|
|
Post by propman on Aug 2, 2019 12:34:43 GMT
Good points.
The optmistic view of the expected defaults is that there credit checking has improved. I have always been wary of P2P re potential to create high defaults due to relatively small issues in credit checking.
P2P generally employs only soft crdit checks prior to loan acceptance. This allows potential borrowers to shop around and bots might make this widely used*. Due to relatively low market share of a P2P company of all loans, an error missing a bad credit marker identified by the other lenders could become catestrophic. Even if this marker only applied to less than 1% of potential borrowers, this might be a very large number of applicants relative to the loans 1 P2P provider is originating. This would lead to giving sub-market quotes to these applicants (ie their quotes do not adequately charge for the credit risk identified by that marker). These would then be accepted by a high proportion of the applicants concerned which could be a high proportion of the loans agreed by that P2P. Hence their entire loan book would under estimate the credit risk.
Such errors will show up in default rates. I would hope that the sites use retrospective analysis to identify common attributes of defaulting (or even late) loans as a minimum, even if they are not utilising an effective AI learning algorithm (ie self correcting credit checking) for credit assessment. Over time this should allow the above faults to be ironed out.
*I do not know whether comparrison sites currently enable automated personalised loan quotes to be obtained from soft checking providers, but if they don't now I expected they will soon!
As I have posted elsewhere, LW has substantially increased the contributions to the Fund. Whether sufficiently is of course a matter of judgement. Personally I stopped lending on RS at one point when lesser excess default issues arose. As it transpired they seem to have weathered that particular bad debt tranche. personally I am struggling to find investments whose return is acceptable, but with low deposit rates (excluding the limited amounts that can be placed in some regualr savings and current accounts), I have reluctantly recommenced lending on RS and am putting a small amount in LW.
1) Is this low risk? Definitely not. 2) Does my over estimation of the impact of increased defaults and lower funds in the past show that I should decrease my risk assessment? No, the nature of fixed income investments is limited potential income against potential total capital losses (picking pennies up in front of steam rollers) so low likelihood risks are potentially critical to the correct risk assessment (ie fat tails can't be ignored) and 1 result is statistically insignificant data from which to assess the future.
So I hold my nose and dive in but make sure that I don't invest criitical funds in P2P. That said, as stated elsewhere I have assessed the likelihood of losing a significant proportion of funds in RS/Zopa to be small, so if I don't need the liquidity, I believe it is a useful investment class that I consider striated into around 90% good capital security and 10% high risk and the premium of returns over risk free only acceptable in the context that most of it is attributable to the 10% tranche.
Hope that makes some sense to some of you!
|
|
r00lish67
Member of DD Central
Posts: 2,692 
Likes: 4,048
|
Post by r00lish67 on Aug 2, 2019 12:56:22 GMT
Good points.
The optmistic view of the expected defaults is that there credit checking has improved. I have always been wary of P2P re potential to create high defaults due to relatively small issues in credit checking.
P2P generally employs only soft crdit checks prior to loan acceptance. This allows potential borrowers to shop around and bots might make this widely used*. Due to relatively low market share of a P2P company of all loans, an error missing a bad credit marker identified by the other lenders could become catestrophic. Even if this marker only applied to less than 1% of potential borrowers, this might be a very large number of applicants relative to the loans 1 P2P provider is originating. This would lead to giving sub-market quotes to these applicants (ie their quotes do not adequately charge for the credit risk identified by that marker). These would then be accepted by a high proportion of the applicants concerned which could be a high proportion of the loans agreed by that P2P. Hence their entire loan book would under estimate the credit risk.
Such errors will show up in default rates. I would hope that the sites use retrospective analysis to identify common attributes of defaulting (or even late) loans as a minimum, even if they are not utilising an effective AI learning algorithm (ie self correcting credit checking) for credit assessment. Over time this should allow the above faults to be ironed out.
*I do not know whether comparrison sites currently enable automated personalised loan quotes to be obtained from soft checking providers, but if they don't now I expected they will soon!
