Normalized Earnings
Normalized earnings estimate what a company would earn in an average year, once cyclical peaks, temporary windfalls, and one-time charges are stripped away. Valuation multiples only carry meaning when applied to a sustainable earnings base, and a single year, whether flattering or ugly, rarely provides one.
The math
A cyclical producer earns $6.00 per share at the top of its cycle, and the stock trades at $42, an apparently cheap 7 times earnings. Across the full cycle, earnings per share averaged $3.50, so the honest multiple is 12.
When the cycle turns and earnings fall to $2.00, the market prices the trough at 15 times: $30 per share. The buyer of the 7x illusion is down 29 percent, a $14,500 loss on a $50,000 position, in a stock that never once stopped looking statistically cheap.
| EPS | The stock | |
|---|---|---|
| Cycle peak | $6.00 | $42, an apparent 7x |
| Full-cycle average | $3.50 | $42 is really 12x |
| Trough | $2.00 | $30 at 15x |
The trap
Cyclicals invert the P/E signal. They look cheapest at the peak, when earnings are inflated and about to fall, and most expensive at the trough, when depressed earnings sit under a tall multiple just before recovery.
Screening for low P/E without normalizing serves up exactly the wrong candidates at exactly the wrong moment, and does it with the confidence of a spreadsheet.
The move
Average seven to ten years of margins, then apply the mid-cycle margin to current revenue for a normalized base. Bring equal skepticism to company-defined adjusted earnings, which tend to normalize away recurring costs with suspicious regularity.
Then locate the cycle before acting: buying a decent business at a high multiple on trough earnings routinely works out better than buying it at a low multiple on peak earnings.