You are now being redirected to Orchard Platform Markets, LLC.
If you aren’t redirected in a few seconds click here.
You are now being redirected to Orchard Indexes.
If you aren’t redirected in a few seconds click here.

Subscribe to our Newsletter

Login

  • Investor Portal

  • Originator Database

  • Manager Database

If you forgot your login email, please contact your Orchard Account Manager or support@orchardplatform.com.

Revisiting Vintage Analysis- How Loans Perform With Age

In this blog post, we revisit one of Orchard’s more popular articles about how loans of different vintages perform with age. We previously highlighted the importance of evaluating a portfolio of loans by their vintages, or the period in which they were originated. As the industry has matured, we have gained further data and insights as to how online-originated loans and their defining characteristics behave over time. We will focus on data from Lending Club and Prosper, but vintage analysis can be done on any portfolio of loans.

 

Vintage Analysis and Performance

In the graphs below, we show the cumulative incidence of charge-off (i.e. when a loan is determined to be a loss) for annual vintages of Lending Club and Prosper loans. The older vintages have longer lines, as they have more months of history. Using this data, we can examine how loans booked at different times compare to each other at equivalent periods in their life-cycle. This can help an investor evaluate their current portfolio and help them make comparative judgments about its performance.

In our initial version of this post, we examined the gross charge-off rates for loans originated on Lending Club and Prosper and found that they were high in the initial years of the platforms’ existence but stabilized quickly. As the industry has matured, it can be seen that charge-offs in newer vintage loans are generally lower. Vintage analysis however, shows us that this does not always hold true. For example, at 60 months, loans booked in 2010 performed better than those booked in 2012.

 

Factors to Consider Within Vintage Analysis

Interest Rates

While basic vintage analysis is useful, it is also important to conduct this type of analysis within the context of the other factors that tend to fluctuate from year to year. For example, by looking at historical data, we can see that the distribution of interest rates varies from origination year to origination year.

The above graphs show us the percentage of loans that were originated each year at every interest rate level. They seem to indicate that interest rate levels in 2015 in particular were lower on the whole than interest rates in other years. This aligns with the facts that market sentiment in 2015 was positive for online lending, supply of investor funding was ample, and that many platforms were competing for customers, driving interest rates lower.

These graphs also show that interest rates for Prosper were historically much higher than interest rates for Lending Club. This difference is highlighted further in the graph below which shows a historical comparison of both firms’ weighted average interest rates. The weighted average interest rate for loans originated by Lending Club has been similar over the observed time period, while the weighted average interest rate for loans originated by Prosper has greatly fluctuated between 2008 to 2016.

This means that when we consider a Prosper loan originated in May 2011 with an average interest rate of over 20% and a similar (in terms of risk characteristics) Prosper loan originated in May 2015, we have to be aware that these two loans may have differing interest rates that may affect their performance.

When considering how interest rates relate to performance over time, we also can not assume that two loans of similar vintages and interest rates charge off at the same rate. Consider the chart below:

As the graph shows, interest-rate bands of loans originated in different years can have very different charge off rates. A loan originated in 2008 with an interest rate between 13-15.9% reached a charge off rate of 20% whereas a loan with the same interest rate originated in 2012 only reached a charge off rate of less than 10%. This means that performance is affected by other factors beyond interest rates. We need to look at FICO scores, debt-to-income ratios, etc. that may have an effect. Secondly, these graphs show that general macroeconomic conditions have a very strong effect on borrower performance.

Credit Grade

Vintage analysis can also help us to see how loans within a particular credit grade perform over time. In our prior analysis, we examined the performance of the top graded loans (A for Lending Club and AA for Prosper). However, as time has passed, these two platforms have increasingly been lending to borrowers with credit just below the top grades. The following is a graph showing the distribution of loans originated by grade in two distinct time periods:

While there is no massive shift in lending by credit grade, both Lending Club and Prosper have increased their B and C rated loans during the 2013-2016 time range. This may be in response to investors demands to increase origination levels.

