|
|
Post by smrutib on May 10, 2016 9:50:30 GMT
Thanks littleoldlady for getting this discussion started. Even though it's very difficult to predict with any certainty the level of defaults, I completely agree with you that having a range of outcomes laid out allows everyone to do a reality check regarding their risk appetite. As @gsv3miac suggested, a Monte Carlo simulation might be the way to go. Has someone done that already? If not then I can have a go as I have some time on my hands. Thanks |
|
|
|
Post by earthbound on May 10, 2016 10:10:43 GMT
Thanks littleoldlady for getting this discussion started. Even though it's very difficult to predict with any certainty the level of defaults, I completely agree with you that having a range of outcomes laid out allows everyone to do a reality check regarding their risk appetite. As @gsv3miac suggested, a Monte Carlo simulation might be the way to go. Has someone done that already? If not then I can have a go as I have some time on my hands. Thanks I think its here.. p2pindependentforum.com/thread/3348/monte-carlo-simulation-losses not read all the posts so not 100% sure.
|
|
homes119
Member of DD Central
Posts: 93 
Likes: 19
|
Post by homes119 on May 10, 2016 10:12:53 GMT
This is my (feeble?) effort but I am sure some of you clever guys can do better. Assumptions: An investment programme covers 5 years. All loans are for the same amount, at 12% Loans repaying are immediately reinvested, creating streams of lending. A stream ends either with a default or when the last loan repays which is assumed to be at the end of year 5. For simplicity, all loans which default do so at the same time after 30 months, and 50% of capital is recovered at the end of year 5. No allowance has been made for any payments from the Provision Fund. According to my calculations: Streams which do not end in a default will return 76% Streams which end in a default will return a loss of 10% So the rate of return for various levels of defaulting streams are: A B (see note) 80% defaults return 7% after 5 years or 1.3%pa 13% 27% 70% 16% 3% 13% 22% 60% 24% 4.4% 9% 17% 50% 33% 5.9% 7% 13% 40% 42% 7.3% 5% 10% 30% 50% 8.4% 4% 7% 20% 60% 9.9% 2% 4% 10% 67% 10.8% 1% 2% Notes The default rate is the percentage of streams, not loans, defaulting. The number of loans failing at each level depends on the mix of loan terms. For example Col A shows the proportion of loans failing if they were all 6 months and col B if they were all 12 months. There are obviously a lot of deficiencies in this model. One is that because SS deduct the interest from the loan up front and no payment is expected from the borrower until the end of the term it is most unlikely that a loan will default before the end of the term. This means that in practice it is not possible to have a neat 5 year programme and there will be a tail of recoveries. However as the objective is to try and get a handle on the return after losses, in a range of default scenarios, I don't think this matters. I too welcome this type of analysis. Thanks! Any chance of stress testing this a little further to see at which point we start getting significant losses. From your table, 80% defaults after 30 months and 50% recovery by the end of year 5 yields 1.3% return PA which is comparable to bank savings rates  For example, all loans which default do so at 15 months and 40% of capital is recovered at the end of year 5. This may be extreme (especially for those who pointed out that there has only been one default so far  ) but it would nevertheless be nice to see losses reflected your scenario. Just a suggestion 
|
|
boundah
Member of DD Central
Posts: 367
Likes: 430
|
Post by boundah on May 10, 2016 10:18:35 GMT
I am intrigued by the comment that SS has yet to be tested compared to others. SS is not some new upstart - it has written more business this year than Assetz and Thin Cats combined. Whilst I admit we cannot be conclusive at this stage, the fact that it hasn't experienced the level of defaults compared to others could suggest that its credit assessment processes are more robust. I agree that SS is so far untested. It has yet to ride out a general downturn in property valuations and demand. Also, just writing lots of new business is no indication of safety - in fact, it could be quite the opposite if lending standards nudge downwards just to meet high demand from lenders. I'm no merchant of doom, and happily continue to invest in SS, but I think it's good to be reminded from time to time that losses are a possibility.
|
|
SteveT
Member of DD Central
Posts: 6,873
Likes: 7,918
|
Post by SteveT on May 10, 2016 10:31:25 GMT
[Mod hat off] I went through a similar thought process recently when preparing my company year-end accounts (for all platforms I lend through, not just SS). I've adopted the accounting principles of IFRS9, somewhat earlier than formally required for the accounting of financial instruments, as it seems well-matched to the core P2P business model of holding loans both to collect contractual cashflows (principal + interest) and also potentially to sell. Assets are valued at Fair Value, with gains & losses recognised through a statement of Other Comprehensive Income. Importantly, IFRS9 requires an "impairment allowance" be applied to financial instruments held, with changes year-on-year to this allowance going to the P&L. In practical terms, this means I had to decide what reasonable level of "impairment allowance" to apply to each loan held across each platform. The only platform where this is straightforward, because estimated annual bad debt rates are provided, is FC. After much mulling over the relative risk profiles of the other platforms (and of differing rate loans on each) I decided that individual loan-level impairment risk assessment was pretty much impossible and instead to treat FC's loan book bad debt projections as the surrogate for all my P2P lending, determining an "equivalent FC risk band" for each loan by the headline interest rate (eg. 10.0% - 10.9% = B, 11.0% - 12.9% = C) and then applying the same level of projected impairment. One can very easily argue that this approach potentially overstates the likely impairment of a particular platform, or a particular loan, but in most cases one can just as easily argue that it potentially understates it. The reality is that we just don't know, but at least I can point to a statistical evidence base that I have applied consistently across all my holdings. And it's surely no less logical than FC's decision to project the same level of annual bad debt for an A+ 12 month 1st-charge property loan as for an A+ 60 month unsecured SME loan. Bottom line is that I rated my entire SS loan portfolio as C-equivalent and therefore set the impairment allowance at 3.3% of current loan principal. Yes, I appreciate that SS loans are all asset-backed (when FC's mostly are not), that FC deducts a 1% fee (which SS does not) and that SS has a provision fund (which FC does not). On the other hand, SS loans are highly correlated, my loan diversification is rather worse than the 1% level that FC projections assume and the platform is probably 2 - 3 years further back on the business curve than FC. Incidentally, applying the same criteria to my FS loan portfolio gave 4.7%, MT 4.1%, AC 3.2% and FC 7.5% (my FC holdings these days being almost all Ds and Es). Overall the effect was to reduce my portfolio return across 2015/16 from a gross IRR of around 13% to a net IRR (after impairment allowance) of around 8%, which feels in the right ballpark to me. Of course, YMMV  |
|
|
|
Post by Deleted on May 10, 2016 10:36:57 GMT
While I consider this sort of analysis is confusing data with information I thought I could just throw in some critieria
1) When bear markets really kick in (and I define a bear market as senior stockmarket indexes dropping by more than 20%, note there is no such thing as an agreed specification in this area), then other asset classes (especially illiquid ones such as property) can suffer value drops in the 25 to 50% range.
2) Bear markets normally need a combination of practical and emotional triggers, so both the actual devaluation of another major asset class and the fear that there may be more
3) Bear markets "appear" faster than Bull markets, so something akin to a cliff appears, this makes most traders go into the fear/flight confusion "is it really a bear market?" etc which will result in any secondary market just about collapsing as fast as it can. In that instance any P2P with limits on how much the seller can reduce his prices will have a problem
4) Cliffs and restrictions actually accelerate the fear so rather than taking say 6 months to crash to 50% of value that might well become 3 months.
5) Bull after a major Bear can be very slow so we are talking 3 years rather than 6 months. Which is a pity for those deals that are shorter than 3 years and 6 months, though I'm guessing others may not agree with that conclusion.
So, we need to factors in 50% loss of value taking place over 3 months and with little chance to get out of the market. Not a great scenario but probably a true one.
SteveT; I do much the same with my "portfolio" but work on the basis that there will 6 years of harvest and 1 year of famine (sometimes the Bible just has the best tunes) and so set aside "excess earnings" each year to cover the coming famine. Not far off the 8% either.
|
|
|
|
Post by GSV3MIaC on May 10, 2016 12:05:29 GMT
Yes, there are/were some 3 month loans IIRC (actually there are some with stated 1-12 month terms, and it isn't clear how many month's interest SS are holding).
|
|
mikes1531
Member of DD Central
Posts: 6,452
Likes: 2,320
|
Post by mikes1531 on May 10, 2016 13:49:01 GMT
I am intrigued by the comment that SS has yet to be tested compared to others. SS is not some new upstart - it has written more business this year than Assetz and Thin Cats combined. Whilst I admit we cannot be conclusive at this stage, the fact that it hasn't experienced the level of defaults compared to others could suggest that its credit assessment processes are more robust. I agree that SS is so far untested. It has yet to ride out a general downturn in property valuations and demand. Also, just writing lots of new business is no indication of safety - in fact, it could be quite the opposite if lending standards nudge downwards just to meet high demand from lenders. I'm no merchant of doom, and happily continue to invest in SS, but I think it's good to be reminded from time to time that losses are a possibility. Like boundah, I'm happy to continue to invest via SS, but I do believe that SS really haven't been tested yet. I look at the handful of SS loans with negative terms remaining and wonder whether we'll get out unscathed. After all, these are loans where the borrower hasn't repaid their loan on time and obviously has been unable to arrange a refinancing in the time they thought they could when they took out their SS bridging loan. Furthermore, they are unable to raise enough cash to extend the loan term to give them more time to arrange their refinance. Many people, myself included, would consider those borrowers to be in default of their obligations, though I recognise that 'default' is a technical term in this business and probably shouldn't be used where SS have not gone so far as to present the borrower with an appropriate notice and a demand for repayment. SS give us the impression that they are happy with these borrowers' progress towards repayment, but they have a huge incentive to keep saying that until they call in the receivers. We also need to remember that the great majority of SS loans weren't written long enough ago to have reached the end of their initial term, so won't really have had any chance for the borrower to default. About 20% of the loans SS have funded are officially 'repaid', but some of those don't really count as they never progressed as far as drawdown or were replaced by follow-on SS loans. A final question for those who are optimistic enough to think that investment via SS will result in a 12% return at the end of the day... If such risk-adjusted returns really are achievable long-term, why don't the mainstream lenders seem to want this business?
