Steven Levitt
Steven D. Levitt is an American economist and the William B. Ogden Distinguished Service Professor of Economics at the University of Chicago, where he directs the Becker Center on Chicago Price Theory.[1] He earned a B.A. from Harvard University in 1989 and a Ph.D. from MIT in 1994 before joining the University of Chicago faculty in 1997.[1] Levitt gained international prominence for co-authoring Freakonomics (2005) with journalist Stephen J. Dubner, a book that applies economic incentives and empirical analysis to unconventional questions, such as the effects of parenting practices on child outcomes and the economics of drug dealing, and which has sold over four million copies.[1] Levitt's research emphasizes causal identification through natural experiments and quasi-experimental designs to test first-principles economic theories against real-world data, spanning topics from campaign finance to school choice and urban crime.[2] In 2003, he received the John Bates Clark Medal from the American Economic Association, awarded to the most influential economist under age 40, particularly recognizing his contributions to the empirical study of crime, including analyses of policing effectiveness and incarceration impacts.[3][4] His work has been extended through sequels like SuperFreakonomics (2009) and media ventures, including the Freakonomics blog and podcast.[1] A defining and controversial aspect of Levitt's research is the hypothesis, developed with John Donohue, that the legalization of abortion in the United States following Roe v. Wade (1973) contributed significantly to the crime decline starting in the 1990s by reducing the number of children born into high-risk environments, a claim supported by state-level variation in abortion access and timing but challenged by critics on grounds of omitted variables and alternative factors like improved policing.[5] Levitt and Donohue have defended the finding's robustness in subsequent analyses, estimating it accounts for about half the observed crime drop, underscoring debates in empirical social science over causal inference amid potential biases in observational data.[5]
Early Life and Education
Childhood and Upbringing
Steven Levitt was born on May 29, 1967, in Boston, Massachusetts.[6] He grew up in St. Paul, Minnesota, attending the private St. Paul Academy and Summit School, from which he graduated in 1985.[7] [8] Levitt came from a Jewish family.[9] At St. Paul Academy, Levitt excelled in academic competitions, captaining the quiz bowl team and leading it to two national appearances.[10] [11] His high school mathematical preparation was limited, extending only to calculus, which he later described as inadequate for advanced economic study.[6]Academic Training
Levitt earned a Bachelor of Arts degree in economics from Harvard University in 1989.[12][1] He then pursued graduate studies at the Massachusetts Institute of Technology (MIT), where he completed a Ph.D. in economics in 1994.[12][1] His doctoral dissertation, titled Four Essays in Positive Political Economy, was supervised by James Poterba.[13] The work focused on empirical analyses of political economy topics, aligning with Levitt's early interest in applying economic tools to non-market behaviors.[13]Academic Career
University Positions
Steven Levitt joined the University of Chicago Department of Economics in 1997 as an assistant professor following his PhD from MIT in 1994.[1] He was promoted to associate professor with tenure in 1998.[14] By 1999, Levitt held the associate professor rank, as noted in his publications from that period.[15] Levitt advanced to full professor status, receiving the named Alvin H. Baum Professorship in Economics by 2003.[16] He subsequently held the William B. Ogden Distinguished Service Professorship in Economics, a position reflecting his contributions to price theory and empirical economics.[1] In March 2024, Levitt announced his retirement from active academic duties, attaining emeritus status while continuing affiliations such as directing the Becker Center on Chicago Price Theory.[17][18] Throughout his tenure at Chicago, Levitt maintained a primary focus on faculty research and teaching in economics, with no recorded positions at other universities post-PhD.[1]
Awards and Honors
In 2003, Levitt was awarded the John Bates Clark Medal by the American Economic Association, recognizing him as the most outstanding American economist under the age of 40 for his empirical contributions to the economics of crime, campaign finance, and related fields.[4][16] The medal, announced on April 25, 2003, highlighted Levitt's innovative use of natural experiments to establish causal relationships in microeconomics.[4] Levitt received several early-career honors, including the National Science Foundation's CAREER Award in 1999 for his research integrating economic incentives with empirical analysis of social phenomena.[19] In 2000, he was selected for the Presidential Early Career Award for Scientists and Engineers (PECASE), the highest honor for early-career researchers from the U.S. federal government, emphasizing his potential for leadership in science and engineering.