As I have posted elsewhere, LW has substantially increased the contributions to the Fund. Whether sufficiently is of course a matter of judgement. Personally I stopped lending on RS at one point when lesser excess default issues arose. As it transpired they seem to have weathered that particular bad debt tranche. personally I am struggling to find investments whose return is acceptable, but with low deposit rates (excluding the limited amounts that can be placed in some regualr savings and current accounts), I have reluctantly recommenced lending on RS and am putting a small amount in LW.
1) Is this low risk? Definitely not. 2) Does my over estimation of the impact of increased defaults and lower funds in the past show that I should decrease my risk assessment? No, the nature of fixed income investments is limited potential income against potential total capital losses (picking pennies up in front of steam rollers) so low likelihood risks are potentially critical to the correct risk assessment (ie fat tails can't be ignored) and 1 result is statistically insignificant data from which to assess the future.
So I hold my nose and dive in but make sure that I don't invest criitical funds in P2P. That said, as stated elsewhere I have assessed the likelihood of losing a significant proportion of funds in RS/Zopa to be small, so if I don't need the liquidity, I believe it is a useful investment class that I consider striated into around 90% good capital security and 10% high risk and the premium of returns over risk free only acceptable in the context that most of it is attributable to the 10% tranche.
Hope that makes some sense to some of you! Makes sense. Would only specifically disagree with "LW has substantially increased the contributions to the Fund". As far as I can see, LW haven't added a penny to the PF, in fact several hundred thousand has been removed in recent months. What you probably mean is that they've increased their forecast for borrower contributions to the fund. All well and good except they don't include a forecast for borrower claims against the fund (RS do, and theirs results in a negative total i.e. borrowers are expected to take more than they put in). This has been the case for LW too in recent months, hence PF cash keeps dropping. So, I'm not sure we can draw anything meaningful from this 'increase'.
|
|
|
|
Post by propman on Aug 2, 2019 14:24:13 GMT
Makes sense. Would only specifically disagree with "LW has substantially increased the contributions to the Fund". As far as I can see, LW haven't added a penny to the PF, in fact several hundred thousand has been removed in recent months. What you probably mean is that they've increased their forecast for borrower contributions to the fund. All well and good except they don't include a forecast for borrower claims against the fund (RS do, and theirs results in a negative total i.e. borrowers are expected to take more than they put in). This has been the case for LW too in recent months, hence PF cash keeps dropping. So, I'm not sure we can draw anything meaningful from this 'increase'. I am drawing a distinction between contributions to the fund and payments from it. I am not commenting on the net movement just that the first (contributions) was increased as a proportion of the loans arranged. Bad debts have crystallised while they have increased the PF contributions which have netted off with the cash raised leading to no net incrase in the size of the PF. Comparing LW to RS, the excess over expected were larger, but the response in increasing the proportion of contributions on subsequent loans was earlier. Also it seems (subjectively) that they have accepted the likely overruns earlier where RS have seemed to ignore the evidience for a while then acknowldge the issue incrementally only after they have arranged the loans that will provide the funding to the PF to cover the shortfall. I thus have a greater confidence that LWs estimates are more realistic.
I am confused about your distinction between LW and RS. I see the estimate of future claims to be their estimate of future outflows. AIUI the future contributions included for both are those on loans already arranged. As a result, if this was > expected bad debt, they would be expecting that none of the current cash would be used against existing loans and so that the PF cash would increase by all net contributions from future loans (ie PF contributions less bad debts). If they structured the PF payments from future loans to achieve this, it would have to be by reducing cash taken for PF upfront and hence increase liquidity risk relative to requiring some of any increase as cash contributions upfront.
|
|
r00lish67
Member of DD Central
Posts: 2,692
Likes: 4,048
|
Post by r00lish67 on Aug 2, 2019 15:35:19 GMT
Makes sense. Would only specifically disagree with "LW has substantially increased the contributions to the Fund". As far as I can see, LW haven't added a penny to the PF, in fact several hundred thousand has been removed in recent months. What you probably mean is that they've increased their forecast for borrower contributions to the fund. All well and good except they don't include a forecast for borrower claims against the fund (RS do, and theirs results in a negative total i.e. borrowers are expected to take more than they put in). This has been the case for LW too in recent months, hence PF cash keeps dropping. So, I'm not sure we can draw anything meaningful from this 'increase'. <snip>
I am confused about your distinction between LW and RS. I see the estimate of future claims to be their estimate of future outflows. AIUI the future contributions included for both are those on loans already arranged. As a result, if this was > expected bad debt, they would be expecting that none of the current cash would be used against existing loans and so that the PF cash would increase by all net contributions from future loans (ie PF contributions less bad debts). If they structured the PF payments from future loans to achieve this, it would have to be by reducing cash taken for PF upfront and hence increase liquidity risk relative to requiring some of any increase as cash contributions upfront.
