AltFi Data Analytics enables us to track the performance of origination cohorts in near real time. The good news is that we are yet to see a marked worsening of credit performance within the sector. However, in the light of the seemingly relaxed attitude that RateSetter investors have to the default performance of the platform, we have taken a closer look at the contingency funds of the two largest platforms that operate them. This has undoubtedly become more pertinent in the light of Brexit which, one would imagine, may result in an increase in defaults.
Thanks to the P2PFA, all member platforms provide standard disclosures covering default rate, and expected default rate, by annual cohort as can be seen in the table in Figure 1. The first thing that strikes us when we look at RateSetter’s figures is the fact that the expected default rate for 2014, 2015 and 2016 lending cohorts is the same – 2.28%. This is particularly surprising when one considers the change in lending mix on the RateSetter platform over the last 36 months – proportionally more lending to business and more secured lending. How has RateSetter been able to create portfolios with exactly the same risk profile across the last three annual cohorts given this change in mix?
These expected default rates are dynamic and we note that this rate has increased of late having been at 2.21% for the 2014, 2015 and 2016 cohorts at the end of April 2016.
Actual Bad Debt
RateSetter Expected Bad Debt
AltFi Data Extrapolated Bad Debt
Extrapolated Bad Debt/Estimated Bad Debt
Figure 1: RateSetter Loan performance table
The level of actual defaults (or bad debt, as they appear to be one and the same here) is above the expected default rate for the 2014 cohort. From Figure 1, we can see that there is still 22.7% of the cohort outstanding. We can therefore expect the bad debt figure to climb further. Extrapolating the default rate until all of the cohort has been repaid, by fitting to a default curve, gets us to a total extrapolated default rate for the 2014 cohort of 3.21%. That’s 1.41x expected bad debt. Using the same curve, the 2015 cohort looks to be heading for a similar overshoot with 3.18% total bad debt i.e. 1.39x expected bad debt. It is too early to perform any meaningful analysis on the 2016 cohort.
The probability of default on a loan begins to reduce once the loan has been outstanding beyond a certain period of time. This is illustrated by the chart in figure 2 which shows historical bad debt curves. On inspection it is clear that the gradient of the bad debt curve flattens significantly after 24 months. Determining the precise shape of this curve is more of an art than a science, however in our extrapolation of the cohort default rate for loans we have tried to fit our assumptions to these curves.
Figure 2: Industry aggregate annual cohort bad debt curves, screenshot taken from AltFi Data Analytics
However one finesses the assumptions it is clear that actual defaults are over shooting expected defaults. But given the presence of a contingency fund do we have anything to worry about? From figure 3 we can see that the 2014 origination cohort has 78.24% bad debt fund usage and 77.28% principal realised. The same figures for 2015 are 51.52% and 49.20% respectively. This means that if bad debts continue at the same rate, then both cohorts will take more from the provision fund than they contributed. However, as discussed above, bad debt rates are likely to trail off and the eventual utilisation may well be below 100% for both cohorts, all things being equal.
Having spoken to RateSetter today, they tell us that they are comfortable with the provision fund’s coverage and believe that more will be paid into the provision fund than is taken out to fund bad debts in each annual origination cohort. This for the below reasons:
Reduction in probability of default for seasoned loans
Income from recoveries and loans in arrangement coming back into the fund
Continual payments into the provision fund by borrowers (as described below)
There is a small buffer in the provision fund. RateSetter’s website tells us that the provision fund is currently 1.25x covered on expected defaults albeit note that this has fallen from 1.63x covered in September of last year. However, that buffer is going to be pushed to the limit. We extrapolate that bad debt performance could be 1.39x and 1.41x expected defaults for the 2014 and 2015 cohorts respectively. If our extrapolations prove correct then the provision fund is very close to becoming depleted. And that’s without factoring in any worsening in credit conditions that the Brexit uncertainty may precipitate.