[19] That same year, he earned the Alfred P. Sloan Research Fellowship, supporting his work as a promising young scholar in economics.[19] For teaching excellence, Levitt received the Llewellyn John and Harriet Manchester Quantrell Award from the University of Chicago in 1998, one of the nation's oldest undergraduate teaching prizes, based on student nominations for his Economics of Crime course.[20][19] He was elected a Fellow of the American Academy of Arts and Sciences in 2002 and a Fellow of the Econometric Society in 2004, acknowledging his advancements in economic theory and empirical methods.[19] Additional recognitions include the Duncan Black Prize in 2000 for a paper on political economy (co-authored with James Poterba) and the Garvin Prize in 2003 from UC Berkeley's Law and Economics Workshop for outstanding research presentation.[19] In 2006, Freakonomics, co-authored with Stephen J. Dubner, was named Book Sense Nonfiction Book of the Year by independent booksellers.[19] Levitt also delivered the Economic Journal Lecture at the Royal Economic Society meetings in 2003.[19]Research Methodology
Natural Experiments and Causal Inference
Steven Levitt's approach to empirical economics emphasizes natural experiments as a primary tool for causal inference, leveraging exogenous variations in real-world data to mimic the randomization of controlled trials. These experiments arise from policy changes, institutional quirks, or unexpected events that create plausibly random assignment to "treatment" and "control" groups, enabling identification of cause-and-effect relationships where traditional randomized experiments are infeasible. Levitt's method addresses the endogeneity problem in observational data—where correlation does not imply causation—by focusing on shocks that are independent of the outcome of interest, thus isolating causal impacts. This technique gained prominence in applied microeconomics during the 1990s, with Levitt among the early adopters who popularized its use for addressing policy-relevant questions.[6][21] A hallmark of Levitt's methodology involves exploiting staggered implementation of policies across regions or time periods to apply difference-in-differences estimation, which compares pre- and post-treatment outcomes between affected and unaffected units while controlling for time-invariant differences. For example, in his analysis of legalized abortion's effect on crime, Levitt treated the varying timing of state-level legalization in the early 1970s—following the 1973 Supreme Court decision in Roe v. Wade—as an exogenous shock, using non-legalizing states as controls to estimate a 15-20% reduction in crime rates attributable to fewer unwanted births. Similarly, he has employed regression discontinuity designs around arbitrary cutoffs, such as school entry ages or election thresholds, to infer causal effects on outcomes like student performance or political behavior. These quasi-experimental strategies, often combined with instrumental variables to further address selection bias, underpin much of Levitt's work on incentives and unintended consequences.[22][23] Levitt's reliance on natural experiments reflects a broader "credibility revolution" in economics, prioritizing transparent identification strategies over purely correlational analyses, though critics note potential omitted variable biases if the exogeneity assumption fails. His papers frequently detail robustness checks, such as placebo tests and synthetic controls, to validate findings, contributing to the field's shift toward defensible causal claims from non-experimental data. This methodology has been applied across domains, from sumo wrestling match-fixing—using performance discontinuities as evidence of collusion—to teacher cheating detection via anomalous test score patterns, demonstrating how institutional details can yield credible inference.[24][25]Incentive-Based Analysis
Steven Levitt's incentive-based analysis frames human behavior as primarily driven by responses to incentives, which he categorizes into economic (financial gains or losses), social (reputation or peer pressure), and moral (ethical considerations or guilt).[26][27] This methodological lens posits that by identifying and varying incentives, one can predict and explain actions that appear irrational or hidden, often revealing unintended consequences. Levitt applies this to empirical data, testing hypotheses where incentive structures change—such as policy shifts or institutional rules—and observing behavioral adjustments to infer causality, complementing natural experiments.