In the standard OLG model, individuals who differ only by age decide how to distribute their lifetime resources between consumption and savings along their lifecycle. The model we use differs from the standard model in several respects.⁴ It is a large-scale model⁵ that incorporates realistic demography⁶ (Bommier and Lee 2003), with households consisting of an adult who may live up to a maximum of 105 years, facing mortality risk as in Yaari (1965), and having a number of dependent children. The composition of each household varies by age and across cohorts based on the observed and projected fertility and mortality patterns. Household heads are assumed to be heterogeneous in terms of their level of education, which is randomly assigned to each individual at birth. The level assigned is one of three ISCED levels developed by UNESCO: ISCED 0–2 (lower secondary education or less), ISCED 3–4 (upper secondary education), and ISCED 5+ (tertiary education).⁷ Hence, in each year, individuals differ according to their age, fertility, mortality, and educational attainment.

Individuals derive utility from consumption goods, leisure, and the time they spend taking care of their children. Consumption goods can be produced in the market (using physical and human capital as input factors) or at home (combining time and intermediate market goods). In addition to consuming home-produced and market goods, the households in our model decide how much time they spend on work (in the market and at home), on childcare, and on leisure.

The government in our model provides public education, health care, pension benefits to retirees, and other public programmes; and distributes accidental bequests to surviving individuals. Both public in-kind and in-cash transfers are financed through a balanced PAYG system via consumption taxes, capital income taxes, labour income taxes, and social contributions.

We explicitly consider not only public transfers, but family transfers to children, including time transfers. Age profiles of all transfers are taken from the NTA and the NTTA datasets (Istenič et al. 2016, 2017). Using historical NA data and NTA-NTTA profiles, the model produces a complete age-time matrix of benefits and taxes for each year and birth cohort, which we then use to develop an analysis from both a cross-sectional and a longitudinal point of view. In particular, the model allows us to trace the annual evolution of expenditures affected by the age structure of the population, while taking into account the evolution of transfers to the past, the current, and the future population.⁸ These transfers affect the tax levels that are required to balance the public budget every year. The model therefore captures the annual tax adjustments that will be needed in the future to meet the projected increase in expenditures caused by the ageing of the population. Hence, we are able to gauge the overall sustainability of the public finances by obtaining a cross-sectional annual version of the tax gap indicator.

In this section, we focus on cross-sectional sustainability indicators. In particular, we analyse the annual evolution of public benefits, which we obtain by combining NTA profiles with the observed historical aggregates of public expenditures, and the tax adjustments needed to achieve an annual balance of a budget with PAYG constraints.⁹

Figure 1 shows the historical evolution (for 1970–2015) and the projected evolution of the ratio of public expenditures to GDP in Austria, France, and Spain through the 21st century. As populations age, an increasingly large share of resources will be devoted to the three main welfare state programmes (retirement pensions, education, and health care). In the three countries, the largest increases in expenditures will be for pensions, which are projected to account for more than one-fifth of GDP by the time members of the baby boom generation have reached retirement ages.¹⁰ Indeed, under current models (see Sánchez-Romero et al. 2017), social security contributions will adjust each year to guarantee that the level of pension benefits is maintained for all retired individuals, but only up to a maximum contribution tax rate of 35%. After this value is reached, benefits are adjusted downwards in order to balance the budget. This adjustment will prevent pension expenditures as a share of GDP from reaching even higher values in the second half of the century. Health care expenditures as a share of GDP are also projected to more than double in the three countries, reaching values of around 11–12% in 2100. Spending on the other main pillar of the welfare state, education, is not projected to change substantially over the 21^st century: i.e. education expenditures will remain at roughly 5–6% of GDP, with slightly lower values in Spain.

Not surprisingly, the programmes involving transfers to the elderly are expected to grow relative to the national income as populations age. In the case of health care, the age profiles have similar shapes in the three countries considered, with higher expenditures at birth and very high expenditures at older ages.¹¹ The projection that the youth dependency ratio will shrink as the old-age dependency ratio rises further confirms that total expenditures are highest at higher ages. Hence, budgets are expected to increase as a consequence of the demographic transition.¹² For education, it appears that the effects of the educational transition compensate for the decline in the number of children. Thus, the weight of education programmes is projected to remain rather constant over time.

