Risk transfer versus cost reduction on two-sided microfinance platforms

Crowdfunding is a specific type of crowdsourcing, a general term that implies a group of people is essential in accomplishing the end goal but does not require the contributions to be monetary in nature.

Source: “Crowdfunding Industry Report: Market Trends, Composition and Crowdfunding Platforms,” 2012, https://​www.​crowdsourcing.​org.

Despite the high interest rates associated with microfinance, the average default rates (across FPs and time) are relatively low in our data (below 2%). However, the delinquency rates are fairly high at the level of about 11%. The distribution of default/delinquency rates, however, is extremely skewed. In some cases the default rates exceed 20% and the maximum is greater than 50%. The delinquency rates reach 100% for some FPs (see Table 1). One possible reason for the relatively low default rates is that other disciplining devices may keep interest rates (and thus default rates) in check, including group liability, dynamic incentives (if borrowers plan to borrow in the future), high default costs associated with reputation effects and/or public shaming, and mechanisms of repayment that often involve frequent, small payments and loan collection occurring in group meetings (Banerjee 2013). Additional disciplining mechanisms which may keep interest rates low include moral hazard on the borrower side if administration costs increase with interest rates (Banerjee and Duflo 2010), and FP reputation regarding social outcomes. Finally, because we are in a setting in which FPs have market power (unlike in many finance settings), the demand elasticity itself can serve as the disciplinary device; indeed, in the rare situations in which interest rates are low, such as the case of the Bolivian microfinance industry, the market is highly competitive - more competition increases demand elasticity, which increases the downward force on interest rates.

Zidisha is the only microfinance platform of which we are aware that connects borrowers directly to lenders without intermediaries. According to their website (https://​www.​zidisha.​org/​why-zidisha), “Zidisha’s direct person-to-person connection results in far lower cost for the entrepreneurs, and a more transparent and interesting experience for lenders. Our first-of-its-kind direct lending platform ensures that the profits go to the entrepreneurs (no cut of it goes to a bank or other intermediary).”

Hulme (2006) provide a nice history of social lending on the internet.

http://​www.​pbs.​org/​frontlineworld/​stories/​uganda601/​video_​index.​html

Factors that alter the FPs’ expectation of loans being funded through Kiva, such as social and earned media exposures studied in Stephen and Galak (2012), will influence FPs’ decisions to provide loans.

The optimal levels can be obtained by maximizing the Lagrangian, \(\mathcal {L}=\log \alpha +\beta _{1}\log K_{0}+\beta _{2}\log L+\lambda (B-K_{0}p_{K}-Lp_{L})\).

We suppress the subscript for indexing FPs to avoid cluttering the notation.

Zidisha started its operations in October 2009.

The overlapping 10 countries and areas are Burkina Faso, Ghana, Guinea, Haiti, Indonesia, Kenya, Mexico, Niger, Senegal, and Zambia.

Note that this percentage of defaulted loans and the delinquency rate are different from the statistics presented in Table 2. The reason is simply that the statistics in these two Tables have different levels of aggregation that lead to rounding differences.

We note that the average APR across all loans is slightly different from the average APR across all FPs as reported in Table 2. This difference is due to differences in weighting when computing the average APRs.

We note that the high default rates of the Middle East may be due to its extensive political instability in the past decade.

We depict another figure only focusing on those Kiva loans issued when Zidisha operates (not reported), which shows similar insights.

If costs C(r) also increase with r, another disciplining mechanism would exist, but we have no cost data to test this relationship. However, this effect is documented in Banerjee and Duflo (2010).

Note that we also consider the same analyses using default rates instead of delinquency rates to measure loan risk and the insights remain the same.

One possible factor that may invalidate Zidisha’s entry as an instrument is if the platform chose countries with more low-risk borrowers to enter, hence the lower APR was the reason of Zidisha’s entry. We think this mechanism is unlikely because Zidisha had fairly high delinquency rates and consequently had to stop its operation for a brief time. Also, we control country and FP fixed effects. The within-FP-country variation related to Zidisha’s entry can further mitigate such a concern. Furthermore, in the appendix, we produce Table 6 to compare Kiva loan characteristics before and after the entries of Zidisha, showing that the loans have very similar characteristics. The similarity of loans implies the FPs tend to use the same screening criteria for borrowers before and after Zidisha, which leads to similar pools of loans.

We find the same using the regression discontinuity approach.

Another potential reason why a loan’s APR has inconsequential effect on the funding probability is that Kiva vaguely labeled the APR as the field partner’s “portfolio yield.” It might be difficult for lenders to perceive it as the interests and fees collected by the FP from borrowers. Recently, Kiva has changed the label to a more transparent “average cost to borrower.”

A full understanding of lender incentives is tangential to this paper’s goals, and there has been significant work in understanding these decisions in microfinance and crowdfunding applications (Galak et al. 2011; Agrawal et al. 2011; Stephen and Galak 2012; Kawai et al. 2014).

An alternative specification with the number of loans as the dependent variable gives similar results.

The rates are equivalent at the FP’s indifference point, which is intuitive since the only remaining force altering rates is due to the discrepancy in lending costs.