Thursday, January 7, 2016

Maximum entropy versus coordination, part 10191

From Evonomics:
Marwell and Ames found that most subjects divided their tokens nearly equally between the public and private accounts. Economics students, by contrast, invested only 20 percent of their tokens in the public fund, on average.
It's another case where a lack of coordination yields the maximum entropy result (equal probabilities to find a token in each of the two accounts), while coordination (groupthink, arising via education in economics) yields a lower entropy result.

Since entropy is output, the economist economy would perform worse than the regular human economy ...

10 comments:

  1. I agree about the disutility of economics for economies. However, I am not sure about your last statement. IIUC, the best personal payout comes when everybody chooses the public option. Wouldn't Sunday School groupthink outperform maximum entropy?

    ReplyDelete
    Replies
    1. Hi Bill,

      I was only saying the economist solution would be outperformed by MaxEnt solution. It is in principle feasible, but computationally complex, to determine the absolute optimum strategy (for a given objective function) when considering thousands of interacting markets. That solution might not be MaxEnt, but MaxEnt should on average outperform any given strategy randomly selected from the set of all possible strategies.

      In the simple system (toy model) case where there only exist two options, you can easily solve the system given its (assumed) behavior.

      But since in real life we can't solve the market allocation problem with linear programming, we can't know if any given strategy is optimal.

      Delete
    2. "That solution might not be MaxEnt, but MaxEnt should on average outperform any given strategy randomly selected from the set of all possible strategies."

      This is reminiscent of John Bogle (founder of Vanguard funds) usual pitch for investing in broad index funds (such as S&P 500 funds) rather than actively managed funds... even if the management service is free. The fact that it's not free is a separate reason. If I recall correctly, he claimed that on average such an approach more or less reliably beats about 2/3 of the managed funds, year after year. It's been a while (> 15 years) since I read that though, so I don't know if the stats have changed.

      Delete
    3. That's a good example of a MaxEnt strategy ...

      Delete
    4. I like the idea of Maximum Entropy investing. A lot of advisors who specialize in passive investing recommend such "kitchen sink" approaches to investing: a smattering of national and international equity index funds, bond index funds, REIT funds, and even a sprinkling of commodities. These tend to have a better chance of hitting your investment goals "Efficient Frontier". Come to think of it, the existence of an efficient frontier for investing tends to reinforce the MaxEnt approach- the more investing "dimensions", the closer one comes to the hypersurface of a constant return, perhaps?

      Delete
    5. I know that Bogle simply attributes the strategy to the efficient markets hypothesis. Or at least he did back when the book of his I read first came out.

      Delete
    6. Get out the darts.

      Monkeys may not be able to outdo Shakespeare, but investment advisors are another matter. ;)

      Delete
  2. I have a feeling that there's a joke embedded in 10191... Googling it reveals a "Dune" reference. Hmm... I'm feeling like Bart again.

    ReplyDelete
    Replies
    1. Yes, there are occasional references to Dune, as well as to The Young Ones and Star Trek.

      In fact I think that is the second reference on the blog to the year 10,191.

      Delete

Comments are welcome. Please see the Moderation and comment policy.

Also, try to avoid the use of dollar signs as they interfere with my setup of mathjax. I left it set up that way because I think this is funny for an economics blog. You can use € or £ instead.

Note: Only a member of this blog may post a comment.