Quant Investing


What would you think of a quant strategy that only invests in the most profitable companies? Would it under perform the market or beat the market? If you’re an efficient market person you may think that higher profitability must be priced into equities and therefore at best the strategy would match the market. Not so. Turns out that profitability is quite a durable factor and is only beaten by momentum and value. In this post I’ll take a look at some of the data on the profitability factor and how it can be applied in a simple quant strategy.

First, let’s look at profitability as a factor. Whenever you look at factors you’re going to run into all kinds of varying opinions on what they are, which ones are real or not, which ones are sustainable, etc. Profitability is no different. See here for a good discussion on profitability. There are 3, 4, 5, and 6 factor models now. The discussion is not just about value, size, and momentum any more.  My favorite take on factors is the AQR factor model . Here is the summary of their factor model.

As the table shows (higher numbers in the intercept column), profitability (RMW) is the 3rd strongest factor, after momentum (UMD) and value (HML-DEV). Now let’s get to the fun part using it in a quant screen.

Let’s build a real simple screen using profitability. Here I use profitability defined as gross profits over assets (from the Novy-Marx 2012 paper . Also, see here for an alternative measure). Since I’m using a quarterly measure of profitability I’m going to re-balance the screen every 3 months. For the universe I’ll use the SP500 stocks to keeps things even more clear. The screen invests in the top 25 Sp500 stocks ranked by quarterly profitability. Below are the results of the backtest from 1999 through today.

Almost double the performance of the total SP500 with less drawdowns. Not bad. Higher profit companies outperform lower profit companies. Now, what about combining profitability with other factors? Turns out that works quite well. We’ll take our Value Composite 2 screen and apply it to the SP500 universe and further screen out the least profitable companies (companies with profitability ranking below 50) and invest in the top 25 companies rebalanced every 3 months as above. Below are the results.

These returns are about 2% higher than just using the VC2 composite on the SP500 universe. Profitability enhanced the performance of the value factor. It can also work similarly with momentum.

In summary, profitability is a strong factor that can be used to enhance quant system performance. It also works in combination with other factors such as value and momentum. It’s worthy consideration to your quant investment toolkit.

Note: all screens and portfolios are simulated in Portfolio123



Tagged , , , , ,

11 thoughts on “ Quant investing: the profitability factor

  1. Paul — when are you going to be bought out by some investment firm? You are smarter at communicating reality than most random tv personalities. Knew a person at one time who was going to use ARTIFICIAL INTELLIGENCE software to control investment. Bet your strategy discussion is as good as that.

    1. Thanks Kevin. I’ll gladly entertain offers…

      Robot/AI/Auto-trading software is already quite common. I can auto-trade in P123. Humans, fortunately, still design the algorithms, but that is changing too.


  2. Thank you for sharing! If I may, do you see the same correlation if you go further back than 1999? Not trying to be a party pooper, but I’m curious as to see if you do a rolling 30 year dating back to the 1930’s, if you’d see a similar pattern on the total returns of the S&P 500 / Wilshire 5000 / DJIA, vs the companies that fall under your metric of profitable.

  3. Hi Paul,

    Thanks as always for posting great info. Very surprised that profitability is nearly as strong a factor as value.

    I know you’ve discussed a few other things that can enhance quant portfolios, such as a debt-change filter. I am assuming that if we combine all of these changes (e.g. debt-change in conjunction with profitability filter, etc.) there would be some overlap and the out-performance wouldn’t be cumulative. Still I would love to see a post where you go over the various enhancements you have found and see how they work together across a variety of quant strategies.

  4. Thanks for the post. Do you include transaction costs in your backtest? Also, what is the turnover?

    1. No transaction costs. Turnover average is about 25% for the pure profitability screen in the first test, and 30% for the value screen with the profitability filter.

Comments are closed.