The Persistence of Growth

It is unlikely that the fastest growing companies from the past five years—think Tesla, Apple, and Under Armor which have grown sales 32x, 4x, and 4x respectively since 2010—will continue to be the fastest growing companies over the next five years.

Some high growth stocks will continue their torrent growth, but future growth rates have always been less impressive than past growth rates for the fastest growing companies.

The relationship between growth and valuation has been exploitable through history. A stock’s valuation is a collectively derived expectation for future growth. A high (expensive) price-to-sales means the market expects sales to grow a lot. A low (cheap) price-to-sales means the market expects sales to grow a lot less. We know that these expectations have often turned out to be overdone on both sides. For growth stocks, expectations have been too lofty (thus their under performance long term), and vice versa for value.

Since the 1960’s, companies trading at the most expensive price-to-sales multiples have had higher trailing sales growth, on average. In this first chart, we group all stocks into ten separate groups by their price-to-sales ratio (average price-to-sales for each group is shown with the blue line on the right scale), and then chart their average trailing growth rates (green bars, left scale). The most expensive stocks (trading at price-to-sales of 4x or so) grew sales by an average of 28% over the previous five years, and the cheapest (trading at 0.16x) grew sales by an average of 13%.

trailing growth by valuation bucket


Next, we sort by trailing sales growth instead of valuation. Now we can look at both the trailing and future growth rates for the highest and lowest growth companies.

The universe of stocks under consideration here is all U.S. companies between 1968 and 2010 (because we need 10 years of data, five trailing years and five future years) with market caps of at least $200MM. Companies are grouped into 10 deciles based on their trailing 5 years sales growth[i]. The universe and time horizon are a bit arbitrary. Results look similar using large stocks only instead of all stocks, or using a 3-year horizon instead of 5-year.

sales growth mean reversion

You can see that on average, the highest growth firms have grown earnings at an impressive 46% per year clip (or, 562% cumulatively over five years). The lowest growth stocks have smaller earnings than five years earlier—their sales fell by -5% per year, -22% cumulatively.

But the forward looking growth numbers look much different. The previously high growth stocks still grow fast, at a rate of roughly 18% per year, but their total growth is much smaller than during the prior five years. The low growth stocks rebound, growing at an average of 12% after their five years of declining sales.

Of course, this is a long term average which various considerably. Here is a time series of the trailing and future 5-year sales growth difference (high growth minus low growth) through time.

time series gaps

Both the long term average and time series charts make this very clear: the growth rate advantage of high growth stocks from the past five years does not persist over the next five years.  Growth stocks tend to keep growing, but their relative advantage over low growth narrows considerably. If and when the market prices stocks by extrapolating past trends into the future, there is an opportunity to bet against that misguided extrapolation (long value, short growth).

In markets, fundamentals level out.


Data Stuff

There are a lot potential biases when building tests like this using historical data, so let me be clear about what I required or didn’t for a firm to be in the sample:

  1. Have a sales number on the starting date
  2. I did not require a sales number five years or ten years in the future, so as not to exclude companies that fall out of the data. I effectively set the future sales for such companies to 0 and assume -99% future sales growth for these companies. Not necessarily the perfect handling, but I think better than excluding them altogether.
  3. Trailing and future growth numbers > 100x were eliminated (that is, if a firm’s sales grew by more than 100 times over five years, I kicked them out).
  4. This is all done with sales—instead of earnings or cash flows—because they are always positive and there are weird issues calculating growth rates when numbers flip from negative to positive.


[i] Simply total sales today / total sales five years ago