It is assumed that all public programmes are financed through a balanced PAYG system. Apart from the pension system’s budget constraint, there is another key constraint on government budgets: namely, that most public expenditures are financed through general taxes (i.e. consumption, capital income, and labour income taxes). Although it would seem more realistic to account for the possibility of public debt, and to introduce debt into the model by considering the government’s intertemporal budget constraint, here we stress the demographic dependency of expenditures by imposing an annually balanced system in all programmes.¹³ Table 1 shows the weight of each public expenditure programme and the value of the different tax rates needed to balance the budget of the programme at three moments in time for the three countries considered: namely, in the year 2015; 30 years earlier (1985), when the ageing process had just begun; and 30 years later (2045), when the effects of the ageing process have spread throughout the economy.

Table 1 highlights the fact that the welfare state was developed later in Spain than in the other two countries. Thus, in 1985, the shares of pension and education expenditures in GDP were lower in Spain than in Austria and France (4.9 and 2.3 in Spain versus 9.1 and 4.1 in Austria and 8.0 and 3.4 in France, respectively). The differences in pension expenditures had shrunk by 2015 and are projected to reverse by 2045. This is largely because even though Spain has relatively low benefit levels, its population is ageing more rapidly than the populations of France and Austria. Moreover, because the educational transition started later in Spain and is projected to continue through the latter decades of the 21st century, public expenditures on education in 2045 are projected to be lower in Spain than in the other two countries. With respect to health care, Austria is the country with the lowest expenditure levels in both 2015 and 2045. This is mainly because in the Austrian age profile, the level of public health benefits relative to labour income is lower than it is in Spain and France.

Regarding tax rates, we assume that the weight of each tax rate (on consumption, labour income, and capital income) observed in the baseline year is held constant in the future, and that this tax mix is adjusted proportionately every year thereafter to fund government consumption. In the case of social security contributions, the differences between the three countries in 1985 again reflect the distinct stages of development of the pension system at that point in time: i.e. the contribution levels in Spain were less than half the levels in the other two countries. Based on the modelling assumptions, social security contributions are projected to reach the upper limit of 35% in Austria and Spain by 2045 (specifically, this limit is reached in 2034 in Austria and in 2041 in Spain); while in France, which is ageing more slowly, contributions are not projected to reach this limit until 2069. In Spain, consumption and labour income taxes were systematically lower in 1985 and 2015 than they were in the other two countries. However, the consumption tax is not projected to be lower in 2045, as larger adjustments are expected to become necessary in Spain as the population ages. Meanwhile, the capital income tax rate is highest in Spain over most of the study period, and is lowest in Austria throughout the period. In Section 4.2, the effects of future changes in the tax bases are discussed.

As Table 1 shows, tax rates are projected to increase in all three countries as the effects of ageing spread throughout the economy. This finding illustrates the challenges facing current public expenditure programmes, as it is clear that these transfer systems would become unsustainable if tax rates remained constant. The next section shows how these tax adjustments are projected to affect different generations.¹⁴

Figure 2 shows the present value at birth of pension benefits and taxes (i.e. social contributions) for cohorts born during the 21st century in Austria, France, and Spain. The three countries exhibit similar patterns for both series, with taxes reaching much higher values than benefits, and the gap between taxes and benefits widening for cohorts born further in the future. In France, this difference is always negative; which means that all generations have paid or are expected to pay more than they have received or are expected to receive over their lifecycle from the pension system. In Austria and Spain, however, there is a small positive gap between the present value of pension benefits and social security contributions for generations born during the 1900–1912 (Austria) and 1900–1940 (Spain) periods. These generations received a windfall gain from the public pension system. Because the system did not exist in the early 20^th century, these cohorts did not have to pay contributions during all of their working years. However, by the time these cohorts retired, the system was in place, and they were eligible to receive pension benefits. The cohorts born in 1900 in Austria and in 1924 in Spain received the largest gains, corresponding to 1.3% and 1.4% of lifetime labour income, respectively.

