The Best Value Stock Screening Ratios – As Shown By Machine Learning
A data-driven and value-focused investment fund publishes some fascinating research in its quarterly letters that we can use to find the best value stock screening ratios.
I’ve mentioned Euclidean Technologies before. It is a machine learning-driven value fund that was founded in 2008. I’m going to talk about two analyses they’ve published, which have helped me develop a new stock screener that’s working out well so far.
The “Magic Formula”: Good Companies or Good Prices?
For those that aren’t familiar, Joel Greenblatt wrote a book called “The Little Book That Beats the Market.” This presented his “Magic Formula” for buying good companies at good prices. He focused on two key ratios:
- Earnings Yield = EBIT / Enterprise Value (where Enterprise Value is Market Value + Debt – Cash)
- Return on Invested Capital (ROIC) = EBIT / Invested Capital (where Invested Capital is Net Working Capital + Net Fixed Assets)
Earnings yield basically tells you how cheap a company is compared to its ability to generate cash for its owners. All else equal, you would want companies with higher earnings yields.
ROIC measures a company’s ability to generate cash for its owners relative to how much capital has been invested in the business. It’s a great indicator of how efficiently a company deploys capital. Again, the higher, the better.
Greenblatt basically tells you to get these ratios for all companies and weight them equally when screening and selecting your investments.
Adding Nuance to Magic Formula Investing
In their Q2 2013 letter, Euclidean asked two questions:
- Should these two ratios really be weighted equally?
- Are there environments where his approach tends to do well or poorly?
Euclidean examined 10 different weightings of the Good Price (using Earnings Yield) and Good Company (using ROIC) ratios over the timeframe from 1973-2012 — much longer than Greenblatt used in his book.
A few nitty-gritty details: they limited their universe to US stocks with market capitalizations > $400M, leaving about 1200 stocks, whereas Greenblatt’s book had a universe of around 3500 stocks to choose from. They also made a small adjustment to EBIT: they added back depreciation and subtracted CapEx, which their research has found to more accurately assesses results. This makes it a bit more like Free Cash Flow, without the changes to working capital. (This ratio then is more like Cash Return on Invested Capital, or CROIC.) They also took the 4-year average of it, rather than a single point-in-time.
They constructed 10 different portfolios with different weightings of the two ratios, each with the top 50 holdings based on the weighted score.
Earnings Yield and ROIC should not be weighted equally
When they ran their simulations, they found that the best performing portfolio was the one that “focused solely on buying companies at very low prices,” i.e., the “100% Earnings Yield” portfolio.
Source: Euclidean Technologies
The 100% Earnings Yield portfolio had the highest compound annual growth, and any portfolio built using ROIC — even just a little — underperformed.
This in and of itself is a fascinating result. But then they looked at the second question, whether these portfolios perform differently during different types of environments.
Defining Optimistic and Pessimistic Environments with the CAPE Ratio
They used Rober Shiller’s Cyclically Adjusted Price to Earnings (CAPE) ratio (or Shiller P/E) to look at periods where markets were “pessimistic” (decreasing CAPE), “neutral”, and “optimistic” (increasing CAPE).
For those that aren’t familiar with the CAPE ratio, it’s just the price of the S&P500 in relation to its 10-year trailing inflation-adjusted earnings of the index’s component companies. It smooths out fluctuations in earnings due to the business cycle and helps indicate whether a broad market or index is over- or under-valued relative to its historic average. Here’s what it looks like today:
You can see that 2001 was a major red flag, and the high CAPE accurately foretold a period of poor returns. But then, during the 2008 crisis, the ratio dropped to pretty solid bargain levels (or at least back to being in-line with its historic average). The market is now off its high from last year, but is still well below its long-run average.
Environments where Earnings Yield and ROIC Outperform
In any case, Euclidean created these increasing/decreasing/neutral CAPE time period buckets, then looked at the 10 portfolios’ performances. They found that during times of increasing P/Es, the 100% ROIC portfolio tended to do best (though not exclusively). In times of neutral or declining P/Es, the 100% Earnings Yield (Good Price) portfolio did the best.
However, they noted that all portfolios tended to do pretty well in the optimistic time periods.
However, in pessimistic periods when price-to-earnings multiples compress, there is an overwhelming advantage to being invested in the least-expensive companies (e.g., those with the highest earnings yields).
Since it’s impossible to predict whether we’re moving into an optimistic or pessimistic period, it seems logical to be highly focused on value all the time.
In terms of my stock screening, this gave me the freedom to not rely as heavily on ROIC measures, and to focus on measures of Value, like Earnings Yield.
A Value Ratio Comparison
The second piece of research that I want to highlight, which I think is even better and more robust, is from their Q2 2017 letter.
Growth vs. Value Stock Performance
As a little bit of an aside, the first thing they highlight in this piece is a reminder of the times when growth has outperformed value in the past:
As we all know, Value outperforms Growth on average, but there are periods when it does not. Like the one we’ve been in since 2010. It still seems very likely to me that we will have the value rebound similar to the ones we’ve seen in the past.
The chart above uses book-to-market to categorize companies. However, despite its wide use in academia and in practical investing, they assert that it book value doesn’t tell you much about of use about a company’s intrinsic value or its ability to generate future cash flows.
Book value makes no distinction between a pile of cash and a company with productive assets, great products, and loyal customers. Many investors, therefore, look at other fundamental qualities to better approximate a company’s future ability to distribute cash to its owners.
Statistical Analysis to Identify Better Value Ratios
So, they decided to analyze a variety of valuation metrics to see which worked best for screening and ranking investment opportunities. Their time period was 1970-2017, and they used the same universe as above (NYSE, NASDAQ, and AMEX stocks with market caps over $400M in 2010 dollars). They invested a hypothetical $100M fund in the 10% cheapest or most expensive stocks as ranked by each metric, holding each position for a year. Here’s what they found:
Owning shares of inexpensive companies, regardless of definition, has been far more fruitful than owning expensive companies over the long term. Indeed, a good route to realizing above-average returns appears to have been adhering to a process — even a very simple one — for buying companies at low prices in relation to their sales, book values, or earnings.
Owning shares of the most expensive companies never paid off overall (though there may have been periods where it did, similar to above).
Book-to-market actually appears to be the least useful standalone valuation metric.
EBIT, EBITDA, and Gross Profit to Enterprise Value showed the best compound annual returns. All of the ratios using Enterprise Value improved upon the ratios that used Market Cap as the denominator. This, again, makes intuitive sense:
To understand whether a company’s shares are mispriced from the perspective of a whole owner, it is crucial to understand first what it would take to own the company outright.
The Best Value Stock Screening Ratios
You may have noticed something about one of the best-performing valuation metrics, EBIT/EV. Well, that’s just Earnings Yield, like we discussed above!
Back in the fall, I constructed an OSV screener that used:
- EV to EBITDA (lower is better, but greater than 0)
- EV to EBIT (or use the Magic Formula Earnings Yield – higher is better)
- Industry – filtering out financials, etc.
- Market cap – keeping my universe to $100M+
- Cash Return on Invested Capital (CROIC) – filtering out anything less than 5% (but not really choosing stocks based on this)
- Piotroski >=5
My primary sort was on EV/EBITDA, TTM, but I also look at which ones also had the highest earnings yield.
This yielded a lot of good ideas that were worth digging more into using the OSV Valuation tools. (I definitely didn’t just buy the top 50 cheapest or anything like that.)
Though it’s too early to tell whether some of my picks since then were good, I’m optimistic!