This "working paper" is now four years old

... it's more of a "working thesis" now.

Previous birthday celebrations focused on the most popular posts or prediction successes. I thought that for this birthday, we could celebrate some of my favorite posts. In no particular order:

[1] Balanced growth, maximum entropy, and partition functions

This post provides the best description of the motivation that goes into treating ensembles of markets (which relates to other work on statistical equilibrium) and changing information transfer indices. It is the theoretical basis for two posts below: [3] and [5].

[2] Dynamic equilibrium (presentation)

This represents a concise summary of the direction that I've found most interesting recently. Coupled with the basic supply and demand/production function unification [4], the ensemble approach in [1], and the matching theory in [6] it forms the core tools needed to understand a sizable chunk of economics.

[3] Stocks and k-states

An application of [1] to the stock market that explains factor models and their issues.

[4] It's a production input. No, it's a market good. Relax, it's both.

A really simple way to understand the scope of supply and demand as well as its connection to growth (i.e. production functions).

[5] An ensemble of labor markets

An application of [1] where the primary factor of production is labor. I originally had looked at the application of the ideas of [1] where the primary factor of production is money (but I didn't put it that way).

[6] Matching theory and employment in information equilibrium

This post connects dynamic equilibrium to mainstream matching models.

[7a] Principal component analysis of jobs data

[7b] Using PCA to remove cyclical effects

A pair of posts that uses principal component analysis and dynamic equilibrium to get a picture of employment by industry/sector.

[8a] Consumption, income, and the PIH

[8b] Curve fitting and relevant complexity

A pair of posts that offer a simple 1st-order explanation of income and consumption data. In the second one I talk about methodology. While I enjoy writing the posts on methodology, I'm not really as proud of or interested in it. I think a good analogy is with movies. I like being a director (making movies), but sometimes I review other movies. The posts on methodology all tend to be "bad reviews" which are fun to write (and read). They're just not fulfilling as doing interesting original work. 

[9a] The basic asset pricing equation as a maximum entropy condition

[9b] The Euler equation as a maximum entropy condition

In this pair of posts I derive the asset pricing equation and the Euler equation from a maximum entropy argument. They both show that the utility description is a bit superfluous, and since they require maximum entropy distributions imply that correlations in agent choices will cause them both to fail. This work is utilized in [11] below.

[10] The Kaldor facts

I show that you can capture the "Kaldor facts" about growth as information equilibrium relationships. I also show that the data is better captured by different relationships.

[11] DSGE, part V (summary)

In order to show that information equilibrium isn't too far afield from mainstream economics, I derived the simple New Keynesian three equation DSGE model in this series using maximum entropy and information equilibrium ‒ or at least something close to it with meaningful differences due the information equilibrium approach.

[12a] Forecasting: IT versus DSGE

[12b] Forecasting: IT versus all comers

Here, the IT model performs much better than far more complex models and policy heuristics using the same tests that economists used.

[13] Causal entropic forces as economic forces

There is a very interesting connection between information equilibrium and causal entropy (which was referred to as "a new equation for intelligence" in the author of the paper's TED talk).

[14] Generative adversarial networks and information equilibrium

This post itself is terrible and consists mainly of notes. However, it might provide an explicit example of the algorithm that the market represents and has a direct connection to information equilibrium. Instead of a rational agent argument where money represents utility, we would transition to an argument where prices represent flows of information.

[15] Is the market intelligent?

My "take down" of prediction markets and the idea that markets perform an information aggregating function rather than just solving an allocation problem. This was actually a summary of the original work on prediction markets (see e.g. here) that never went anywhere for me in my regular job (no one wanted to fund it), but did lead to this blog. Basically, this post was part of my notes when I put up these three posts ‒ the original application was going to be metrics for prediction markets.

...

Update 24 April 2017: As I mentioned here, writing this post inspired this graphic that captures the interconnections described above: