|
|
Post by ranjeb on May 17, 2014 18:30:42 GMT
Hello,
I was doing some statistics work and out of curiosity I did some calculations on the number of loans defaulted that had a loan id under 1000 (unfortunately I can't show the workings but can easily be worked out by those here who download loan book). What I noticed was default rates for B and C loans with a loan id <1000 appeared to be approx 12% and 16% respectively. Should we be worried?
I've looked through the forums but didnt find anything but would have thought someone else would have noticed this (unless my workings are wrong)
|
|
|
|
Post by captainconfident on May 17, 2014 22:31:57 GMT
Welcome to the forum, ranjeb. I do hope you have calculated that wrong. What is probably more pertinent is when these loans defaulted, as a default in early stages takes with it most of the investment, while a default in the final stages only captures the residual remainder. A statistical analysis of when to sell would be useful.
|
|
|
|
Post by GSV3MIaC on May 18, 2014 6:28:53 GMT
Cs less than loan 1000 must be a very small sample? Bs were well dodgy in 2012 iirc. But yes, you really need to look at how much got lost vs how much was invested, for how long. If the loan defaults with 100% recovery it is a minor issue, or if it defaults in last 2 payments. There are historical discussions on t'other forum .. Although that place is still acting weird with posts which are listed not being there when you look.
|
|
agent69
Member of DD Central
Posts: 6,037 
Likes: 4,435
|
Post by agent69 on May 18, 2014 8:20:46 GMT
If the loan defaults with 100% recovery it is a minor issue. FC loan defaults with 100% recovery? That would probably be front page news in every news paper in the country (probably only going to see that on 1st April)
|
|
|
|
Post by ranjeb on May 18, 2014 10:40:53 GMT
Will try and show my workings shortly, I had considered that these loan ids are very old and close to completion however I was just comparing the default rates to fundingcirlces advertised default rates, the actual default rates can be skewed (temporarily) by increased quantity of loans on the market.  |
|
andy2001
Member of DD Central
Posts: 361
Likes: 34
|
Post by andy2001 on May 19, 2014 4:11:27 GMT
Will try and show my workings shortly, I had considered that these loan ids are very old and close to completion however I was just comparing the default rates to fundingcirlces advertised default rates, the actual default rates can be skewed (temporarily) by increased quantity of loans on the market. The annual default rate will mean Amount lost due to default over a 12 month period. This is not the same as number of loans defaulting.
|
|
|
|
Post by ranjeb on May 19, 2014 7:28:18 GMT
Will try and show my workings shortly, I had considered that these loan ids are very old and close to completion however I was just comparing the default rates to fundingcirlces advertised default rates, the actual default rates can be skewed (temporarily) by increased quantity of loans on the market. The annual default rate will mean Amount lost due to default over a 12 month period. This is not the same as number of loans defaulting. Ah, that would be the flaw in my thinking.
|
|
|
|
Post by GSV3MIaC on May 24, 2014 19:14:10 GMT
I just ran my stats on the first loans (numbers <1000) and got PA losses of 7.1% for Bs and 6.1% for Cs, although there are only 42 of the latter, so the stats are dodgy. This number is loss, divided by pound- years of loan, although I can't tell if some were repaid early. Iirc I am also correcting for the known effect where loans usually don' t fail in first few months.
Yes, this is not good. It got better since then, and even these numbers may get revised better-wards if there are further recoveries - despite sarcastic comments, the facts show that there are some loans with 100% recovery, eventually.
I can show my working, but it complicated. I wish FC would show theirs!!
I normally adjust this to add in an estimated badness for ones which are just late, as well as those which are acknowledged to be defaulted - a fudge factor, if you like.
|
|
merlin
Minor shareholder in Assetz and many other companies.
Posts: 902
Likes: 302
|
Post by merlin on May 24, 2014 20:37:12 GMT
It would be interesting to see the results for those loans in the 1001 to 2000 range and 2001 to 3000 range as most of these have been around for awhile now and comparison may provide some indication as to whether FC are getting better or worse at recovering defaults. That being said recently FC took debt chasing in house and in time that may also bring improvement if only to keep on the right side of the FCA.
|
|
|
|
Post by GSV3MIaC on May 25, 2014 3:34:40 GMT
Right, here goes .. this is actually NOT corrected for 'infant health' (young loans don't usually default) - I only need to do that, I recall, if looking for significant differences, which I'm not. There are some errors in the loan book (699 claims to be a C-, which makes no sense, although the date matches for a 699 number, the risk band didn't exist then - that one was easy to spot, others would go undetected).
0-999 1000-1999 2000-2999 3000-3999 4000-4999 Everything (0-7000)
Risk Band: A+ 1.2% 2.7% 0.0% 0.0% 0.0% 1.1%
Risk Band: A 4.7% 1.8% 0.4% 0.0% 0.0% 1.5%
Risk Band: B 7.1% 2.1% 3.1% 2.0% 0.0% 3.1%
Risk Band: C 6.1% 3.8% 1.1% 0.3% 0.8% 2.2%
Risk Band: C- ! 7.0% 1.0% 4.6%
This is LOSS (at present time, i.e. net of whatever recoveries have been made so far) divided by SUM OF (amounts outstanding each month) (i.e. pound-years) for that selection (in this case risk band, but I sometimes look by region, or loan size, or whatever. In future I shall be looking carefully at 'vs number of loan parts = 1'. 8>.).
We still have some 'apples and oranges' going on, because later loans offer more different loan terms, larger loan sizes, and different security arrangements. Almost all sub-samples are too small to do any meaningful significance testing, which makes it possible to say things like 'there is no significant proof that C-s are any worse than Bs'.
This is JUST defaults .. normally (as I mentioned) I'd toss in some fractional correction for loans which are repeatedly, or persistently, or seriously, late (although the recent massacre of zombies has cleared the air a bit).
|
|
|
|
Post by GSV3MIaC on May 26, 2014 11:02:13 GMT
Ah, it occurred to me later (sorry, the 'flu has me in its grip) that we do (some of us) want to compare between the categories, so here is some more data. the 'problem loan count' and 'expected problem loans' basically takes account (now) of the size of the loss - i.e. 50% recovery is 0.5 of a problem loan. The 'expected' is what you'd get allowing for the effect of infant health .. i.e. loan which are still in the first few months are less likely to have defaulted.   for comparison, here is the table showing 'infant health'. In this case the 'expected' problems assume that loans fail uniformly over time (i.e. just as likely to die in each month .. demonstrably not true): Be aware though that I don't think there ARE any 36, 35 or whatever payment loans - these were all repaid early, and it's not possible to extract the truth from the loan book. 60 months? Interest only?? over to the crystal ball! |
|
merlin
Minor shareholder in Assetz and many other companies.
Posts: 902
Likes: 302
|
Post by merlin on May 26, 2014 19:34:03 GMT
At first sight you numbers give the impression that maybe, just maybe FC are getting to grips with defaulters. However the time factor and length of loan in this have a very profound influence in the overall picture.
Thanks a lot for your work on this as most of my ideas on this tend to be rather subject.
|
|
