Author Topic: 2015 & recent loan quality  (Read 45026 times)

Fred93

  • Hero Member
  • *****
  • Posts: 2164
    • View Profile
2015 & recent loan quality
« on: October 10, 2016, 04:40:32 PM »
Here's a chart showing delinquency (gross dollars delinquent divided by originally funded amount) for recent (the last 14 quarters) of LC loans.  (The legend is on the right.  Some of you will need to scroll to see it.)  This is plotted from data LC produced and distributed.  I've shown only 60 month loans.  I chose to do this because the curves get more smeared out if you include both 36 and 60 month loans in the same chart.  People normally chart this all the way out to 60 months, but I want to draw your attention to the left side of the chart.

By this measure, the last 4 vintages (2015Q3 thru 2016Q2) are each the worst among all their recent peers at their current age.  In other words, the four arrows point to dots that are each the highest so far.



This situation is much easier to see in delinquency curves than in chargeoffs.  I believe that is because delinquency precedes chargeoff, so we're seeing the early warning here.  If that is right, chargeoffs for these vintages will be getting worse in coming months.

LC has acknowledged "pockets of underperformance" recently.  It is discussed in their Q2 earnings call slides.  They describe changes in underwriting designed to address this, and predict that Q3 loans will perform better.  They have not told us when these underwriting changes occurred, but looking at the data, I'd say they didn't arrive in time to improve the Q2 vintage.

 I hope to hear more on loan performance from them during the Q3 earnings call a few weeks from now.
« Last Edit: October 10, 2016, 04:44:42 PM by Fred93 »

storm

  • Full Member
  • ***
  • Posts: 134
    • View Profile
Re: 2015 & recent loan quality
« Reply #1 on: October 10, 2016, 05:04:15 PM »
This chart certainly has me recontemplating my strategy.  Thanks for putting it together.

Rob L

  • Hero Member
  • *****
  • Posts: 2064
    • View Profile
Re: 2015 & recent loan quality
« Reply #2 on: October 10, 2016, 06:32:58 PM »
Wow!. Very nicely done. I especially like that you have segregated 36 and 60 month loans. It's a pet peeve of mine when they are lumped together as you may know from comments I've made many times before. But accolades aside, delinquencies rather than charge offs are with out a doubt a canary in the coal mine. 16Q2 is specially disappointing. The $64k question is simply what is going on here? Can't see where things for average borrower has changed much the past year or two, yet we are all being handed our heads on a platter so to speak; from delinquencies through charge offs. "Something is rotten in Denmark" ... What could it be?

jennrod12

  • Full Member
  • ***
  • Posts: 132
    • View Profile
    • Email
Re: 2015 & recent loan quality
« Reply #3 on: October 10, 2016, 10:08:50 PM »
Thanks for the great chart!  What is included in "delinquent"?  Is that IGP + both late statuses + defaults?

Thanks,

Jenn

PhilGD

  • Full Member
  • ***
  • Posts: 150
    • View Profile
    • Email
Re: 2015 & recent loan quality
« Reply #4 on: October 10, 2016, 11:02:54 PM »
That chart is great but I think we need more clarity into the underlying borrowers. Have the loans written so far in 2016, on average, been assigned higher interest rates that would compensate investors for the higher delinquency performance?

AnilG

  • Hero Member
  • *****
  • Posts: 1095
    • View Profile
    • PeerCube
Re: 2015 & recent loan quality
« Reply #5 on: October 10, 2016, 11:20:48 PM »
Average Interest Rate has been declining since 2013. https://www.peercube.com/histperf/index

That chart is great but I think we need more clarity into the underlying borrowers. Have the loans written so far in 2016, on average, been assigned higher interest rates that would compensate investors for the higher delinquency performance?
---
Anil Gupta
PeerCube Thoughts blog https://www.peercube.com/blog
PeerCube https://www.peercube.com

twigster

  • Jr. Member
  • **
  • Posts: 95
    • View Profile
    • Email
Re: 2015 & recent loan quality
« Reply #6 on: October 10, 2016, 11:40:53 PM »
Quote
Average Interest Rate has been declining since 2013.

