What I like about the paper is, that Matthias seems to be a great blend of an academic with nice practitioner approach. From what I understand is, that he puts all the papers of the last 70 Years together and looks what is working the best…
Would love to create a ranking system based on his paper (page 24 following…)
Got too much projects now, so lets work together on it?
Blogpost from p123 based on the paper? Interesting because he tests it outside of the US and we could test it for the US.
Its probably not possible to get the model 100% replicated but an 80/20 approach would be cool!
I can find a lot of stuff from Matthias in my own main ranking system, but some stuff is different.
I could put together a ranking system based on the six-factor model that Hanauer favors. However, the only value factor he uses is book-to-market, which is probably the least useful value factor of all. I’m not sure why he only uses this one. He doesn’t even mention the price-to-sales, price-to-earnings, or EV-to-EBIDTA ratios. Anyway, I’ll post again once I’ve created the ranking system, sometime in the next few days. But I would think that our own Core Combination ranking system, especially if combined with a size filter, would outperform it.
There’s a sixth factor, which is the market factor (index return minus 1-month T-bill return), but I didn’t include it because it would make no difference to ranking. That’s just if you want to do multiple regression.
I called it Barillas because that’s how Hanauer refers to it.
Very nice Yuval. I learn a lot studying other’s RSs. This RS did well on my microcap universe, weekly. And performed freakishly well for more recent 1 - 3 year tests. I see all your sub nodes are set to zero. I assume that gives them equal weight?
Yuval… thanks for putting that RS together. Just out of curiosity I ran a test on the P123 MicroCap Univ subbing out the Book Value base Value Factor with P123’s Value Ranking System, and the results were interesting. In particular, the 1-3 year outperformance of the Barillas model referenced by abwillingham above appears to be almost entirely driven by the outperformance of a simple Book Value based Value Factor from the March 2020 low. Replacing the Barillas Value Node with P123’s more complex Value Ranking System eliminates a significant amount of this performance.
Not sure what the takeaway is, but found it very interesting nonetheless.
I believe Clifford Asness of AQR has commented that price to book value has performed better than most value metrics during the “value inversion” of the last 1-3 years simply because it’s not a very good indicator of value.