Yesterday I talked to another person in the finance industry who told me she thought 45% was the number we should expect. So that's two knowedgable people who have given me similar numbers now.

So we gotta get the collections numbers from the present 5-7% up to 30-50%, and if we want to do better than average at collections, then higher still.

My thoughts on this-


Having had a lot of experience dealing with collection agencies, I can tell you that Prosper's loans are small potatoes to these agencies, and aren't going to be getting much attention from them - at least not until the accounts get large enough for them to be making the agencies any real money.

Over the years, I've probably turned over 100 deadbeats to various collection agencies, and have had very little luck collecting much money. I didn't have the volume for the agencies to work hard for my accounts. Big agencies like NCI are going to work hard for the Citi and Verizon collection accounts, they're not going to be driving themselves crazy over a few grand that's owed to Prosper.

In time, one or more of these agencies might prove to be good at collecting on Prosper loans, but if you're looking for results soon, don't hold your breath.

30-50% might be realistic for a Citi Visa account that's been sent to collections, but you're going to have some hurt feelings if you think that these collection agencies are going to be pulling 30-50% on Prosper loans.

Yep. Undoubtedly this is the most difficult aspect managing the collection agency relationship. We're small potatoes.

I was reading one of the agency's web sites recently and it said they had 600 seats in their call center. (I assume they had telephones and desks and computers and human employees to go with them, but of course I don't really know.) I'm not sure how many loans per day need to be dumped in the input hopper of this machine to keep it fed, but clearly Prosper only has maybe 1 per day to give, and that is spread among multiple agencies! If I'm the guy running the operation at the agency I can't imagine being excited about the call from Prosper where they give me 1 more loan, and want to take a lot of my time talking about how prosper is unique and I should be customizing my phone scripts to push the p2p aspects of prosper loans, etc. I'd probably feel like I lost money just takin' the call.

Perhaps doing more in house is the answer. Or, maybe prosper can get one of 'em hooked on the concept and the potential, as many of us are. Maybe prosper has to pay 'em extra. There is no question that the current volume isn't gonna excite 'em.

I suggested to a prosper employee that the agencies probably don' t realize that their performance statistics are posted on the prosper web site for 4000 lenders (soon to be 40,000 lenders) to see! I'll bet no other customer does that. For all their other customers, collections is a secretive back-room operation. Would they work harder if they understood this? Depends on how they view the prosper concept I guess.

However its done, its gotta be done. We can't survive on 5% collection.

Isn't it just numbers for the collection agencies?

They don't make money unless they are able to collect, so I would think that a single account is still a revenue stream. Now if over the years when they run their reports and realize that Prosper accounts are harder to collect on, then maybe they will turn away the business.

I have never been on the collection side of the equation, just on the other side of trying to be collected from, so my knowledge is limited.

...Gulp

From that perspective, Prosper ought to be great for them. We deliver loans at only 30 days late, much earlier than most of their business, so they ought to be able to collect more and earn more on them. Our loans ought to be easier to collect.

But easy or hard, there's still the overhead of dealing with a very small customer. No big company puts a lot of effort (mental capacity of best people etc) into a small customer unless the boss is hooked on the potential. Now if we get the guy who owns the company hooked as a prosper lender...

pninen summed it up. The big customer gets the attention, and the small ones don't get much.

For right now, lenders need to be aware that they are the canary in the coal mine, and there will be negative consequences of being in that position.

Years from now, I bet that Prosper will have great success with the collection agencies, and all of these early bugs in the system will be a distant memory.

You gotta realize too that a bunch of these loans are puny HR loans. When the delinquent amounts are less than $100, the most they can collect (I think) is about $20. If a guy's late on his $35 monthly payment, it's not even worth picking up the phone.

The CITI's and AmEx's are seriously past due accounts in the several thousand dollar range. More money to be made.

Check this out...

http://prosper.spreebb.com/index.php?showtopic=7710&hl=

wow, prosper made a change, huh?


actually it makes sense, if there is insufficient collection volume at this time, they should channel it to a single collection agency and see what results they get.

only after they have enough volume should they start giving the lenders "a choice"

Compared to a small credit card company, Prosper is small potatoes. But it is a growth business which is more rare than you would think. Credit cards in this country are fairly mature. And they are getting loans very early. I'm surprised they don't do a better job- some of these ought to be curing by themselves at a greater than 5% rate.

