Freakonomics: A Rogue Economist Explores The Hidden Side of Everything
By Steven D. Levitt and Stephen J. Dubner
(William Morrow, 242 pp., $25.95)
Much of the influence of modern economics derives from the claim that human beings respond rationally to incentives. It is a deceptively simple claim with many implications. If welfare pays more than work, many people will avoid work in favor of welfare; if the price of gasoline suddenly jumps, more people will use mass transportation; if the minimum wage is significantly increased, some workers will lose their jobs, simply because it is more expensive to pay them. These examples are pretty obvious, but the interest in incentives can have more surprising ramifications. Some economists contend, for example, that laws forbidding discrimination will actually increase discrimination: if employers have to make expensive accommodations for disabled people, or if they face a risk of age-discrimination claims from disgruntled older workers, they may be less likely to hire those who are disabled and older. Or consider the question of how to explain declining rates of marriage and increasing rates of divorce: the focus on incentives suggests that the growing economic opportunities for women are a key contributing factor.
This unusual book is an outgrowth of a profile of Steven D. Levitt, a young economist at the University of Chicago, that appeared in The New York Times Magazine. The profiler, Stephen J. Dubner, lionized his subject, and found him to be entirely fascinating. Dubner depicted Levitt as a maverick or an iconoclast— a nerd but also a rogue, as the book’s subtitle describes him. More, Dubner saw Levitt as someone who could use data to sort out the deepest riddles of society—what the book’s subtitle describes as “the hidden side of everything.” When several publishers approached Levitt with the suggestion that he should produce a book for a popular audience, he asked Dubner to collaborate with him. While much of the lucid, non-technical, and occasionally annoying exposition appears to be Dubner’s, the findings and the analyses in Freakonomics are Levitt’s, based on his academic writings; and so the book may be rightly treated as Levitt’s own.
What makes Levitt’s work distinctive is that he does not rest content with theory. In many domains, he offers remarkably creative empirical tests for theoretical claims, and he shows where they stand or fall. But there is an irony here. The book’s title and presentation disguise the extent to which Levitt is not really a “rogue economist” at all. On the contrary: Levitt’s central theoretical claim is completely conventional. As he puts it, “Incentives are the cornerstone of modern life.” Much of his book is built on this premise. But the very idea that people respond rationally to incentives is under intense pressure from some of the most exciting developments in contemporary economics.
Levitt is widely admired among social scientists, and his focus on incentives certainly leads to a number of striking findings. Here is an example. Many sellers of houses think that real-estate agents can be trusted to obtain the highest possible offer. After all, real-estate agents earn commissions, and hence their incentives should be perfectly aligned with the interests of sellers. But is this really so? Levitt calculates that a typical agent stands to gain only about $150 from obtaining an additional $10,000 for the seller—which means that haggling over $20,000, or even $30,000, isn’t worth much for the agent. On the other hand, real-estate agents tend to invest a lot of money in houses that they purchase for resale. How does their behavior shift when they are selling their own houses? It turns out that it shifts significantly. Real-estate agents keep their own houses on the market an average of ten days longer, and they sell their houses for about 3 percent more, producing a $10,000 gain on a $300,000 house. Contrary to the conventional wisdom, it turns out that the incentives of agents are not always allied with the incentives of sellers.
Levitt is particularly interested in the problem of cheating, and he wants to investigate the possibility that when there is a strong incentive to cheat, human beings will be more inclined to do so, in a way that will undo important social programs. His most interesting finding involves the performance of schoolteachers under President Bush’s No Child Left Behind law. In Chicago, public school administrators were told that if their students’ reading scores were low, their schools would be placed on probation and perhaps eventually shut down; if so, staff would be dismissed or reassigned. Levitt hypothesized that the new system created a strong incentive for cheating. He suspected that an easy way to cheat would be for teachers to collect students’ answer sheets and replace wrong answers with right ones.
To test his suspicion, he obtained an amazing database: the test answers of every Chicago public school student from third grade to seventh grade during the period from 1993 to 2000. Examining nearly 100 million individual answers, Levitt developed and applied a “cheating algorithm.” This was essentially a computerized method for detecting suspicious patterns—in which, say, a large group of students in a given class managed to produce the same six correct answers in a row, even though they were otherwise making many blunders. Levitt was able to show that widespread cheating occurred. His analysis was actually used by the Chicago public school system, which fired about a dozen demonstrated cheaters. The threat of discharge is a strong disincentive, and so it might not be surprising that the following year cheating fell by over 30 percent.
