The starting points for this project are the demographic and economic tools that are used in analysing population aging and its consequences, namely stochastic population projections, models of stable populations, and economic models with life-cycle structures. Another starting point is household projections, especially their probabilistic form.
Probabilistic household projections
Probabilistic household projection (PHP) extends probabilistic population forecasts with a household dimension. At present, only three empirical applications of PHP models are known of. Improving on early attempts by Alders (1999, 2001) and Scherbov and Ediev (2007) who based a large part of their uncertainty distributions on intuitive grounds, Alho and Keilman (2008) computed a PHP for Norway with empirically estimated uncertainty. Their deterministic multistate projection combines fertility, mortality, migration, household formation and household dissolution. The projection results in expected values for population shares in six private household positions: a person is either a dependent child, lives alone, lives in consensual union, lives with marital spouse, is a lone parent, or is an unrelated adult in a private household. Observed forecast errors in an old household forecast are used to specify the statistical distributions of the household shares, and the way these develop over time.
The current project will produce first-ever PHPs for Denmark and Finland. Both countries have good register-based data on households. We will combine the predictive distribution of the population of a certain age-sex group from the UPE project (see Alho et al., 2006) with random shares that distribute that population over various household statuses. As in Alho and Keilman (2008), point forecasts of the shares are produced by the multi-state program LIPRO (van Imhoff and Keilman 1991). The project will move beyond the current state of the art in PHP by testing the method for different data situations, and estimating more realistic random shares models when the data allow this, and developing methods for estimating uncertainty parameters when the data are scarce.
Life-cycle models of household behaviour
Since the work of Auerbach and Kotlikoff (1987) the long-run consequences of tax and social security systems around the world have been analyzed quantitatively with dynamic general equilibrium models that feature overlapping generations (OLG). The original model has been extended in various directions. French (2005) and Määttänen and Poutvaara (2007) analyze individuals’ retirement decisions within models where individuals can partially self-insure against uninsurable idiosyncratic earnings risks by accumulating financial assets. There are still relatively few papers that take into account family position. Gustman and Steinmeir (2004) find that considering retirement decisions within a family, instead of just at the individual level, substantially increases the explanatory power of the model.
Two versions of life-cycle models are specified and calibrated. The first is a structural life cycle model for analysing retirement decisions within a family. The model has incomplete insurance markets, following French (2005). Such models are good platforms for analyzing how different social security policies affect individual welfare and welfare inequality because there is a realistic role for welfare increasing social insurance. We extend the model in two ways. We include family positions, like Gustman and Steinmeier (2004), but in our model couples may also divorce. We will also consider old age care within a family.
The second model will consider changes in family patterns due to increased immigration. The model is in many respects similar to the first model but differs substantially in transition probabilities. Immigrant families in the Northern European countries have different family patterns from the natives. This leads to changes in the average size of households and the average fertility rates. Such changes will in turn affect consumption, demand for government services and education.
Uncertainty in population aging projections
Stochastic projections of Social Security are now routinely being done in the United States. The pioneers have been Lee and Tuljapurkar (1994, 1998 and 2001). Europe has come later to the field, but in using stochastic models in policy analysis in connection with population aging Europe has been the forerunner. Most of this was done in EU’s 5th framework research project “Demographic uncertainty and the sustainability of social welfare systems” (DEMWEL), see Alho et al. (2005), Fehr and Habermann (2006) and Alho, Jensen and Lassila (2008). Building on Alho et al. (2005), similar approach in policy analysis has recently been applied in the U.S., see Auerbach and Lee (2006).
Aging projections rarely include considerations about household type changes and their effects. Yet there are several links between changing household structures and public expenditures, especially in the Nordic countries. Availability of an informal caregiver, particularly a spouse, is among the most important factors explaining variation in long term care expenditure growth in 15 OECD countries (Yoo, et al. 2004). A comprehensive account of household effects has not been previously done.
We use the probabilistic household projections to explore the economic consequences with spreadsheet models. We attach the average amount of labour supply, saving, or some other economic variable, in a specific age, sex, and household position, to the predictive distribution of persons in private households, specific for age, sex, and household position. Although the procedure is rather straightforward, the amount of data is vast and the number of dimensions makes it complicated.
Stochastic life-cycle models are calibrated to replicate the data concerning the economic issue considered, such as labour supply distribution over the life-cycle and across different household types at a certain period. The replicated distribution is based on fixed transition probabilities. Replacing these initial probabilities with changing transition probabilities yields a new distribution. Differences between these two distributional outcomes describe the economic effects of changing household structures.
Effect of migration on household structure
Migration has been seen as a possible means to decelerate population aging. Alho (2008) introduced a stable, open-population model in which cohort net migration is proportional to births. Studying the migration-fertility trade-off he showed that although migration can increase the population growth rate, which tends to make the age distribution younger, it also has an opposite effect because of its typical age pattern. The effect of the age pattern of net migration is captured in a migration-survivor function.
In a stable population, the age-structure is (unrealistically) invariant over time, but the useful contribution of the theory is to show how population growth interacts with survival to produce the age-structure. The extension of the theory to open populations allows us to describe analytically the effect of migration on the age structure. In the proposed research we will extend stable population theory further, to include household shares. Our goal is then to use the open population case, to describe the effect of migration on the shares, under simplifying assumptions.