The Peer to Peer Lending Default Curve

Most p2p investors will likely experience a borrower default at some point. I have been curious for some time to see a distribution of when these defaults occur during the lifetime of a loan. Now, Michael from Nickel Steamroller has produced a chart that shows this very thing.

He has taken just Lending Club data for now and he has mapped every default and when it occurs during the loan term. To make the chart more useful he is only including 36-month loans and has excluded all loans that are less than a year old. The Lending Club chart is below and here is a link to the charts page on Nickel Steamroller which is updated daily.

The chart is probably what most people would expect. The vast majority of defaults occur in the first 18 months, with the biggest spike occurring between month 5 and month 11. And once you get past month 21 there are very few defaults even for the higher risk loans. Also, as expected the C-G grade loans are defaulting in much higher numbers than the A & B grade loans even though they are approximately half of all loans.

The astute reader may question the X-axis numbers. We all know it is impossible to have a default in month one (or month two, three or four for that matter) because it takes 120 days of non-payment to create a default. Lending Club does not record a default date on their data export so what Nickel Steamroller has done is estimate the default month by taking the total payments to date and dividing by the monthly payment.

So if someone had a monthly payment of $100 and the total payments made is $300 dollars and there is a default or charged off status on their account, it is estimated that the default occurred in month four. In other words there were three successful payments before the default.

Less Than 0.2% of P2P Borrowers Are DeadBeats

Since inception through September 6th, 2010 Lending Club had originated approximately 16,000 loans. Of that total just 26 loans defaulted without a single borrower payment. I hear many investors who worry about these kinds of borrowers but each note you own has literally one chance in 600 of that happening.

One point to keep in mind when looking at this chart is that it doesn’t represent all loans. The average age of loans here is around 19 months so there will be some loans that will default that are not included here. So, I will be revisiting this chart from time to time to see if the same trends continue.

Nickel Steamroller is working on Prosper data as well but has nothing ready for release yet. For Prosper investors you can check out the Lendstats charts page which has some interesting charts on late loans at Prosper.

[Update: After I published this post some people indicated they would like to see the above defaults chart in a percentage form, with defaults expressed as a percentage of the total number of loans available at each age period. For example, the number of defaults at 10 months would be expressed as a percentage of all loans that are 10 months old. This gives a clearer indication of the default curve for every loan. What is interesting is that the C-G grade loans have kept the same curve but the A&B loans are relatively flat through month 28. The new chart is below.]

Subscribe
Notify of
32 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Roy S
Roy S
Sep. 6, 2011 3:14 pm

I just have to be the first to say that I do not like this chart. It gives you the total number of defaults per category but not the percentages. Obviously, if there were 10,000 notes issued with a C – G grade, but only 100 for A’s and B’s, then grades C – G have a lower chance of default. Also, how many loans are only a year old versus the entire population on this chart. You can’t create a chart with 50% of loans 1.5 years old and then say that after 18 months defaults drop in half when half of the loans haven’t matured beyond 18 months! I’m just sayin’…

Roy S
Roy S
Sep. 6, 2011 4:29 pm

It appears as though I read over the “approximately half of all loans” on my first read through. I apologize. The chart does give an accurate picture of the first year. Once we start seeing the loans age more I would expect the remaining two years to lift up (relative to the first year) a bit more. So there would not be as steep a decline as we are currently seeing.

I’m wondering whether Michael could use percentages, though. My thoughts are that the first month of the graph would be divided by the total number of loans over one year. Then each month after, the total number of defaults would be divided by the total number of loans that have aged that far. For the first year, the bottom number would only be reduced by the defaults in the prior month, but once you get beyond a year the bottom number would also be reduced by the number of loans that haven’t aged as long as the time-frame. I’m not sure whether this would yield a more accurate picture, though.

Dan B
Dan B
Sep. 7, 2011 1:55 am

Peter………. Perhaps a more interesting distinction would have been borrowers who default within the first couple of payments. Although it certainly is annoying there really isn’t much difference between a borrower who defaults without making a single payment & one who defaults after making a couple of payments. The loss to the lender is between +/- 94% & 100%.

PS………I know you & I define words like “few”, “several” etc. differently. So just to be clear, when I say a “couple” of payments I mean more than 1 but never more than 3. 🙂

Michael
Sep. 7, 2011 9:46 am

@Roy S

In response to your first comment, the chart would stay the same if it was expressed in percentages since all the values would be divided by the same total loan number. As Peter mentioned the idea was tossed around, but I feel you get a better way to conceptualize the data using numbers.

In response to your second comment, that is an interesting idea. There are a lot of moving parts in that approach but that might be one way to “normalize” the data. I really like this because in theory you could include all loans. I would argue understanding the nature of defaults is the most important aspect of increasing returns in p2p lending. Let me work on this tonight and I will update you tomorrow. Thanks for the feed back everyone, and thanks Peter for posting this.

Michael
Sep. 7, 2011 9:55 am

@Roy S
I think I understand now what you are saying. The A+B defaults should be divided by the total population of A&B loans. and likewise for C,D,E,G & F. I believe that if this chart was expressed in terms of percentage that would be the most honest way to present the data.

Let me work on this tonight. Percentage might be the way to do this since then we can get a number relative to the total population of loans in that pool.

Thanks again.

Charlie H
Charlie H
Sep. 7, 2011 10:03 am

Michaels use of the word “Normalize” and DanB pointing out lender loss got me thinking.

Can we see a chart of defaults based on Total Capital Lost?

A loan that defaults after 20 payments have little capital lost.
These defaults don’t “hurt” that bad.
A loan that defaults after 2-3 payments has ~95% of the capital lost.
These defaults hurt and have an outside impact on returns.

Michael
Sep. 7, 2011 10:08 am

@Charlie H.

Take a look at this blog post of mine:

https://www.nickelsteamroller.com/blog/2011/08/default-rate-of-36-month-loans-vs-age/

Would you like to see that chart integrated into the main application? I’m thinking there is enough feed back here to create a whole new part of the application dedicated to default analysis.

Roy S
Roy S
Sep. 7, 2011 2:07 pm

@Michael, First, thanks! Second, I think that actually using all loans that have aged for the certain time period would be better, regardless of whether the loan defaulted several payment cycles ago. After all, a loan that was taken out September of 2010 is still 12 months old regardless of whether it defaulted in June. So the first 12 months should have the same denominator if you are sticking to all loans that have aged at least 12 months. I’m not sure whether this overall strategy would give a more accurate picture of the default curves, but it would at least give another perspective. Both your original chart and my suggestions on a different chart will give a good account of the first year’s default curve. Beginning at month 13, it’s really just a matter which becomes less useful faster.

, I would like to think that something happened that was beyond the borrowers ability to control, but I’m not naive enough to think that there aren’t deadbeats out there trying to con the system. I am curious as to whether any of those were due to identity theft. I know Prosper has an identity theft guarantee, but I don’t know whether LC does or whether it will still include those loans in its statistics. I would also be interested in knowing how Prosper and LC treat those who have taken out loans on their platform and defaulted, especially to those “deadbeats.” Do they ban those people from their platform for a certain period of time? Or do they just calculate that into the loan grade on the next loan they try to take out?

Moe
Moe
Sep. 7, 2011 6:29 pm

Roy, if the loan went into collections there’s no way their credit score would qualify for a loan in the next 7 years.

Michael
Sep. 7, 2011 9:16 pm

@Roy S
https://www.nickelsteamroller.com/analysis/chart#default_age_normalized

@Charlie H
Visit the link above and hover over the data points, it will tell you the ROI for each month. Let me know if this will work or if you would like a table as well.

Charlie H
Charlie H
Sep. 8, 2011 10:28 am

Thanks Michael. Will play with it tonight.

Dan B
Dan B
Sep. 8, 2011 8:37 pm

Peter………..I think that if you go to Prospers.org you will find a number of individuals who would vigorously argue (with supporting evidence) that Prosper should have already paid out on a number of cases.

Phillip Cohen
Phillip Cohen
Sep. 13, 2011 10:00 pm

Hi Peter,

Given this information, do you think it’d be a viable strategy to use the secondary market and only purchase notes that have at least 12 (or 20) completed payments? It’d be more work (and the LendingClub service fee might have a bigger impact — I think) but obviously it’d be well worth if it cuts down your default rate.

Thanks,

Phillip

Michael Phillips
Sep. 14, 2011 6:31 am

@Phillip Cohen

I don’t believe you will have much success with that approach. You may in fact reduce defaults but the amount of interest receive in interest will be much less at around the two year mark. The high interest rates you earn with p2p lending are predicated on the initial “interest only” payments.

