|
|
Post by vaelin on Mar 25, 2018 12:16:04 GMT
I'm trying to get a better look at whether loan performance really is degrading as many people here seem to believe. To do a meaningful analysis, I need a few old copies of the loan book so I can compare loan performance at similar stages in the life cycle of the loan.
If anyone has downloaded an old copy of the loan book which is still sitting on their computer, please attach it to a comment for me.
|
|
Gruff
Posts: 63 
Likes: 68
|
Post by Gruff on Mar 26, 2018 12:03:17 GMT
Oops, ignore this
|
|
baldpate
Member of DD Central
Posts: 549
Likes: 407
|
Post by baldpate on Mar 26, 2018 16:14:04 GMT
|
|
cobi
Member of DD Central
Posts: 76
Likes: 80
|
Post by cobi on Mar 27, 2018 8:52:01 GMT
Is it not possible to do all the analysis from the current loan book?
You know when each loan was accepted and on defaulted loans how many payments remaining so when they defaulted.
|
|
|
|
Post by vaelin on Mar 28, 2018 8:07:41 GMT
Is it not possible to do all the analysis from the current loan book? You know when each loan was accepted and on defaulted loans how many payments remaining so when they defaulted. I hadn't noticed that. I will see what i can do.
|
|
cobi
Member of DD Central
Posts: 76
Likes: 80
|
Post by cobi on Mar 28, 2018 10:30:12 GMT
Here is a chart showing the number of loans that defaulted within the first 6 months by week of issue. Starting in 2015 and going up to the end of 2017 ![]()  |
|
|
|
Post by vaelin on Mar 28, 2018 11:56:49 GMT
I found the percentage rate of defaults for loans accepted within the trailing 12 months once per quarter since the start of the loan book. I then found the average age of the most recent 12 months of loans, then weighted the results from past years to reflect the average age of current loans. Using that model, it does not appear that defaults have increased as a percentage of total loans over the last 12 months. In fact it suggests defaults have decreased. However, my model assumes that loans are equally likely to default at any point during their first 12 months, but that isn't the case. Loans are almost 3x more likely to default in month 6 than month 1, and are just under twice as likely to default in month 12 than month 1. I don't know how to adjust my results for that. It'll be much easier to do this analysis with older loan books where the most recent 12 months are already matching weight. Fortunately, baldpate is supplying me with some. I will post more detailed findings when I have done that. |
|
cobi
Member of DD Central
Posts: 76
Likes: 80
|
Post by cobi on Mar 28, 2018 12:12:23 GMT
This is what I did.
Add two columns to loan book
1. Year and week number of loan acceptance
2. If defaulted how many months loan was paid before default ( term - payments remaining )
Pivot table on year/week with count of defaults and a filter on months paid
|
|
