Freaks of Nature

Freakonomics is bad too

hey apple (2000s kids remember)

In some quarters of our profession, the level of discussion has sunk to the level of a New Yorker article (…) The authors of these papers are usually unclear about the economic questions they address.

Nobel Laureate James Heckman for The New Republic

After going after Why Nations Fail (or attempting to), here I come for Freakonomics, the father of the pop-econ book genre in its modern form. Focusing on the economics “of the unusual and bizarre”, and it was successful enough to spawn a sequel, SuperFreakonomics, and a movie, as well as a website, podcast, and consulting firm. It also launched its authors, Stephen Levitt and Stephen Dubner, into economics superstardom. The book is a massive bestseller and many people have cited it as what got them interested in economics to begin with.

Super freak, super freak, he’s super freaky

Freakonomics and SuperFreakonomics are, at their core, a series of vignettes illustrating how economic behavior and rationality permeate all aspects of life. By “rationality”, what we (generally) assume is that we mean “behaving in a way that is as if a perfectly informed agent optimizes behavior per some rule”. This is somewhat reasonable, I think. And as someone who has written a ton of pieces about pop culture, economics, the intersection of the two, and various freaky occurrences, you’d think I like this book. But I don’t. And it’s for a simple reason: it is very frequently wrong.

The core five “tenets” of Freakonomics, per the UC Berkeley review of the book, are:

• Incentives are the cornerstone of modern life.

• The conventional wisdom is often wrong.

• Dramatic effects often have distant, even subtle, causes.

• “Experts”—from criminologists to real-estate agents—use their informational advantage to serve their own agenda.

• Knowing what to measure and how to measure it makes a complicated world much less so.

Regarding incentives, yeah, I agree, I’m not a hater. The second one is dicey. I’m gonna nope out on the third one, on principle. Fourth, of course, rationality. And fifth, vital for everyone to understand.

Overall, the “Freakonomics mindset” is not disagreeable or wrong - look deeply for incentives and hidden causes, even if counterintuitive or unconventional, and what matters is the result you obtain and not whether the cause seems “too small” for the consequence. Empirical carefulness is key. And experts (who determine conventional wisdom) might just be in it for themselves. The Freakonomics method, however, is more under fire (more on this later).

What is used to argue for the mindset and method are the vignettes themselves, on various topics.

Welcome, welcome, welcome to the house of freaks

Freakonomics volume 1 (or whatever you want to call it) is a pretty succint book, which draws on research close to its publication date (the book is now old enough to vote, or to drink in most civilized nations) to prove the five tenets above. The content list is as follows:

• Chapter 1: the economics of cheating - sumo wrestlers, bagel shops, teachers.

• Chapter 2: information frictions - the Ku Klux Klan, real estate agents

• Chapter 3: the economics of drug trafficking (hey! I’ve seen this one)

• Chapter 4: abortion legalization and crime

• Chapter 5: good parenting and education

• Chapter 6: the effect of names on the economy

Overall, chapters 1 and 2 seem fairly simple and straightforward: people cheat when cost > benefit, information frictions benefit people with more information. Florals in spring are more creative. Chapter 3 is very very interesting (some critiques, though). Chapter 5 is good for causal inference: a lot of “good parenting” trends are simply useless, and the reality is that “good parenting-ers” are probably just likelier to have successful kids. Chapter 6 - just ask Emily and Lakisha.

There was a minor controversy involving policing and crime. Since cities with higher crime rates also tend to hire more police, it’s impossible to determine whether crime causes policing, policing reduces crime, or policing does nothing. All the variation is endogenous - the variables respond to each other. But Levitt found that cities tend to hire more police officers during election years, with no relationship to crime rates - meaning there’s a random positive shock to policing, and Levitt found that this lowers crime rates. Except during these times, not enough police is hired - too small a sample size means the experiment is underpowered. Levitt did a no no. However, actually exogenous studies sorta vindicate Levitt - for instance, after the 1994 AMIA bombing in Buenos Aires, police presence was heightened around synagogues, which resulted in lower crimes near Jewish places of worship than other similar areas. Other crime-related claims are equally debatable, at least for the evidence presented.

Chapter four is the most controversial one of the entire book. The core thesis is that the drop in crime rates during the 1990s was caused, largely, by the legalization of abortion through the 1972 Supreme Court decision “Roe vs Wade” (Z”L). The core argument made by the Stephens is that the aborted fetuses (or murdered babies, your pick) of the 1970s would have grown into no-good hoodlums two decades later, as a result of their impoverished upbringings and dire life circumstances. The identification strategy here is simple: men aged 18 to 24 are the likeliest to commit crimes. One possibility is that there were simply fewer crime-aged men post abortion; another, that the remaining ones were less likely to commit crimes. The correlation to find is that states with higher abortion rates pre-Roe would have lower crime rates, that states with ealier legalization had lower rates still, and that post-Roe the high-abortion, high-crime states converged. Controlling for all influencing factors, they find a strong relationship between the two variables. The book also uses Ceaușescu’s Romania as another example: banning abortion resulted in soaring crime a generation later (rational expectations might explain Biden’s crime wave, then) - but in fact the paper they cite has the opposite conclusion, that there was a positive relationship.

