Here's yet another way to look at the LC loan performance data. I start with the chargeoff by vintage data that LC publishes. In the past I've used delinquency data rather than chargeoff, because delinquency comes before chargeoff, so you get an earlier view. There are difficulties with that choice. In particular, nobody knows quantitatively what delinquency predicts. Chargoff on the other hand, is pretty final, so we know its quantitative impact.
We can't wait until loans are "done" to look at chargeoff, because that would be years after we've invested. I'm going to look at chargeoff during early months of loans. I start at month 5 because that's when the action begins. The first payment is due 1 month after origination, and the loan can be 4 months late and therefore eligible for chargeoff at 5 months. I end at month 10 because I don't want to show too many curves, and also because this is the last age that is available for vintage 16Q1. I could plot more months, but they would be shorter curves.
I'm going to plot the numbers the "other way". Ie, instead of the usual way of plotting one curve for every vintage, and making the horizontal axis be months since origination, I'm going to draw one curve for each "months after origination" and make the horizontal axis be the vintage. Then as you scan from left to right you can see how loan behavior has changed vs date of origination.
The first chart is all LC 36 month loans. Data is from the Feb2017 chargeoff file.
You can see that as we move from 2012 to 2016 things got worse. Frankly in this view, it doesn't look all that bad. Things are still much better than they were back in 2008/9. You can see tho that the curves jump up a significant fraction of the distance up to where we were in 2008/9. This is different than say the federal reserve consumer loan data where there is a bump up as we hit 2016, but its very very small relative to the big bump back in 2008/9.
So now lets look at this by grade. I'm not going to do a separate chart for each grade. I'm just gonna plot "E" loans. The data for F&G is very noisy, and I'm not very interested in F&G because I have always avoided them. So lets let "E" be our proxy for the riskier end of the spectrum at LC...
Well! That's a horse of a different color. We can see that performance is getting worse as we move from 2011 thru 2016 vintages, and it is getting worse in a much more dramatic way than the numbers for all loans. Depending on which curve you like to look at and your starting point, it appears that chargeoffs have doubled or tripled! That's a pretty damn big change. And get this... At month 10, the recent vintages are worse than 2008/2009 vintages!
For many years, I used the 2008/2009 crisis as my benchmark for a worst case recession. We are at present not even in a recession, and yet E grade notes of recent vintage are performing worse than 2008/2009.
And note that this behavior is strongly associated with vintage. What would make behavior strongly correlated to vintage? Well of course the underwriting rules the originator had in place at the time of loan acceptance. This is the sort of thinking that leads me to believe that a 2015/6 "E" is not the same animal as a 2011/12/13/14 "E".
I am not saying rawraw is wrong about consumer loan performance on the whole getting some worse in 2015/6. I suspect that many things have contributed, so perhaps we're debating degree of each contribution.
By the way, I suppose some of you have seen Peter Renton's blog this week about performance. Its a bit of a confusing read, because he headlines all-history performance, but inside the article he mentions several other metrics, including trailing 12 month, and most-recent-quarter. When thinking about how returns have changed vs time, I the most recent quarter data is most revealing, so I suggest as you read his blog you concentrate on that, and don't get confused by the other numbers.
At one point he says
Overall my defaults in Q4 were the highest they have ever been. Several of my Lending Club accounts had negative months where the charge-offs were more than the total interest earned. December seemed to be a little better and January was also decent so I am hoping the worst is behind us. We are all paying for the mistakes made by Lending Club and Prosper in 2015 as they clearly mispriced many of their loans, particularly those in the higher risk loan grades where my portfolio is focused.
Yep.
They have since increased interest rates and tightened underwriting considerably so we should see better results from 2016 and 2017 vintages
Here I differ a bit with Peter's view. I don't see any evidence that they have "tightened underwriting considerably". I know that they've repeatedly said that, but I just don't see the evidence that they've really done it. In fact, the charts above show that loans have continued to get worse, even after they made that statement the first few times. Like Peter, I hope things have leveled off. Maybe they have. I just don't see evidence yet.
