Abstract: This paper investigates how criminals learn about criminal laws. It uses the case of a natural experiment in which sentences were drastically increased for a specific type of recidivism in France. In the short run, media coverage of the reform did not seem to affect criminal behavior. However, people who had first-hand experience of the law later committed significantly fewer targeted crimes, but the same number of non-targeted crimes. This pattern is consistent with a strategic reaction based on a better understanding of the reform. This learning effect does not spread easily. While co-defendants also learned, other criminal peers and defendants in the courtroom at the same time but for a different case did not.

We document judicial leniency on defendant birthdays across 5 million decisions. French sentences are 1% fewer and 3% shorter. U.S. federal sentences are 33% shorter in the day component of sentences (the month component remains unaffected). New Orleans sentences are 15% shorter overall. No leniency appears on the days before or after a defendant’s birthday. Federal judges using deterrence language in opinions, are unaffected, isolating the judicial as opposed to defendant channel. The effect is doubled when judge and defendant share the same race. Our courtroom setting rules out many models of social preferences with reciprocity motives.

Abstract: We study spatial variation in criminal sentencing. We show the existence of local sentencing practices varying widely even across geographically proximate areas. Using compulsory rotation of judges in North Carolina, we find that judges arriving in a new court gradually converge towards local sentencing practices. We provide evidence that convergence in sentencing corresponds to a process of learning about local practices, which is accelerated by the presence of senior judges in the district, and that these sentencing practices are correlated with local norms of behavior. Finally, we discuss the theoretical implications of these results for the optimal design of judicial systems. 

Abstract:  We add to our understanding of the optimal method of judicial selection by ex- ploiting an unusual feature in North Carolina: judges rotate location every six months. This allows us to identify the existence and source of sentencing variation over the electoral cycle. We show that when elections approach, felony sentences rise. This increase is found exclusively when judges are sentencing in their district of election, and only when elections are contested. When judges hear cases outside their home district, sentences do not significantly vary over the electoral cycle. Our results show that electoral sentencing cycles can be explained by strategic sentencing by judges in an attempt to please voters. The unique setting allows us to reject alternative behavioral or contextual explanations for the rise in sentences as elections approach. 

Abstract: This paper presents the results of a field experiment designed to reduce gender discrimination in student evaluations of teaching (SET). In the first intervention, students receive a normative statement reminding them that they should not discriminate in SETs. In the second intervention, the normative statement includes precise information about how other students (especially male students) have discriminated against female teachers in previous years. The purely normative statement has no significant impact on SET overall satisfaction scores, suggesting that a blanket awareness-raising campaign may be inefficient to reduce discrimination. However, the informational statement appears to significantly reduce gender discrimination. The effect we find mainly comes from a change in male students’ evaluation of female teachers.

Abstract: We study how local labor market conditions and information about jobs affect recidivism among former inmates. Our identification strategy exploits daily variations on new job vacancies and news coverage of job openings and closings at the county level, merged with individual-level administrative data on inmates released from French prisons. Overall job creations do not affect recidivism, but inmates released when more jobs in manufacturing are created are less likely to recidivate. We also show that media coverage of job creation reduces recidivism, beyond actual employment opportunities, suggesting implications for crime-control policies: information about employment contributes to reduce recidivism.

Abstract: This paper uses the universe of convictions in France between 2000 and 2003 to document the gender gap in sentencing. It reveals three main findings. First, during this period, and after controlling for very precise descriptions of the offences as well as other observable characteristics, women received prison sentences that were 15 days shorter, on average, than those of men who committed comparable offences. This represents a 33% decrease relative to the average sentence length in the sample (45 days in the entire sample). Second, this gender gap is also observed for mixed-gender pairs of criminals (one man and one woman) who are convicted together, on the same day, by the same person and for the same crime. Finally, this paper presents robust evidence that the gender gap is affected by the gender of the judge, but not that of the prosecutor. Using the evolution of the courts’ composition between 2000 and 2003, the results show that a one-standard-deviation increase in the number of women in the court decreases the gender gap by 10%.

Abstract: We explore how television broadcasting of unrelated criminal justice events affects sentencing. Exploiting as-good-as-random variation in news content before a verdict, we find that sentences are 3 months longer when the verdict is reached after coverage of crime. Sentence increase with media exposure to crime, not crime itself, and the effect tapers off quickly. Our results suggest that professional experience and expertise mitigates the effect of irrelevant external information. This paper highlights the influence of noise in the news cycle: media can temporarily influence decisions by changing what is top-of-the-mind, rather than signaling deeper changes in offending or societal concerns. 

Press: France culture, Ouest France, 20 minutes

Abstract: What is the effect of introducing jury members in criminal courts? While surveys regularly point out a demand by citizens for harsher punishment, the differences between surveys’ and real decisions’ conditions are large enough to cast a doubt on the results. The introduction of two jurors into a court composed of three professional judges in two French regions and for a subsample of crimes during sixteen months offers a good natural experiment. Difference-in-differences or triple-difference methods do not permit me to identify any major change in the probability of being convicted or in sentences given by a court including jurors. If some characteristics of the reform could partly explain those null results, they clearly go against the hypothesis of a major disagreement between professional judges and citizens when they have to make real decisions in criminal cases.

Why do women experience a persistent drop in labor earnings upon becoming mothers, i.e. a “child penalty"? We study a new mechanism: search frictions. We analyze data on job applications sent on a popular online platform linked with administrative data for 350,000 involuntarily unemployed workers in France. First, we highlight differences in job search behavior between mothers and similar women with no children. Mothers send 12.2% fewer job applications and are more selective regarding wage and non-wage amenities. Consistently, they have a lower job finding rate. Second, we analyze the exact time when applications are sent and highlight differences in the timing of job search. We find that mothers’ rate of applications decreases by 20.3% in the hours and days when there is no school. We also show that mothers responded to a re-form that introduced school on Wednesday by smoothing their search across weekdays and narrowing their search timing gap with other women. In a simple search model, we show that our results imply that mothers both face lower incentives and higher costs to search. We conclude that search frictions disproportionately prevent mothers from improving their labor market situation and contribute to the child penalty.

This paper examines the impact of prison connections on re-incarceration, using comprehensive data on cell assignments in France. It documents that having one additional cellmate with a drug-related conviction increases re-incarceration for drug crimes (+7.1%) in the year after release), while having an extra cellmate with a property crime conviction raises the probability of committing a property crime (+5.2%). The number of cellmates with other convictions has no effect, and other types of recidivism remain unaffected. Peers from prison also affect the place where offenses are eventually committed. Lastly, cellmates have more influence when they share similar characteristics.

What is the effect of incarcerating one member of a group on her criminal partners? I answer this question using administrative data on all convictions in France between 2003 and 2012. Data on joint convictions are exploited to identify 34,000 groups and build a 48-month individual panel that records later criminal activity and sentencing. Using event study designs and panel regressions with individual fixed effects, I find that the incarceration of a peer is associated with a 9%decrease in the conviction rate in groups of two individuals. Offenders who have the characteristics of leaders are not affected by their followers but exert influence over them. Lastly, I show that the effect derives from lower criminogenic behavior and not from a loss of criminal human capital or from better information on the risks associated with crime.