Author Topic: 2015 & recent loan quality  (Read 35779 times)

Fred93

  • Hero Member
  • *****
  • Posts: 2102
    • View Profile
Re: 2015 & recent loan quality
« Reply #105 on: May 07, 2017, 03:07:46 PM »
This and your 36 month version were really nice!
Any plans to update them now that the loanstats files have been updated to include 2017 Q1 (maybe you already have and I missed it)?

The files have not yet been updated to include 17Q1.  The official timing is 6 weeks after end of quarter, which puts it at mid May.  Sometimes they're a few days earlier than that.

Edited to add: Clarification ... The loan status files have been updated, but the loan payments files have not yet been updated.  One could theoretically produce the graphs from the loan status files, but that's not how I set it up, so we have to wait a bit more.
« Last Edit: May 07, 2017, 03:50:58 PM by Fred93 »

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: 2015 & recent loan quality
« Reply #106 on: May 07, 2017, 08:32:39 PM »
This and your 36 month version were really nice!
Any plans to update them now that the loanstats files have been updated to include 2017 Q1 (maybe you already have and I missed it)?

The files have not yet been updated to include 17Q1.  The official timing is 6 weeks after end of quarter, which puts it at mid May.  Sometimes they're a few days earlier than that.

Edited to add: Clarification ... The loan status files have been updated, but the loan payments files have not yet been updated.  One could theoretically produce the graphs from the loan status files, but that's not how I set it up, so we have to wait a bit more.

Thanks for the response. The "big file" pmthist. Should have remembered that schedule.
I'll very much look forward to seeing the latest.

Fred93

  • Hero Member
  • *****
  • Posts: 2102
    • View Profile
Re: 2015 & recent loan quality
« Reply #107 on: May 08, 2017, 06:26:20 AM »
I know you're waiting to see the 17Q1 vintage appear, but I went ahead and charted the April2017 delinquency file, as I haven't updated vintage delinquency curves in awhile.  I believe the May2017 file will be where 17Q1 makes its appearance.



Each curve represents one "vintage", ie loans issued during one calendar quarter, as indicated by the legend on the right of the chart.

Vintages from 2015 and 2016 are mostly toward the upper edge, ie near worst ever. 

However, 16Q4 is in the middle of the pack, ie better than one would expect from recent vintages.  Its still quite young tho.  Month 3 doesn't dictate what happens later.  For example, 15Q4 started out good, then bent upward.

Hard to say what this data means.


Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: 2015 & recent loan quality
« Reply #108 on: May 08, 2017, 11:11:49 AM »
Thanks! All kinds of interesting stuff there.

It looks like after month 13 the curves begin to converge in performance. There has been and continues to be a problem with delinquent payments in the first 13 months starting with 2015 Q2. After 13 months the difference narrows significantly. Simply put, the older the loan then the less difference vintage makes in payment performance. It would be interesting if the chart extended another two months to 20 to confirm this thesis.

If I'm reading the colors right then the difference between 2015 Q1 and 2015 Q2 is remarkable. At month 11 their comparative performance was actually inverted. Afterwards Q1 remained pretty flat and Q2 just took off.

Not looking to make additional work :), but a constant time look at these delinquencies, like you did for charge offs in another thread, would also be very cool! Lets us know if something was happening with borrower performance at a certain moment in calendar time across all vintages.

bluto

  • Newbie
  • *
  • Posts: 22
    • View Profile
    • Email
Re: 2015 & recent loan quality
« Reply #109 on: May 11, 2017, 05:39:43 PM »
The 13 month convergence seems likely to be because delinquency is only a temporary state (they would be further apart if delinquency + default were charted). 

JohnnyP

  • Newbie
  • *
  • Posts: 34
    • View Profile
Re: 2015 & recent loan quality
« Reply #110 on: May 12, 2017, 12:46:00 AM »
They converge because the denominator is the originally funded amount and does not change. The numerator is the gross amount dilingquent which eventually starts to decrease as the outstanding principle decreases (as loans are paid off). If the graph showed gross amount dilinquent divided by outstanding principle, the slopes would continue in a positive direction and would even increase as months on the books increases. I would bet the lines would not converge either.

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: 2015 & recent loan quality
« Reply #111 on: May 12, 2017, 10:47:16 AM »
I think Fred93 defines delinquent as Late / (Late + Current) so the denominator isn't the originally funded amount.

Fred93

  • Hero Member
  • *****
  • Posts: 2102
    • View Profile
Re: 2015 & recent loan quality
« Reply #112 on: May 12, 2017, 02:29:31 PM »
I think Fred93 defines delinquent as Late / (Late + Current) so the denominator isn't the originally funded amount.

