Normalised Earnings for Cyclical Multiples Explained

When you apply a price-to-earnings (P/E) multiple to a cyclical business, you risk vastly overstating or understating value if you use peak or trough earnings. Normalised earnings—a smoothed average of profitability across a full economic cycle—let you apply multiples consistently, showing what the business is truly worth at mid-cycle profitability.

Why Raw Earnings Mislead for Cyclical Companies

Cyclical companies—mining, steel, automotive, oil and gas, retail—earn dramatically differently depending on where the economy sits in its cycle. During booms, a mining company might report earnings that look cheap at a 5× P/E multiple. During a bust, identical assets trade at 30× because earnings have collapsed. Neither snapshot tells you the company’s real earning power.

If you valued both using raw earnings, you’d conclude the bust-era stock is massively overpriced. In reality, it’s the same business; the cycle has simply shifted. Normalised earnings correct this by replacing one year’s distorted figure with an average across a full cycle—typically 3–10 years depending on the industry—to estimate what the company earns at neutral economic conditions.

How Analysts Calculate Normalised Earnings

There are three main approaches:

Cycle average. Sum after-tax earnings (or EBIT, or EBITDA) over 5–7 years, divide by the number of years. Simple, transparent, but assumes the historical cycle is representative. If the business has structurally changed (say, a steelmaker closed half its mills), the older data misleads.

Trough-to-peak midpoint. Identify the most recent full cycle—from one trough to the next—calculate the average of the peak and trough earnings, then use that midpoint as normalized. This method is faster but ignores the many intermediate quarters.

Regression or trend-fitting. Plot earnings over time and fit a line or curve to remove cyclical noise. More sophisticated, but requires judgment about which method fits the data and whether underlying growth trends have shifted.

Whichever method, the goal is the same: replace “this year’s earnings” with “what earnings would be if we were at mid-cycle.”

Example: A Mining Operator

Imagine a gold miner whose annual earnings per share over a 7-year cycle were:

Cycle average: ($0.50 + $1.80 + $3.20 + $2.50 + $1.10 + $0.40 + $1.90) ÷ 7 = $1.49 per share

Now suppose the stock trades at $22.35.

• If you use Year 4’s peak earnings ($3.20), the P/E is 7×—tempting, but misleading; the company won’t sustain that.
• If you use Year 6’s trough ($0.40), the P/E is 56×—terrifying, but also false; earnings will recover.
• Using normalized earnings of $1.49, the P/E is 15×, reflecting the company’s mid-cycle earning power and letting you compare it fairly to non-cyclical peers or to historical valuations.

Applying the Normalized Multiple

Once you’ve calculated normalized earnings, you apply a mid-cycle multiple—a P/E or EV/EBITDA ratio that reflects what the market typically pays for the business when it’s not at an extreme of the cycle.

If the market has historically valued the miner at 12–15× normalized earnings, then a current P/E of 15× suggests fair value. A P/E of 10× suggests the stock is cheap (either the cycle is bottoming, or the market has permanently downgraded the company). A P/E of 20× suggests it’s dear.

The discipline here is crucial: you’re not picking a single “magic multiple.” You’re anchoring to a range the market has paid over a full cycle, then asking whether today’s valuation sits above, within, or below that historical band.

When Normalization Fails

Normalized earnings assume the past cycle repeats. But if a cyclical business fundamentally changes—new technology, regulatory shift, permanent demand collapse—the old cycle is not predictive.

A coal producer normalized over 2000–2010 would appear cheap if valued at pre-2010 multiples in 2020, yet coal demand has permanently weakened. Similarly, a carmaker that outsources production might have a structurally lower cyclical range than its historical data suggests.

Always check: has the business model, market structure, or competitive position shifted materially? If so, weight recent cycles more heavily, or rebuild your normalized estimate from first principles.

Normalized Earnings vs. Forecasted Earnings

Don’t confuse normalized earnings with analyst earnings forecasts. Forecasts predict next year’s results; normalization estimates what the company earns at a steady-state cycle. Both are useful:

• Use forecast earnings if you’re betting that the next few years will deviate from the cycle (e.g., a new mine starts production, or tariffs hit steel demand).
• Use normalized earnings if you’re valuing long-term intrinsic value and believe the business will return to its typical cycle.

Professional analysts often use both: they forecast the next 3–5 years, then assume the company “normalizes” to mid-cycle profitability thereafter, using that normalized figure to value the “terminal” or “perpetual” period in a DCF (Discounted Cash Flow) model.

See also