Levitt’s attention to incentives also plays a key role in his best and most provocative chapter, an inquiry into the massive decline in crime in the 1990s. Between 1975 and 1990, the rate of violent crime shot up by 80 percent, and experts predicted that the 1990s would see a continued increase, even a crime epidemic. But to almost everyone’s surprise, the crime rate rapidly decreased, ultimately to the levels of forty years earlier. Levitt’s innovative claim is that part of the decline is a direct result of the Supreme Court’s legalization of abortion in Roe v. Wade. The claim has outraged many people across the political spectrum, but Levitt’s basic idea is simple. After abortion was legalized, a number of likely criminals were not allowed to be born in the 1970s, and as a result the crime rate went down twenty years later. Levitt’s striking conclusion is that legalized abortion played a much larger role than capital punishment in reducing the rate of violent crime.
But Levitt is not only interested in abortion; his more general goals are to debunk failed explanations of declining crime and to isolate the factors that really made a difference. Many people think that the strong economy was responsible for the downturn in crime, but Levitt finds that a look at the data “destroys” this theory. At most, economic growth slightly reduces nonviolent crime, and it has little or no effect on violent crime. Hence the high growth rates of the 1990s cannot possibly explain the 40 percent decrease in violent crime in that period. Tougher gun laws have been unimportant, in part because there is a flourishing black market in gun sales. Levitt concludes that the much-publicized Brady Act, requiring a criminal check and a waiting period prior to gun purchases, “has proven to be practically impotent in lowering crime.” Nor has capital punishment had much of an impact. Levitt notes that the execution rate was and remains extremely low; hence he believes that the fear of execution is unlikely to be “a driving force in a criminal’s calculus.”
Drawing on Mayor Rudy Giuliani’s highly publicized successes in New York City, many people have attributed the drop in crime to innovative new strategies by the police. These include the “broken windows” approach, which attempts to reduce minor nuisances as a way of discouraging the more serious criminal activity that supposedly follows them. But Levitt shows that in New York City and elsewhere, the drops in violent crime actually preceded the adoption of the new strategies, which cannot be shown to have had a significant beneficial effect. Some people speculate that the reduction in crime came from the aging of the population, and in fact there was real population growth in the 1990s among the elderly (who are a lot less likely to be murderers and thieves). But Levitt concludes that “the graying of America did nothing to bring down crime in the 1990s. Demographic change is too slow and subtle a process—you don’t graduate from teenage hoodlum to senior citizen in just a few years—to even begin to explain the suddenness of the crime decline.”
What, then, does account for the decline? Levitt points to four factors. The first is the bursting of the crack bubble, which greatly reduced the market for crack, a source of a lot of violence. During the 1990s, the profits for selling crack decreased dramatically, and “younger dealers decided that the smaller profits didn’t justify the risk.” As a result, “the violence abated.” Levitt concludes that the “crash of the crack market” produced no less than 15 percent of the crime decrease in the 1990s. Even more important, in his view, was the second factor: the government’s increased reliance on incarceration. In the 1990s, courts started putting far more people in jail, so that by 2000, “more than two million people were in prison, roughly four times the number as of 1972. Fully half of that increase took place during the 1990s.” Levitt finds that the high imprisonment rates accounted for no less than one-third of the crime drop. The third factor is the increased number of police on the streets, producing increased deterrence and ultimately counting for as much as 10 percent of the diminution in crime in the 1990s.