The numbers have changed again, due to Prosper dumping AllianceOne, and moving loans that were in AllianceOne's hands to Penncro. This move makes sense, as (expletive deleted) AllianceOne never cured any loans! Penncro is now the only agency attempting to collect any delinquent loans for Prosper.
(IMG:http://img.villagephotos.com/p/2006-6/1187065/prospercollections16.JPG)

3.92% = 4/102
4.90% = 5/102

From this we know that 102 loans have now been sent to Penncro, and that 5 have been cured. (Previously there were 74 at Penncro, so they must have transferred 28 from AllianceOne.)

Prosper recently confirmed that the "cured in 2 months" line is cumulative, in other words it includes the count of loans cured in 1 month. Therefore we don't need to add the two lines together. Prosper has already done it for us. One obscure detail... Prosper presently leaves loans with the collection agency for 3 months, yet there is no "cured in 3 months" line. Also as a prosper employee recently mentioned in another thread, they are leaving some loans at the agency for longer than 3 months. So it seems that some loans might not be counted in the two lines shown. An oversight they'll eventually fix. My guess is that there is 1 cured loan in that invisible category, but I can't be sure.

Bottom line: The cure rate is now about 5%. I continue to believe we need it to be 8x to 10x better.

Will the cure rate improve by itself as the delinquent loans "age"? Some people have tried to tell me it will, but I see no evidence of this so far.

Will the cure rate stay the same? If it does, it seems to imply that the default rates are going to be much higher than the Experian numbers, and as a result my loan portfolio is going to have a much lower return than I had hoped. If so, I don't see how I can continue to invest here. The numbers just don't work.

Will some magic occur which changes the entire process? If you think you know how to make this happen, please call Prosper and tell 'em!

Why are the cure rates so bad? Does this say something about our borrowers? Or something about our lenders? Or something about the collections process? Seems like the first step in finding out is to work the collections process hard.

I continue to scratch my head about how to use these numbers in a carefully analytical way. The loans at the agency have a mix of ages (time at the agency), but we want to know what fraction will be cured during the whole 90 days. I don't quite have a method for doing that yet. The average age seems to not quite be enough information to use as a normalizer. Might be possible to discover the age distribution by milking the information from the "performance" page. Also we should be careful in the near future not to penalize Penncro too much for those 28 loans they got after AllianceOne ignored them for a couple of months. Real life data is always messy. But even after various adjustments, I don't see how I can possibly get to a cured loans fraction that would produce default rates that look anything like the Experian numbers. :(

pninen,

I hear what you're saying. However, for these numbers to mean anything, I need to see a Markov analysis of the changes of payment state. IOW, I need to know how many loans from a traditional bank go 1 mo late to 2 mos late to 3 mos late to default (or become current and at what point).

The other thing that makes the Experian default data useless in comparison to late payment tracking is that a traditional lender won't send a collection agent after a borrower until the borrower is 90 days late.

Finally, I have my own somewhat restrictive SO criteria. I am happy with my own portfolio performance, and I am happy with the Prosper-wide performance of loans that fit my characteristics.

Referred to as "roll-rates" (just migration analysis), but, even there, as you point out, there's not really a comparable credit product to take roll-rates from to use as any kind of a standard for Prosper... at least if you consider it to be a uniquely new asset class...

Anything you did find would likely be about as correlatable as the existing Experian default data, don't you think? Or can you think of a model that might be a better fit (and have suff. stats available) for this type of unsecured sub-prime lending than bank card products? I've tried, but failed.

Maybe, at the least, Prosper could ask Experian for the roll-rates belonging to the default data they've already given us, so that we'd be a little more informed...

-t

I keep hoping you'll do it.

Lemme see if I understand your thinking. Are you thinking that if we knew the state transition probabilities (between states (current, 1molate, 2molate, ...)) at a traditional bank, we could apply same here, and use them to get from the late rates we can observe to the prosper default rates which we can't yet observe?

As for the state transition probabilities on prosper, almost nothing has flipped back after being 1 month late, so the numbers aren't very entangled yet.

Again not 100% sure I get your drift. Are you saying that the lenders who participated in the Experian data set likely had more patience? ...and if so that our default numbers will be worse, just due to the lack of patience? 'Cause if that's so, then we should adapt and have more patience!

