QUANTIFYING THE INFLUENCE OF DEMOGRAPHIC TRANSITION

Transcription

1 QUANTIFYING THE INFLUENCE OF DEMOGRAPHIC TRANSITION ON PUBLIC FINANCES IN FINLAND Jukka Lassila 1 The Research Institute of the Finnish Economy (ETLA) October 6, 2015 Abstract: We study the effects of demographic transition on public finances. The novelty is that we quantify separately the effects of mortality, fertility and migration, using realizations of stochastic population projections as inputs in an economic model. Economic projections based on stochastic population simulations are often difficult to explain and require substantial time and attention from the users. Intuition about the reasons and mechanisms behind the results may not always be obtained. By using decompositions, we aim at gains on pedagogic aspects and intuition while giving up as little as possible on quantitative realism. Our example country is Finland, but the results should be relevant to all aging societies. The dynamics of the fiscal position is complicated by the fact that different vital rates influences different items of the public expenditure and tax bases with different timing. Mortality changes affect health and long-term care financing relatively quickly. Fertility variation has mainly expenditure effects during the coming decades, before the impact on tax bases grow, while migration affects the number of taxpayers already in the first decades. National tax and expenditure rules and fiscal strategies are of course important for the results. A key issue also is how typical lifecycles change as longevity increases. Key words: Demographic transition, public finance, stochastic population simulations JEL: H63, H68, J11 1 Jukka.Lassila@etla.fi. This research was funded by the Commission s Seventh Framework Programme (FP7-SSH /No ) within the MoPAct Project. I thank Eija Kauppi for carrying out the computations and making the graphs. I also thank the participants of the MoPAct workshop in Oslo, June 16, 2014, for comments, and especially Tarmo Valkonen for many detailed and useful recommendations for improvements. 1

2 TABLE OF CONTENTS 1. Introduction and motivation Forecasting demographic transition Changing population forecasts Stochastic population projections Decomposing the variation in stochastic projections From demography to public finances Demographics in the economic model Mortality and public finances Fertility and public finances Migration and public finances How do different demographic paths lead to different tax and public debt outcomes? Longer working lives and public finances Conclusions and future directions References Appendix: The economic model INTRODUCTION AND MOTIVATION This paper explores the economic consequences of population ageing in Finland. We ask how future fertility, mortality and migration affect public finances. We try to give insights on the quantitative significance of these different elements in demographic transition and on the timing of the effects. The results can rather easily be generalized to other welfare states with large agedependent public expenditure and high tax rates, and hopefully provide also lessons to ageing societies with less prominent role of the public sector. The strong conviction that the Finnish population is ageing is based on projections. Qualitatively there is no reason to doubt this view, it has a solid foundation in the current age structure. Quantitatively the issue is different. Populations have not developed as they have been projected in the past, and will not do so in the future. We account for this by using stochastic population projections. We use these projections as inputs in an economic model, whose outcomes then describe the economic effects. The results provide a probabilistic characterization of the impacts of the future demographics on public finances, assuming that the uncertainty related to vital rates is similar to the one observed in the past. 2

3 A traditional way of handling uncertainty would be to present alternative variants (e.g., a high- and a low-fertility scenario) in addition to a benchmark projection. One can then evaluate the effects of population ageing under alternative demographic variants so as to produce different economic scenarios. The method produces results that are easy to present and grasp. Thus it is pedagogically and intuitively appealing. The problems with this approach are that it is difficult to know what aspect of future demographics should be varied and how to do that in a manner that realistically acknowledges uncertainty. Results based on stochastic simulations are often difficult to present and grasp, and require more time and attention from the users. Intuition about the reasons and mechanisms behind the results may not always be obtained. We use here a somewhat novel approach of studying each of the vital rates separately with stochastic simulations, aiming at giving up as little as possible on quantitative realism while gaining on pedagogic aspects and intuition. In Section 2 demographic projections are discussed. In Section 3 the auxiliary population paths are used in the economic model. The aim is to describe how fertility variations affect some economic variables, especially tax rates and receipts, public expenditure items, public deficits and public debt, in different time periods, how mortality variations affect them and how migration variations affect them. These descriptions and the results from Section 2 are then used in Section 4 in trying to understand the relevant connections between future demographics, taxes and public debt. Section 6 discusses longevity and the length of working lives. Section 6 considers methodological implications and ponders on future directions of research. 2. FORECASTING DEMOGRAPHIC TRANSITION 2.1. CHANGING POPULATION FORECASTS All quantitative evaluations of the effects of population ageing must utilize long-term demographic forecasts. To illustrate how these forecasts can change substantially in a relatively short time, Figure 1 shows five forecasts, made between 2002 and 2012, for the future population in Finland. The total population was forecasted in 2002 to be about 5 million in The view has changed gradually, and the 2009 forecast is about 6.1 million in That means a 22 percent difference between forecasts made during seven years. The 2012 forecast for the total population is almost identical to the 2009 forecast. 3

