Carlos Andrés Hernández García Intra and Intergenerational Transfers

Transcription

1 Carlos Andrés Hernández García Intra and Intergenerational Transfers The Case of the Netherlands for the First Pillar of Pensions MSc Thesis

2 UNIVERSITY OF TILBURG Intra and Intergenerational Transfers: The case of The Netherlands for the first pillar of pensions Carlos Andrés Hernández García Company Supervisor: Lambrecht Van Eekeleen University Supervisor: Prof. Lex Meijdam Abstract Since the creation of the Old Age Pension scheme (Algemene Ouderdomswet, henceforth; AOW) in 1957 some important changes in its regulation have taken place. The purpose of this study is to analyse the redistribution of income for different individuals in the AOW taking into account these changes in regulation and different groups of individuals according to their gender, civil status and income. This comparison is made by using the Generational Accounts approach, initially presented by Auerbach (1991), in which the Net Present Value of the pension benefits received by every individual from the system is subtracted from the Net Present Value of the contributions paid to the system. In this study the database IPO, Documentatierapport Inkomenspanelonderzoek from CBS, Statistics Netherlands, has been used for some specific years after For the data of 2008 onwards the future contributions and benefits of the individuals will be projected according to the forecasts of the wage increases and pension benefits. A sensitivity analysis is also performed in order to check the changes in the Generational Accounts of the individuals with different discount rates and including the effects of the New Pension Contract. Document presented as part of the studies of the Msc. In Economics and Finance of Ageing at University of Tilburg. This document was realized during the internship at Sociale Verzekeringsbank, SVB. August 2011.

3 Table of Contents 1 Introduction General description Redistribution from young to old Redistribution between groups: inside generations AOW briefly historic description and regulatory changes Before the AOW Introduction of the AOW Current Tax system in The Netherlands Literature review How to measure the redistribution? Theoretical point of view Generational Accounts Advantages and disadvantages of Generational Accounting Micro-simulation models - MSM Description of the data and methodology Methodology of calculating the Generational Account Assumptions and procedure Descriptive statistics Results Gender: Men and women Single and Married Percentile analysis Interest rate changes Conclusions and final considerations References Appendix 1. Principal regulation changes AOW Appendix 2 AOW premium percentages, taxable maximum and maximum premium table (from Mahieu. R Master Thesis, 2007)

4 1 Introduction The purpose of this study is to analyse the transfers between individuals with different characteristics and birth cohorts in the first pillar of pensions in The Netherlands. This analysis is of particular interest because of the changes in the regulation that took place since AOW creation in Some of these changes include the abolition of the unequal treatment for married women in 1985 (who before that time didn t get an individual pension even when they were married, only if they were breadwinners), and the introduction in the same year (1985) of the Supplementary Allowance for pensioners with younger partners. All these regulatory changes affected different groups or cohorts of individuals in a different way. The main idea of the public PAYGO system is to redistribute income from young to old individuals, but these historic regulatory changes affected the redistribution of income for every cohort and group in different ways. Currently, there are some other regulatory changes on the table. The New Pension Contract includes the increase in the retirement age from 65 to 66 years (in 2020) and to 67 (in 2025), additional to a raise in pension benefits from 2013 to All these changes affect the individual s welfare in a different way and it is of interest to public policy makers to check how the average individual of every group (men, women, low income, high income, singles and married) is affected in order to evaluate the effects of the policies. The problem statement of this thesis is therefore: what is the pattern of the transfers in the first Dutch pension pillar? And what are the effects of some possible public policies on these transfers and consequently on the sustainability of the AOW? Some other questions that are intended to be answered with this study are: i) How is the redistribution within the first pillar in a pension system usually measured in the literature? ii) How are the transfers between and within generations affected by the aging population? What are the determinants of these transfers? iii) How will these transfers develop under population and regulatory changes? And iv) How will these changes affect the AOW? 2

5 The main idea of this exercise is to analyse these transfers using the Generational Account approach for the average individual using the information from income and pension benefits from CBS historical data. Some extrapolations are used for the years with no available data and for future years. 1.1 General description In The Netherlands the Old Age Pension scheme (Algemene Ouderdomswet, henceforth; AOW) is a Beveridgean system 1. This means that there is no direct link between salary and pension benefits. All individuals contribute to the system according to their salary and the benefits are flat. The benefits only depend on the household composition. The total population is composed by different groups of individuals and the net benefit of each person depends on their specific characteristics. In other words, what they put into the system as contributions is actuarially different to what they get as benefits from the system. The aim of this study is to analyze the redistribution in the first pillar pension for different kind of individuals. The differences in life expectancy and in contributions are taken into account in order to choose the groups for the study. For those groups with different life expectancies there are some discrepancies between the total contributions and the total benefits received. We consider the case of different groups according to: gender, household composition, income and age. In the case of gender, women and men have different life expectancies 2 ; also there is an increasing participation of women in the labour force (feminization of work) 3. Both women and men contribute to the system according to their salary, but they get the same amount of pension benefits since the system only differentiates between married people and singles. 1 A Beveridgean system is characterized by wage dependent contributions, a flat rate of benefits and low actuarial fairness. On the other hand a Bismarckian system is characterized by contribution - benefits linked to the salary, and high actuarial fairness. See Ponds, E (2010, pp ). 2 In 2010, the age expectancy for women in The Netherlands is years and years for men. 3 The proportion of women in The Netherlands labor force has increased from 28% in 1957 to almost 44% in 2009, according to the data from CBS. 3

