Forbes article about a new monkey experiment

On the Forbes website today there is an article about some new research about to be published by Research Affiliates.  It involved making up 100 random portfolios of 30 shares from the top 1000 shares every year from 1964-2011. It then compared the performance of these portfolios with the index. It claims that 98 our of 100 of these portfolios beat the market.

Although I have seen consistent out-performance of monkeys vs the index and professionals, the size of this out-performance seems slightly incredulous. I fear that they may not not have used survivorship bias free back data for this simulation. I await the full report from Research Affiliates to see.

4 thoughts on “Forbes article about a new monkey experiment

  1. Following up on Pete’s blog mentioning an article in Fortune about research by Rob Arnott at Research Affiliates.

    Here is a link to the paper

    The Surprising “Alpha” from Malkiel’s Monkey and Upside-down Strategies1 Rob Arnott2
    Research Affiliates, LLC
    Jason Hsu3
    Research Affiliates, LLC UCLA Anderson Business School
    Vitali Kalesnik4 Research Affiliates, LLC SDSU
    Phil Tindall5 Towers Watson Limited
    First Draft: October 31, 2011 Current Draft October 10, 2012
    The latest index literature is bursting with new innovations based on quantitative strategies that are predicated on sensible investment beliefs. Empirical studies confirm that these strategies do indeed deliver economically large and statistically significant excess returns over the cap- weighted market benchmarks in nearly all geographical and country studies. To the casual observer, it will be shocking to learn that inverting the portfolio construction algorithms does not reverse the alphas. Embarrassingly, the inverted strategies often outperform the originals.
    This paradoxical result is driven by the phenomenon that seemingly unrelated and non-value- based strategies and their inverse strategies often have unintended and almost unavoidable value and small-cap tilts. Even Burt Malkiel’s blind-folded monkey, throwing darts at the Wall Street Journal, would produce a portfolio strategy with a significant value and small-cap bias that would have outperformed historically. The value and small tilts stem from the fact that these new weighting schemes sever the link between a company’s share price/capitalization and its weight in the portfolio. Generally, an investment thesis where price does not drive the weight in the portfolio will have a value tilt and an investment thesis where company size/capitalization does not drive portfolio weights will have a small-cap tilt. As a result, these strategies produce outperformance against the cap-weighted benchmark due to the often unintended value and small cap tilts and independent of the investment philosophies that drive the product design.

    • Looking at the draft paper, the exact method for the monkey strategy was:
      “we simulate a dart-throwing monkey, picking a random 30-stock portfolio out of the top 1,000 largest stocks by market capitalization once a year. We then equal weight the random stock selections to form the portfolio. We repeat the exercise 100 times, and examine both the individual 47-year trials and the average of the trials.”
      They use CRSP/CompuStat merged database (for their US analysis).

      It is not clear from the paper, but I assume for each year’s calculation that only companies that existed at both the beginning and end of the year are included. It will therefore exclude those that get taken over and those that exit the top 1000 database either by going bankrupt or lower capitalisation. If this is the case, there is a lot of academic evidence that this will have affected the results they found due to survivorship bias. It could account for a sizable part of the effects they proclaim for this and the other strategies discussed in the paper.
      Don’t get me wrong, I still think there is merit in a ‘monkey strategy’. Genuine monkey tests (not historically reworked ones like this) have shown it. However the size of the effect is probably not as large as that proclaimed by this paper.

    • Good point. I’d like to think though that if they did, they also used a Total Return index to compare against. I doubt they’d make that error, but…

Add Comment Register

Leave a Reply