduce spending on essential goods and services that can have substantial negative long-run consequences on household wellbeing. Moreover, households often adopt agricultural production processes that are less risky but also less productive in order to limit their exposure to these types of shocks (Janzen and Carter, 2018). Unfortunately, it has proved challenging to develop financial tools that reduce exposure this risk. Traditional insurance is often absent in developing countries because of moral hazard and adverse selection. Furthermore, weather-index insurance, which was designed to help farmers increase their resilience to extreme weather events, has suffered from low demand (Cole and Xiong, 2017).

In my job market paper, I investigate the effectiveness of a new financial product that I designed in collaboration with BRAC, Bangladesh to help households manage this risk. The product, marketed as the “Emergency Loan”, is a pre-approved loan that is made available to households after a flood. However, eligible households are made aware of this opportunity before the flood season occurs. They know ahead of time that they will have access to additional funds to cover any losses incurred from a flood, and they can invest in new activities that will generate additional income.

This guaranteed credit line offers a number of benefits relative to traditional insurance products. First, it provides risk-coverage to eligible households without requiring any up-front purchases. Households can make productive investments in the planting period because they know they have the option of taking the loan should a flood occur. This by-passes the demand problem that has limited the appeal of traditional insurance products (households do not want to pay a certain premium up-front for an uncertain payoff in the future (Casaburi and Willis, 2018)). Second, households can assess their own needs and the alternative options that may be available to them after the shock (e.g. assistance from relatives) before they choose whether or not to incur the cost of loan repayments.

We tested the impact of the Emergency Loan using a randomized control trial with a sample of 200 BRAC branches serving approximately 300,000 micro-finance clients spanning the entire country. Eligible clients in treatment branches were informed that they were pre-approved for a loan equal to 50% of their last previously approved loan from BRAC. The loan would be made available to them if a flood hit in their branch service area.

Household benefits pre and post flood

I examine two main categories of changes that result from the offer of guaranteed credit. First, I examine how households’ decisions change in the planting season before flooding occurs. Differences between treatment and control in this period will be driven by a reduction in their perceived exposure to risk, which leads them to re-optimize their investments. Second, I examine outcomes for households at harvest in both flooded and non-flooded areas. Here, differences between treatment and control households come from 1) the different payoffs they can expect as a result of the decisions they made during the planting season; and 2) differential access to credit arising from the availability of the Emergency Loan in flooded areas.

In the planting season, I find that informing households that they were pre-approved for credit:

• Increases the amount of land dedicated to agricultural cultivation by 15%
• Increases non-agriculture business investments by 30%
• Effects are concentrated among the most risk-averse households

This suggests that the guaranteed credit line reduces exposure to flood risk, and households respond by increasing investment. After harvest, treatment households:

• Have 33% higher crop production when there is no flood.
• Spend 10% more on food and own 18% more livestock when flooding does occur

Sustainability for the lender

These first results suggest that the Emergency Loan helped households better manage flood risk. However, for this product to be a viable risk-management solution, it must also be sustainable for the lender (from the lender’s perspective, extending loans to households just hit by a shock is risky). Using BRAC’s administrative data, I find that extending guaranteed credit to clients improves overall MFI performance in treatment branches. Borrowers with access to the Emergency Loan improve their overall repayment rates by four percentage points relative to the control group, especially after a flood. Branch profits increase by one dollar per-person, with the largest increases in profits coming from “marginal” clients – i.e. clients that just barely qualified for the loan. This result suggests that BRAC could have been less selective in targeting clients for the Emergency Loan while still improving their returns.

Policy Implications

This research suggests that guaranteed credit can be a useful tool to address uninsured risk in places where traditional insurance markets have failed. Guaranteed credit is an appealing solution because microcredit is already used in many locations. Furthermore, because the product does not require any up-front payments, eligible households can benefit from the guaranteed offer and improve ex-ante investments - even if they ultimately choose not to take the product after the shock. This expands the potential number of beneficiaries beyond those who normally opt-in to insurance products.

However, there is still an open question about whether the results documented in Bangladesh will replicate in other settings. First, my partner BRAC has been successfully operating in Bangladesh for many years and is highly trusted by their clients. In settings without as much trust, clients may be less likely to respond to an MFI’s guarantee. Furthermore, in settings with lower baseline loan repayment rates, the overall effect on the lender’s performance could be significantly worse than what BRAC’s experienced. Finally, guaranteed credit may not be helpful for very large or long-lasting shocks. Rebuilding assets in these cases is difficult, which makes subsequent loan repayments infeasible. Therefore, it is necessary to test guaranteed credit products in other settings before advocating for widespread adoption.