4 Figure 1. Population in Finland, as forecasted by Statistics Finland Figure 2. Population in ages in Finland, as forecasted by Statistics Finland There were large and systematic changes also in the size of the age groups that dominate the working-age population, as Figure 2 shows. These changes can be traced back mostly to changing views on migration. The projections of the age groups of the elderly, in Figure 3, follow largely the changes in views of future longevity. These changes have affected empirical evaluations of public finances in various ways. There are more people working (good for tax revenues), more retirees (costly for public finances) and people live longer (good for individual welfare but costly for public 4

5 finances). The described large changes in population projections are of course not just a Finnish peculiarity. It is well known that long-term population projections have performed poorly in the history and have not become more precise in time (see e.g. Keilman et al., 2008). Figure 3. Population in ages 65+ in Finland, as forecasted by Statistics Finland Although the changes in forecasts for the number of working-age people and people aged 65 and over have been significant, they still show the basic feature of an ageing society: the share of the elderly is growing. The issue is quantitative the population is ageing but we don t know how much STOCHASTIC POPULATION PROJECTIONS Demographic uncertainty is often included in analysis by using stochastic population projections. Statistical methods of expressing demographic uncertainty have been developed by many researchers (see e.g. Alho & Spencer, 2005, Lee & Tuljapurkar, 1998). These methods quantify uncertainty probabilistically, based on analyses of past demographic data and the views of experts. Fertility, mortality and migration are considered as stochastic processes. The parameters of these processes are fitted to match the errors of past forecasts (see Alho, Cruijsen and Keilman, 2008). The uncertainty estimates related to fertility are based on a statistical analysis of the Finnish total fertility rate since The relative error of a naive forecast that assumes fertility to remain constant in the future was determined empirically. A naive forecast approximates closely the 5

6 medium forecasts made in Finland. For mortality, the analysis of uncertainty was based on the relative error of the naive forecast with data for 5-year age-groups from 1900 onwards. The naive forecast assumed that the recent past decline in mortality continues indefinitely. Alho (2002, p.9) explains how migration is dealt with in stochastic projections: The forecasting of migration differs from that of fertility or mortality in at least three ways. First, migration can be influenced by government policies to a higher extent than fertility or mortality. Second, although out-migration can be reasonably analyzed via out-migration rates, it is typically difficult to define a meaningful risk population for in-migration. Third, data on migration are poor even in a country like Finland that has a well-functioning population register. Because of these problems, migration forecasts are typically judgmental, and given in terms of the net number of migrants one expects. On the other hand, a probabilistic approach is well suited to the handling of the uncertainty of judgment concerning future migration. The primary difficulty is in finding a robust way to elicit judgments. After the processes for fertility, mortality and migration have been modeled, sample paths for future population by age-groups are simulated. In the following, we use a stochastic population projection with 500 simulated paths. As an example of what they predict, the stochastic projection for population in ages 65+ in Finland is presented in Figure 3. Half of the simulation outcomes in each period are in the shaded area around the median. 10 % of the outcomes are above the 90 % line and 10 % are below the 10 % line. Figure 4. Predictive distribution of population in ages 65+ in Finland 6

7 Since all vital rates (fertility, mortality and migration) vary stochastically, the simulated paths differ from each other in multiple ways in all age groups and all periods. Together they aim to provide a quantitatively realistic picture of uncertainties that grow with the forecast horizon. Used as inputs in economic models, they can be especially useful if and when they inform the users about the possibility of such future developments that are not the foremost in the minds of the planners and politicians. For examples of economic analyses using stochastic populations, see Alho, Jensen and Lassila (eds., 2008). There is, however, a downside of dealing with all vital rates simultaneously. The stochastic results are often difficult to present and discuss, and intuition about what is really relevant may not be obtained. Results are obtained but the pictures, so to speak, are blurred. A concrete example of a blurred picture will be presented and discussed in Section DECOMPOSING THE VARIATION IN STOCHASTIC PROJECTIONS To gain on pedagogic aspects and intuition aiming at while giving up as little as possible on quantitative realism, we use here a somewhat novel approach of studying each of the vital rates separately with stochastic simulations. To do that, every path of the stochastic population projection is decomposed into three auxiliary population paths. One auxiliary path describes the fertility variations, one the mortality variations and one the migration variations in the original simulated path. We first make a no-change version of Statistics Finland s 2012 population forecast. The no-change population starts from the prevailing age structure in the beginning of 2013 and assumes a constant inflow of persons below 20 years of age in the future, no variation in net migration, and the latest (2012) observed age-specific mortalities in all future periods. This population evolves so that new young generations are all of the same size, and the cohorts diminish in time by mortality rates that are the same as currently observed. In the auxiliary population paths, we start from the no-change projection and allow future vital rates to deviate from those currently observed. These deviations follow the assumptions used in the stochastic population projection. Using these auxiliary paths we describe how fertility variations affect the sizes of different age groups in different time periods, how mortality variations affect them and how migration variations affect them. We also describe how one age ratio, the ratio of 7