Oh ... I have when I computed the numbers for my own account ... but the chart above is made from data provided by LC, in their delinquency file, and they indeed used originally funded amount as the denominator.  See the label on the vertical axis.

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: 2015 & recent loan quality
« Reply #113 on: May 12, 2017, 04:03:09 PM »
I think Fred93 defines delinquent as Late / (Late + Current) so the denominator isn't the originally funded amount.

Oh ... I have when I computed the numbers for my own account ... but the chart above is made from data provided by LC, in their delinquency file, and they indeed used originally funded amount as the denominator.  See the label on the vertical axis.

Thanks for the correction.

Fred93

  • Hero Member
  • *****
  • Posts: 2102
    • View Profile
Re: 2015 & recent loan quality
« Reply #114 on: May 26, 2017, 02:27:23 AM »
Well, LC has delivered their May update to the charegoff file, so here are my latest chart updates.

Quick review:  The vertical axis is cumulative net chargeoff, ie ($ charged off minus $ recovered) / original amount of loan.  Most people plot vintage chargeoff curves with one curve per vintage and a horizonal axis of loan age.  That may be the intuitive way to do it, but it doesn't really let you see the differences between vintages well.  I've turned that around.  There is one curve for each loan age, and the horizontal axis is vintage.  For example, the green curve shows chargeoff at 10 months age.  If every vintage of loans performed the same, these curves would each be a horizontal line.  Instead, they go up and down, as you move across vintages, showing how performance of vintages differ.  I start with month 5, because that's the first month where a loan can normally charge off.  I've only charted a few ages, because if I show them all, the chart gets cluttered.

The first chart is for all 36 month loans.  The second is for all 36 month grade "E" loans.



In spite of repeated assurances from LC, 16Q1 and 16Q2 look worse than all the (very bad) 2015 loans.  Still not nearly as bad as 2008/9, that is until we look at the higher risk grades.



OMG.  "E" loans issued at the end of 2015 were as bad as 2008/9 loans at the 10 month point, and Q1 & Q2 of 2016 are much much worse!  Thank goodness the early data from Q3 doesn't continue the upward trend... but it is still worse than any data from 2015. 

60 month loan charts look similar, except that the chargeoff numbers are bigger.
« Last Edit: May 26, 2017, 02:29:33 AM by Fred93 »

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: 2015 & recent loan quality
« Reply #115 on: May 26, 2017, 11:10:12 AM »
Thanks for the update. Certainly brightened my morning as a holder of a mostly D & E portfolio.

There are two factors that contribute to changes in charge off rates; borrower performance and LC underwriting standards.

Over time borrowers change their behavior, typically in response to economic conditions, based on their willingness and/or ability to pay. Generally it would seem that the economy has done nothing but improve since the 2008-2009 great recession, with unemployment now extremely low (top line number 4.5% I believe) and real personal income very near an all time peak (equaling real personal income all the way back in 2002). This would support an argument that borrower performance should at this time be quite good, but there is evidence that there has been some degradation recently, looking at auto loans for example.

Also over time LC changes its underwriting standards, choosing to accept or reject borrowers with stronger or weaker credit history and they assign interest rates to loans that are accepted. Obviously if LC lowers the bar to accept weaker borrowers then charge offs go up. The reverse is also true when they raise the bar. From a charge off perspective lowering the bar only affects the most risky loan grades. All borrowers that qualified before the bar was lowered still qualify after. Borrowers that would have been rejected before the bar was lowered enter the loan pool into the highest risk grade(s) and are the most likely to charge off. Lowering the bar and increasing charge offs isn't necessarily a bad thing; it all depends on interest rates. If the interest rate is high enough and if a lender is able to diversify across a large enough number of loans these riskier loans could be attractive to lenders wiling to take on the added risk. And there's a lot of risk, considering what would likely happen in an economic downturn for example.

My theory is that three of four things have gone badly with the riskiest loans. Considering the example of auto loans, borrower performance has gone down. LC lowered the bar and began accepting less credit worthy borrowers. LC actually dropped interest rates while lowering the bar, thus LC threw lenders to high risk borrowers under the bus. That's three and I'm glad the economy / lending cycle hasn't produced another recession to make it a quadfecta.

