Hey pninen, you'll get a kick out of this.

I was re-reading the thread and I saw that 'Prosper Karen' was also viewing it.

So hopefully she abosrbing all your good gripes.

Also, Question: sorry to detract from thread a bit.

But is debt/income on listings worth any valuable information?

When i test drove the borrowing process, i saw that users manually enters their income level with no verification. Also, however prosper is doing the calculation seems to be a low-ball of how extended people they are. for example, I'm almost 100% sure the calculation does not include mortgage payments. I have a fairly expensive, new home and I just closed on a mortgage 4 months ago. During that time my debt/income ratio calculated by the bank which included the cost of the new mortgage and used the averages from all 3 credit unions were at 35%.

propser shows me at 8% on my current listing. so great for me as a borrower, but as a future lender that concerns me

Also, the other thing that was 'weird' was that i relisted a couple of times within day. And my debt/income went from 10% to 8%. you can see this in my listings.

I don't see anything weird. Since the D/I ratio in the posted listing includes the proposed Prosper.com loan, when you borrow less, the D/I ratio is expected to be smaller.

Supposedly, the D/I includes your debt payments except for housing expenses (mortgage), divided by your gross income.

Also, at least on a spot basis, Prosper.com will verify the income you specified before they originate the loan.

Several borrowers have reported on the boards that their loan was cancelled by Prosper and they were invited to relist with their personal income specified.

hey so does anyone have some info on this?

One of my questions is addressed to lenderguy and wisecrack(claus).


How are you tracking individual listings and their payment history? Is this just from your own portfolio of loans? Performance data I see from prosper is only available at aggregate levels.

I'm not doing anything on a loan-specific basis. What eric does is get the data from the group pages.

The best that I can (currently) do is get data for loans issued during a specific time period (year, quarter, month, week, day) and watch them progress from month to month. In a "whole numbers" sense, I don't care about the performance of any specific loan -- what I want to know is general trends. After all, this really is a numbers game.

Really, the underlying question is the one pninen keeps harping on -- recovery rates for delinquent loans. The way I have my program configured, my data is actually quite behind (it was design choice I had to make). But, my analysis supports pninen's assertion that recovery rates are quite poor.

how did wiseclerk get info on individual loans, was he lender on all of those loans?

or everything that is listed on this page:

http://www.wiseclerk.com/reporting-t-7-m-all.html

He reads the prosper web site group pages, where status of loans that are in groups are shown. Same as ericscc does.

pninen,

This is for you. It's recovery rates for loans originating in March, April, and May. I'll tell you how to put this in "format friendly" text, and then I'll explain it:

1. Copy the following into notepad (include the leading comma):

2. Save the file with the extension ".csv"

3. In Microsoft Excel (or some other spreadsheet), open the file. It should paste into a pretty little grid.

Now for an explanation:

What I want to know is how many loans that go one month late become current, how many loans that become two months late become current, and how many loans that become three months late become current.

For loans originating in March, I took observations on June 1, July 1, August 1, and September 1. For loans originating in April, I took observations on July 1, August 1, and September 1. For loans originating in May, I took observations on August 1 and September 1.

For each month, I observed the progression of the lateness of the loans. For March originations, for example, I observed how many loans were 1 mo late on June 1. I then looked at July (for only those loans) and observed the loans that were 2 mo late. I looked at August 3 for 3 mo late, and September 4 defaults. (I also took 1 mo late readings in July, 2 mo late in August, and 3 mo late in Sept. Further, I took 1 mo late readings in August, and 2 mo late readings in September).

If there were 6 loans 1 month late in June, but only 4 loans two months late in July, I counted that as two loans that were brought current. I applied the same algorithm for all subsequent observation months for a particular calendar month of originations, and summed up all of the data. Some of the data doesn't total properly, because in some months, I observed an INCREASE in late loans (ie in July there was 1 loan two months late, but in August, there were 2 loans THREE months late. I haven't figured out an explanation for that yet.)

C 1-2 mo reflects loans that were 30 days late but were brought current before 60 days.
C 2-3 mo reflects loans that were 60 days late but were brought current before 90 days.
C 3-4 mo reflects loans that were 90 days late but were brought current before 120 days.

1 mo late reflects the total number of loans that were ever one month late.
2 mo late reflects the total number of loans that were ever two monhts late.
3 mo late relfects the total number of loans that were ever three months late.

Edited to fix an error I found in pappism's quote

LenderGuy, great analysis, I concur your findings and agree that there is no explanation other than Prosper not having good stat information for the problem you described above.

so approximatley 3/56 that went 1 month late were brought current or approximately 5.36% 'curing' rate?

Also to penin, thanks for the info. so info is only available on group loans at level of detail by scraping the data off ALL the group pages?