8 the number of people aged 60 years or more to the number of people in ages 20 to 59 years, varies in time due to variations in each of the vital rates. The horizontal axis in Figure 5 depicts the number of people in 20-year age groups in the original simulated population paths for Finland. The left-side box describes the situation in and the right-side box in The vertical axis in Figure 5 depicts the number of people in the auxiliary population paths. For every original simulated path, there are three points in the graph which all have the same x-axis value - the number of people in the age group in the original simulated population path but whose y-axis value is different. One of these three points (red) depicts the population in the age group when only mortality changes, one (green) when only fertility changes, and one (grey) when only migration changes. Concerning the age group 20-39, Figure 5 shows that only migration is likely to change its projected size in Fertility effects come later, and can be seen in the size of the age group in Migration-based variation is also strong in Mortality changes mostly occur in older ages, and thus cause only negligible variation in the size of this age group. In age group 40-59, migration dominates the variation both in and in In the latter period fertility still has significant effects. Mortality effects are still small, but can be seen in the Figure. Future migration is likely to happen in working ages. Thus its effects on age group in are likely to remain small. Mortality changes appear to be more important. In migration again dominates the variation. Future fertility effects come too slowly to be visible any more. In ages 80+, future changes in mortality are the main mechanism that affect the size of the population. Migration has minor effects and fertility has no effect. Figure 5 provides snapshots of the three features of demographic transition. Fertility changes, by definition, take time to affect older age groups. Mortality changes affect nowadays mostly old age groups. This has not always been so. Migrating people are mostly in working ages. This may of course change in the future. 8

9 Figure 5. Population in 20-year age groups 9

10 The variation in the sizes of different age groups described above generates corresponding variation in the age ratio, which is illustrated in Figure 6. The horizontal axis depicts the age ratio in the original stochastic simulations, and the vertical axis depicts the age ratio in the auxiliary population paths. In , the numerator varies by mortality (ages and especially 80+) and, to a lesser extent, by migration (ages 60-79), as explained in connection to Figure 5. The denominator variation in is due to migration (ages and 40-59) and, to a much lesser degree, to mortality (ages 40-59). In , mortality variation affects the number of older persons (the numerator) and fertility variation the number of younger people (the denominator). Migration affects both. Figure 6. The ratio of population aged 60+ to those aged Should we look further in the future we would of course notice that the uncertainty in each of the three vital rates would affect both the numerator and the denominator. In the next section, the auxiliary population paths are used in the economic model. The aim is to describe how fertility variations, mortality variations and migration variations affect selected economic variables, especially tax rates and receipts, public expenditure items, public deficits and public debt, in different time periods. 3. FROM DEMOGRAPHY TO PUBLIC FINANCES 3.1. DEMOGRAPHICS IN THE ECONOMIC MODEL Table 1 summarizes the conclusions drawn from Figures 5 and 6, and adds some economic comments that describe the Finnish economy, but are limited to the structure of our economic 10

11 model. While the economic channels in the model have counterparts in the real world, there may be important channels from demographics to the economy which are not included in the model. Table 1. Demographic transition and its economic reflections Finnish population aged Main demographic drivers 20 years from now 50 years from now Current population (under 40) and future migration Current population (under 10), future migration and future fertility Main economic roles of working-age population: Workers, savers, consumers, tax payers (state and municipal earned income taxes, consumption taxes, pension contributions, payroll taxes, capital income taxes), users of health care services and receivers of family transfers. Their children are using day care and schooling services. Note: Country-specific tax rules, changing work careers Finnish population aged 60+ Main demographic drivers 20 years from now 50 years from now Current population (40+) and future longevity Current population (10+), future longevity and, to a lesser degree, future migration Main economic roles: Some still working but mostly retired persons, consumers, tax payers (especially municipal taxes, consumption taxes, capital income taxes), users of health care and long-term care services Finnish age ratio 60+/20-59 Main demographic driver 20 years from now 50 years from now Current population structure, future longevity (numerator) and future migration (denominator) Current population structure, future longevity (numerator), future fertility (denominator) and future migration (both) Main economic consequences: Ratio of contributors to beneficiaries, tax and contribution rates, share of workers producing health and long-term care services Note: Country-specific tax and transfer rules and provision of publicly financed services, changing work careers 11