[28] A prominent example is Levitt's examination of teacher cheating in Chicago Public Schools during the late 1990s, where high-stakes testing tied teacher evaluations and school funding to student performance, creating strong economic and professional incentives to manipulate results. Collaborating with Brian Jacob, Levitt developed algorithms to detect anomalies in standardized test data, such as unusual patterns of erasures favoring correct answers, particularly when scores hovered near passing thresholds. Their 2003 study estimated cheating occurred in at least 4-5% of classrooms, with incidence rising in schools facing severe consequences for low performance, demonstrating how accountability incentives could perversely encourage fraud rather than improved teaching.[29][30] Levitt extended this approach to other domains, such as sumo wrestling, where the incentive to fix matches—gaining a critical win for promotion versus minimal cost for losing—explained statistical improbabilities in close bouts; analysis of tournament data showed wrestlers with 7-7 records improbably winning 80% of decisive matches against 7-7 opponents, far exceeding expected odds.[26] In real estate, he found agents sell clients' homes quicker for lower prices due to commission-based incentives prioritizing volume over maximization, but hold their own properties longer for higher returns, highlighting asymmetric incentives between fiduciary duty and self-interest.[26] These cases underscore Levitt's emphasis on marginal incentives: small changes near decision thresholds often trigger disproportionate behavioral shifts, enabling rigorous testing against null hypotheses of random variation.[31] While effective for uncovering hidden motives, Levitt's framework has faced scrutiny for potentially oversimplifying complex social dynamics by prioritizing incentives over cultural or psychological factors, though his data-driven validations—via statistical anomalies and comparative statics—provide empirical robustness absent in purely theoretical models.[32] In experimental settings, such as performance incentives for students, Levitt's co-authored work shows immediate rewards boost effort more than delayed ones, aligning with incentive responsiveness but fading over time, informing policy design to sustain effects.[33]Key Research Areas
Economics of Crime
Steven Levitt's research in the economics of crime emphasizes empirical identification of causal factors influencing crime rates, often leveraging natural experiments to isolate policy effects from confounding variables. His work demonstrates that increases in police presence and incarceration levels have statistically significant deterrent and incapacitative effects on criminal activity. For instance, Levitt estimated that a 10% increase in police force size leads to a 3-4% reduction in violent crime and a smaller effect on property crime, using variations in police hiring tied to mayoral election cycles as an instrumental variable to address endogeneity concerns. Levitt's analysis of incarceration highlights its role in crime reduction primarily through incapacitation rather than deterrence. In a 1996 study, he found that a 10% expansion in prison populations correlates with roughly a 4% decline in serious crimes, attributing most of this to removing active offenders from society rather than altering potential criminals' behavior via fear of punishment. This conclusion stems from panel data regressions across U.S. states from 1970 to 1990, controlling for factors like unemployment and demographics, and using instruments such as sentencing policy changes to establish causality.[34] Distinguishing between deterrence and incapacitation, Levitt examined arrest rates' apparent crime-reducing impact, concluding that measurement error and incapacitation explain the correlation more than behavioral responses to punishment risks. His 1998 paper used time-series data from 59 U.S. cities, showing that while higher arrest probabilities coincide with lower crime, this largely reflects capturing repeat offenders rather than scaring others away, as short-term arrest spikes do not produce sustained deterrence beyond incapacitative effects.[35] Levitt also investigated specific policies like California's three-strikes law, collaborating with Daniel Kessler to show that sentence enhancements reduced serious crime by 20-30% in the early 1990s through both incapacitation of high-rate offenders and modest deterrence, as evidenced by discontinuities in crime trends around policy implementation dates compared to other states. Overall, these findings underscore the marginal returns of expanding criminal justice resources, with Levitt estimating that police and prisons accounted for about 40% of the U.S. crime drop in the 1990s, challenging narratives minimizing enforcement's efficacy.[36][37]Cheating and Corruption
