Print

[leporello]

For the past few months, Prosper CEO Chris Larsen has been claiming that Prosper loans have defaulted at a 2.7% rate, which was technically correct, because the dollar-weighted volume of loans that had in fact defaulted and been sold off was about 2.7%.

As of the December 2007 debt sale, the number is no longer 2.7%, but 5.6%. Anybody who sees the old 2.7% number in the press should be aware that the number is wrong.

The attached spreadsheet shows that the dollar-weighted volume of loans that have now defaulted is 5.6%. It also shows that the Leporello-projected defaults now stand at 12.4%.

The Leporello-projected defaults are an estimate of how many of the loans currently in the marketplace will default. The assumption is that
*   50% of currently "Late" loans will default,
*   85% of currently 1-month-late loans will default,
*   90% of currently 2-month-late loans will default,
*   95% of currently 3-month-late loans will default, and
* 100% of currently 4-month-late loans will default

[ira01]

If anything, this is probably somewhat too optomistic, especially in the 2 and 3 month late sections. 

A much more glaring omission, however, is that you failed to address what percentage of currently "current" loans will default.  Certainly if the Leporello-projected defaults is meant to estimate "how many of the loans currently in the marketplace will default," then you must account for the certainty that some now-current loans will go down the tubes.

[leporello]

If anything, this is probably somewhat too optomistic, especially in the 2 and 3 month late sections. 

A much more glaring omission, however, is that you failed to address what percentage of currently "current" loans will default.  Certainly if the Leporello-projected defaults is meant to estimate "how many of the loans currently in the marketplace will default," then you must account for the certainty that some now-current loans will go down the tubes.

Point well taken. I agree completely. So, let's look at it this way.

12.4% of all loans are headed for default. So let's add 12.4% as an expected default figure for current loans. That brings us to a Leporello/Ira-projected default figure of 21.8%.

[Navy-slush-fund]

For the past few months, Prosper CEO Chris Larsen has been claiming that Prosper loans have defaulted at a 2.7% rate, which was technically correct, because the dollar-weighted volume of loans that had in fact defaulted and been sold off was about 2.7%.

As of the December 2007 debt sale, the number is no longer 2.7%, but 5.6%. Anybody who sees the old 2.7% number in the press should be aware that the number is wrong.

The attached spreadsheet shows that the dollar-weighted volume of loans that have now defaulted is 5.6%. It also shows that the Leporello-projected defaults now stand at 1

The Leporello-projected defaults are an estimate of how many of the loans currently in the marketplace will default. The assumption is that
*   50% of currently "Late" loans will default,
*   85% of currently 1-month-late loans will default,
*   90% of currently 2-month-late loans will default,
*   95% of currently 3-month-late loans will default, and
* 100% of currently 4-month-late loans will default

 It also shows that the Leporello-projected defaults now stand at 12.4%.


Your one scary wabbit....

[fishcreek]

I have 100 loans current out of 129 and counting.

[cowdog]

12.4% of all loans are headed for default. So let's add 12.4% as an expected default figure for current loans. That brings us to a Leporello/Ira-projected default figure of 21.8%.

So at this point, it is fair to state that 20% of all Prosper loans will default.

Yikes.

[ira01]

If anything, this is probably somewhat too optomistic, especially in the 2 and 3 month late sections. 

A much more glaring omission, however, is that you failed to address what percentage of currently "current" loans will default.  Certainly if the Leporello-projected defaults is meant to estimate "how many of the loans currently in the marketplace will default," then you must account for the certainty that some now-current loans will go down the tubes.

Point well taken. I agree completely. So, let's look at it this way.

12.4% of all loans are headed for default. So let's add 12.4% as an expected default figure for current loans. That brings us to a Leporello/Ira-projected default figure of 21.8%.

At first I was going to say that might be too pessimistic, because the loans that aren't currently headed for default might be less likely to default (i.e., "better") than those that already headed for default.  But OTOH, I believe that the average loan age for Prosper as a whole is somewhere less than a year, so there is still a whole lot of time left for loans to continue to go south.  Thus, I think that the 21.8% "Leporello/Ira-projected default figure" is probably as good an estimate as any for now.  Of course, there is probably a somewhat lower default risk on loans originated more recently, given the extra information now available to lenders than long ago, so this number may go down over time (assuming Prosper is still around to originate new loans, which seems unlikely in the medium term of 2-3 more years).