12 The 500 population projections and the 1500 auxiliary population paths are used as inputs in an economic model. We use a perfect foresight numerical overlapping generations model of the type originated by Auerbach and Kotlikoff (1987). It is modified to describe a small open economy and calibrated to the Finnish economy. The model consists of five sectors and three markets. The sectors are households, enterprises, a government, two pension funds and a foreign sector. The labor, goods and capital markets are competitive and prices balance supply and demand periodby-period. There is no money or inflation in the model. The model is described in more detail in the Appendix. Demographics enter the model through the household sector. There are overlapping generations of households, 16 adult generations living in each period (the unit period is five years). They all make lifetime decisions about consumption and leisure, according to the life-cycle hypothesis. The lifetime budget constraint says that discounted lifetime wage and pension income and discounted received bequest and transfers equal discounted consumption expenditure and the given bequest. People consider the possibility of early death by discounting future consumption and incomes by a factor that includes both the interest rate and the age-specific survival probability. Labor input is determined partly by exogenous assumptions and partly due to endogenous adjustments in the model. Hours of work are decided by households. Retirement occurs at the age of 65 at the latest. At ages below 60 an exogenous share, increasing with age, of persons retire due to disability. Average retirement age follows life expectancy, and in the model this is achieved by changing in each age group the share of those retired. The household variables, such as consumption, labour supply and wealth, are aggregated using population weights by age to obtain aggregate consumption, labour supply and household wealth. Similarly, aggregate wage bill, pensions, various taxes and transfers are obtained by analogous aggregation. Demographics also affect the labour markets indirectly. Part of the labour input is used to provide health and long-term care services. This share depends on the number of elderly people, weighted by per capita need of these services in different age groups. Similarly, part of the labour input is used for education work, whose demand depends on the sizes of young cohorts. These parts of the labour input, which vary from one population path to another, reduce the labour available for private production and affect the wages that balance the demand and supply of labour. 12

13 The growing number of people in old age and near death increases the demand for health and old age care, as described earlier. We assume that these demography-driven additional services are produced in private sector, but production costs are paid totally by the public sector. These services are produced using labor and intermediate goods as inputs. There is no productivity growth in the production. The shares of employees in private and public sector are kept constant. Real wage adjusts to equalize the value of marginal product of labor and labor costs in the production of private goods and services. The rest of the workers, who provide tax-funded services produced in private and public sector, earn the same wage. The model includes thereby the Baumol effects. Real wage growth varies between population paths, even though there is a common trend growth in total factor productivity in the production of the private good. Due to only partial indexation of pension accruals and benefits, this directly affects the replacement rates. Thus there would be considerable variation in all pension outcomes even without longevity adjustment of pensions. Public expenditures have strong connection to the age of individuals in Finland. Provision of public services is allocated mainly either to the early part of the life cycle (day care and education) or to the last years (health care and old age care). Similarly, income transfers are distributed mainly either to young families or to retired individuals. We assume that all income transfers (except the earning-related pensions) are fully indexed to wages because any other assumption would have dramatic consequences to income distribution in the very long term analysis. Other than agerelated public expenditure is assumed to grow at the same rate as the GDP. A majority of the revenues of the public sector are accumulated by income taxes, consumption taxes and social security contributions. Another noteworthy revenue source in Finland is the yield of the public sector wealth. This is important especially for the pension funds, but also the central government has substantial amount of financial assets. In the model the municipal sector, the public and the private sector pension fund and the national social security institute have their own budgets, which are balanced either by social security contributions or earned income taxes. The only exception is the state budget, which is balanced by borrowing until 2145, and after that by using a lump sum transfer to households. Earned income tax brackets are adjusted with the growth of the economy. The pension funds follow their current prefunding plans, and pension contributions are endogenous. Households are modeled to react to 13

14 the income and substitution effects of taxation and social security contributions. Firms choose the optimal amount of intermediate goods, investment and use of labour to maximise the price of their shares MORTALITY AND PUBLIC FINANCES In Section 2 it was shown that mortality variation affects mostly the number old people, those over 60 and especially those over 80 years. These people are often retired from working life, although more and more of them continue to work past earliest eligibility age as longevity increases. They pay earned income taxes especially to the municipalities. The state gets less income tax revenues from them but collects a lot of consumption taxes. The old also use more health care services per capita than younger people, and especially those over 80 are heavy users of long-term care services. Along with age, also proximity to death influences the need for services, as shown by several studies. This approach limits the increase in the needed public services when mortality rates decline. It is useful to acknowledge that life expectancy does not uniquely define what happens in different age groups in the population, and using expectancy calculated for one period leaves out variations in other periods. The horizontal axis in Figure 7 depicts the value of the variable in question e.g., in the top boxes, the revenue from the state s income tax - in the economic model simulation with all three demographic uncertainties present. The vertical axis in Figure 7 depicts the value of the variable in question in the economic model simulation with auxiliary population paths that include mortality variations only. Thus the top boxes aim to show the role of mortality variation in state earned income tax revenues, in relation to the role of all demographic variations in these revenues. There is also a colour code, describing the variation in mortality. The measure we use is period life expectancy of a 30-year old person (average of male and female). Simulation results from auxiliary paths where life expectancy is among the lowest quartile are denoted by green dots, whereas those in the highest quartile are marked with blue dots. Yellow dots represent the remaining 50 % of paths with life expectancies in the middle of the distribution. There are two main channels through which mortality changes directly affect public finances in our economic model. First, we assume that working lives on average become longer when people live longer. According to the estimates of Määttänen (2014), a 3-year to the life expectancy of a 30-14

