Impact Research: How the National Institute of Statistics Uses the Work of Three Bocconi Scholars

THE ITALIAN INSTITUTE IS THE FIRST IN EUROPE TO SHIFT FROM A DETERMINISTIC TO A STOCHASTIC METHOD FOR POPULATION FORECASTING, ADOPTING A TECHNIQUE FRANCESCO BILLARI, EUGENIO MELILLI AND REBECCA GRAZIANI HAVE BEEN WORKING ON FOR NINE YEARS

Three Bocconi researchers are the driving force behind the latest National Institute of Statistics’ (ISTAT) revolution. At the end of April, ISTAT released the latest document on our country’s demographic future. For the first time ever, it was based on stochastic forecasts (CLICK HERE to download the document – in Italian) that were elaborated in collaboration with Bocconi researchers. ISTAT is therefore the first European institution to put aside the deterministic method used until recently. This is the result of a process ignited eight years ago by Francesco Billari, current Vice Rector of Human Resources at Bocconi, Eugenio Melilli (Department of Decisions Sciences) and Rebecca Graziani (Department of Policy Analysis and Public Management). “This is a remarkable outcome for us”, the latter says, “and an important recognition of the effectiveness of our method”.
 
Billari, Graziani and Melilli started working on population projections in 2009. These projections are traditionally calculated by national statistical institutes using a deterministic method which means that three scenarios – main, high and low – are based on different assumptions on fertility, mortality and migration rates. In Stochastic Population Forecasting Based on Conditional Experts Opinions (Journal of the Royal Statistical Society, 2012, doi: 10.1111/j.1467-985X.2011.01015.x), Billari, Graziani and Melilli have proposed a stochastic forecasting method based on expert elicitations (CLICK HERE to read Bocconi Knowledge’s piece about the article).
 
In this method, a questionnaire is submitted to demographers. They are asked to express opinions on future trends of fertility, mortality and migration. “This method” Graziani says “allows us to elicit their opinions on the variability of rates, their inter-temporal association and the association of different rates at different times, without asking them directly to quantify these rates, which would be difficult to do anyway”. The Bocconi researchers, together with Marco Marsili and Gianni Corsetti (ISTAT) created the pilot questionnaire that was used in the aforementioned paper. Two years later, Billari, Graziani and Melilli proposed an alternative method of combining expert judgments in Stochastic Population Forecasting based on combinations of Expert Evaluations within the Bayesian Paradigm (Demography, 2014, doi: 10.1007/s13524-014-0318-5).
 
In 2015, given the experimentation done and thanks to a funding by the Carlo F. Dondena Centre for Research on Social Dynamics and Public Policy, a new questionnaire was created. This is the one used by ISTAT. The institute has then elaborated results provided by the Bocconi researchers and released population projections by age and sex. For the first time, the user is provided with the level of likelihood that a particular future population value will occur. The ISTAT website now provides a main scenario, but also 75%, 80%, 90% and 95% confidence intervals for population statistics. “It brings remarkable added value”, Graziani says, “since these data are used to adopt high-impact policies, for instance in the field of welfare”.

by Claudio Todesco