Pursuit of Truthiness

Steve Levitt has said that he became a successful economist by coming up with a lot of ideas and quickly discarding the 98% that are bad.  Assuming 98% of my ideas are also bad, I should make sure to have at least 50 ideas to sort through to have a chance of a good one.  Here’s where I am now:

Insurer Mandate Research Program: How do mandates affect insurance prices and quantities? Knowing this, we can determine the price elasticity and the health benefits of health insurance.  Examine specific mandates to determine the elasticities and health benefits of specific procedures.  Determine optimal composition of the “essential health benefits package”.

Rational Lotteries: Determine how often there is a positive expected monetary value to playing the lottery after implicit marginal tax rates are taken into account.  Compare behavior with predictions of prospect theory.

Value of a Statistical Life: Calculate using insurance fraud

Fraud and the Business Cycle: Empirically test the hypothesis that fraud is a cause / leading indicator of recessions.

Elections and Inequality: Expand on my earlier paper about how inequality affects voting.

Size of Nations: What is the optimal size of a nation (empirically, which size maximizes various good outcomes like gdp), and is it decreasing over time as the world gets more peaceful?

How to Make Money: One of the many ironies of economics is that Homo Economicus would keep market anomalies to himself, while economists regularly publish their discoveries.  I will write a blog post of general ideas, but the low-hanging fruit here is perpetually disappearing.

Testing Hansonianism: Robin Hanson is always noting ways in which people care more about buying medicine than health.  One way to empirically investigate this is to see how much demand for hospitals is driven by high subjective evaluations verses objectively good health outcomes using different variations of hospital report cards.

CEO Pay:  Expand on a previous paper of mine that cast doubt on the highly-cited Gabaix and Landier  paper.

Alternative Currencies:  How much are money substitutes expanded in response to central bank tightness (as measured by some kind of Taylor rule).  Theory question- why do some money substitutes (alternative currencies) expand in response to tightness while others (credit) are reduced.

Drug Laws: What is the effect on poverty and other economic outcomes.

I Just Ran 3 Billion Regressions: Update Sala-i-Martin 2001 using another decade of data, and using a modern supercomputer to run the specification that Sala-i-Martin proposed but was unable to do with the computers of the time.

Data-Driven Investigation: All the above ideas basically start with an idea and look to data to test it.  But nowadays there are huge amounts of data being generated that have yet to really be investigated. So an alternative strategy is to be on the lookout for interesting-looking, unexamined datasets.

….not quite to fifty yet.  Perhaps I have lucked out and one of my ideas is actually decent.

Michael Lewis’ book The Big Short: Inside the Doomsday Machine has been a popular bestseller even as it received unanimously good reviews from economists- the harshest review I saw was Eric Falkenstein’s “entertaining but doesn’t get to the heart of the issues” take.  So of course a part of me was secretly hoping the book’s arguments would be wrong or oversimplified, and I could explain why and feel superior to everyone else.  But I am afraid it was simply an excellent book.  It is very entertaining, gives a good explanaiton of the crisis (though not mutually exlusive of other potential explainations), and explains complicated financial ideas in a way that most people should be able to grasp.

One of the big themes of the book (though it is not explicitly stated this way) is that the Efficient Markets Hypothesis often fails to apply to U.S. financial markets.  This is often because its assumptions are not met, particularly in bond markets where there are few buyers and sellers and little information is publicly available.  But sometimes large, liquid markets seem to take a surprisingly long time to incorporate publicly available information.  For instance, one small hedge fund claimed that they were the only people really examining the financial statements released by subprime mortgage companies in early 1997, and that it took the market months to look at the same data and realize the firms would soon be bankrupt; the fund’s accountant said “it made me feel good that there was such inefficiency to this market… if the market catches on to everything, I probably have the wrong job.”

Another small hedge fund, Scion Capital, made huge returns when their “decision-making apparatus consisted of one guy [Mike Burry] in a room… poring over publicly available information”.  Lewis tells the entertaining and inspiring stories of how three small, wildly successful hedge funds got started.  My favorite was Cornwall Capital, “two guys in a garage in Berkeley with $110,000 in a checking account”.

One small part of the book gave me an idea for a great economics paper (which has probably been written already).  There is now a great debate about whether “bubbles” can be meaningfully defined, discovered in advance and deflated.  Many argue that once you have a definition of a “bubble” that is meaningful and testable, you will not be able to find any.  But in the book, Mike Burry puts forward the thesis that

“It is ludicrous to believe that asset bubbles can only be recognized in hindsight.  There are specific identifiers that are entirely recognizable during the bubble’s inflation.  One hallmark of mania is the rapid increase in the incidence and complexity of fraud… the FBI reports mortgage-related fraud is up fivefold since 2000.”

I would love to get fraud and financial data by sector and look for relationships.  I am sure others have done this already, but if not it would be cool to test the theory, and if it holds up start making money on it.  (of course, making money by killing the Efficient Markets Hypothesis only serves to make it stronger going forward… it is like Obi-Wan Kenobi.)  Another potential way to make money described in the book is to buy options (which at least as of 2007 were usually priced by assuming a normal distribution of potential prices) on stocks which should have a bi-modal distribution of future prices; it made me wonder if this still works.

