Close

“What Works”: Key New Findings on Stock Selection

by James O'Shaughnessy

Hear James O'Shaughnessy's Keynote Opening speech at the AAII Investor Conference!

When I began the research for what would eventually become the first edition of “What Works on Wall Street” in 1995, I sought to identify which individual factors delivered the best alpha (risk-adjusted performance) over time and did so with the greatest consistency (base rates).

What I have found is that there is no “best” factor, per se; rather, the strongest individual factors come in and out of favor. The price-to-sales ratio and EBITDA-to-enterprise-value ratio vie for top billing, but it depends on the time period under review. (EBITDA is earnings before interest, taxes, depreciation and amortization. Enterprise value is a combination of a company’s market capitalization, debt, minority interest and preferred stock at market value, less any investments in associated companies at market value and all cash and cash equivalents.)

Figure 1 illustrates the changing nature of individual factors over rolling 10-year periods. Look at how the free-cash-flow-to-enterprise-value ratio briefly added 4% more value than the value composite, but then took a nose dive.

This new research led to what can be considered the key new findings in the fourth edition of “What Works on Wall Street” (McGraw Hill, 2012): You get much better and more consistent results by using several factors together in a composite than you get using a single factor alone. Here I discuss our value composite first, then move on to others that look for financial strength and earnings quality. What’s more, by combining these three factors you greatly improve your chances of finding high quality, financially strong companies that offer excellent value. Let’s start with a look at O’Shaughnessy Asset Management’s (OSAM) value composite.

Assessing Whether a Stock Is a Value or Expensive

Sometimes a stock with a low price-earnings ratio is not a true value stock. The way to determine that is to rank the stock on other key value variables. As you can see in Figure 1, a composite made up of five individual factors—price to sales, price to earnings, EBITDA to enterprise value, free cash flow to enterprise value and shareholder yield—outperforms its individual components in 85% of all rolling 10-year periods between 1972 and 2012.

We create the composite in this manner: For each combined group of factors, we assign a percentile ranking (from 1 to 100) for each stock in two universes: All Stocks and Large Stocks. If a stock has a price-earnings ratio that is in the lowest one percent for the universe, it will receive a rank of 1; if a stock has a price-earnings ratio in the highest one percent for the universe it will receive a rank of 100. We follow a similar convention for each of the factors—thus, if a stock is in the lowest one percent of the universe based on its price-to-sales ratio, it gets 1; if the stock is in the highest one percent, it gets a 100. If a value is missing for a factor, we ignore the factor; however, we require at least three of the factors to have a value to be included in the value composite. For shareholder yield—which is dividend yield plus buyback yield—those stocks in the one percent of the universe with the highest yields will be ranked 1, whereas those within the lowest one percent will be ranked 100. Once we’ve assigned a rank to all the factors, we add up their rankings and assign the stocks to deciles. Those with the lowest scores are assigned to decile one, while those with the highest scores are assigned to decile 10.

Source: OSAM Research. Past performance is no guarantee of future results. Complete disclosures can be found at www.aaii.com/files/journal/OSAMdisclosures.pdf.

Thus, the stocks in decile one would feature the best combined score and would have the lowest price-earnings ratios, price-to-sales ratios, etc., whereas the stocks in decile 10 would have the highest price-earnings ratios, price-to-sales ratios, etc. Figure 2 shows the results. From January 1, 1963, through December 31, 2012, the stocks in the best decile of the value composite earned an average annual compound return of 17.3%, approximately 6.2% better than the All Stocks universe, which earned 11.1% over the same period.

You can see in Figure 2 that as a stock becomes increasingly expensive, it stops providing excess returns—indeed, the high valuation actually detracts from the return you would earn by simply investing in the All Stocks universe. The worst decile by value subtracts 9.6% annually from the All Stocks return, returning a miserable 1.5% annually over the period. To put into perspective how badly the stocks in the worst decile performed, consider that an investment in U.S. Treasury bills earned 5.18% annually over the same period! What’s more, if you look at how consistently the best decile by value beat the All Stocks universe—again featured at the bottom of Figure 2—you will see that it beat the All Stocks universe in 99% of all rolling five-year periods and 100% of all rolling 10-year periods. Conversely, the stocks in the worst decile for value beat the All Stocks universe in just 5% of all rolling five-year periods and never in any rolling 10-year period.

