|
Post by westonkevRS on Dec 11, 2015 22:07:06 GMT
I've completely bailed out my £70k out of the monthly market a couple of hours ago having seen the coverage ratio dip to 139%. I had set myself a minimum of 150% coverage since joining RS about 9 months ago, and the sudden drop over the past week was more that a little worrying. I suggest that if you could communicate underlying changes in bad debt assumptions on RS it would be helpful - to distinguish from actually losses incurred affecting the coverage ratio. This would help lenders understand that your ratio reflects a more prudent assumption. Hi woodbury . Sorry that you felt the need to remove the funds. Always sad to lose a lender. I wasn't going to respond because I don't want to provide reassurance on the Internet, in reality P2P is a risk and it would be wrong for me to try and give an otherwise impression. I don't feel comfortable doing this. I obviously have an inside knowledge of the risks at RateSetter, which makes me doubly sad when a lender leaves due to risk or liquidity worries. I appreciate it would be good if I could explain the coverage mechanics in more detail online, but it's hard. The mechanics of the expected loss calculations are not easily explained in words, and that's why I offer any lenders the opportunity to come into the office to demonstrate. But I appreciate that isn't practical for everyone. For the record and the financially technically amongst you, although we are not a bank we decide to follow quite closely Basel II risk rating methodology, using a 12:18 approach for PD and 12:24 for LGD. We've been through this with the institutional investors including the BBB, who continue to lend £ms with RateSetter. There are plans very shortly to enhance Provision Fund reporting, respond to questions regarding governance and perhaps periodical commentary. I won't comment more here, but I think you'll be pleased. We may actually include a type of BoE fan of estimates, but if you've seen their inflation estimates you'll see this fan is quite rightly a very wide set of projections, basically an honest admission that they don't know. I honestly and simply don't know what the future RateSetter losses will actually be. A nice lender came into the office today and I discussed some of the prudence within the expected loss model. And like several other lenders I've met before said we should be more public with the mechanics. Especially as of all the scenarios we model to determine expected losses (and therefore coverage), we have chosen one of the most prudent. Here are some of the reasons our modelling is prudent: 1) We use a straight line expectation of future losses that will equal losses already experienced in the loan, which simply isn't true. Loans are much less likely to default in the second half of a loan than the initial phase. For 5 year loans, peak default is 9 months and after 16 months has a small percentage of defaults.
2) The model treats all loans as if they are on month 1, the mature clean segment of the book is treated as a newbie with higher risks.
3) Higher risk paying customers are not reclassified after x months, they stay high risk until they complete. A 'Z' grade customers isn't reclassified using behaviour scoring into an 'A' grade customer even if they make 20 clean perfect payments!
4) Income from assets held by the Provision Fund are not built into model, for example the income from Debt Management Plans and Arrangement to Pays is ignored. This income is becoming increasingly large and stable.
5) Income from debt sale on assets in recoveries is not considered. More news on this in 2016.
6) Provision Payments from loans with fees over the lifetime rather than up-front are excluded, despite us shifting to this model therefore booking lots of loans with no corresponding up-front payment. This is causing a real short-term drag. 7) LGD only considers a 24 month income period. The majority of "bad" debt is on debt management plans so in theory given enough time we will collect a majority of the debt. This just isn't true for a 24 month period, especially loans originally on 5 year terms. 8) A bank would give a customer a payment holiday or two and the loan is classified as good. RateSetter has to "default" the loan by putting it into the Provision Fund and our model ignores the obvious future income we'll get. Ignoring these factors feels more honest as we'd rather over predict losses. But in turn this causes a lower coverage ratio and worries with lenders such as yourself. We could quite easily choose another less prudent (dare I say accurate, based on rear view mirror analysis) model which would give a high coverage ratio. Then lenders such as your good self might not leave. Sitting on my pedestal is losing us customers. Which is why I'm writing this, because I'm in a position of perfect moral hazard. Be honest and prudent, risk losing lenders. Or use a less prudent model, build in some optimistic outcomes and hey presto the coverage is over 200 and lenders get a false sense of security and invest more. I've taken the righteous path and lost a lender. Doh.Regards, Kevin. P.S. For the record, the Provision Fund just hit a new record, £16,430,662. Cash in hand sometimes speaks stronger than forecasts that inherently unforeseeable. P.P.S. This reminds me of the economists joke. There are only two, those that know they can't predict the future, and those that don't know they can't predict the future.
|
|
jonah
Member of DD Central
Posts: 2,031
Likes: 1,113
|
Post by jonah on Dec 11, 2015 22:36:01 GMT
Thanks for putting that out there. An interesting read. It looks like moving from upfront to month by month is providing a lot of the issues and I assume that will take years to work though?
