Print

[RateLadder_com]

Am I reading this graph right that 60% of 451 day old HR loans are/were late?

Yes.

Yes if you meant in terms of dollars and not # of loans.  On a # of loans basis it is slightly less bad...  (larger loans carry a higher risk of default)

Be careful about putting too much weight behind the very ends of the curves... the lines end with only 250 loans in the basket.  I include the ends (1000 to 250 loans) of the curves in order to watch the long tail...

I imagine the the curves will improve slightly over time as the effect of some of the changes catches up with the older loans.

Don't expect too much improvement.

For me improvement in the graphs would be a stable curve...  Until such time as there is plenty of 38 month old loans a complete stable curve will be not be possible...  Hence, I leave the long tail on the curve in order to watch it.

[Staneslav]

a

[Loan_shark_74]

Damn, AA's are approaching/at 20% late in under 2 years.

Not many will make money at prosper...

This does not make sense. If you go to lendingstats, there are only 60 AA loans that are in a status of late through defaulted. There have been 2421 AA loans so far. That is only 2.5%, not 20%. Am I missing something here?

[RateLadder_com]

Am I reading this graph right that 60% of 451 day old HR loans are/were late?

Yes.

Yes if you meant in terms of dollars and not # of loans.  On a # of loans basis it is slightly less bad...  (larger loans carry a higher risk of default)

Be careful about putting too much weight behind the very ends of the curves... the lines end with only 250 loans in the basket.  I include the ends (1000 to 250 loans) of the curves in order to watch the long tail...

I imagine the the curves will improve slightly over time as the effect of some of the changes catches up with the older loans.

Don't expect too much improvement.

For me improvement in the graphs would be a stable curve...  Until such time as there is plenty of 38 month old loans a complete stable curve will be not be possible...  Hence, I leave the long tail on the curve in order to watch it.

Even if you trim the tails back to 191 days, which should be more than enough, the default rate does not appear to be one that could be profitable.

Smoothing should occur, but don't expect too much of a drop in the default rate as more loans mature. It will most likely be a minimal effect.

Keep in mind this is using only credit grade as a filter.  But if you look at my portfolio you will see my bias away from anything lower than C.

I like seeing the tails -- I like being able to begin to see the jitteriness of the curves...  

[RateLadder_com]

Damn, AA's are approaching/at 20% late in under 2 years.

Not many will make money at prosper...

This does not make sense. If you go to lendingstats, there are only 60 AA loans that are in a status of late through defaulted. There have been 2421 AA loans so far. That is only 2.5%, not 20%. Am I missing something here?

For these curves the data of both the numerator and demoninator are reduced as the days past origination increases.  and the curve shown is money lent based not loan count based.

[Staneslav]

a

[RateLadder_com]

Am I reading this graph right that 60% of 451 day old HR loans are/were late?

Yes.

Yes if you meant in terms of dollars and not # of loans.  On a # of loans basis it is slightly less bad...  (larger loans carry a higher risk of default)

Be careful about putting too much weight behind the very ends of the curves... the lines end with only 250 loans in the basket.  I include the ends (1000 to 250 loans) of the curves in order to watch the long tail...

I imagine the the curves will improve slightly over time as the effect of some of the changes catches up with the older loans.

Don't expect too much improvement.

For me improvement in the graphs would be a stable curve...  Until such time as there is plenty of 38 month old loans a complete stable curve will be not be possible...  Hence, I leave the long tail on the curve in order to watch it.

Even if you trim the tails back to 191 days, which should be more than enough, the default rate does not appear to be one that could be profitable.

Smoothing should occur, but don't expect too much of a drop in the default rate as more loans mature. It will most likely be a minimal effect.

Keep in mind this is using only credit grade as a filter.  But if you look at my portfolio you will see my bias away from anything lower than C.

I like seeing the tails -- I like being able to begin to see the jitteriness of the curves... 

Bias nothing. Its hard to turn a profit on AA with a 10% default rate.

Preliminarily it appears that with 2 extended credit filters you can 1/3 the default rate in the higher credit grades... (current dq and inquires)  But the curves for that data is even thinner and hence even more jumpy...

[ira01]

Damn, AA's are approaching/at 20% late in under 2 years.

Not many will make money at prosper...

This does not make sense. If you go to lendingstats, there are only 60 AA loans that are in a status of late through defaulted. There have been 2421 AA loans so far. That is only 2.5%, not 20%. Am I missing something here?

For these curves the data of both the numerator and demoninator are reduced as the days past origination increases.  and the curve shown is money lent based not loan count based.

Furthermore, Prosper "disappeared" 66 loans in the New Agency Test (its test litigation project).  These loans -- all of which should be treated as at least Late-4's (if not defaults) -- are instead treated as if they had never originated, so they currently don't affect Prosper's performance results.  Nifty way to goose the stats, huh?  IIRC, a number of these loans are AA.

[Loan_shark_74]

Damn, AA's are approaching/at 20% late in under 2 years.

Not many will make money at prosper...

This does not make sense. If you go to lendingstats, there are only 60 AA loans that are in a status of late through defaulted. There have been 2421 AA loans so far. That is only 2.5%, not 20%. Am I missing something here?