The fourth factor is legalized abortion. Before Roe, daughters of wealthy and middle-class families were usually able to arrange for a safe abortion, even if it was unlawful. The real effect of Roe was felt mostly by poorer women, and especially those who were unmarried, teenaged, or poor. After the Supreme Court’s decision, Levitt says, such women had much easier access to safe and lawful abortions, and hence the abortion rate, for them, increased significantly. If we are trying to explain the decline in the crime rate, this is an important change, because childhood poverty and a single-parent household “are among the strongest predictors that a child will have a criminal future.” Levitt thinks that it is no coincidence that the decline occurred in the early 1990s, “just as the first cohort of children born after Roe v. Wade was hitting its late teen years.” The key point is that this cohort was missing those “children who stood the greatest chance of becoming criminals.” Hence it is unsurprising that the crime rate fell. Levitt’s extensive supporting evidence includes the striking finding that early reductions in violent crime occurred in the five states that legalized abortion at least two years before Roe v. Wade (New York, California, Washington, Alaska, and Hawaii). Indeed, studies in Australia and Canada have found a similar link between legalized abortion and crime.
Levitt likes to puncture myths. Turning from crime to democracy, he finds that campaign expenditures do not much affect elections. True, victorious candidates tend to raise a lot of money; but maybe they raise money because they are popular, rather than the other way around. To test this hypothesis, Levitt examines what happens when the same two candidates run against each other in consecutive elections, as has happened in almost a thousand congressional races since 1972. “Here’s the surprise; the amount of money spent by the candidates hardly matters at all,” he writes. “A winning candidate can cut his spending in half and lose only 1 percent of the vote.” In Levitt’s view, what matters is who the candidate is, not how much money he spends.
Levitt thinks also that myths pervade the topic of parenthood. For one thing, parents are scared of the wrong things. Most parents are unwilling to keep a gun in their house, believing that guns impose a significant fatality risk. But the same parents are also entirely willing to have a swimming pool in the backyard. In Levitt’s view, this disparity makes little sense. For every 11,000 residential pools in the United States, one child dies—meaning that about 550 children drown each year in pools. By contrast, only one child is “killed by a gun for every 1 million-plus guns.” Hence any particular child is more than 100 times more likely to die in a pool than to be killed by a gun.
Emphasizing that people fear guns far more than they fear pools, Levitt concludes that most people are “terrible risk assessors.” People fear flying more than they fear driving, but the per-hour death rates of driving and flying are about equal—an impressive point in favor of flying, because you can get a lot farther in an hour in an airplane. And when a risk produces outrage, people’s concern is greatly heightened, whatever the statistics show. Levitt believes that this creates genuine problems both for individuals and for governments as they respond to short-term scares. Far more lives would be saved by some modest precautions for swimming pools than are now saved by flame-retardant pajamas and child-resistant packaging.
Can parents do anything to promote academic success? Levitt finds that high test scores are not correlated with having an intact family, having a mother who worked while the child was between birth and kindergarten age, or having attended Head Start. None of these factors matters. What does matter is whether the parents are highly educated, whether the mother was thirty or older when her first child was born, whether there are many books in the home, and whether the parents have a high socioeconomic status. It follows that for purposes of producing high academic achievement, child-rearing techniques are greatly overrated. Of course parents count. But “by the time most people pick up a parenting book, it is far too late. Most of the things that matter were decided long ago—who you are, whom you married, what kind of life you lead.” Children are more likely to succeed if their parents are well-educated and well-paid. “It isn’t so much a matter of what you do as a parent; it’s who you are.”
Continuing his focus on parenting, Levitt also offers an extended and entertaining (and somewhat obsessive) discussion of children’s names. He is particularly interested in the difference between the naming decisions of African Americans and those of whites. Before the 1970s, black and white names overlapped far more than they do today. Now, more than 40 percent of African American girls born in California receive a name that is shared by none of the 100,000 white girls born in the same year. Distinctly black names are most likely to be given by “an unmarried, low-income, undereducated teenage mother from a black neighborhood who has a distinctively black name herself.” The “whitest” names include Molly, Amy, Claire, Jake, Connor, and Tanner; the “blackest” names include Imani, Ebony, Shanice, DeShawn, DeAndre, and Marquis. Moving beyond race, Levitt finds that there is a strong correlation between babies’ names and the socioeconomic status of parents. Wealthy white families are especially likely to select such names as Alexandra, Lauren, Katherine, Benjamin, Samuel, Jonathan, and Andrew; poorer white families favor such names as Amber, Heather, Kayla, Cody, Brandon, and Anthony. Highly educated parents show systematically different naming patterns from those of parents with little education.