6 By household composition, the number of single households has increased in the recent years in comparison with multi-person households 4. This case is of concern for the pension system in The Netherlands because the pension benefits depend on the household composition. According to CBS married persons live longer on average than single ones, and even longer than those divorced or widowed 5. By income, low (blue collar) and high income (white collar) workers make different contributions to the system according to their salary. However, the pension benefits are equal if they have the same household composition. Also, the life expectancy is different because it is expected that white collar workers have higher life expectancy than blue collar workers 6. According to Van Herten et. al. (2002) the life expectancy for highly educated individuals in The Netherlands is higher than for low educated ones, for males (4.9 years) than for females (2.6 years). Since there is no academically accepted / standard way to mark the difference between low and high income, the proposal is to make the analysis by percentiles. According to the CBS, the proportion of blue collar workers within the agricultural and industrial economy (over the total of workers) has decreased from 55% in 1960, to 28% in Consequently, the proportion of white-collar workers in services has increased from 45% in 1994 to 72% in Considering age, young people pay to sustain the elderly. Intergenerational solidarity is one of the main foundations of the first pension pillar. According 4 According to the CBS data, the percentage of one-person households (from the total of households) has increased from 33.4% in 2000 to 36.1% in Source: CBS Statline 5 In 2002 married, fifty year old men live 4 years more than those never married and 3 years more than that same age and divorced. For women the case is similar, fifty year old women live 2 years longer when married in comparison with widows or divorced. GB/menu/themas/bevolking/publicaties/artikelen/archief/2002/ wm.htm 6 The case of Germany Kunzler (2002) shows differences between mortality rates according to occupations, and Austria Klotz (2007) also addresses this difference. It is also addressed by Bovenberg (2007), p.p CBS Statline 4

7 to CBS data, the percentage of persons who are 65 years or older is going to increase from 15.6% in 2011 to 24.8% in Assuming a constant contribution rate, the PAYGO scheme has an internal rate of return equal to i that depends on the wage growth rate, x, and the population growth rate, n. In such a way that the internal rate of return of the scheme is. If the economy is dynamically efficient, young individuals find the PAYGO scheme less attractive for their retirement than a funded plan, because the real interest rate r is higher than the internal rate of return of the PAYGO i. If the economy is efficient then it holds: In this way, intergenerational transfers will depend on the parameters of the PAYGO scheme, such as the wage growth rate and population growth rate. For example, if population decreases in a lower rate n, then every time less young affiliates will pay for the old pension benefits. As a result contributions would need to increase in order to make the system sustainable and this would decrease the value of the Relative Generational Account of young affiliates, GA < 1. Where GA is equal to the present value of pension benefits divided by the present value of contributions: Also there are other intergenerational effects due to specific demographic changes, such as increasing longevity. For example, if there is a dynamic efficient economy and people live longer then currently is the case, young affiliates will need to pay for elderly benefits for a longer period. This increasing longevity will generate two effects on the Generational Accounts, for example for current young: by one side, a negative effect because they will need to pay increasingly more contributions to compensate the rise in longevity and pay for the current old benefit s. By the other side, there is a positive effect over the current young GA s because they will receive benefits for a longer period and this will increase the GA. Besides demographic changes that affect the GA there may also be some institutional changes that affect the GA. For example: changes in the contribution rate, changes (increases) in the retirement age or changes in pension benefits. Each one of those will generate different effects on GA s, depending on the sign and magnitude of 5

8 the change. In a mixed policy, for example, the net effect on the GA will depend on the mix effects: of an increase in the contribution rate (negative effect on GA) accompanied by an increase in the retirement age (positive effect on GA), and by an increase in pension benefits (positive effect on GA). The pension scheme in The Netherlands implies that contributions are different according to the income of the individuals, but a flat benefit leads to implicit transfers from one group to another. According to these differences in the characteristics that I have given above, there are some transfers between groups. The first type of transfer is the one from younger to older persons, as stated as the main goal of the pension systems (Barr, 2008). These are named Intergenerational transfers: from younger to older persons. Those transfers are justified explained by the young paying back the previous transfers in education, health care, alimentation, and others to the elderly. The other kinds of transfers are those inside every generation, called intragenerational transfers. From men to women, from high income workers to low income, and from married to singles. The purpose of this thesis is to analyze these redistributions at different points in time (according to main changes in regulation) since the implementation of the AOW in The years initially proposed for this are 1957, 1964 (65), 1990, 2002 and The implicit redistribution in the first pillar of pensions can be analyzed in two different aspects: from young to old, and between groups. In every case, this redistribution of income is a matter of concern because of demographic changes such as the decrease in the fertility rate and population ageing; and also because of the increasing proportion of public funds available to pay the AOW benefits due to a lack of sufficient contributions 8. These reasons are explained separately below. 8 SVB (2008). The Dutch State Pension, Facts and Figures. Amstelveen. p.p.13 6

9 1.2 Redistribution from young to old The redistribution from young to old (intergenerational transfers) in the first pillar is one of the main purposes of the system 9. The government also has social purposes with the PAYGO system in order to avoid poverty, to help the old when they are facing budgetary or health problems, as an insurance against old-age risks (like income risk and longevity), and to help individuals smoothen their consumption over their life cycle ( Bovenberg, 2010). The study of the patterns of redistribution is of crucial importance because of the increasing longevity of the population in the recent years. Since people are living longer, current elderly will receive pension benefits for more years than the elderly of the previous 20 or 30 years. In this way the transfers from young to old take an even more important role inside the system. As part of the demographic changes in the population, since there are fewer births, the current number of young workers is lower than before. If the benefits to the pensioners need to be paid for more years (because of the higher longevity), then the government needs more funds in order to pay those benefits for more time. The government currently has the option to increase the amount of public funds (which has been a fact in recent years in The Netherlands 10 ). By one side it could be argued that this is actuarially unfair to the new young generations in comparison with the previous generations because they will need to pay more contributions to the system than the previous generations did. But by the other side, because of the increasing longevity the new young generations will also receive benefits for a longer period. In terms of the labour market, an increase in the contributions (as a response of the increasing necessity of funds) will generate a negative effect in the labour supply since the mandatory contributions to the first pillar are considered an implicit tax. An increase in the tax will generate a disincentive to work (substitution effect between labour and leisure). 9 World Bank (1994), Averting the Old Age Crisis: Policies to Protect the Old and Promote Growth, World Bank, Washington, DC. 10 SVB (2008). The Dutch State Pension, Facts and Figures. Amstelveen. p.p.13 7

