I relate individuals' level of schooling obtained from three rounds of the Nigeria Living Standard Measurement Survey (LSMS) to spatial distance too university measured by pairing residential and university campuses, GPS coordinates. To identify the effect of the distance to the university, I exploit the theory of residential sorting to instrument residential proximity to the university. Specifically, I instrument distance to the university drawing on variations in households’ proximity to state boundary posts and neighborhood population density. The instrumental variable estimates show a negative and significant effect of distance, revealing that geographical constraints during teenage years represent a barrier to the subsequent human capital acquisition. Additional results from a difference-in-difference estimation strategy indicate that a large-scale establishment of universities had beneficial trickle-down effects by decreasing the intention to drop out of secondary school, supporting evidence of the role of geographical constraints in the accumulation of human capital in Nigeria.

The failure of the socialist utopia, the economic crisis, the increasing lack of primary resources, and the dictatorial turn have caused the third, most dramatic, and complex Venezuelan out-migration wave in the past decade. Drawing on exclusive and georeferenced survey data collected in Venezuela and providing information on 21,382 individuals, this presentation investigates the role of the police force militarization in the Venezuelan migration crisis of 2018. I find that the higher the level of authoritative violence—proxied by the share of homicides committed by the security forces—the higher the likelihood is for an individual to migrate. The effect is significant only among males with a lower level of education. Estimates that rely on the travel time from the capital to each state’s most populated city as an instrumental variable are robust to the inclusion of several households and environmental and sociodemographic characteristics, including the overall level of violence represented by the number of violent deaths per 100,000 inhabitants.

In this presentation, I use original data from Albania on mobility intentions and elicited risk aversion to provide causal estimates on this relationship. My identification strategy relies on the occurrence of two earthquakes during data collection that unambiguously led to upward shifts in risk aversion as shown in an article by Beine et al. (2021). While OLS estimates fail to capture a (negative) relationship between risk aversion and emigration intention, a control function strategy using the two earthquakes as instruments uncovers such a relationship. I argue that my results highlight a new channel through which risk preferences explain the trapped population phenomenon documented in the climate change and migration literature.

However, most analyses have focused only on female labor-force participation and gaps at the median. The Great Recession questions this approach, not only because of the major shift in male employment that it implied but also because women’s decision to participate seems to have been different along the distribution, particularly because of an “added worker effect”. This presentation uses the methodology proposed by Arellano and Bonhomme (2017) to estimate a quantile selection model over the period 2007–2018. Using a tax and benefit microsimulation model, I compute an instrument capturing the male selection induced by the crisis as well as female decisions: the potential out-of-work income. Because my instrument is crucially determined by the welfare state, I consider three countries with notably different benefit systems—the U.K., France, and Finland. My results imply different selection patterns across countries and a sizeable male selection in France and the U.K. Correction for selection bias lowers the gender wage gap and, in recent years, reveals an increasing shape of the gender gap distribution with a substantial glass ceiling for the three countries.

I want to assess to what extent temporary contract workers face higher risk of poverty than standard workers and how factors such as the family structure and the welfare states influence this risk. I study the implication of being under temporary contract on the risk of poverty in a longitudinal perspective in order to investigate further the association between atypical work and poverty not only by contract type and individual characteristics as done in two previous dynamic analyses—Debels (2008) and Amuedo-Dorantes and Serrano-Padial (2010)—but also by the households’ financial situation, the role of the partners’ earnings and benefits, while controlling for feedback effects of contract type and state dependency of poverty as done by Amuedo-Dorantes and Serrano-Padial (2010). In order to do that, I use two large panels for Germany (SOEP) and the U.K. (BHPS-UKHLS). Those panels allow me to cover extended periods: SOEP goes from 1984 to 2017 and the U.K. from 1991 to 2019.

