|
|
Post by solicitorious on Sept 17, 2015 20:19:38 GMT
I love doing simulations with spreadsheets, so I thought I'd try a Monte Carlo Analysis of the SS loanbook. What is a Monte Carlo Analysis? Basically, it is a way of simulating randomness in complex systems, and deriving useful information as a result. We run the model say 10,000 times to obtain meaningful and quite accurate results. What I am trying to model here is likely investor losses (and a histogram), for certain input parameters. Our parameters are:-D - probability of a default for any loan. Since there are currently 43 loans the model must cope with the possibility of more than one default. L - % loss, given a default. I'll base this on the valuation figure. We note at this juncture that since the maximum LTV is 70% we could have a 100% Default rate and a 30% loss rate without any investor losing a penny. PF - provision fund, fixed at 2% of the loan book, but increasing in absolute terms as the loan book expands. Currently standing at £877.060k. We notice that 27 of 43 loans are currently fully protected by the PF in the case of single default.But life is rarely that straightforward... We could have two or more defaults, including the 27, which when added together have losses which breach the PF. The loans are all different sizes as well, giving millions of different combinations of losses. What we are trying to find? The overall probability histogram of losses for various D and L (we know the current PF already, and the model will take account of future changes as they occur). Model Assumptions:The investor is fully invested in all loans. Literally they own the entire loan book, or are at least invested in perfect proportion to loan size. The model can cope with alternative investment strategies, which we may look at later (equal amounts, inversely proportional, just the largest/smallest loan, etc.) The quoted LTVs are perfectly accurate at the time of the loan and do not vary over time. The increased flow of DFLs undermines this assumption. The investor does not have an opportunity to exit or scale back any of their investments, to mitigate losses. e.g. defaults occur unexpectedly, more or less simultaneously and/or the secondary market becomes frozen. The Provision Fund is fully applied to the point of exhaustion to cover any losses. The Provision Fund is not replenished or increased via the origination of new loans between any defaults which may occur. D is a uniform random variable, and every loan has the same independent probability of default. There may be hard-to-model effects, such as varying probabilities given the different remaining terms of each loan, and new loans replacing old, etc. We'll just assume the current loans all start and end on the same day, and then the experiment ends. L is a fixed %loss (of value) applicable to all assets in the event of default. We may be able to enhance this in future to be quasi-normally distributed around a mean loss of L. Further general reading on Monte Carlo Methodswww.riskamp.com/files/RiskAMP%20-%20Monte%20Carlo%20Simulation.pdfwww.goldsim.com/Web/Introduction/Probabilistic/MonteCarlo/Now, who would like to start me with sensible worst-case values of D and L? |
|
registerme
Member of DD Central
Posts: 6,196 
Likes: 6,006
|
Post by registerme on Sept 18, 2015 1:19:47 GMT
Ask SS?
They may not answer, they may refuse to answer, but to not ask the question would be....
|
|
|
|
Post by solicitorious on Sept 18, 2015 1:25:36 GMT
Ask SS? They may not answer, they may refuse to answer, but to not ask the question would be.... Well, AFAIR there's been no losses so far. I need worst case estimates, anyhow, to plug into the model.
|
|
|
|
Post by supernumerary on Sept 18, 2015 8:24:17 GMT
I love doing simulations with spreadsheets, so I thought I'd try a Monte Carlo Analysis of the SS loanbook. Now, who would like to start me with sensible worst-case values of D and L? OK, to get the ball rolling... For simulations purposes, I have provided four scenarios, for your spreadsheet anaylsis; [1] Low asset value. Low LTV. PBL49 - Land 30, The Chase, Essex AV = £790,000, LTV = £316,000 40%[2] Low asset value. High LTV. PBL42 - Gun Boat Sheds, Gosport AV = £300,000 LTV = £210,000 70%[3] High Asset value. Low LTV. PBL51 - Home Farm House, Crowborough AV = £2,350,000, LTV = £250,000 11%[4] High Asset value. High LTV. PBL43 - Axnoller Farm, Beaminster, Dorset AV = £2,000,000, LTV = £1,400,000 70%BTW, to cover myself and you here, I know of NO problems with these loans and I have chosen these, as they are fully funded and are not at the end of their loan period, as of writing. FOR SIMULATION PURPOSES ONLY, they default in that order, what happens using your formulations and spread sheet simulation analysis?
|
|
webwiz
Posts: 1,133
Likes: 210
|
Post by webwiz on Sept 18, 2015 8:36:58 GMT
Well the worst case figures are clearly 100% for both, but you did ask for sensible estimates. I will ignore doomsday scenarios which are not directly financial, such as WW3 or a meteor strike. If we are talking worst case we have to consider the worst case market conditions, ie a full scale collapse in the property market coincident with some sort of financial crash. Then there is the possibility of fraud, most likely in the valuation where the valuer has relied on information which turns out to be untrue. I really cannot get my worst case estimate below 100% although the probability is very low.