SPECIAL OFFER: Get AAII membership FREE for 30 days!

Value Traps and Actual Bargains

Let’s look at two examples where using the value composite helps avoid value traps and two examples where seemingly pricey stocks were actually well valued. We’ll use the S&P 500 index as our starting universe.

Source: OSAM Research. Past performance is no guarantee of future results. Sharpe ratios and complete disclosures can be found at www.aaii.com/files/journal/OSAMdisclosures.pdf.

On January 31, 2010, Merck and Co. (MRK) had a price-earnings ratio of 10.02, making it the 17th cheapest stock in the S&P 500 based upon price-earnings ratio alone. Its low price-earnings ratio might have made it very appealing to a value investor, yet when looking at Merck through the lens of the value composite, it was actually more expensive than 65% of the stocks in the S&P 500 and should have been avoided. Had you bought the stock on January 31, 2010, you would have lost 35% of your money 12 months later. Over the same time period, the S&P 500 was relatively flat, losing just 42 basis points (0.42%).

Another example is Hartford Financial Services Group (HIG). On May 31, 2011, the stock had a price-earnings ratio of just 6.86, making it the sixth cheapest stock in the S&P 500 based on price-earnings ratio. Yet when measured by the value composite, it was more expensive than 56% of the names in the S&P 500. Again, this is a an indication that the stock should have been avoided. Had you invested on May 31, 2011, you would have lost 9.40% of your money 12 months later, while the S&P 500 gained 22.18%.

Now, let’s look at two stocks that appeared to be grossly overvalued based upon their price-earnings ratio, but were in reality very attractively priced based on their value composite score. On February 28, 2001, Office Depot (ODP) had a price-earnings ratio of 57.44, making it more expensive than all but 94 other stocks in the index. Yet its value composite score revealed it to be cheaper than 74% of the other stocks in the index, making it an attractive stock to a value investor. Had you bought Office Depot then, you would have realized a gain of 106.63% over the next 12 months from owning the stock, versus a loss of 9.52% for the S&P 500.

Another example is Dillard’s Inc. (DDS). On April 30, 2004, the stock had a price-earnings ratio of 153, making it more expensive than all but 76 other stocks in the index. Yet Dillard’s ranking by the value composite found it to be cheaper than 74% of all the other stocks in the index. Had you bought the stock then, you would have a gain of 39.21% 12 months later, versus a gain of just 6.33% for the S&P 500.

Finally, as I write this on August 27, 2013, two stocks that might appear cheap based upon price-earnings ratio are actually very expensive when ranked by the value composite: M&T Bank Corp. (MTB) has a price-earnings ratio of 13.29, making it cheaper than 418 stocks in the index. Its value composite ranking, however, shows it to be more expensive than 67% of the other stocks in the index. SCANA Corp. (SCG) has a price-earnings ratio of 14.07, which is less expensive by price-earnings ratio than 400 other stocks in the index, whereas the stock’s value composite score finds that it is more expensive than 64% of the stocks in the S&P 500. Thus, the value composite on its own proves to be a very valuable tool for investors hunting for bargains in the stock market.

Financial Strength and Earnings Quality Composites

Our next two composites also add excess return on their own, but are actually better used to identify which stocks to avoid. We’ll start with the financial strength composite.

Financial Strength

The financial strength composite ranks stocks on the following four factors: external financing, debt-to-cash flow ratio, debt-to-equity ratio and one-year change in debt. The composite is created exactly like the value composite, with stocks with the best levels for each factor receiving a 1 and stocks with the worst levels for each receiving a 100. In the case of the financial strength composite, stocks missing a value for a factor are ignored; to be included in the composite, a stock must have a value for at least two of the four factors.