10 Since life expectancy has generally increased, in The Netherlands exist some policy proposals to increase the retirement age. Such proposals should generate effects in the sustainability of the system and also affect the redistribution from younger to older people. 1.3 Redistribution between groups: inside generations In order to consider the redistribution within generations, there is an implicit assumption that groups of individuals with equal characteristics are homogenous and constant over time. For example to analyse the transfers between single women and married women, the average ratio of (average pension benefits/ average income contributions) per age of all single women is taken, and compared with the ratio (average pension benefits/ average income contributions) for all married women. There are two implicit assumptions in this procedure; the first one is that the group of all single women remains constant over the time horizon (no marriages or divorces considered). And the second assumption is that the representative individual from the group is the one with the average income and pension benefits (in this way all members of the group have constant participation). In this case, the intragenerational transfers are going from a group of individuals with specific characteristics to another group with the same age but different characteristics. In the available literature it is common to find the analysis for cross-gender redistribution (between males to females) and intereducational redistribution (between groups with difference in their educational background). Those transfers are explained by differences in life expectancy, income and labour force participation 11. In the case of The Netherlands for example, in the collective pension funds, Bonenkamp (2009) finds redistribution from low to high-educated people and in this case this transfer can be seen as unintentional. For the PAYGO it seems that the transfer happens the other way around (from high educated to low educated). If there are transfers from the high to the low 11 Bonenkamp, J (2009). p.p.54. 8

11 educated this can be seen as a result of the disparities in the labour market and inequalities in the wages. Since The Netherlands system is a Beveridgean system, one can argue that these transfers are not actuarially fair for any of the individuals. By gender redistribution, those kinds of transfers are also unintended. The fact is explained because of the higher life expectancy of women, and fewer contributions because of the lower average income for women, and pregnancy and maternity leave. In that way, it responds to an external fact not managed by the pension system. The labour market has presented some changes since the 1950 s. The first one has to do with the composition sectors, and the second one with the feminization of work. i) The proportion of jobs for highly educated workers has increased considerably in comparison to the jobs for blue collar workers. ii) The proportion of women working has increased. There are different effects that affect the intra - gender transfers. On one side, since there are more women working nowadays, they make more contributions to the pension system and this decreases the men - women transfers. But also women tend to work more in part time jobs; this increases the transfers from men to women, taking into account that the benefit is flat. In actuarial terms there does not exist fairness for men or women. But it is also necessary to take into account that most women that work in part-time jobs are taking care of their children in the meantime. In that way, transfers can be seen as fair since they can be perceived as a payback to them. 2 AOW briefly historic description and regulatory changes The AOW is provided by the Social Insurance Bank (Sociale Verzekeringsbank, henceforth; SVB). There is no link between labour history (wages) and benefits. The sources of the General Old Age Pension come from affiliate contributions and public funds. In order to be insured, the affiliates need to fulfill two conditions: i) to reside in The Netherlands or ii) to not be resident in The Netherlands but liable to Dutch taxation (SVB, past present and future). 9

12 2.1 Before the AOW In the previous years before the introduction of the AOW there were some laws that introduced the concept of the scheme and were the first approaches to the actual system. From 1901 to 1956 there are two recognizable periods: The first period goes until In the first years of the century, the first social security regulation aimed to cover the workers against the bad situation resulting of the industrialization process. The second period comprises the years between 1945 and In this stage social security was established in The Old Age Pensions emergency Act was set up in Introduction of the AOW The AOW was introduced in The benefits have always been uniform, regardless of the previous income. The choice of criteria as residency and no income limit reinforce the principle of solidarity (also principle of the breadwinner). In order to check the implicit transfers between one generation and another, some specific years when changes in regulation took place are chosen 12 : In 1957, the AOW was created. The pension age was 65 years and the contribution ages between 15 and 64. The level of pensions depended on the marital status (if the person was married but not if they shared a house 13. ). The benefit was flat but the contributions depended on the income of the person. There was not a requisite for a minimum number of years insured. There was an unequal treatment for married women because they did not pay the contributions individually; instead their income was added to their husband s. Also, married women did not receive a pension unless they were breadwinners. The premium will only be levied up to an income of 2723 Euros. There were no premium claims for income under 681 Euro for unmarried individuals. The AOW adjustment in pensions took place after consecutive increases 12 For a further description of the history of the AOW see Mahieu, R (2007) 13 This policy changed in 1985; in 2011 two cohabiting persons are considered as married in order to receive the AOW. 10

13 in wage index 14. In this way, one initial hypothesis is the redistribution from married women (not breadwinners) and high-income individuals to lower income individuals. The years when it is considered that main regulations took place are: 1964 (65), 1985, 1990, 1997, 2002 and In 1964 the AOW was increased to a social minimum level. Minimum pension level was equal to 70% of the gross minimum wage 15. Around these years the AOW pension increased 15 percentage points until achieving the 85% of minimum gross wage. This reform can change the generational account of the individuals for this and further cohorts because it increases the future benefits received by the pensioners. In 1985 the abolition of the unequal treatment by law came into play. In the next years this change was realized. People cohabiting were treated as married for pension purposes. The AOW supplementary allowance for pensioners with a younger partner was also introduced. These changes could lead to an increase in benefitswelfare perceived by married women and married persons in general. In 1989 there was a change in the system of income tax and social insurance premiums. The number of tax brackets passed from 9 in 1989 to 3 in The range of the brackets was lower and social security premiums merged in the first bracket of taxes. This change in regulation is important because it will affect the contributions of the affiliates and it will increase the total value of contributions. In 1998 a tax change was introduced. The maximum AOW Contribution rate was increased from 15.4% (in 1997) to 18.25% (in 1998). Also, the Tax brackets were changed, and the first bracket was reduced by 1.78 percentage points. There was a big change in 2001 in the taxation scheme. The current system was created in which there are 3 income boxes: income from employment, profit from an owned company and capital income. The tax credit was introduced (instead of the tax free allowance) and this negatively affected low-income individuals and consequently their generational accounts. Finally in 2001 and in future plans, there are some proposals to increase the retirement age and eliminate the supplementary allowance in the upcoming years. 14 Standard Wage Index: measure average changes in contract wages as set out in collective agreements. 15 SVB (2008). The Dutch State Pension. Past, Present and Future. Amstelveen p.p