My investigation essentially extends the traditional Oaxaca-Blinder decomposition technique (1973). It also proposes a counterfactual analysis along the entire wage distribution. The fundamental basis of this method is to set up a counterfactual distribution that allows us to estimate the male-female wage gap at each quantile. As shown in my empirical analysis, the gender pay gap does not refer to the differences in observable characteristics between males and females. It is rather the outcome of discriminatory practices against women.

Using French survey data, I estimate the effects of automation risk on workers’ well-being and mental health. Implementing the propensity-score matching method to solve the issue of endogenous exposure to automation risk, I find that workers exposed to the risk of automation have a higher probability to declare anxiety or depression and to have a low self-assessed well-being by five to six percentage points.

There are potential endogeneity concerns both for the probability of remaining employed during the pandemic and, conditional on employment, for the choice of work arrangements. My research design overcomes these problems by exploiting the occupational variations in the technical feasibility of remote working and sectoral differences in the legal restrictions on in-presence work. I find that remote working increases the probability of reporting feelings of sadness or depression. This effect is larger for women, respondents with children at home, and singles, as well as in regions with more restrictive containment policies and low-excess death rates. My results should alert policy makers to the potential adverse consequences of remote working for mental health in the post-pandemic situation.

To face this societal challenge, governments worldwide are increasing public expenditures focusing on healthy aging. The objective of these investments is to increase quality of life among older people. However, there is a lack of studies focused on understanding the extent to which a wide range of demographic, socioeconomic, and health characteristics are associated with quality of life in advanced ages. Therefore, the objective of this presentation is to explore the role of a variety of factors toward quality of life, with a particular focus on health. Structural equation modeling (SEM) is employed using Stata 16 to explore these associations, using data drawn from the Survey of Health, Ageing and Retirement in Europe (SHARE). Contrary to many studies that use self-assessed single-item questions or additive indices to measure unobserved concepts, such as health and quality of life, this presentation models such constructs as latent variables. Moreover, a minor contribution of this presentation is to employ standard statistical techniques using additive indices along with the main SEM estimation. As the theory predicts, estimates found with additive indices are downward biased compared with latent variables, but so far, there are no studies showing this empirical exercise. The overall findings suggest that nonpecuniary factors, especially physical health status and participating in social activities, play a larger role in enhancing quality of life in advanced age compared with pecuniary factors such as income and financial assets. Therefore, greater attention should be paid on non-economic factors to enrich quality of life among an increasingly aging population.

I illustrate how to obtain these treatment effects using GMM. Additionally, I show how some other proposed estimators for heterogeneous treatment effects can be fit using GMM.

Combined with a surprisingly small variation in unemployment rates over this period, some press articles have nicknamed the impressive German economic recovery the "second German miracle". In this presentation, I produce empirical evidence of the role played by short-time work in the "German miracle". By exploiting firm-level data, I show that short-time work programs should target firms facing huge financial constraints and difficult business conditions. To these conditions, short-time work programs can preserve employment during a crisis and allow a greater take-up afterward.

Based on a structural estimation of the labor supply and using the European tax-benefit microsimulation model EUROMOD, I simulate various scenarios of reform. The results show that the labor supply response to the introduction of an EMU-UI differs substantially across countries and depends on the design of the EMU-UI. I find that a flat EMU-UI scheme implies very strong disincentive to work but reduces poverty. On the contrary, a fully contribution-related EMU-UI system limits much more the distortions on the labor market in most countries but has limited effects on poverty and inequality. An EMU-UI with a common replacement rate, articulated with floor and ceiling amounts, would allow for upward convergence because it would strongly reduce poverty and inequality in several countries while not inducing important labor supply reduction.

