Author Topic: LC Assigned Interest Rates and Probability of Default, D&E Grades  (Read 2874 times)

Rob L

  • Hero Member
  • *****
  • Posts: 2028
    • View Profile
It's been a year since I updated this chart. These numbers are from LC, not me.
So, FWIW:



Fred93

  • Hero Member
  • *****
  • Posts: 2129
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #1 on: April 29, 2017, 08:08:32 PM »
Beautiful.  Prob of default has gone way up.  Interest rates haven't.  Elephant in the room.

Fred93

  • Hero Member
  • *****
  • Posts: 2129
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #2 on: April 29, 2017, 08:48:54 PM »
Rob, which data source are you using?

Rob L

  • Hero Member
  • *****
  • Posts: 2028
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #3 on: April 30, 2017, 12:03:08 AM »
Rob, which data source are you using?

For quite a long time now I've kept an archive of the "feeding time" browseloans.csv files.
The files contain the columns INTEREST_RATE, EXPECTED_DEFAULT_RATE and SUBGRADE that I used in the chart.
I think this is the only source for EXPECTED_DEFAULT_RATE as it is not preserved in the loanstats files.

Fred93

  • Hero Member
  • *****
  • Posts: 2129
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #4 on: April 30, 2017, 02:19:05 AM »
For quite a long time now I've kept an archive of the "feeding time" browseloans.csv files.
The files contain the columns INTEREST_RATE, EXPECTED_DEFAULT_RATE and SUBGRADE that I used in the chart.
I think this is the only source for EXPECTED_DEFAULT_RATE as it is not preserved in the loanstats files.

Brilliant. 

I have been aware of that column, but have ignored it.  I don't think that number appears anywhere else, and I don't think they define exactly what calculation they're doing anywhere, ie what exactly that number means.  I always start by presuming that the word "rate" means something per year, tho that is uncertain. 

They do publish regularly in 8Ks, and in the monthly chargeoff stats file, some annualized net chargeoff rate numbers by grade.  (not by subgrade)  Those numbers don't match the numbers you have, so this area is a bit confusing.

None of this confusion detracts from your chart.  Whatever they mean exactly, they've gone up, and interest rates haven't gone up to match.

rawraw

  • Hero Member
  • *****
  • Posts: 2757
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #5 on: April 30, 2017, 09:15:14 AM »
It's been a year since I updated this chart. These numbers are from LC, not me.
So, FWIW:


Very interesting chart. The one caveat I would have is you need a benchmark line of what is acceptable rate per risk (which differs from investor to investor).  It is a great way of showing trend, but there are situations where things are mispriced too high.  In that situation, the decline sucks but that high rate also cannot be expected to continue
« Last Edit: April 30, 2017, 09:17:06 AM by rawraw »

nonattender

  • Hero Member
  • *****
  • Posts: 713
  • I am not here.
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #6 on: April 30, 2017, 09:33:04 AM »
Does "probability of default" factor, in any way, the timing of the default within the lifecycle of that loan grade (or cohort of graded loans)?

If not, then this is a handy heuristic to look at (I love the quant ephemera, here), but I'm not sure the picture it gives will be all that clear.
A little nonsense now and then is relished by the wisest men.

Rob L

  • Hero Member
  • *****
  • Posts: 2028
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #7 on: April 30, 2017, 09:51:57 AM »
Does "probability of default" factor, in any way, the timing of the default within the lifecycle of that loan grade (or cohort of graded loans)?

If not, then this is a handy heuristic to look at (I love the quant ephemera, here), but I'm not sure the picture it gives will be all that clear.

Typically there would be a "Loss Given Default" percentage associated with "Probability of Default".
At least for D & E loans, the last time I checked, LGD was about 50%. It's been a while since I looked at it.
I've never seen LC publish any LGD numbers (historic or forecast).

nonattender

  • Hero Member
  • *****
  • Posts: 713
  • I am not here.
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #8 on: April 30, 2017, 09:57:08 AM »
Does "probability of default" factor, in any way, the timing of the default within the lifecycle of that loan grade (or cohort of graded loans)?

If not, then this is a handy heuristic to look at (I love the quant ephemera, here), but I'm not sure the picture it gives will be all that clear.

Typically there would be a "Loss Given Default" percentage associated with "Probability of Default".
At least for D & E loans, the last time I checked, LGD was about 50%. It's been a while since I looked at it.
I've never seen LC publish any LGD numbers (historic or forecast).