Figure 3 shows the results by decile for the period starting January 1, 1963, and ending December 31, 2012. If you simply bought the decile of stocks with the greatest financial strength, you would have earned an average annual compound return of 13.2% versus the annual 11.1% return earned by the All Stocks universe, whereas an investment in the worst decile of financial strength earned 5.5% annually, just slightly ahead of Treasury bills. In terms of consistency, the best decile by financial strength beat All Stocks in 89% of all rolling five-year periods and 98% of all rolling 10-year periods. Conversely, the worst decile by financial strength beat the All Stocks universe in just 5% of all rolling five-year periods and in none of the rolling 10-year periods.

In this case, two examples show how combining the financial strength composite with the value composite helps you weed out stocks that look great in the value composite, but whose score on the financial composite would have removed them from consideration. At the end of February 2005, General Motors Co. (GM) was cheaper than 84% of the stocks in the index based on the value composite and sported a dividend yield of 5.42%. Based just on the value composite and the high yield, this would have been a very attractive purchase for either a deep value or dividend investor. Yet, consulting the financial strength composite, you would have seen that GM was in the worst decile by financial strength, thus eliminating it from consideration. Twelve months later, GM had lost 39.46%, compared to a gain of 8.39% for the S&P 500.

A similar thing occurred at the end of November 2007 when Citigroup (C) was cheaper than 81% of the stocks in the value composite and had a dividend yield of 6.69%, making it very attractive on value and yield. Yet its financial strength rank of 96.20 put it in the worst decile for financial strength and eliminated it from consideration. Over the next 12 months, the stock lost 74% of its value, whereas the S&P 500 sank 39.46%. The power of the combined composites is undeniable.

Two stocks that currently face the same problem are Laboratory Corporation of America Holdings Group (LH) and Jabil Circuit (JBL). Both stocks are cheaper than 85% of the stocks in the value composite, but have financial strength ratings in the second-worst decile. This characteristic would eliminate them from a prudent value investor’s portfolio.

Earnings Quality

Finally, let’s look at our earnings quality composite. It is composed of the following four factors: ratio of current accruals to assets, change in operating assets, ratio of total accruals to total assets and ratio of depreciation to capital expenditures. [Current accruals to assets is defined as the difference in accruals earnings over the last 12 months minus the cash earnings over the last 12 months. Total accruals to total assets is calculated as the change in working capital accounts on the balance sheet (change in current assets – change in current liabilities – change in cash). Click here for more detailed explanations.] Obviously, we would expect companies with high earnings quality to perform much better than those with low quality, as low quality infers that the chief operating officer (COO) of the company is pulling every trick out of the bag in order to make earnings appear more solid than they are. That is exactly what we find.

Between January 1, 1963, and December 31, 2012, the best decile by earnings quality earned an average annual compound return of 15.7%, approximately 4.6% better than the All Stocks universe. The worst decile by earnings quality earned just 4.1% per year, making Treasury bills a better investment choice. Figure 4 shows all 10 deciles. In terms of consistency, the best decile of earnings quality beat the All Stocks universe in 95% of all rolling five-year periods and 100% of all rolling 10-year periods. The worst decile by earnings quality beat the All Stocks universe in just 5% of all rolling five-year periods and in no rolling 10-year period.

Source: OSAM Research. Past performance is no guarantee of future results. Sharpe ratios and complete disclosures can be found at www.aaii.com/files/journal/OSAMdisclosures.pdf.

As with financial strength, earnings quality is best used to help you avoid stocks that can be disastrous investments. For example, poor earnings quality scores would have kept you out of three world-class frauds: Enron, Tyco International and WorldCom. Each of these companies was in the worst decile for earnings quality, and their losses ranged from 63% and 99%. In each case, their earnings quality scores would have prevented you from buying them. Two current examples of stocks in the worst decile for earnings quality are CarMax (KMX) and Monster Beverage Corp. (MNST).

Use Composites Instead of Individual Factors

We have seen that using composites rather than individual factors can add real value to an investor’s portfolio selection process. Finding companies with excellent value, financial strength and high earnings quality leads to much better performing portfolios than those constructed using single factors or single composite factors alone. When used in combination, they allow for the best of stocks to earn a way into your portfolio and they eliminate stocks that are often positioned to do poorly.