As far as an explanation of the financial crisis, the book has two main explanations.  One is that the bond market is opaque and oligopolistic.  “The presence of millions of small investors had politicized the stock market.  It had been legislated and regulated to at least seem fair.  The bond market, because it consisted mainly of big institutional investors, experienced no similarly populist political pressure… bond traders could exploit inside information without worrying that the would be caught… in the bond market it was still possible to make huge sums of money from the fear, and the ignorance, of customers.”

The second, and related explanation is that the bond-rating agencies were very bad at their jobs, and only some people realized that.  The Wall Street banks making securities could exploit the flaws in the rating agencies models to get their product rated too highly.  Meanwhile, some investors stupidly trust the rating agencies and buy bonds at high prices assuming their ratings to be correct.  There were many flaws in the rating agency model, perhaps the most surprising to me is that “both Moody’s and S&P favored floating-rate mortgages with low teaser rates over fixed-rate ones”.  Many people focus on the potential corruprtion at the rating agencies and Lewis does bring that up, but his main focus is on stupidity.  “You know how when you walk into a post office you realize there is such a difference between a government employee and other people.  The ratings agency people were all like government employees.  They’re underpaid.  The smartest ones leave for Wall Street firms so they can help manipulate the companies they used to work for.”

In the epilogue Lewis turns to a third explanation for the crisis: a faulty incentive structure for employees in financial firms.  ”What are the odds people will make smart decisions about money if they don’t need to make smart decisions about money- if they can get rich making dumb decisions?”

Before this post gets any longer, I’ll just say: read the book, it tells some great stories.  Also, check out Lewis’ article on the Greek debt crisis.  He manages to interview a lot of important Greeks.  The article is fascinating but somewhat depressing as it makes me wonder how Greece can possibly fix its government and social order; I can understand why Paul Romer proposes giving the EU a bigger role in governing Greece.

We discussed public goods in Micro Theory yesterday, and it was asserted that they are always consumed equally by all parties, whether they are the two consumers in a Kolm Triangle or everyone in a nation or world.  Our professor usually points out assumptions that seem very strong or unrealistic, but he seemed to endorse this one without reservation.

Myself, I don’t see how this comes close to holding for any definition of “public good”, “consumed” or “equally”.  Consider national defense, the archetypical public good.  I don’t think anyone would contend that all Americans pay for national defense equally or derive the same utility from its provision.  The assumption is that we all receive the same “amount” of national defense, though this is admittedly hard to measure (perhaps we get two Major Theatre Wars worth).

In reality though, some members of a nation receive more military protection than others, and often the scarce resources of the military are used to protect some people at the expense of others.  Historically, armies were used to protect the people near a frontier, often from threats that had no chance of harming those in the interior.  The US used forts and raids to protect frontier Americans from Native Americans who posed no threat to those in major east coast cities.  Today our War on Terror is supposed to prevent terrorism, and so benefit people in New York City and Washington more than those in rural Kansas or Montana.  In a two-front war, the military must allocate resources between fronts.  For instance, in WWII the US prioritized the European theatre over the Pacific, so that east coast citizens received more defense resources than those in the west.  Today US forces arrayed against North Korea provide much protection to South Koreans but very little to the majority of Americans one would think to be the relevant “public”.

Other kinds of “public” goods are even more obviously consumed at different levels.  Those without cars don’t “consume” public roads as much as those with them, those without televisions don’t consume public TV, those without boats don’t get as much out of lighthouses.  Even knowledge, especially when broken up into various subdomains, is not consumed equally and in some cases perhaps cannot be consumed directly by all people (if they are not capable of understanding it and it has no “practical” uses).

This seems to be an open and shut case of a super-unrealistic assumption.  So the next questions for an economist are, how much does the math and the theory of public goods rely on it?  Can we generate results if we relax it, and if so what are they?  Have people done this already?  These are the questions I will be thinking about in our further study of public goods.

Using the ideas of all major schools of the philosophy of science, determine the scientific status of economics and its subfields.  If economics is not scientific, how could it become so and should it become so.

Just kidding, I would like to get a job.

Find an obscure dataset that has been used only by historians who don’t know statistics.  See what can be learned from using statistical methods and economic reasoning.  The problem being that economists will think most history-type projects irrelevant, while historians may ignore outsiders on general principle and aren’t on your dissertation committee in an case.  The other problem being that actually finding a dataset that 1)hasn’t been really examined; and 2)has the potential to shed much light on an area; is likely to be very time consuming.

It has, however, been done successfully as a dissertation.

Taking everything that psychologists and behavioral economists have learned about how utility actually works, and build a relatively realistic utility function.  Then plug this utility into several classical theoretical results and see how they are affected.  Of course, economists refrain from using realistic utility functions because of mathematical intractability as much as lack of knowledge about what they look like, so this may be a doomed project.  Hard to know for sure without losing a few months of life to it first.

Take some bigass panel datasets like MEPS and AHA.  Test every theory about rising U.S. healthcare spending.  If you take long enough to write it you can see the effects of the new regulatory regime and subsidies.

If you want to do something really useful, build a dataset about the regulatory regime in each state and use this to analyze the effects of state-level policies.  This would be an extremely useful dataset to many researchers, but could also take a very, very long time sifting through state legal codes to build.