Figure 3: Screenshot of RateSetter returns performance
Lending is a risky business and at certain points in the credit cycle bad debt rates increase and returns fall. For a platform that does not operate a contingency fund, a fall in returns, whilst unwelcome, would not cause any structural changes for investors. However, for those platforms operating contingency funds, the fund becoming depleted has knock on effects. In RateSetter’s case this could cause a Resolution Event – described on RateSetter’s website: “This would mean that all outstanding loan contracts would be automatically assigned to the Provision Fund, and all loan repayments would be collected by the Provision Fund on behalf of investors. Repayments would then be shared out (pro rata) to investors to ensure diversification of default risk. There would be a material delay in repayments being made.” – i.e. RateSetter closes for new business and effectively becomes a collective investment scheme.
Of course, RateSetter is monitoring the provision fund and loan portfolio performance on a daily basis and its website tells us that the platform would take steps to shore up the fund:
“RateSetter actively manages the Provision Fund, by determining how much borrowers pay into it. If the Provision Fund was falling significantly in value, it’s likely that RateSetter would increase contributions into the Provision Fund. This is not reflected in the chart.
Borrowers pay into the Provision Fund regularly, over the lifetime of the loans. This ongoing contribution into the Provision Fund is not reflected in the chart.”
Increasing the size of the contributions to the fund, thereby reducing the returns of new investors to boost the returns of existing investors, might be frowned upon.
It is important to point out that we believe that RateSetter lenders are currently a long way from losing capital as the graphic taken from RateSetter’s website in figure 4 demonstrates. However, the provision fund looks to be on less firm ground.
Figure 4: RateSetter graphic showing what happens in the event that losses exceed expectations.
Zopa has a reputation for conservative lending. It is the only UK platform that was around during the 2008/2009 credit crunch and whilst bad debt rates picked up in that period, gross lending rates also increased and investor returns suffered only marginally. Net returns, as shown in figure 5 below, remained healthily positive over the period. In fact, the low point, according to our Liberum AltFi Returns Index methodology, was just under 5%.
Figure 5: Liberum AltFi Returns Index – the line prior to 2010 is exclusively Zopa returns as Zopa was the only platform in existence up until that point.
Zopa also operates a contingency fund - known as its Safeguard Fund - which protects investors choosing the classic or access products. Historically Zopa’s loan cohorts have undershot their expected default rates. However, in the past few years the actual default rate has tracked much closer to the expected default rate. This is likely due to adjustments in the credit model rather than any deterioration in credit or underwriting quality.
Indeed, in 2014 and 2015, using the same extrapolation methodology as we did for RateSetter, it is possible that bad debt will exceed Zopa’s forecasts. Using this methodology, Zopa’s extrapolated bad debt is 0.92x expected bad debt for the 2014 and 1.26x for the 2015 cohort. These are smaller multiples than those seen at RateSetter and close to, or below, the 1.2x Safeguard fund coverage. Zopa does not appear to be sailing as close to the wind as RateSetter. However again this assumes no worsening in credit conditions – the impact of Brexit on credit conditions remains a big unknown. Full analysis can be seen in the table in figure 6.
It is not possible to do an analysis on the utilisation of the Safeguard fund contributions by cohort as, unlike RateSetter,Zopa do not provide that information. Zopa do tell us that only 31p of every £1 put into the Safeguard fund has been used thus far. However, given the rapidly growing nature of the platform, this figure is fairly meaningless.
If Zopa’s Safeguard fund were to become depleted, the impact on the platform does not seem to be as profound as the impact on RateSetter in a similar situation. There is no resolution event, no orderly wind down, investors will just begin to take losses if the Safeguard fund is no longer able to pay out.
Actual Bad Debt
Zopa Expected Bad Debt
AltFi Data Extrapolated Bad Debt
Extrapolated Bad Debt/Estimated Bad Debt
Figure 6: Zopa loan cohort performance table
The bottom line is, whilst contingency funds appear to make for a very simple product offering to retail investors, the rate you invest at is in fact only the rate that you get when the fund is operational. As such, investors would be well served by understanding the true risk of the situation. This requires more thought and is potentially more confusing than the non contingency fund models operated by Funding Circle or ThinCats. A depletion of a contingency fund would be a serious setback for the UK industry, much more so than a similar rise in defaults at a non contingency fund platform. With actual bad debt increasingly tracking at, or above, anticipated levels, the protection provided by these funds may soon be tested.