a view of LC's deteriorating investor returns

[JohnnyP]

I appreciate the conversation. A recent post brought up the Orchard Index which I have attached. This is fascinating to me.

https://www.orchardindexes.com/

The Orchard Index represents all outstanding principle on consumer loans for the largest companies (Lending Club, Prosper, etc.). It also reduces returns based upon estimated defaults on late loans, so it does not wait for charge offs to occur before affecting the returns.

When you look at the web site, you see scary negative slope from 0.45% monthly return in July, 2016 to 0.04% monthly return in December. How can this happen? The numbers are not by vintage. The numbers represent the entire universe of outstanding principle. That means there could not be much "portfolio" turnover in 6 months. This was not caused by underwriting changes! What does the Forum think?

[SLCPaladin]

Well, here's another way to look at the situation.  We usually look at returns.  Many of us calculate returns different ways, but all of them
I've calculated a chargeoff rate for two portfolios.  These happen to be two portfolios that I have access to.  The first is my own Fred93 portfolio.  The second is Lending Club's Broad-Based Consumer Credit Fund (which I refer to as BBQ). 

For my own account, each month I take the net chargeoff $ (which means chargeoffs minus recoveries) and divide by the account balance at the beginning of the month.  This produces a monthly chargeoff rate.  I then multiply it by 12 to get an annual rate.  Nothing fancy.  No freaky math like IRR or ANAR.  Everybody can follow along.

Since the beginning of 2016, the BBQ fund has published their monthly net chargeoff rate.  I've multipled this by 12 and plotted it on the same chart as the Fred93 chargeoff rate.

As always Fred, your posts are very insightful. I've stopped reinvesting in LC for now because I have not been able to discern any leveling out of the default rate, and hence I don't know where the ultimate return rate will be. I'm in de-accumulation mode right now until I see evidence that I've seen a floor, or that LC raises interest rates enough to convince me that I'll be compensated for the "new normal" of defaults.

I do have a question about your calculation though: Wouldn't you expect the charge off rate to be highly correlated to one's average portfolio maturity? What I mean is, whenever I look at the slope of cumulative defaults, I tend to see a lot of them happening in the first 12 months, and certainly within the first 18 months. I would expect that charge-offs trend up according to the natural decay of one's portfolio, so couldn't increasing charge-offs simply be the natural aging associated with the decay process of a portfolio? How much of the uptick to attribute to lax underwriting standards and how much to attribute to normal aging seem to be the problem to solve. To be certain, the slope of "cumulative charge-offs" has definitely increased in the past several quarters and years, so we know that defaults are happening much more frequently than in the past, and, as a result, that returns have dipped dramatically. No surprise there. I certainly think the calculation you've highlighted is a handy tool to quickly calculate returns as a function of interest rates and charge-offs, since they are obviously joined at the hip. But I seem to think that the more useful comparison is by vintage of cumulative charge-offs.

[Fred93]

A recent post brought up the Orchard Index which I have attached. This is fascinating to me.
https://www.orchardindexes.com/

The Orchard Index represents all outstanding principle on consumer loans for the largest companies (Lending Club, Prosper, etc.). It also reduces returns based upon estimated defaults on late loans, so it does not wait for charge offs to occur before affecting the returns.

Thanks for bringing that up.  Orchard plots this as a cumulative return (ie one of those "if you invested $1000 back when it would be worth $xxx now" charts).  However, they give the monthly return numbers they've computed, so I multiplied by 12 to annualize (consistent with the other charts I've recently shown) and charted Orchard Index annual return...



This is helpful for comparison to our own returns, including the other charts I've shown recently.  If I were to pick nits with them, the nits would be ... They don't say which companies are in the index, so I can't compute, for example, how much of the index is LC.  I can guess of course.  They have taken the time to write out a document explaining their methodology, which is great, but they include "mark" term to mark loans to market, and then their document fails to discuss how this mark is computed, or even what is included in it.  Because it is an all-inclusive index, it doesn't address the question of whether one could easily do better by choosing some subset of the loans offered.

Nevertheless, it is another view, and a big-picture view, which gives us something to compare against. 