15 year-old would extend working lives by 6 months, assuming that any health problems are likewise postponed by 3 years. This estimate has been used in our numerical OLG model in such a way that the change in life expectancy automatically affects the length of working lives in accordance with the ratio depicted. Longer working lives mean larger income tax bases and larger income tax revenue. Higher wage bills facilitate bigger consumption expenditure and thus increases in consumption tax revenues. Both revenue increases are clear in in Figure 7. The blue dots are on top, followed by the yellow ones and the green ones are on the lowest position. In these tax base effects are small, however, and other effects dominate. One is that deaths may occur in different age groups and cause different bequests and inheritance taxes, in a way our onedimensional longevity measure does not capture. The second main channel is publicly-financed health and long-term care. In the economic model, the expenditure on health care depends both on the number of people in each age group and the number of people who will die within the next five years. This is explained in more detail in the Appendix. Since mortality variation is the main factor behind the variation in the sizes of old age groups, as Figure 5 showed earlier, it causes substantial variation in health expenditure also (Figure 8). Because our modelling includes also the proximity of death as a cost factor, however, there are interesting dynamics. In the need for health care declines with longevity. This is a matter of timing, namely when do the big cohorts, the baby boom generations after WW2, die. If mortality declines slowly, or even increases, there will be lot of deaths in , and thus lot of health expenditure. If mortality declines rapidly, the deaths of the baby boomers are postponed beyond Longevity becomes cheaper, but only temporarily. In the baby boom generation is gone and the population age structure is more balanced. Living longer then increases the need for health care services, aggregated over all age groups. Long-term care is modelled similarly to health care but with different expenditure weights. The need for long-term care services increases rapidly with increasing longevity in In the connection to longevity is more blurred, the reason being the same as with health care: on average, there are additional long-term care costs that occur in the last years of life. Health and long-term care expenditure is financed by municipal taxes and grants from the state to municipalities. The grant is based on the age structure and morbidity of the population in the municipality. State tax rates are held constant, so variation in expenditure, caused by variation in mortalities, shows in variation in the municipal tax rate and the state debt. 15

16 Figure 7. Mortality and public finances: revenues and rates 16

17 The effect of mortality on pension contribution rate declines gradually, as more and more generations are affected by longevity adjustment of benefits (see Lassila and Valkonen, 2008). In our simulation results, both public debt and municipal tax rate are affected by mortality variation through tax base variation (length of working lives) and through expenditures (health and LTC costs). The net outcome of these opposing channels depends on more detailed features of the population structure and its economic consequences than our longevity descriptor, period life expectancy, is capable of grasping. Note that in all these figures we should look at the variation among the simulations, not where they are in the plane. Thus concerning the long-term care need in (Figure 8 top left), we note that variation in mortalities (or longevity) alone produces roughly similar variation that all demographic variation produce together (actually mortality variation here causes almost the whole variation, because future fertility and migration does not really affect the size of old cohorts in , as we noted from Figure 5, and old cohorts use the bulk of long-term care services). The fact that the dots are all below the diagonal is not important here, it is a result of how we created the auxiliary population paths. The variations were decomposed, the paths themselves were not decomposed so that the originals would be sums or products of the auxiliary paths. 17

