Forecast by Asking Opinions
IN A FORTHCOMING ARTICLE BILLARI, GRAZIANI AND MELILLI FORECAST CHANGES IN ITALIAN POPULATION TREATING THE ITALIAN NATIONAL STATISTICAL OFFICE AS A PROVIDER OF EXPERT OPINION IN A STOCHASTIC POPULATION FORECASTING EXPERIMENTPopulation forecasts are crucial ingredients in long-range planning, both for government and for private institutions. Trends in population by age are needed to forecast the demand for education and to plan the provision of education at all levels. Similarly, forecasts of demographic dependency ratios are essential to the design and reform of social security systems. In the past, most population forecasts were based on relatively simple mathematical extrapolations of current trends, giving rise to so-called scenario-based population projections. Global as well as national population projections have been produced following this approach by various agencies. All scenario-based forecasts are produced by developing several variants for each of the three main determinants of demographic change (i.e. fertility, mortality and migration). Uncertainty is not incorporated in the projection method. The expected accuracy of the forecasts cannot be assessed: prediction intervals for any population size or index of interest cannot be computed.
More recently, stochastic (or probabilistic) population forecasting has received great attention from researchers. The main reason for the development of stochastic forecasting methods lies in the awareness that only in this way can forecasting uncertainty be fully and coherently managed. Still, stochastic population forecasting has not yet influenced most official forecasting agencies.
In Stochastic Population Forecasts Based on Conditional Expert Opinion (forthcoming in Journal of the Royal Statistical Society Series A Statistics in Society, doi: 10.1111/j.1467-985X.2011.01015.x), Francesco Billari, Rebecca Graziani (both Department of Policy Analysis and Public Management) and Eugenio Melilli (Department of Decision Sciences) introduce a method that lies in the framework of the so-called random-scenarioapproach and makes it possible to derive stochastic population forecasts on the basis of evaluations provided by experts. Roughly speaking, the stochastic population forecasting method theypropose proceeds through a series of subsequent expert-based conditional evaluations on thefuture of demographic components, expressed through summary indicators, given the valuesof the indicators at previous time points. Conditional evaluations given by the expert are in the traditional form of medium–low–high scenarios.
In the example of application, the three scholars discuss how toderive a stochastic version of the official population projections for Italy which was developed by ISTAT (the Italian National Statistical Office), i.e. they treat ISTAT as the expert providing information to inform the model. Their results show that, although the Italian population is likely not to undergo sizable changes up to 2030, considerable uncertainty exists when the time horizon is extended towards 2050. There is, however, less uncertainty in the fact that the elderly dependency ratio will rise. The importance of the migration component is, however, fundamental in determining population dynamics.
by Rebecca Graziani