14 2.3 Current Tax system in The Netherlands The contributions to the pension system (first pillar) in The Netherlands come from the tax payments. The system in the Netherlands is EET (Exempt Investments, Exempt Returns, and Tax Withdrawals). That means that contributions are exempt from taxes, as well as the returns and gains from investment. Instead of that, the pension benefits are taxed from personal income tax. Remunerations in cash are usually taxed, while remunerations in kind are not always taxed (as reimbursements or provisions that may be exempt of tax). Table Ranges Contributions to the Social Security Range Salary Range Tax total (percentage) 1 Until 18, % Tax Social Security 2.3% Tax 31.15% social security (the 17.9% correspond to AOW) 2 From 18,218 to 32, % 10.8% Tax 31.15% social security (the 17.9% correspond to AOW) 3 From 32,738 to 54,367 42% It is not labeled explicitly the percentage directed to the AOW 4 More than 54,367 52% It is not labeled explicitly the percentage directed to the AOW Source: Ministry of Finance The Netherlands In the case of the contributions to the first pension pillar, there are some ranges specified according to the income of the individuals. As can be seen in Table 1, those individuals with higher income contribute to the pension system with a higher percentage of their salary, and consequently the amount transferred to the System is also higher. 3 Literature review 3.1 How to measure the redistribution? Theoretical point of view. 12

15 There are two main ways to measure redistribution. The first one is to calculate the Generational Account (GA) of the groups of individuals (men and women, high income and low income, married and single) per cohort, and after compare the results per cohort and group. In this way, those groups with higher values of Relative Generational Accounts get higher implicit benefits from the first pension pillar; this methodology is explained in Section 3.2, and Section 3.3 describes the pros and cons of this analysis. The second method consists of the use of micro-simulation models in order to consider the specific case of every individual. Those models (as the one for the Netherlands) can be of 2 types: static (with the characteristics of the individuals fixed) or dynamic (with variable characteristics of the individuals). The review of the Micro simulation Models is presented in Section Generational Accounts This methodology was initially presented by Auerbach (1991). The Generational Accounts of currently living generations are calculated based on the projected net tax payments (benefits and tax payments) deducting the time t (the year considered). Borsch-Supan (2001) follows this analysis and confirms the redistribution interpreted as divergences in the rates of return within (or between) a generation. There are 3 different definitions of transfer share from one group (or cohort) to the other one: i) Deviations from Absolute Equivalence, DAE: differences between the presented discounted value of the benefits and the liabilities (contributions). ii) Differences between the present value of the benefits and the contributions are considered as transfers that the individual receives or gives to the system. If DAE is equal to zero it implies that the social security system does not have transfers. Deviations from Relative Equivalence: if the discounted present value of the benefits is proportional to the one of the contributions, this methodology is applied in this document. The proportionality between benefits and 13

16 contributions is evaluated for some specific groups and cohorts. If the present value of the benefits is at least proportional (not necessarily equal) to the net present value of the contributions then the system presents relative equivalence. iii) Concept of No opportunity Costs Foregone: Under this concept the alternative systems for retirement provision (besides PAYGO) are considered. According to Borsch-Supan (2001): The difference between the implicit rate of return in the PAYG system and alternative systems for income provision in old age could then be interpreted as tax (or, if negative, as a subsidy). In this case, the cheapest outside alternative defines the insurance premium, the remainder is a transfer Ter Rele et al. (1998) calculated the generational accounts for the Public Sector in The Netherlands. They present a measure of positive or negative effects for the members taking into account the total tax system and public benefits; this is made in order to check intergenerational effects of fiscal policies. Even when this study does not measure redistribution between groups or cohorts for the pension system, this is an application of the Generational Accounts approach for The Netherlands. The individual s lifetime benefits from the government are considered for individuals born since The study finds that post-war cohorts perceive negative net benefits in the first years, which highest benefits are reached among those cohorts born between because of the different social programs present in this period, and that after 1990 the benefits start to decrease because of fiscal expenses reduction programs. For the case of the Canada Pension Plan, Morrison (2007) finds differences in the internal rates of return between cohorts and groups. In most of the cases the rates of return in the pension system for the individuals decrease from 1930 to Between groups he observes that females have higher internal rates of return than males. This is also the case between immigrant females and immigrant males, and low earning individuals in comparison with the rest of the individuals. 14

17 The Generational Account approach has also been used in the case of The Netherlands to evaluate the redistribution within and between generations in the Second Pillar (see Bonenkamp, 2009) They classified the pension fund affiliates by educational level, gender and age. The present value (for every group) of the average benefit received from participating in occupational pensions is calculated. They found redistribution from males to females (Inter-gender transfers) and from low educated to higher educated workers (Inter-educational-workers). It seems that higher educated workers have higher life expectancy than the lower educated, so lower educated die earlier and receive less benefits. It is interesting to notice that the differences between the characteristics of the population are translated in redistribution/transfers in the second pillar (and probably in the first one as well). For individuals with similar characteristics the gains or losses in pension funds are approximately divided in equal size at individual level. 3.3 Advantages and disadvantages of Generational Accounting According to Bovenberg (2010) among the limitations of the Generational Accounting we find that this approach includes strong assumptions, such as perfect markets (no liquidity constraints and no uncertainty). Also, in this approach the possible behavioural responses of the individuals in front of policy changes are not included. Indeed most of the critics about GA point out the main fact that factor prices (wages and interest rates) are constant and do not depend on fiscal policies. On this subject, Van Dalen (2000) suggests the use of interest rates as stochastic and timevarying for example. Another limitation is that it does not include the possibility of private intergenerational transfers, and perhaps this effect can be contrary to the public transfers. Other authors as Haveman (1994) mention some main deficiencies in the Generational Account methodology: 15

