When we look at a stock, we tend to treat the financial statements as objective truth. A company reports $100 million in net income, and we accept that as how much money the business made.
For most industries, that's a fair approximation. For some, such as insurance and banking, it isn't, and the gap between reported earnings and real earnings is where investors get hurt.
The reason is structural: both industries get paid up front and pay out later.
Insurers collect premiums today against claims that may arrive in five, twenty, or fifty years.
Lenders book interest income today against loans that may default in two years or five. GAAP doesn't ignore this, it requires companies to estimate the future costs and reserve against them. But those estimates are guesses, and guesses about rare or distant events tend to look generous when times are good.
Two examples make the problem concrete.
Insurance: when the bad year hasn't happened yet
Insurance lines fall along a spectrum from high frequency to low frequency.
High-frequency lines: pet insurance, auto, basic health, are relatively easy to read. Claims happen often enough that an insurer can build a reliable picture of true loss costs within a reasonable period of time. Reported numbers can track reality pretty closely.
Low-frequency lines are different. Take Palomar $PLMR, which writes earthquake coverage. A major quake might strike tomorrow or not for eighty years. Palomar collects premiums every year either way, sets aside loss reserves based on its catastrophe modeling, and reports the difference as profit.
If the modeling is right, the reported profit is real. If it's wrong, or if the company has been pricing a once-in-a-century event as once-in-two-hundred, then every year of premium collection makes the reported P&L look better and the underlying business look more profitable than it is.
Cash from operations looks especially clean: money comes in, very little goes out, and the eventual claim payment may be a decade away.
A classic example of this is asbestos liability. General liability policies written from the 1950s through the 1970s never priced for asbestos claims because no one knew they were coming. Insurers reported profits on those policies for decades. Claims didn't peak until the 1990s and 2000s. Lloyd's of London nearly collapsed in the early '90s as the bills finally came due. A whole generation of "profitable" underwriting turned out to be a wealth transfer to past policyholders.
GAAP isn't lying in these cases. It's faithfully reporting management's best estimate. But the estimate is just that. The question is whether the estimate is right.
Banking: when the loan looks good until it doesn't
Banks face the same problem. They make a loan, project expected credit losses over its life, book interest income against that expectation, and report the spread as profit.
For short-duration, high-volume consumer lending, such as credit cards, this works reasonably well. Losses develop quickly and management adjusts. For long-duration loans, or any portfolio exposed to a macro cycle, the estimated loss rate can drift far from reality before anyone notices. The gap between projected and actual losses tends to widen during the boom and snap shut during the bust.
The 2020 accounting standard known as CECL (current expected credit loss) was supposed to help by forcing banks to provision for lifetime expected losses up front rather than waiting for losses to materialize. It did help, but it didn't solve the underlying problem, which is that lifetime expected loss is still a guess.
Goeasy, a Canadian non-prime consumer lender, gave a textbook demonstration in March 2026.
The company's stock fell roughly 60 percent in a single day after it disclosed about $331 million (Canadian) in Q4 charge offs from its LendCare auto and powersports financing division, suspended its dividend, and admitted it would need to correct historical reporting practices going back to 2024. Specifically: some payments had been recorded as received when they hadn't been.
Nothing fundamental about the borrower base changed between the previous quarter and the quarter where the stock collapsed. The borrowers were already in trouble. What changed was that the accountants stopped letting management treat past-due loans as if they would still be paid. The economics revealed in March had been the actual economics for at least a year. The reported economics had been a fiction the whole time.
What this means for investors
The takeaway is not to entirely avoid insurance and banking stocks, though it might be a good idea.
Both industries can produce fantastic long-term returns. Berkshire Hathaway and JPMorgan, for example.
The takeaway is that reported earnings in these businesses are less reliable than they look, and the standard analyst toolkit (P/E, EPS growth, last quarter's return on equity) tells you less than it does for, say, a software company.
If you are researching one of these stocks, three things are worth checking:
For insurers writing long-tail or catastrophe lines, look at prior-year reserve development. Insurers are required to disclose whether their previous loss estimates turned out to be too high or too low. A pattern of "adverse development" where reserves that repeatedly prove inadequate is a warning that current reserves are probably also too low. Combined ratios should be averaged across a full underwriting cycle, not read off a single year.
For banks and consumer lenders, compare the loss provision rate to actual charge-offs over time. A widening gap between what management is reserving and what's actually defaulting is exactly the red flag goeasy was flying. Loan growth is another tell: aggressive growth late in a credit cycle usually means the marginal new loan is worse than the average old loan, even if the numbers don't show it yet.
For both, be skeptical of cash from operations as a measure of profitability. In businesses that collect money up front and pay it out later, strong operating cash flow mostly tells you the business is growing. It does not tell you the business is profitable on the policies and loans it is writing today.
None of this requires becoming an accountant. It requires holding the reported numbers a little more loosely than you would for a normal business and remembering that in insurance and banking, the difference between a great year and a slow-motion accident is often invisible until it isn't.
Our content is provided subject to our Disclaimer.