Decomposing a Return Into Earnings, Multiple, and Dividend

The Three Components

Any stock return over any period of time is divided into three parts. Earnings growth: the company made more money. Multiple swing: investors decided to pay a different price for each dollar of that profit. Dividend yield: cash paid out along the way. That's the full list nothing else is hiding there

Total return is roughly earnings growth plus change multiple plus dividend yield. It's an approximation rather than a perfectly exact identity in that simple additive form but it's close enough to be really useful and you can run the entire exercise on the back of an envelope in less than a minute

Why the Split Matters

Here's why it's important to bother splitting up a return: Each of the three pieces answers the question "can it happen again?" completely differently

Earnings growth is long-lasting. If a company continues to grow earnings growth can continue to appear in the results year after year without a natural limit

Multiple expansion doesn't behave that way. It reflects a change in sentiment or in the discount rate the market applies to future earnings and it can't last forever. A stock that rerated from 15 times earnings to 30 can't do the same trick again by going up to 60 unless something genuinely new justifies it. This piece is self-limiting by construction

The Exact Version, and Where a Cross Term Sneaks In

The additive version is a shortcut. The actual ratio is multiplied. If earnings grow by a factor of one plus g and the multiple changes by a factor of one plus m the price performance is equal to one plus g multiplied by one plus m minus one. Expand that product and you get g plus m plus an extra piece: g times m

When g and m are small that extra piece is rounded to zero. Gains increase by 5 percent multiples increase by 4 percent: the cross term is 0.05 times 0.04 or 0.002 two-tenths of a percentage point. That's why the simple additive shortcut works well most of the time and why no one bothers to teach the multiplicative form first

The cross term stops being a rounding error when both pieces are large in the same period and that happens more often than people expect usually in exactly the ranges that are worth analyzing carefully. Suppose earnings grow 40 percent and the multiple expands 50 percent over the same period. The cross term is 0.40 times 0.50 or 0.20 a full 20 percentage points. IfIf you ignore it you will underestimate the price return by a fifth. Dividends bring their own small correction on top of all this but that correction is usually small compared to the earnings multiplied by the multiple term and I treat it as noise unless the dividend yield itself is unusually high

A Full Multi-Year Decomposition, Worked Out

Let's build one from scratch with a hypothetical company I'll call Vantage Corp so no one mistakes it for a real symbol. Start with earnings per share of $5 and a price-to-earnings multiple of 20 which gives an initial share price of $100. Let's assume Vantage pays a constant dividend worth 1.5 percent of the initial price each year a clear assumption not a figure taken from a presentation.real

In five years let's assume earnings growth and multiple moves move like this. I chose numbers that rotate on purpose because a smooth path would hide the exact interaction I want to show

First year: Earnings grew 12 percent multiple fell 3 percent. Cross term: 0.12 times minus 0.03 or minus 0.36 percentage points small enough to round off. Price return: 12 minus 3 minus 0.36 which is 8.64 percent. Add the 1.5 percent dividend and the total return for the year is10.14 percent

The second year is where the cross term starts to matter. Earnings grew 8 percent and the multiple jumped 15 percent. Cross term: 0.08 times 0.15 or 1.20 percentage points. Price return: 8 plus 15 plus 1.20 which is 24.20 percent and the total return with the dividend is 25.70 percent

The third year is moving in the same direction. Earnings grew by 10 percent and the multiple appreciated another 20 percent. Cross term: 0.10 times 0.20 or 2.00 points. Price return: 10 plus 20 plus 2 or 32 percent total return 33.50 percent

The fourth year is the reversal. Earnings still grew 9 percent but the multiple plummeted 25 percent the kind of move that a rate shock or growth scare can produce in a single year. Cross term: 0.09 times minus 0.25 or minus 2.25 points. Price return: 9 minus 25 minus 2.25 equal to 18.25 percentnegative percent. Even with the dividend the year loses 16.75 percent

The fifth year calms down. Earnings grew 11 percent and the multiple recovered a modest 5 percent. Cross term: 0.11 times 0.05 or 0.55 points. Price return: 11 plus 5 plus 0.55 or 16.55 percent total return 18.05 percent

Now complicate the five years instead of adding them together because that's the part people skip. Multiplying the five annual price return factors 1.0864 times 1.242 times 1.32 times 0.8175 times 1.1655 gives you a cumulative price return of about 69.70 percent. Combining the five total return factors which include the dividend gives you about81.64 percent in the five years

It's worth doing one more check because it's the part that really separates someone who understands the identity from someone who recites it. The cumulative earnings growth over the five years from an EPS of 5.00 to 8.05 comes to about 60.98 percent. The cumulative multiple change from 20 to just over 21 is about 5.41 percent. Add those two together and you get 66.39 percent not the return onprice of 69.70 percent calculated above. The gap of 3.31 percentage points is the cumulative effect between terms. It is not the same as adding the five annual cross terms listed above negative 0.36 1.20 2.00 negative 2.25 and 0.55 which add up to only 1.14 points. The identity is exact within each year. It is not chained by a simple addition over the yearsonly by combining and that distinction is exactly where many people including myself the first time I tried this quietly get the arithmetic wrong

Back to the Original Two Scenarios

Go back to the simplest version of this exercise the one where this whole idea began before cross-terms came into the picture. Take a stock that returns 60 percent in three years. If earnings grew 15 percent a year and the multiple remained stable the return came almost entirely from the business and there is no obvious reason why the same performance couldn't be repeated if growth continues

If instead earnings remained stable and the multiple ranged between 12 and 19 the total return came from investors repricing the exact same earnings. Repeating that trick requires the multiple to then reach 30 which requires an increasingly optimistic story

