Right in the Short Run

the 5 best models of drug  (incl. alcohol) consumption?

Earlier this year Andrew Gelman and co. came up with a new argument solving the voting paradox, based on social preferences (the paradox of voting: if you are rational, you will realize that the probability of affecting the outcome of a vote is negligible, while the costs are often considerable. Yet still people vote regularly).

Social preference is basically the idea that individuals incorporate other people’s utility into their considerations.

Gelman argues that if individuals take into consideration the impact of a vote’s outcome on other individual’s in society, as well as themselves, voting may indeed be rational.

A recent experimental paper by Fisman, Kariv and Markovits shows that some people do indeed have social preferences.

Voters, congratulations.

For all the crap Steve Levitt gets about his abortion paper due to coding errors and what not, I feel bad for him on this count.  Kiki Pop-Eleches did similar work in Romania, utilizing a “natural experiment” in which the Romanian dictator Nicolae Ceausescu banned abortion in 1966.  In any event, he finds that:

The restrictive policy disproportionately affected disadvantaged women and created telltale signs of the “unwantedness” effect - a rise in infant mortality and criminal behavior later in life.

The paper can be found here.  Essentially, Kiki found similar results.  People seem to forget this when smashing Levitt. Sure, a few coding mistakes were made, and that certainly isn’t a good thing.  But that certainly does not take away from the ingenuity of his original paper, which seems to be corroborated by other independent studies such as Kiki’s.

Things to read up on:

1. Relative utility functions. Let’s go find the literature on u_1(x_1) = f_1(x_1,x_-1) where x_-1 = vector of others’ wealth or consumption vectors.

2. Addiction literature. Under what conditions does it make sense to describe drug consumption in the expected utility framework?

3. Causal inference tools employed in labor economics. See this text by Angrist (MIT) and Krueger (Princeton).

Roland Fryer strikes again, this time with Michael Greenstone. The paper entitled “The Causes and Consequences of Attending Historically Black Colleges and Universities” as the following abstract:

Until the 1960s, Historically Black Colleges and Universities (HBCUs) were practically the only institutions of higher learning open to Blacks in the US. Using nationally representative data files from 1970s and 1990s college attendees, we find that in the 1970s HBCU matriculation was associated with higher wages and an increased probability of graduation, relative to attending a Traditionally White Institution (TWI). By the 1990s, however, there is a wage penalty, resulting in a 20% decline in the relative wages of HBCU graduates between the two decades. We also analyze the College and Beyond’s 1976 and 1989 samples of matriculates which allows us to focus on two of the most elite HBCUs. Between the 1970s and 1990s, HBCU students report statistically significant declines in the proportion that would choose the same college again, preparation for getting along with other racial groups, and development of leadership skills, relative to black students in TWIs. On the positive side, HBCU attendees became relatively more likely to be engaged in social, political, and philanthropic activities. The data provide modest support for the possibility that HBCUs’ relative decline in wages is partially due to improvements in TWIs’ effectiveness at educating blacks. The data contradict a number of other intuitive explanations, including relative decline in pre-college credentials (e.g., SAT scores) of students attending HBCUs and expenditures per student at HBCUs.

A fun read.

There seems to be a lot of recent backlash against “cuteness” in economics paper. Steve Levitt and Ulrike Malmendier, in particular, have taken a lot of heat. So has Emily Oster. In his April 2nd article at The New Republic entitled “How Freakonomics is Ruining the Dismal Science”, Noam Scheiber identifies a “cleverness problem” in economics, arguing that students in the premier economics departments in the US end up playing an “academic parlor game”.

Several thoughts on the matter.

First, there seems to be an assumption that “cute” papers lack academic value. Certainly Levitt’s bagel and donut research led to his investigation about the (lack of) profit maximization by “legit” firms such as airlines. Similarly, Malmendier and DellaVigna’s research on gym memberships offer methodological insights from the perspective of psych and econ. The point is, “cute” research may allow us to learn important fundamental points about more “important” issues. As Matthew Rabin pointed out, decades ago there was short paper about used cars that was chastised for being “cute”.

