Note: Don’t miss today’s other post on AM Best’s new innovation standard!
Punchline:
10K reserve disclosures need to require earned premium data.
I’m going to channel my inner Chris Berman and quote a 50 year old song and act like everyone should get the reference. So here goes: 10K reserves…what are they good for? Absolutely nothing!
How We Got Here…
OK, maybe that’s a little unfair, but only a little. First, let’s review how we got here. It all began in the early 2010s. A desolate time when the FASB wanted to overhaul accounting standards for all insurers and create pain and misery for investors and corporates alike.
In the industry’s darkest hour, a band of heroes 🦸♂️ came together, who would become known as the “coalition of the unwilling”, to try to restore balance to the universe and limit this destruction with an alternative called “targeted improvements”.
After much begging and pleading, this brave band of heroes were able to turn back the advancing forces and roll back or delay most of the accounting changes and balance was restored to the universe. At least that’s how I remember the story!
One of the compromises all sides agreed to is there should be more disclosure about P&C reserves and the assumptions behind them. Who could object to this? It would be like saying you were against puppies or cute kids selling lemonade (actually, there are people against the latter).
So, FASB mandated more disclosure on reserves…and it turned into a great exercise in futility.
Where are we now…
First, the good: we get incurred and paid triangles by product lines. This is potentially very useful and it can be interesting to know the $ of adverse development by accident year. This is the end of the good section.
Next, the bad: There is one really important piece of information missing. Earned premium by product line. Without EP, I can’t calculate any ratios so the disclosed reserve triangles are largely useless. Ratio analysis is at the heart of reserve analysis. (Aside: if you read my entry about quant investing, you might be thinking to yourself, dividing loss $ by premiums to create a ratio is an attempt to make the data stationary…if so, you’re exactly right!).
The Importance of Ratios
Why do ratios matter? Because nearly all the interesting conclusions about whether reserves are getting better or worse come from looking at sequential changes in loss ratios.
For example, if I see the reserves for commercial auto written in 2016 (aka accident year 2016, or AY16) went up by $50M in 2017 and $30M more this year, that doesn’t tell me anything other than “reserves got worse”. There is no context. I want to know did the 16 AY LR go from 68.3 to 68.5 or 70.2.
Additionally, I want to know, if AY16 went up to 70.2, did AY18 also start around 70? Or did the company ignore the recent trend and go back to 68 again? While this alone isn’t enough information to draw strong conclusions (maybe they raised pricing, maybe they shed bad accounts), it is the launching point for further analysis of say paid or IBNR levels. Without EP, I can’t identify which parts of the triangle merit further work.
Example #1: Progressive
Let’s do a couple examples of why the missing EP is so important. We’ll start with Progressive (PGR). Observers of the company know they have grown a lot in recent years. So when you see first year incurred loss for direct auto liability grow from $2.95B in AY14 to $4.90B in AY18 (66% cumulative), what might seem a startling growth in reserves may not be a sign of anything out of sorts. However, without the ability to see if premiums also grew near 2/3 we are unable to tell if PGR’s view of profitability has changed.
| Direct Auto Liability | AY14 | AY15 | AY16 | AY17 | AY18 |
| Initial Incurred | $2.947B | $3.331B | $3.819B | $4.210B | $4.905B |
I did have one brief moment of hope when I saw PGR actually does provide earned premium by business (personal lines agency, personal lines direct, commercial, and property). However, when I flipped back to the reserve tables, I remembered (with the exception of property) that they break the business segment reserves into liability and physical damage (e.g. agency liability and agency physical damage). Alas, back to square one.
There is one useful tool we can derive from the 10-K. We can use the paid loss triangle to calculate paid/incurred ratios and, indeed, those have improved for PGR in auto liability over the last five years. However, they have undergone mix changes which might have skewed the timing of payments so it is preferable to have other methods of corroborating the paid improvement.
| Direct Auto Liability | AY14 | AY15 | AY16 | AY17 | AY18 |
| Initial Paid/Incurred | 48.0% | 46.4% | 46.6% | 45.4% | 45.6% |
Example #2: Chubb
The next example is Chubb (CB), specifically their North America liability line. This line has shown initial adverse development (i.e. from year one to year two of a given policy year) for the better part of a decade. For most accident years, the development continues adverse for about three or four years and then starts to reverse. In fact, on a cumulative basis, every AY from 2013 and prior has unwound all of its initial adverse development to become cumulatively favorable.
The question before the house is whether 2014-18 will continue this “start bad, end good” trend. There is one concerning element in the K as the initial adverse development (so from the initial pick to 12 months later) is greater than in the AY13 & prior period. However, without any premium data to compare this to, it is impossible to say anything definitive.
| NA Liability | AY11 | AY12 | AY13 | AY14 | AY15 | AY16 | AY17 |
| % Adverse @12 mo. | 2.4% | 2.1% | -0.1% | 1.4% | 4.2% | 1.7% | 5.3% |
I did search for any premium disclosures at a product level and we again got close, but no cigar. CB discloses premium by workers’ comp, commercial casualty, professional liability, etc. but there are a few problems.
First, this is written premium, not earned premium which means it is on a different basis than reported losses. Second, they are global premiums by product where the reserve tables are split into North America and Overseas. Finally, Chubb shows the reserves by “liability” and “other casualty” not commercial casualty, professional liability, etc. So long story short, there is no useful premium data that reconciles with the reserve disclosure.
Recommendation
We’ll leave it there. I could do similar examples with many other companies. Suffice it to say, my recommendation to the FASB is as follows:
Require earned premium disclosures in the reserve triangles!
That’s it. It’s a simple fix. Please make it happen!