When analysing transfers from younger to older individuals, as in the case of the pension system, it is generally found that the present value of taxes paid is higher than that of benefits. This standard result is partly due to the effect of discounting.¹⁶ To understand the intuition of this effect, consider a transfer system in which an individual pays a fixed amount of money for 20 years during his/her working life, and then receives exactly the same amount of money during the following 20 years. When computing the present value at birth of taxes and benefits for this transfer system, we obtain a higher discounted value of tax payments than of transfer receipts, even though the monetary amount paid by an individual during his/her lifecycle is the same as the total amount s/he receives at the end of his/her life. This is because in an upward transfer system, tax payments occur earlier in life than transfer receipts, and are thus less discounted.

The finding that the gap between taxes and benefits is growing over birth cohorts is attributable to the share of beneficiaries increasing (due to population ageing) relative to the share of contributors, together with the assumption that the PAYG system is balanced annually by adjusting social security contributions, up to the maximum level of 35%. In an ageing population, the decline in the support ratio implies that each generation of young individuals has to support an increasing number of elderly people. Hence, the net present value of pension benefits becomes smaller, and the rate of return in a PAYG pension system is reduced.

Figure 3 displays results for a public education system financed with the tax mix mentioned above. Since education benefits represent a transfer from parents to children, whereas pension benefits represent a transfer from children to parents, the picture changes completely, and the lifecycle accounts are mostly positive and increasing for future cohorts. The cohorts born between 1907 and 1943 in Austria and between 1927 and 1977 in Spain pay more in taxes for education than they receive in transfers, and thus have a negative net present value at birth for this programme. The largest losses correspond to 1.6% and 4.4% of lifetime earnings for the cohorts born in 1933 in Austria and in 1961 in Spain, respectively. The largest gains in these two countries are achieved by the cohorts born in 2100, amounting to 30.0% and 34.4% of lifetime labour income, respectively. In France, cohorts born before 1968 are almost in lifecycle balance, with a net present value near zero. The largest gains from public education in this country amount to 25% of lifetime earnings for the cohort born in 2100. In all three countries, cohorts born further in the future increasingly benefit from higher education, which translates into higher wages.

The lifetime-discounted survival-weighted benefits received from the health system and the corresponding taxes paid are displayed in Figure 4. As is the case for the pension system, the present value of taxes for this public programme is above the present value of benefits received by the majority of cohorts and countries. This is because the age profiles for health expenditures, derived from the NTA data report, show substantially higher values for the elderly. The only groups who do not follow this pattern of a higher present value of taxes are a few generations born at the beginning of the 20th century in Austria and France; although the maximum positive gap found for these cohorts is only 0.4% of lifetime labour income. In all countries, the present value at birth of benefits and taxes increases at a faster pace up to the 1970 birth cohort, and more gradually for the later cohorts.

In the case of Spain, the evolution of both benefits and taxes is less smooth due to the effects of the ageing of the baby boom generation and the lower fertility rate in the 21st century. In Spain, the highest present values of benefits (18.0%) and taxes (24.5%), and the largest gap between them, are reached in the generation born in 2009, who are projected to suffer a loss equivalent to 11.3% of lifetime labour income. Spaniards born between 1990 and 2020 are projected to suffer the greatest losses, as they will have to pay larger tax bills to fund the health care needs of the Spanish baby boomers.

In France, the present value at birth of health benefits rises up to 15.4% of lifetime earnings for the 2100 birth cohort, while the corresponding value for taxes in 2100 reaches 22.2% of lifetime labour income. The generations born after 1980 suffer similar losses (the difference between taxes paid and health benefits received), amounting to around 7% of lifetime earnings. Earlier generations have smaller gaps between health benefits and taxes. The maximum loss is observed for the 2007 birth cohort (7.8% of lifetime labour income). The lower rate of ageing in France explains these results.

