Monday, February 29, 2016

Draft slides for BPE 2016 talk in DC

Noah Smith likes slides (as do some of the commenters and twitterers (?) ...). A downloadable pdf version is [here, pdf] (let me know if my google drive settings aren't appropriate); here are some (DRAFT) slide images. Here you go:





























References:

Smith, Jason. Information equilibrium as an economic principle. arXiv:1510.02435 [q-fin.EC]
** and references therein

Lee Smolin Time and symmetry in models of economic markets. arXiv:0902.4274 [q-fin.GN]

Becker, Gary S. Irrational Behavior and Economic Theory. Journal of Political Economy Vol. 70, No. 1 (Feb., 1962), pp. 1-13

Chen, M. Keith and Lakshminarayanan, Venkat and Santos, Laurie, The Evolution of Our Preferences: Evidence from Capuchin Monkey Trading Behavior (June 2005). Cowles Foundation Discussion Paper No. 1524. Available at SSRN: http://ssrn.com/abstract=675503

Fielitz, Peter and Borchardt, Guenter. A general concept of natural information equilibrium: from the ideal gas law to the K-Trumpler effect. arXiv:0905.0610 [physics.gen-ph]

Fisher, Irving. Mathematical Investigations in the Theory of Value and Prices. (1892).

Shannon, Claude E. (July 1948). A Mathematical Theory of Communication. Bell System Technical Journal 27 (3): 379–423.

Okun, Arthur M. (1962). Potential GNP, its measurement and significance.

35 comments:

  1. Personally, I do like slides. There's nothing like a good set of slides when it comes to getting a quick understanding of a model. Papers are usually superfluous while slides, if done properly, can give just the right amount of explanation without seeming overbearing.

    Needless to say, more slides would be great; perhaps for each of the ITM versions of specific orthodox models -- e.g., Solow-Swan, IS-LM, AS/AD, etc.

    ReplyDelete
    Replies
    1. In my experience, the quality of writing has more to do with my ability to finish a paper than the length of that paper. In each case, your working paper is better than most.

      Delete
  2. I agree with John. However slides are way, way better if they are in a PDF, not images in HTML.

    ReplyDelete
    Replies
    1. I concur. It'd also be great if there was one location where all the relevant slides were made available.

      Delete
  3. I hope that your audience is open to the idea of macroeconomics without microfoundations. :)

    ReplyDelete
  4. Hey Jason, I haven't gone through them in detail yet, but I think so far they've turned out really nice!

    ReplyDelete
  5. Its interesting work and I wanted to see some of your references.

    ReplyDelete
    Replies
    1. Cheers, thanks.

      There are references in the original draft for the slides here:

      http://informationtransfereconomics.blogspot.com/2016/01/draft-paper-for-talk-this-summer.html

      But I added them above and they will be in the conference proceedings paper.

      Delete
  6. One thing I'm missing are some of your great plots, like that of P vs time over several countries (model vs data). You do have some (Okun's law and interest rates), but I'd put a few more examples of empirical success in there. Maybe you can buzz through those pretty fast... just examples of how this framework can work and be successful. Just a thought.

    ReplyDelete
    Replies
    1. There's insufficient time to present partition function approach that makes those graphs make sense.

      Actually, many of those other empirical successes require a bit more background than is currently in the presentation ... e.g. changing k, NGDP-M0 path.

      I chose those two examples because Okun's law is fairly well established and the interest rate picture is quite powerful. And to leading order they both can be represented with a simple IE model.

      Delete
    2. Sure, I agree, the partition function is too much... but "changing k" is pretty easy, isn't it?

      Delete
    3. Not really. If there aren't any constraints on how k changes, you could get anything you want ... and the way that I know how to explain slowly changing k that is the most rigorous is with the partition function approach. Picking a model for k is pretty ad hoc.

      Delete
    4. I was thinking of this post, where κ=logm/logd seemed to follow directly from κ≡logs/logd.

      But you have a point: you can only do so much, and that might be a bit of a distraction.

      Delete
    5. I imagine that post might become the wrong theory equivalent of these airfoil arguments:

      https://www.grc.nasa.gov/www/k-12/airplane/wrong1.html

      Not that I think there is anything currently wrong with it as an ansatze.

      :)

      Delete
    6. This comment has been removed by the author.

      Delete
    7. Ok, so you're saying that it was a good educated guess that happened to match the data pretty well, but it's now been superseded (by the partition function, etc).

      Delete
    8. I wouldn't say superseded so much as I feel like that's a more satisfying explanation that would hold up under assault better. Ad hoc has "it works" in its defense.

      There are two ways to approach the 1/2 in the kinetic energy formula. In one, in purely non-relativistic mechanics, it is kind of a circular argument -- by assuming p = mv. In the other, it follows from a coefficient in the Taylor expansion of the Lorentz factor in relativistic mechanics.

      Neither is wrong per se. But the second one is a better theoretical argument.

      Delete
  7. On slide 6 ("Maximum Entropy?"), I'm curious what your words will be there at the top. That's where you'd presumably tie together what you just presented in the previous slides ("Aggregate economic forces like supply and demand follow from properties of the state space..."), to what you'll be presenting over the next set of slides ("natural information equilibrium").

    Maybe it's just me, but I always feel like we never clearly or explicitly tie the thread of logic back to those important "properties of the state space" which was the takeaway message in the 1st five slides.

    For example, right after slide 13 or perhaps slide 22, I'd love to see a bullet asking:

    o But how does all this relate to the state space (opportunity set) we mentioned here on slide 5? Here's what...

    ReplyDelete
    Replies
    1. I'm not quite sure I understand.

      "o But how does all this relate to the state space (opportunity set) we mentioned here on slide 5? Here's what..."

      My answer would be to define Shannon information entropy -- which is exactly what I do.

      Delete
    2. I could probably add "... over the opportunity set" to the distribution labels in the communication channel diagrams.

      Delete
    3. This comment has been removed by the author.

      Delete
    4. labels... yeah, I like that!

      Delete
  8. Jason,

    In slide 4 you introduce the maximum entropy point.

    Can you explain maximum entropy and then the maximum entropy point.

    ReplyDelete
    Replies
    1. Maximum entropy is generally assuming the least informative prior distribution. In this case, a uniform distribution.

      The maximum entropy point is just the expected value of the state position operator with this distribution.

      But in simpler terms, it is just the average of all the points inside the triangle.

      Delete
  9. Jason,

    Could you also explain consumption state.

    Henry

    ReplyDelete
    Replies
    1. An agent with consumption c = p1 c1 + p2 c2 < B where B is the budget constraint is occupying the consumption state (c1, c2).

      Delete
    2. Erratum: Leave off the prices p1 and p2.

      Delete
  10. Jason,

    In slide 5 you mention the properties of state space.

    What are these properties?

    In slide 6 you mention self information - could you define this.

    Henry

    ReplyDelete
    Replies
    1. The properties of the state space include (but are not limited to): its dimension, the number of states, and its shape.

      In the budget constraint example, the budget constraint and the dimension (number of goods) d = 2 are properties of the state space.

      That is the definition of self information in slide 6; the self information I(p) = - log p.

      Delete
  11. "Maximum entropy is generally assuming the least informative prior distribution. In this case, a uniform distribution."

    So this is the state of maximum uncertainty and minimum information?

    ReplyDelete
    Replies
    1. It's actually maximum information entropy distribution -- given the prior knowledge of a budget constraint.

      It's not an absolute maximum or minimum of anything, however. A Pareto distribution is the maximum information entropy distribution over x that constrains E[log x] (and has the minimum prior information).

      If the ensemble of agents is consistent with being drawn from the maximum entropy distribution (given constraints), then that ensemble is in a maximum entropy state.

      Delete
  12. Jason what is your view on "bounded rationality"?

    https://en.wikipedia.org/wiki/Bounded_rationality

    ReplyDelete
    Replies
    1. Sorry Random -- I missed this comment. Better late than never, I guess.

      My personal opinion is that bounded rationality is an attempt to reconcile the fact that agents aren't rational with the basic econ theory that they are rational.

      In the information equilibrium approach, information equilibrium is kind of like rationality and non-ideal information transfer is like bounded rationality (agents don't fully optimize). So the general idea is incorporated -- i.e. I don't think it is wrong. In a sense, Info Eq approach incorporates 'bounded rationality' at a fundamental level (level of model definition) rather than as an "extra effect" modifying econ rationality.

      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.