I have 395 loans, with an average age of 2.2 months, and 7 of 'em are >1mo late. Now is that good or bad? That's 1.77% late. Doesn't sound bad all by itself.

To relate that to an annual default rate, we have to annualize and then multiply by some fraction we expect to not be cured. Here's a hack: Annualizing the 1.77% using the 2.2 month average age produces a 9.29% annualized late rate. Now if I compute an expected default rate for this portfolio using the Experian default rates for each credit grade, and the number of loans in each grade in my portfolio, I get a 3.3% annual default rate. To get from 9.29% annual late to 3.3% annual default would require a 64.4% cure rate. This is so much higher than our present 5% cure rate, that it makes me think we're in trouble.

I will admit that there are some handwaving steps in there. Many things could be wrong with this extrapolation. Maybe the annualization of lates is bogus. (I could do something more firm perhaps with the markov model you mentioned.) Maybe I had a run of bad luck. The loans that are going late now in my portfolio are mostly from very early, before we had better credit data. Maybe those loans aren't representative of my portfolio as a whole. Also the above calculation is based on loan counts. Dollar weighted I'm doing better.

Do you have any lates? Ericscc doesn't show any for you.

nonattender,

Yeah, we're messing around in uncharted waters. The reason I don't like the "bank card product" comparison is that the structure is 100% different. We've all heard the stories about how if you pay you min payment only, you'll pay hundereds of times your principal balance in interest paymetns and it will take 20 years. If you default halfway through that 20 year period, the bank has still made a killing off of you. In that respect, it really doesn't matter that you defaulted -- the bank has still made a profit off of you. Same goes for those PDL companies -- you pay those rollover fees every month for a year and then check out, they've still made bank and then some.

Nobody does fixed-term amortized subprime loans. However, I still would like to see roll rates for prime grade fixed term amortized loans. Like I said, running my SO criteria through the performance page returns results I can live with.

I will when I have sufficient data. As you allude below, "almost nothing" has flipped back yet. You saw some stats I've posted earlier -- we have a whopping one or two loans become current. The number of loans getting cured or whatever you want to call it is greater than zero, but not by much.

I think we might be able to draw some general comparisons, but nothing more. IOW, we might be able to hypothesize that half of all 1 mo lates from Experian's data will actually default, the same could apply here. But what the exact % of one month lates will be is uknown. I'm trying to get a handle now on the actual Prosper state transition probabilities.

I'm saying that the lenders who participated in the Experian data set have their own internal collections (aka "accounts receivables") that handles delinquent accounts before they get turned over to a collection agency. It's not that Prosper is "impatient" but that they outsource what other companies have the infrastructure to handle internally.

However, I do think Prosper's policy of selling off the account at 120 days is industry standard. Don't pay your credit card bill for a couple of months, and see how long it takes them to shutoff your credit line. Then see how long it takes until the account is closed. OTOH, everybody does things a little differently.

But no, I don't think that our numbers will be worse due to lack of patience. They could be worse for other reasons, but I don't think lack of patience will be it. Prosper Andrew (in another thread) mentioned that they're not agressively sending loans to the jdb market -- if they think that they can still recover them, they are going to try. Chris Larsen alluded to something similar at the SoCal M&G.

Not at all, presuming you're not "averaging" AA and A loans. If you want to sell your portfolio on the secondary market when it becomes available, contact me, I'll be interested (but I might have to take out a loan to do it).

Here's the point I DON'T try to drill home (I need to compete with a few scared lenders): The most important comparison of your personal late/default rate should be with your personal returns. If you're averaging 25% return on your loans, you can definitely afford 5% defaults. My personal backwards (but not really handwavy) math suggests that I can fare worse than the Experian default rates indicate and still make a healthy return. I NEED a handful of defaults and some scared lenders to keep the rates up.

Honestly, I'd like to see some complete cycles (ie wait three years) before I come out with some "complete" data. Some people suggest that the bulk of defaults will occur towards the end of the loan cycle. If that happens, and the rates are in line with the data that Experian has provided us, we'd be doing very, very well. I suspect that the defaults won't be skewed towards the end, though. Why? If you get to the last few months of payments and hit hard times (or tough times anyway) would you really let the Prosper loan default? It will hurt your credit just as much as if you would have let the entire balance default. That type of logic doesn't apply to "bank card products" because there really is no end in sight for that type of credit when the borrower is "barely getting by. "

I have one <15 and that's it. I've just ramped up my monthly transfers, so of the 36 you see, 21 are > 30 days old. I have 11 loans coming up with their first payment by the end of the month, I just had four originate, and I'm waiting for 7 more to originate (or finish verification). So, the next couple of weeks and the next month will be a bit of a nail biter for me -- these "no payments yet due" loans are mostly SO's.