The purpose of this presentation is to estimate the economic impact of nonpharmaceutical interventions in Italy by taking advantage of timing differences in their implementation across regions. To achieve this, I estimate one-way and two-way fixed effects models on a large sample of Italian provinces. I also isolate a set of well-defined natural experiments in which one region goes from a lower to a higher tier of restrictions, while a neighboring region remains in the lower tier, for which we can estimate difference-in-differences and continuous treatment models. Moreover, in order to observe whether the impact of restrictions has changed over time, I split the sample around December 2020 and replicate the analysis in each subsample. My case studies indicate that an Italian province moving from tier 2 to tier 3 in the system of restrictions can expect a fall in mobility of between 12 and 18 percentage points. Thus, I provide evidence of the negative effects of nonpharmaceutical interventions on economic activity. Finally, I provide some evidence that the effectiveness of nonpharmaceutical interventions in reducing mobility is likely to reduce over time, which has important policy implications.

First, I compute a time-varying fiscal space and show that complex fiscal space numbers resulting from solving the model provide crucial information in assessing fiscal sustainability; thus, they should not be ignored. I propose three different scenarios to deal with complex numbers in an empirical framework. Second, I provide a new determinant for fiscal sustainability: sustainable development. Using data from 24 OECD countries from 1998–2015, I find that sustainable development, proxied by an environmental, social, and governance performance index, has a robust positive impact on fiscal space.

The proposed framework has several advantages over existing de facto classifications. First, it offers a unified framework based on the most divergent classifications, the RR and LYS classifications, leading not only to a broader coverage but also to a broader spectrum of exchange systems. Second, it fits better with the known history of exchange rate regimes developments in the post-Bretton Woods era. Among other things, it brings an interesting nuance to the so-called hollowing-out hypothesis by showing that the evolution of de facto regimes—especially in emerging economies since the late 1990s—has essentially involved movement toward more tightly “managed” intermediate regimes and not a shift away from such regimes. As an illustration of the insightfulness of our classification, I empirically revisit the nexus between currency crises and exchange rate regimes. In addition to associate a higher probability of currency crisis with both intermediate and floating regimes, my classification, also displays better statistical performances than other classifications in predicting currency crises.

While this approach leads to highly efficient estimators when the distribution is Gaussian, GMM might yield more efficient estimators if the distribution is misspecified. For the SAR model, Lee (2004) proposes an alternative QML estimator that is less sensitive to the violation of the normality assumption. In this presentation, I derive an estimator that is highly efficient for skewed and heavy-tailed distributions. More precisely, I here assume that the distribution of the errors is a Tukey g-and-h (Tgh). However, because the density function of the Tgh has no explicit form, the optimization program for the MLE needs a numeric inversion of the quantile function to fit the model, which is a computationally demanding task. To solve this difficulty, I rely on the local asymptotic normality (LAN) property of spatial econometrics models to propose an estimator that avoids such a computational burden. My Monte Carlo simulations show that our estimator outperforms the ones available as soon as the distribution of the errors departs from Gaussianity either by exhibiting heavier tails or skewness. I illustrate the usefulness of the suggested procedure relying on a trade regression.

As an alternative, I propose to use partial linear models that are inherently interpretable. Specifically, this presentation introduces GAM-lasso (GAMLA) and GAM-autometrics (GAMA), denoted as GAM(L)A in short. GAM(L)A combines parametric and non-parametric functions to accurately capture linearities and nonlinearities prevailing between dependent and explanatory variables and a variable-selection procedure to control for overfitting issues. Estimation relies on a two-step procedure building upon the double residual method. I illustrate the predictive performance and interpretability of GAM(L)A on a regression and a classification problem. The results show that GAM(L)A outperforms parametric models augmented by quadratic, cubic, and interaction effects. Moreover, the results also suggest that the performance of GAM(L)A is not significantly different from that of random forest and gradient boosting.

Estimating a panel-ordered probit model, my results indicate that the impact of relationship lending (in the form of duration and intensity) on loan quality is different according to the firm’s profitability level. The intensity of the relationship lending positively (negatively) impacts the loans of high (average or low) quality. When intersecting intensity of a banking relationship with firm balance sheet indicators, the link between the intensity of the bank-firm relationship and loan quality is lower (higher) for good- (low-) quality firms. In addition, the length of the bank-firm relationship increases the probability of poor-quality loans. These results show that perverse and opportunist effects, in the form of strong moral hazard, are persistent for firms at different levels of profitability.