Does the petri dish forum have a suggestion that prosper have a loan page that shows loan status for groups and non-groups?

It's actually four loans recovered, so divide 4/56 -- it's that totalling problem that you mentioned in your quote. You can see that by reading the row all the way across.

Will the numbers keep getting worst?


Looking through some of the performance data for march and april originations:

April has 2.6 million in loan originations and over 6% of that is 2 and 3 months late. 150k. If these loans default, only 10-14% is typically recovered from collection agencies?

What happens 1 yr from now, 2 years from now. when irresponsible borrows have more and more easy mechanisms to borrow cash and continue to over-extend themselves.

Will the default rates climb to 10%? 15%? 25%? Maybe the rates being offered are way, way to low if prospers numbers don't trend anyway near experian historical default rates.

This is a thread about collection agency performance (or lack of performance) not default statistics. Please take the default statistics discussion elsewhere.

The correct calculations are more complex than you describe. The fraction lf loans late doesn't mean anything unless you also consider the age of the loans in the portfolio, as late and default events occur over time. Therefore, direct comparison of late rates vs the experian default rates (for example) is an apples vs oranges kind of thing. I've written about this in other threads.

The bottom line after doing what calculations can be done carefully is that we're right in line with expectations, IF something like 75% of the 1 month late loans can be converted to current and thereby avoid default. I believe this is a completely reasonable objective, but it is very different than the collection agency performance to date. Therefore it is my position that attention needs to be paid to the collection agency performance problem.

This thread is kinda deep for me. I switched to NCI as you suggested. I feel you are taking care of this side of Prosper. Is there anything you need from us lenders to help with your stats?

I don't fully understand what you've done, but it looks like you're saying that you found 6 loans that have been brought current after being 1 month late, from among the within-a-group loans that you can observe. That may or may not be consistent with prosper's collection agency statistics, which either show 5 or 9, depending on how you interpret them. You might be inclined to think that your number shows that the correct interpretation of the prosper stats is that the two lines are not cumulative, ie they show 9 loans total. However, I don't have any faith that they are measuring the same way you are. In other words, just because you saw a loan show 1 mo late doesn't mean that loan REALLY went to the collection agency. We don't have any visibility. So your 6 could be 5 to them for all we know. So all I can see in the end is that your numbers are in the same ballpark as prosper's CA statistics.

Is that how you interpret this? ... or am I wandering in the dark somewhere?

Yeah. When I collect my data, as I said before, I do it on a monthly basis. I was actually quite surprised by March loans as observed on Sept 1.

I posted the info as I did, because it's the only way I know how to guage recovery rates aside from Prosper's self-reporting. Either loans are becoming current or they're not. If they're not, we have a collection agency issue that we NEED resolved.

Thanks.

Can you join the protest march in front of prosper HQ at 2PM tomorrow?

Seriously, I would appreciate it if when given the chance, you make known to prosper your concern that the collection agencies are not working properly, based on your observation of the statistics provided by prosper.

Good to see someone else trying to get a handle on this. Somebody, I think it was Eric of ericscc.com, said recently that he was gathering this info daily and hoped to do some analysis at some point.

My biggest problem is how to manage this data in a meaningful way. The nice thing about the way Prosper has set things up is that you can back-gather the daily data by changing the "observation" data.

My data appears to lag, but that's because any more recent data is useless for the most part. If I observe loans for the period of March 1-March 31st on April 1, there are NO loans that could be one month late. If I do it on Apri 15th, only half of the March loans could be one month late. Even if I do it on May 1, I'm STILL not getting any 1-month late loans showing up, although there are some "lates." So I just wait until June 1 for my first data point. Then, I have to pick a monthly reference point and stick with it.

Pninen and LenderGuy,

I wanted to draw your attention to Prosper_Karen's posting in this thread earlier this afternoon.

http://prosper.spreebb.com/index.php?showtopic=7126

If Prosper is in the process of re-thinking how collection stats are presented, this would probably be a good time to buzz her with an outline of how you would like to see it done.

Pninen - curious as to your math/stats background. I am guessing more than just a casual hobby for you?

Thanks. Indeed I have talked with her recently about this.

Getting a bit more clarity in the stats will be nice, but the problem is deeper than that. We've gotta make the collections process work better. At this point prosper folks understand that there are a few bugs and presentation oversights, and they appear to be working to fix those things. However none of them have yet admitted understanding that the collection process isn't working right. ...so my work here is not done.

Math was a hobby from an early age, and work is best when it is an excuse to practice the hobby.

Add this to the conversation and controversy...

Our "Fireman" loan contacted me this morning (spouse) and indicated his wife came clean about the loan to him. We are working on getting the log-in updated so he can take care of this. One thing that made me VERY upset in this conversation was the fact his wife was concerned as NO collection attempts had been made via phone, mail or email in the 3 months they've been late!!!!!!!