The Austrian public health system achieves the highest degree of intergenerational fairness. As Figure 4 shows, the differences between the benefits received and the taxes paid for each cohort are relatively small (and rather similar within cohorts), ranging from 2% to 3.3% of lifetime labour income for all generations born in the 1950s onwards. The present value at birth of taxes paid to fund public health in Austria reaches a maximum of 17.3% of lifetime earnings for the 2100 birth cohort, while the analogous figure for benefits is 14%. These values are lower in Austria than in the other two countries because the scale of the Austrian age-specific health care profile is smaller.

So far, we have analysed the net contributions of each generation relative to lifetime earnings for the three main pillars of the welfare state – pensions, education, and health care – that are of varying importance for different age groups. Next, we sum up the lifecycle accounts obtained for each public programme, and focus on the total net impact of those public transfers.

Figure 5 shows the present value at birth of net benefits received from the public sector for cohorts born in the 1900–2100 period in Austria, France, and Spain. In all three countries, the generations born at the beginning of the 20th century have the highest values. In general, these cohorts were not responsible for funding the expansions of the public pension, education and health care systems. The initial positive values for Austria are due to the modest net benefits obtained from the pension and health care systems. In the case of Spain, these early cohorts benefited greatly from the introduction of the social security programme, as they received pension benefits despite having contributed little or nothing to the system. Moreover, while these cohorts paid taxes to support the health care system, the combined net effect for these cohorts is positive. Positive values are observed in Austria and Spain for the birth cohorts up to 1913 and 1927, respectively.

Cohorts born in later years have lower net present values, mainly because benefit costs rose as the pension and health care systems were introduced and expanded. The decline is larger in Spain because it is amplified by the rapid change in the population structure and by the expansion of the educational system (see Figure 3). The downward tendency continues in all countries up to the cohorts born in 1963 (Austria), 1967 (France), and 1972 (Spain). For the cohorts born in the 10–20 years thereafter, the net effect increases because the benefits they receive from public education help to offset their losses attributable to pensions and health care. In Spain, where the expansion of the education system was delayed, an increase in the net effect is not observed. In all three countries, the subsequent birth cohorts have decreasing net present values, primarily as a result of the large increase in lifetime-discounted social security contributions and the modest decrease in corresponding pension benefits. The peak loss is observed for the 2003 birth cohort in Austria, amounting to 10.8% of lifetime labour earnings. The analogous figures for the other countries are a loss of 20.7% for the 2006 birth cohort in France and a loss of 27.6% for the 2009 birth cohort in Spain. For the later cohorts, relative gains in education and health benefits can be seen in all three countries, most notably in Spain; and the net impact of public transfers becomes decreasingly negative for each successive generation. For the cohort born in 2100, the net losses due to public transfers are estimated at 1.4%, 11.1%, and 2.5% of lifetime labour income in Austria, France, and Spain, respectively.

In line with the observed differences within countries in the present value of benefits and taxes of each public programme displayed in Figures 2–4, the shape of the total net effect by birth cohort follows a similar pattern in all three countries. In Spain, however, the evolution is much more volatile because the country’s demographic transition has been more dramatic. This pattern suggests that in Spain, the higher degree of redistribution of income from younger to older generations is induced by public policy. Another distinct feature of the evolution of the Spanish present values is that they occur later than in the other two countries. This pattern reflects both the later introduction of the welfare state and the more gradual expansion of public programmes in Spain. In particular, the positive effects of public education are postponed in Spain, even though they reach higher values once they are achieved. In France, there seems to be a greater degree of intergenerational fairness, as the fluctuations in net present values are smaller. However, as Figure 5 shows, public programmes provide lower returns for future generations in France than in the other two countries. Of all of the generations in all three countries, the Spanish cohorts born between 2000 and 2017 suffer the greatest losses, amounting more than 25% of lifetime labour income.