James O'Shaughnessy is chairman, CEO and CIO of O’Shaughnessy Asset Management (OSAM) and author of “What Works on Wall Street” (4th edition, McGraw-Hill, 2012).


Discussion

Andrew Beaudoin from GA posted about 1 year ago:

Does AAII have a screen for this method?


t re from nj posted about 1 year ago:

Does AAII have a screen for this method?


Richard Oetting from AZ posted about 1 year ago:

An excellent article.


Charles Egerton from MS posted about 1 year ago:

The earnings quality composite sounds like a very valuable tool for a stock investor's armamentarium. However, for casual purchases the mathematics is prohibitively tedious. Is this tool available on-line?


Charles Egerton from MS posted about 1 year ago:

The earnings quality composite sounds like a very valuable tool for a stock investor's armamentarium. However, for casual purchases the mechanics of its calculation seems dauntingly tedious. Is this tool available on-line?


james keller from california posted about 1 year ago:

How can I learn to calculate this composite method for picking stocks


Vaidy Bala from AB posted about 1 year ago:

Is there a software available to these analysis for each stock? Or any of the free AAII screens would do them. One thought is what is the time frame you look through this window?


Ronald Raecek from CT posted about 1 year ago:

Does AAII have a screen for this method?

Is there a software available to do these analysis for each stock?


F Krasowski from CT posted about 1 year ago:

Why doesn't AAII have a stock screen for this method?


Robert Black from FL posted about 1 year ago:

How do I get this score?


Maurice Peel from AR posted about 1 year ago:

The article is great. I want to try it but where do I get the screen? Without the screen it isn't usable for me.


Daniel Ekstrand from IL posted about 1 year ago:

I agree the article was good and makes sense, but without some way to put it, or other similar methodologies to use, it is just a read. What should we do next? How do we use it? Also, if one is able to use this type of filter against the world of good dividend paying stocks - BINGO in my world!


William Bentley from KY posted about 1 year ago:

Early last year, I attempted to put O'Shaun's very good work to practice, adding to it some of the work done by another researcher. I used Stock Investor Pro to generate data, first filtering to meet the authors' market cap filters. I downloaded the data into Excel. Very few of the exact metrics were provided; some of those could calculated in Excel, in many cases data was missing, requiring that I try to collect the raw data from other sources. After sorting, I then computed composite metrics using multiple factors. In summary, you can't screen directly using SIP, and the amount of work required to do on your own is quite substantial. Add to that the inability to effectively back test, and I shelved this concept. Too bad, I think the author is on to something.


Gerald Sensabaugh from TX posted about 1 year ago:

What is AAII's answer to the above questions/observations ??????


Wayne Maybach from VA posted about 1 year ago:

Excellent article - and I, like many other AAII members. would like to know if AAII can provide a stock screening tool using this methodology. It sounds like a lot of research for individual investors to perform and just not practical.


Charles Rotblut from IL posted about 1 year ago:

Hi,

I'm looking into seeing what we can do in terms of providing instructions on how to replicate Jim's criteria in Stock Investor Pro. I think it will likely require the use of SI Pro and a spreadsheet.

Keep in mind that a big focus of this article is not to rely on one single criteria. Rather, consider valuation, financial strength and the quality of earnings.

-Charles


David Huong from IL posted about 1 year ago:

This is and excellent piece of information.
However, if aaii can have a screen to share, that will be fantastic, if not reading these information is great as a general knowledge.
Appreciate it very Jim.


Scott Strobeck from GA posted about 1 year ago:

I love it! A composite of composites, but they only suggested combining composites without suggesting a method or providing results.

Anyway, I ditto the request to have AAII put together a screen. Maybe just average the 3 composites and take the top 5%?

Scott


James Ranum from FL posted about 1 year ago:

That seems to be the problem with AAII. They throw info like this out there and walk away.
Over and over, I've seen this.