18 Figure 8. Mortality and public finances: outlays and debt 18

19 3.3. FERTILITY AND PUBLIC FINANCES Analogously to Figures 7 and 8, the horizontal axis in Figures 9 and 10 depicts the value of the variable in question in the economic model simulation with all three demographic uncertainties present, and the vertical axis depicts the value of the variable in question in the economic model simulation with auxiliary population paths that include variations in births only. The colour code describing the variation in fertility is based on the number of births. For the period we use the number of births during , and for the number of births during Simulation results from auxiliary paths where the number of births is among the lowest quartile are denoted by green dots, whereas those in the highest quartile are marked with blue dots. Yellow dots again represent the remaining 50 % of paths. The main fact concerning fertility effects is that it takes decades before newborns improve public revenues, whereas expenditures are affected immediately. Childhood and youth cause outlays such as child allowances, schooling and education, and children also use health care services. Public revenues start increasing only after future cohorts have entered the labour force. Part of public outlays accumulate into the state s debt, and in our time frame this effect dominates the added tax revenues that will come later. Perhaps surprisingly, the top left box in Figure 9 shows that high fertility brings more income tax revenue to the state in , although none of those born in the future are yet working. The reason is that children need care and teaching, which require more workers in these services. That leaves fewer workers for the firm sector, which drives wages up. The model does not have inflation, but wages and domestic prices can change relative to the numéraire (import prices). The total wage bill is higher, and so is the income tax revenue. The same wage effect explains why pension contribution rates in 2030s are marginally lower with high number of children. Higher wages increase the wage bill, from which the contributions are collected, and as pension benefits are only partially indexed to wages, a slightly lower rate suffices to pay the pensions. In 2060s high fertility has already affected the labour force and increased the contribution base substantially, which facilitates distinctly lower contribution rate. By contrast, the municipalities run initially into problems with higher fertility. Even though they also benefit from the higher tax base, the increase in the expenditures dominate and they have to raise the municipal income tax rate the more the higher is fertility

20 Figure 9. Fertility and public finances: revenues and rates 20

21 Figure 10. Fertility and public finances: outlays and debt 21

22 3.4. MIGRATION AND PUBLIC FINANCES As in previous graphs, the horizontal axis in Figures 11 and 12 depicts the value of the variable in question in the economic model simulation with all three demographic uncertainties present. The vertical axis depicts the value of the variable in question in the economic model simulation with auxiliary population paths that include variations in migration only. The colour code describing the variation in migration is based on the size of population including future migration relative to the size of the population without any further migration. We term this relative migration. For the period we use relative migration in , and analogously for the period Simulation results from auxiliary paths where relative migration is among the lowest quartile are denoted by green dots, whereas those in the highest quartile are marked with blue dots. Yellow dots again represent the remaining 50 % of paths. In the following simulations, we have made a strong assumption concerning migration: the migrants are exactly as natives in all economic issues. As discussed in Lassila and Valkonen (2014c), this assumption is probably more accurate in the long run than in the short run, but alternative short-run modelling would be more complex and require data and knowledge that are not currently available. With the above assumption, the economic consequences of migration are straightforward. The more there are migrants, the more there are taxpayers and the larger are tax bases. The state s revenues are large, and the municipalities and the pension system manage with low rates. The migrants of course use also health and LTC services and receive pensions and other income transfers, but as most of public expenditure is generated during retirement years and most tax revenues during working lives, the public sector benefits from the expansion of population. This effect is similar to gains of expanding a pay-as-you-go pension scheme by adding young participants in the system. The exact gain depends on the lag between paid contributions and received pensions and the difference between the interest rate and growth rate of the economy. With immigration, educational costs are also saved. Thus, within the timeframe we consider, their effects on public finances are generally positive. 22

23 Figure 11. Migration and public finances: revenues and rates 23

24 Figure 12. Migration and public finances: outlays and debt 24

25 4. HOW DO DIFFERENT DEMOGRAPHIC PATHS LEAD TO DIFFERENT TAX AND PUBLIC DEBT OUTCOMES? Lassila, Valkonen and Alho (2014), in studying fiscal sustainability in Finland, used exactly the same fiscal strategy that is used in this paper. Welfare transfers and services are provided according to current Finnish rules and practices. Mandatory pension contributions adapt to pension expenditure. Aggregate health and long-term care costs depend on the population age structure and proximity to death. They are partly financed by municipal taxes, partly compensated for via block grant from the state. State tax rates are held constant, so variation in expenditure and tax bases causes variation in public debt. The total tax measure includes the pension contribution rate, municipal taxes and all state taxes. Debt and taxes are related to GDP. Figure 13 is from Lassila, Valkonen and Alho (2014). Each dot again represents the situation in one population path. The figure describes what kind of outcomes in the 2060s we can expect the given fiscal strategy to produce under demographic uncertainty. This section tries to explain, or at least give some intuition about how different demographic paths lead to different public debt and tax revenue outcomes. Figure 13. Public debt and taxes in Finland in 2060s The simulation setup used in Lassila, Valkonen and Alho (2014) was more complicated than in the current paper. The authors used stochastic population projections with embedded regular 25

