Sabermetrics for asset managers

[tede02]

This article was posted in another thread.  It is an excellent article that raises a lot of issues. 

 

http://www.slate.com/articles/arts/books/2016/12/how_daryl_morey_used_behavioral_economics_to_revolutionize_the_art_of_nba.html?wpsrc=sh_all_dt_tw_ru

 

One that came to mind, and perhaps its been discussed, but is there any merit to attempting sabermetrics to asset managers?  I guess firms like Morningstar kind of attempt this, but it is really more on a fund level.  Could individual asset managers (people) actually be statistically evaluated in better ways just as Daryl Morey in the the article had to figure it out with NBA players?  Or is there simply not enough information to measure.  For example, I doubt measuring someone's height has any correlation to their success as an investor.  ;)

 

 

 

[Bluffy]

If it would be measurable, I would attempt to look at the correlation between time read and investment success.

[Jurgis]

For example, I doubt measuring someone's height has any correlation to their success as an investor.  ;)

 

Possibly OT, but you may know that someone's height has a correlation to their success in life (I forget if this was done for business people only or some wider population). We have a positive bias for tall people and they have a positive bias (confidence?) for themselves. Same for some other physical attributes.  ::)

[Travis Wiedower]

^^ Makes sense. I remember seeing something years ago that the average male CEO in the US was 6' while the average height for a male was 5'9" (numbers may be off, barely remember it).

[LC]

I would find some way to measure marketing presence ($s spent as % of AUM, or perhaps some other metric) and compare that to historical returns. My guess is heavy marketers underperform, so you could use it as a tool to disqualify asset managers (rather than select).

[Guest Schwab711]

http://www.starvinecapital.com/moneyball-how-is-baseball-related-to-value-investing/

 

@Jurgis: Looks like it was from the book Blink

-------

 

As to the actual thread topic, I work on this idea occasionally. Most of "Moneyball" started with finding correlations for accepted rules-of-thumb in baseball to see which were true and which weren't (and the magnitude between perception and reality). That's probably the place to start with this (not that I have done that, but it probably is).

 

I think we would need the annual returns and transactions for thousands of investors to replicate Beane's work for investing. There is a breakdown between baseball sabermetrics and hypothetical investormetrics due to sample size. In baseball, a hitter will have an order of magnitude more samples (plate appearances) than an investor will have investments (hopefully!). Most folks agree that it takes 6-36 months for a trade/investment to have time to play out (though it can extend to 10-20 years in some cases). We are always going to have a sample size issue. I would think the best way to approach the topic at first is to look at accuracy and allocation metrics but maybe there's better ways.

 

Here's a few of the ideas I can remember off the top of my head. The idea was to segment returns in luck vs. skill. I'm guessing there are issues with some or all of them and the decent ones still need to hammer out the specifics. You could probably weight the metrics based on position size.

 

• 1) The % of investments with positive returns over some period (I'll just use 1-3 years below to avoid being overly generic)
  
• 2) What % of investments declined > 50% over the next 1-3 years (regardless of whether you held during that period or not). Basically, are you investing in areas of high-risk (swinging at balls).
  
• 3) What % of investments return > 100% over 1-3 years (regardless of if you held or not). The idea with #2 and #3 would be to capture your outlier tendencies.
  
• 4) Portfolio turnover: Did you create the compounding or did the companies you invested in? There may be arguments on both sides but it would be good to know what an investor does. It's similar to the MBS issue. You get slightly higher return for your risk because when interest rates decline your higher yielding investment starts to get called.
  
• 5) Some circle of competence measurement: Probably something along the lines of #1 for each industry to determine where you have success. Of course, 90%+ accuracy in one sector may be better than 60%+ in 10. Who knows.
  
• 6) Concentration factor: Some type of measure of how concentrate you run on average (to weight any other metrics, if necessary)
  
• 7) Diversification factor: Similar to #6
  
• 8) Multiple expansion: How much of your returns come from multiple expansion versus underlying profits realized
  
• 9) Number of transactions (buy + sell): The idea would be to determine how thin you are spread from a research perspective.

 

Even if you begin to quantify the luck/skill of returns, there will always be a qualitative factor involved. All else equal, it's probably better to have financial institutions in your circle of competence as opposed to furniture manufacturers. I think 5 or so years would smooth out issues like that, for the most part. I'm sure there are other issues I'm not considering right now.

 

Anyone else have any ideas of how to attack this question?

[stylized_fact]

In a recent interview, Michael Lewis remarked that when you look for a doctor, you should seek out someone completely unmatched to the stereotype. The argument goes that such a person would have had a really hard time in medical school and therefore is probably really good.

 

I have a hard time understanding how this thinking might apply to investment management. An example that springs to mind is Michael Burry, but he didn't have a big struggle soliciting capital. I guess VIC and maybe twitter serve as an "alternative" farm system for the industry, complementing CBS, Wharton and other traditional sources.

[Jurgis]

I agree with Schwab711: this is a tough problem to tackle.

 

Couple observations from various angles.

 

Like Schwab711 alluded, everyone wants to be evaluated long term. So if we do evaluating using 1-3 year measurements, even this board would rise up in protest. Come on, how many threads did we have about known investor underperformance in last 5+ years and there's always a defense that this is too short, it's all bull market, blame indexers, blame FANG, etc. But if we go for 5-10 year timeframes, the number of data points shrinks a lot. The universe of managers shrinks a lot too (assuming your goal is to pick a great manager).

 

Related to last sentence, your goal probably would be not just to find a great manager, but to find an underappreciated great manager. I.e. if manager is obviously great (20%+ returns for decade, minimal drawdowns), everyone would want to invest with them and they'd either be running superhuge monster fund or be inaccessible to new investors or both. If your approach just tells that Seth Klarman or Stan Druckenmiller are great investors, it is pretty useless. How would you find an underappreciated great manager? Are there measures that show superlative managers while conventional measures miss them?

 

Maybe the paragraph above is less applicable if we evaluate just mutual fund managers and limit the analysis to funds open to new investors. Then perhaps it's enough to find great manager. The next question then is how you separate the results of a fund manager from the results of their organization. Perhaps this is easy for solo hedgies. But for a mutual fund, there will be analysts, etc. who contribute to the fund stock selections. If analysis assumes everything is done by a fund manager, but in reality most work is done by analysts, what happens if some of the good ones leave?

 

Anyway, possibly it could be an interesting project. However, even if someone did it - likely on behalf of some big client - they'd never announce the results, since that pretty much kills their competitive advantage. Maybe if Morningstar did it...

 

To ask a side question: if we are talking about sabermetrics on investment managers, why not attempt it on companies or CEOs? Figure out a way to measure successability of a company (or CEO), invest in the great-but-unappreciated, ... , profit.  8) There's probably more data and payoff might be bigger. And then we wouldn't need 10 or 100 page threads about XXX and XXXX. Anyone game?  8)