Guy Carpenter had an interesting idea buried in its recent post, Workers Compensation Reserve Risk Development: The Cat That May Be Lurking in Your Balance Sheet.
GC notes that companies have a hard time getting their comp reserves right. If they are inadequate, said inadequacy may hurt earnings as much as a catastrophe would.
Thus, GC concludes, you should spend as much money and time getting your comp reserves right as you spend getting your cat exposures right. Naturally, they have a product that helps you do that.
I’m sure they put out a fine product, but I’m actually drawn to their representation of the issue:
The chart shows AY industry results at different evaluations, relative to the first evaluation of the accident year. The 1-2 line shows the second evaluation as a % of the first evaluation. The 1-3 line shows the third evaluation as a % of the first evaluation, etc. (I should add that this is Schedule P data, so each AY is evaluated 10 times. Obviously, comp’s tail is considerably longer than that, but after 10 evaluations, you do have a pretty good idea of how a year will turn out.)
It’s a bit hard to read, so I’ve cleaned it up a bit:
(My data, from SNL Financial, only go back to 1987. )The blue line is the second estimate of workers’ comp reserves, relative to the initial estimate. The red line is the latest estimate, relative to the initial estimate. The dotted line represents accident years that haven’t been evaluated 10 times yet.
Take AY 1987, to the far left. After the second evaluation (occurring in 1988), the new estimate was 1.2% higher than the original estimate. And the latest estimate was 8.2% higher than the original.
One remarkable thing, I think, is how the chart looks like two coincident sine waves of different amplitudes. It’s the best evidence I’ve seen of the old saw that bad years keep getting worse while good years keep getting better.
It also seems to indicate there may be some redundancy remaining in the 2003-2007 reserves, while the bad news from 2008-2010 may have just begun.
Another remarkable fact is the way that the ultimate for a given year tends to either continually improve or continually deteriorate:
This table shows the change in an AY estimate from one period to the next. (I’ve highlighted favorable development in green.) The change in AY1987 from the 24-month evaluation to the 36-month was 1.3%, for example. A year later, the estimate was 0.9% higher.
Recall that these are estimates of ultimate losses, so the increases and decreases should be more or less randomly distributed across the grid. But of course, they are not. Once a year starts to improve or deteriorate, it tends to continue on that path.
Of the 23 years, only 10 exhibit any change in direction at all. Four of those change direction exactly one time.
In other words, the bad years keep getting worse, and almost every year they get incrementally worse. The good years keep getting better, and almost every year they get incrementally better.