[ducks]

12.4% of all loans are headed for default therefore 21.8% of all loans are headed for default? What?   

[sb92075]

Code: [Select]

mysql> select CreditGrade, count(*) from Loan group by CreditGrade;
+-------------+----------+
| CreditGrade | count(*) |
+-------------+----------+
| A           |     1807 |
| AA          |     1846 |
| B           |     2311 |
| C           |     3200 |
| D           |     3130 |
| E           |     2385 |
| HR          |     2688 |
| NC          |      141 |
+-------------+----------+
8 rows in set (0.07 sec)

mysql> select CreditGrade, count(*) from Loan  where Status like 'Defaulted%'  group by CreditGrade;
+-------------+----------+
| CreditGrade | count(*) |
+-------------+----------+
| A           |       25 |
| AA          |       11 |
| B           |       56 |
| C           |      114 |
| D           |      169 |
| E           |      340 |
| HR          |      633 |
| NC          |       60 |
+-------------+----------+
8 rows in set (0.04 sec)

mysql>  select CreditGrade, count(*) from Loan  where Status like '%Late%'   group by CreditGrade;
+-------------+----------+
| CreditGrade | count(*) |
+-------------+----------+
| A           |       60 |
| AA          |       38 |
| B           |      119 |
| C           |      243 |
| D           |      278 |
| E           |      286 |
| HR          |      409 |
| NC          |       12 |
+-------------+----------+
8 rows in set (0.04 sec)



[ira01]

12.4% of all loans are headed for default therefore 21.8% of all loans are headed for default? What?   

12.4% of all loans are already defaulted or are already late loans that will default (assuming certain percentages of the loans of each degree of lateness will default).  That represents the percentage of loans (by dollar value) that are already headed down the path towards default.  But, we know that some portion of the now current loans will also default.  If we assume that 12.4% of the now current loans will eventually default, that gives us a total of 21.8% of Prosper loans (by dollars) that has or will default.

[LoanChimp]

anyone care to convert this to an annual default rate?

[ducks]

12.4% of all loans are already defaulted or are already late loans that will default (assuming certain percentages of the loans of each degree of lateness will default).  That represents the percentage of loans (by dollar value) that are already headed down the path towards default.  But, we know that some portion of the now current loans will also default.  If we assume that 12.4% of the now current loans will eventually default, that gives us a total of 21.8% of Prosper loans (by dollars) that has or will default.

That assumption seems irrational to me.

[leporello]

12.4% of all loans are already defaulted or are already late loans that will default (assuming certain percentages of the loans of each degree of lateness will default).  That represents the percentage of loans (by dollar value) that are already headed down the path towards default.  But, we know that some portion of the now current loans will also default.  If we assume that 12.4% of the now current loans will eventually default, that gives us a total of 21.8% of Prosper loans (by dollars) that has or will default.

That assumption seems irrational to me.

I think it's actually a bit pessimistic, because for the first year or so of PMI's existence there was no ScoreX floor, and now there is a ScoreX floor of 520, so it's reasonable to conclude that the worst of the worst borrowers are no longer borrowing.

[NewHorizon]

I'm not following why, in the spreadsheet, the "Defaulted (Bankruptcy)" row has 0% and $0 in the "expected default %" and "$ total" columns, respectively?  Shouldn't it be 100% like the "Defaulted (Delinquency)" row?

And is it assumed that the odds of a borrower who reverses a payment which PIF'd the loan and then goes on to default too insignificant to consider?  I would agree that this would be quite rare.  But I thought I'd mention that technically, I think PAID loans still have a 60 day window of defaulting.

[NewHorizon]

The assumption is that
*   50% of currently "Late" loans will default,
*   85% of currently 1-month-late loans will default,
*   90% of currently 2-month-late loans will default,
*   95% of currently 3-month-late loans will default, and
* 100% of currently 4-month-late loans will default

From the spreadsheet, I gather these are percentages of the original loan values, not of the outstanding loan balances at the time of default?  If yes, I think on the whole, the resulting numbers are a bit too pessimistic.