But Levitt isn’t really interested in providing answers to trivia contests. He also wants to know whether names matter to economic success. As he is aware, some impressive studies show that people with distinctly black names tend to receive fewer callbacks from employers than otherwise identical people with distinctively white names. The economists Marianne Bertrand and Sendhil Mullainathan sent out numerous resumes and found that employers were far more interested in people with such names as Emily and Greg than in those with such names as Lakisha and Jamal. Consistent with this evidence, Levitt finds that “on average, a person with a distinctively black name—whether it is a woman named Imani or a man named DeShawn—does have a worse life outcome than a woman named Molly or a man named Jake.” But unlike Bertrand and Mullainathan, Levitt’s evidence suggests that the name is not the source of the problem. What matters is the family and economic circumstances, not the name itself: “If two black boys, Jake Williams and DeShawn Williams, are born in the same neighborhood and into the same familial and economic circumstances, they would likely have similar life outcomes.”
What accounts for the differences in names over time? Levitt does not have a complete answer, but he does demonstrate that “once a name catches on among high-income, highly educated parents, it starts working its way down the socioeconomic ladder.” The most popular names of any given year were particularly popular among high-end families of fifteen years prior. Levitt thus predicts that in 2015 we will find more girls with such names as Annika, Emma, Fiona, Grace, and Kate, while more boys will be called Aidan, Asher, Beckett, Carter, and Liam. Levitt concludes that “an overwhelming number of parents use a name to signal their own expectations of how successful their children will be.” While names are not “likely to make a shard of difference,” parents appear to “feel better knowing that, from the very outset, they tried their best.”
Readers of Freakonomics should not be misled by the book’s slick packaging and stupid title. Every one of the chapters is embarrassingly prefaced by an excerpt from Dubner’s Times profile, often congratulating Levitt for his wonderfulness and sometimes depicting him as a demigod. (I am not exaggerating. One chapter is introduced thus: “[Levitt] has shown other economists just how well their tools can make sense of the real world. `Levitt is considered a demigod, one of the most creative people in economics and maybe in all social science,’ says Colin F. Camerer, an economist at the California Institute of Technology.”) The authors seem to fear that their substantive material might be boring, and they try too hard to create major drama, cuteness, or perhaps a reality television show, with one- or two-sentence paragraphs. Here’s one paragraph: “Was Temptress actually `living out her name’ …? Or would she have wound up in trouble even if her mother had called her Chastity?” Another paragraph: “Has there ever been another art so devoutly converted into a science as the art of parenting?” Another paragraph: “Right?”
There is no such thing as freakonomics. Nor is there anything remotely freaky about Levitt’s approach. In assuming that people are rationally responsive to incentives, he is a perfectly conventional man of his trade. Levitt stands out not because of any large claims about human motivations, but because of his remarkable ingenuity, creativity, and sheer doggedness in investigating empirical questions about which no one seems to know much at all. Unfortunately, some of his findings are not terribly exciting. Are you amazed to learn that well-educated parents produce children with high test scores? Are you stunned to hear that many parents give their children names that they consider to signify high status?
But some of Levitt’s inquiries do illuminate important problems. We now know, for example, that economic growth does little or nothing to reduce violent crime; that innovative policing strategies haven’t much mattered; and that if states really want to reduce crime rates, they will put a lot more police on the street. But these points do not entirely explain the excitement generated by Levitt’s book. (The New York Times Magazine has just introduced a regular freakonomical column by Dubner and Levitt.) One factor, I think, is that the book cleverly combines Dubner’s journalistic gushing with Levitt’s academic findings, thus encouraging the reader to gush along with Dubner even before the findings are presented and assessed. More charitably, it is fun and even exhilarating to see how a real social scientist goes about testing competing hypotheses. Some of Levitt’s inquiries read like good detective stories. With respect to many urgent social questions, people engage in noisy and vehement debates without a sense that helpful illumination is available, in principle and maybe even in practice, by means of a careful investigation of the facts. Many of the issues that sharply divide us are, after all, empirical matters, and it is real progress simply to recognize them as such. And Levitt does far more than that.