18 i) It is not possible to compare the levels of current generations with other levels from different years. ii) It is necessary to include a lot of projections and forecasts for the calculation of the GA s. These projections can bias the analysis. iii) It is necessary to make assumptions on long-term economic growth and population. 3.4 Micro-simulation models - MSM The other possibility to measure redistribution is using micro-simulation models. In these models the population is initially classified in different groups according to socio-demographic characteristics, welfare or level of real income. In the models, the information for every specific individual is taken into account. In this analysis, some probabilities are calculated, for example the probability to get married, to die, etc. These probabilities affect the trends of the Contributions or the Pension Benefits and consequently, the total net benefit of the individuals. There are 2 types of micro-simulation models: dynamic (characteristics of a record can change over time) and static micro-simulation models (does not allow characteristics to change). These types of models are widely used in some countries for policy analysis. For example, Cassels, Harding and Kelly (2006) present an overview of 6 models: USA, Canada, Norway, Sweden, UK and Australia. It only holds for income differences up till 32,000 Euro (in 2009). In the available literature comparisons between some group s indices are also found. Mazzaferro, C (2008) compares three kinds of indices 16 in order to examine the intra and inter-generational redistribution effects in the Italian pension system, taking into account the reforms of the social security system in 1992 and This study uses a discrete time based dynamic population model to project benefits through time to Mazaferro finds a decreasing replacement rate in the Italian system from 70% in 2008 to 45% in The gross replacement rate between the first year pension benefit and the last year wage/earning, the GINI Index computed on the annual distribution of old age pension benefits and a poverty index computed among pensioners and among the whole population as well. 16

19 The Ministry of Social Affairs of The Netherlands created a model tailored to forecast pension expenditures and investigate budgetary effects. The name of the model is SADNAP (Social Affairs Department of The Netherlands Ageing and Pensions Model). This model is used to analyse income redistribution. It uses a new micro data source, and individual data based on private pensions (including differentiation in mortality rates). The SADNAP model includes the redistribution effect of different policy measures. This model uses data from 2 main sources: first, Statistics Nederland (CBS) micro data file concerning state pension entitlements and 2005, for 11 million data. And second, it considers information from 2.6 million persons receiving the AOW (gender, Birth Date, country of residence, marital status, birth date of the partner, number of entitled years for the AOW and entitlement to a partner allowance). In comparison with the GA approach used here, it is possible to distinguish some advantages and disadvantages of micro-simulation. As advantages, it is possible to realize that Microsimulations include more detail information for individuals with specific characteristics; also it is possible to evaluate special cases when individuals divorce, married, start working or get an increase in their salary. Usually it is not possible to distinguish between groups but the micro simulation allows doing this. Some features can be taken into account, like income changes, migration and trends in households. Information does not get loss when creating groups. As disadvantages, this micro simulation models can be complex and require a lot of information that can be difficult to obtain. Also, it is necessary a high maintenance of the database and the model, the latter one can be seen as an advantage of the GA approach used here. After seen the GA and MSM separately, it is possible to summarize the similarities and differences between the static models and Generational Accounts, see Table 2. 17

20 Table 2. Similarities and differences MSM and GA Similarities- Differences Characteristics of individuals Variable analized to check redistribution Microsimulation Static Models, MSM Fixed Generational Account. In some exercises the Share of pension Benefits (Sonsbeek, 2011) Generational Accounts Can include changes Net present value of the contributions (benefits - contributions) Individual or aggregated level Individual cases Aggregated for characteristics of individuals Uses Use, access and maintaining of the data Differentiation for groups For analysis of the immediate effects of Tax-Benefit policies High Yes Measures the fiscal burden that government policies impose on future generations Low Usually not. Purpose of this analysis. 4 Description of the data and methodology Here it is calculated the Generational Account (Net Present Value of the Contributions to the AOW Net Present Value of the Expected Pension Benefits received) for some specific groups of individuals in the years during which the main changes in the regulation of the AOW took place. Graph 1. Generational Accounts in The Netherlands AOW For example, as can be seen in Graph 1, those individuals born in 1892 (vertical axis) did not pay contributions to the system since the AOW was created in Nevertheless, they started receiving pension benefits in 1957 (horizontal axis) until 18

21 they died (grey area). In order to calculate the Generational Account of this cohort (divided by sub-groups) it is necessary to obtain the Net Present Value of contributions (equal to zero) minus the Net Present Value of the benefits (from 1957 onwards). In a different case, for those individuals born in 1972 the net present value of the contributions (green area) will include an historical part (from 1987 when individual age is 15 until 2011 when he is 39 years old). The remaining part of the contributions (from 2012 to 2037 when the person reaches the age of 65) needs to be calculated according to the growth forecast of the contributions/salaries of the individual. The grey area corresponding to the present value of the pension benefits will depend on some variables for the remaining years (2038 onwards, according to life expectancy at birth of that cohort). These variables are: i) The marital status of the person. ii) The monthly/yearly amount of the AOW received and, iii) The number of years that the person receives his pension. The individuals are classified by different variables to compare: by household composition, by income and gender. 5 Methodology of calculating the Generational Account For the currently living generations the generational accounts are calculated based on the projected net tax payments discounting the values to the time t when the exercise is performed. In order to make the results of the GA comparable between years, it is possible to obtain the ratio between Contributions and Benefits. This Contributions/Benefits ratio is comparable between years and gives ideas for the analysis between groups. Example, for individuals born in 1892: In this case, the value of the contributions (from 1907 to 1956) and pension benefits (from 1957 onwards) are discounted to the moment when the individual is 15 years old. The interest rate used is 3% in the basement case (this interest rate is applied to 19