|
|
|
|
Post by GSV3MIaC on May 10, 2016 14:06:58 GMT
A final question for those who are optimistic enough to think that investment via SS will result in a 12% return at the end of the day /mod hat off ARE there any such people? Many? Any? Maybe CD could run a poll to see what sort of annual returns (we'll settle for un-compounded, simple interest) the forumites are actually expecting. I'm with SteveT on this 6-8% would be good, more would be wonderful, 12% would be miraculous. If you are getting 12%, you better be putting some by for the forthcoming lean years, as predicted by uncle Moses. If you really think you''ll keep getting 12% with no losses, I've got some building land in Florida.. oh wait, no financial promotions on the forum .. ok, scratch that .. "the tooth fairy will see you now" .. 8>.
|
|
mikes1531
Member of DD Central
Posts: 6,452
Likes: 2,320
|
Post by mikes1531 on May 10, 2016 14:09:50 GMT
I do agree that correlation is a major issue, but I could not think of how to cater for it. I don't know how to cater for it either, but ISTM that the impact must be to make a bad situation worse and increase volatility. So if the average 'expected' return were to be 8%, the most likely returns might be either 4% or 12% depending on whether or not a correlated property price crash occurs. Or 0% and 12% if the two outcomes are not equally probable.
|
|
|
|
Post by earthbound on May 10, 2016 16:10:19 GMT
A final question for those who are optimistic enough to think that investment via SS will result in a 12% return at the end of the day /mod hat off ARE there any such people? Many? Any? Maybe CD could run a poll to see what sort of annual returns (we'll settle for un-compounded, simple interest) the forumites are actually expecting. I'm with SteveT on this 6-8% would be good, more would be wonderful, 12% would be miraculous. If you are getting 12%, you better be putting some by for the forthcoming lean years, as predicted by uncle Moses. If you really think you''ll keep getting 12% with no losses, I've got some building land in Florida.. oh wait, no financial promotions on the forum .. ok, scratch that .. "the tooth fairy will see you now" .. 8>. Sheepishly... I will admit to being one of the wild optimist's , who for 2 an half years has only had the miraculous 12%+, all interest has been re-invested and compounding. I work 3 platforms FS, SS and MT and have not lost a single penny. I was involved with the SS care-home which they handled superbly well. I have a very specific way of investing, using specific amounts and only with loans that have a specific criteria, which i revealed mostly( ) in another thread and was basically told i was doomed. (your plan falls flat on its face) but that was because i did not reveal all, so i got what i expected. However i am not wildly optimistic enough to expect never to be caught out, in fact i am involved with the boatyard saga, so my first loss could be here quite soon, and interestingly, it is the only loan in my portfolio, where i broke 2 of my own rules, If i get out of the boatyard saga intact, which i doubt, then i will be even more optimistic. How much is invested... enough to buy a nice 3 bed semi round these parts.
