Quant investing: a better value composite?

In quant investing there is always an urge to continuously look for ways to improve a model. Trying better ways of doing things can be a worthy and profitable endeavor. After all, markets do change. You learn new things, etc… But it can also be fraught with pitfalls. Data mining is a constant worry. With these caveats in mind in this post I’ll take a look at a potential new value composite metric and it’s performance over the last 16 years.

My idea for a new value composite comes from Professor Damodaran at NYU, his blog is at Musing on Markets. His academic site is here. If you’re a stock valuation person or a fundamental investor you’re probably familiar with his work (if not you should be). He wrote THE BOOK on stock valuation. He also does some of the best work on equity risk premiums (my favorite measure of ERP is his implied ERP adjusted for sustainable payout). You can also take his valuation course on-line or view past archives of his classes in various formats. A great resource. In a recent post Damodaran goes through the pricing of wordwide markets based on four ratios. The ratios must meet his consistent multiple rule which he describes as follows:

The “Consistent Multiple” Rule:   If your numerator is the market value of equity (market capitalization or price per share), your denominator has to be an equity measure as well (net income or earnings per share, book value of equity. For example, a price earnings ratio is consistent, since both the numerator and denominator are equity values, and so is an EV to EBITDA multiple. A Price to EBITDA or a Price to Sales ratio is inconsistent, since the numerator is an equity value and the denominator is to the entire business, and will lead to conclusions that are not merited by the fundamentals.

That got me thinking that the value composite that I use for many of my quant portfolios doesn’t meet this rule 100%. What if I changed VC2 to meet the rule requirements? Would that help performance? Or is this just some academic concern that doesn’t matter in real world portfolios? Let’s see.

I created a new value composite, VC Damodaran, that combines P/E, P/B, EV/Sales, EV/EBITDA and P/FCF. I replaced P/S with EV/Sales and dropped SHY. I then tested the VC2 quant portfolio over as long a time as possible on P123 with the original VC2 composite and the new VC2 Damodaran. Universe is All Stocks, 25 stocks equally weighted, re-balance the portfolio once a year. Results are shown below.

The new value composite results in increased absolute and risk-adjusted performance over VC2.  The other important thing I noticed during the testing of VC Damodaran is that it interacts better with a quality overlay in the portfolio. Using VC Damodaran I was able to tighten my debt change quality overlay in the portfolio. Doing this increased returns to 19.3% per year while keeping drawdowns the same. Risk measures improved in line with these results. The revised screen also work with various stock universes, more concentrated portfolios, more frequent re-balancing, and risk reduction strategies such as using SPY-UI to reduce drawdowns. Rolling backtests were also just as strong. Just like the original VC2. Impressive.

Let’s take a look at what the top 10 holdings of each portfolio would be today. As with any pure value portfolio expect to see some puke worthy recommendations. That’s always my test for a good value portfolio. If you don’t want to puke after reading it’s recommendations then it’s probably not value enough and won’t work so well. Here are the top 1o holdings for each. Original VC2 on the left, VC Damodaran on the right.

Only 4 of the top 10 holdings are the same. And many of them are definitely puke worthy like any good value portfolio should be.

The key question is are the results real? Are they persistent and sustainable? Or are they the result of data mining? I can’t answer those questions with the tools at my disposal. But they are encouraging. They are based on some solid fundamentals. For now, I’ll track the results of the VC Damodaran composite vs the original VC2 composite.

In summary, a slightly different twist on the original VC2 value composite may be a better way to capture the value premium in individual quant stock portfolios.

 

Full Disclaimer - Nothing on this site should ever be considered advice, research or the invitation to buy or sell securities. These are my personal opinions only.

14 thoughts on “Quant investing: a better value composite?

  1. Nice read, as usual!

    I’m wondering about two of the choices, though: Why did you drop SHY? and why did you keep P/B? There wasn’t any discussion of these choices in the text.

    There’s been a lot of talk about P/B not being a useful measure anymore, since buybacks started becoming a big thing. Does your research contradict this?

    1. Simon, I dropped SHY because it wasn’t one of Damodaran’s valuation indicators. I then went back and added it to VC2 Damodaran and results weren’t as good.

      I did a whole post on dropping P/B or not. Based on my results, I decided to keep it in most of my strategies.

      Paul

  2. Good job Paul … great info.

    I’m wondering why you dropped the Shareholder Yield (SHY) from your modified VC2 (it is now a modified VC1)?

    I believe O’Shaunessy in WWOWS added SHY to VC1 because this improved draw-down and volatility (and it also added an insignificant performance increase). Do you think SHY may add the same characteristics to this new composite?

    Again, thanks for the info. I was not aware of the Professor and his book.

    B

    FYI – I posted a link that Pat OShaunessy did on P/S vs EV/S that I remember reading.
    http://investorfieldguide.com/using-the-price-to-sales-ratio/

    1. B, I left it out because it wasn’t in Damodaran’s valuation indicators. I then went back and added it and results weren’t as good.

      Thanks for the link.

      Paul

  3. Paul, great stuff, original research for sure, congrats! When you say you tightened the debt quality filter, do you mean you filter any stocks that increased over the period? Just wondering what you mean by tightening.

    Thanks,

    Victor

    1. I mean I filtered out stocks further down the ranking in terms of debt change. So, where previously the cutoff for the performance peak was a debt change ranking below 70, now it’s below 90.

      Paul

    1. I found the same Matt. I also found it doesn’t play as well with momentum but as I mentioned in the post it works better with quality.

      Paul

  4. On a somewhat related topic: Do you have a metric to screen for O’Shaughnessy’s value composite 2 screen on AAII’s stock screener? Thanks for any help you can give me.

  5. Is there a stock screening program that you recommend that can calculate O’Shaughnessy’s value component 2 composite? I’d like to employ it to invest in his staples/utilities portfolio and an allcap and microcap portfolio. Thanks.

  6. I’m signed up for the free trial on P123. Now…where the rubber meets the road–I can’t figure out how to implement a screen for VC2 with O’Shaughnessy’s value composite criteria. Any chance you could point me towards something that could help me? THanks.

    If I can make this work, given the added over-market earnings, the fee will be peanuts…

    1. No easy way around that part. Just need to dive in, learn the tool, and learn how to implement the strategies. You just need to learn how to do screens and custom ranking systems to get strategies up and running.

      Paul

Comments are closed.