Overall, our results indicate that the positive effect of public education never offsets the negative effects of public pensions and health care for the generations born after 1927 in the three countries we focus on. This finding contrasts with the results obtained in Bommier et al. (2010) for the US, who found that cohorts born between 1947 and 2060 attain positive net present values. The differences in these findings can be explained in part by the lower benefit levels of the American pension system. Moreover, when comparing our results with those of Bommier et al. (2010), it is important to note that we included general equilibrium effects in our analysis. For example, the future interest rate used in the calculation of the present value at birth of benefits and taxes is not constant in our framework, but instead results from the general equilibrium model, and evolves according to the demand and the supply of capital in each year. The rate oscillated quite a bit in the past due to capital accumulation, which was driven in part by demographic change. From 2060 onwards, the rate stabilises at around 5%, with slight differences between countries. Higher interest rates give more weight to downward transfers like education, as they imply that the benefits people receive earlier in life have higher values than the taxes they pay later in life. The opposite pattern is observed for a transfer system that redistributes from young to old, such as the pension system. Therefore, a higher interest rate improves the net present value at birth for downward transfers, and worsens it for upward transfers.¹⁷

There is a large body of literature on pension system reform. Besides decreasing pension benefits and increasing contributions — both of which are reflected in our baseline scenario — most studies have argued for the postponement of the statutory retirement age. This policy appears to be very effective, as it affects both the revenue side and the expenditure side of the system by increasing the period when people are making contributions and decreasing the period when people are receiving pension benefits. This measure is likely to be especially effective if it is implemented at a time when relatively large numbers of baby boomers are retiring. Moreover, raising the retirement age seems desirable given that people are living longer and entering the labour market later as they spend more time in education (see Lee (2016)).

We have simulated the effects of delaying the retirement age from 65 to 67, and from 69 to 71. The increase in the retirement age is implemented step-wise from the year 2020 to the year 2040. Figure 6 reflects the variation in the pension expenditure to GDP ratio with respect to the baseline in the three counterfactual scenarios. After 2020, the year in which the reform is first introduced, public pension expenditures decrease in the three countries as a percentage of GDP. Further reductions are observed in the following 20 years as the policy change is being implemented. After 2040, the difference with respect to the baseline shrinks progressively into the distant future, although a negative net effect still remains. The pattern varies somewhat across the three countries. For example, the changes are the smallest in Austria, and are the longest lasting in France.

The sequences are similar for the three chosen retirement ages within each country, with the later ages always having a stronger effect. When retirement is delayed by two years, the ratio decreases up to 10–12% in the most critical years. Moreover, the ratio decreases up to 18–24% if retirement is delayed by four years, and up to 27–36% if the new retirement age is set at 71. The largest effects occur in Spain, where the percentage change in pension expenditure to GDP is further reduced in the 2070s if retirement is postponed to age 71. This pattern is attributable to the dissipation of the effects of the baby boom generation in that decade.

These changes derived from the postponement of the retirement age translate into variations in the net present value of public programmes presented in the previous section. Figure 7 displays the change in the total net present value of the combined pensions, education, and health systems for cohorts born from 1950 to 2100 relative to the baseline scenario. Although the main impact of a policy of delaying retirement is expected to be on the present value of taxes paid and benefits received from the pension system, several effects on pensions and contributions may offset each other, and could thus minimise the net result for this programme. On the one hand, delaying retirement implies that pension benefits are received for a smaller number of years and contributions are paid for a larger number of years. Both factors push down the net present value at birth of participating in the pension system. On the other hand, the general equilibrium nature of the model, together with the assumption that the system is balanced every year on a PAYG basis, entail a reduction of the contribution rates and a potential increase in benefits received. In this case, the result is an improvement in the net present value of public pensions.

The evolution of the variations in the different reform scenarios shown in Figure 7 also reflects the changes in the other public programmes (education and health care) due to general equilibrium effects. Overall, for any postponement of retirement, members of the older generations who are near the end of their working lives at the time the policy is implemented — roughly, those born between 1950 and 1980 — lose as a result of the reform. Future generations — or those who are very young when the reform is first implemented in Austria and Spain — also lose from the delay in retirement. However, the younger generations benefit, as they obtain a higher net present value at birth compared to the baseline. How large these changes in benefits and losses are depends on the number of years the age of retirement is postponed. In the most extreme scenario (retirement at age 71), the maximum losses for the earlier generations are between 2% and 3% of lifetime labour income, and are larger for the future generations (around 5% in Austria and France, and around 7% in Spain). The largest gains amount to 3.5% of labour earnings for the 1995 birth cohort in Austria; 5.4% for the 1997 birth cohort in Spain; and 7.1% for the generation born in 2009 in France.