|
|
Investor
Member of DD Central
Posts: 662
Likes: 590
|
Post by Investor on Sept 18, 2015 9:36:05 GMT
D 10.3%
L 43%
|
|
|
|
Post by ramses505 on Sept 18, 2015 10:37:43 GMT
To give the ball a little push
D = 7.5%
L = 25%
|
|
registerme
Member of DD Central
Posts: 6,196
Likes: 6,006
|
Post by registerme on Sept 18, 2015 11:34:57 GMT
So worst case guess for me would see an environment with a significant interest rate hike and a property market crash. Assuming both of these in play simultaneously:-
D - 20%
L - 40%
|
|
sam i am
Member of DD Central
Posts: 697
Likes: 555
|
Post by sam i am on Sept 18, 2015 11:49:25 GMT
OK, to get the ball rolling... For simulations purposes, I have provided four scenarios, for your spreadsheet anaylsis; [1] Low asset value. Low LTV. PBL49 - Land 30, The Chase, Essex AV = £790,000, LTV = £316,000 40%[2] Low asset value. High LTV. PBL42 - Gun Boat Sheds, Gosport AV = £300,000 LTV = £210,000 70%[3] High Asset value. Low LTV. PBL51 - Home Farm House, Crowborough AV = £2,350,000, LTV = £250,000 11%[4] High Asset value. High LTV. PBL43 - Axnoller Farm, Beaminster, Dorset AV = £2,000,000, LTV = £1,400,000 70%BTW, to cover myself and you here, I know of NO problems with these loans and I have chosen these, as they are fully funded and are not at the end of their loan period, as of writing. FOR SIMULATION PURPOSES ONLY, they default in that order, what happens using your formulations and spread sheet simulation analysis? I fully appreciate that these examples were given purely to give the model something to work with. But they also neatly demonstrate why a detailed understanding of each loan is essential. PBL42 also has a charge over a flat in Gosport valued at £110k, so the LTV could be considered to be 51%. PBL51 is a second charge. There is a first charge of £1.16m. So the true LTV is 60%. And in all cases the value that could be realised from a quick sale will be lower than the full LTV, sometimes markedly so. Go and check out the 90-day auction value of PBL 58. And then there are all manner of unknowns that could impact on security. I'm not being critical, just pointing out that in reality there are many difficulties in assessing the risk and that the risk is probably higher than we would like to think. That's why we are paid 12% and the borrower is charged significantly more.
|
|
|
|
Post by supernumerary on Sept 18, 2015 12:34:37 GMT
OK, to get the ball rolling... For simulations purposes, I have provided four scenarios, for your spreadsheet anaylsis; [1] Low asset value. Low LTV. PBL49 - Land 30, The Chase, Essex AV = £790,000, LTV = £316,000 40%[2] Low asset value. High LTV. PBL42 - Gun Boat Sheds, Gosport AV = £300,000 LTV = £210,000 70%[3] High Asset value. Low LTV. PBL51 - Home Farm House, Crowborough AV = £2,350,000, LTV = £250,000 11%[4] High Asset value. High LTV. PBL43 - Axnoller Farm, Beaminster, Dorset AV = £2,000,000, LTV = £1,400,000 70%BTW, to cover myself and you here, I know of NO problems with these loans and I have chosen these, as they are fully funded and are not at the end of their loan period, as of writing. FOR SIMULATION PURPOSES ONLY, they default in that order, what happens using your formulations and spread sheet simulation analysis? I fully appreciate that these examples were given purely to give the model something to work with. But they also neatly demonstrate why a detailed understanding of each loan is essential. PBL42 also has a charge over a flat in Gosport valued at £110k, so the LTV could be considered to be 51%. PBL51 is a second charge. There is a first charge of £1.16m. So the true LTV is 60%. And in all cases the value that could be realised from a quick sale will be lower than the full LTV, sometimes markedly so. Go and check out the 90-day auction value of PBL 58. And then there are all manner of unknowns that could impact on security. I'm not being critical, just pointing out that in reality there are many difficulties in assessing the risk and that the risk is probably higher than we would like to think. That's why we are paid 12% and the borrower is charged significantly more. Actually being critical tests the robustness of any modeling, so don't pull your punches here. It will be interesting, to see how solicitorious incorporates this additional information that you have provided into his simulated analysis with spread sheets, which he has confessed that he loves doing. I just thought I would give him a scenario to get the ball rolling, as after 12 hours nobody had provided one.
