Book Review: O’Shaughnessy – What works on Wall Street (4th edition)

A first remark: I haven’t read the previous editions and I am a maybe biased active value investor ­čśë

For many years, O’Shaughnessy’s book “What works on Wall Street” has been the Bible of many quantitative value investors who wanted to avoid analysing single companies and prefer an automated startegy to select stocks.

The book starts with some chapters loosely summarizing current behavioural finance knowledge which implies that human selection is inferior to automated systems.

For the large quatitative part of the book which follows, he seems to have used a new data set on US equities which wasn’t available before which is going back to 1926. He clearly lays out the methodology, with the major features being

– avoiding illiquid small caps

equal weighting of stocks in the portfolio (we’ll come to that one later !!!)

– reinvesting dividends

– enhanced rebalancing (not only on July 1st per year but the average of 12 annually rebalanced portfolios starting each month of the year)

– no transaction costs assumed

He then uses his data to run many different strategies. Some of the most interesting results for me were:

– small caps outperform large caps and indices pretty consistently by 1-2% p.a.

– low P/E stocks (cheapest decile) outperform by almost 5% p.a., interestingly with a lower beta

– low EV/EBITDA works even better as single factor model (5,5% p.a. out-performance)

– low Price/Cash Flow works, but “only” around 3.5% p.a. out-performance. Price to free cashflow and Price to operating cashflow work slightly better but not as good as EV/EBITDA

– the “old favourite” price to sales does not really work well (only 1.5% p.a.)

– Price to book: DOES NOT WORK for the cheapest decile !!!!

– Dividend yield DOES NOT WORK

– however stock buy back yield works with ~3% p.a. excess return

– Stock buy back yield and dividend yield work better, 3.3% p.a.excess return

– certain accounting ratios work also well, such as low accruals (2.8%), Asset to Equity, asset turnover etc.

– composite accounting ratios (4 factors) work even better, close to 5% p.a.

– composite ratio with only the 25 best stocks scores even better (7% p.a. out performance)

– composite Value factor (P/B, P/E, P/S, EBITDA/EV, P/CF) with 5.8% p.a. out-performance

– Earnings changes, profit margins and ROE DO NOT WORK

– Price momentum (6 and 12 months) works, with 2-3% p.a.

combining value and price momentum works best, some strategies yielding 10% p.a. excess returns or more


I don’t want to sound arrogant, but from earlier discussions I knew that O’S favoured the P/S ratio in the prior editions. I always thought that this doesn’t make any sense because then you end up with a portfolio of supermarket stocks and wholesale companies.

It is also not surprising that P/B doesn’t work if you don’t adjust for debt. So no wonder, EV/EBITDA does a much better job. But this is something many active value investors know without having to go through 90 years of data.

For any active investor it is also no surprise that it is better to look a various factors as single factors can always be influenced by certain special effects. So welcome to the club, Mr. O’S !!!

Maybe in his 5th edition, O’S then starts to look at the historical developments, who knows ?

For me, the relevance of price momentum is still difficult to really understand but I am willing to learn !!!!

Conceptual issue: Equal weighted performance

I have one big conceptual issue with the whole book: as described in the beginning, O’S just briefly notices that he uses equal weighted portfolios. He doesn’t test if this alone has an impact on the performance of the strategies.

Some recent papers indicate that equal weighted portfolios themselves create significant out-performance vs. market cap weighted indices.

So we do not know for sure, how much of the out-performance of O’S strategies is due to his equal weight assumption as he benchmarks against a market cap index and not against an equal weight index.


O’S book is of course a interesting read. Many of the newer combined strategies make intuitive more sense than some of the older strategies. He also correctly states that strategies work only long term and you have to avoid market timing at all costs !!!

Many investors will underestimate what it means to invest in a mechanical strategy which underperforms for 3 or more consecutive years.

Unfortunately, the conceptual issue with the equal weighted portfolio takes away a lot of my personal “trust” into those mechanical strategies.

For me the key take aways are:

1. Any more or less fully invested value startegy with some basic analysis will perform quite well against the general market

2. Also the most famous mechanical models can become outdated


4. Price momentum should not be ignored


  • With regard to stay invested: Well, although I agree with O’S in his basic assumptions, I do not draw the conclusion from what O’S writes, or rather: what you conclued, that you should not try to “time” the market. He only says that one should persue one strategy and stick to that – within the equity part of the portfolio. But why not reduce the equity part and increase cash, when the whole market is overpriced? “What works on Wall Street” is all about RELATIVE valuations – relatively cheap is relative better than relatively expensive. That does not mean, that you should buy the relatively cheap stocks, when all stocks are expensive! I also do not fully agree, that you necessarily always have to stick to the same strategy – for exampe, you can and should ignore RS after a market crash. You will find the most intersting stocks among those who suffered from the greatest declines. You can also ignore the p/e ratio, since otherwise, you will select defensive stocks, and not the more interesting depressed cyclical stocks.
    I agree with your second conclusion: I do not believe, that it is necessary to pursue a strategy exactly in the way that O’S tested it. I think one can look “behind the spirit” of the strategies – and this is, generally said, to buy cheap stocks (and a positive medium term RS), and to ignore margins and so on, and maybe watch some red flags.
    I also agree with your third conclusion. I conclude from O’S data, that it is much more important to avoid the falling knives, than to invest in momentum stocks (only rather small outperformance vs. greater underperformance). That’s also sometimes emotionally difficult for me, as a contrarian at heart. By the way, the RS turns already negative after 16 months! That’s why momentum stocks (high RS) are a hot ride – when the momentum breaks, then it turns relatively quick into the opposite…

  • Yes, that’s also the most important thing for me! It is not SO much important, which single strategy worked best, because I also would not rely on one single valuation metric. In the third edition, there was some hint, that combining the strategies improves the result, but I am really happy, that he now tested the rank-sum, and that this strategy is effectively the best over all (only four microcap strategies had a better compund return, all of them with a small number of selected stocks, 10-50, which indicates, that it could be just chance, and three of them are based on the p/b). Unfortunately and surprisingly, the compound return of the value composites was only about 1 percent better than the best single strategies (e.g. EV/EBITDA), but with a much higher consistency and with the worst decile performing with a negative return! Together with the relative strenght, you get around 9 percent outperformance vs. the “All Stocks” universe (or even more vs. the S&P500, over 10%). Even if the overall stock market won’t perform with 9-11% p.a. from now on (the real return was significantly lower, of course) – is it really awarding to ignore the statistics, when you earn 5-6% + 10% with such a strategy, or rather: can one afford to ignore it?

    • Hi Winter,

      no, one should definitely not ignore it. Howver, I still refuse to go for “fully automated”. I think one of the big lessons is to stay invested. Secondly you should buy stocks which are generally cheap. thirdly you should not catch falling knifes.

      The third part is the hardest for me.


  • Winter,

    thanks for the comment. As you know, I am a “newbie” with regard to this. I still like the combined criteria as well as the accounting criteia. This might make a good screener.


  • Well, rest assured, it’s clearly not due to equal-weighting, because the “All Stocks” Universe also is an equal weighted portfolio, not a capital weighted like the S&P (at least I understood it in that way). That’s why the All Stocks perform much better than the S&P500. (Ok, Large Stocks nearly had the same performance, although it is nearly a universe of the same stocks, so there was no small cap effect within the large caps – but the small cap effect is relatively small, anyway – most of it disappeared in the 4th edition, and he clearly pointed out that much of the microcap-outperformance is due to pennystocks which twentyfolded in one month – whatever that was, just a data error or a obscure mining company, or pushed stocks.)

    I also always had my difficulties with price momentum, but I solved this conundrum: This is no contradiction to an anticyclical approach, in fact, it even does confirm this approach. It only means, that a trend lasts over a certain time (in the medium term). So in the shorter time frames, you actually will see that momentum works. The ideal time frame is 6-7 months, it also works for 1 year, but with two years, the outperformance disappears. And remember, the holding period is just one year! Whereas over the past five years, a bad relative strenght is better!

    The price-to-book also works over the whole time period, and also for the first decile, but not only is the outperformance rather small, it is not consistent, not just within the deciles, but also over time. That really came like a bombshell. The debt ratio does also not work, but maybe in conjunction? At least it seems, that the p/b still adds value in a combined strategy.

    The P/S made sense to me, because besides wholesale stocks, you always get stocks with a very low P/S relative to their sector – and the P/S works also within one industry sector; so the average of that should work (and it still does, it still adds value!). I’m sure, in the next edition, there will be different single strategy which worked best. In Richard Tortoriellos book, it was the P/FCF ratio.

    Nevertheless, there are still plenty of questions open about the book…

  • Assuming that no taxes are deducted to reduce the yield metrics, why would share buy backs perform much better than reinvested dividends?

    It is possible that decreasing floats contribute disproportionately to increase valuation metrics. I know that screening for low float, high institutional and insider ownership metrics is an old trick in the bag of momentum investors.Low float (though not illiquid) could be interesting metric to test for.

    An other explanation would be that investors typically use dividend proceeds for other purposes than to reinvest in stocks, like discretionary spending or investing in fixed income. This would mean that the temptation of using dividends to suplement income is a pretty big value destroyer if it means that you’ll be compounding at 3% less over many years.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s