Thanks for that and the graph, trying to square this declining interest rate data with what I've heard about lc increasing interest rates this year.  The data they show here:
https://www.lendingclub.com/public/rates-and-fees.action

has been increasing in 2016 as announced has it not (increased in most loan grades, especially lower grades)?

Fred93

  • Hero Member
  • *****
  • Posts: 2164
    • View Profile
Re: 2015 & recent loan quality
« Reply #7 on: October 11, 2016, 12:07:43 AM »
Thanks for the great chart!  What is included in "delinquent"?  Is that IGP + both late statuses + defaults?

As LC put it ... "A loan is categorized as delinquent when loan status falls in one of the following categories: (Late 16-30 Days, Late 31-120 Days, Default)"

Fred93

  • Hero Member
  • *****
  • Posts: 2164
    • View Profile
Re: 2015 & recent loan quality
« Reply #8 on: October 11, 2016, 03:02:58 AM »
Quote
Average Interest Rate has been declining since 2013.
...trying to square this declining interest rate data with what I've heard about lc increasing interest rates this year.

Both are true.  LC reduced rates several times during 2014 and 2015, and has recently increased rates.

Fred93

  • Hero Member
  • *****
  • Posts: 2164
    • View Profile
Re: 2015 & recent loan quality
« Reply #9 on: October 11, 2016, 04:00:07 AM »
Have the loans written so far in 2016, on average, been assigned higher interest rates that would compensate investors for the higher delinquency performance?

You've asked a seemingly simple question and I apologize in advance for giving a complicated answer.

My opinion is "no", the rate increases don't yet compensate lenders, but then I'm a lender, so may have a biased view, and there are many issues.  Some of them are discussed below.

To compensate lenders means to make lenders' "return" adequate, so we have to talk about return.

First, any calculation of "return" for a loan portfolio is going to depend to some degree on predicting the future.  Some of those loan payments will be made, and some will not.  You should note that the chart above shows "delinquency" which does not directly affect return.  We have to wait to see how much of that increased delinquency turns into "default", and whether we're in a bump, or a deteriorating trend.  Delinquency is an early warning system for portfolio health, but doesn't lead directly to knowing where we'll end up.

Next, there are many different ways to calculate return.  We used to debate them here, but that's a waste of time.  Each method has its own merits.  They differ in two fundamental ways.  First, the philosophy.  Different people set out to accomplish different things.  Second, implementation.  Most approaches require estimating future payments, and there are a variety of details surrounding how one does that.

LC started with a very simple calculation they called NAR.  It was based on what has happened so far, ie the payments and the chargeoffs.  It would be a reasonable estimate of return if all the loans in your portfolio made all their future payments, but that is way optimistic, so NAR is too high for most purposes, especially for young loans and young portfolios.  As loans age, NAR slides down toward something less optimistic. 

To improve on NAR, LC added a term which adjusts for loans that are late.  They call this ANAR.  The idea is that if a loan is late, there is some considerable probability that it is going to chargeoff, so lets take part of that chargeoff right now.  They treat the expected value of future chargeoffs as if they occurred right now.  They based the expected value on only the number of days the loan is late.  Simple, and a good first step.  However, ANAR makes no adjustment for loans that are now current.  We know that a considerable fraction of them won't make all their payments, so ANAR is still optimistic, especially for young loans and young portfolios.  ANAR still slides down over time, as its overly optimistic presumptions are replaced with reality.

The ROI computed by the excellent NSRPLATFORM.COM web site is very close to LC's ANAR.  I show some of their numbers below, because many of you are familiar with these numbers, and that will give you a point of reference.

For my own work, I use a different kind of calculation, called IRR.  It computes a return number based on ALL cash flows, not just the ones that have already occurred.  For past payments, I use the history files to tell me which payments were made.  For future payments, I use the expected value of each payment, ie the amount of the future payment multiplied by a probability that the payment will actually be made.  These probabilities get smaller and smaller as we go farther into the future.  I base the calculation of these probabilities on LC's historical loss rates, and some of the credit variables of the loan.  The intent is to compute a return which is a good estimate of where I'm going to end up.  If the loans match the historical loss rates, I should be able to predict return far in advance.  This kind of return estimate doesn't slide down over time, like NAR and ANAR.

