Why Are Some Companies So Expensive?
As a value investor, I often puzzle over why other investors are willing to pay such high prices for companies. Our own portfolio trades at around 6x EBITDA, so you can imagine my consternation when I see that Visa, for example, trades for 17x revenue.
Count me doubly confused when I look at the track records of the growth investors who have been paying these nosebleed prices—phenomenal! Turns out buying the most popular, highest-quality companies no matter the price has been a winning strategy the last few years, particularly when contrasted with the performance of us sober skeptics in the value investing community.
So I set out to answer a simple question: why are some companies so expensive?
This being Verdad, I started by running a few regressions. I looked for variables that I thought would explain why a company was expensive or cheap. From there I honed the list down until I had a concise set of predictive variables that explained a big proportion of the variation in multiples among S&P 500 constituents.
The list of variables I found most predictive were net income margin, dividends as a percentage of sales, two-year forward analyst estimates of revenue growth, and the average return on assets over the trailing three years. These variables capture the full array of fundamentals: margins, growth rates, return on investment, and capital allocation. And together they explain about 44% of the variance in EV/Sales multiples (which is quite high for explaining any market phenomena).
Figure 1: Regression Coefficients for Explaining EV/Sales
Source: Capital IQ, Verdad Analysis.
I then used these variables to develop a simple model that predicted a company’s EV/Sales based on the input variables. Below you can see a scatter plot of the largest 100 companies in the S&P 500 comparing the predicted EV/Sales to the actual EV/Sales.
Figure 2: Predicted vs. Actual EV/Sales
Source: Capital IQ, Verdad research.
We can think of this graph as showing what a purely quant model thinks a company is worth as opposed to what the market thinks it’s worth. The delta in between is explained by perceptions about the quality of the business that are not showing up in the financials. And we can see where the greatest incongruence is by comparing the predicted multiple to the actual multiple.
First, here are some of the companies that score as the most undervalued relative to a quantitative prediction. I’ve shown companies that trade at low multiples that the model thinks should be higher as well as companies that trade at high multiples that the model thinks should be even higher. Note, we haven’t done any research on these companies beyond running this quantitative model.
Figure 3: Undervalued Companies
Source: Capital IQ, Verdad research.
We next looked at stocks, trading at both high and low multiples, where the model predicted a much lower trading multiple than reality.
Figure 4: Overvalued Companies
Source: Capital IQ, Verdad research.
These gaps in valuation are driven by narratives that transcend the numbers. The reasons investors really love MarketAxess, Verisign, and Visa—and why they hate Capital One, Micron, and Marathon Petroleum—cannot be explained purely by a quant model.
But what would cause these companies to re-rate? We looked historically at what explained changes in valuation by regressing the change in each driver variable against the change in observed EV/Sales multiple. Only one variable was statistically significant: two-year forward revenue estimate.
Figure 5: Drivers of Changes in EV/Sales
Source: Capital IQ, Verdad research.
In other words, re-rating of multiples is primarily driven by changes in forecasts of future sales growth. How Microsoft’s multiple in 2020 changes relative to today will be dictated by what analysts in 2020 think Microsoft’s revenues will be in 2022. Historically, during expansions, revenue growth estimates have had a 60–70% correlation year over year, which is significantly less stable than return on assets (70–80% correlated year over year) or dividend yield (80–90% correlated year over year).
In sum, companies with high dividend yields, high return on assets, high margins, and high expected revenue growth trade at high multiples, and the most frequent reason for a company’s multiple to change significantly is changes in revenue growth expectations. To beat the market buying high multiple stocks, an investor needs to accurately predict what analysts will expect forward revenue growth to be in the future - a difficult if not impossible task. We believe that given the uncertainty embedded in growth projections and the volatility of margins, investors should weight those factors less strongly than the market does—a conclusion that leads ineluctably to the logic of value investing.