Valuing Cyclical Companies

A shipping company earns $500 million in a year when freight rates spike, and loses $100 million two years later when rates collapse — same fleet, same routes, same customers, wildly different numbers, purely because global freight rates move in a cycle nobody fully controls. Apply a standard P/E multiple to the peak year's earnings and the stock looks absurdly cheap; apply it to the trough year's loss and the model can't even produce a positive number. Neither is a sensible valuation, because neither year is a sensible base to extrapolate from.

Fixing the base, or fixing the multiple

There are two broad ways to handle this, and they address the same problem from different ends.

Normalize the earnings. Replace the current year's margin or earnings with an average across a full cycle (peak, trough, and years in between), then apply a normal multiple to that averaged, "mid-cycle" figure. This keeps the multiple stable and lets the earnings base absorb the cyclicality.

Adjust the multiple instead. Some practitioners apply a lower multiple to peak-year earnings (recognizing they won't repeat) and a higher multiple to trough-year earnings (recognizing the business is worth more than its current depressed earnings suggest, since the cycle will turn). This is the logic behind why cyclical stocks often look "expensive" on trailing P/E right at the bottom of a cycle and "cheap" right at the top — the market is already doing this adjustment, which is exactly why naive multiple screens flag cyclicals at the wrong times.

Worked example

A steel producer's operating income has ranged over the last cycle from a trough of $40 million to a peak of $260 million, currently sitting at the peak of $260 million. Comparable mature industrials trade at 10x normalized operating income.

1. Naive approach: apply 10x directly to the peak $260 million: 260 \times 10 = \2.6$ billion — almost certainly overstated, since this year's earnings won't repeat.
2. Mid-cycle normalization: averaging operating income across the full cycle (including the $40 million trough, the $260 million peak, and intermediate years) gives a normalized figure of, say, $130 million.
3. Normalized valuation: 130 \times 10 = \1.3$ billion — exactly half the naive figure, and a far more defensible estimate of sustainable enterprise value.

An analyst who values this company at $2.6 billion using peak earnings, then watches the cycle turn and earnings fall back toward $40–80 million, will find the "valuation" was really just a snapshot of one good year.

What this means in practice

Cyclical valuation should always disclose which stage of the cycle current earnings represent and how the normalized base was derived — the specific years included in the average matter, since a cycle that hasn't yet seen its next trough will bias the average upward. Asset-based valuation is also a useful cross-check for capital-intensive cyclicals, since the replacement cost of plant and equipment provides an anchor independent of where earnings currently sit.