Do Peer-to-Peer Lenders Understand Risk?

Picture yourself in Kenya. You have your first good job. For the first time, you have a little spare money. You might be looking for somewhere to invest it, and you might decide to lend it out.

Traditionally, lending between people in emerging markets like Kenya has been strictly informal. It is usually done in cash with little or no documentation and is generally limited to a close network of people in a lender’s community. That could be changing fast. New peer-to-peer lending platforms allow consumers to lend via mobile money and to do so in very small amounts. When these loans are intermediated by a peer-to-peer lending platform, a third party handles credit scoring, loan origination and collections, and shares a portion of the profits with the lender.

Man and woman in Nairobi
Man and woman in Nairobi. Photo by Sarah Farhat, World Bank.

The proliferation of new lenders on these platforms makes it important to ensure people understand the risks associated with lending. So how well do Kenyans who use these platforms understand investment risk? CGAP and the Busara Center for Behavioral Economics conducted a laboratory experiment with 148 participants in Nairobi to explore this question. The answer suggests that providers need to rethink how they ensure their customers are informed investors.

How well do peer-to-peer platform lenders in Kenya understand risk and uncertainty?

It is well established that few people understand risk. Institutions and actuaries may be comfortable talking in terms of percentage probabilities, but even the best-educated lay people can struggle to comprehend risk in those terms. College-educated people repeatedly fail tests on their ability to calculate probability in percentages. In our experiment, participants (University of Nairobi students) were presented with three risk investments and asked to select the option with the highest expected payoff. What we found was that they selected the correct answer only 30 percent of the time — lower than the 33 percent success rate one would expect for random guessing.

Past research suggests a couple of promising techniques to promote better risk comprehension. Risks can be expressed as natural frequencies (i.e., a 1 in 10 chance). They can also be expressed as icon arrays, in which a proportion of a 100-box grid is shaded (see example below). Unfortunately, in our experiment neither technique helped the participants improve their results. Performance in calculating risk was equally low across treatments.

Icon array chart
Icon array chart

Participants’ difficulties with math and unfamiliarity with formal investment decision scenarios — both factors that apply in the real world — may have contributed to these numbers. Part of the explanation may also lie in people’s limited understanding of risk as a standalone concept. When asked about the most important things to consider when lending money to a stranger, just 16 percent of the students discussed risk directly or mentioned traditional metrics for risk like creditworthiness, financial history or collateral. Notably, these participants performed better than others at the quantitative questions.

Many more (36 percent) discussed risk indirectly, citing factors such as the purpose of the loan and the borrower’s occupational status that could be used to estimate a borrower’s ability to repay. But 48 percent did not discuss risk at all. These participants’ top priority was obtaining basic personal information, such as address and contact information, that they could use to find the borrower in the future but that did not concern risk. 

Chart: Risk perception

This new class of investors will need more than passive disclosures of risk

All this suggests a number of challenges for enhancing consumer protection as more people become investors by joining peer-to-peer lending platforms. Our experiment shows that tweaks to the way traditional information is presented are unlikely to be sufficient in explaining formal financial risk to someone accustomed to informal lending to family and friends. Simply changing “10 percent probability of default” to a natural frequency (“1 in 10 borrowers in this group is not expected to repay their loan”), for example, might not be enough. Providers will likely need to do more than update passive disclosures if they want to empower and inform new lenders.

Peer-to-peer lending platforms could consider providing lenders with the type of personal information that interested our participants in addition to specific risk factors and estimated probability of default. This might help communicate risk in a more tangible way than financial risk factors alone for people used to informal lending. However, providing this type of information opens up challenges around discrimination and prejudice. Also, if lenders are interested information that has not proven to be a good predictor of risk, should platforms enable them to decide based on these factors? Or would doing so hurt lenders?

Digital peer-to-peer lending platforms could be useful for people who are looking to invest in small amounts but cannot typically access traditional investment options like stocks or bonds. Yet a new approach to disclosure and oversight might be needed for these investors to be protected. This approach should be rooted in an understanding of how people actually comprehend and behave in the face of risk. Simply providing traditional financial risk metrics does not seem to be enough. 

This blog is available in Mandarin on the World Bank's Voices blog.


21 April 2017 Submitted by Norbert Mumba (not verified)

May be to suggest that concerning one self with address of a debtor is not risk is either missing the point about risk or expecting risk to fit in academic defined categories. Failure to locate a debtor in the context of low value peer to peer lending is perhaps one of the most significant risk in the context of time value of recovering debt

28 April 2017 Submitted by Maria Fernandez... (not verified)

I definitely agree that being able to locate the debtor is a key factor in being able to get the money back. It makes complete sense that this was top of mind for the respondents. We should probably have referred to quantifiable measures of risk, or something along those lines, instead of just risk. We were looking for a metric that could be standardized to help select from different investments in a platform: if people lend to other people they don’t know, what would be the best way to convey the risk associated to each potential borrower, so that the lender can compare to the payoff and decide between borrower A or borrower B in an informed way?

Add new comment