Edited:
There is a turn around in charge offs shown on the E chart for 16Q3 that is consistent for MOB 5, 6, 7. That's what I would expect to see when the bar has been raised.  If true then the question for us lenders is whether or not the interest rates assigned to loans in 16Q3 and later were (are) high enough to yield acceptable returns. I really have my doubts but time will tell.
« Last Edit: May 26, 2017, 02:11:45 PM by Rob L »

Fred93

  • Hero Member
  • *****
  • Posts: 2102
    • View Profile
Re: 2015 & recent loan quality
« Reply #116 on: May 26, 2017, 06:11:54 PM »
My theory is that three of four things have gone badly with the riskiest loans. Considering the example of auto loans, borrower performance has gone down. LC lowered the bar and began accepting less credit worthy borrowers. LC actually dropped interest rates while lowering the bar, thus LC threw lenders to high risk borrowers under the bus. That's three and I'm glad the economy / lending cycle hasn't produced another recession to make it a quadfecta.

I spent a lot of time over the past few months trying to determine whether the cause was LC or external, and I've concluded that it was a combination, precisely as you describe.

I will add that LC was slow to realize what was going on, so didn't correct quickly.

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: 2015 & recent loan quality
« Reply #117 on: May 26, 2017, 06:34:38 PM »
My theory is that three of four things have gone badly with the riskiest loans. Considering the example of auto loans, borrower performance has gone down. LC lowered the bar and began accepting less credit worthy borrowers. LC actually dropped interest rates while lowering the bar, thus LC threw lenders to high risk borrowers under the bus. That's three and I'm glad the economy / lending cycle hasn't produced another recession to make it a quadfecta.

I spent a lot of time over the past few months trying to determine whether the cause was LC or external, and I've concluded that it was a combination, precisely as you describe.

I will add that LC was slow to realize what was going on, so didn't correct quickly.

Hey, you did the heavy lifting. All I did was look at your charts and say what seemed pretty obvious.
We can only speculate as to why LC has been so slow to correct. They make money on originations and servicing, not the loans (no skin in the game), so they had little incentive to correct. Also, they were preoccupied with recovery from the events of May 2016 and were up to their necks in alligators. There's more blame game themes that I could tack on, but it is only speculation.

Addition:
In my best years leading up to the demise of LC high risk lending I made well over 10% per year. However, it's an well accepted truth that recessions are not predictable. Certainly they are not predictable over the 3 - 5 year time scale commensurate with the time scale of the principal I invested in these high risk loans. I gotta ask what rate of returns is appropriate and sufficient for lenders during the good times to compensate them for the inevitable bad times that will someday follow?  Hint, it ain't 6.72% which is my primary notes ANAR today. If we have a recession in the next 3 years I'll be lucky to break even. Signed from under the bus ...!
« Last Edit: May 26, 2017, 07:06:31 PM by Rob L »

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: 2015 & recent loan quality
« Reply #118 on: May 30, 2017, 09:24:41 AM »
From Economic Policy Journal.com:

"The average credit score nationwide hit 700 in April, up one point from last fall, according to new data from Fair Isaac Corp. That is the highest since at least 2005. That was the year Fair Isaac, the creator of widely used FICO credit scores that range from 300 to 850, began tracking the data.

Meanwhile, the share of consumers deemed to be riskiest, with a score below 600, hit a new low of roughly 40 million, or 20% of U.S. adults who have FICO scores, according to Fair Isaac. That is down from 20.5% in October and a peak of 25.5% in 2010."

FWIW: http://www.economicpolicyjournal.com/2017/05/credit-scores-hit-record-high.html

The article has a nice chart of the FICO of the riskiest borrower scores over time.

This hardly supports the notion that LC borrower performance has played a major part in the degradation of returns we've experienced over the past 18 months.
« Last Edit: May 30, 2017, 09:28:34 AM by Rob L »

nonattender

  • Hero Member
  • *****
  • Posts: 713
  • I am not here.
    • View Profile
Re: 2015 & recent loan quality
« Reply #119 on: May 30, 2017, 11:15:04 AM »
Dude.  Those scores should come with a pamphlet weighing 10 pounds that explains how the models have progressed over time and what they actually reflect, today.  It's like how the SAT used to be highly correlated with intelligence quotient scores - and now Junior gets 1000 points for bubbling in the letters of his name properly.  They're getting ready to fuck us, again, in July, when they stop counting civil suits / medical debt / who knows what else.  Anyone who thinks they have a good handle on any FICO algo should be quite careful in a month...

LOL at that journalist's line about 2005 being "the first year FICO tracked (average credit score) data", too.  Yeah, before, they'd just toss darts all day and say "my, we have data on 200 million US consumers... anybody ever done basic analytics?  Nah, you're right - lunchtime!"

 ::)
A little nonsense now and then is relished by the wisest men.