Since AO is down again, I might as well post my question here.
This triangle has the issue of partial first development and partial last calendar period, but I don’t fully understand the proper adjustments to the ODP bootstrap model to address these issues yet.
To adjust the last diagonal, CF says “Annualize the losses to fit the model and produce residuals. Calculate LDF’s for each sample triangle based on the annualized residuals and then adjust the latest diagonal of each sample triangle back to a three-month period before producing sample reserves (using interpolated LDFs)”
Here, for the bolded part, I think the data (except AY 2016) on the last diagonal is 9 months old, so it would be inaccurate to adjust the entire diagonal back to a 3-month period which only makes sense for AY 2016?
To adjust for the first evaluation period, CF says “divide the latest accident year sample reserves by two to account for the fact that it is a partial year”
So why are we not cutting off 75% of the reserves to reflect the fact the data is only 3 months old for AY 2016?
This exam is many years behind me, but I’ll deduce from your posted solution
When you create sample triangles, every value in the triangle is using the same set of annualized residuals, so your first development period value for AY 2016 will actually be something like 4000.
All it’s saying is that when you use the sample triangles to calculate the final reserve for AY 2016 with these sample triangles, be sure to divide by 2 to account for the fact that AY2016 is only 3 months instead of 6 months.
Regarding 2017 #12, the entire diagonal is as of March 31, 2016. So, every accident year has an additional 3 months of data. For example, AY 2016 is evaluated as of 3/31/16 (thus, 3 months old). AY 2015 is evaluated as of 3/31/16 (thus, 15 months old). And so on and so forth. So, the adjustment back to a 3-month period applies to all years.
Now, regarding your second point about adjusting the reserves for AY 2016, I completely agree with you. This was something I noticed when I read through the examiner’s report when it was released back in 2017. The sample solutions only show “divide by 2.” In my opinion, the AY 2016 figures should be divided by 4 to reflect that it is only 3 months old. But, I used the solution shown in the examiner’s report to be consistent with the CAS. My guess is that they were okay with a “general solution” as described in the paper rather than using something specific to the problem. Hope this helps!