However, if loans perform differently than they have in the past, then even my IRR will change with time.  We have some hints that several quarters of recent vintage are performing worse than historical data, so my IRR is likely also optimistic for these recent vintages.

Finally, the chart.



The orange curve: Simple.  It is the dollar weighted average of interest rate (WAIR) for LC's portfolio for the quarter in question.  You can see that interest rates came down big time in 2014 and 2015, and have recently moved up some.  2016Q3 will be even higher, due to LC's most recent rate increase.

The blue dots: These show what WAIR would have been if LC had been using today's interest rates.  You can see, for example, that interest rates were most recently raised too late for Q2 loans.  Q3 loans will have significantly higher rates.  ... but in the big picture, the recent increases seem modest.

The yellow and green curves:  These curves show investor return, measured two different ways, as described above. 

Toward the left side of the chart, many of the payments are already made, so there aren't a lot of future payments to estimate.  The left side is mostly cast in concrete now.  That washes out the different estimation philosophies of the two calculation methods, and the yellow and green curves look about the same.  Yellow is higher, but that is mostly because IRR is inherently compounded, where as NSR's ROI is simple interest.  If you compound NSR's ROI, the numbers come out very close. 

Even tho the two methods begin with a different philosophy, for old loans the numbers come out almost the same.  That's why I stopped arguing about which method is "right"!

Toward the right side of the chart, something different is going on.  The green curve is wildly optimistic, because it takes no account of the possibility that some fraction of future payments on current loans will not be made.  If I recompute the green curve a few weeks from now, it will be lower on the right side, for sure, because some dude somewhere will have stopped paying.  If you recompute the green curve a couple of years from now, it will almost surely be below where the yellow curve is now.

The yellow curve is a serious attempt to use all we know to avoid that kind of optimism by calculating and using probabilities that future payments will be made.  However, this depends on those probabilities being fair estimates of future paying.  We know, from the delinquency statistics, that the last several quarter vintages are performing more poorly than the historical average.  Unfortunately, we don't know how much more poorly they will perform.  We will learn that in the coming months. 

The yellow dotted line is a projection of the sort of thing that could happen, if the underperformance we see produces significantly increased loss rates for these vintages.

LC has told us that they've made some underwriting changes, which we should expect to improve Q3.  We don't know what those changes were, except in the vaguest terms, and don't know how effective they will be.  We'll have to wait several weeks to get the first hints of Q3 performance.

« Last Edit: October 11, 2016, 04:16:40 AM by Fred93 »

rawraw

  • Hero Member
  • *****
  • Posts: 2768
    • View Profile
2015 & recent loan quality
« Reply #10 on: October 11, 2016, 08:28:12 AM »
Is this true across all grades or is this masking mixed performance by grade? Thanks for sharing, you continue to be one of the most valuable contributors

Sent from my SAMSUNG-SM-G935A using Tapatalk


SLCPaladin

  • Full Member
  • ***
  • Posts: 202
    • View Profile
Re: 2015 & recent loan quality
« Reply #11 on: October 11, 2016, 09:55:52 PM »
This is perhaps the most detailed and understandable explanation of forecasting returns I have read on this thread. I am quite grateful for your insight. I kind of feel like a big wild card right now is the political climate we're in. We need some clarity on how policy might affect broader economic trends and how those might contribute to better, or worse, macro environment.

Fred93

  • Hero Member
  • *****
  • Posts: 2164
    • View Profile
Re: 2015 & recent loan quality
« Reply #12 on: October 11, 2016, 11:14:20 PM »
Is this true across all grades or is this masking mixed performance by grade?

The effect appears to be larger in the riskier grades.  (on an absolute basis, perhaps not on a relative basis)  I have included below a chart from the excellent INSIKT.COM web site.  They have a tool that you can use to chart this data broken down many different ways.  They don't have 15+ day delinquency numbers, as I showed before, but do have 30+ and 60+.  For some reason this is most visible in their 60+ charts.  They will display quarterly data, but this makes awfully busy charts, so I showed the yearly breakout, and restricted the curves to B thru E grades and 3 years.  Each color is one grade.  Start by just looking at one color.  The three curves of the same color represent the different years.



