merlin
Minor shareholder in Assetz and many other companies.
Posts: 902 
Likes: 302
|
Post by merlin on Jan 12, 2014 18:41:20 GMT
There weren't many comments on my December analysis so presume others don't have the same interest in the subject as I do. Not a huge number of viewings either. Well, I did it for myself in the first place and am happy to share it with that minority here that ARE interested. The attached is a development of that first analysis with more rigorous comparisons with FC's estimated default rates. For the sceptics out there who doubted my conclusions the first time, I must forewarn you that little has changed in that part. But check out the analysis and give me your arguments if you think I am wrong. For my part, I believe FC's estimated default rates are realistic, except possibly for the C- loans. No one really knows how they are likely to pan out. I don't intend to repeat this analysis so a plea to FC if you come to this place - publish your default model so that those who are interested (not many!) can set up their own calculation. Well done, a great piece of work. This certainly points up a number of significant factors that are not immediately available from FC. I find your Table 3 Number of Defaulted Loans particularly interesting as this is much closer to my experience than the FC claimed default rate. I share with you concerns that the actual default rate over time is likely to prove to be larger than that currently claimed by FC. However my experience indicates a falling off of the large number of "late payments" so it is possible that things could improve.
Again Well Done. I look forward to future updates.
|
|
maxmarengo
Member of DD Central
Posts: 96
Likes: 28
|
Post by maxmarengo on Jan 19, 2014 8:49:09 GMT
Another way to look at this topic is to follow the trend in Bad Debts compared to the expected bad debts based on FC's suggested rates:  The expected bad debts (blue line) are calculated using a formula suggested by SL75. Using FCs Estimated Annual bad debt by risk category, for each loan you calculated what amount of default would be expected based on age of the loan and average principal outstanding. The good thing about the formula is that it takes account of the changing age and risk profile of the FC portfolio. The main weakness is that it assumes that defaults are spread evenly across the life of a loan - there is some evidence that risk rises with age. The red line shows the actual bad debts (defaults-recoveries). You can see that after a bad run up to the end of summer 2013, bad debts have reduced considerably, so that, for the first time since I have been monitoring, they have dropped below expected. The green Ratio% line should be read against the RH scale. This is the ratio between the Actual and expected. You can see there has been a downward trend for more than a year. The orange line is recoveries. There seems to be little regular recovery, but there are occasional steps up, when a big win is made. Finally, the light blue line shows the quantity of loans late > 90 days. This has increased significantly in the last month and is now more than 30% of the defaulted figure. I think this may explain some of the apparent success in "reducing" defaults. More on this in my next post. |
|
maxmarengo
Member of DD Central
Posts: 96
Likes: 28
|
Post by maxmarengo on Jan 19, 2014 9:05:32 GMT
When I saw the post on Zombie loans yesterday, I decided to revisit this analysis:  For illustration I have added the red dashed line. This is bad debts + 50% of the late loans. I know, 50% is an arbitrary figure, but it illustrates the impact of the lates. Now there is still a gap between bad debts+lates and expected. The total ratio (green dashed, RH scale) is above 110%. BUT the picture is improving, even when you add in a proportion of the lates. Even adding in all the lates the picture has improved over the last 6 months. Conclusions 1) The bad debt picture has improved over the last year and over the last 6 months, even allowing for lates. 2) There is still some evidence that the bad debts may be 10 to 20% higher than FC's predictions. On the positive side, the improving trends may continue. Historical performance may be partly due to some early bad loans that are still impacting the bad debts total and FC's performance in avoiding bad loans seems to have improved, gradually diluting the impact of the early failures. On the negative side, the portfolio is still young, so we may not have seen the worst period for defaults yet. |
|
maxmarengo
Member of DD Central
Posts: 96
Likes: 28
|
Post by maxmarengo on Jan 19, 2014 9:22:06 GMT
Final chart, then I will go quiet for another month. This one is about the age of the FC portolio:  Here I am looking at the age of the FC portfolio, calculated in principal days. The red line shows the age I have observed, since I have been logging the total portfolio. Over the last year, age has risen from about 200 to about 220 days. This relatively flat trend has been because new loans have been added to the portfolio at an increasing rate. To flesh out the picture, I have simulated the portfolio back to FC's creation. The green line shows how the age will have grown with the actual portfolio. Rapidly at first, then flattening off as the smaller principal on the older loans is outweighed by the new loans. The blue line shows what would have happened to age if FC lent the same amount every month. I have used the actual mix of loans/loan durations, scaled so there is no growth in monthly lending. The age stabilises at around 355 days. The current age is around 2/3 of the age we would expect from a mature, stable FC loan portfolio. At some point, the FC portfolio will head that way. If, as I believe, defaults increase as loans age, an ageing portfolio will increase the default rate. For comparison - the current age of my loan portfolio is 151 days! |
|
|
|
Post by GSV3MIaC on Jan 19, 2014 21:41:32 GMT
Another way to look at this topic is to follow the trend in Bad Debts compared to the expected bad debts based on FC's suggested rates: The expected bad debts (blue line) are calculated using a formula suggested by SL75. Using FCs Estimated Annual bad debt by risk category, for each loan you calculated what amount of default would be expected based on age of the loan and average principal outstanding. The good thing about the formula is that it takes account of the changing age and risk profile of the FC portfolio. The main weakness is that it assumes that defaults are spread evenly across the life of a loan - there is some evidence that risk rises with age. I did the sums and there are significant (to 5% confidence level) variations in failure rate with repayment month .. the first two or three are significantly better, months 9 and 20 are significantly worse, than the average. Exactly when the problems start(ed) is unclear, obviously (usually) before the last payment made, which is all you can determine from a quick look at the loan book data. By my sums the failure rates (and I didn't use as much as 50% for most of the bad debts) are mostly worse than FC estimates, particularly on the earlier A/A+/B loans - the whole thing is (somewhat) rescued by the better than predicted performance (so far) or C and C-. Even backing out the effect of repayment month (i.e. allowing for the good performance we expect in months 1-8) the C- loans are doing way better (significantly better) than expected/predicted. Whether it will last when they get into the month 9+ period .. we shall have to see. I do think FC have tightened up their credit checks in the last year or so .. there are some earlier periods (I only analysed by quarter) where the failure rates were best described as 'B****y awful' - again, even after you comb out the analysed effects of the loans being older.
|
|