What it seems to say is that we're all goin' down.

Edited to add...

PS: This view does seem to say that we (collectively) are doin' better in the early months of 2017 than we did in 2016.

PPS: Another thing that struck me about this is how straight the curve is.  I added a linear trendline, and you can see that except for the 2nd half of 2016, we were hugging close to a linear trend.  Unexpected.  Has this trend been broken?  Hope so.

[JohnnyP]

This trend is not our friend. Nice graph though!

[RaymondG]

Since the beginning of 2016, the BBQ fund has published their monthly net chargeoff rate.  I've multipled this by 12 and plotted it on the same chart as the Fred93 chargeoff rate.

I do have a question about your calculation though: Wouldn't you expect the charge off rate to be highly correlated to one's average portfolio maturity? What I mean is, whenever I look at the slope of cumulative defaults, I tend to see a lot of them happening in the first 12 months, and certainly within the first 18 months. I would expect that charge-offs trend up according to the natural decay of one's portfolio, so couldn't increasing charge-offs simply be the natural aging associated with the decay process of a portfolio? How much of the uptick to attribute to lax underwriting standards and how much to attribute to normal aging seem to be the problem to solve. To be certain, the slope of "cumulative charge-offs" has definitely increased in the past several quarters and years, so we know that defaults are happening much more frequently than in the past, and, as a result, that returns have dipped dramatically. No surprise there. I certainly think the calculation you've highlighted is a handy tool to quickly calculate returns as a function of interest rates and charge-offs, since they are obviously joined at the hip. But I seem to think that the more useful comparison is by vintage of cumulative charge-offs.

I guess that Fred93's portfolio is well aged.  It must have been in a relative stable state where loans are close to evenly distributed by age after having been reinvesting for a long time.  In this state, the charge-off rate per month would not change significantly by time passing by.

[rawraw]

Fred, you seem pretty savvy -- do you calculate vintage past due and loss curves?  It may be more useful for you to spot trends

[Fred93]

Fred, you seem pretty savvy -- do you calculate vintage past due and loss curves?  It may be more useful for you to spot trends

Sure.  You can get such traditional curves on www.insikt.com .  Right now they haven't updated their database since Feb17, whereas the Apr17 files are now available.  Other than that detail, they're fine.

Meanwhile, I've been charting loss (cumulative net chargeoff) a different way than the normal one.  Usually people plot "vintage curves" with one curve for each vintage of loan, and a horizontal axis that represents months since start of loan.  LC just published their April update, so I updated my chart.

{FYI: Vintage means a group of loans issued within a given quarter.}

I've turned the data around, with one curve for every age (ie months since start of loan), and a horizontal axis which represents vintage.  If every vintage performed the same, then all these lines would be strictly horizontal.



These curves end at different vintages, because that's where time ends.  In other words, 16Q3 loans haven't all reached 8 months age yet, so LC doesn't yet publish that number.

The curves are all tilting up toward the right end.  This means that recent vintages are performing more poorly than prior vintages.

Another thing worth noticing is that the curves don't all have the same shape.  In other words, chargeoffs at 6 months don't predict chargeoffs at 36 months very well.  The early months give a hint, but you can't extrapolate very well.  Things change.  Perhaps economic conditions, seasonality, etc.  Vintage is important, but it isn't everything.

Whether this way of looking at the data is more insightful than some other way is unclear to me.  Each way of looking at the data allows you to see different things.  Attributing the deteriorating performance to cause is more difficult in any case.

[brycemason]

Nice alternative graph, Fred93. I like the iso-time-through concept.

[Fred93]

Thanks Bryce.  Glad somebody understood what I was trying to do.

As we know, the recent deterioration is heaviest on the high risk end.  With that in mind, I've charted the same thing, but this time for "E" grade 36-month only loans...


The thing I find interesting about this chart is that it shows that E loans are ALREADY doing as poorly as they did in 2008!  We called that time "the financial crisis" or the "great recession".  Today economic indicators today are MUCH better than in 2008/9, but "E" grade loan performance is back to what it was at that time, and are heading toward even worse.  I don't claim to understand this, but the data is clear.