Personnally my biggest recent issue with RS has been the collapse in 3yr rates, but I am glad that the 'backend' continues to evolve.
|
|
|
Post by Deleted on Dec 11, 2015 22:40:46 GMT
You're definitely creating a rod for your own back somewhat by having the coverage ratio so prominent, but important details like recovery rates nowhere to be found.
Makes things seem far more volatile than they are, especially if defaults cause a large provision fund drop up front which are then replaced over time by asset recoveries.
I actually pay little attention to the coverage ratio myself, since the mechanics are so open to assumption. I prefer the provision fund/total outstanding loan amount ratio personally, which has been more stable (a slight decline since I started), but I have to work this out myself each time.
It would be nice to see that ratio more prominently, plus maybe some ability to see historical ratios more easily. And more info on things like historical asset recovery rates on defaults would be nice, definitely.
From a marketing perspective, for customers unaware of all the assumptions, the big coverage ratio swings must be disconcerting.
|
|
|
Post by westonkevRS on Dec 12, 2015 10:09:08 GMT
I actually pay little attention to the coverage ratio myself, since the mechanics are so open to assumption. I prefer the provision fund/total outstanding loan amount ratio personally, which has been more stable (a slight decline since I started)... It would be nice to see that ratio more prominently, plus maybe some ability to see historical ratios more easily. Personally, as a lender, I agree. Its the absolute amount and the ratio that I find important. Certainly the size gives RateSetter time to change tack when other factors change. The ratio has flat lined with a small decrease over the last 6-months. Nothing serious, but performance hasn't been optimal compared to the strong growth in 2013 and 2014. But I'm confident 2016 will be back to growth for a multitude of reasons, and as I indicated web page reporting will be enhanced based on feedback and discussion. Kevin.
|
|
|
Post by closetotheedge on Dec 12, 2015 13:27:12 GMT
Thank you westonkev for taking the time to provide such a full answer. I would not pretend to understand very much of it but the honesty to your writing gives me confidence. Your answers on this forum seem the best advertisment for Ratesetter. Contrary to your advice I have far more with Ratesetter than I could afford to lose so the bare headline of a drop to 139% grabs my attention.
I do like the idea of looking at the absolute size of the provision fund compared to the amount outstanding this is much more my level of understanding.
|
|
|
Post by westonkevRS on Dec 12, 2015 20:29:12 GMT
Thank you westonkev for taking the time to provide such a full answer. I would not pretend to understand very much of it but the honesty to your writing gives me confidence. Your answers on this forum seem the best advertisment for Ratesetter. Contrary to your advice I have far more with Ratesetter than I could afford to lose so the bare headline of a drop to 139% grabs my attention. I do like the idea of looking at the absolute size of the provision fund compared to the amount outstanding this is much more my level of understanding. Glad to be of some service. One extra point is that a reasonable portion of the outstanding balances do NOT have access to the Provision Fund, this is basically some of our large lenders and partners that have chosen to accept the risk. Think like FC or the original P2P concept. I can't say how much for commercial reasons, but it is not immaterial. These partners do not contribute into the Provision Fund and hence another reason it doesn't seem to grow at the same rate as our lending or outstandiung balances. As a result, the £16.4m covers a smaller percentage of the outstanding balance that you might think - hence the actual ratio is actually better than you could ever calculate. I know this is only marginally helpful at this stage, but in the New Year we are planning to improve the Provision Fund reporting to add more depth. Kevin.
|
|
|
Post by westonkevRS on Jan 6, 2016 9:26:57 GMT
Comrades, There is a new blog that you might like to read: www.ratesetter.com/blog/article/further-enhancements-to-our-loan-book-transparencyThere are some new agreed reporting requirements from the P2PFA that are useful, and perhaps go some way to improving on lender's ability to interpret the coverage ratio and actual losses. Often forum members have only been able to see the % of bad debt lending, when actually what's important is the " spend" of the provision Fund: www.ratesetter.com/aboutus/statisticsThese are actuals, whereas the coverage rate is estimated: The screenshot as @ today, but the web page is live. Hopefully this goes some way to demonstrate continued transparency and respond to comments on the forum. Kevin.
|
|
oldgrumpy
Member of DD Central
Posts: 5,087
Likes: 3,233
|
Post by oldgrumpy on Jan 6, 2016 9:31:40 GMT
Interesting figures, Kev. Thank you
|
|
|
Post by p2plender on Jan 6, 2016 10:18:10 GMT
Thanks.
|
|
spiral
Member of DD Central
Posts: 910
Likes: 456
|
Post by spiral on Jan 6, 2016 11:28:15 GMT
Kev, do defaults get added to the year of default or the year of loan origin. Thanks
|
|
|
Post by westonkevRS on Jan 6, 2016 14:23:08 GMT
Kev, do defaults get added to the year of default or the year of loan origin. Thanks To the year of origin, it's all tracked within specific application cohorts. Any unused Provision Fund is passed kept in the fund for lenders, so an unused Provision Fund from one annual cohort is available should there be an over-spend in a subsequent cohort. Although to date all individual application annual cohorts have under spent the Provision Fund contributions. Long may this continue. Kevin.
|
|
jlend
Member of DD Central
Posts: 1,817
Likes: 1,444
|
Post by jlend on Jan 15, 2016 8:01:54 GMT
Good to see the provision fund over 17m now. A sizable fund to manage.
The percent of loans 1 month late is about 0.5%, similar for loans 2 months late.
I dont look at this normally. Is this the usual sort of figures over the years for late loans?
Apologies if there is a graph of this over time somewhere already that i have missed.
|
|
|
Post by westonkevRS on Apr 6, 2016 19:44:20 GMT
We've updated the methodology for the Provision Fund estimation of losses and therefore the coverage ratio. This will be changed on the 13th April 2016. It's blogged in summary here: www.ratesetter.com/blog/article/improving-the-accuracy-of-the-provision-fund-coverage-ratio-2With more detail in the members area when logged in: members.ratesetter.com/noticeboard/improving_the_accuracy_of_the_provision_fund_coverage_ratio_explaining_the_methodologyThe main change is: " The main update we are making is for loans to individuals, where we will move the likelihood of default from a ‘straight line’ over the term of a loan to a ‘curve’. So, for example, rather than a 60 month loan with a perfect repayment record being assigned a fixed likelihood of default over its whole term, our data suggest that the likelihood of default declines after 12 months, so we will apply that lower rate for the rest of the loan term." We always knew this would be the case. Anyone that works in consumer lending is well aware of this dynamic delinquency feature of maturing loans. We just wanted to wait until we had enough data to model it for ourselves. Kevin.
|
|
jimc99
Member of DD Central
Posts: 284
Likes: 115
|
Post by jimc99 on Apr 6, 2016 21:24:02 GMT
So will a higher expected default rate be assigned to the earlier life of a loan to reflect the corresponding increased risk of loans in their initial period?
I assume that the actual overall default rate you expect will be unchanged at the current 2.3% as this is based on all your data for the last 6 years of lending?
But perhaps I am being a bit naive. No doubt the PF will look better after the change!
|
|
pip
Posts: 542
Likes: 725
|
Post by pip on Apr 6, 2016 21:51:30 GMT
Thanks for the update Kev, I get the logic, probably best to review this regularly to check repayment profile matches expectations.
For me I prefer my way of calculating the coverage ratio using the metric bad debt as a % of repayments. I then apply the metric % to the current loan book. This is giving me £16.3m at present, which gives a lower coverage ratio than ratesetter has. No moving goalposts and provides a pretty accurate picture of the current level coverage at today's rate of bad debt. You could argue it is artificially high as it doesn't capture the end of repayments on recently made loans (which based on the above logic should be lower), but counterbalancing this is that newly made loans won't have many defaults as even overdue repayments won't have had time to default.
Whether it provides a guide to future bad debt, well clearly anything can happen, but it provides me with the best picture based on current trends.
Now I am sure Kevs calculations of future defaults are more sophisticated than mine, maybe mine is too prudent on the bad debt of mature loans and maybe the current loan book has a lower risk profile than previous ones did. However I like to know what goes into the numbers I use, it makes sense to me, I think it's pretty prudent and I am happy with it.
|
|