22 discount the real values base 1957 = 100. In that way it can be understood as a real interest rate 17 ) Since the system was established in 1957, those individuals born in 1892 do not contribute to the system. In that way, only pension benefits are discounted. Since the AOW was established in 1957, this generation has the biggest positive gain because the added value of the Contributions of the system is zero. The Net Present Value of Pension Benefits starting in 1957 until the average year of death (estimated also with the life expectancy at birth data). 5.1 Assumptions and procedure The procedure to calculate the Relative Generational Accounts is summarized in the following steps: a. Initially the value of wages and pension benefits, per age and gender, is taken in nominal prices for some years from the IPO survey. For the other years the Wage Index 18 growth is used from the historical data (1957 to 2011) to extrapolate the wage value, per year per age. Since 2012 the yearly growth in the GDP per cápita is used as a proxy for the growth in wages. Central Plan Bureau, CPB forecast 4 different scenarios of economic growth for The Netherlands up to In those scenarios the yearly economic growth is equivalent to: Scenario 1: 2.05%; scenario 2: 1.15%; scenario 3: 1.73% and scenario 4: 0.73%. Since 2013 onwards it is assumed long run inflation equal to 2%.This taking into account the forecast of CBS. Since all the calculations are done in real terms a change in the inflation rate does not have any effect on the generational accounts. 17 This interest rate of 3% in real terms is used in some studies. A recent one developed for Generational Accounts in the UK. See McCarthy, D (2011) 18 Source: CBS info service 20

23 In terms of the Aaron condition, for the long term and according to the assumptions taken in this exercise we have four different cases responding to the yearly economic growth scenarios (since the GDP per cápita is used as a proxy for the growth in wages): Long term assumptions: Real Wage Growth depends on the scenario evaluated, x will be equal to (Scenario 1) 2.05%; (Scenario 2) 1.15%; (Scenario 3) 1.73%; or (Scenario 4) 0.73%, and the Real Rate of Return, r is between range 2% to 4% according to the case (basement case 3%). Population growth, n is not implicitly included in the forecast. But it is included through the increases in the expenses in the AOW; I used the CBS population forecast for all population average between (n = 0.94%) and average between (n = -0.2%). And for people older than 65 years, average between (n = 10.2%) and average between (n = -1.6%). Replacing the first growth scenario values for the period we have that the economy is dynamic inefficient, in such a way that PAYGO scheme seems to be slightly more attractive than funded plan, in this way we have: n = 0.94%; x = 2.05% (real wage growth) and r = 3% But for the next period, , the difference between the internal rate of return of the PAYGO and the real rate of return increases. Indeed, the economy becomes dynamic efficient because of the decrease in population growth, the Aaron condition becomes: 21

24 According to this data, in the further years with the only change in population growth the funded scheme becomes more attractive than the PAYGO 19. b. The nominal salaries per year, per age, are compared with the Maximum Salary contribution rule for AOW. The historical Tax Rate and the Maximum Salary Value can be found in the Appendix 2. According to the value of the nominal salary and it s comparison with the Maximum Salary Value, there are 2 cases: i) If the average nominal salary is lower than the Maximum Salary Value, then the value of the Contribution to the AOW system, is equal to (2): ii) If the average nominal salary is higher than the Maximum Salary Value, then the value of the Contribution, is equal to (3): For the period the historical Tax AOW Rate is used. For 2012 onwards the value for Tax Rate AOW is assumed constant and equal to 17.9%. c. Also, the Average Contribution Tax for the AOW is calculated as the remaining part from the Total AOW expenses minus Total Contributions, divided by the total population between 15 and 65 years old, in this way for every year the Total AOW Tax value is, according to (4): 19 For the remaining scenarios, the economy is dynamically efficient in both periods (1+n)(1+x) < (1+r). Period : Scenario 2 (1.021 < 1.03), Scenario 3 (1.026 < 1.03), Scenario 4 (1.016 < 1.03). And for the period we have: Scenario 2 ( < 1.03), Scenario 3 (1.015 <1.03), Scenario 4 ( < 1.03). 22

25 And the Average Contribution Tax per person: The values of the AOW Tax, Total AOW expenses and AOW contributions, are dependent variables of the population over 65 years and the Wage Increase forecast (economic growth). There are 4 different values of Average Tax Contributions per year, according to the 4 scenarios. d. The pension benefits for historical data are also extrapolated using the Wage increase. And for the future years it is calculated using the GDP per Capita increase according with each of the 4 scenarios. e. At this moment, Contributions and Pension Benefits are presented in nominal prices. Those are converted into real values using the IPC 1957 = The interest rate used in the basement case is 3%. In other cases for sensitivity analysis the values 2% and 4% are changed, those results are presented in Section 7.3. f. Pension benefits are added (in real values) until the year of Life Expectancy Value at 65 years is achieved (for men and women). The Life Expectancy year at 65 value for the next years is obtained from CBS Statline 21. g. In this way, the demographic changes that affect The Netherlands population are included by 2 ways: i) With the continuous increase in life expectancy for men and women since the introduction of the AOW system. 20 There are 2 Index used here, Indexcijfers cao-lonen, total base 1972 = 100 sent by CBS info service, and the Index for 2000 = 100 available in the CBS webpage. 21 According to the different cases evaluated, the Life Expectancy for single men and women, married and single, high and low income individuals is evaluated. 23

26 ii) Through the total value of future AOW expenses, this includes the values of pensions forecast for the next years. And this responses to the changes in population composition. The total AOW expenses increase when the number of affiliates younger than 65 decreases and the number of pensioners increase. This is a result of the aging of the population. h. As other assumptions, in this analysis the Supplementary Allowance for younger partners is not included. i. The AOW insures all the individuals in The Netherlands from 15 to 65 years, even if they are working or not. Also, if you do not have a residence in NL but work in NL then you are insured. In this exercise implicitly is taken into account the group of individuals that are working. This is the case of individuals that work during all the lifetime and receive pension benefits. In this case unemployment probabilities are not considered, and only persons working in The Netherlands are considered. Also marriage/divorce probabilities are not taken into account. j. The case when individuals are not working in The Netherlands and/or living for some years in The Netherlands is not taken into account. When this occurs, per year, the pension benefits decrease by 2%. k. The New Pension Contract: the policies included in the New Pension Contract and included in this exercise are: i) raising of the AOW pension age from 65 to 66 in 2020 and 67 in 2025, ii) raising of the pension benefit in the period 2013 to 2028 with 0.6% per year (this is an extra raising, the AOW will also be raised with the wages). 5.2 Descriptive statistics As can be seen in Graph 2, in the case of The Netherlands, in 2000 for example the peak average salary was 67,764 Euros reached at the age of 54 years. The sample confirms the life cycle theory. After the age of 57, the salary starts to decrease until it is equal to the value of 43,738 Euros at age