But does it work? I answer this question by exploiting the announcement by the World Health Organization (WHO) in October 2015 that processed meat products have been classified as carcinogenic to humans. My findings are based on two datasets, a representative consumer panel data and aggregated product-level market data of meat purchases in Israel. I apply two different methods of natural-experiments: regression discontinuity in time and difference-in-differences; both yield similar results. It turns out that the WHO warning caused a negative, sizable, statistically significant and persistent change in the equilibrium quantities of processed meats, which have dropped by 164 grams per household per month (-18%). To produce an equivalent demand reaction, prices of processed meat would have had to increase by 24%. The effect lasts for at least two years, long after media coverage has faded. The response is affected by income, ethnicity, and education. Low-income households and immigrants from the former USSR did not significantly respond to the announcement. Interestingly, I find that secondary education on the part of one parent is a necessary threshold for reducing long-term consumption. I evaluate two values: (1) the price increase that would have induced processed-meat consumption reduction equivalent to that of the WHO announcement at 3.3-4.05 $/kg and (2) the marginal expected cancer cost (through illness and mortality) evaluated at 1.1-4.0 $/kg. As the two ranges overlap, I conclude that the announcement successfully internalized the costs associated with cancer risks into consumers' considerations with respect to consumption of processed meat.

I use a French epidemiological cohort. Two types of health shocks are considered: lung cancer and smoking-related cancer. First, I exploit heterogeneity in the age of the individual at the moment of the parent’s health shock to analyze the influence of the cancer diagnosis on the offspring's smoking behavior. Second, I propose a Cox proportional hazards model to study the impact of the age of the offspring at the date of the diagnosis on the probability of quitting smoking. Finally, I use the individual's smoking history to build a retrospective panel and estimate an individual fixed-effects model to identify the impact of the parent’s diagnosis on the probability of smoking. In line with the existing literature, I find in all cases very limited impact of the parent’s health shock on the offspring's behavior.

I study firm effects in individual earnings dynamics for the Italian private sector population, using the covariance structure of co-workers for identification. My model allows for dynamics of both worker and firm effects, worker-firm sorting, worker segregation, and correlation among connected firms. While firms explain most of the earnings variance when workers are young, workers explain most over the life cycle. Sorting of workers across firms is substantial, especially for younger workers. Standard earnings dynamics models overstate the relevance of individual heterogeneity.

I compute concentration measures relying on new hires, finding that LMs aren’t on average concentrated, despite showing relevant heterogeneity. I then investigate the relationship between concentration and wages and employment, finding negative correlations. I then develop an IV strategy based on M&As to explore whether mergers increase concentration at a market-level and to find a reliable source of variation to identify their effect. First-stage estimates indicate that only mergers raise concentration significantly, while other events don’t. Estimated elasticities with different IVs range between -0.09 and -0.14 p.p for wages and between -0.68 and -0.77 p.p for hires.

Based on a speech by the president of France, Emmanuel Macron, I examine how political leaders can influence the willingness of a country’s citizens to get vaccinated by transmitting scientific insights into a clear and vivid message as well as by threatening unvaccinated people with future restrictions. In a difference-in-differences (DiD) framework, it is shown that a televised address of Macron increased the vaccination rate in France by roughly 5%. I test the robustness of this result by applying an event-study design. My findings imply that leadership communication is an effective weapon to change the beliefs of unvaccinated citizens and to overcome COVID-19 vaccination hesitancy.

Flexible models that allow substantial treatment-effect heterogeneity—with and without covariates—are easy to estimate in a variety of ways, including pooled OLS, doubly robust methods, and matching estimators. The framework can be used to both test for and correct for failure of the parallel trends assumption.