Still, Levitt’s claims raise questions. Many parents know that unsupervised swimming is dangerous; they prefer pools to guns because they think that for most children pools are a lot more fun (and swimming is good exercise). In short, parents think that pools are much more likely to meet the standards of cost-benefit analysis, the economist’s lodestar. Nor does Levitt’s investigation offer anything to contradict the important finding that those with distinctly African American names are less likely to be called back for job interviews than are people with more stereotypically white names. If Jake Williams and DeShawn Williams have sent out the same resume, employers will show more interest in Jake. Even if the two have similar economic prospects in life, it remains a problem if DeShawn is subject to greater discrimination at an early stage in the employment market.
Levitt’s claim that campaign expenditures do not much affect elections is not established by the evidence that he marshals, which suggests, much more narrowly, that such expenditures do not have a substantial impact on outcomes between candidates who have run against each other in a prior election. Suppose that, having defeated Rick Lazio in 2000, Hillary Clinton runs against Lazio again in 2006. It is highly plausible to think that Clinton would win again, even if Lazio raised a lot more money this time. Levitt’s evidence suggests that significant changes in funding for two candidates who have opposed each other before are not likely to make much difference. But for candidates who are running for the first time, the accumulation of money is usually indispensable. In any case, Levitt generally insists that people respond rationally to incentives. If money is essentially irrelevant, why do candidates spend so much of their time trying to obtain it?
Levitt has been severely attacked for his most notorious claim, which is that the legalization of abortion significantly contributed to the decrease in violent crime in the 1990s. Liberals fear that the claim is racist, and conservatives despise the idea that abortion might be an effective crime-control strategy. But Levitt is not defending abortion, or arguing on its behalf for women who are poor and single. His claim is simply about what the facts show; and he does a great deal to make that claim more than plausible. To know whether he is right, however, we need to know a great deal not just about the law, but also about the actual abortion rates in the 1960s and 1970s. Roe v. Wade was decided in 1973, but numerous abortions, both legal and illegal, occurred before that date. One problem for his analysis, and even for the collection of data, is that there is no comprehensive evidence about the rate of illegal abortions, and the figures are much disputed, based as they are on extrapolations rather than direct evidence. If the Supreme Court’s decision did not by itself produce a significant increase in abortions—if the actual abortion rate spiked, say, in 1965 or 1980—then Levitt cannot demonstrate that the decrease in crime was significantly affected by Roe v. Wade. In his technical writing, Levitt has tried to address this complication; but in his book he does not adequately support the bold suggestion that “abortion was one of the greatest crime-lowering factors in American history.”
And there is a larger question in the background of all this. In general, Levitt assumes and sometimes shows that people respond rationally to incentives. This assumption is central to his claim that the death penalty is unlikely to deter; it also motivates many of his hypotheses about real estate, cheating, and crime. But Levitt also emphasizes and sometimes shows individual and social errors. He suggests that people choose “high-end” names with the (false) expectation that their children will be helped as a result. Is that rational? More broadly, he emphasizes that people are “terrible risk assessors”; hence his claim that parents are afraid of the wrong things. If people are terrible risk assessors, they will not always respond rationally to incentives. Maybe real-estate agents, criminals, and prospective cheaters are terrible risk assessors, too.
Much of the most exciting work in contemporary economics rejects the idea that human beings are entirely rational. It systematically investigates our difficulties in assessing risks. It finds pervasive human errors and illusions. It emphasizes problems of self-control. It also suggests that many people will often refuse to cheat, or to behave unfairly, simply because they fear social sanctions—and that many other people will play fairly because of their own conscience. In short, the real “rogue economists” of our day have been rethinking the theoretical foundations of modern social science, replacing homo economicus with homo sapiens. Many economists have known for a long time now that homo economicus is the real freak.
But Levitt is an empiricist and not a dogmatist, even about his own assumptions. In his profile and in this book, Dubner quotes Levitt as saying, with a nerdish guileless confidence, that “I’d like to put together a set of tools that let us catch terrorists. I don’t necessarily know yet how I’d go about it. But given the right data, I have little doubt that I could figure out the answer.” This is more than the confidence of economic theory, it is the confidence of social science; and it is admirable. One of the virtues of this decidedly unfreakish book is its demonstration that such confidence might well be vindicated—that with the right data, we might well be able to make progress in resolving some of society’s most intractable perplexities.
This article originally ran in the July 25, 2005 issue of the magazine.