When we compare the three countries, we see differences not only in the levels of maximum gains and losses with respect to the baseline, but in the time paths when these changes occur. In Austria, the generations who gain from the increase in the retirement age are those born from 1980 to 2006 in the age 67 scenario, those born from 1979 to 2009 in the age 69 scenario, and those born from 1979 to 2011 in the age 71 scenario. In France, the corresponding cohorts who gain from the reform are those born from 1985 to 2034 (spreading to 5–6 more years for each additional step in the retirement age increase). In Spain, the corresponding cohorts who benefit from the reform are those born between 1978 and 2013 (1979–2014 and 1979–2016, respectively). Hence, the benefits of the reform are spread out across more generations in France than in other countries. Recall that the maximum contribution rate for the social security system is set at 35%, and that once that limit is reached, any adjustments that are made must come from reducing pension benefits. Because France is ageing more slowly than the other two countries, this limit is reached later in France (and is never reached in the age 71 scenario), which explains the previous results.

While public expenditures mainly flow to children and the elderly, taxes are mostly paid by the working population; although some taxes are paid by children (consumption tax through parents) and the elderly (consumption and capital income tax). This suggests that one way to alleviate the pressures governments are facing as a result of the ageing process is to make changes to the tax bases. As Table 1 shows, the tax bundle used to finance public health and education differs to some extent in each of the countries in our study. Since this may be among the factors that influence the results we obtained, we developed two alternative scenarios that we use to investigate to what extent the choice of a different tax mix alters our results so far. To keep things simple, we focus on only one public programme, and implement a gradual reform by using one tax base as the funding source, instead of a mix of the three, as we assume in the baseline.

In particular, the policy reform we analyse consists of changing the tax base to finance future health care spending from the year 2020 onwards, and it is implemented progressively over a period of 20 years. We consider two policy options: financing future health expenditures using only taxes on consumption; and, alternatively, using only taxes on labour income. In both policy options, the specific tax rate chosen increases while the other two tax rates decrease. We decided not to use the capital income tax in these counterfactuals because of the high deadweight losses this policy would imply.

Table 2 reports the percentage change in all tax revenues (of consumption, capital income, and labour income) as a share of GDP and with respect to the baseline as a result of changing the tax base for financing health care. The results are displayed for the years 2050 and 2100; i.e. at two points in time after the reform has been fully carried out and the strongest effects have been realised. If health is financed by taxing only consumption goods, the corresponding tax revenues on consumption are projected to increase compared to the baseline scenario in which a mix of the three tax bases is used by 5.9% of GDP in Austria and France and by 6.4% of GDP in Spain in 2050. The analogous values in 2100 are somewhat higher for the first two countries and are slightly lower for Spain. Both the capital income and the labour income tax revenues decline relative to the benchmark case in the decades following the full implementation of the reform. The labour income tax revenues decrease from between 4.3% of GDP in Spain to 6.0% of GDP in Austria; while the adjustments in capital income tax revenues are smaller, ranging from 1.1% of GDP in Austria to 2.5% of GDP in Spain.

If health care is financed by taxing only labour income, the increases in tax revenues for this specific tax are slightly lower than in the first reform using consumption taxes in Austria, but are higher in Spain and in France in particular, reaching 8.2% of GDP in 2100. Tax revenues from consumption taxes are reduced by between 4.7% and 6.3% of GDP, with a stronger effect in France. The effects on revenues from capital income are similar to those in the previous scenario.

The effects on the total net present value at birth of pensions, health, and education are shown in Figure 8, which displays the changes with respect to the baseline. For all three countries, there are large gains for all generations if public health is financed exclusively through labour income. This is because a tax on labour income depletes more capital per worker than a tax on consumption. As a consequence, the interest rate increases, which gives more weight to transfers received earlier in life (education) than later in life (pensions and health care), and thus increases net benefits. In Austria, these benefits monotonically increase with birth cohorts, and the maximum gains amount to 11.7% of lifetime labour income for the generation born in 2100. In France, the benefits increase up to the 2048 birth cohort (with net gains of 13.5% of lifetime earnings compared to the baseline), and then remain rather constant, reaching 13.8% of lifetime labour income for the last cohort born in 2100. The evolution of net present values is somewhat different in Spain, where a maximum difference of 14% is observed for the generation born in 2011. Thereafter, the difference decreases to 10.2% for the 2042 birth cohort, increases modestly to 11.1% for the 2061 birth cohort, and then decreases again to 9.2% for the 2100 birth cohort.

Alternatively, if the tax base used to finance public health care consists of consumption taxes, the generations born between 1976 and 2023 in Austria gain compared to the baseline. Similarly, the 1972–2047 birth cohorts in France and the 1979–2030 birth cohorts in Spain benefit from this policy scenario. The largest gains are observed for the generations born in 2002–2004 (depending on the country), and are highest in Spain (11% of lifetime earnings) and lowest in Austria (4.9% of lifetime earnings). In France, the maximum gains correspond to 8% of lifetime labour income. The other generations lose, and the future generations in particular. The losses for the 2100 birth cohort amount to 4.6%, 1.6%, and 6.6% of lifetime earnings in Austria, France, and Spain, respectively. The losses for the previous cohorts are much smaller, reaching less than 1% (in absolute terms) in all cases.

One of the most distinguishing features of the model we used to derive the previous results is the degree of heterogeneity with respect to the educational level of the population. There are important differences in the educational composition of workers across countries, which generate diverse patterns of behaviour, and which in turn lead to distinct patterns of economic development.

Table 3 reports information on the actual and predicted educational composition for several birth cohorts in Austria, France, and Spain. We consider three levels of education based on the UNESCO classification: ISCED 0–2 (lower secondary education or less), ISCED 3–4 (upper secondary education), and ISCED 5+ (tertiary education). The pace of the educational transition differs in each country. The share of the 1980 birth cohort who have attained tertiary education is 44% in France, 24% in Austria, and 33% in Spain. The shares of the 1980 birth cohort with ISCED 0–2 and ISCED 3–4 educational levels are, respectively, 15% and 61% in Austria, 15% and 41% in France, and 28% and 39% in Spain. This pattern reflects the fact that the modern educational system was introduced later in Spain than in Austria and France. These differences between the three countries shrink by 2040, and it is assumed that by the end of the century, all individuals will attain at least upper secondary education. The share of the 2100 birth cohort projected to attain tertiary education is five points higher in France than in Austria, and is two points higher in Austria than in Spain.

In order to assess the impact of education on our results, we develop a scenario in which the educational attainment is held constant. In the experiment, we simulate the effects of stopping the educational transition by fixing the educational levels achieved by the 1980 cohort, who completed their education in the base year.

Figure 9 shows the impact of the fixed-education scenario on the net present value at birth of public programmes for cohorts born between 1920 and 2100 in Austria, France, and Spain. In all three countries, some cohorts realise net fiscal gains from this scenario (those born between 1952 and 2007 in Austria, between 1949 and 1995 in France, and between 1990 and 2018 in Spain), but the positive effects are small.¹⁸ If educational attainment is fixed at the 1980 birth cohort levels, these generations will have to pay lower taxes for the education of their descendants, but they will also lose some benefits, as they will have received less education. In this case, the former effect dominates. For previous generations, there is little variation with respect to the benchmark scenario. Future generations are projected to lose from the interruption of the educational transition in all three countries, as they will have lower educational levels, and thus will receive lower returns from education in the labour market. The largest losses are in Spain and France. For the cohort born in 2050, the loss in the present value of lifetime labour income is 2.7% in Austria, 3.6% in France, and 3.7% in Spain. For the cohort born in 2100, the corresponding loss is 5.2% in Austria, 6.9% in France, and 9.7% in Spain.