For these curves the data of both the numerator and demoninator are reduced as the days past origination increases.  and the curve shown is money lent based not loan count based.

If you filter for only the first 4 months, there were 31 AA loans. Only 1 has defaulted. that is a 3.2% default rate. I go by number of loans, not total funds. Plus, the 1 default was a small loan of 2000. If I bid evenly over all 31 loans, I only lost 1/31 of my initial investment. at the average interest rate of 7.99% for those 31 loans, the lender is still ahead. I remember a saying my Grandmother always told me, "never put all your eggs in one basket." If you diversify your bids, you limit your risk. This is a prime example, http://www.lendingstats.com/lenders/weshays. I do not claim to be able to obtain returns in the double digits, like some on here, but I think a lender can maintain a positive return on Prosper. even some who have stopped bidding believe a lender can win; they chose to stop lending because of Prosper's business tactics, or lack there of. People use statistics to persuade others. Politicians do this all the time. For example, 76 AA borrowers take out loans totaling 100,000. 75 of them at 1000 and 1 at 25000. The only default is the 25000 loan. One statistician will say AA's have a default rate of 25%. Another will say AA's have a default rate of 1/76 or 1.3%. Who is right? Who is wrong? For the lender, it depends on how you bid. If you put 50 on all 76 loans, you have lost only 1.3%. If you put 50 on each of the smaller loans, and 1250 on the 1 large loan, you lost 25%.

[Loan_shark_74]

Damn, AA's are approaching/at 20% late in under 2 years.

Not many will make money at prosper...

This does not make sense. If you go to lendingstats, there are only 60 AA loans that are in a status of late through defaulted. There have been 2421 AA loans so far. That is only 2.5%, not 20%. Am I missing something here?

For these curves the data of both the numerator and demoninator are reduced as the days past origination increases.  and the curve shown is money lent based not loan count based.

Furthermore, Prosper "disappeared" 66 loans in the New Agency Test (its test litigation project).  These loans -- all of which should be treated as at least Late-4's (if not defaults) -- are instead treated as if they had never originated, so they currently don't affect Prosper's performance results.  Nifty way to goose the stats, huh?  IIRC, a number of these loans are AA.

This I did not realize. Is there information on this somewhere? Does anyone have a list of these "disappeared" lates?

[ira01]

Damn, AA's are approaching/at 20% late in under 2 years.

Not many will make money at prosper...

This does not make sense. If you go to lendingstats, there are only 60 AA loans that are in a status of late through defaulted. There have been 2421 AA loans so far. That is only 2.5%, not 20%. Am I missing something here?

For these curves the data of both the numerator and demoninator are reduced as the days past origination increases.  and the curve shown is money lent based not loan count based.

Furthermore, Prosper "disappeared" 66 loans in the New Agency Test (its test litigation project).  These loans -- all of which should be treated as at least Late-4's (if not defaults) -- are instead treated as if they had never originated, so they currently don't affect Prosper's performance results.  Nifty way to goose the stats, huh?  IIRC, a number of these loans are AA.

This I did not realize. Is there information on this somewhere? Does anyone have a list of these "disappeared" lates?

There are at least a couple of threads on here somewhere with at least the loan numbers of the NAT loans, and I vaguely remember someone posting the listing numbers of them as well.  You can also get pretty close on Lendingstats, by looking at all of the repurchased loans from California.  Of course not all of these are the NAT loans (for one thing, there are the 11 or 12 identity theft Crawford loans that Prosper had to repurchase, all of which were from Los Angeles).

[NPX]

RL, how do paid loans figure in these curves?  When I look at the performance page for, say, the months from October 2006 through January 2007 (approximating the period at the tail end of the AA curve), I find that approximately 15% of the presently "active" AA loan value is late, but I also see that this represents only about 5% of the originally loaned principal.  What am I seeing on your graphs (pointing me to a link that describes methodology would be fine)?  Thanks.

http://www.prosper.com/lend/performance.aspx?af=0&esba=63&gm=0&gr=0%2c1%2c2%2c3%2c4%2c5&hw=0&iba=255&ibid=0&iwatch=0&lc=0%2c1%2c2%2c3%2c4%2c5%2c6%2c7&lq=&maxAmt=25000&maxDTI=1000000&maxFund=1&maxGrpTLC=1000000&minA=0&minAA=0&minAmt=0&minB=0&minC=0&minD=0&minDTI=0&minE=0&minFund=0&minGrpTLC=0&minHR=0&minNC=0&occ=&od=04%2f18%2f2008&oer=02%2f01%2f2007&osr=10%2f01%2f2006&plcgd=&plp=0&sf=10&sh=0&sn=&tg=0

[RateLadder_com]

Here is the link to the sql I use to generate Prosper Vintage curves...

The data used in the sql is from the Prosper Data Export both public and lender files...  In addition I have a table I added called all days which has an aday column and a row for every day from the beginning of Prosper time until now.

Once I have the raw data from that sql statement...  I do an excel pivot on credit grade and days since origination.  The graph is the result.

The data is also from the beginning of the month...  and the lender table LoanPerformance has a 10 day lag...

[Staneslav]

a

[Staneslav]

a