I have what I thought was a relatively conservative portfolio. Based on the current cure rates, essentially everything I have over 1 month late is a goner. Once I write off 95% of that, my overall portfolio return drops to around 4%, even without counting other expected defaults (because I cannot imagine that there will not be more). If only 95% of my >1 months are collected, my expected return becomes negative despite the fact that I have an estimated ROI of ~10.5% and an average yield of ~14.75%.

This is fun, but it is not worth paying to play.

The more I think about these collection statistics, the more I think they're biased downward by a significant number of borrowers who never intended to repay, or who knew they would be lucky to make more than a payment or two. The collection agencies can't do much if there's nothing to collect.

It will be interesting to see whether 1. loans that make the first few payments are less likely to go late than new-minted loans, and 2. loans that go late in the later payments are more likely to be "cured" than loans that go late in the first few payments.

so you're saying that the "scam rate" is a significant portion of the current delinquent rate?


--does anyone have any idea how to determine the payment histories of the delinquent loans from Ericss page? In other wordes, which of those delinquent loans failed to make the first payment?

You can back into it based on the expanded credit data (okay, you can't due that for originations prior to Apr 24) and figure out when it orginated. Odds are that there is only one loan that matches that specific criteria. Once you find the origination date, you can watch the loan at various observation dates. It's not a quick way to do it, but it works.

I'm talking about something looser than the strictly-defined "scam rate" (borrowers who never make the first payment). I'm thinking more along the lines of borrowers who didn't want to not repay, but who didn't have a plan in place to pay back the loan and just sort of hoped to buy time until they got lucky. The financial equivalent of burning the furniture for heat -- you're almost certainly screwed, you don't want to do it, but at least it will buy you a few days if you have nothing else left. I think there were a lot of people like this borrowing in the early days, and because we couldn't see their current delinquencies, we didn't realize how bad things really could be.

Or, I could just be wrong.

according to erics, out of the 98 group loans in total, over 31 (some are over 3+months late but remain at that level) never made first payment.
and only 4 of those 31 are from before the expanded credit era, which means there are quite a bit of dishonest borrowers here or gullible lenders, or combination of both.

[edit]: actually many of the 3+ months are from the "wild west days" but I can't tell if they have made at least one payment because prosper does not "default" them after 3 months

Huh?

Out of those 98, I count 21 that are default or 3+. Do you mean 31 of 98 that are late on eric's never made first? As in, there are 10 that are fairly recent which didn't make the first payment?

31 of 1632 would be a "scam rate" of ~2%. Assuming some of those 1632 haven't become a month old yet, call it 3%.

Hmmm.

sorry for my bad math. (trying to juggle 50 things at once while hypercaffeinated.)

yes you are correct about the 2-3% scam rate. What I meant to suggest was that the numbers suggest your assessment about the downward bias. A good portion of the lates were lates the moment they listed.

So if 1/3 or more of the lates are scam listings, and my assumption that CAs have no chance of getting them to pay up is correct, then that increases the Penncro "cure rate" to about 8-10%.

Still crappy.

but on the other hand, doesn't that mean good news for lenders when a borrower makes a first payment?

Speaking of roll rates, I was reading one of the zopa blogs, where Zopa Dave said...

Well, starting with loans that are 1 month late, we multiply 50%x60%x75%x90% = 20.25%, so he's saying that of the loans that go 1 month late, 20% should default, in other words 80% should be cured.

80% cured is very different than Prosper's 5% cured statistic.

How can this be? Something is very different at Prosper.

I found this picture tonite. For those of you who were wondering what lenderguy has been talking about when he has used the term Markov models, this picture is worth 1000 words.
(IMG:http://img.villagephotos.com/p/2006-6/1187065/markovmodelgraph2.GIF)

I found it in http://www.philadelphiafed.org/pcc/confere...eph_Breeden.pdf

Well, another week has gone by with no collections.

