Note · Opinion, not rated
The one-time item that happens every year.
Nobody is lying. The item really is non-recurring, it really was disclosed, and the accounting is correct. It just makes this period look like something the next one will not be, and the multiple you compute on top of it inherits every bit of that.
Two companies, one mechanism
The last two reports we published turned on the same move, in industries with nothing in common. Neither involved anything improper. Both involved a number that was real once.
Nike, fiscal fourth quarter
a one-time recovery of previously paid tariffs
- Reported earnings per share
- $0.72
- Of which the recovery
- $0.52
- Earnings excluding it
- $0.20
- Share of the quarter
- 72%
Roughly $986M, worth about 900 basis points of gross margin in the quarter. Across the full year, reported earnings of $2.10 become $1.77 once normalized, and the multiple moves from 18.3 times to 21.7 times, which is 19% higher than the screen said.
UnitedHealth, second quarter
net favorable prior period development, meaning money released from reserves set aside for earlier claims
- Reported medical care ratio
- 86.7%
- Favorable development
- $860M
- Ratio excluding it
- 87.7%
- Difference
- +100 bps
A point of medical care ratio is the difference between a good year and a bad one for a health insurer. The reported improvement is real. Roughly a point of it came from reserves rather than from the business getting better at pricing risk.
Different industries, different accounting, identical shape. Something genuinely non-recurring lands inside a reporting period, the headline number improves, and every ratio built on that number improves with it.
Why small items move multiples so much
The arithmetic is worse than intuition suggests, because the item is removed from a base that is already small. A gain worth a quarter of reported earnings does not overstate them by a quarter. It overstates them by a third, because you take it out of the numerator and the denominator shrinks with it.
| A one-time gain worth | Reported | Real | Overstated by |
|---|---|---|---|
| an item worth a tenth of reported earnings | $1.00 | $0.90 | 11% |
| a quarter of reported earnings | $1.00 | $0.75 | 33% |
| Nike's fourth quarter | $0.72 | $0.20 | 260% |
| half of reported earnings | $1.00 | $0.50 | 100% |
Nike's quarter is the extreme case and it is not hypothetical: the item was 72% of reported earnings, which means the underlying business earned about a quarter of what the headline said.
When adjusting is honest, and when it is laundering
We use adjusted figures constantly, including in the UnitedHealth report, where we treated a 2.8 billion dollar restructuring and cyberattack charge as genuinely non-recurring. So this note is not an argument that adjusted numbers are fake. It is an argument about which direction the adjustment runs.
Honest adjustment removes items in both directions. A company that strips out one-time costs but quietly keeps one-time gains is not normalizing, it is selecting. The tell is not the existence of an adjustment. It is the asymmetry of them.
The three-year test. The single most useful question is whether the same category of one-time item appears three years running. Restructuring charges every year for five years are not restructuring charges, they are operating costs with a better name. A tariff recovery is credibly a one-off. A reserve release in every quarter is a reserving policy.
Watch who benefits from the framing. When a company excludes a cost, it is arguing its earnings are better than reported. When an analyst excludes a gain, they are arguing the opposite. Both can be right, and the direction of the adjustment tells you whose interests it serves before you evaluate the merits.
Cash is the referee. An adjustment that never shows up in cash flow, year after year, is describing something other than economics. Reserve releases are a good example: real money, correctly accounted, and not a signal that the business improved.
What we actually do with it
Normalize first, then value. Never the reverse. If you form a view and then decide which items to strip, you will find exactly the adjustments that support the view you already had, and you will be able to defend every one of them individually.
This is the same discipline as normalizing across a cycle, applied to a single period rather than a decade. A cyclical peak flatters earnings through the business cycle. A one-time item flatters them through a single quarter. Both produce a multiple that measures the moment rather than the business, and both are invisible on a screen.
It is also why our reports state the treatment explicitly rather than burying it. If we exclude something, the report says what and why, and a reader who disagrees can add it back with a number we have given them. An adjustment you cannot reverse is not disclosure, it is an assertion.
The headline number is not a lie. It is an answer to a question about last quarter, being used to answer a question about next year.