Education differentials and Population Growth on Pension Systems

The accuracy of mortality forecasts shapes the government decisions on expenditures. For instance, overestimation of future mortality affects fiscal accounts because more people staying alive may require access to health services and public non-contributing pensions. Underestimation of fertility may distort the child dependency ratios and planned expenditures on education. Therefore, better projections are required for informed policymaking.

This proposal considers, first, reconcile two schools in Demography, the school of various mortality outcomes depending on educational attainment, with the approaches of forecasting mortality using Lee-Carter model. Second, calculate the impact of the forecast on public health and pension expenses. The inclusion of educational differentials on top of the traditional variables of sex and age will lead to improved projections because the level of education is closely associated with changes in mortality and fertility. We can define three types of risks related to the components of population change: mortality, fertility and migration.

The mortality risk has been widely studied. Lee and Carter (1992) proposed a statistical model used to forecast mortality. Despite being the most widely used in forecasting mortality, the model has been criticised and many improvements have been proposed (e.g. Bergeron-Boucher et al. 2017; Pascariu, Canudas-Romo, and Vaupel 2018). This proposal identified three further shortcomings: sensitivity to the length of series, modelling assumptions and inclusion of auxiliary variables. That is, the model does not forecast well without long time series and the results can be sensitive to additional data if short series are used. This is challenging as many countries do not have sufficiently long series, or the existing series have measurement problems. The model also imposes strong assumptions about the age patterns of mortality (Girosi and King 2008). Finally, it has not been tested to what extent the inclusion of socio-economic and educational differentials may improve the forecasts (Lutz and Samir 2013).

Fertility risk and migration risk are a long-term risks not as widely explored as mortality. Literature typically focuses on fertility risk using an overlapping generation model, in which the number of children is deterministic, and abilities are stochastic, while both depend of background and education, (Cremer, Gahvari, and Pestieau 2011). Also, household's fertility decisions affect the rates of population growth and, thus, the sustainability of the pension system. On the other hand, (Pânzaru 2015) shows how migration in Romania becomes the only solution for labor market deficit.

Finally, in my preliminary analysis, I have demonstrated that the models show improvements in mortality and fertility forecasts by adding variables like socioeconomic differentials. In the proposed research, I will test further the relevant and available variables such as education and expenditure (as a proxy for wealth) to forecast mortality, fertility and migration, to further study their effects on pensions. A preliminary result of my work suggests that there are changes in mortality rates according to education changes.