(IMG:http://img.villagephotos.com/p/2006-6/1187065/prospercollections17.JPG)

3.64% = 4/110
4.55% = 5/110
So there are now 110 loans at Penncro, and they still show only 5 loans cured. In other words, no progress that we can see in the last week. I hope there is some progress that we cannot see, because otherwise the situation is just depressing.

Last night's maintenance gave us some enhanced agency stats (but no more collections :( )

Why does Penncro have $45,000 in loans so far?

If they have 110 loans, that works out to ~$420/loan, smaller than the Prosper minimum.

$43000 worth of accounts spread across 116 individuals is beyond insignificant to a big collection agency. Our recovery numbers won't scale and there's insufficient leverage to improve the situation. As for the quality of the debt, we frequently handed out lumps of cash to people who had been turned down elsewhere. That means we took on scammers with bad intentions or more commonly gave too much money to well-meaning people without resources to revolve such amounts. We took chances the people who compiled the Experian tables wouldn't take. It's not recoverable.

That's not what people want to hear but it's real life. We gave money for car loans but can't repo the car. We gave money for tuition but can't lock the transcript. We gave money for housing but can't evict. We gave money for businesses but have no claim to assets. I know this thread is filled with intelligent people but have you thought of how naked we are?

I agree Prosper should be doing more and it was good to consolidate the action by removing Alliance One from the equation. I'd like to see Prosper report loan status to all three credit agencies (not just Experian) in a timely manner. That's the best leverage we have of someone eventually making good. There are definitely collectable loans out there just dribbling away due to mismanagement, but the majority of our cash is long gone and IMO we need to look at other mechanisms used by traditional creditors.

I think they're only trying to collect the overdue payments- not the total loan. In the loan agreement, Prosper says they will not accelerate the loan until it is 3-4 months old (I forget which).

Great to see the improved/clarified display of collections data. It is clearer. I think more lenders will be able to "get the picture" now.

Picture is of course the same: Almost nothing being collected.

Have you noticed that loans are recovering nicely from "<15 day late" and ">15 day late" states. I see loans in my portfolio pop in and out of those states every few days. But once they get to 1 month late, we're screwed. Does this say something about the borrowers? ...or about the collections process?

Well, a couple more weeks are under the bridge, so I thought I'd sample the collection situation again. The new display has more information but they left the date out, so note that this picture was sampled on 10/07/06.
(IMG:http://img.villagephotos.com/p/2006-6/1187065/prospercollections18--2006-10-07.JPG)
Of the 142 loans that have gone to collections, a total of 8 have been brought current.
In 1st month 7/142 = 4.93%
In 2nd month 1/142 = 0.7%
In 3rd month 0/142 = 0%

A common argument I have heard is "well you have to let them sit there awhile." Obviously many of the loans in the collection agency's hands have been there a relatively short time. It takes time for the agency to contact people, extract promises, etc. So maybe we're being harsh by including those loans that have freshly entered the collection agency's portfolio. I tried to figure out how to avoid counting loans that had been in collections only a short time, until I realized that was fruitless. I could invent a new denominator which excluded loans that have been in collections only a short time, but to be consistent I would need to exclude those loans from the numerator as well. It never makes sense to compute a ratio with overlapping sets of objects counted in the numerator and denominator. I have the ability (although it is painful) to find out when each loan went 1 month late (and therefore should have gone to collections, and be counted in the denominator) using prosper's "performance" page, but I don't have any such ability for the loans that the collection agency has "fixed", and therefore go in the numerator. Of course there is one date for which I can bifurcate the data. That's at the point where loans have been in the collection agency's hands for 1 month. I can easily exclude loans that are fixed in the first month. However, as you can see, almost all of the loans fixed so far have been fixed during that first month! Excluding the first month would therefore give us a very small ratio. The "have to let them sit awhile" theory is apparently wrong. The next guy who tries to give me that argument had better have some quantitative data to back himself up.

So here we sit. The collections situation, if not improved, will result in defaults far in excess of the Experian numbers, on which many lenders have based their bids. If this continues, a great many lender's portfolios are going to be unprofitable. Prosper may not survive.