26 demographic forecast revisions as inputs. This allowed them to separate, in each demographic outcome, the expected and the realized effects of population ageing on public finances. With this novel method, they studied different policy rules based on demographic point forecasts. Here we use a simpler approach, where model agents know future demographics. It is likely that this simplification does not alter our results essentially. Figure 14 shows the debt-tax outcomes with variation in one vital rate at a time. The grey dots denote the full demographic variation, whereas the coloured dots depict mortality variation in the top boxes, fertility variation in the middle and migration variation in the bottom boxes. Variations are measured as before: with period life expectancy of a 30-year old, with the number of births during and , and with relative migration. For each vital rate, the lowest quartile is denoted by green dots, the highest quartile with blue dots, and yellow dots again represent the remaining 50 % of paths. Fertility effects on public debt are strongly correlated in This is so because births affect at and before that period only expenditures such as child support, education costs and health care services. Child support is a transfer from the state to households, and its variation goes to public debt. Education and health care costs are mainly paid by municipalities and affect municipal tax rate, but partly included in the block grants from the state to municipalities, and thus affect the state s debt. The expenditure effects cumulate into the debt before, further in the future, fertility affects also labour supply and start bringing in tax revenues. It emerges from Figure 14 that fertility variation is the main source for variations in the debt/gdp ratio. Mortality variation has also rather effects on the debt ratio, via the need for health and long-term care services and the timing of the need, as explained in section 3.2. Migration affects mostly the size of the working-age population, and this is reflected especially in municipal tax rates and the pension contribution rate, which are included the total taxes/gdp ratio. The direct effect from migration to the state s expenditure and to the public debt comes mostly from health care expenditure, and is smaller than the effects coming from fertility and mortality. Considering the public debt/gdp ratio, dynamic aspects are important. All vital rates affect the state s income and expenditure. Debt accumulates from deficits, which vary in time. Fiscal policy strategies and rules are thus very important. Understanding the debt/gdp ratio is hampered by the fact that deficits or surpluses and GDP do not decompose fully to different vital rates, due to e.g. the firm sector. 26

27 Figure 14. Public debt, total taxes and vital rates 27

28 5. LONGER WORKING LIVES AND PUBLIC FINANCES When we simulated the effects of mortality variation (section 3.2.), the change in life expectancy automatically affected the length of working lives, in a ratio of 1/6. An additional 3 years to the life expectancy of a 30-year-old would extend working lives by 6 months. In this section, which draws heavily on Lassila and Valkonen (2014b), we briefly explore two alternatives. The first is simple: the ratio is zero, the length of work careers remain the same irrespective of changes in life expectancy. The second alternative describes the economy with a reformed pension system, where the link between longevity and working life length is stronger. In the retirement age reform, the earliest pensionable age is linked to the adulthood life expectancy. Adulthood is defined as having begun when the legal coming of age takes place at age 18. The pensionable age adjusts every year to changes in mortality so that it divides the expectancy for time lived as an adult to working lives and retirement years at the same ratio (roughly 2:1). If the life expectancy of a 63-year-old grows by just over six years over a period of 50 years, this method of linking would raise the pensionable age by four years. The earliest eligibility age for the part-time pension and the unemployment pathway are changed to the same degree as the pensionable age. Linking the retirement age to life expectancy affects the length of working lives. Raising the eligibility ages of the pensionable age, the unemployment pathway and the part-time pension by two years will extend working lives by 7 months in our model. The estimates are based on Määttänen (2014), who studied policies aiming to extend working lives on the labour supply decisions and income distribution of employees close to the earliest eligibility age for retirement. He used a stochastic life-cycle simulation model that depicts the decision-making of wage earners in different situations. Individuals are grouped in the model according to their age, gender and education. The decision to continue working, or to use one of the various exit routes from working life, is made with the insecure future in mind. Wage earners face the risk of losing their jobs, the risk of becoming disabled and the risk of a surprisingly long life. The size of these risks has been evaluated based on statistics. For instance, people with a low education have a higher disability risk and shorter average life cycle than others. The results of the base option (current retirement rules) and the alternatives (no change in careers, retirement age reform) on total taxes and public debt are shown in Figure

29 Figure 15. Longevity, work careers and public finances The results are both unsurprising and clear. Longer working lives makes tax bases bigger and bring in more tax revenues, which, under our policy assumptions, leads to lower municipal tax rates and lower pension contribution rates and to lower level of the state s debt. 29