Need a synthetic metric which takes into account both Probability of and Loss Given --- this might be what the consultants from the Squeeze-A-Nickel Corporation who just recommended that "hardship" plan / state were looking at when they suggested a change. :)
A little nonsense now and then is relished by the wisest men.

brycemason

  • Hero Member
  • *****
  • Posts: 801
    • View Profile
    • P2P-Picks.com
    • Email
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #9 on: April 30, 2017, 10:42:30 AM »
LGD is hard to predict and doesn't correlate with many underwriting variables, as if it did, one could incorporate that information into the pricing. Pretty much whatever is known up front is used for decisioning and pricing, and exogenous events like job loss, death, and family issues determine the severity. Therefore, P[D] is the most common metric. Blending the probability of default and loss given default help you construct an annualized expectation of loss, but it will correlate very highly to P[D]. I simply used the average LGD in my P2P-Picks work. Therefore, Rob's charts aren't flawed in the way contemplated by nonattender.
« Last Edit: April 30, 2017, 10:44:01 AM by brycemason »

Rob L

  • Hero Member
  • *****
  • Posts: 2028
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #10 on: April 30, 2017, 10:44:43 AM »
Okay, here's a chart that shows LC assigned interest rates and the resulting effective interest rates given probability of default and loss given default.

I post this with some fear I've made an error since the effective APR rates in 2013 were so high.
On the other hand I've looked at the calculations closely and don't see any error. Also the probability of default numbers were dramatically lower in 2013 and, when raised to current levels, results in effective APR values in line with the current ones. If somebody has a few hours to kill then all the numbers required to reproduce or refute the results I've shown are in the charts here.



Note that both charts use forward looking numbers from LC; i.e. projections from their model and assigned to loans at time of listing.

nonattender

  • Hero Member
  • *****
  • Posts: 713
  • I am not here.
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #11 on: April 30, 2017, 11:20:19 AM »
LGD is hard to predict and doesn't correlate with many underwriting variables, as if it did, one could incorporate that information into the pricing. Pretty much whatever is known up front is used for decisioning and pricing, and exogenous events like job loss, death, and family issues determine the severity. Therefore, P[D] is the most common metric. Blending the probability of default and loss given default help you construct an annualized expectation of loss, but it will correlate very highly to P[D]. I simply used the average LGD in my P2P-Picks work. Therefore, Rob's charts aren't flawed in the way contemplated by nonattender.

One should not confuse tight correlation between P[D] and LGD to mean that one cannot implement real practice/policies to reduce LGD, at the platform level.  You may not have figured out how to do this at the individual selection level, but the platform has its own levers...

If LGD can be reduced - through any means (say, interstitial state introduction - ie, let's squeeze a few more nickels), returns will go up, over time, and those gains can be offset directly against perceived pressure to raise rates --- because "what else can we do?" Whether things like that will work or not are to be seen, but you don't seem to be looking at the components as a "string", with room to wiggle...

I sense an impending nerd fight --- I will pre-opt-out.

Yay, more charts from RobL - just in time to save me.
A little nonsense now and then is relished by the wisest men.

Rob L

  • Hero Member
  • *****
  • Posts: 2028
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #12 on: April 30, 2017, 11:56:11 AM »
How much difference does a reduction of 5% in Loss Given Default (from 50% to 45%) make in effective APR?

         LGD         LGD
         50%        45%
MAY-13
        EFFECTIVE APR       DELTA
D1   12.37%   12.52%   0.15%
D2   12.82%   12.98%   0.16%
D3   12.92%   13.09%   0.16%
D4   13.01%   13.18%   0.17%
D5   13.30%   13.48%   0.18%
E1   13.74%   13.93%   0.19%
E2   13.77%   13.97%   0.20%
E3   13.88%   14.09%   0.21%
E4   14.09%   14.31%   0.22%
E5   14.31%   14.53%   0.22%

APR-16         
D1   7.21%   7.47%   0.27%
D2   7.25%   7.55%   0.29%
D3   7.29%   7.61%   0.32%
D4   7.21%   7.54%   0.34%
D5   7.34%   7.70%   0.36%
E1   7.46%   7.84%   0.39%
E2   7.49%   7.91%   0.42%
E3   7.52%   7.96%   0.44%
E4   7.52%   7.98%   0.46%
E5   8.11%   8.58%   0.47%

APR-17         
D1   8.09%   8.35%   0.26%
D2   8.14%   8.42%   0.29%
D3   8.18%   8.49%   0.31%
D4   8.23%   8.56%   0.34%
D5   8.27%   8.64%   0.37%
E1   8.52%   8.92%   0.40%
E2   8.52%   8.96%   0.43%
E3   8.58%   9.03%   0.45%
E4   8.77%   9.23%   0.46%
E5   8.86%   9.34%   0.48%

brycemason

  • Hero Member
  • *****
  • Posts: 801
    • View Profile
    • P2P-Picks.com
    • Email
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #13 on: April 30, 2017, 11:59:38 AM »
I was only coming from a place of trying to show the points that Rob's charts made didn't have some fatal flaw due to some defaults being systematically worse. :)

rawraw

  • Hero Member
  • *****
  • Posts: 2757
    • View Profile
Re: LC Assigned Interest Rates and Probability of Default, D&E Grades
« Reply #14 on: April 30, 2017, 12:41:15 PM »
I assume LGD is 100% in unsecured lending for simplicity sake.  Also, the timing aspect comes through in the EAD (exposure at default).  PD (%) * LGD (%) * EAD (%) = expected loss ($)