[Rob L]

I'd like to second the very nice work. A fresh perspective, but paints an ugly picture and for now it's only getting worse.
It's pretty clear by contrasting your two charts that the erosion is most pronounced on the high risk end.
What is it with borrowers these days? It's not just LC but consumer debt in general (autos, etc.).

[AnilG]

The implicit assumption in your analysis is that the credit and loan profile of E Grade loans has stayed similar across vintages. Lending Club has expanded credit criteria several times since 2008 in attempt to increase volume but still assigns these newly eligible borrowers to same 7x5 Grade buckets. I don't believe today's E Grade loans are same as the E Grade loans of 2008-2012, they are more like E+F+G+H of  yesteryears ...

Thanks Bryce.  Glad somebody understood what I was trying to do.

As we know, the recent deterioration is heaviest on the high risk end.  With that in mind, I've charted the same thing, but this time for "E" grade 36-month only loans...


The thing I find interesting about this chart is that it shows that E loans are ALREADY doing as poorly as they did in 2008!  We called that time "the financial crisis" or the "great recession".  Today economic indicators today are MUCH better than in 2008/9, but "E" grade loan performance is back to what it was at that time, and are heading toward even worse.  I don't claim to understand this, but the data is clear.

[Fred93]

The implicit assumption in your analysis is that the credit and loan profile of E Grade loans has stayed similar across vintages.

I believe, like you, that they have not stayed the same.  I don't understand how they have changed, but the performance seems to tell us that they have changed. 

Other folks have done analysis of average FICO, average inquiries, ... in each grade vs time, and those studies seem to conclude that the grades haven't changed much.  I haven't repeated those studies myself.  I accept their results. 

However, I believe, like you, that LC has changed its underwriting over time in a way that changes the meaning of the grades.

This is the simplest explanation of the loan performance data.  Being the simplest explanation doesn't make it right, but it surely puts it on the short list.

[Rob L]

The implicit assumption in your analysis is that the credit and loan profile of E Grade loans has stayed similar across vintages.

I believe, like you, that they have not stayed the same.  I don't understand how they have changed, but the performance seems to tell us that they have changed. 

Other folks have done analysis of average FICO, average inquiries, ... in each grade vs time, and those studies seem to conclude that the grades haven't changed much.  I haven't repeated those studies myself.  I accept their results. 

However, I believe, like you, that LC has changed its underwriting over time in a way that changes the meaning of the grades.

This is the simplest explanation of the loan performance data.  Being the simplest explanation doesn't make it right, but it surely puts it on the short list.

I've lost all perspective on what grade/sub-grade is even supposed to mean. It's very clear the definition has changed over time and there's nothing consistent about it. Maybe it serves a psychological "anchoring" purpose for LC regarding lenders but in truth there simply is no meaning.

[AnilG]

Shoot... I lost everything I wrote before, my desktop reset on its own. Anyway, the gist is averages don't always tell the whole story. I quickly looked at box plot for a few credit attributes for E Grade 36 month loans. Annual Income, DTI, Bankcard Utilization, and Accounts opened in last 24 months stood out. The inflection point seems to be in 2012 which kind of agrees with your charts. I also seem to recall that these were the time when LC started expanding some of these credit attributed for loan eligibility. Whatever model LC switched to 2012 may not be working.

The implicit assumption in your analysis is that the credit and loan profile of E Grade loans has stayed similar across vintages.

I believe, like you, that they have not stayed the same.  I don't understand how they have changed, but the performance seems to tell us that they have changed. 

Other folks have done analysis of average FICO, average inquiries, ... in each grade vs time, and those studies seem to conclude that the grades haven't changed much.  I haven't repeated those studies myself.  I accept their results. 

However, I believe, like you, that LC has changed its underwriting over time in a way that changes the meaning of the grades.

This is the simplest explanation of the loan performance data.  Being the simplest explanation doesn't make it right, but it surely puts it on the short list.

[lascott]

April statements are on your LC site now: https://www.lendingclub.com/account/lenderDocuments.action

I was negative on my Taxable acct and positive on my ROTH acct ... and overall negative. And I thought I was one of the more conservative investors here.