Abstract :
[en] This PhD thesis traverses an array of research domains, predominantly focusing on deciphering and tackling societal issues through a socioeconomic perspective. Nevertheless, it is essential to regard each chapter as an independent study, as they engage with distinct research questions, necessitating diverse datasets and methodologies.
Chapter 1 – Understanding trends and drivers of urban poverty in American cities.
Published in Empirical Economics.
Urban poverty arises from the uneven distribution of poor populations across neighborhoods of a city. Over a span of four decades, we critically examine the trends and factors driving urban poverty in American cities. Our approach involves the utilization of several urban poverty indices that account for the incidence, distribution, and segregation of poverty across census tracts. Built on solid normative foundations, these indices provide a more detailed understanding than the concentrated poverty index. We analyze tract-level data to determine how demographics, housing, education, employment, and income distribution impact the levels and changes in urban poverty. Through a decomposition analysis, we differentiate between the effects of changes in the distribution of these factors across cities and changes in their correlation with urban poverty. Our findings highlight the significant role of demographics and income distribution in shaping urban poverty, a result that markedly differs when using concentrated poverty indices.
Chapter 2 – Urban poverty and the onset of the Coronavirus pandemic: Evidence from American cities.
This study empirically explores the extent to which urban poverty in American cities influenced the propagation of COVID-19 during the pandemic's initial phase and the impact of mobility restriction measures on this dynamic. Leveraging ACS data, along with mobility and confirmed case data, and accounting for potential bias from measurement errors and unobserved confounders, we ascertain that an increase in urban poverty by one standard deviation corresponds to an escalation of 0.55-0.7 COVID-19 cases per 100,000 residents at the county level. This represents roughly a quarter of the COVID-19 incidence reported in the median American city by the end of April 2020. Intriguingly, we observe that stay-at-home orders fail to curb the contagion, instead inadvertently accelerating it in cities where poverty is less uniformly spread across neighborhoods, attributing this to the underlying factors of urban poverty.
Chapter 3 – The pandemic's toll on domestic violence: Investigating the effect of COVID-19 public health measures.
This research examines the influence of COVID-19 public health measures on domestic violence, focusing on the evolution of their effects throughout the pandemic. Given the underreporting issues associated with domestic violence due to the health measures, we employ the relative trajectory of domestic violence-related Google searches across 31 countries as a proxy indicator of domestic violence. By integrating this data with the Oxford COVID-19 Government Response Tracker, I can harness the fine-grained timing and intensity of COVID-19 public health interventions across countries. The results show that the measures significantly contribute to the rise in domestic violence, with evidence of impact as early as two weeks post-implementation of lockdowns. While these effects gradually decline over time, successive variations in the stringency of the measures persist in influencing domestic violence for several months following their initial introduction. The research further uncovers that while economic support policies intensify the effects of these measures, diminishing compliance to the measures by individuals helps to alleviate them.
Chapter 4 – Psychotropic drug consumption during the COVID-19 pandemic in Luxembourg: Excess consumption and socio-demographic profile.
This chapter inspects the evolution in psychotropic drug purchases in Luxembourg amid the COVID-19 pandemic, contrasting it with the preceding trends. Leveraging extensive administrative data – incorporating quarterly purchases of psychotropic drugs from all non-hospital pharmacies in Luxembourg between January 2016 and December 2021 – and anonymized individual-level socioeconomic details from the Inspection générale de la sécurité sociale (IGSS), the study allows for a detailed analysis of the characteristics of the population, aged 18 to 79 years, that resided in Luxembourg in February 2020. The study considers factors such as sex, age, household size, household income, and employment status. Although the findings reveal no clear excess purchases of psychotropic drugs following the pandemic's onset, the usage patterns varied by medication type, showing a rise in antidepressant consumption but no marked increase, or even a decrease, in the use of anxiolytics, hypnotics, and sedatives. Additionally, the research surfaces notable disparities in drug consumption patterns across various demographic strata, with the most pronounced relative change seen among younger individuals. The study underscores the need for further investigations into the pandemic's repercussions on mental health across diverse demographic groups, even as it confirms the pandemic's impact on mental health in Luxembourg, primarily evidenced by increased antidepressant use.
Chapter 5 – Emotional barometers: Twitter emojis and emoticons as tools to gauge temperature's effect on mood.
Contrary to the prevalent notion that weather significantly influences mood and well-being, empirical findings have shown mixed evidence. This paper makes two key contributions. First, this study introduces a framework for gauging mood on Twitter by leveraging emojis and emoticons, addressing methodological concerns such as omitted variable bias and small sample issues. Second, the paper offers a fresh perspective on the weather-mood nexus using alternative mood data, with a specific focus on temperature. The data consist of geotagged tweets randomly selected among Twitter users in the United States in 2014, representing different types of profiles and capturing various short-term weather variations. Overall, the findings suggest that individuals are nonlinearly affected by weather conditions. This study provides a versatile, cost-effective, language-neutral, and instantaneous approach to analyzing involuntarily disclosed mood indicators on a global scale, offering a significant complement to conventional methods.
Chapter 6 – Commuting time and absenteeism: Evidence from a natural experiment.
This research investigates the effect of commuting time on absenteeism using a natural experiment. This relationship is notoriously difficult to assess without exogenous shocks to commuting and with the survey data typically exploited. The study uses detailed administrative data for Luxembourg to measure the impact on work absences of a temporary shock to commuting time caused by large-scale roadworks at the border between Belgium and Luxembourg. The roadworks affected the commuting time of cross-border workers from Belgium, leaving cross-border commuters from France as a natural control group in a difference-in-difference setup. The findings reveal a positive – but quantitatively relatively small – effect of commuting time on absenteeism, driven mainly by increased absences due to reported illness or family reasons. Male workers appear to respond more than female workers to the shock in commuting time.
