[i]. Since each decile represents several hundred stocks (and is therefore unrealistic for individual investors to manage) I’ve also included a concentrated, 25 stock strategy which selects the best stocks in the market by this measure. So does this simple measure help you find stocks that will outperform the market? The answer is a qualified yes. The top 20% of the market by the P/B-to-ROE measure does tend to outperform by a decent margin…but the concentrated portfolio isn’t as impressive.
Results
Here are the results for the 10 deciles—and the 25-stock strategy—versus the market. Both of the top two deciles outperform the market, but the second decile provides the highest Sharpe ratio.
More curious is that the concentrated version doesn’t do all that well. While it does beat the market over the very long term, it is not at all consistent. In rolling 10-year periods, it only beats the all stock universe 26% of the time—before costs. If transaction costs and taxes were factored in, it wouldn’t be worth it. As for the top decile, you’d have to be very patient to win with this strategy. There has been a 1-in-5 chance, historically, that the best decile underperforms the market in a 5-year period-and again, that is before costs.
I should note that price-to-book is the least impressive value factor I’ve tested, and ROE is the least impressive measure of quality/profitability. EBITDA/EV, for example, yields much better results than P/B, and ROIC yields much better results than ROE. Is this just historical coincidence, a small bit of data mining? It is impossible to say, but there are compelling reasons for using factors other than P/B and ROE—but that’s a topic for another post.
Let me know if you have other ideas for testing, email me at [email protected]
[i] Deciles are rebalanced on a rolling annual basis, and companies with negative earning and/or negative equity are excluded.
/rating_off.png "1 Star")
Loading...
Thanks for the post. When you back test any strategies, would you also show the alphas( CAPM, FF three factors and FF three factors+ momentum ) and their t-tests ?
need to expand my framework a little to get carhart alpha and t-stats, will do so when I get the time
Hi Patrick, I am a recent graduate and have been very interest in the Quantitative Equity research, and was wondering if you can share some insights about how you do those factors tests, like what tools you use and what datasets you rely on?
For me, I mainly used the WRSD database and used Matlab for data analysis… but from your post, it seems like you used excel add-in tools, can you share what commercial tools you use ? Thanks a lot!
I’m working with a strategy similar to Greeblatt’s with the addition of a "shareholder return" factor: (dividends + share repurchases + debt paydown) / Market Cap. This three-factor system of value (EV/EBITDA), quality (ROIC), and yield seems to me a comprehensive analysis method. Any thoughts on this approach?
Hey Matt, it works fantastically well. Three of my favorite selection factors.
Hi Patrick,
Any thoughts on the ‘Asset Growth Effect’ as a factor?
I’m trying to implement a system with ‘value’ (EV/EBITDA or EV/EBIT) + ‘momentum’ (12-1mo price change or price to 52w high) + Asset Growth [ (Assets (t) / Assets (t-2))-1] and/or Shareholder Yield (CFF /Mkt Cap).
Regards from Spain
Shareholder Yield has worked historically much better than asset growth, although asset growth captures two important effect: capital intensity (growing/high capex) tends to lead to weaker future returns, and growth in current accrual accounts (AR and Inventories) can be a sign of poor earnings quality (which also leads to weaker future returns historically)