Like everything in economics, this is debatable. Another paper with two same-name authors (the “Johns”, Lott and Whitley) posits another Freakonomics-esque possibility: if abortion increases the number of out-of-wedlock births, then there would also be a positive effect on the births of “future criminals”. Which effect is bigger, the Stephen effect or the John effect? The Johns, naturally, find that their effect is bigger - legal abortion increases crime rates, rather than decreasing it. Explains Romania, then.

A second rebuttal comes from a paper focusing on the crack epidemic (unfortunately, only one author, Ted Joyce). Instead of comparing between states, the study compares between the crime outcomes of people whose states legalized abortion before and after Roe, and between those states and the same states post-Roe. This is because, naturally, a negative relationship would show itself as a change in some states before Roe versus the rest of states, and then a change of the rest of the states versus the early repealers post Roe. The author finds no such relationship, and instead shifts blame to the crack epidemic and gun ownership for the drop in crime rates: controlling for those, any effect disappears.

Levitt and his coauthor, Donohue (named John), shot back that using Joyce’s methodology but more numbers, then their original results still stood (why would extending a specific cohort further down the line make sense, though). Joyce counterargued that, with a similar but slightly tweaked methodology (comparing cohorts instead of states), you also got no result. But it turns out that an embarassing blunder was committed by the Stephens all along: a different paper, by the Chrises (Foote and Goetz ) found that a coding error in Levitt’s work explained his findings, and that fixing it erased his results. Oops!

Future replications for England and Wales found no impact, but two different ones for Europe found conflicting results. And Levitt had another set of results confirming his original findings in 2020, which he summed up on his podcast. Twenty years from now, we’ll be able to repeat the experiment the other way around, at least. Lastly, an alternative theory emerged: lead. Exposure to lead affects behavior, and leaded gasoline was banned in the early 70s too; meaning that the Roe effect might be a Clean Air Act effect instead. (This is an even more Freakonomics).

When everyone’s super… no one will be

SuperFreakonomics was a sequel to Freakonomics (woah!), and expanded the concept into new fields of superfreaky inquiry. The chapter structure is:

• Chapter 1: the economics of prostitution (future blog post?)

• Chapter 2: patterns and details as predictive tools

• Chapter 3: altruism and selfishness

• Chapter 4: unintended consequences and simple fixes

• Chapter 5: externalities

• Chapter 6: can monkeys use money? (NOPE alert)

The book is, again, mostly okay on the topic, and as good as the papers it cites - and the two big controversies come from chapter two and chapter four.

Chapter two has a smaller one: does birth month matter for baseball success? Mirroring a famous paper by (Nobel Laureate) Joshua Angrist and (if he hadn’t died he’d have been a Laureate) Alan Krueger, the Stephens try to tease out whether kids born in December/July (i.e. “older”) are more likely to become sports pros than kids born in January/August (i.e. “younger”) in the same class. The answer is that yes, except the effect is very very small (and might be underpowered), but this goes against the Stephens’s claims that effort, not talent, matters - especially given that kids with pro fathers are 800 times more likely to go pro than just random spawn.

Another small controversy involves driving drunk versus walking drunk: when the goal is self preservation, driving drunk is actually safer than walking drunk. The problem here is that this is a ceteris paribus argument, but choice of transportation is mutatis mutandis: there are likely large, systematic differences between those who drive and those who walk - perhaps gender, perhaps age, perhaps drunknenness, perhaps distance. So it’s not really possible to say whether, controlling for everything, driving is safer than walking - gender wage gap denial logic strikes again.

The biggest SuperFreakonomics controversy was about claims involving climate change. The book’s climate science is pretty widely criticized (here’s the New Yorker, or a climate scientist called, I shit you not, Raymond Pierrehumbert) so I don’t want to dwell into that. But the core argument the Stephens seem to be making is that the cost of fossil fuels is too low compared to renewables, so fighting climate change is extremely costly - and climate change is just not dangerous enough to justify it. None of this is right.