Confess I'm not quite sure I catch your drift either! Tricky to explain.. IMV, Ratesetter are slightly clearer in their stats currently on this: RS say their PF cash is £12.7m, their expected future provision fund inflows are £27.2m and their expected future losses are £33.0m. As a result (although they don't explicitly state this), their net forecast PF position is surely just £12.7m + £27.2m - £33.0m = £6.9m (0.78% of their £888m loans under management). LW say their PF cash is £1.045m, their expected future provision fund inflows (contractual future income) are £2.082m, but do not state their expected future losses explicitly as one number. They do however provide their actual lifetime bad debt rate for each year, as well as their forecast lifetime bad debt rate. So, I think we can therefore take the forecast bad debt rate and subtract the actual bad debt rate for each year in order to work out how much more call there is supposed to be on the shield. If we sum all of those differentials up, I calculate that as being £3.5m. If that's correct, then that is interesting. Ratesetter's forecast PF losses are 121% of their forecast PF inflows (£33.0m / £27.2m) Lendingworks forecast PF losses are 168% of their forecast PF inflows (£3.5m/£2.082m) That feels about right because LW's PF cash has been dropping faster than RS's in recent months. The interesting bit though is that whilst RS's net forecast PF position is a very slender amount (0.78%, as above), LendingWorks' is actually negative: £1.045m + £2.082m - £3.5m = -£0.373m (-0.41%)I will add a big caveat to the above that I'm not sure my methodology is correct. If it is correct, then that's a little concerning frankly as it would suggest LW are effectively implicitly forecasting that they're going to run out of PF cash before the end of their currently held contracts. Matthew - as it's not like you have a full-time job doing anything else (  ) , I don't suppose you'd like to check my logic there? propman , or anyone else, perhaps you'd like to check my working?
|
|
r00lish67
Member of DD Central
Posts: 2,692
Likes: 4,048
|
Post by r00lish67 on Aug 2, 2019 15:40:54 GMT
Friday afternoon maths challenge for RS/LW investors! (who doesn't like a maths challenge on Friday afternoon  ). Looking at both platforms, assume for the moment that they both have written their last loan and all we have to witness is the paying off of all of their respective loans over the next few years: As a £ amount, how much do Ratesetter + LendingWorks respectively forecast their Provision Fund balances to be once all of their current loans have run to term?
The reason I'm asking is that I've attempted the maths above and the answer is 'interesting', but not sure I've interpreted the stats correctly, esp. in LW's case.
|
|
|
|
Post by Ace on Aug 2, 2019 19:50:19 GMT
Friday afternoon maths challenge for RS/LW investors! (who doesn't like a maths challenge on Friday afternoon ). Looking at both platforms, assume for the moment that they both have written their last loan and all we have to witness is the paying off of all of their respective loans over the next few years: As a £ amount, how much do Ratesetter + LendingWorks respectively forecast their Provision Fund balances to be once all of their current loans have run to term?
The reason I'm asking is that I've attempted the maths above and the answer is 'interesting', but not sure I've interpreted the stats correctly, esp. in LW's case. Your logic looks sound to me. Will be interesting to see if LW are willing to comment. I shall take silence to indicate that you are correct. I'm already drawing down from RS. I'm holding on LW for now, hoping for some positive news on the PF situation. Not really sure why, but I tend to trust LW more. Though that hasn't stopped me diverting funds that could have been sent there to CP and UB instead, as these two platforms are currently even further up my trusted scale.
|
|
p2pmark
Member of DD Central
Posts: 218
Likes: 187
|
Post by p2pmark on Aug 3, 2019 8:06:46 GMT
|
|
djay
Member of DD Central
Posts: 121
Likes: 87
|
Post by djay on Aug 3, 2019 8:24:31 GMT
Friday afternoon maths challenge for RS/LW investors! (who doesn't like a maths challenge on Friday afternoon ). Looking at both platforms, assume for the moment that they both have written their last loan and all we have to witness is the paying off of all of their respective loans over the next few years: As a £ amount, how much do Ratesetter + LendingWorks respectively forecast their Provision Fund balances to be once all of their current loans have run to term?