This would be my SECOND loan (loan codenamed "reason" gave me a tip a month ago on this issue) that is more than one month late that has confirmed for me that ZERO effort has been given by the agencies to collect on behalf of lenders!

RED FLAG! :angry:

Anyone at Prosper want to touch on this one? Kind of hard for me as a GL to stand by the collection agencies if they aren't even contacting late people consistently (once a week isn't too much to ask). I've done my part to establish contact with late people on a weekly or greater basis, now Prosper has to fulfill their obligation to borrowers and lenders to keep collectors accountable.

Concerned,
Jeff

Makes perfect sense. The collection agencies know that a fair amount of the <90 day lates will self-correct. So, it's better for them to just sit back and consider Prosper a source of passive income.

This is VERY interesting, Jeff. But how do you know it's accurate? Isn't it possible that a collection agency had indeed been contacting "fireman" or fireman's wife and either they didn't tell the other or are not telling you? Even so, it's cause for concern. Now, if only I had a trikke to ride when the rain clears up! KK

This is a travesty if true. In my few weeks on here I would say the collections is the weakest link for this idea. Lenders would be better served to have no collections group since it appears they are just allowing loans to self-correct and then keeping 30-40% of our money. I want to hear a prosper response on this.

I emailed this thread to Prosper Karen. Hopefully they'll shed some light on this.

I'm a little skeptical about the claim of no contact.....

On June 1, two weeks after I sent the recipient of the "fireman" loan a query, I received the following:

"hey jane,

thanks for the kind words. we have been having a few dificulties. payment will be made this weekend."

-- Sadly, that assertion was not correct.

Jane

I missed this post somehow.

I looked for loans that were 1 month late but brought current before they were 2 months late, loans that were two months late but brought current before they were three months late... but yes, the summary is that these are loans that were "1+ months late" (Prosper term) but brought current.

I actually found 8 -- the data is computer generated and has some errors, some visible, some not (which I didn't "touch up").

I can't draw any conclusions as to the VALIDITY of the reported data from Prosper, as my data only cover the time period Mar-May. There is also some confusion to me as to the accuracy of some of the data presented on the performance page. I'm showing that it is possible to have 4 loans 2 months late in July, yet in August there are 5 loans 3 months late. That's something I don't understand, and it impacts the data that I have.

Because of that, I can't swear to the complete accuracy of my data, although I postulate that it is a somewhat accurate relection of the current state of things. Also, you are correct, I have no idea if any of these loans was actually handled by a collection agent.

In some senses, I don't care if a collection agency ever touches them -- either they're becoming current or they're not. However, if a CA is grabbing a piece of the action (fees) and not doing anything about it, there is a problem.

Good luck I can't even get an email response from Prosper.

Do you know the secret handshake?

has anyone heard anecdotal evidence or forummy complaints from any late borrowers about whether they're actually getting hit with negatives on their CR's? that's a *cough* moderate payment performance motivator for some people.

my thought is: who's responsible for dinging naughty borrower's CR's?

Is it prosper - or is it the CA's? if it's outsourced to the CA's, the only real non-performance penalty that exists probably doesn't even have the (milk-)teeth it's been advertised to have...

do we know.. is anyone doing it? are our first few month defaulters being dinged appropriately? how about the 30 lates? do we have any data on how long it's taking the loan tradeline itself to show up on a borrower's CR (it was taking several months at one point there..)? can a late show up before the TL itself does (again, are our first few month lates/defaulters being dinged appropriately)?

And, just to add another voice to the chorus above: Some modest amount of guidance and transparency regarding roll rates and general collection activity are greatly needed...

-t

if you really need things done fast, i suggest bypassing customer service emails and emailing the employees directly.

doing it through the "send customer service an email" channel will not get it done, as they are completely swamped with verification procedures.

My position is that I do care whether or not collection activity takes place on delinquent loans. I don't care who does it, but I do care whether it gets done. All the evidence we have indicates that it is not getting done. That's a problem.

Personally ... I'd rather they focus on the verification ... Let's get some of this $$ originated.

Perhaps a lender should borrow on a $1000 loan and let it go delinquent just to see how it is being handled? Any volunteers?

I have a $3,000 loan out. Don't think my lenders want me to miss a payment. Oh chit! I already made 4 advance payments.

Ms Ava's loan is my most current! It shows 9 payments & the first one isn't due yet ;)

Well I warned my lenders that I was only borrowing the money to see the all sides of the Prosper program. Probably just hold the loan for 6 months so everyone makes a few dollars.

MsAva, what have you learned?

