Mailbag column!

So there’s a reason people do these. They’re so easy to write. I think I’ll do more of these in the future.

I would also like to preface this column by asking anyone who enjoys this to consider getting a paid subscription. I hope to keep my blog unpaywalled, and simply rely upon everyone to pay what they believe to be their valuation. Now, on to the questions!

O.H. Murphy asks the following four questions: 

Have you thought about the relationship between video/board game design and mechanism design in economics? Social deduction games especially seem almost like reverse revelation mechanisms.

No, I have never thought about this. This is an interesting way of thinking about it, though. What I wonder is if, within any game, there is any way to make it optimal to reveal someone's true state. Imagine you’re playing werewolf or mafia, where you possess private information about what you are, and are trying to guess who the bad guys are. Is there a way to compel everyone to reveal true preferences? Werewolf and mafia, by the way, are the same game with slightly different names. Two players are secretly chosen as werewolves, and decide on a villager to kill during the night; all players vote on who to kill during the day. If werewolves care about them in particular winning, and not their team winning, you could say to someone whom you believe to be a werewolf, “choose to kill your teammate — if they are a werewolf, you win with some small probability — if not a werewolf, we kill you”. If you’re sure, at least, that that one person is a werewolf, it should improve your winning chances. I may be misunderstanding the question, but it’s just where my mind goes.

What do you think are the most interesting results at the intersection of Computer Science and Economics?

A lot of stuff on algorithms and auction efficiency in close-to-optimal auctions is basically the same thing as theoretical computer science (I think). Tim Roughgarden’s book “20 Lectures in Algorithmic Game Theory” is absolutely fantastic here, extremely clean, clear, and precise. My ability to answer what is at the intersection of computer science and economics is limited by my almost total ignorance of computer science; I know rudimentary python, and little more.

If AI counts, then AI is going to be a massive deal in data collection. When people use satellite datasets, it’s all machine learning — you need to tell it what you’re looking for, and then it can characterize terabytes of data. Sentiment analysis lets you do what would have taken an army of research assistants months to compile. Optical character recognition is enabling the use of previously hard to work with archives. 

As someone not intending to do a masters/PhD in Economics, I'd appreciate hearing what you think are the biggest topics that are missing from undergraduate courses (i.e. that I would have missed)?

I actually do not know! My problem is that I do not remember where I really learned things — I generally don’t learn all that much from the classes directly, they are mainly there to point me to the frontiers of the literature. What’s worse, I didn’t even take introductory economics in college, but took it as an AP class in high school. From what I gather from tutoring AP economics, though (which may not overlap with what is taught in college) is that there may not be enough on growth, and too much on some very Keynesian business cycle management. All of it is good, but there is some question of emphasis. I cannot really pontificate, though, it having been so long since I have had any familiarity. Check back after I teach an introductory class.

What do you think of the quality of economics as presented in fiction? Can you think of any particularly good or bad examples?

It’s generally really bad, though it is difficult to think of examples. Most fiction is written by innumerate people with no sense of scale whatsoever. So, most of them are very hostile to business and production. 

Did you ever read Hoot? Maturing is realizing that yes, it is just a couple of owls, and people being able to eat pancakes is good.

Sleeping Aristocrat asks: Why does the labor market fail to clear the way other markets do if not for active monetary policy?

Uhhhhhhh … wage stickiness. Don’t look too closely at it.

In all seriousness, wages have two problems. First, the labor market tends to be the most regulated market out there, and second, wages get charged with a moral significance that other prices don’t. Your wages influence your dignity and worth as a person. So, people are probably less willing to take a cut. In addition, because you build up a lot of firm specific capital, refusing to cooperate with wage cuts might be the optimal bargaining strategy. 

I think the main reason is just how profoundly regulated labor markets are, though. Could you imagine unemployment insurance for the product market? “Oh, you didn’t sell your product cause your asking price was too high, guess we’ll give you money for that?” Is that not a crazy idea in any other context than the labor markets. Besides that, the whole reason you have frictional unemployment is just cause it’s really expensive to get rid of bad employees – if firing were genuinely cheap and easy, you’d have a lower full employment unemployment rate. So, I think most of our inflexibility is created, not natural.

Of course, while I can cite massive wage regulations for the Great Depression (this Lee Ohanian paper from 2009 makes a convincing case that labor market regulations were indeed responsible for the Great Depression – the unregulated sectors, like agriculture, saw next to no unemployment, while the ones which saw extensive interventions by the government in an effort to keep wages up saw extensive unemployment) I cannot give you the same answer for the Long Depression of the 1870s! It seems not implausible to me that readjusting wages and prices simply took longer back then, as compared to now – information was far costlier. Thus, a recession as a consequence of a fall in aggregate demand is explicable as due to price stickiness, not merely wage stickiness.

O.H. Murphy asks again: Why do companies with rating systems (yelp, Uber, etc) often not make stars, etc limited in order to make them more credible signals?

The question being – why do rating systems converge to you either get a perfect score, or you don’t? Doesn’t this make it impossible to indicate when quality is exceptional? I think the answer is pretty easy – it converges to all perfect scores, for merely adequate services, if people do not care about exceptional experiences. When people do, the rating system becomes meaningful deviating up and down.

So, in restaurants, Yelp has meaningful star ratings, because you care about good experiences. Uber doesn’t, because people don’t care about exceptional service. They only care about avoiding the worst outcomes.

Sam Harsimony asks: If a monetary authority had to commit to a certain algorithm for its decision making, which algorithm is best?

What weird ideas have you heard about for dealing with patents and intellectual property?

I really don’t want them to commit to an algorithm. I want them to commit to certain goals – specifically nominal GDP targeting. Any particular algorithm would need to have its parameters regularly updated. Think about it like this – pure money supply targeting, ala Friedman or other early monetarists, depends upon a given and constant velocity of money. If the velocity of money is changing, then your algorithm will lead to unwanted results.

You should check out my recent article on Kremer patent auctions.

Arya Biss asks: What is the topic of the PHD you are working on?

I have no idea. I’m just gonna follow the papers, and probably cram three of them with at least a  tenuous connection to each other together. I’m not sure how attached I am to economic history. I feel like I have a strong tendency to cover many things, rather than dive deeply into only one topic. I really like blogging for this reason – I can cover the breadth of my knowledge, rather than spend several months compiling a better dataset.