Reassessing Microcredit Impact: Uncovering Limitations of Randomized Controlled Trials

In my previous article, I provided a summary of 6 RCTs that found microcredits were not as transformative as assumed by several development practitioners in the 2000s. However, a relatively recent study published by New York University Professor stated that the RCTs failed to answer key questions on the impact of microcredit, such as its overall contribution to the economic and social lives of its customers. This claim is significant, considering that RCTs are known as the Gold Standard for measuring impact.

The professor highlighted this primarily because the RCTs focused only on newly targeted customers, leaving out long-term customers who were already using microcredit before the studies started. This oversight made it difficult to estimate the impact of microcredit on all customers.

While stating this, the professor also pointed out that the studies were well-designed but faced the standard issues when dealing with RCTs in terms of execution suitability. The randomization for the study was conducted in the order in which microcredit lenders entered new locations. The researchers didn't choose specific places deliberately; instead, they let chance decide the order of expansion.

The approach might not have captured the average impact of microcredit in those areas due to a phenomenon known as "self-selection bias." Self-selection bias occurs when individuals or groups choose to participate in a program or intervention based on their preferences, characteristics, or circumstances. In our context, it means that the new customers who decided to take microcredit loans might have different characteristics from the first ones who used the services before the randomization began.

This could be an issue for several reasons:

💡 Differences in Borrowers: The new customers who signed up for microcredit after the lenders entered new locations might have different financial needs, and business ideas compared to the first customers who chose to use microcredit before the randomization.

💡 Risk Tolerance: People with varying levels of risk tolerance might self-select into microcredit. For instance, those who are more risk-averse may have been hesitant to join the program early on but later decided to participate when they saw positive outcomes from others. This could influence the impact of microcredit on the new customers compared to the first customers.

💡 Entrepreneurial Experience: Early adopters of microcredit might have been more experienced entrepreneurs, while later adopters might include more individuals with less business experience. This difference could affect their businesses.

As a result, it becomes challenging to generalize the findings to the entire population or customer base. While the approach using randomization is useful for studying the specific impact on the treated locations, it might not fully capture the overall average impact of microcredit in those areas.