Louis Lasday from FL posted about 1 year ago:

The theory is great but AAII never seems to cross the finish line. Let's be more specific. Some of your readers don't have time for study hall. A list or screen would be helpful


Jim Hamilton from FL posted about 1 year ago:

To answer Ronald's question:
Yes, AAII has not one but six O'Shaughnessy screens. Here they are and how they rank (out of 77 total screens, where 1=best) this year:
Small Cap Growth and Value (8)
Tiny Titans (13)
Growth Market Leaders (33)
Growth (45)
Value (65)
All Cap (74)


Frederick Joffe from KY posted about 1 year ago:

Can you show us a list of stocks that are in the top decile of all 3 composities now?


Frederick Joffe from KY posted about 1 year ago:

Can you show us a list of stocks that are in the top decile of all 3 composities now?


Thomas Conley from IL posted 12 months ago:

I know of three ways to screen for this data. S&P Capital IQ, Zacks Research Wizard, and Morningstar. A paid membership is required for all three, but in my view it's well worth the cost for all but the smallest investors.


G Shaner from IA posted 12 months ago:

I spent about two hours attempting to see if I could create a screen with SI Pro. Some of the parameters were all ready there. i.e. price/sales or price/earnings. Others were elusive even with a custom formula i.e. accruals/assets.


Dean Copeland from CT posted 12 months ago:

Back in May of 2010, John Bajkowski wrote a First Cut article the explained how to implement Tortoriello's Quantitative Strategies in Stock Investor Pro. I think a similar First Cut article would be great as a follow-up to O'Shaughnessy's article. In particular, I like the idea of measuring Earnings Quality by looking at accruals, but I can't find Accruals in SI Pro.


Doug from NY posted 12 months ago:

Dean Copeland wrote:
In particular, I like the idea of measuring Earnings Quality by looking at accruals, but I can't find Accruals in SI Pro.
-----
It should be straightforward to compute: See
http://www.investopedia.com/university/accounting-earnings-quality/earnings5.asp


Mark Seibert from CA posted 12 months ago:

I am going to suggest that if this method works so well then why doesn't everyone use it? It is evidently not a secret. I will also suggest that if it works well everyone will want to use it and it will therefore cease to work.
So what does work, probably no set formula but I suspect doing what the crowd is not doing i.e. buying when a reasonably priced stock is pummeled for missing its earnings by 1 cent, that should be 1 cent out of many of course. Another case is buying when the next finance panic arrives. The even harder question though is when do you get out?

Not trying to be negative, just pointing out that everyday people are hatching systems to beat the market, many will work for a time and fortunately some can even be recycled as they fall out of favor.


Jim Linnemann from MI posted 12 months ago:

What wasn't clear to me is whether these results are pure back-testing based, or whether they work in the real world going forward.

It it's just backtesting, well one looks at lots of plausible things and only reports the ones that look good in retrospect. That's often called data snooping or data dredging (see the wiki). Notice that the screens Jim Hamilton cited above, based on earlier work by the same author, are all over the map when applied in real time.


I'm a lot more impressed if the methods have been applied successfully in the real market for some time, after having been previously developed.


Robert Koch from MO posted 11 months ago:

I have two questions:

How is "external financing" calculated? This is one of the financial strength factors.

The last sentence in the first paragraph under "Financial Strength" says, "In the case of the financial strength composite, stocks missing a value for a factor are ignored; to be included in the composite, a stock musts have a value for at least two of the four factors." Which way is it?


Doug from NY posted 11 months ago:

It's curious when a composite outperforms all its components (this wouldn't happen with an arithmetic average, for instance). It makes me think that there are probably other plausible composites which UNDERPERFORM all their components. The question is, how to determine, in advance, which kind you are dealing with....


Arnold Siemsen M D from HI posted 11 months ago:

In the 2012 edition of O'Shaughnessy's "What Works on Wall Street" starting on page 284, there is a lot of data on results from ranking the four factors in earnings quality. These factors are mentioned in the current article, however, there is no mention of how to screen for them. If we could get some AAII help with the definitions, perhaps we could write the screen in SIP.


Charles Rotblut from IL posted 11 months ago:

We are working on a SI Pro screen based on this article and will discuss it in a future AAII Journal article.