Second, it isn’t obvious to me that “sexy data-set” papers cannot tackle big issues. Consider Ted Miguel’s utilization of rain as an instrument for income to analyze the relationship between income and civil unrest. Hardly a “small issue”! Now Scheiber also levies a criticism at Oster’s work about hepatitis B and missing women, citing “a snot-nosed grad student from Berkeley [who] pointed out that hepatitis B couldn’t possibly explain the missing women problem”. Fair point, I suppose (I admittedly have not read either papers), but the subject of her research obviously wasn’t a parlor game. Even if she was wrong, at least criticisms levied against her paper help strengthen the point that “Sen feared: [that] people were taking dramatic steps to avoid ending up with two girls”. Female infanticide ain’t no small issue!

In any case, people much much smarter than me have commented extensively on this article. For instance, see Marginal Revolution’s take and subsequent comments.

Steven Levitt and friends answer some interesting questions for students planning on entering PhD programs in the next few years: 

Do graduate students who do well in core microeconomics (Micro) courses also do well in core macroeconomics (Macro) and econometrics (Metrics) courses? Are students who achieve higher grades in their first-year core classes or general exams more likely to complete their Ph.D. and to obtain higher ranked positions in the job market?

Apparently answering these not very difficult, though somewhat interesting quesitons takes the efforts of not one, not two, but five of the world’s top labor economists.

Lest you think they were engaged in heated debate on intricate points of methodology, let me reassure you. They were not. They produced a single sixteen page paper on the topic.

It is probably not a coincidence that the authors are each from a different school, and each represent one of the five schools whose data is analyzed within the paper.

The content of the paper itself is interesting, but for those who truly want to succeed in the world of economics its implications are clear: if you want good data, first become a good politician.

A little bit of background on myself: I love computers, and I love economics. Unfortunately, the two disciplines mix pretty infrequently. That’s why I let out a little yelp when I found this paper, called “Fact-Free Learning” by Enriqueta Aragones, Itzhak Gilboa, Andrew Postlewaite, and David Schmeidler. The paper is trying to present a model of how we understand information that is already available to us. Sometimes, we figure out something new by looking at the data we have in a new light — fact-free learning!

But fact-free learning is tough. The real breakthroughs are often unexpected, and they may happen slowly. The authors argue that some familiar tools of economics and computer science, when brought together, can explain this phenomenon. Before I get into their argument, I need to explain something called complexity, because the central argument of the paper relies on it.

Computational complexity is the study of how computer algorithms scale as the size of the problem they are trying to solve increases. Usually we are looking for some sort of bound on the amount of time it will take the algorithm to finish, given a problem of size n.

Some algorithms may not scale so well. Seriously, they might not scale so well. We’re not entirely sure! (Why? Wikipedia explains.) This class of algorithms is called NP, and we are pretty sure that any implementation of an NP-hard problem will end up taking a few lifespans-of-the-universes to complete for even a small size input.

Now let’s get back to the paper. Remember how I said that it brought together tools of both computer science and economics? Well, the economic tool these fellows bring to the table is the regression. Say you have information on lots of variables, and you want to see which variables explain some phenomenon. If you pick out a few of these variables, you can regress the measure of the phenomenon on those variables to determine which variables are relevant.

But what if you can only use a few variables — say k — in the regression at once? Then you might want to run the regression with every possible combination of k variables, looking for the one that does the best job explaining the phenomenon. The authors argue that this process — finding the set of k variables that does the best job explaining a phenomenon in a regression — is a lot like fact-free learning. But there’s a catch:

Linear regression is a structured and relatively well-understood problem, and one may hope that, using clever algorithms that employ statistical analysis, the best set of k regressors can be found without actually testing all (mCk) subsets. Our main result is that this is not the case. Formally, we prove that finding whether k regressors can obtain a prespecified value of R2, r, is, in the language of computer science, NP-Complete. Moreover, we show that this problem is hard (NP-Complete) for every positive value of r.

The implication is that fact-free learning is really difficult for computers. And if it’s difficult for computers, it’s probably really difficult for people too!