Both actions print 60 percent on one screen. They describe two completely different situations and the screen can't tell you which one you're looking at

Notice something about both scenarios: In each only one component is doing anything. In the first the multiple is flat so it contributes exactly zero when multiplied by earnings growth. In the second earnings are flat so the movement of the multiple is multiplied by zero. Neither scenario ever touches the cross term. That's a big part of why they look so clean side by side. Actual periods are rarely so considered which is whywhich Vantage Corp uses numbers that really shock

What Drives Multiple Changes

Multiples move for several reasons and are worth the effort to separate them. Interest rates matter mechanically: a lower discount rate increases the present value of the same future earnings stream justifying a higher multiple with zero change within the business itself

Expected growth also matters. A company whose growth prospects improve deserves to trade at a higher multiple than before for the same current earnings. Perceived risk matters on top of that as a business that the market now considers more durable earns a higher multiple for identical current earnings

Then there's sentiment the component that is neither rational nor stable. Distinguishing a type-based revaluation from a sentiment-based revaluation is really difficult and honestly that's where most of the analytical value of this whole exercise lies

Applying It to a Period

The exercise becomes especially illuminating over a full cycle. A lot of the market's strength during the low-rate years came from multiple expansion driven by falling discount rates rather than anything companies were doing differently. When rates rose sharply that component reversed which is why 2022 was painful even for companies whose earnings held up well

Periods where performance is due to earnings growth rather than appreciation are structurally longer lasting and that is exactly why analysts spend so much energy checking whether a market rally is accompanied by actual earnings growth rather than taking the index level at face value

Case Study: Nvidia's Two Very Different Eras

For a real-world illustration of a great crossover term I keep coming back to Nvidia mostly because I watched both halves as an outside observer not because I traded any of it

In the years leading up to the rise of generative AI roughly the second half of the 2010s Nvidia's business was already growing rapidly. Gaming demand was strong and a data center segment was building beneath it as cloud providers began buying GPUs for machine learning workloads.perform well because the profits themselves were doing real work

Then came the period around 2023 after the launch of big language model products made the appetite for AI training hardware sudden and visibly huge. Nvidia's data center revenue increased in a way that is well documented and doesn't need a precise figure from me to make it clear.The beginning of something much bigger and more lasting than a good year. Earnings grew a lot and the multiple expanded a lot in the same period. That is precisely the condition under which the crossover term stops being a footnote and begins to be a significant part of the performance

I deliberately don't attach exact percentages to Nvidia's move here. The specific figures become stale quickly and the number isn't really the point. The direction is the point and the direction is not in dispute: earnings grew very quickly and the market simultaneously chose to pay more for each dollar of those earnings. Anyone who treats that performance as pure earnings growth or pure multiple expansion ignores that it was really both of those things interacting

The Trap: Reading Mean Reversion Into an Accounting Identity

This is where I think this whole framework is constantly misused and I include my first attempts in that criticism

The decomposition is an accounting identity. By itself it says nothing about what will happen next. But the instant someone sees that "this performance was primarily a multiple expansion" the temptation is to jump straight to "so the multiple has to go back down and the performance will reverse".instead of a free use of arithmetic that is true by construction

Sometimes the claim of mean reversion is correct. Sentiment-driven reratings actually fade because sentiment isn't anchored to anything durable. But sometimes the old multiple was simply too low for what the business had become and the rerating is a one-time correction rather than a loan against the future

Microsoft under Satya Nadella is the case I find most compelling. Starting in the mid-2010s the company shifted its center of gravity from one-time perpetual software licenses toward recurring cloud and subscription revenue including Azure and Office 365. Investors rerated the stock significantly higher during that period and within the framework of this article a real portion of Microsoft's performance in that period reads as multiple expansion. If the reversal instinct toIf the average was the whole story that higher multiple should have faded over time. It didn't. It held up roughly for the better part of a decade because underlying earnings had become genuinely more predictable and higher-margin and the market's willingness to pay more for that predictability was no fad. It was a good update

The lesson I learn from comparing Nvidia's run to Microsoft's lasting rerating is that identity can't tell you on its own what kind of multiple shift you're looking at. That judgment lives outside of arithmetic in the particulars of the business

How I Actually Use This When I Look at a Stock

None of this is investment advice and I'm not telling anyone what to buy. This is the process I actually go through in my head and I think it's useful primarily as a way to ask better questions about a return I'm already looking at

First I separate the return into three parts plus the cross term if both parts seem large. I do this before I give myself an opinion on whether the stock is a good business because otherwise I will unconsciously adjust the story to the conclusion I already wanted

Second I ask which piece did the work. If it's earnings growth I ask whether the conditions that produced that growth are still in place. If it was multiple expansion I ask the tougher question from the previous section: Is it a swing in sentiment that's plausibly fading or a genuine revaluation of a business that's changed below the stock? I don't always know the answer. Sometimes I really find it hard to say and I'd rather say it clearly than feign confidence that I don't have

Third I check to see if both pieces were large at the same time. If they were I know that simple additive history whatever headline return is cited is underestimating how much of the movement came from the interaction rather than from any of the pieces separately

The way I would actually use this in a room defending a thesis is to start with the decomposition not the conclusion. Showing the division shows that I looked under the headline number. What I do with that information afterwards whether I believe the multiple holds or fades is a decision I would explicitly mark as my own reading rather than a fact that the identity handed me

The Bottom Line

Divide any performance into earnings growth multiple variance and dividends before judging it and remember that the first two components interact: When both earnings growth and multiple variance are large in the same period their cross term can be a significant share of the total not a rounding error. Earnings growth can be repeated.from Microsoft.Confusing a revaluation that will fade with one that will remain is how people extrapolate a revaluation to a forecast that the identity never supported