The reason I'm asking is that I've attempted the maths above and the answer is 'interesting', but not sure I've interpreted the stats correctly, esp. in LW's case. Looks reasonable to me based on information currently available. The factor that I'm unsure of is the interaction of LW's provision fund with the insurance policies. For example, operationally, does the provision take an initial hit which we may have seen in recent figures and is effectively replenished by inflows from insurance policies.
|
|
r00lish67
Member of DD Central
Posts: 2,692
Likes: 4,048
|
Post by r00lish67 on Aug 3, 2019 9:05:00 GMT
Looks reasonable to me based on information currently available. The factor that I'm unsure of is the interaction of LW's provision fund with the insurance policies. For example, operationally, does the provision take an initial hit which we may have seen in recent figures and is effectively replenished by inflows from insurance policies. Not an exact answer to this question, but Matthew did touch on this the other week. saying "Historically, the insurance has roughly covered itself i.e. the level of claims has generally covered premiums paid. This is to be expected - after all, if claims were significantly in excess of premiums, the insurer would not be happy. The insurance is there primarily to help protect some of the downside risk, for example in the event of widespread unemployment" I'm assuming the net position forecast includes any rebate from insurance, and so can effectively be disregarded. As Matthew suggests above, we shouldn't be expecting LW's insurance to be much more than a contingency for drastic times. If it was going to pay out in any significant number of cases, then the cost of the insurance would start to outweigh it's risk event.
|
|
djay
Member of DD Central
Posts: 121
Likes: 87
|
Post by djay on Aug 4, 2019 11:10:39 GMT
Looks reasonable to me based on information currently available. The factor that I'm unsure of is the interaction of LW's provision fund with the insurance policies. For example, operationally, does the provision take an initial hit which we may have seen in recent figures and is effectively replenished by inflows from insurance policies. Not an exact answer to this question, but Matthew did touch on this the other week. saying "Historically, the insurance has roughly covered itself i.e. the level of claims has generally covered premiums paid. This is to be expected - after all, if claims were significantly in excess of premiums, the insurer would not be happy. The insurance is there primarily to help protect some of the downside risk, for example in the event of widespread unemployment" I'm assuming the net position forecast includes any rebate from insurance, and so can effectively be disregarded. As Matthew suggests above, we shouldn't be expecting LW's insurance to be much more than a contingency for drastic times. If it was going to pay out in any significant number of cases, then the cost of the insurance would start to outweigh it's risk event. Its probably correct that the net position could be disregarded in a steady state loanbook, I'm no so sure for a rapidly growing loanbook with a lag to insurance claim inflows. It would be good to get an explanation of what is happening with the LW provision fund form lw, it's certainly holding me back with investment on the platform.
|
|
r00lish67
Member of DD Central
Posts: 2,692
Likes: 4,048
|
Post by r00lish67 on Aug 4, 2019 11:58:19 GMT
Not an exact answer to this question, but Matthew did touch on this the other week. saying "Historically, the insurance has roughly covered itself i.e. the level of claims has generally covered premiums paid. This is to be expected - after all, if claims were significantly in excess of premiums, the insurer would not be happy. The insurance is there primarily to help protect some of the downside risk, for example in the event of widespread unemployment" I'm assuming the net position forecast includes any rebate from insurance, and so can effectively be disregarded. As Matthew suggests above, we shouldn't be expecting LW's insurance to be much more than a contingency for drastic times. If it was going to pay out in any significant number of cases, then the cost of the insurance would start to outweigh it's risk event. Its probably correct that the net position could be disregarded in a steady state loanbook, I'm no so sure for a rapidly growing loanbook with a lag to insurance claim inflows. It would be good to get an explanation of what is happening with the LW provision fund form lw, it's certainly holding me back with investment on the platform. I agree. I also think some stats on insurance claims would be useful. Reading between the lines of Matthew's previous comments, I think the significance of the recoveries from insurance claims in £ terms seen so far are nowhere near as significant as some might think. Matthew you mentioned a stats overhaul coming shortly, is that still on the way? All I see now is that the many graphs have been replaced by one. I'm not quite sure of the value of that one either as all it shows is a steepening lifetime default rate for each cohort year. Without the context of increased borrower APR's, that probably actually paints probably a worse picture then is really the case..