Is the situation improving? No. Here's a chart of Prosper's collection stats vs time, from the images I've captured. These are all Penncro's stats, as no one else has ever collected anything.
(IMG:http://img.villagephotos.com/p/2006-6/1187065/prosper-collections-rates_29592_image001.gif)
The big dip in early September occurred when Prosper fired AllianceOne and transferred their loans to Penncro. That increased the denominator, but not the numerator (because AllianceOne had never collected anything). Penncro's stats naturally recover over time as more freshly late loans are added.

To get a proper perspective, we have to change the vertical axis and add a couple of landmarks. If you've read my earlier comments you know what's coming.
(IMG:http://img.villagephotos.com/p/2006-6/1187065/prosper-collections-rates_30287_image001.gif)

Many questions should come to mind:

Q: Is anything happening that will improve these numbers?

A: I don't see anything coming that will improve the numbers.

Q: Are these bad collections numbers a result of incompetent collections activity, or lender behavior? In other words are we dealing with bad loans or bad collectors?

A: This is the most important issue, and I don't have an answer.

I do have one clue. As I posted in another thread, the late rates for loans originated in March and April '06 are consistently running about 1.5x the late rates for loans issued in May, June, July..., when compared on a consistent basis, ie same number of months since issue. Therefore it seems likely that either the improved credit data which Prosper made available to lenders, or else the improved verification that Prosper instituted around that time, resulted in a measurable change in the quality of loans. This gives us a hint that additional improvements might be possible with improved informaton and/or verification. Indeed, such improvements might be necessary for the continued existance of Prosper.

Meanwhile, Prosper contracted with the collection agencies, so Prosper is in control of their level of activity, and has the power to make it anything they want. A higher level of activity may also be necessary for the continued existance of Prosper.

My impression is that much of the low level of collection activity is due to bad decisions by lenders. Given the significant differences in performance within credit grades when one controls for variables like the number of current delinquent accounts, I think a lot of loans were made on Prosper which would not have been considered by more experienced lenders.

As our collective level of experience rises, I would expect the collection rate to rise, as fewer doomed loans are made.

I'm planning on tracking late rates for each month of originations at the 6 month mark to see if there's a learning curve effect, but it'll be another 6 months to a year before there are enough data points to mean much...

Have you seen my attempt to do this? Charts and explanation in this thread: http://forums.prosper.com/index.php?showtopic=8842&hl=

There is a distinct difference between the Mar & Apr originated loans vs the later months. Other than that, no experience effect is in evidence at the all-Prosper level.

This article on loan aging indicates that we may have been too aggressive in lending. :unsure: A few caveats: this article is from 2002.

I did see that. I was thinking of making the data points the late rate at the 6th month point (and probably the 1 yr, 2 yr, and 3 yr) for each month of originations. So just a single line, rather that plotting out each month as you have done. But so far, I just have the March 2006 point, which makes for a rather boring chart.

I would like to know more about the loans that went late, then were cured.

I would't be surprised to hear that they have better credit numbers than the uncured loans, small sample size be damned.

I don't know any way to figure out which loans those were. Sometimes a loan number slips out in conversation, as when Prosper sent various folks email about the software problem on the late payments on #515, but other than that I think this data will be hard to find.

And oh it is a small sample. Eight loans.

As a general statement this seems way too high. Maybe AA & A loans that are only slightly late will be turned around. Lower grade loans that are behind more than 30, 60, etc days late will not see a 75% collection rate.

Maybe I do not get the point being made. The industry averages for collections does not show 75%.

What am I missing or are some people mis-priciing based on unrealistic expectations for collections?

I expect that if the collection rate if going to be average or above (and you seem to want way above) the Prosper lenders will need to using similar criteria as the traditional lenders.

A bit of garbage in garbage out if the loans that were funded were bad to begin with.

When you talk about 75% or other numbers above the average for collecting unsecured loans that have defaulted what are you assuming about the quality of the decisions when Prosper lenders make the loan?

Excuse me, but I am not sure we are talking about the same things. Lets look at your sentence above.

You say "for collecting unsecured loans that have defaulted"...

I am not talking about loans "that have defaulted". I'm talking about loans that have become "1 month late". Specifically, I'm talking about what fraction of loans that have become one month late should we expect to be cured, so that they do not default.

If you have some industry data which shows in some other context what fraction of loans that have become 1 month late one should expect to recover, vs what fraction one should expect to default, please share it.