Learned:

1. The better the credit the quicker the funding. My loan was fully funded within 30 seconds of listing.

2. Not everyone is verified after funding. My money was in the account in 2 days. No further verification.

3. Really quite easy to make payments.

4. Tempting to put the money on other loans rather then quick repay.

5. Even with the 8% rate I finally got, it is too expensive for an AA borrower.

6. It is much too easy. Once this well-kept secret is known, God only knows how it will morph.

5. Then you haven't tried to get an unsecured loan from a bank...

6. Good. The more credit worthy borrowers we have, the better.

No. Never tried unsecured at bank. My loans at bank are 0-4% credit cards and 4.5% mortgage.

If you wanted to finance $25k over more than a year, I suspect you'd find the credit card option would be a bit difficult. For a few thousand dollars, I'd probably go the CC route -- but if I didn't have the capability to pay off $25k right away, I'd hesitate to use the cards. Too much at stake if a payment burp/short term cash flow issue gets in the way. Also, the odds are that your utilization would be too high to get mountains of 0% BT offers.

I have open line of credit from real estate. Usually runs about 7%. Haven't bought anything lately, so it not being used.

Oh yeah, forgot to mention... people in extremely high cost of living cities would be great candidates for Prosper loans. They can't afford a house, so they have no HELOC to tap :)

Start small. Condo in up and coming area.

It appears we actually may have a recovery by a collection agency. Stay tuned to this thread:

http://prosper.spreebb.com/index.php?showtopic=7381&hl=

The numbers have change slightly once again. The scandal continues.

(IMG:http://img.villagephotos.com/p/2006-6/1187065/prospercollections15.JPG)

The percentages are now 5.88% and 7.35%. Because these are ratios of integers, and have the same denominator, and there is only one combination of integers that works, we know these are 4/68 and 5/68 respectively.

Previously I was interpreting the cured rows as noncumulative. In other words, I was adding them together to get all loans cured. Yankeefan suggested that they are more likely cumulative, in other words that the "cured in 2 months" line includes loans that were cured in 1 month. Prosper is silent on this issue. There is no help text explaining any of this, and a request to prosper for clarification went unanswered. Therefore I am now interpreting these lines the way Yankeefan suggests.

This tells us: Thare now 68 loans at Penncro. Four of those loans were cured in 1 month, and one more loan was cured in 2 months, for a total of five loans out of 68 cured.

After an average of 42 days at Penncro, only 7.35% of loans have been cured.

I believe this is such a low number that it most likely represents only a placebo effect. After being 1 month late, some borrowers will pay during the next month or two with no prompting at all. It seems reasonable that 7% of 'em might do so.

The results are even worse at AllianceOne, which has never cured anything.

Observing these horrible numbers over several months leads me to the belief that hte collections process is mismanaged, and I ask Prosper to correct the situation pronto. We need 75% of the 1-month-late loans to be cured. Not 7.35%.

Perhaps a better measure than 'cured' is the 'recovery' line. That counts $ recovered. Here the numbers are 6.21% for AllianceOne and 5.56% for Penncro. Because the delinquent loans are split between the two agencies, the overall ratio must be something in between. Perhaps around 5.9%. We need something much higher. This number needs to be at least a factor of 10x larger.

Any time something is off by a factor of 10, it should be receiving a tremendous amount of management attention.

If this is not corrected, prosper is dead. Left unchecked it means the default rates on our loans will be far higher than the Experian suggestions, and we're all charging interest rates that are far too low.

If this is corrected, then there's hope.

But I've been talking about this for several months now, and there are no signs of improvement, no signs of change, no signs of understanding.

That is bad.

Pninen, I completely concur with your findings and appreciate your research. A response from Prosper is definitely needed on Collection agency performance and transparency.

We're not even sure that PENNCRO has done ANYTHING!!!. It's very possible that Prosper is calculating the statistics on their own, based on loans that are late and the default agencies selected for those corresponding loans. That 7.35% might be 100% from borrowers catching up on payments on their own without any collection agency intervention. Better transparency is needed. When did PENNCRO receive the loan. When did they send a letter. How many calls did they make? What was the borrowers response. We should be able to look at their actual loan files for individual loans. If privacy information is an issue there is plenty of technology with redaction features for confidential documents.

What's your thoughts?

Pninen-

Thanks for keeping this front of mind.

It would be interesting, and maybe even useful, to know the results split by how many payments had been made before the borrower went late. The "never paid a dime" crew might well be a tougher nut than the "I paid a couple and then went away" guys.

and thanks for the credit. My suggestion is just a guess as well until we hear from the source.

Thank you.

If we ask for more statistics, they might just give us more statistics. I do not want more statistics.
I want collections.

More statistics would allow you to know in more detail that collections is mismanaged, but you have enough data to know that already. Prosper is our loan servicing agent. And we are captive customers. We can't take our loan and walk. Therefore, prosper must perform. We must insist on it.

An idea whose time has come:

http://prosper.spreebb.com/index.php?showtopic=7462

-t