|
|
|
|
Post by Matthew on Aug 5, 2019 15:41:20 GMT
Hi all Apologies for the delay in coming back to this thread – I was out of the office last week. I’ll try to address the various points raised in one post, so apologies in advance for the length of this one… It’s clear that some cohorts are performing significantly worse than expected, and the reasons for this are being managed closely, with poor performing segments being removed and credit rules tightened, in addition to various ongoing updates and improvements being made to our credit modelling. We’ve performed extensive analysis on the existing portfolio and note that these changes would have resulted in a significant improvement in loss-performance against expectations for these cohorts. It’s important to note that while the UK has experienced a benign credit environment generally for the past few years, this is not necessarily reflective of our comparatively small lending book – Lending Works (currently) has an extremely small share of the personal loans market, so our lending performance is not necessarily going to conform to the wider market trends. However, we are focused on ensuring that our credit performance and Shield remain robust throughout the cycle. Lending Works takes its responsibility for managing credit risk very seriously. We are continuing to strengthen our credit, analytics and risk management teams in order to further improve our ability to originate high-quality loan portfolios and provide predictable returns to investors. In February 2019, we appointed a new Head of Risk, Ines Maia, who previously managed credit risk at TSB. Ines has made a very positive impact so far and will continue to strengthen our already experienced credit and underwriting functions significantly over the coming months. Regarding the updates to the stats pages, there will be some changes happening over the next few months, but the hard deadline for the full refresh is going to be around 1 December. I think one of the major flaws of the existing stats pages is the lack of context/explanation, as r00lish67 mentions – the new pages will provide much more narrative in the form of quarterly Shield and credit ‘outcomes statements’ which explain graph variances and performance/portfolio changes over time. Regarding the logic for the calculations presented in a previous post on future losses, this is broadly reasonable, and we are working on bringing coverage levels back in line with our targets. The new stats pages will include more clarity and granular detail on expected losses etc, so you will be able to see the exact figures I think you’re looking for. It’s important to note that the Shield and the level of cover it provides is neither static nor binary – while it is possible to analyse on the basis that the last loan has been written i.e. no further contributions to the fund are possible, the reality is that it is much more fluid, both in terms of new contributions from new loans and adjustments to contributions from the spread on the existing portfolio. The Shield should help to smooth returns to investors throughout the cycle, so there will of course be points in time where static performance metrics are better or worse. The new stats pages will provide more information on our stress testing so you’ll be able to better understand the likely impact of increases in default rates on the Shield and its coverage ratios. Regarding the loss performance versus APR point, while it is indicative, using APR as a sole proxy for risk is not always recommended – for example pricing a loan at 3% APR doesn’t make it low-risk (it might be poorly assessed and underpriced) and pricing a loan at 9% APR doesn’t make it high-risk (it might have been well assessed with a sensible provision for stressed losses built in). The losses in the higher-APR loans are for sure likely to be higher, however provided that’s been priced in with a sensible buffer for losses it’s likely to offer at least the same or better loss-adjusted returns for investors. We have no interest in lending at 3% APR as we would not be able to offer our investors anywhere near the projected returns currently offered and remain a sustainable business. The final point I wanted to make is that it’s important to remember that P2P lending is an investment and investments necessarily carry risk. We sometimes receive questions from investors about whether the Shield guarantees returns – investment returns are never guaranteed, despite the use of provision/buffer funds. The highest-level objective of the Shield is to fairly distribute losses between investors across the portfolio. This helps prevent a situation where ‘unlucky’ investors end up with a poor-performing individual portfolio of loans they didn’t actively choose, while ‘lucky’ investors beat the average purely because they ended up with a good portfolio. We fully back the Shield model for this reason, as we believe it provides a much fairer outcome for investors throughout good and bad times. Apologies again for the long post but hope it helps clarify some of the points raised. |
|
|
|
Post by propman on Aug 5, 2019 17:19:52 GMT
<snip>
I am confused about your distinction between LW and RS. I see the estimate of future claims to be their estimate of future outflows. AIUI the future contributions included for both are those on loans already arranged. As a result, if this was > expected bad debt, they would be expecting that none of the current cash would be used against existing loans and so that the PF cash would increase by all net contributions from future loans (ie PF contributions less bad debts). If they structured the PF payments from future loans to achieve this, it would have to be by reducing cash taken for PF upfront and hence increase liquidity risk relative to requiring some of any increase as cash contributions upfront.
