Normalized Earnings and Cyclical Adjustments: Valuing Through-The-Cycle Performance

If you value a homebuilder at the top of a housing boom or an oil driller at a hundred dollar crude, you're going to lose money.

Cyclical companies present one of the most seductive traps in valuation work. The math looks incredible at exactly the wrong time and terrible at exactly the right time. A homebuilder trading at four times earnings during a construction boom isn't cheap. It's expensive. The steel manufacturer at twenty-five times earnings during a trough isn't broken. It might be a gift.

The problem is that reported earnings in cyclical businesses reflect a moment in time within a cycle that will inevitably turn. If you anchor your valuation to peak earnings, you're capitalizing an unsustainable profit level. You're building a discounted cash flow model on quicksand. If you anchor to trough earnings, you're assuming permanent impairment in a business that might be structurally sound but temporarily depressed.

Normalized earnings solve this. The concept is straightforward but the execution requires judgment. You're attempting to estimate what a company earns through a full economic cycle, averaging the peaks and troughs to arrive at a sustainable baseline. This becomes your foundation for valuation. You're not predicting next quarter. You're not even predicting next year. You're asking what this business can reasonably earn in a mid-cycle environment when conditions are neither euphoric nor disastrous.

Start with revenue normalization. Look back at least ten years, preferably fifteen if the data exists. You want to capture a full cycle, which for most industries runs seven to ten years. Take a company like Caterpillar. Construction and mining equipment sales swing wildly with infrastructure spending, commodity prices, and global growth rates. In 2012, Caterpillar did seventy-two billion in revenue. By 2016, it was down to thirty-eight billion. By 2022, it was back over sixty billion. If you valued Caterpillar in 2012 using that year's revenue as a baseline, you would have dramatically overpaid.

The method here is to calculate the average revenue over the full period, but not a simple arithmetic mean. You need to adjust for inflation and for structural changes in the business. If the company acquired or divested significant operations, you need to normalize for that. If the entire industry grew or shrank due to technological shifts rather than cyclical factors, that matters. A coal company's declining revenue isn't cyclical. It's structural. A copper miner's swings are cyclical.

Once you have normalized revenue, move to margins. This is where the real work happens. Cyclical companies see margin expansion at peaks because fixed costs get leveraged over higher volumes and because pricing power temporarily increases. They see margin compression at troughs for the inverse reasons. A chemical manufacturer might run at fifteen percent operating margins at cycle peaks, eight percent at troughs, and something like eleven percent on average.

Pull the operating margin for each year in your lookback period. Calculate the median, not the mean. The median filters out extreme outliers that might distort the picture. If the company had one catastrophic year with negative margins due to a one-time event, the median handles that better than the mean. Now you have a normalized operating margin.

Apply that normalized margin to your normalized revenue. This gives you normalized operating income. From there, you work down the income statement using the current tax rate and current interest expense unless those are also cyclical, which they can be if the company levers up and down with the cycle. Adjust accordingly. The output is normalized earnings, either at the operating income level or net income level depending on where you want to enter your valuation.

Let's use a real example. Take Ford in 2021. The company earned eighteen billion in net income, driven by vehicle shortages, pricing power, and inventory normalization after pandemic disruptions. The stock traded up to twenty-five dollars. Someone valuing Ford on 2021 earnings at a ten multiple would conclude the equity was worth a hundred and eighty billion. Seemed reasonable in the moment. By 2023, Ford earned four billion. The business didn't break. The cycle turned. Chip shortages eased, inventory returned, pricing power evaporated, and competition intensified. If you had normalized Ford's earnings using the prior fifteen years, averaging periods of strong profitability and periods like 2008 where the company nearly collapsed, you would have arrived at something closer to six or seven billion in sustainable earnings. That normalization would have suggested Ford was trading at fifteen or sixteen times normalized earnings in 2021, not ten times. Still not outrageous, but far less compelling.

The same logic applies in reverse. Take US Steel in 2020. The stock traded below six dollars. The company was losing money. Steel prices had collapsed. Demand was weak. Someone looking at current earnings would conclude the business was uninvestable. But looking at normalized margins from prior cycles, you could see that US Steel historically earned mid-single-digit operating margins through the cycle. When steel prices recovered, as they cyclically do, margins would recover. By 2021, steel prices spiked, the company earned billions, and the stock ran to thirty dollars. The opportunity was visible in 2020 if you normalized instead of extrapolating the trough.

Margin mean reversion is the underlying principle. Margins in cyclical industries revert to historical averages over time. High margins attract competition, increase supply, and erode pricing. Low margins force capacity closures, reduce supply, and stabilize pricing. The cycle feeds itself. Your job as an analyst is to recognize where in the cycle a company sits and adjust expectations accordingly.

One refinement is to segment normalized earnings by division if the company operates across multiple end markets with different cycle timings. A conglomerate like 3M has businesses exposed to industrial cycles, consumer cycles, healthcare, and electronics. Each moves differently. Normalizing at the consolidated level can obscure what's happening underneath. Break it out. Normalize each segment. Roll it back up.

Another refinement is to stress test your normalization. Don't just take the fifteen-year average and call it done. Ask whether structural factors have permanently shifted the cycle. The auto industry today has different economics than the auto industry of 2005. Electric vehicle transitions, software content, and battery supply chains are reshaping margins and capital intensity. A straight historical average might overstate or understate what's achievable going forward. Adjust your normalization for structural shifts while filtering out cyclical noise.

The error most people make is recency bias. Whatever happened in the last twelve months feels permanent. The homebuilder that just printed record earnings will keep printing record earnings. The airline that just lost two billion will keep losing two billion. Neither is true. Cycles turn. They always turn. The companies that look the best on a trailing basis are often the most dangerous to buy. The companies that look the worst are often the most interesting.

This doesn't mean you blindly buy every beaten-down cyclical or short every elevated one. Normalized earnings tell you what the business can earn, not what the cycle will do next or when it will turn. Timing the cycle is a different skill. But valuation discipline requires that you don't overpay for peak earnings or dismiss a business because it's in a trough. Normalize first. Value second. Then decide whether the cycle timing and the current price create opportunity.

See you Friday. When everyone else is extrapolating the present, you normalize through the cycle.