Renewables are cheaper than almost all forms of fossil fuels now, which is due to accelerated innovation in the sector - a basic Malthusian tier mistake. This means there’s upside, not downside, in reducing emissions - replacing expensive energy with negative externalities with cheaper energy with positive ones. And the costs of climate change are underestimated, not correctly calculated, in traditional economic models - while (Nobel Laureate) William Nordhaus’s DICE claims there would only be a loss of 8.5% of global income with 6° Celsius (42° American ones) of warming - which would be hot enough to make most of the Caribbean uninhabitable¹, so I think that’s on the low end. Climate economist (and Nobel snub) Martin Weitzman countered saying that higher global temperatures entailed higher tail risks, which Nordhaus doesn’t account for based on outdated climate science, and that even lower levels of warming might prove catastrophic. The Nordhaus-Weitzman debate deserves a closer look on its own, but it should be noted that the book quotes Weitzman, who is literally arguing for the opposite position as the authors: climate action does pass a cost-benefit analysis because its costs are so high and unpredictable. And the book’s solution, geoengineering (aka making aritifical volcanoes go off to cool the Earth²) only makes sense if actually tackling emissions is too expensive - though there is a point to be made in this general direction about the collective action issues stemming from rich countries being responsible for climate change but developing countries having to do the legwork of cutting emissions, this time involving carbon capture.

To quote the comments on one of Brad Delong’s blog posts on the controversy:

While eco-cultists like Al Gore keep referring to a 6 degree C rise in global temperature as some sort of doomsday scenario, the agnostics grumble that the temperature of the earth is already 287 Kelvin, so we're talking about a mere 2% increase.

I can haz best-seller?

Freaks versus Geeks

A good example of the Freakonomics mindset in actual practice comes from the PhD thesis of COVID thought leader Emily Oster. Nobel Laureate Amartya Sen mentioned that Asia has around a hundred million “missing women” due to sex-selective abortions; Oster, noticing a pattern involving when the missing women were happening (after the first birth), found a series of instrumental variable hijinks that attributed the difference to hepatitis B instead of sexism. Except some “snot-nosed grad student from Berkeley” found that there wasn’t a high enough rate of hep B in Asia to explain this, so sex-selective abortions remained a more parsimonious explanation. This isn’t to shade Oster, who graciously admitted her mistake after it was pointed out. The problem is that there was an emphasis on the cleverness and counterintuitiveness of the answer, which led to it getting pointed out in the Freakonomics blog.

I don’t think that exploiting clever quirky scenarios to answer hard questions is necessarily bad (Levitt’s paper on car alarms and crime is very good), or that all counterintuitive results must be wrong; rather, that there is a point where only rigor can answer big questions. The Angrist and Krueger education paper I cited, for instance, showed that children born just below the cutoff for each school cohort had an extra year of education, which resulted in higher earnings compared to children born right after the cutoff (July and August, respectively). The issue is that you can’t really say much about the education-income relationship this way, simply because most of the questions don’t have easy, quirky experiments you can do - though it can take you some of the way, say on class sizes, or on college admissions.

A big opponent of this “cleverness” approach is Nobel Laureate James Heckman (fun fact: that glowing review of his work was written by Stephen Levitt), who opposes the “cleverness” approach due to its, to him, lack of rigor - and attributes its popularity to the extreme incentives in academia to publish fast, and a lot, which means it’s harder to just sit down and grind through an extremely thorny subject. Raj Chetty said “people think about the question less than the method” on this same issue. And of course there’s David Card:

It is exactly like postmodernism in the humanities. What is there to say about Beethoven anymore? (...) Every moron can’t understand technical orchestration, doesn’t know the history of music. So you write about him having a gay affair with his nephew.

The point here is that doing “bullshit empirical work”, per Heckman, doesn’t really answer big questions - it just answers small ones. This is a double-edged sword: you can’t really make big inferences about, say, inequality if all you have are quirky studies about bathrooms in Lesotho. The conclusions you draw are likely going to be ideological (because all humans are ideological), and Freakonomics is a book with a very strange ideology indeed. But the problem is that invading other disciplines to evangelize on instrumental variables while not knowing the basics of it are problematic (see: drugs), but also just miss the basics of economics. From another review of the book:

"When future generations ask the economics profession 'What were you doing while the great bubble built up ahead of the Second Great Depression?', and we have to reply 'Lots and lots of quirky little working papers about sumo wrestling and speed-dating', it is going to be really, really, [sic] f-ing embarrassing"

And we did, and it was; thank God nobody told the truth to HM The Queen [Maia note: she asked why nobody saw 2009 coming], or the high brows of the economics profession might be decorating a series of pikestaffs outside Traitors' Gate.

The basic problem with the Freakonomics era was that the profession abandoned the study of production, consumption and exchange. I don't wholly agree with Lord Skidelsky, but he is right - economics is the study of the economy, it's not the study of "rational choice" or "behaviour" in the abstract, and the fact that econometricians have invented a huge part of the toolkit of modern statistics doesn't mean that anything you can estimate using an econometrics package is thereby "economics".

Conclusion

Just do economics guys. It’s hard, but it’s good.

1. Well it wouldn’t be inhabitable you would just get third degree burns from going outside for 8 out of the 24 hours of the day. Sounds like “less than COVID” in output losses to me!

2. The science here is that volcanic eruptions are known to release substances that block out sunlight and thus reduce temperatures (see the Year Without A Summer), so they would do something similar except without harming the environment. I’m not sure if it would work.