There is a blog post about this, you’ll most likely be interested in the summation of the article.

https://www.nickelsteamroller.com/blog/2011/08/total-roi-a-look-at-diminishing-returns/

I’d be curious do hear you thoughts.

Dan B
Dan B
Sep. 14, 2011 9:12 am

Actually it would be fairly easy to set the filters for x amount of months left, a non declining credit score & a “never late” payment record. I suppose if those were your only selection criteria you could even make your selections without having to look at each individual loan. The other thing to keep in mind is that the trading platform is a lot less crowded these days due to the elimination of the “over 70% premium” notes & the implementation of the shorter 7 day listing duration. Those 2 factors have made navigating a bit easier.
All of the above apply to Lending Club, I’m not sure about Prosper.

Phillip Cohen
Phillip Cohen
Sep. 14, 2011 3:32 pm

Actually, I’m not worried about the trading platform because I’ve been developing my own little tool which eliminates the crap from that experience 🙂 (I’m in TX so the platform is the only way I can invest.) It fetches the loans and does most of slicing and dicing for me, so I don’t even see the loans that don’t match my criteria (price >$25, markup up > 1%, etc.) It was pretty trivial to add a slider based on # of completed payments.

I’m not worried about the reduced number of available loans, since my goal is to invest only in the top 0.x% percent anyway. If it takes longer for me to find those loans, I don’t mind.

But, if buying later-stage loans will cut into the ROI, now that’s a bummer. Thanks for the link, I didn’t know the payment proportions changed over time. Guess we’ll have to find the sweet spot on the # of payments to balance between defaults and YTM.

Roy S
Roy S
Sep. 15, 2011 10:34 am

@Phillip, Have you considered asking someone who lives in another state to be the initial buyer and then have them sell it to you on the secondary market? You might have to pay them a small fee, which would reduce your ROI. But you’d at least get to pick and choose your loans prior to the loan being funded. I’m not sure whether this would actually work, but I thought I’d mention it…

Roy S
Roy S
Sep. 15, 2011 12:05 pm

I was rethinking my idea, and probably setting up a trust, business or mutual fund-like operation outside of TX for investors may be a better option instead of having to pay a fee for every transactions. I do not know the legalities of any of these options, but I can see someone attempting this from a business perspective to offer this additional option to states like TX while charging a lower fee and possibly offering a broader exposure of risk than a small individual investor can obtain alone.

Phillip Cohen
Phillip Cohen
Sep. 15, 2011 9:22 pm

I’ve always wondered if somebody created a mutual-fund-like instrument that invests in P2P notes. Unfortunately, the concepts don’t really transfer. The idea of an “actively managed” fund would involve the owner picking loans himself, and there’s a fixed limit of how many loans you can go through in a day and a fixed limit of how much money can go into them.

I’m OK getting everything off the trading platform. It is rare that brand new notes are listed, but honestly, I don’t buy them even when they do appear. Being able to see at least a 3-4 month on-time payment history before investing is worth it to me. Also, if you did buy a brand new note you wouldn’t have to wait for it to fund.

Aside: Given that the earlier payments of a loan are more interest-based, would this be another reason why the ROI for new LendingClub accounts tends to be higher than older ones?

Dan B
Dan B
Sep. 15, 2011 9:31 pm

Philip………On average what type of a premium (if any) are you paying?

Phillip Cohen
Phillip Cohen
Sep. 15, 2011 9:40 pm

None. I buy notes at discount.

Phillip Cohen
Phillip Cohen
Sep. 15, 2011 9:40 pm

(If any premium, it’s usually not more than 2% and then only in the case of a really good loan)

Roy S
Roy S
Sep. 16, 2011 11:33 am

@Phillip, Your aside is partially correct. The other reason is that it takes time for a loan to be listed as a default and ultimately charged off. So no loans under 5 months are really ever charged off. As loans start defaulting, your return starts to drop. And as loans age, more and more of them default. I place the difference more heavily influenced by loans being charged off than due to the interest decreasing over the life of the loan.

LendStats.com
Oct. 21, 2011 12:41 pm

Effective interest rate does not change. Interest rate always stays the same. The decrease in principal balance is what is causing the decrease in interest paid out. If you buy a 12 month old note which pays 12% interest you will earn interest at a rate of 12% on the remaining principal balance for the remaining life of the loan.