Let us examine how credit grade has impacted the performance of these loans over time. Below is the performance of the top graded loans (A for Lending Club and AA for Prosper):

The first observation we can make by comparing the scale in the graphs to the scale of the graphs from the first two performance charts is that “Grade A” (Lending Club) or “Grade AA” loans (Prosper) charge off at a much lower rate than the overall portfolio. The next observation is that in comparing the 2008-2012 time frame to the 2013-2016 time frame, charge-offs have more or less decreased for top graded loans. Another observation is that “A” or “AA” from one vintage is not necessarily an “A” or “AA” from another vintage. For instance, by the 15th month post-booking, 0.83% of 2014 Lending Club loans and 0.95% of 2014 Prosper loans charged off whereas 0.73% of 2015 Lending Club loans and 1.01% of 2015 Prosper loans had charged off by the 15th month. This illustrates how predictions of future credit risk can be volatile over time. The loans might have looked the same at the time of issuance, but they performed differently, perhaps due to broader economic circumstances, shifts in the “mix” of loan applicants, or variables not accounted for in Lending Club’s or Prosper’s credit models.

It is important to note that grades are not necessarily comparable between originators — a Lending Club “B” and a Prosper “B” may have differing risk characteristics. Different credit grades also have varying interest rates from year to year as seen in this graph:

The graph above shows the interest rate for a credit grade A Lending Club loan has largely stayed the same while the interest rate for credit grade F and G loans has increased precipitously over time. Through vintage analysis we can see how different credit grades from different years can perform very differently though they technically have similar risk characteristics.

Charting credit grade by FICO score over time further illustrates the point that a grade A loan from 2010 is not the same as a Grade A loan from 2016. For a 36-month loan the average FICO score for grade A loans has dropped but charge offs have not increased. This suggests that over time Lending Club’s credit models have improved at predicting risk across FICO scores.  

FICO Score and Debt-to-Income Ratio

Other important metrics to examine in conducting vintage analysis are FICO scores and debt-to-income ratios. It is particularly important to examine these factors when conducting vintage analysis on a pool of loans originated by a platform that does not assign credit grades. From the data below we can see how loans from Lending Club charge-off over time controlling for the debt-to-income ratio of the borrower.

At the 12 months we can see that for the most part, loans with higher FICO brackets and lower debt-to-income ratios charge off at a lower rate which is to be expected. However, by adding the time dimension we can see that in more recent years this trend has become more clear as the industry has matured. There are fairly high charge off rates in all debt-to-income brackets in 2008 and 2009.

After 24 months, the loans originated in 2008 and 2009 have charged off at a much higher rate across all debt-to-income brackets than loans originated in later years. Therefore, when trying to predict performance of loans, while it is tempting to rely solely on factors like FICO scores and debt-to-income ratios as a major predictor of charge off rates, volatility and other economic factors can cause a certain vintage to under or over perform.

 

Conclusion – Vintage Analysis as a Tool, Volatility as a Certainty

Our initial conclusions on the importance of vintage analysis still hold true and are reinforced by the data that Orchard has collected over time. Ultimately, vintage analysis is only one tool in a broad spectrum of tools typically used to evaluate a pool, or portfolio, of loans.

Vintage analysis is a tool available to investors seeking to better forecast the performance of loan portfolios over their lifetime. Understanding how “seasoned” loans are may help investors distinguish between the mature and new portions of a portfolio. Investors may want to consider how recently-booked loans are trending relative to earlier vintages at similar points in their lifecycle. We can look at the various factors that impact loan performance through a temporal lens to make predictions on the types of returns we might see from a portfolio of loans.

As seen in the graphs above, volatility is a fact of life. Vintages will vary in their performance for reasons beyond your foresight as an investor, even when you account for fluctuations in credit grade, interest rates, and other factors. When investing in loans, it is advisable to leave yourself a buffer between the interest rate and your expected loss in the event that your loans default at a higher-than-expected rate. Most importantly, diversification across vintages and other factors can help to manage your exposure to volatility.