Print

[Fred93]

Fred93 has updated his blog with latest Prosper late loan statistics.

The main course is the updated chart...

[gelt4u]

well the breaks are becoming clearer

[mhs505]

What are the breaks again?  When something significant changed?

[Mark12547]

October 30, 2007 introduced, among other things:

Improved bidding guidance: When a lender places a manual bid, Prosper will subtract estimated losses and fees from the lender's bid rate to give a better idea of the estimated return for this loan and others of its kind.

See What's New on Prosper? and scroll down to October 30, 2007.

That might explain why the November and December loans show a far flatter slope than any of the months that preceded them. That does look like good news and hopefully will translate to a better lender experience.

I still have issues with Prosper, but I think they made a good step in the right direction with the improved bidding guidance.

Thank you, Fred, for continuing to keep track of these numbers! That's greatly appreciated!

[Mtnchick]

Thank you, Fred, for continuing to keep track of these numbers! That's greatly appreciated!

+1

[112233]

fred, did you categorize the test loans as repurchased (as Prosper said it would)?

.. not that it would necessarily make a visible difference in the chart, but I am curious anyway

[Fred93]

fred, did you categorize the test loans as repurchased (as Prosper said it would)?

Explained in more words in the blog entry.  Prosper set them "repurchased", hence missing from the performance page, but I added 'em back.

[LoanChimp]

If it weren't for the f-in maintenance window right now, I'd be working on some goodies for y'all. I'll post them tomorrow... 

[LoanChimp]

If it weren't for the f-in maintenance window right now, I'd be working on some goodies for y'all. I'll post them tomorrow... 

Maintenance window is over, but I ain't gettin' nuttin on the performance page...   

[Staneslav]

a

[DakotahFury]

I hate to be a stickler for all things statistical but this needs to be said:

The graph above is a representation of the entire population of loans on prosper.

The population has been slowly changing over time, most importantly by credit grade, which we know to have a profound effect on default rates. Since we know that HRs ahve a higher default rate, we would expect that if we took the averages of two month's loan originations, the month with the larger amount of HR loans (all other thigns being equal) wouldhave the higher default rates.

Simply put, fewer HRs are being orginated these days, which is going to drop the average default rate of the more recent months.

This does not mean that defaults are decreasing overall. A stratification by credit grade would remove the undesired credit grade weighting that is naturally occuring in this graph and would demonstrate whether or not default rates for AA and A's are changing. Again this is cruical, because the population features are changing over time. We can no longer simply look at Oct 06 and Jan 08 and say that an overall comparison is valid.

I do not mean to belittle anyone's work or efforts, I simply think additional information is necessary to keep the desired picture clear.

Would it be possible to get a reading of the average credit-grade during these months, like the "C-, 4.02" readings we get at LendingStats? Another possibly is the average interest rate. That might give a better picture of how the risk/reward ratio is changing in the market, and how that has changed the late-rates as a result.

[Staneslav]

a

[zcommodore]

Late loans by credit grade are, due to the smaller sample sizes, very noisy.  Some efforts towards this goal have been attempted as can be seen here though:
http://www.prospers.org/wiki/Data/LatesByMonth

I've got more that haven't been posted but, as I said, they are very noisy.

[Staneslav]

a

[Urbi_et_Orbi]

It would also be interesting to over-lay the data (if broken out by credit grade) with historical average interest rate by credit grade.