Levitt's research on cheating emphasizes how incentives can lead individuals to subvert rules in high-stakes environments, often using statistical anomalies to detect dishonest behavior. In collaboration with Mark Duggan, he analyzed professional sumo wrestling tournaments, where wrestlers compete in 15-match cycles and require eight wins to maintain or advance in rank.[38] Their 2002 study documented systematic match rigging, particularly in "crunch-time" bouts between two wrestlers each holding a 7-7 record, where the probability of victory for the wrestler needing the win to reach eight was 76.1%, far exceeding the baseline 50% expected under fair play.[39] Evidence of reciprocity emerged in subsequent tournaments, with the previously "helped" wrestler achieving a 60.2% win rate against the benefactor, compared to a 48.5% rate in non-reciprocal pairings.[38] Wrestlers from the same heya (stable) showed elevated collusion rates, suggesting institutional facilitation, while cheating declined markedly during periods of heightened media scrutiny, such as post-scandal coverage.[39] In education, Levitt partnered with Brian Jacob to investigate teacher and administrator cheating on standardized tests in Chicago Public Schools from 1994 to 2000. Analyzing over 95,000 classrooms, they identified cheating through unusual patterns, including disproportionate erasures changing wrong answers to correct ones, especially near passing thresholds.[29] Their model flagged serious cheating in at least 4-5% of elementary classrooms annually, with higher incidence in grades facing promotion gates and schools under probation.[29] Factors like frequent teacher absences and principals incentivized by test scores correlated with elevated cheating rates, though economic training among administrators reduced it, implying awareness of detection risks.[29] Levitt's approach extended to experimental validation; in a 2003 field test, altering test security in Chicago reduced detected cheating by over 40% in targeted schools, confirming incentive responses to monitoring.[40] These studies underscore corruption's prevalence where marginal benefits outweigh risks, with empirical detection reliant on deviations from random error distributions rather than direct observation.[29][38]Other Empirical Studies
Levitt, in collaboration with Chad Syverson, analyzed real estate transactions using data from Chicago-area sales between 1992 and 2002, finding that homes owned by real estate agents sold for 5.1% more and remained on the market 10 days longer than comparable client-owned homes, suggesting agents exploit informational asymmetries by providing less effort on clients' properties compared to their own.[41] This study employed a natural experiment by matching agent-owned and client-owned properties on observables like location and characteristics, isolating the effect of agency incentives.[41] In a 2004 study with Roland Fryer, Levitt examined California birth certificate data from 1961 to 2000 alongside socioeconomic outcomes, determining that distinctively Black names are chosen by lower-income families but do not independently cause adverse life outcomes such as reduced wages or test scores; instead, differences stem from correlated parental characteristics, with no evidence of employer discrimination via lower callback rates in audit-like analyses of name signals.[42] The research used regression discontinuity and sibling fixed effects to control for family background, challenging assumptions of direct name-based bias.[43] Levitt's work on education includes a 2006 analysis with Brian Jacob and Jens Ludwig of Chicago school choice lotteries, revealing that winning a spot in a selective high school increased college enrollment by 6.5 percentage points but had no significant impact on earnings or incarceration rates 8-10 years later, attributing limited effects to non-compliance and peer influences.[44] Separately, a 2016 field experiment with John List and Sally Sadoff in Chicago elementary schools tested performance incentives, where promising high-stakes rewards to students boosted math scores by 0.15-0.20 standard deviations, though effects faded without sustained motivation, highlighting short-term behavioral responses over long-term skill building.[45] Levitt and John List's 2004 examination of the game show Weakest Link used over 2,500 episodes to test taste-based versus statistical discrimination, finding evidence of the former: contestants discriminated against women and minorities beyond performance differences, with exit probabilities 5-10% higher for targeted groups conditional on skill, supporting Becker's model where prejudice imposes costs on discriminators.[46] These findings drew on sequential elimination data to disentangle beliefs from preferences, with robustness to controls for contestant demographics and game dynamics.[46]The Abortion-Crime Hypothesis
Origins and Evidence