I showed typical roll rates I got from Zopa Dave in this posting http://forums.prosper.com/index.php?showto...indpost&p=87282 . They multiply out to show that 80% of loans tha go 1 month late should recover before our cutoff at 4 months late.

In this post http://forums.prosper.com/index.php?showto...indpost&p=70668 motleygunner says he owns a collection agency, and he says "These agencies should be recovering 30-50% of this bad debt."

I had a telephone conversation with a person in the collections business who would probably rather remain anonymous, who told me she would target 45%.

If you have different industry data, please share it.

It may well be that "prosper is different" for legitimate reasons. If so, we need to understand them, because if trend continues, the prosper averages will blow past the Experian default rates by maybe a factor of 2.

I work in the Banking industry and deal with collections and workouts for business loan as part of my role. I also have a brother who works as a collector on credit cards for Wells Fargo. In my knowledge of the industry if the % below were based on loans sent to them 60 or especailly 90 days past due I would like it is pretty good for unsureced debt, but since they get the loans at 30 days this is horrible.

Agency: Penncro Associates, Inc.
Collection fee (1st mo): 15%
Amount collected (net): 6.7%
Accounts cured: 5.63%

One question I have, if the customer becomes current is it then transferred back out of the collection agency, and if so how do they get their % (15% if collected fee if cured in month 1) Sorry if that was discussed already.

I believe the fee is simply deducted from the payment they collect, and the remainder is then distributed among prosper (for its fees), and then to the lenders.

Thanks, so technically each lender would take a 15% loss on their % of that payment. That's not so bad if they would have a 25-40% cure rate.

I am not sure that comparisons to other lenders and collection processes are valid.

The large firms run in-house collections in many cases.

The external agencies collect for different lenders. How much of the commentary about collect rates applies to a Prosper borrower is unclear. There are different types of loans, different lender policies, etc. Debt for a medical bill vs. debt for a car, etc.

I suspect (pure speculation) that the ability to collect will vary per loan based on the quality of the lender review. Some of the loans that have been funded would not be funded any where else. Hence the lending community has judge the borrower higher risk than the lenders who funded a loan here at Prosper. If too many naive lenders fund enough loans the collection process will reflect this in lower recovery figures.

While I believe the debate is healthy I am not sure there is enough history and enough detail to draw conclusions at this point. We know that a number of lenders are hear to learn, that they expect to have complete write offs on loans and that state they are prepared to lose all the money they have brought to Prosper. Unlike a traditional lender they are saying they can handle 100% loss and still consider the opportunity to participate as worthwhile.

Hence having such people participate in the loans that form the basis of the collection debate leads me to think that we will have higher defaults and write-offs.

We will likely have a bit of a shake out where the lenders who are hear to do good or to have fun while learning pull back when they lack the capital. Then we should see the ratios trend back to what other lenders can produce.

Comments?

That's a poor excuse. If an in-house collection department is better, then that's what we should have!

Above I provided numbers I got from three individuals who have industry experience and are intimately familiar with Prosper. Its not like I went to another planet and got numbers from loans on Abelian Zorks. If you have better data, please share it.

Difficult issue. I hope you would agree that while this may apply to SOME loans, there are surely some other loans that are of high quality. We have a mix. If you hypothesize that we have half of each, and the good loans should see a cure rate of 45% while the bad loans should see a cure rate of 5%, then the average would come out 25%. Unfortunately the cure rate not 25%, it is 5.63%. What assumptions would one need to make about the mix to get 5.63%? Answer: Really ugly ones.

Nobody in his right mind is here to lose their money. Therefore, when there's a problem ya' gotta try to fix it. To just ignore it and say well everybody planned to lose all their money anyway is nuts.

A week has passed, so here is another archived snapshot of the collection statistics. This image was grabbed on 10/15/06.

(IMG:http://img.villagephotos.com/p/2006-6/1187065/prospercollections19--2006-10-15.JPG)

What has changed in the last week? Very little. No new loans have gone to collections. The performance stats page says a some loans went past 1 month late this week, but apparently they didn't make their way to the collection agencies.

The cured numbers haven't changed, but the gross amount collected did go up from 6.88% to 7.73% .

$50,140.98 x (7.73% - 6.88%) = $426.20

so Penncro collected $426.20 this week.

Have lenders seen this money arrive? Apparently wasn't on any of my loans.