Security Analysis Ch. 6: Normalizing Earnings and the Cyclical Trap

"At the peak of prosperity a cyclical business shows its lowest price-earnings ratio; at the depth of depression, its highest. He who invests on that ratio alone will buy at systematically wrong moments." — Benjamin Graham and David Dodd

Investment Context

This is the most practical and most counterintuitive lesson in Security Analysis. It explains a phenomenon that has confounded investors for nearly a century: the simplest valuation metric gives exactly the wrong signal precisely when you most need it.

The Wall Street Translation

1. The Mechanism of the Trap

Consider a steel company across a full cycle:

| Cycle position | EPS | Price | P/E | Surface reading | |---|---|---|---|---| | Peak prosperity | $10 | $80 | 8× | "Very cheap!" | | Mid-cycle | $4 | $50 | 12.5× | Reasonable | | Trough | $0.50 | $25 | 50× | "Far too expensive!" |

Note the right-hand column: acting on P/E alone, you buy at $80 and sell at $25 — precisely inverted.

The cause is that cyclical earnings swing far more violently than prices, so a low P/E typically marks the peak of earnings rather than a low in price.

2. How to Normalize

Graham's remedy replaces current earnings with average earnings across a full cycle:

• Take average EPS over 7–10 years, spanning at least one complete cycle;
• Compute the P/E from that normalized figure;
• In the example above, normalized EPS is roughly $4, making the true peak P/E 80÷4 = 20×, not 8×.

That single adjustment converts "very cheap" into "expensive."

3. Which Industries Require Normalization

• Must normalize: commodities, steel, chemicals, shipping, semiconductors, autos, real estate, banks (credit loss cycles).
• Less necessary: utilities with stable revenue, consumer staples, subscription businesses.

The test: look at the last ten years of earnings. If the best year exceeds the worst by more than threefold, the business is cyclical.

4. An Apparent Conflict With How to Make Money in Stocks

O'Neil demands quarterly earnings growth above 25%, while Graham warns that rapid growth may signal a cycle peak. These do not conflict — they apply to different holding periods and exit rules. O'Neil pairs his criteria with a mechanical 8% stop; Graham intends to hold for years.

The key: any valuation method must be used together with your holding period and exit discipline. Extracting a single metric in isolation guarantees errors.

5. Normalize Margins as Well as Earnings

Beyond earnings, margins require normalization too. A company's operating margin may be 25% at a cycle peak against a ten-year average of 12%.

Extrapolating peak margins forward is among the most common systematic errors in analysis — it converts a cyclical boom into an apparent structural improvement.

In practice: compute the ten-year average margin and measure how far the current figure sits from it. The larger the deviation, the more conservative your forward assumptions must be.

Actionable Trading Rules

1. Always use ten-year average earnings for cyclicals: Never value a cyclical business on a single year.
2. Treat an unusually low P/E as a warning: A ratio far below industry and historical norms should first be read as a signal that earnings are peaking, not as evidence of cheapness.
3. Look at the earnings curve, not the earnings number: Pull the ten-year EPS series; the shape of the variation carries more information than the latest figure.