The abortion-crime hypothesis was developed by economists John J. Donohue III and Steven D. Levitt as an explanation for the sharp decline in U.S. crime rates beginning in the early 1990s, a trend that conventional factors such as increased incarceration, more police officers, and economic improvements failed to fully account for.[47] Levitt conceived the core idea while reviewing historical data in the Statistical Abstract of the United States, noting that national violent crime rates peaked in 1991 before falling precipitously, approximately 18 years after the Supreme Court's Roe v. Wade decision legalized abortion nationwide on January 22, 1973—a lag aligning with human gestation (about 9 months) plus 16-18 years until the affected cohort reached typical peak crime ages of 18-20.[48][47] Donohue and Levitt formalized the hypothesis in their paper "The Impact of Legalized Abortion on Crime," published in the Quarterly Journal of Economics in May 2001, arguing that legalized abortion reduced the births of children born into high-risk environments—such as poverty, single-parent households, or maternal youth—which are statistically linked to elevated future criminality.[47] To establish evidence, Donohue and Levitt exploited the natural experiment created by varying state-level abortion legalization dates prior to Roe v. Wade: five states (Alaska, Hawaii, New York, Washington, and the District of Columbia) and California (via judicial ruling) permitted elective abortions starting in 1970, allowing earlier exposure compared to the other 45 states post-1973.[47] Their regressions of crime rates (including murder, violent crime, property crime, and arrests) on lagged abortion rates, using data from 1985-1997 FBI Uniform Crime Reports and CDC abortion statistics, showed that states with higher post-legalization abortion rates experienced larger crime reductions 15-20 years later; for instance, a predicted 10% reduction in unwanted births correlated with a 6-11% drop in murder rates among affected cohorts.[47] Early-legalizing states saw crime declines begin 3-4 years sooner than elsewhere, with national models attributing 0.5-1.0 fewer arrests per capita by 1997 per 1,000 abortions in 1973.[47] Further supporting their causal claim, Donohue and Levitt incorporated controls for confounders like per capita prison populations (which rose through the 1980s but did not align with the 1990s crime timing), unemployment rates, income inequality, police per capita, and the crack cocaine epidemic's decline, finding the abortion effect persisted robustly across specifications.[47] They estimated legalized abortion accounted for about half of the observed crime drop between 1991 and 1997, with disproportionate impacts on crimes committed by 18-24-year-olds, who comprised the post-Roe cohorts; for example, murder arrests for this age group fell 35% nationally from 1991-1997, exceeding declines in older cohorts unaffected by abortion access.[47] International parallels, such as crime drops 15-20 years after abortion reforms in Canada (1969), Australia (early 1970s), and Romania (post-1989 restriction reversal), were cited as corroborative, though U.S. data formed the primary basis.[47]Empirical Support and Challenges
The Donohue-Levitt hypothesis posits that the legalization of abortion following Roe v. Wade in 1973 reduced crime rates approximately 18 years later, as cohorts exposed to higher abortion rates were less likely to engage in criminal activity due to reduced unwanted births associated with socioeconomic risk factors. In their seminal 2001 analysis, Donohue and Levitt estimated that legalized abortion accounted for roughly half of the decline in violent crime and property crime observed in the 1990s, with an elasticity implying that a 10% increase in abortions per live birth reduced crime by 5-15%.[47] Supporting evidence included the temporal pattern where crime rates began falling earlier in states that legalized abortion before 1973, and cross-sectional correlations between abortion exposure and lower arrest rates for relevant birth cohorts.[47] A 2004 follow-up by Donohue and Levitt extended the analysis to later data, reaffirming the link by showing sustained crime reductions aligned with abortion exposure, even after controlling for factors like incarceration rates and economic conditions.[49] Subsequent work by Donohue in 2020, updating the original models through 2017, maintained that legalized abortion explained 47% of the violent crime drop and 33% of the property crime drop over three decades, attributing robustness to the hypothesis's ability to withstand alternative explanations such as the crack cocaine epidemic's decline or policing innovations.[50] The analysis incorporated state-level fixed effects and lagged abortion rates as instruments for cohort unwantedness, yielding statistically significant negative coefficients on crime outcomes for affected generations.[50] Anecdotal consistencies, such as lower crime among smaller, post-legalization birth cohorts from high-risk demographics, further bolstered the causal interpretation in these models.[47] Challenges to the hypothesis emerged prominently from econometric critiques. Christopher Foote and Christopher Goetz (2005, revised 2008) identified a coding error in the 2001 paper where state and year fixed effects were inadvertently omitted from violent crime regressions, rendering the abortion coefficient statistically insignificant once corrected; the effect persisted weakly for property crime but was deemed non-causal due to omitted variable bias from unmodeled trends like lead exposure reductions.