The disease: extrapolating a point on a sine wave
Steel, cement, chemicals, mining, oil & gas, autos, capital goods, shipping, semiconductors, airlines. For every one of these businesses the latest year is a location on a cycle, and you can't use it as a base. Prices are set by global supply-demand balances that no single firm controls. Margins swing from famine to feast and back. Every valuation input you learned to take from "the most recent fiscal year" (margin, growth, even ROIC) is contaminated by cycle position. The cure is normalization. You estimate the business's mid-cycle earnings power, meaning what it earns across a full cycle. Then you value that, and you refuse on purpose to capitalize either the crest or the trough.
Damodaran's mnemonic for all of this is to value the company as if it were frozen at mid-cycle, and to let the current cycle position show up in the price. Keep it out of your estimate of value.
Advertisement
The peak-P/E trap (drill this until reflexive)
Because the market partly anticipates mean reversion, cyclicals look optically cheapest at the peak and dearest at the trough:
- At the peak the earnings are 2 to 4× mid-cycle, so the P/E prints 4 to 6× and the screens scream "value!" Then earnings halve, and the "6× bargain" becomes a 15 to 20× stock on the way down. The price follows earnings.
- At the trough the earnings are tiny or negative, so the P/E prints 40×, 90× or "N.M." and the screens scream "avoid!" Historically, that's where the best cyclical purchases were made. Buyers used P/B, EV/capacity or normalized earnings, never the meaningless trough P/E.
The old floor-trader's rule, buy cyclicals at high P/Es, sell them at low P/Es, is just this arithmetic. Here's a real-shaped illustration, modeled on the scale of Tata Steel's FY22 to FY24 actuals (pull the filings and verify). In FY22's global steel-price spike, consolidated profit reached roughly ₹40,000 crore and the trailing P/E compressed to ~4 to 5×. By FY24 profit had collapsed toward ₹5,000 crore. The stock had fallen a long way from its peak price, yet on trailing P/E it looked several times more expensive. Anyone who bought at "4×" bought the crest. The US mirror is Nucor, which printed ~$23 of EPS in 2021 (~5× P/E at the time) against a plausible mid-cycle EPS nearer $10 to 12. On normalized earnings that "5×" was really ~10 to 11×, which is ordinary for a well-run cyclical.
Normalization method 1: through-cycle averages, anchored to today's size
- Average margin method. Take 7 to 10 years of operating margins, which should span one full cycle, so check that your window really contains a peak and a trough. Average them and apply the average to current revenue. That gives you normalized EBIT, and you run the standard machine on it.
- Average ROE method, Graham's updated:
normalized EPS = 10-year average ROE × current book value per share. Current book keeps the estimate in step with a company that has grown. Average raw historical EPS instead and you'll understate a firm whose capital base has doubled.
Normalization method 2: mid-cycle unit economics × current capacity
For commodity producers, rebuild earnings from physical units: mid-cycle EBITDA = mid-cycle margin per unit × current volumes. Take the mid-cycle price assumption from the industry cost curve. Over time, commodity prices gravitate toward the cash cost (plus sustaining capital) of the marginal producer, roughly the 90th percentile of the global cost curve. Prices above that attract supply, and prices below it force closures. A steelmaker's mid-cycle EBITDA/tonne, a cement company's mid-cycle EBITDA/bag, an airline's mid-cycle spread per seat-km: it's the same template. Worked Example 2 runs a full one, valuation included. You take mid-cycle EBITDA × mid-cycle EV/EBITDA, subtract net debt to reach equity value, and compare that with the price. For commodity converters 5 to 6× is typical. They deserve low multiples at mid-cycle because their ROIC barely clears WACC across the cycle.
Normalization method 3: asset-based cross-checks
- EV per unit of capacity against replacement cost. Say it costs ~$700 to 900 for each tonne of integrated steel capacity you build, which is ₹6,000 to 7,500 crore per million tonnes. A company trading at an EV of 40 to 50% of replacement cost is then priced for permanent famine, and historically that's the zone where cycles bottom. An EV above replacement cost invites new supply (why build a moat when rivals can buy a greenfield cheaper?) and marks froth. Cement analysts speak fluent "$ per tonne EV", and so should you.
- P/B through the cycle. Book value moves slowly, which makes P/B the stable-denominator ratio for cyclicals. Historically, buy zones cluster at deep discounts to book. Remember the caveat from the earnings-games work, though. The book has to survive an impairment inspection.
What is the reinvestment trap for cyclical companies?
Then there's the reinvestment trap, the value-drivers lesson in cyclical dress. Across a full cycle most commodity businesses earn ROIC ≈ WACC or worse. The boom years fund capacity additions at the top, with capital allocated at peak prices, and those additions crush returns in the bust. So when you normalize earnings, normalize for how management allocated capital too. A management that expanded counter-cyclically, or returned the boom cash, deserves a better multiple than one that broke ground at the peak. The capstone rubric grades that.