Quant Investing


The performance of value factors varies over time. Sometimes value is in favor. Sometimes it is out of favor. But overall value overall is one of the two single factors, along with momentum, that has withstood the test of time. But what if one way of expressing value in stocks has simply stopped working or is just nor working as well as in the past? That’s is what I’ll consider in this brief post. In particular, I’ll look at whether it is still worthwhile to use P/B in individual quant stock portfolios.

Many of the quant stock models discussed on the blog use a composite of value factors to find the best value stocks. One of the factors is Price to Book, probably the oldest and most used of the value factors. It is well documented that Price to Book is not as effective as it used to be. One way to look at the effectiveness of value factors is to rank stocks by a certain factor and then look at the spread between the most expensive and cheapest stocks. For example, the charts below show the spread between the cheapest and most expensive stocks for three value factors (see this OSAM paper for details).

A narrower spread between the cheapest and most expensive stocks means lower performance for the value factor. The chart below shows rolling excess returns for the value factors above.

Something change in the 1999-2000 downturn. Price to Book ceased to match the effectiveness of the other value factors. Great. Let’s dump Price to Book from the value composite and be done with it. Sounds like a good plan. Let’s see how that works out in various quant portfolios .

I created a new value composite that eliminates P/B and compared it to the original VC2 in the VC2 value portfolio , the Utilities value portfolio , and the Trending value portfolio . I calculated returns of the portfolios in P123 from 1999 through 2016. A lucky coincidence is that the earlier P123 backtest data happens to align with the time Price to Book started underperforming, All portfolios consist of 25 stocks and a 1 year holding period as usual. Below are the results of the various backtests I performed with the best performing portfolio highlighted in green.

Hmm. Not as straight forward as you would expect. As originally constructed the returns are higher for portfolios formed with the value composite that does not contain Price to Book as a factor. This confirms what the graphs above told us. But, a few things to note. One, the outperformance is not as high as you might expect. The analysis presented above is for large cap stocks whereas the quant portfolios are run using the All Stocks universe. The inclusion of small caps changes things a bit. Also, the quant portfolios are equal weighted instead of market cap weighted which also seems to influence the results.

Most importantly, real world portfolios usually combine other factors that when combined with value, may influence the performance of value. For example, many of the quant portfolios I present here use some quality factor to either improve returns and/or reduce drawdowns. In the table above you see the strategies implemented with a quality metric and then run with and without Price to Book in the value composite. These results change the conclusion. When quality is incorporated into the portfolios Price to Book is still quite effective in 2 of the three portfolios, with only a very slight advantage in the Utilities value portfolio. Maybe we shouldn’t be too quick to throw our Price to Book?

In summary, based on the recent performance of Price to Book as a single factor it makes sense to discard it in quant portfolios. However, when combined with other factors the decision is not so clear. You need to run through the analysis in your own portfolios. For my portfolios, I will continue to use Price to Book in the VC2 value portfolio and the Trending Value portfolio. On the next re-balance of the Utilities Value portfolio I will switch to using the value composite without Price to Book.


Tagged , , ,

10 thoughts on “ Quant investing: is it time to get rid of Price/Book?

  1. Your blog entries on Quant Investing peaked my interest, so I borrowed and just finished What Works on Wall Street Fourth Edition, and I am about to go back and re-read your blog entries and comments.

    One thing I may revisit from the book is the amount of correlation between the factors that make up VC2. O’Shaughnessy gives all of the composite factors equal weighting, but if two factors are too highly correlated then the driver behind those factors essentially gets a 2x multiplier in the composite. It might be interesting to do a sweep of different factor weightings to see how much the balance of factors makes a difference. Do you think that would cause too much of an overfit to the data?

    I may try a few of the strategies once I figure out the most cost-effective way to get the necessary screener data.

    1. Hey John, yes I do think trying to optimize weighting would lead to overfitting of the historical data.


  2. Paul,

    Thanks again. We’re lucky you read all those books, blogs, and articles and share with us what you see as pertinent to quant investing.

    Figure 3 intrigues me. Ignoring the obvious dropoff of P/B since 1999, the last 10 years also shows that Ebitda/EV is slightly more significant than P/E. While WOWS and others suggest sticking with a strategy over the long haul, having data and computers readily available opens the door for all of us to monitor the effectiveness of factors over rolling periods and to tweak our quants accordingly.

    Any plans to monitor all the factors you use and then apply weighting to each? For example, we might be better off giving a little more weight to Ebitda/EV than to P/E in a value composite. Any interest in analyzing this and developing unequally weighted value composites that change every year based on the prior various rolling periods? Could fit into that new business you are thinking about.

    1. In general, non equal weighting of the factors opens things up to unnecessary complications. Not worth it in my opinion.
      The only thing I may look at is timing factors by where we are in the economic cycle.


  3. Paul,

    I might be missing something. In your chart of 3 quants titled, “Is using P/B still worth it” you show Utils w/ Qual. What is the Qual? I only show rank on VC2, sector=util, and yield>0. Is the yield >0 the quality factor?

    1. No. I tried to add a debt change quality filter. But it doesn’t have to be. Several quality indicators work about the same.


  4. Hi Paul,

    I think when you wrote the post awhile back that discussed using a quality filter you were using it in the context of a value-only portfolio. It looks like you’ve started to use it for quant strategies that also include momentum now. Would be interested to see a post on which strategies you find the quality filter to be helpful.

    1. I use quality everywhere I use value. So the only quant that doesn’t use quality is pure momentum.


      1. Thanks for clarifying that. Going back and looking at the post where you discuss this (for those interested: https://investingforaliving.us/2014/10/03/improving-the-performance-of-quant-value-portfolios/ ) you say, “I rank stocks by percentage debt change (net debt change in $$$ divided by total debt) and only keep the top 3 deciles (least change in debt) of stocks (30%).”

        Just to be sure I am understanding this correctly, are you running this quality filter in conjunction with the other universe-limiting parameters? For example, with the TV2 portfolio, limit the universe to U.S. stocks, $250M+ market cap, no OTC, and then take top 30% of those by smallest debt change. Then follow up by ranking based on VC2, and finally momentum. Is that right?

        1. Pretty much. I run it as a rule instead as a specific reduced universe but effectively that’s the same thing.


Comments are closed.