The research project is coordianted by Professor Francesco Bartolucci, Department of Economics, Finance and Statistics, University of Perugia (Italy). Five research units make up the research group: Bologna (PI: Silvia Cagnone), Firenze (PI: Alessandra Mattei), Perugia (PI: Francesco Bartolucci), Milano-Bicocca (PI: Fulvia Pennoni), and Roma-La Sapienza (PI: Paolo Giordani).
This webpage mainly describes the research activity of the Florence research unit. See the Project Web page for further information on the project.
The Florence research unit will focus on both methodological issues and substantive empirical research related to the identification and estimation of causal effects in complicated settings under the potential outcomes framework for causal inference. The following objectives will be the focus of the research:
Bayesian inference for causal effects with multiple outcomes in the presence of intermediate variables. We will focus on causal inference in studies where treatment comparisons need to be adjusted for post-treatment intermediate variables, which are potentially affected by the treatment and also affect the response. We will address these issues using the principal stratification framework within the potential outcomes approach to causal inference.
Bayesian inference for regression discontinuity designs (RDDs) . A fully Bayasian approach will be also used to analyse fuzzy RDDs and their generalizations. We will derive posterior distributions of causal estimands, modelling directly the distribution of potential outcomes for each sub-population defined by the joint potential values of an intermediate variable describing the treatment status around the threshold. The proposed methodologies will be applied for the evaluation of Italian university grants.
Estimating causal effects when data show a hirarchical structure. We focus on the following specific issues.
Causal inference in the presence of interference. We aim at developing innovative techniques to incorporate a-priori information on possible interactions among units and sensitivity analysis techniques to evaluate how different departures from assumptions affect the estimated causal effect of interest.
Contextual effects and school effectivness. We will discuss the challenges in conceptualizing and obtaining reliable estimates of the causal effects of the context, focusing on the ability of value-added models to capture institutional causal effects. These issues will be discussed and addressed in the context of a study on the effectiveness of Italian schools, using INVALSI data based on standardized tests and questionnaires on the familial background. The main objective of this study is to estimate the value added by each school, namely, the contribution of the school to the learning progress of their students.
Alessandra Mattei (PI), Department of Statistics, Informatics, Applications, University of Florence (Italy)
Bruno Arpino, Department of Political and Social Sciences, Pompeu Fabra University (Barcelona, Spain)
Leonardo Grilli, Department of Statistics, Informatics, Applications, , University of Florence (Italy)
Fan Li, Department of Statistical Science, Duke University (NC, USA)
Fabrizia Mealli, Department of Statistics, Informatics, Applications, University of Florence (Italy)
Barbara Pacini, Department of Political Science, University of Pisa (Italy)
Carla Rampichini, Department of Statistics, Informatics, Applications, University of Florence (Italy)
The research team will collaborate closely throughout with researchers of the
ESRC Research Centre on Micro-social Change (MISOC) - Institute for Social & Economic Research (ISER) - University of Essex, and CEPS/INSTEAD (Centre d'Etudes de Populations, de Pauvreté et Politiques Socio-Economiques/International Networks for Studies in Technology, Environment, Alternatives, Development). The research team will especially collaborate with Professor Stephen Pudney, the Director of MISOC, and
Dr Michela Bia, researcher at CEPS/INSTEAD.