Confess I'm not quite sure I catch your drift either! Tricky to explain.. IMV, Ratesetter are slightly clearer in their stats currently on this: RS say their PF cash is £12.7m, their expected future provision fund inflows are £27.2m and their expected future losses are £33.0m. As a result (although they don't explicitly state this), their net forecast PF position is surely just £12.7m + £27.2m - £33.0m = £6.9m (0.78% of their £888m loans under management). LW say their PF cash is £1.045m, their expected future provision fund inflows (contractual future income) are £2.082m, but do not state their expected future losses explicitly as one number. They do however provide their actual lifetime bad debt rate for each year, as well as their forecast lifetime bad debt rate. So, I think we can therefore take the forecast bad debt rate and subtract the actual bad debt rate for each year in order to work out how much more call there is supposed to be on the shield. If we sum all of those differentials up, I calculate that as being £3.5m. If that's correct, then that is interesting. Ratesetter's forecast PF losses are 121% of their forecast PF inflows (£33.0m / £27.2m) Lendingworks forecast PF losses are 168% of their forecast PF inflows (£3.5m/£2.082m) That feels about right because LW's PF cash has been dropping faster than RS's in recent months. The interesting bit though is that whilst RS's net forecast PF position is a very slender amount (0.78%, as above), LendingWorks' is actually negative: £1.045m + £2.082m - £3.5m = -£0.373m (-0.41%)I will add a big caveat to the above that I'm not sure my methodology is correct. If it is correct, then that's a little concerning frankly as it would suggest LW are effectively implicitly forecasting that they're going to run out of PF cash before the end of their currently held contracts. Matthew - as it's not like you have a full-time job doing anything else ( ) , I don't suppose you'd like to check my logic there? propman , or anyone else, perhaps you'd like to check my working? Apologies I haven't managed to explain my disagreement with what I think you were saying about RS's Expected inflows, but i believe that this ignores all inflows from future loans although there wording is not definitive on this.
Matthew has covered much of this. Yes if LW made no further loans after the position currently shown and bad debts and pre-payments continued as expected, it would not have sufficient funds to repay all bad debts. the difference is somewhat less than a month's interest. However they have continued trading. The estimates are much more subjective for recent years. RS is showing 2018 PF usage at 44% against LWs 26% and for 2019 3% against LWs 1%. That and the fast that they have not fudged the expectations to show an adequate PF give me more confidence that LW's numbers are more objective and that they will take the necessary action to deal with the issue.
Personally I think both need some luck not to require the imposition of haircuts within the next 3 years.
- PM
|
|
r00lish67
Member of DD Central
Posts: 2,692
Likes: 4,048
|
Post by r00lish67 on Oct 17, 2019 11:46:01 GMT
New stats up for LendingWorks. The provision fund cash is continuing to drop and is now sitting at £582k (from £1.6m in February). This is not surprising, as a rough calculation shows that expected future income is less than expected future outflows in a ratio of 3:2, and has been similar for some months. i.e. more is forecast to be taken out then put in. So, Matthew , what happens from here? The maths continues to imply that the LW PF will run out of cash-in-hand in the near future (maybe 3-4 months). Are LendingWorks going to stump up some funds from somewhere to keep the provision fund going? |
|
macq
Member of DD Central
Posts: 1,934
Likes: 1,199
|
Post by macq on Oct 17, 2019 13:09:08 GMT
Hopefully that's not why the new portfolio projection tool was added recently with the performance based on Good-Moderate & Poor options?
|
|
) , I don't suppose you'd like to check my logic there?
).
