Ongoing research

Abstract: This paper examines the impact of ethnic fractionalization on public service use by exploiting a major constitutional reform in Kenya. Following an important period of inter-ethnic conflict, responsibility for local health services was decentralized to 47 newly created county governments. Crucially, this changed the ethnic composition of the administrative area responsible for healthcare, while leaving the composition of the local population unchanged. Using an event-study design, we find that use of public clinics for births increased significantly after the reform, but only in counties that were relatively ethnically homogeneous. We also find a significant increase in the correlation between county ethnic fractionalization and a range of other measures of public health service use. Using within-county variation to investigate mechanisms, we find healthcare use increases were concentrated among individuals of the same ethnicity as members of the new county government executives. Overall, the results suggest that more ethnically homogeneous sub-national jurisdictions can rapidly increase public service use.  

Abstract: We study the relative impact of custodial and non-custodial sentences on later crime and labor-market outcomes in Denmark, a country where detention conditions are particularly good. To do so, we take advantage of a large-scale reform of the Danish legislation implemented in 2000, whereby incarceration was replaced by a non-custodial sentence for most drunk-driving crimes, which represented a quarter of the custodial sentences inflicted prior to the reform. Our first key finding is that stakeholders anticipated the consequences of the reform: around the time of the reform, the number of cases tried dropped and the nature of the cases changed significantly. To measure the relative impact of incarceration, we therefore resort to a novel instrumental variable approach exploiting quasi-exogenous variation in the probability of being tried after the reform, and therefore incarcerated, based on the crime date. We find that incarcerated offenders commit more crimes and have weaker ties to the labor market after release. The pattern of results suggests that part of the explanation for this increase in offenders’ criminal activities can be found in their greater precariousness.  

Abstract: Choosing between vocational or academic education at the end of secondary education has important long-run effects, and is taken at an age where peers’ influence might be paramount. In this paper, we investigate the effect of neighbors’ track choices on 9th graders choices at the end of lower secondary education, in Paris. This question is central to understand the extent to which residential segregation can reinforce social segregation across vocational and academic tracks. We rely on neighbors from the preceding cohort in order to bypass the reflection problem, and use withincatchment- area variation in distance between pairs to account for residential sorting. We use a pair-wise model that enables us to study carefully the role of distance between neighbors, and to perform detailed heterogeneity analysis. Our results suggest that close neighbors do matter in track choices at the end of 9th grade, particularly so for pupils pursuing a professional track. This effect is driven by neighbors living in the same building, and is larger for pairs of boys. Our results also suggest that neighbors’ effects tend to accentuate social segregation across high school tracks.

Abstract: We study job displacement in France. In the medium run, losses in firm-specific wage premium account for a substantial share of the overall cost of displacement. However, and despite the positive correlation between premium and productivity in the cross-section of firms, we find that workers are reemployed by high productivity, low labor share firms. The observed reallocation is therefore productivity-enhancing, yet costly for workers. We show that destination firms are less likely to conclude collective wage agreements and have lower participation rates at professional elections. Overall, our results point to a loss in bargaining power.

Abstract: This paper explores the relationship between unemployment rate and crime victimization at the neighborhood level, using data from the French victimization survey. The very local nature of the data enables me to tackle the endogenous location selection issue: once I control for the characteristics of a larger area into which household select their location, the remaining variation of observables across neighborhoods within this larger area can be considered as exogenous. The contribution of this paper to the economics of crime literature is then twofold. First, I show that, at the very local neighborhood level, unemployment rate is an important factor explaining victimization. Second, I take advantage of the precise localization of the data to compare the effect of unemployment rate in the reference neighborhood and in adjacent neighborhoods. The results support the idea that criminals are mobile across neighborhoods for more serious economic crimes, but that petty crimes and vandalism do not involve any mobility.

Abstract: Residential segregation matters as it may lead to a socially inefficient equilibrium because of externalities. Crime is one of the main outcomes that has been shown to be strongly affected by peer and network effects. At the city level, a decrease in residential social segregation might lead to simple displacement effects with no net effect on crime, or to a decrease in crime due to peer and network effects. This paper studies the effects of a common national urban renewal program aiming at decreasing social segregation at the city level. An important feature of this program consists in demolishing deprived public housing buildings in the poorest neighborhoods to rebuild new ones on site as well as off site in other neighborhoods in the city. The other main feature of this program is to renew the remaining housing stock and public spaces in the targeted neighborhoods. The intensity of this program as well as the number of affected cities allow us to study the context in which such a desegregation policy leads to a net decrease in crime. To analyze the effects of this program, we use very geographically precise income tax data as well as crime data at the city level. Our identification strategy relies on variations in the year the policy started in each city and on the existence of credibly untreated cities. We show that the policy led to an overall decrease in residential income segregation in treated cities, and that this effect is strengthened when the predicted intensity of the demolitions increases. We are now finishing to clean crime data, which will be used in the main part of the analysis.

In this project, we aim to understand the dynamic effects of contact with diversity on trust and cooperation. In order to address this question, we will rely on the real-world setting provided by tutorial groups of economics students in a large European university. In this institution, first year students are allocated randomly to their tutorial groups, providing us with a natural experiment-type of setting where group diversity is exogenous. Using this setting, we will rely on experimental games and implicit association tests in order to measure students’ level of interpersonal trust and biases at several points in time throughout their freshman year. We will thus be able to investigate how attitudes and biases evolve over time, and whether they evolve differently for students in more diverse tutorial groups.

We have now designed the questionaire and games to be administered to students in a pilot stage, and we are currently going through the administrative process necessary before running the pilot (University approval and IRB authorization). 

Abstract: Although ethnic minorities face adverse labor market conditions, the pace of immigration has remained unchanged over the past decades. This issue is adressed here through a model in which disadvantaged minorities engage into the underground economy to overcome unemployment. More specifically, the present paper develops a dynamic model (based on Calvó-Armengol, Verdier and Zenou, 2007) in which agents from two different ethnic groups form a network through which they can exchange information on formal or informal job offers. Networks play a central role, as information about formal and informal job opportunities can only be obtained through word of mouth communication. The formal labor market discriminates against the minority group, which thus faces higher unemployment in equilibrium. However, I find that the minority group is more likely to undertake informal jobs than the majority group, hence avoiding too large unemployment rates. In addition, and contrary to the predictions of the benchmark model, tighter networks turn out to reduce informal employment. Numerical simulations finally allow us to confront various policies aiming at lowering undeclared employment: compared to an increase in the audit rate to detect informal jobs, anti-discrimination policies appear to be less costly in terms of unemployment.

Abstract: Using the French Labor Force Survey, which contains exhaustive information on close neighbors as well as employer administrative identifiers, we show that workers residing in the same block are more likely to work in the same firm than those living in nearby blocks, suggesting the presence of local referral networks. We provide several robustness checks against two main threats to identification: sorting and reverse causality. Finally, focusing on transitions, we show that referral networks are used most intensively for out-of-unemployment transitions rather than job-to-job transitions while most the existing work has focused on the latter.