[51] Theodore Joyce (2004, 2009) replicated the models using arrest data for younger cohorts and found no significant crime reductions attributable to abortion exposure, arguing that abortion rates served as a noisy proxy for underlying fertility trends and that pre-Roe legalization effects were confounded by contemporaneous policy changes.[52] Joyce further contended that the hypothesis failed robustness checks against international data, where abortion liberalization did not uniformly predict crime drops, and emphasized aggregation issues in measuring effective abortion exposure across demographics.[52] John Lott and others proposed alternative mechanisms, such as abortion's potential to select for higher-risk pregnancies proceeding to birth or amplifying crime through family instability, though empirical tests yielded mixed results with some analyses suggesting a positive crime-abortion link in specific contexts.[53] Donohue and Levitt countered these critiques in replies (2004, 2006), reinstating fixed effects and arguing that the core lagged-cohort effect held for arrest rates over full lifespans, but critics maintained that endogeneity from unobserved confounders—like differential state responses to 1960s crime waves—undermined causality.[49] Overall, while initial correlations supported the hypothesis, methodological refinements and sensitivity analyses have reduced its estimated magnitude and raised doubts about unconfounded causation, with no consensus emerging in the literature as of 2020.[50][52]Freakonomics and Public Engagement
Book Series Development
The Freakonomics book series emerged from an unlikely collaboration between University of Chicago economist Steven D. Levitt and journalist Stephen J. Dubner, sparked by Dubner's 2003 New York Times Magazine profile of Levitt's unorthodox research into incentives and hidden social patterns.[54] This encounter highlighted Levitt's empirical approach to questions like cheating among sumo wrestlers and real estate agents, prompting the duo to expand these ideas into a full-length book. Published on April 12, 2005, by William Morrow, Freakonomics: A Rogue Economist Explores the Hidden Side of Everything applied basic economic principles—particularly incentives—to counterintuitive topics, blending Levitt's data-driven analyses with Dubner's narrative flair.[55] The volume eschewed traditional economic jargon, instead posing provocative questions such as "Why do drug dealers still live with their mothers?" to reveal causal mechanisms in everyday behavior.[56] The debut book's commercial triumph—topping The New York Times bestseller list for two years, selling over 4 million copies in the United States by 2009, and translating into more than 35 languages—drove the series' expansion.[57] Levitt and Dubner capitalized on this momentum by launching a New York Times blog in 2005, which served as a testing ground for new ideas and reader engagement, feeding directly into sequel content.[56] This iterative process allowed them to refine their formula: Levitt supplied rigorous, often proprietary datasets and econometric insights, while Dubner structured the material into accessible, story-driven chapters that prioritized empirical surprises over policy advocacy. The series evolved from standalone essays in the first book to broader thematic explorations, maintaining a focus on marginal incentives and unintended consequences without rigid theoretical frameworks.[58] Subsequent titles built on this foundation amid growing public appetite for "freakonomics-style" thinking. SuperFreakonomics: Global Cooling, Patriotic Prostitutes, and Why Suicide Bombers Should Buy Life Insurance, released on October 20, 2009, delved into larger-scale puzzles like prostitution economics and climate engineering, incorporating collaborations with other researchers but drawing criticism for selective data emphasis in non-Levitt studies.[59] Think Like a Freak, published June 3, 2014, shifted toward prescriptive advice on problem-solving, urging readers to embrace "thinking like a child" through experiments and quitting bad strategies, based on Levitt's incentive models.[60] A 2015 compilation, When to Rob a Bank: ...And 131 More Warped Suggestions and Well-Intentioned Rants, aggregated blog posts into thematic clusters, reflecting the series' maturation into a multimedia ecosystem while preserving its core empirical skepticism.[60] By 2025, the series had sold tens of millions globally, influencing popular economics discourse, though Levitt has noted in interviews that sequels demanded balancing accessibility with analytical depth to avoid diluting original rigor.[28]Podcast and Media Presence
