Research

Publications:

We investigate the medium-term health effects of prenatal exposure to moderate air pollution matching satellite P M10 concentration estimates with longitudinal data on hospitalisations and filled prescriptions for the universe of live-births in a large Italian region. We employ exogenous variation in PM10 in a multiple fixed- effects model to show that prenatal exposure to pollution leads to worse birth outcomes and to more hospitalisations and filled prescriptions in the first ten years of life, especially at the bottom health quantiles. Women do not leave their residence to avoid exposure to air pollution. We use these estimates to quantify the monetary cost of prenatal exposure to air pollution.  


Many empirical studies have focused on the direct health consequences of COVID-19 for infected individuals, but little attention has been dedicated to its indirect consequences for patients with nonrespiratory medical conditions. We employ a combination of machine learning and regression analysis techniques and administrative records on the universe of inpatient hospitalizations in Italy from 2012--2022 to investigate the effects of the outbreak on non--COVID-19 patients in one of the countries most acutely affected by the pandemic. A comparison of hospital- and population-level excess mortality suggests that 53.7% of COVID-19 deaths occurred outside hospitals. We interpret this as evidence of limited hospital resources, and we show that a higher number of hospital beds per capita is associated with a higher proportion of in-hospital deaths. We also document a 22.6% decrease in hospitalizations of nonrespiratory patients, a shift towards treating nonrespiratory patients with more severe conditions, and a conditional decrease of 0.5 days in the average stay length for nonrespiratory patients. We attribute these changes to fear of infection and hospital resource limitations, and we show that the drop in admissions was more pronounced in areas that were more impacted by COVID-19 and that had fewer hospital beds per capita. Our findings suggest that the pandemic's direct impact corresponds to a small fraction of the broader health losses in the population.

Empirical studies on the impact of weather and policy interventions on Covid-19 infections have dedicated little attention to the mediation role of social activity. In this study, we combine mobile locations, weather, and COVID-19 data in a two-way fixed effects mediation model to estimate the impact of weather and policy interventions on the COVID-19 infection rate in the US before the availability of vaccines, disentangling their direct impact from the part of the effect that is mediated by the endogenous response of social activity. We show that, while temperature reduces viral infectiousness, it also increases the amount of time individuals spend out of home, which instead favours the spread of the virus. This second channel substantially attenuates the beneficial effect of temperature in curbing the spread of the virus, offsetting one-third of the potential seasonal fluctuations in the reproduction rate. The mediation role of social activity is particularly pronounced when viral incidence is low, and completely offsets the beneficial effect of temperature. Despite being significant predictors of social activity, wind speed and precipitation do not induce sufficient variation to affect infections. Our estimates also suggest that school closures and lockdowns are effective in reducing infections. We employ our estimates to quantify the seasonal variation in the reproduction rate stemming from weather seasonality in the US.

Working papers:

Employing over 2 million Emergency Department (ED) records, we combine machine learning and regression discontinuity to document novel distortions in triage nurses' assessments of patients' conditions and investigate the short- and medium-run consequences for patients. We show that triage nurses progressively become more lenient over their shifts, and identical ED patients arriving just after a shift change are thus assigned a lower priority. We show that these patients receive lower levels of care and are more likely to demand further emergency care over the following months.  We conclude that distortions in nurses' initial assessments of urgency bias physicians' perceptions.

Recent years have seen an increase in the frequency and severity of climate change-related natural disasters and the rise of a global movement against climate change. These natural and socio-political events are accompanied by peaks in the prominence of climate change issues on media and social media. Existing evidence on how climate change salience on social media impacts real-world environmental behavior is scarce. Exploiting exogenous variations in the prevalence of climate-related content on Twitter, due to natural disasters and climate strikes, and electricity usage data from a large sample of 1.5 M Italian households between 2015 and 2019, we show that increases in climate change salience on social media causes a reduction in monthly electricity consumption. This effect is short-lived, as is typical of salience-induced behavioral changes. Sentiment analysis suggests that natural disasters and climate strikes are associated with emotions that are strong motivators for action, such as anger and positive feelings of solidarity. These results suggest that episodes that heighten attention to climate change may lead to actual behavioral change, but that their effect is only short-lived.


Other projects: