A random physicist takes on economics A personal view by Jason Smith A Random Physicist Takes on Economics by Jason Smith. Published by arandomphysicist. www.arandomphysicist.com © 2017 Jason Smith All rights reserved. No portion of this book may be reproduced in any form with permission from the publisher, except as permitted by U.S. copyright law. For permissions contact: [email protected] Cover by Jason Smith Single Blueberry The book cover contains a modified form of by Kevin Payravi, Wikimedia Commons. Creative Commons Copyright Attribution-ShareAlike 3.0. Contents Introduction The critique Physicists Random people Another dimension Advantage: E. coli Great Expectations Rigid like elastic SMDH The economic problem Economics versus sociology Are we not agents? Conclusions Acknowledgments Introduction I am a random physicist. If you were to select at random a physics PhD from the population of the United States, you'd probably end up with someone like me. I went to two large state schools (The University of Texas and the University of Washington). I am a male in field with an over-representation of males (like economics). Like most physicists, I didn't pursue an academic career. Instead I ended up working in the defense and aerospace sector. One way that I'm not typical is that I was a theoretical physicist (the ones writing the equations) as opposed to an experimental physicist (the ones running the experiments) — there are many more of the latter. My field was nuclear and particle physics — my thesis was titled “Quarks and antiquarks in nuclei” if that gives you a flavor. As I was finishing up that thesis and deciding against going into academia, I looked around at my other options. This was 2004 and the big thing physicists were doing in the real world was building financial models on Wall Street — if “real” is the appropriate adjective. I looked into becoming what they call a “quant”. I studied some quantitative finance. I had a couple interviews. I was offered a job at what I'll call Major Financial Company (MFC) helping to develop mortgage risk models in 2005, but (luckily, it turns out) I turned it down. However this is when my interest in economics started. It was also the heyday of the economics blog on the Internet. Instead, I took a research and development position in signal processing which for professional reasons will go by the name of Large Defense Contractor (LDC). At its most basic level signal processing the study of methods of extracting meaningful information from data. There's a new term out there for people who do this with enormous data sets called a “data scientist”. That's not what I do. I stick to smaller data sets. At this point in my life it looked like economics was in the rear-view mirror, but that changed after two events. First, the biggest financial crisis since the Great Depression hit in 2008. I was glad I didn't take that job; MFC went bankrupt. However, the economics blogs were inundated with all kinds of explanations, criticisms of economics, and claims of predictions. I asked myself: How could I tell if any particular economic theory was adding value? Second, LDC sponsored me for a government fellowship in 2011 which temporarily relocated me to Washington, DC. I was introduced to some research into the use of prediction markets for intelligence problems as part of the Intelligence Advanced Research Projects Activity (IARPA) Aggregative Contingent Estimation (ACE) program. I actually participated in the Decomposition-Based Elicitation and Aggregation (DAGGRE) project as a test participant, which was a forerunner of the now- defunct SciCast. Prediction markets are a kind of options market that lets you place a bet on an outcome. You have option contracts that, for example, would pay $1 if Hillary Clinton won the 2016 US Presidential election. If you thought that wasn’t going to happen, one of those contracts is worth close to zero for you. If you thought it was definitely going to happen, it would have been worth closer to $1 to you. With enough people participating in the prediction market, the value will be somewhere in between — say $0.70. If you thought Clinton had a better than 70% chance, you'd likely make money if you bought at that price (or higher). If you were holding a contract and you thought Clinton had a worse than 70% chance, you could have sold at the market rate (or lower). For intelligence prediction markets, those options contracts would be for events like whether Iran will test a nuclear weapon in the next year. The first time I had ever encountered prediction markets was during the 2004 presidential election (via the Iowa Electronic Markets). Research has shown that prediction markets do just about as well as polls. This actually makes them advantageous over polls in one sense; prediction markets are asynchronous polls like those dials given to focus groups as they watch debates these days. One weird property of prediction markets are sudden shifts in the market price — almost as if a wave of new conventional wisdom sweeps over the participants. While these markets seemed like a new form of polling in regards to elections (regardless of what their proponents' advocacy says), it could be problematic in the world of intelligence. We had already seen herding behavior in intelligence without prediction markets in the run-up to the Iraq War and the 2008 financial crisis wasn't exactly a big point in favor of markets in general. You could have made a killing shorting Clinton in 2016 in How could we online prediction markets. I had a serious question: tell if prediction markets are working? One of my proposed solutions to find out if prediction markets were working was to try to check the performance of a prediction market against a random person (in economics jargon, “agent”) model. Assign people random probabilities they believe an event will occur known only to them (a random valuation of an options contract as private information), and let them randomly encounter each other to trade. If the distribution of final options contract payouts was consistent with empirical data from prediction markets and was directly related to the initial random distribution of probabilities (valuations), then the real information is in that initial probability distribution. If