AN APPLICATION OF KRIGING TO ITALIAN MORTALITY RATES

In population studies, it is customary to calculate the mortality rates as the ratio of the number of deaths to the number of exposures to risk for a given set of ages and calendar years. The demographic data sets usually involved are sometimes affected by some problems, i.e. missing data or high volatility. In order to better manage these problems, we propose to represent the mortality process as a Gaussian random field and to use the Kriging methodology for building up mortality values. We analyze the quality and the flexibility of this approach into some demographic applications based on Italian mortality data.