|
|
gurberly
Member of DD Central
Posts: 168
Likes: 98
|
Post by gurberly on May 10, 2016 18:09:50 GMT
Default correlation is notoriously hard to model (and even harder to calibrate given the paucity of the observable data). The standard technique in quant finance is to use a copula (multivariate probability distribution). However, applying that with a Gaussian distribution will not allow for correlation asymmetries in the upside vs. downside regimes. Canonical pair (vine) copulas can be used to remedy this is when combined with a initial Monte Carlo simulation. With regard to the broader issue of modelling loan defaults, my view is that ideally one would attempt to leverage the infrastructure available. Unfortunately most finance analytics are either proprietary or expensive. Last night I played for a couple of hours with QuantLibXL. link. QuantLib is a (free!) open source analytics library (static C++ object classes). The QuantLibXL spreadsheet addin allows the user to access around 1000 wrapped financial functions. It's similar in spirit to the one I use day to day though less well documented and rather clunky. To be fair that might be because I played with it for just a few hours, rather than almost 20 years for the one I use most days! Attached is a spreadsheet I put together using a few of the functions available. I've constructed a simple discount curve, default probability curve and some loan cashflow objects. Given a recovery rate assumption, loan NPVs can be extracted, both in the "credit risky" and "risk-less" scenario. For simplicity, I've used flat term structures for both the interest rate curve and default curves. . This probably isn't of interest to most so I apologize but a few may be interested. Obviously this is just a simple example of what can be done and is not to be used "in anger" E&OE etc etc I think I just learnt a few new words.... I'm now beginning to regret taking the social sciences options at school rather than the "hard stuff" G
|
|
|
|
Post by earthbound on May 10, 2016 18:25:12 GMT
Default correlation is notoriously hard to model (and even harder to calibrate given the paucity of the observable data). The standard technique in quant finance is to use a copula (multivariate probability distribution). However, applying that with a Gaussian distribution will not allow for correlation asymmetries in the upside vs. downside regimes. Canonical pair (vine) copulas can be used to remedy this is when combined with a initial Monte Carlo simulation. With regard to the broader issue of modelling loan defaults, my view is that ideally one would attempt to leverage the infrastructure available. Unfortunately most finance analytics are either proprietary or expensive. Last night I played for a couple of hours with QuantLibXL. link. QuantLib is a (free!) open source analytics library (static C++ object classes). The QuantLibXL spreadsheet addin allows the user to access around 1000 wrapped financial functions. It's similar in spirit to the one I use day to day though less well documented and rather clunky. To be fair that might be because I played with it for just a few hours, rather than almost 20 years for the one I use most days! Attached is a spreadsheet I put together using a few of the functions available. I've constructed a simple discount curve, default probability curve and some loan cashflow objects. Given a recovery rate assumption, loan NPVs can be extracted, both in the "credit risky" and "risk-less" scenario. For simplicity, I've used flat term structures for both the interest rate curve and default curves. . This probably isn't of interest to most so I apologize but a few may be interested. Obviously this is just a simple example of what can be done and is not to be used "in anger" E&OE etc etc I think I just learnt a few new words.... I'm now beginning to regret taking the social sciences options at school rather than the "hard stuff" G I went to the quantlib.org website, i couldn't even get through the download procedure... some very clever people round here.
|
|
poppyland
Member of DD Central
Posts: 237
Likes: 243
|
Post by poppyland on May 10, 2016 18:32:21 GMT
Sheepishly... I will admit to being one of the wild optimist's , who for 2 an half years has only had the miraculous 12%+, all interest has been re-invested and compounding. I work 3 platforms FS, SS and MT and have not lost a single penny. I was involved with the SS care-home which they handled superbly well. I have a very specific way of investing, using specific amounts and only with loans that have a specific criteria, which i revealed mostly( ) in another thread and was basically told i was doomed. (your plan falls flat on its face) but that was because i did not reveal all, so i got what i expected. However i am not wildly optimistic enough to expect never to be caught out, in fact i am involved with the boatyard saga, so my first loss could be here quite soon, and interestingly, it is the only loan in my portfolio, where i broke 2 of my own rules, If i get out of the boatyard saga intact, which i doubt, then i will be even more optimistic. How much is invested... enough to buy a nice 3 bed semi round these parts. Hey Earthbound, nice to hear some optimism, along with the fact that in 2.5 years you have lost nothing. I will see if I can find your thread where you reveal your investment strategy. Good luck with the boatyard!
|
|
|
|
Post by earthbound on May 10, 2016 18:43:35 GMT
Sheepishly... I will admit to being one of the wild optimist's , who for 2 an half years has only had the miraculous 12%+, all interest has been re-invested and compounding. I work 3 platforms FS, SS and MT and have not lost a single penny. I was involved with the SS care-home which they handled superbly well. I have a very specific way of investing, using specific amounts and only with loans that have a specific criteria, which i revealed mostly( ) in another thread and was basically told i was doomed. (your plan falls flat on its face) but that was because i did not reveal all, so i got what i expected. However i am not wildly optimistic enough to expect never to be caught out, in fact i am involved with the boatyard saga, so my first loss could be here quite soon, and interestingly, it is the only loan in my portfolio, where i broke 2 of my own rules, If i get out of the boatyard saga intact, which i doubt, then i will be even more optimistic. How much is invested... enough to buy a nice 3 bed semi round these parts. Hey Earthbound, nice to hear some optimism, along with the fact that in 2.5 years you have lost nothing. I will see if I can find your thread where you reveal your investment strategy. Good luck with the boatyard! its here .. p2pindependentforum.com/thread/5182/diversity?page=5&scrollTo=112423 where i had another good argum... i mean discussion.
|
|
) but it would nevertheless be nice to see losses reflected your scenario. 