The effect is difficult to see in grades A & B.  After that, each grade you go up, the gap between the years 2014, 2015, 2016 gets larger.

Although F & G are not shown, the gap there is even bigger.

LC's statements about recent vintage performance speak of "pockets of underperformance", which sounds like "Oh, he must mean those risky guys I don't invest in".  I'm still scratching my head on their use of the word "pockets", because at least when you break down loans by grade, the effect seems spread across many grades.  The effect is substantial at "C" grade, which is the center of my portfolio.  Perhaps there is some other criteria which will isolate the "pocket".   I don't know what that would be.

James Wu at Monja wrote a blog entry about this awhile back.  He filtered by various credit parameters vs time, and found that around mid April 2016, there is some sort of change in LC's population where there are fewer high DTI and high Inquiries_last_6_months.  That's evidence of a change in underwriting of some sort, but it isn't clear exactly what changed.
https://www.monjaco.com/blog/changes-in-lending-club-underwriting-looking-beneath-the-headlines/

Wu's observations aren't the final word on the problem, because I never invested in that corner of credit variables, yet my account is certainly affected (ie my delinquent fraction has grown).

Quote
Thanks for sharing, you continue to be one of the most valuable contributors

You are most welcome.
« Last Edit: October 11, 2016, 11:46:37 PM by Fred93 »

Fred93

  • Hero Member
  • *****
  • Posts: 2164
    • View Profile
Re: 2015 & recent loan quality
« Reply #13 on: October 12, 2016, 04:42:31 AM »
Sometimes my memory is imperfect.  I went back and looked at what was in LC's slides from the Aug 8 2016 second quarter results presentation.



Here he does actually say what they changed and when they changed it.  Its the little list on the bottom left.

There's also something amazing at the top of this slide.  I thought I was going out on a limb a few messages ago when I projected that the last few quarters loans might end up at an all-LC average around 5% return.  Well, son of a gun, a similar statement is right there on slide 8.  He says "Portfolio level returns expected to increase from 4-5% to 6+% for vintages after June."  In other words, they seem to admit that recent vintages are headed toward around 5% returns (all-LC average).

Of course we don't know how the 2015Q3 thru 2016Q2 loans' performance will evolve, ie how bad they will actually end up.  Our projections are tentative, because after something has changed, so can't reliably use history to guide us.

The big thing we want to know is - How will 2016Q3 loans perform?  LC made an adjustment, and it will take a few months before we know if they got it right.  We'll get our first glimpse when Q3 loans are added to the historical file a few weeks from now.  Early Nov I think.  I suspect we won't have enough data in that release to make a judgement.  By the end of the year we should have enough payments (either paid or not paid) on these loans to plot some nice delinquency curves like above, and see whether early performance of the Q3 loans "pokes out" like recent quarters have.

Here's a fact that hints that maybe they have not fixed the problem:  The two changes they list above were made in April and June.  We don't know when it actually took effect, so lets assume mid-April.  Second quarter is April, May, June, so 80% of Q2 is after that change!  And yet... We know (although not enough time has elapsed to have really good data), 2016Q2 delinquency numbers are looking bad.  That's a thumbs down.  Then they made another change in June, but pulling in the DTI limit from 40% to 35% doesn't seem like it would be a major fix.  Count me skeptical.

On a related note: A quick look at Prosper shows that their recent quarters have suffered a similar reduction in quality.  I haven't done any detailed crunching on Prosper data tho.



« Last Edit: October 12, 2016, 04:46:01 AM by Fred93 »

SLCPaladin

  • Full Member
  • ***
  • Posts: 202
    • View Profile
Re: 2015 & recent loan quality
« Reply #14 on: October 12, 2016, 01:21:06 PM »
I would think that if total portfolio returns stay in the 4-5% range for too long, LC is going to have a very difficult time attracting more capital to fund their loans, from whatever source. I don't think that is near enough of a risk premium for retail or institutional money. If they don't quickly bump their rates or get their underwriting back to where returns get back to around 7%, I won't matter that they hired a "Chief Capital Officer".