|
|
|
|
Post by nickthefool on Sept 18, 2015 12:46:00 GMT
Assuming I've understood how 2nd charges work, having a 2nd charge after a much larger 1st charge means that the % returned is very different in event of a default (if less than total LTV). For example in the one mentioned above, if the property sells for 50% of its valued price, then there would be only £15k left after repaying the first charge, so SS investors would only get 6% returned (assuming no provision fund intervention).
|
|
|
|
Post by solicitorious on Sept 18, 2015 13:49:52 GMT
OK, taking into account the repayment of PBL32, we now have 42 loans and PF of £865.160k. Model says: Chance of any loss 7.05%, no loss 92.95% loss <0.5% 4.73% [these are cumulative, i.e. the chance of a loss between 0.5% and 1% is 2.32%-0.56% = 1.76%, etc] loss >0.5% 2.32% loss >1% 0.56% loss >2% 0.03% loss >3% 0.00% loss >5% 0.00% loss >10% 0.00% loss >20% 0.00% loss >30% 0.00% loss >40% 0.00% loss >50% 0.00% overall loss 0.03%, including times when there's no loss average loss 0.43%, if there is a loss Comparison with when there's no PF Chance of any loss 96.31%, no loss 3.69% loss <0.5% 29.67% loss >0.5% 66.64% loss >1% 37.01% loss >2% 7.05% loss >3% 0.56% loss >5% 0.00% loss >10% 0.00% loss >20% 0.00% loss >30% 0.00% loss >40% 0.00% loss >50% 0.00% overall loss 0.89%, including times when there's no loss average loss 0.93%, if there is a loss
|
|
|
|
Post by solicitorious on Sept 18, 2015 13:53:04 GMT
To give the ball a little push D = 7.5% L = 25% These parameters will obviously result in no losses... [no loan has an LTV higher than 70%]
|
|
Investor
Member of DD Central
Posts: 662
Likes: 590
|
Post by Investor on Sept 18, 2015 14:07:28 GMT
OK, taking into account the repayment of PBL32, we now have 42 loans and PF of £865.160k. Model says: Chance of any loss 7.05%, no loss 92.95% loss <0.5% 4.73% [these are cumulative, i.e. the chance of a loss between 0.5% and 1% is 2.32%-0.56% = 1.76%, etc] loss >0.5% 2.32% loss >1% 0.56% loss >2% 0.03% loss>3% 0.00% loss >5% 0.00% loss >10% 0.00% loss >20% 0.00% loss >30% 0.00% loss >40% 0.00% loss >50% 0.00% overall loss 0.03%, including times when there's no loss average loss 0.43%, if there is a loss Comparison with when there's no PF Chance of any loss 96.31%, no loss 3.69% loss <0.5% 29.67% loss >0.5% 66.64% loss >1% 37.01% loss >2% 7.05% loss>3% 0.56% loss >5% 0.00% loss >10% 0.00% loss >20% 0.00% loss >30% 0.00% loss >40% 0.00% loss >50% 0.00% overall loss 0.89%, including times when there's no loss average loss 0.93%, if there is a loss Which leads us back to ' The conclusions I draw are that the LTV risks for all loans are on the minimal side. In the event of a default it looks like the chance of any capital loss is very low, and a serious capital loss virtually non-existent.
Saving Stream are to be congratulated for not exposing lenders to such risks. LTVs are however just ONE aspect of the risks involved in lending.'** **attributed to Solicitorius 20/03/2015
|
|
|
|
Post by solicitorious on Sept 18, 2015 14:14:22 GMT
So worst case guess for me would see an environment with a significant interest rate hike and a property market crash. Assuming both of these in play simultaneously:- D - 20% L - 40% Model says: Chance of any loss 12.27%, no loss 87.73% loss <0.5% 8.52% loss >0.5% 3.76% loss >1% 0.80% loss >2% 0.00% loss >3% 0.00% loss >5% 0.00% loss >10% 0.00% loss >20% 0.00% loss >30% 0.00% loss >40% 0.00% loss >50% 0.00% overall loss 0.05%, including times when there's no loss average loss 0.39%, if there is a loss Comparison with when there's no PF Chance of any loss 99.73%, no loss 0.27% loss <0.5% 13.10% loss >0.5% 86.64% loss >1% 58.21% loss >2% 12.28% loss >3% 0.80% loss >5% 0.00% loss >10% 0.00% loss >20% 0.00% loss >30% 0.00% loss >40% 0.00% loss >50% 0.00% overall loss 1.21%, including times when there's no loss average loss 1.21%, if there is a loss
|
|
