Investors spend a lot of time modeling the earnings of insurance companies and I was no different. However, I never found traditional GAAP models satisfying, so I had a number of alternate ways of looking at the quality of an insurer’s earnings.
Today, I thought I’d share one as it leads to some very interesting conclusions vs. traditional analysis.
Many of you are likely familiar with the concept of an accident year combined ratio (AY CR), which is the reported CR adjusted for reserve development. I extended this to create accident year earnings.
Now, you might think, oh that’s just regular earnings backing out the development. Sure, that was the first step, but there were a number of further adjustments I’ll run you through.
But before I do, let me share the conclusion. The gap between reported earnings and the likely ultimate earnings is probably as large as it’s been in 20 years (and in a positive way).
This is happening for three reasons. Recent accident years are likely conservative, reported investment income is well below new money investment yield, and cash flow should accelerate which will drive additional NII growth.
Conservative Underwriting Margins
Let’s take a company reporting a 92 CR with 3 points of releases. Most people would say their AY CR is a 95. While definitionally accurate, it’s not really meaningful.
If the company has a long history of releasing three points every year, is its AY a 95 or 92? The answer is both. 95 is its initial AY estimate, but its developed AY will end up around a 92.
Most people just look at the 95. In reality, you need to make an estimate of what you think developed will be. If a company tends to have little development, then using the initial AY is OK, but some companies consistently pick 10+ points redundant. In that case, you need to use what you think a “normalized” ultimate AY will be.
Once you’ve done this, now you need to go back to evaluate how the current environment compares to normal. What are pricing trends? How are loss trends changing? How are those being incorporated into picks?
We have seen loss reporting come in very light since Covid, even as pricing has risen. While many expect late reporting to ultimately lead to claims catching up, the reality is there is a good chance that is an issue for future renewals not the business written during Covid.
Thus, if normally our sample company picks at a 95 with an expectation of releasing to 92, given improvements in pricing that 92 should maybe be a 90. But due to uncertainty over when loss reporting returns to normal, they keep the pick at a 95.
So now we have 5 points of cushion, not 3. So when you’re looking at AY earnings, you should probably assume a 90 CR. That’s something like a 15% increase to EPS (depending on the company’s leverage).
New Money Investment Income
Analysts are terrible at modeling NII. Frankly, a lot of times the companies are too, even though they have real time access to every bond holding vs. my rough estimates.
Why the struggle? Most people default to looking at portfolio yields and don’t adjust properly for new money. We used to have the problem where portfolio yields were 5% and companies were investing new cash at 2-3%. Now, we have the opposite problem.
This leads to big errors in modeling future investment income as investors tend to lag how the portfolio rolls over and is reinvested.
But there is a bigger issue. Even those investors who thought about AY earnings, never adjusted the NII. They combined an AY CR with a CY NII (aka the portfolio yield).
AY NII is very simply the new money yield (well, not quite so simple because you have to think about investment of surplus and case reserves waiting to be paid but let’s punt that for now).
If you want to understand the profitability of new business, you need to re-cast the income statement with a) your best estimate of ultimate AY CR and b) restated NII as if you invested all new float at today’s yields.
Note, the last part is a little tricky because of the reasons I hand waved away. The best way to think about new float is NPW less first year paids less cash expenses. That’s probably 1/2-2/3 of NPW depending on payment patterns.
You can try to then estimate the yield on maturities and model the delta between the old and new, but this is hard to do accurately and there’s a much simpler approach.
If you assume all bonds matured immediately (making sure to unwind and unrealized gains or losses in book value), you can then re-cast the entire portfolio to current yields as a shortcut.
While this may be unrealistic, it is also unrealistic to assume old bonds will be reinvested in the future at the same rate as they were bought for years ago. It is more likely that maturing bonds will be re-invested at today’s rates than last year’s rates.
For some companies, this would lead to a doubling of NII. Most companies are probably looking at a ~50% bump. The increase in EPS will depend on the investment leverage (and how strong the CR is), but it is easy to model this as a 25% increase in earnings.
Improved Cash Flow
But there is still one more piece to the puzzle. If paid losses are slowing, pricing is still strong, and investment yields are rising, then cash flow generation is improving.
This means there will be more float to invest, which will raise NII even further. While this impact on AY EPS is more marginal (it builds up over time), it can make a difference (especially for shorter tail businesses) and most investors completely overlook it.
Additionally, the improvement in GAAP earnings as the AY strength earns through results in increased capital generation which allows for higher buybacks which is another small gain for EPS.
Modeling It Out
So let’s try to put it all together and look at a sample income statement on a GAAP basis and then an adjusted basis.
| GAAP | Accident Year | Change | |
| GPW | $1B | $1B | |
| CR | 92% | 90% | |
| U/W Income | $80M | $100M | 25% |
| NII | $60M | $90M | 50% |
| Corporate Expense | -$20M | -$20M | |
| Pre-tax Income | $120M | $170M | 42% |
| Net Income | $96M | $136M | 42% |
In other words, earnings power is 42% higher than what is being reported. This will take time to emerge and, by the time it is does, sentiment may be more negative on pricing or other issues.
However, we know GAAP earnings and book value are often not representative of underlying fundamentals. That 42% number is a hypothetical, but you can certainly find companies that look like that.
If many names are actually earning 25-50% more than they are reporting, then they deserve multiples above historical norms. Many of them are already at those valuations, which is a separate discussion, but hopefully this framework is helpful in thinking about how to evaluate the quality of earnings.
Trying to subscribe to future posts!