-Charles


Raymond Vance from CA posted 11 months ago:

Mr. O'Shaughnessy has managed my stock portfolio since 1997. The portfolio is always fully invested. Since then he has beaten the S & P 500 by an annualized 6.2 percent per year. I am more than satisfied.
Everyone doesn't accept a successful strategy because they want to time the market and will abandon the strategy if it does not beat the index in every short period they might consider.


Raymond Vance from CA posted 11 months ago:

Mr. O'Shaughnessy has managed my stock portfolio since 1997. The portfolio is always fully invested. Since then he has beaten the S & P 500 by an annualized 6.2 percent per year. I am more than satisfied.
Everyone doesn't accept a successful strategy because they want to time the market and will abandon the strategy if it does not beat the index in every short period they might consider.


Arnold Siemsen M D from HI posted 11 months ago:

Charles, glad to hear you are working on a screen. In the article he discusses the percentile ranking system. For example if a stock has a price-earnings ratio that's in the lowest % it gets a one. On page 302 of the latest "What Works on Wall Street" he says that if a stock has a PE ratio in the lowest 1% it receives a rank of 100 and it gets a 1 if it is in the highest %. This is confusing. Which way is it?


MarkP from CA posted 11 months ago:

Hi Arnold,
In the book and in the subsequent AAII Conference presentation actually he has used different scales (100 being the best, or 1 being the best). You can set it up either way. But, "best" needs to be consistent in your calculations, obviously.
And "best" in these calculations means "highest value". So if you call 100 "best", then for PE, a high PE would get a low percentile value, and a low PE would get a high value. So, the stock with the lowest PE would get 100. In the book actually he talks about EP (Earnings to Price). But you just need to be consistent about "best".

Mark


Richard Fisel from CA posted 10 months ago:

I am trying to understand the criteria in O'Shaughnessy's screens. What is the difference between Total Accruals (I presume for trailing 12 months) and Current Accruals? The definitions seem to be the same, cf. http://www.investopedia.com/university/accounting-earnings-quality/earnings5.asp


Thomas Ford from NY posted 10 months ago:

This does not seem to differ that much from what Piotroski was doing. For the accruals piece, one can use the calculations from Sloan, which appear a bit simpler. I am curious why no longer term valuation metrics were utilized such as P/B or a company-specific CAPE style valuation metric in order to get away from y/y earnings volatility. The problem with O'Shaughnessy is he measures every stock on whether it is absolutely cheap or not, but high return, wide moat companies will often not show up as being absolutely cheap, but can be historically cheap and as a result great investments.


John Givens from FL posted 10 months ago:

While I am typically quick to lash out at AAII for their total lack of customer service (responding to emails, backtesting 1st cut and other strategies instead of just putting them out there, etc. etc.) they are still the best service and value in existence. Painfully, I think they deserve a break. While they could do us a great service by working with O'Shaughnessy and setting up fields and screens for "What Works" the main problem here is the book. So far, I have not been able to find definitions for "financial composite", external financing, change in operating assets, etc. in the book. How can O'Shaughnessy expect us to take his book seriously if he doesn't define his terms. Now, I can find definitions from other sources, but why should I have to and how would I know if they are the same definitions that he used?


Michael Dimillo from GA posted 9 months ago:

I have been using SI Pro and Excel to replicate the Trending Value Screen (aka VC2) since April of last year. A few of the metrics require custom formulas, but it isn't that difficult. O'Shaughnessy has some material on his site that help with the calculations. You have to translate it from Compustat, but it is doable. Here is the link - http://www.whatworksonwallstreet.com/supplement.html.

So far the top 25 I generated in April is up 40%, excluding OMX, DELL, VMED, and STRZA, which were all acquired. In the real world you would have reinvested, but it was too much effort to track it. He recommends a yeah holding period, but I have been running it every month when the new data sets come out as well so I can compare it to the other screens on AAII that I use.

I think it will be hard to replicate this in SI Pro alone, but definitely think it is worth the manual labor. I am re-tweaking my spreadsheet to plug some data holes, but I would be happy to share it with others when I am done.


You need to log in as a registered AAII user before commenting.
Create an account

Log In