a model with random agents could explain prediction market results, then prediction markets really aren't much more than polls of that initial probability distribution. The relevant information is in the initial probability distribution, not the series of transactions and prices. Of course that is just a test of the “information aggregation” mechanism — aggregating that private information (each agent's private event probabilities) into the overall probability distribution. There's another important piece of prediction markets:  the ability to reward (with money) high quality knowledge and punish (by losing money) low quality knowledge (or errors). If they are functioning correctly in theory, prediction markets act as a one-way valve pumping good knowledge in and bad knowledge out. That would take a bit more theoretical work to understand. Incidentally, around the same time both of these questions came up in my head, I discovered by a short preprint by physicists Peter Fielitz and Guenter Borchardt at arXiv.org (a repository for pre- publication physics and other technical papers) while looking for references about a signal processing topic called compressed sensing. Its title (at the time) was "Information transfer model of natural processes". It was in the general physics (gen-ph) section. The first line of the abstract tells the basic story: Information theory provides shortcuts which allow [one] to deal with complex systems. The paper is an attempt to answer a technical question about applying the so-called principle of maximum entropy expounded by Edwin Jaynes in 1957 in situations when you don't have things like the conservation of energy that you have in physics. Markets are complex systems — and there aren't conservation laws like conservation of energy in economics; I saw in this paper the potential to answer my questions about prediction markets and economic theory. I was temporarily relocated in Washington, DC, in a long distance relationship with my future wife, with few friends in the area, and lots of time outside of work on my hands so I dove in. I tried to work out how to couch the concepts of supply and demand as well as the price mechanism in this very simple theoretical framework. I later started thinking out loud on a blog. Peter Fielitz and Guenter Borchardt eventually updated their preprint with a new title (“A general concept of natural information equilibrium”) and to reference to “non-physical ... economical processes” along with my blog. Peter Fielitz got in touch with me and we have an ongoing discussion about concept of information equilibrium. I eventually put up a preprint myself in the quantitative finance economics (q- fin.EC) section of arXiv.org titled "Information equilibrium as an economic principle". I showed some of the early results to Jason Matheny at IARPA in a pitch for funding to see if it might be a useful way to produce metrics for prediction markets. He was very encouraging, but told me that I should put it in front of some academic economists. I tried to publish my preprint in an economics journal, but that resulted in immediate desk rejections. A desk rejection is where a paper is rejected before even going to peer review (based on the title and author's name and affiliation, I guess). I can't say I blame them — economics journals probably receive a choice selection of the collective id’s theories of money. There are a lot of crackpot theories about economics out there, and I try to remind myself of physicist Sean Carroll’s “The Alternative-Science Respectability Checklist” blog post from June of 2007 to keep from starting down the dark path. A version of the information equilibrium model paper from Fielitz and Borchardt was published in Physics Essays as “A generalized concept of information transfer”. However the material in my paper hasn't gone through peer review yet, so I'm not going to talk about it very much. It covers the technical application of the principles I am going to talk about here to economic systems — and actually comes up with many of the same results as standard economics but without a lot of the additional assumptions about rational behavior (to give one example). Luckily David Glasner, an economist at the Federal Trade Commission and in my opinion one of the best economics bloggers on the Internet, recognized some of my arguments as very similar to ones advanced by economist and Nobel laureate Gary Becker in 1962 in a paper titled “Irrational Behavior and Economic Theory”. It is from that starting point that I will take on economics. I know that lots of people out there are probably saying to A physicist? Oh, no! There are going to be a lot of themselves: equations!   Math is a tool, but I've been working with these topics for long enough that I can (hopefully) explain them in plain language. It used to be that economics was discussed with long dry prose, with occasional snappy political tracts (even a graphic novel The Road to Serfdom of sorts made from Friedrich Hayek's ). Even The General Theory of Employment, John Maynard Keynes' seminal Interest and Money The General Theory (hereafter, ) only has a few equations among its hundreds of pages. * * * I going to add some notes before we begin. If you are familiar with economics jargon, you can probably safely skip to the next chapter. I'm not going to be too formal about citations, but plan to give credit where it's due. I'll cite articles by authors, title and year but keep it within the narrative. Everything here is discoverable on the internet. I will refer to a “Nobel Memorial Prize in Economic Sciences” as a Nobel Prize and a winners as a Nobel laureate. It doesn't matter that Alfred Nobel didn't choose economics as a field to be honored; it matters that people today generally think of the prize as if he did. As it will occasionally be used as a sarcastic honorific, sticklers for detail should be satisfied. Economics is divided into two broad fields: microeconomics and macroeconomics (maybe three if you consider growth economics). Economics has basically become the general study of systems that involve money, although it has branched out to include cases where people make strategic decisions of any kind (via game theory, and Macro- the work of the aforementioned Gary Becker). is a prefix