Saturday, March 12, 2016

The unreasonable effectiveness of information transfer

The past couple posts (here and here) tout the performance of the information transfer framework, but I would like to clarify something. I'm somewhat surprised by its effectiveness. It's a bit like looking at a bunch of data about gasses under pressure and finding they all essentially follow the ideal gas law with most of the deviations being just noise.

As a framework, the IT model tries to encapsulate the successes of economics and macroeconomics -- mostly supply and demand and marginalism. Writing down an IT model is a shorthand for these successes. And if you use a supply and demand or marginal argument, you can write your argument as an IT model.

The thing is the IT framework is a first order theory -- it almost certainly is the limit of a more accurate theory (or large scale agent-based models). But to a surprising degree, most quarter to quarter changes in macro aggregates like NGDP or core CPI seem to be just noise. Sure, there are recessions, but those are sporadic not continuous. Seasonal effects are one of the only other observable macro fluctuations, and that derives from human reactions to the orbit of the planet.

Where are the second order effects?

I guess I should rephrase that: where are the other successes of macroeconomics? If there were other successes out there that weren't at their core marginal or supply and demand arguments (this includes most diagrammatic and marginal utility arguments), then it is likely the IT framework would fall short in explaining them.

You might think it should also surprising that no one else has stumbled onto this relatively simple theory. I don't think so -- from my interactions with the econoblogosphere I've realized that a lot of people have a hard time letting human decisions average out most of the time. Only when we least act like individuals does human behavior seem to enter the picture ... and it's nearly always a bad sign. It's when our plans fall apart that we become (panicked) humans making (bad) economic decisions rather than an abstract collection of  complex entities with aggregate properties.

That simplicity also seems to work against acceptance. It even keeps me from, say, being like Scott Sumner with NGDP level targeting. As I mentioned in the first paragraph, I'm still surprised it works as well as it does. I would love to find some serious limitations. I have two: 1) the quantity theory of labor doesn't seem to work for the UK and 2) the interest rate model when countries hold large quantities of foreign currency denominated debt (e.g. Australia). But in the first case, the model is already a massive simplification (I don't even really take it seriously -- I tend to use it as a reductio ad absurdum) and in the second, it has a plausible explanation as "interest rate importation" (at the link).

Update 22 Nov 2016: Solved 1) [here], and have a better understanding of 2) [here]. Now I need new limitations.

So when commenters or other bloggers say to me that the theory doesn't take into account property X or effect Y or say there's more to the story, I tend to be in agreement. However: there also doesn't seem to be very much else in the data!


  1. Don't sell yourself short, Jason. Have a look at this blog- there are 100s of posts, put there by someone with a high level of sophistication for mathematics and economics. Probably not too many other people in the world who could have done what you did.

  2. I love these "deep thoughts" posts.


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.