Levitt hosts the podcast People I (Mostly) Admire, produced by the Freakonomics Radio Network, in which he interviews accomplished individuals on their personal motivations, decision-making processes, and unconventional insights into success and failure.[61] The series emphasizes revealing, often contrarian conversations that probe guests' obsessions and life choices, featuring figures from economics, science, policy, and entertainment, such as economist Dambisa Moyo and game show host Yul Kwon.[61] Launched in 2020, the podcast has released over 100 episodes by mid-2023, with ongoing production as of 2024, and maintains a 4.6-star rating on Apple Podcasts based on approximately 1,860 reviews.[62] Episodes are distributed across platforms including YouTube, Spotify, and Amazon Music, often exceeding 45 minutes in length to allow for in-depth dialogue.[63] Beyond his hosting role, Levitt frequently appears as a guest on Freakonomics Radio, the flagship podcast hosted by co-author Stephen J. Dubner, discussing topics from child-rearing economics to behavioral incentives, with archival episodes dating back to the series' early years around 2010.[64] His media presence extends to public speaking and interviews, including a 2007 TED Talk analyzing the economics of drug dealing versus legal employment, which has garnered millions of views and highlighted data-driven myths about urban poverty.[65] Levitt has also featured in C-SPAN programming, such as a 2005 speech on applied economics, and conducted interviews promoting Freakonomics sequels, like a 2023 discussion on decision-making frameworks from Think Like a Freak.[66] [67] These appearances underscore his role in popularizing empirical economics through accessible, narrative-driven formats rather than traditional academic channels.Criticisms and Methodological Debates
Statistical and Interpretive Flaws
Critics have identified several statistical shortcomings in Levitt's analyses, particularly regarding omitted variables, coding errors, and overreliance on correlations interpreted as causation. In the seminal 2001 paper co-authored with John Donohue on the abortion-crime link, a key regression omitted state-specific year fixed effects due to a computer programming error, which artificially inflated the estimated impact of abortion legalization on crime rates; correcting this reduced the effect size for violent crime by approximately 75% and rendered it insignificant for property crime.[51] Levitt and Donohue addressed this in a 2004 response by arguing that alternative specifications, such as including prison population controls, restored robustness, but subsequent analyses, including those by Ted Joyce, found the hypothesis lacked support in earlier U.S. data or international contexts, attributing residual effects to model misspecification rather than causation.[49] Interpretive flaws often stem from Levitt's emphasis on novel incentives without sufficient robustness checks, leading to claims vulnerable to alternative explanations. For instance, the Freakonomics chapter on real estate agents, drawing from Levitt's 1997 study, concluded agents sell their own homes faster (by about 3 weeks) but at a 3.5% lower price due to reduced effort, based on Chicago-area data; however, this inference overlooks potential selection bias, as agents' personal sales may involve unique market knowledge or property types not controlled for, and fails to account for transaction costs or liquidity differences that could explain the patterns without invoking shirking. Similar issues appear in the analysis of crack gang economics, reliant on a single Chicago gang's financial records from 1989-1992, which represent a non-random sample of approximately 6,000 individuals; extrapolating nationwide trends from this limited dataset ignores heterogeneity in drug markets and risks ecological fallacy, where gang-level data misrepresents individual behavior. Levitt's detection of cheating in contexts like sumo wrestling and standardized tests has also faced scrutiny for statistical power and multiple testing concerns. The sumo rigging study used win probabilities for wrestlers on the tournament's final day needing a 8-7 record, finding suspicious patterns (e.g., 7-7 wrestlers winning 80% against 7-7 opponents but 40% against 8-6), but critics note the method's sensitivity to unobserved match quality or wrestler form, and the lack of falsification tests across non-critical bouts weakens causal claims of collusion. In school test cheating, Levitt applied z-score thresholds to flag anomalous score improvements, identifying patterns in Chicago data from 1993-2000; yet, without adjusting for multiple comparisons across thousands of classrooms, the approach risks false positives, as random variation could mimic cheating signals, a flaw compounded by post-hoc interpretive emphasis on incentives over data artifacts. These examples highlight a broader pattern where Levitt's instrumental variable or reduced-form approaches prioritize intriguing stories over exhaustive sensitivity analyses, potentially overstating empirical support.[38]Ideological and Policy Critiques
Critics from the political left have argued that Levitt's emphasis on individual incentives in Freakonomics reflects a neoliberal ideology that reduces complex social problems to market-like behaviors, sidelining structural factors such as inequality and institutional failures. This approach, they contend, portrays incentives as the primary driver of human action—"the cornerstone of modern life"—while downplaying the role of systemic barriers in perpetuating issues like crime or poverty, thereby justifying minimal state intervention and market-oriented solutions over comprehensive reforms.[68] The abortion-crime hypothesis has drawn ideological fire for its policy implications, with detractors claiming it endorses a utilitarian view of abortion as a mechanism for social engineering to curb crime rates, potentially echoing eugenic rationales by linking legalized abortion in the 1970s to the 1990s crime drop through the non-birth of "unwanted" children from high-risk demographics. Conservative analysts have highlighted how this framing undervalues traditional moral considerations of fetal life and overlooks evidence suggesting abortion legalization correlated with rises in illegitimacy and certain crimes, as argued in studies by economists John Lott and John Whitley.[69] On policy grounds, Levitt's minimization of policing innovations—such as New York's "broken windows" strategy under Mayor Rudy Giuliani, which he attributed only about 10% of the 1990s crime decline to—in favor of demographic explanations has been faulted for potentially misleading policymakers away from effective deterrence measures toward passive reliance on birth trends. This perspective, critics assert, ignores data showing sharper crime reductions in areas with aggressive enforcement, like New York's 65% drop despite modest per-capita police increases, and could foster underinvestment in law enforcement infrastructure.[69][70] Levitt's broader incentive-focused analyses, including on education (e.g., teacher cheating scandals) and real estate, have been critiqued for promoting a cynical, data-driven policymaking that prioritizes tweaking individual behaviors over addressing root causes like economic disparity, potentially entrenching status quo policies that benefit elites while appearing apolitical.[70][68]Publications and Bibliography
Academic Works
Steven D. Levitt's academic output spans applied microeconomics, with a focus on empirical analyses of crime, incentives, discrimination, education, and market behaviors, often employing natural experiments to establish causality.[71] His publications appear in premier journals such as the Quarterly Journal of Economics, American Economic Review, and Journal of Political Economy, reflecting rigorous peer-reviewed scrutiny.[72] By 2019, Levitt had produced over 140 research works, garnering more than 24,000 citations, underscoring their influence in economics.[73] Levitt's early scholarship emphasized the economics of crime, including deterrence and unintended policy effects. In "The Impact of Legalized Abortion on Crime" (co-authored with John J. Donohue), published in the Quarterly Journal of Economics in 2001, they argued that the legalization of abortion in the 1970s contributed substantially to the crime decline of the 1990s by reducing the cohort of high-risk youth.[74] This paper estimated that legalized abortion accounted for about half of the crime drop, using state-level variation in abortion rates post-Roe v. Wade.[47] Similarly, "Understanding Why Crime Fell in the 1990s: Four Factors That Explain the Decline and Six That Do Not," in the Journal of Economic Perspectives (2004), attributed the U.S. crime reduction primarily to increased police presence, incarceration, the waning crack epidemic, and legalized abortion, while dismissing factors like gun control or innovative policing.[75] Other notable crime-related works include "An Economic Analysis of a Drug-Selling Gang's Finances" (with Sudhir Alladi Venkatesh), in the Quarterly Journal of Economics (2000), which used internal gang records to reveal low net earnings for foot soldiers—averaging $3.30 per hour—highlighting organizational parallels to legitimate firms despite high risks.[76] Levitt also examined corruption and cheating, such as in "Rotten Apples: An Investigation of the Prevalence and Predictors of Teacher Cheating" (with Jacob Goldin), which analyzed standardized test anomalies to detect educator manipulation, finding higher rates in high-stakes environments.[77]| Year | Title | Co-authors | Journal |
|---|---|---|---|
| 2001 | The Impact of Legalized Abortion on Crime | John J. Donohue | Quarterly Journal of Economics 116(2):379–420[74] |
| 2004 | Understanding Why Crime Fell in the 1990s: Four Factors That Explain the Decline and Six That Do Not | None | Journal of Economic Perspectives 18(1):163–190[75] |
| 2000 | An Economic Analysis of a Drug-Selling Gang's Finances | Sudhir Alladi Venkatesh | Quarterly Journal of Economics 115(3):755–789[76] |
| 2004 | The Causes and Consequences of Distinctively Black Names | Roland G. Fryer Jr. | Quarterly Journal of Economics 119(3):767–805[71] |
| 2011 | Checkmate: Exploring Backward Induction among Chess Players | John A. List, Sally Sadoff | American Economic Review 101(2):975–990[19] |
| 2006 | White-Collar Crime Writ Small: A Case Study of Bagels, Donuts, and the Honor System | None | American Economic Review 96(2):290–294[77] |