30 6. CONCLUSIONS AND FUTURE DIRECTIONS It appears we can say something on the effects of different vital rates on public debt and taxation. They follow from the effects on population age structure how many people are in working-age population, how many are in younger and how many in older age groups. Migration affects mainly the working-age population, without delay, and the number of elderly further in the future. Mortality affects the number of the elderly, and fertility affects first the number of young, then those in working ages, and finally, far in the future, the number of the old. Age structure movements are slow to appear and can be foreseen to some extent, but with declining accuracy as the forecast horizon increases, because surprises in all vital rates will occur. While age groups are well defined, working-age population is a concept whose economic content varies. Schooling and education are main determinants of the entry ages to labour markets, and pension systems and retirement affect the average exit ages. Education and pension systems are not constant and may well vary with demographics. In our simulations, average retirement is affected by longevity changes. The role of the public sector in the different phases of an average life-cycle is obviously important for public finance outcomes. In our Finnish model, schooling, education, health and long-term care, and many transfers are financed by the public sector. The money is collected by income and consumption taxes and estate tax. Earned income tax brackets are adjusted with the growth of the economy. Private-sector earnings-related pensions are financed by contributions. Longevity affects benefits via cohort-wise adjustment factor. How the public sector is organized and what kind of fiscal strategy it follows is also important. In the model the municipal sector budget, is balanced by earned income taxes, whereas the state budget is balanced by borrowing. Our approach may bring to the fore some implications of our economic structures. An example here is the fact that retired persons pay more municipal taxes than state taxes. The importance of this varies with the number of retirees and the specifics of taxation and fiscal policies. Households are modeled to react to the income and substitution effects of taxation and social security contributions. Firms take taxes and contributions also into account when choosing the optimal amount of investment and use of labour. Demographic transition causes complicated 30

31 economic adjustments, since wages and prices balance supply and demand in the labour, goods and capital markets period-by-period. There are difficult measurement issues involved, especially concerning the determinants of the need for health and long-term care services. Our assumptions and the research they are based on, are explained in the Appendix. The assumptions are of course simplified but in some respects more complicated than in many evaluations of fiscal sustainability. Although some insights concerning the economics of demographic transition can be offered, a complete and fully intuitive explanation seems out of reach. The demographic transition itself is not very complicated and its evolution can be mentally tracked. But adding economics, at least in the form of the current model, inevitably complicates the outcomes. Concerning public finances, deficits which vary in time, and debt which accumulates from deficits, do not decompose fully to different vital rates, because of, e.g., the firm sector, and because the economy adjusts in many ways to different pressures and opportunities that demographics create. Stochastic projections v scenarios We must ask whether the basic idea of using stochastic population projections, rather than variants that describe the effects of one vital rate at a time, is lost when we decompose the variation in them. Economic assessments of ageing costs commonly rely on one basic demographic projection, supplemented by high and low demographic variants for sensitivity analysis. Lee and Edwards (2002, p. 11) offer four ways in which this scenario-based approach to assessing the uncertainty of forecasts of total population is seriously flawed. The first concerns assumptions about cross correlations between fertility and mortality, which by the nature of the method are usually either +1.0 or 1.0. The second is that fertility and/or mortality will always follow the highest plausible path or the lowest; long-run fluctuations like the baby boom are ruled out. Thirdly, the method provides inconsistent indications of uncertainty to differing outcome variables such as population size, life expectancy and old age dependency ratios. Fourthly, the method is intrinsically unable to assign probabilities to its high-low ranges. Using conditional stochastic projections, as we have done in Section 2, avoids the three lastmentioned flaws, up to a point. Thus it clearly is an improvement over the high-low variant method. But as it is always conditional on two vital rates, it does not provide a realistic picture of the possible future outcomes. The co-existence of fertility, mortality and migration uncertainties, their 31

32 dynamic nature, and their interplay in time, are factors that explain the scattered outcomes in Figure 13. A downside in conditional stochastic projections is still their complexity compared to the traditional way of handling uncertainty with alternative variants (e.g., a high- and a low-fertility scenario) in addition to a benchmark projection. One can evaluate the effects of population ageing under alternative demographic variants so as to produce different economic scenarios. These are simpler to present than e.g. our stochastic fertility results. Pöysti (2014) discusses the difficulties in presenting uncertainties in an understandable way, using a version of Figure 13 as an example: The communication of uncertainties and probabilities is a considerable challenge. An argument against the stochastic sustainability analyses is that they are too complex and very difficult to communicate to the general public and, also to the political decision-makers. However, from the communicational rights points of view such analyses and communications include the background information and access to multiple assumptions and knowledge about uncertainties that form the rational public debate perspective the public should be aware of. Perhaps a practical line can be drawn based on whether we are discussing new policy ideas, where qualitative and pedagogic scenarios are in order, or quantitative effects of well-defined policy proposals, where blurry pictures such as Figure 13 give more realism to the assessments. Using conditional projections as in this paper does not take into account forecast revisions. These could be included, at the cost more time-consuming simulations. No obvious counterpart for this method is found among the variant approach. The key novelty in Lassila, Valkonen and Alho (2014) is that, in a setup of stochastic population projections, revisions of demographic forecasts are included. This allows separating expected and actual outcomes in fairly realistic situations. Importantly, that allows studying sustainabilityenhancing fiscal policy rules based on expected demographics. This will make clearer the relationship between deterministic population projections and the process of policy decision making. While deterministic demographic projections have shortcomings which severely limit their usefulness, the stochastic approach with revisions shows that decision makers can adapt to errors in forecasts and mitigate their effects by updating their decisions. This is an area of research that has not received adequate attention in policy studies yet. 32