Piero Cocevar and Luigi Di Falco provide an international perspective from the Italian Actuarial Professions Working Party on Pensions and Annuities
Actuaries in the UK are very familiar with the presence of cohort effects in UK mortality data. However, despite similar mortality experience observed in Italy for both pensioners and the general population, the concepts of mortality improvements and in particular cohort effects have not been widely discussed within the Italian actuarial community.
The presence of cohort effects in Italian pensioner data is only one of many findings contained in the report published by the Italian Actuarial Profession's Working Party on Pensions and Annuities in July of this year. This article highlights a few findings of this study - interested readers may view the full report written in both Italian and English on the Italian Actuarial Profession website.
The Working Party analysed data provided by several public and private pension providers, including INPS (the Italian National Social Security Institute) and INPDAP (the Italian National Insurance Institute for Civil Servants), the main state pension providers. As a rough indication of the amount of data, the exposed to risk for 2009 alone covered 10 million lives or 80% of the over-65 population, with payments totalling over 140 billion Euros. INPS alone pays pensions to over 18 million people and has the second largest balance sheet in Italy, behind the Italian state.
Italians are living longer
Figure 1 shows the period life expectancy at age 65 for Italian pensioners from 2000 to 2009 by type of employment compared with the general population. Note that while the general population data includes pensioners receiving disability benefits, the pensioner data examined does not. The general population data would be expected to exhibit lower life expectancy.
Pensioner life expectancy has continued to improve in the last few years. Over the nine-year period, the life expectancy of retired public employees has seen an increase of 14% for males and 9% for females, compared with 9% and 6% respectively for the general population.
The data shows that there are clear differences in survival prospects by occupation at retirement. It can also be seen that the self-employed and public employees have a longer life expectancy in retirement compared with private employees.
The presence of cohort effects in the Italian population and in other European populations has been known for several years (an account is given in Richards et al, 2007, and Cocevar, 2007). The Italian population in particular has cohort effects very similar to those seen in the UK population.
Detecting the presence of cohort effects in data sets smaller than full population data has proved challenging in the UK due to insufficient amounts of data. For example, the mortality improvements used in the CMI projection model are derived from population data rather than from the CMI's own data. The dataset available for Italian pensioners contains a considerable number of lives and so analysis of the pensioner data is carried out to see if it was possible clearly to detect mortality improvements in that data to compare meaningfully with the general population.
Mortality improvements were calculating by applying the P-spline model (as described in CMI Working Paper 20, 2007) to the Italian pensioner mortality data. These results were then compared with the data for the general population. The results displayed in the heat map on the left-hand side of Figure 2 show mortality improvements for the Italian male population, where a cohort effect for 1930-1940 generation can be clearly detected. The diagram on the right-hand side of Figure 2 shows the same heat map as the left-hand side but overlapped by a box showing mortality improvements from only the pensioners data. It can be seen that the historic improvement in pensioner mortality is similar to that observed in the general population. This result is not surprising considering that the size of the pensioner data examined is a significant proportion of the population.
To gain further insight, the pensioner data was further segmented by the amount of pension paid. The diagrams in Figure 3 show the equivalent heat maps to the right hand side of Figure 2, but for pensioners receiving less than 1 200 Euros a month on the left and for pensioners receiving more than 1 200 Euros a month on the right.
It is evident that Italian pensioner mortality improvements have been driven by rapid improvements in mortality for more highly paid pensioners. If these cohort trends continue into the future, the cost of pensions will increase disproportionately for the more wealthy members in that cohort.
So what's next?
While the Working Party report does not aim to go beyond a statement of the key trends observed, some implications can be deduced. In the current regime, pensions are calculated by observing historic mortality data in the general population, without consideration of any potential future improvement, selection by occupation category or weighting by the amount of pension in payment. A degree of imbalance is embedded within the Italian pension system, which would be anticipated to lead to some mismatch between actual and expected pensions paid.
Since the improvement patterns observed in the Italian population are similar to the UK population, to the extent that the cohort effect in the two populations are driven by the same underlying causes, it would be reasonable to expect that some of the results and implications for an analysis of the Italian pensioner data might also apply to UK pensioner-only data. Analysis of this type in the UK has not yet been possible due to the lack of sufficient volume of data from private sector pension providers.
Perhaps some readers of this article might also be interested in carrying out the same analysis for other state pension providers.
Piero Cocevar is a reporting actuary at AdminRe UK
Luigi Di Falco is head of life and pensions at ANIA, the Italian Association of Insurers
CMI (2007). Working Paper 20: Stochastic projection methodologies: Further progress and P-Spline model features, example results and implications, The Faculty of Actuaries and Institute of Actuaries.
Cocevar, P. (2007). An analysis of recent mortality trends in the Italian population using penalised B-spline regression. Giornale dell'Istituto Italiano degli Attuari70,21-43.
Richards, S.J., Ellam, J.R., Hubbard, J., Lu, J.L.C., Makin, S.J. & Miller, K.A. (2007). Two-dimensional mortality data: patterns and projections. British Actuarial Journal13(III), 479-555.