Abstracts

P. Burauel, "Controlling for discrete unmeasured confounding in nonlinear causal models" (with F. Eberhardt and M. Besserve) Unmeasured confounding is a major challenge for identifying causal relationships from non-experimental data. Here, we propose a method that can accommodate unmeasured discrete confounding. Extending recent identifiability results in deep latent variable models, we show theoretically that confounding can be detected and corrected under the assumption that the observed data is a piecewise affine transformation of a latent Gaussian mixture model and that the identity of the mixture components is confounded. We provide a flow-based algorithm to estimate this model and perform deconfounding. Experimental results on synthetic and real-world data provide support for the effectiveness of our approach.

F. Serti, "Assessing the heterogeneous impact of economy-wide shocks" (with M. Duēnas, F. Nutarelli, V. Ortiz-Giménez, and M. Riccaboni)  Our paper presents a methodology to study the heterogeneous effects of economy-wide shocks and applies it to the case of the impact of the COVID-19 crisis on exports. This methodology is applicable in scenarios where the pervasive nature of the shock hinders the identification of a control group unaffected by the shock, as well as the ex-ante definition of the intensity of the shock’s exposure of each unit. In particular, our study investigates the effectiveness of various Machine Learning (ML) techniques in predicting firms’ trade and, by building on recent developments in causal ML, uses these predictions to reconstruct the counterfactual distribution of firms’ trade under different COVID-19 scenarios and to study treatment effect heterogeneity. Specifically, we focus on the probability of Colombian firms surviving in the export market under two different scenarios: a COVID-19 setting and a non-COVID-19 counterfactual situation. On average, we find that the COVID-19 shock decreased a firm’s probability of surviving in the export market by about 20 percentage points in April 2020. We study the treatment effect heterogeneity by employing a classification analysis that compares the characteristics of the firms on the tails of the estimated distribution of the individual treatment effects.

J. Sprenger, "Causal and counterfactual inference"   The question of how to make causal inferences is tightly connected to the question of evaluating counterfactuals. Specifically, in Causal Modeling Semantics (CMS, developed by Galles and Pearl 1998; see also Briggs 2012), counterfactuals are evaluated relative to causal models and interventions in these models. Two main questions are, however, still unresolved: (i) how to extend CMS from a fragment allowing for conjunctions of atomic interventions to a general theory of counterfactuals, and (ii) how to evaluate their probability. This work tries to make some progress toward both of these aims.

A. Dominici, "The limits and perils of gentle communication against vaccine hesitancy: an informational trial" (with E. Bilancini, L. Boncinelli, and F. Panizza)   Motivational Interviewing (MI) is a gentle communication technique effective in doctor-patient interactions and is now recommended by the WHO to reduce vaccine hesitancy. Nevertheless, it entails high training and implementation costs. We test MI in the context of a flu vaccine video informational campaign by conducting a survey experiment on a representative sample of 8310 Italians aged above 40. MI significantly improves the informant's perception in line with theory. However, it reduces willingness to vaccinate (-7.4%, equivalent to approximately 3 p.p.) and does not impact actual vaccination behavior. Causal forest analyses also reveal that despite an overall null behavioral effect, a minority who increase vaccination uptake after MI is selected in terms of worse health status, older age, and higher distrust of vaccines. Conversely, MI decreases the uptake of a minority with more vaccine-compliant positions at baseline. While MI can be helpful for specific groups who might be more used to interacting with health professionals, it could backfire when addressed to the general population: this suggests caution in pushing MI as a large-scale policy.

F. Chiaromonte, "COVID-19 in italy: characterizing pre-vaccine epidemic waves through Functional Data Analysis"  (with T. Boschi, J. Di Iorio, L. Testa, and Marzia A. Cremona)  We use data from 107 Italian provinces to analyze and compare mortality patterns during the first two COVID-19 waves before vaccines were introduced. These patterns are examined alongside mobility data, timing of government restrictions, and various socio-demographic, infrastructural, and environmental factors. Using Functional Data Analysis techniques, we document differences between the two waves, focusing on their magnitude and variability, and analyze the association between local mobility and mortality. We retrieve province-level differential mortality data from public sources, and local mobility data from Google (“Grocery & Pharmacy” and “Workplace”). This data is processed with smoothing splines to generate curves, which are then aligned through landmark registration. Mortality curves are clustered to identify distinct patterns, and regression models are used to evaluate the impact of restrictions, mobility and other factors on mortality patterns. Our analyses reveal significant differences between the two waves. While both were characterized by a co-occurrence of “exponential” and “mild” mortality patterns, the first had higher and more concentrated mortality peaks, while the second was more widespread and asynchronous. We find statistical support for the effectiveness of timely restrictions in curbing mortality, and a strong positive association between local mobility and mortality in both waves. However, the quality of available data poses limitations to our study. Despite limitations in data availability and quality, the use of Functional Data Analysis techniques allows us to effectively characterize and contrast the first two waves of the COVID-19 pandemic in Italy. Our study highlights the importance of timely restrictions for mitigating the pandemic’s impact. The significant association between mobility and mortality captured by our analyses adds to prior evidence on the role of mobility restrictions as a control measure during pandemics.