27 On average the salary of men is higher than the salary of women. In the case of women, the highest average salary (36,642 euros) is reached at 29 years of age. Pension benefits for men at 65 years old are higher than for women, some hypothesis than can explain this is because of the Supplementary Allowance for younger partners. Because, according to Harmsen (1999) in The Netherlands men on average are 2.5 years older than their wives, in that way when they reach the retirement age they will receive the Supplementary Allowance and this will increase their pension benefits until his wife retire. Graph 2. Annual average salary men and women 2000 (euros) Average Salary per year (eur.) Age Mean Men Women Graph Mean, Annual average salary men and women 2000 (euros) 25

28 *The average salary for men and women can be found between the intervals 1* and 2* with a 95% of confidence. Graph 3. Annual average AOW pension benefits 2000 (euros) Pension benefits yearly (eur) Age Pension Men Pension women In average when both men and women are 72 years old, the pension benefits for women are higher than those for men, see Graph 3. 6 Results 6.1 Gender: Men and women Men In the case of men, the value of the GA decreases with the time, because the first generations pay fewer contributions (before 1957). This decrease in the GA with the time is compensated by the increase in life expectancy, according to CBS life expectancy at 65 years increase from 11.2 years in 1892 up to 17.8 years in As can be seen in Table 3, the average GA changed from 0.44 in average between to 0.23 for those that entered to the system after 2000 (born in after 1985). Table 3. Average GA for men. ( ) Birth year (or year enter to the system) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Average GA Men

29 As we can see in Table 3, for the present value of the contributions for men is higher than the present value of the benefits (GA = 0.23 < 1). This can be explained because pension benefits increase every year with the increase in wages, and contributions increase every year as a result of the changes in population and also the increase in wages. As a result in the Section 5.1.a we got that the economy is dynamically efficient for the period onwards, this result of the Aaron condition confirms the outcome of Table 3 for GA of men that enter to the system in because in present values they will receive less in benefits than what they put into the system as contributions. In that way, contributions increase more than benefits because of demographic changes (young population decrease) and this implies that affiliates need to contribute more every year in order to pay old pension benefits. Therefore men that entered to the system in will find more attractive the fully funded system than the PAYGO scheme. Women For women, they have a decreasing GA over time according to the year when they enter to the system. From 2.98 in average between to 0.63 between Women start to have less relative contributions in comparison with the benefits after That means that they (on average) are contributing more to the system than what they are receiving. See Table 4. Table 4. Average GA for women ( ) Birth year (or year enter to the system) Average GA women ( ) ( ) ( ) ( ) ( ) ( ) ( ) , This decrease in GA is also compensated by the higher increase in life expectancy in comparison to men. For women, their life expectancy at 65 years at 1892 was 11.8 years, and it increases up to 21.2 years in

30 Comparison by gender According to Table 3 and Table 4, the value of the generational accounts is higher for women than for men in all the periods. The decrease in the value of GA is faster in the case of women than for men, Graph 4. In the year of enter to the system 1960 the value of generational accounts (GA) for men was 0.39 and for women it was 2.3. For those entering the system in 2011, GA is equal to 0.21 for men and 0.6 for women. In this way, GA for women decreased 74% between 1960 and 2011 in the meanwhile it decreased 46% between 1960 and 2011 for men. Also the gap between women s and men s GA has decreased since the beginning of the system in For example, on average the difference was equal to 2.54 points for those who were 15 years old between 1950 and According to Table 3 and Table 4, between men GA value was on average 0.44 while women s GA equalled 2.98 in the same period. For those who were 15 years old between 2000 and 2011 this difference between averages GA s decreased to 0.4. For that period GA for men was 0.23 while GA for women equal to The policies included in the new pension contract initially generate a gain for both women and men from the cohorts of until However, for those born since 1955 and onwards, these policies generate a GA s loss for both. The values are calculated according to the growth scenario 1. Women have a higher return than men from the system and this difference will decrease in the upcoming years. As we can see in Graph 4, there is a drop in the relative generational accounts in birth cohort 1955 (those who enter to the system in 1970) and in birth cohort 1959 (those who enter to the system in 1974). This can be explained because of the increase in the retirement age up to 66 in 2020 and to 67 in Those born in 1955 are 65 years old in 2020 but because of the reform they will need to postpone their retirement one extra year. Also, those born in 1959 are 66 years old in 2025 but because of the reform they will need to postpone their retirement one more year until they reach 67 years old. 28

31 Graph 4. Relative Generational Accounts at 2011 Men and Women. Scenario 1 and New Pension Contract (NPC) 22 New Pension Contract: Men There are 2 policies included in the New Pension Contract and included here: i) the increase in the pension benefit in the period 2013 to 2028 with 0.6 % per year (extra to the increase in the AOW by the wages), ii) and the increase in the AOW pension age from 65 to 66 in 2020 and 67 in In the case of men, as we can see in Graph 5, birth cohorts 1940 to 1954 have an extra gain because of the extra increase in the pension benefits. For the next years this initial gain is compensated (negatively) by the increase of the retirement age, in that way birth cohorts 1955 onwards will lose in GA with the new policies. On average for birth cohorts the gains in the generational accounts of the individuals are close to 1.2 % (between 1.1% to 1.2% according to the different economic growth scenarios). For birth cohorts 1955 onwards, it is possible to see 2 different phases: initially, from 1955 to 1957 losses are -0.3% in average, and starting in 1958 those start to decrease in different ways according to the different scenarios. 22 For the analysis by Gender and by Civil Status there were some differences in the availability of the data. For example there was less historic information available about income for women classified by married and single, in this way these data were obtained by extrapolating the latter information of income for the years available. This creates some differences between the results for average women and for women (single and married). 29

32 3% Graph 5. Men, Differences between Relative Generational Accounts New Pension Contract usual case. 2% Decreasing loss because of the increase in the retirement age 1% Percentage 0% -1% -2% -3% Birth year Scenario 1 Scenario 2 Scenario 3 Scenario 4 As we can see in Graph 5 there are some differences between the effects of NPC with the different scenarios. Growth Scenario 1 presents the biggest gains for men born between and the biggest (decreasing) losses in the long run scenario (from 1958 onwards). In contrast, low growth Scenario 4 present the lowest gains between , and also the lowest (decreasing) losses in the long run. These differences can be explained according to the high grow scenario, where the yearly increase in contributions will have a bigger effect than the yearly increase in benefits. For the initial years, the effect on contributions will be compensated because of the yearly extra increase in benefits of 0.6%. But for the next years the effect of increases in contributions will be higher that the effect on benefits. This effect will be higher because individuals will contribute for more years (40 to 42 years) than the years that they will receive pension benefits (around 20 years). The NPC has different effects on Generational Accounts, through changes in benefits and contributions. For the first period, because of the extra increase in benefits, there is a positive gain for birth cohorts 1948 and Because of the increase in the retirement age, there are some different effects on GA. On one side, affiliates will contribute for a longer time to the system (1 or 2 more years), and this will increase the present value of contributions. On the other side, there are 2 effects on benefits: a negative effect because individuals will contribute for more years (1 or 2) instead of getting pension benefits; and a positive effect because of the increase in life expectancy for both men and women. In this 30

33 way, GA with NPC will decrease for birth cohorts after 1955 and individuals will perceive a loss with the new policies. In summary, contributions and pension benefits will change because of different effects: The effects on contributions are the result of: demographic changes (less young and more old), economic growth (via yearly increase in wages and taxes), New Pension Contract (one or two extra years of contributions). Also, the effects on pension benefits are the consequence of: economic growth (via yearly increases in pension benefits), New Pension Contract (extra increase of 0.6% for the initial years of the NPC) and increase in life expectancy compensated by an increase in the legal retirement age. New Pension Contract: Women For women, the results are similar to men. Women that were born from 1940 to 1954 receive an initial gain because of the New Pension Contract, on average 6.9% (1940 to 1954), and 2.1% (from 1955 to 1957). There is also a loss on average equal to -5% for those that were born between (1958 and 2011). As we can see in Graph 6, the Scenario 1 presents the initial biggest gains between and the biggest (decreasing) losses after birth cohort For the years after 2011 it is assumed that contributions and benefits increase with the economic growth, with the NPC (between ) contributions and benefits increase, but the effect on benefits is higher because of the extra 0.6% yearly increase. After birth cohort 1955, the effect with the NPC for women is similar than the effect with the NPC for men: Scenario 1 presents the biggest losses because the total effect in contributions is higher than the total effect in benefits. 31

34 Graph 6. Women, Differences between Relative Generational Accounts New Pension Contract usual case. 15% 10% 5% Decreasing loss because of the increase in the retirement age Percentage 0% -5% -10% -15% Birth year Scenario 1 Scenario 2 Scenario 3 Scenario 4 In comparison of the effects of the NPC for men and women we can see that women are more affected than men as a response of the New Pension Contract. Women get a bigger gain in the initial years, when they retire before the first increase in the retirement age (6.9% women vs. 1.2% men) Single and Married In the single and married situation there are also some generalized differences in the value of the generational accounts. According to the CBS data, women benefit more than men from a marriage. Married women on average live 4 years longer than single women, and in the case of men their life expectancy increases only in 2 years when they are married in comparison with those that have never been married 23. This effect generates an increase in the benefits and consequently in the generational accounts. For married individuals there are 2 opposite effects affecting their generational accounts. On one hand, on average they get pension benefits for a longer period (4 years for women and 2 years for men) when they are married because they live 23 De Jong, Andries, Married People live Longest. CBS Web Magazine, 03 June Available online at: 32

35 longer. But on the other hand, these benefits are smaller than the ones received by singles per month. In this exercise changes in the civil status are not assumed with time, so for the representative individuals (men and women) it is assumed that if they were single when they contribute to the system, then they are also single when they retire. Consequently, if they are married when they contribute to the system they also remain married when they retire until one of them dies. According to the Dutch system, pension benefits for singles are higher than for the married. The initial effect increases the value of the Generational Accounts of married individuals and the second one decreases it. As we can see in Graph 7, in the sample for single individuals the value of the Generational Accounts decreases with time. A single woman that entered to the system in 1960 (born in 1945) has a return of the AOW system of 2.27 with the basement case in the scenario 1, and a single man (born in the same year) gets a return of 0.42 from the system. Single women get a higher return than men because of their higher life expectancy. In both cases return decreases with time when they enter to the system in 2011, it is equal to 0.23 in the case of single-men and equal to 0.59 in the case of women. As we can see in Graph 7, there is a drop in relative generational accounts for birth cohorts 1955 and 1959 because of the NPC. Graph 7. Relative Generational Accounts at 2011 Single Men and Women. Scenario 1 and New Pension Contract (NPC) 33

36 Graph 8. Relative Generational Accounts at 2011 Married Men and Women. Scenario 1 and New Pension Contract (NPC) In contrast, for married women that enter the system in 1960 (born in 1945) the value of their Generational Account is 0.9 with the basement case, and for a married men it is For those that enter the system in 2011 those values are 0.22 for married men and 0.24 for married women. As can be seen in Graph 8, for married individuals the generational accounts value decrease with time as well, with the basement case and the New Pension Contract. In this case, there is also a drop in relative generational account for birth cohorts 1955 and 1959 because of the increase in the retirement age in NPC. With the New Pension Contract, the effect is similar in both cases for singles and for married, to the one for the generalized gender case. For the generations that enter to the system between 1960 and 1970 (born between 1945 to 1955) there is an initial gain according to the extra increase in benefits. And after, in the next years, there is a loss in welfare for the next generations according to the increase in retirement age (see Graphs 9 and 10). For married men that were born from 1940 to 1957 the initial gain on average is equal to 1.36% (taking the 4 different scenarios into account, values range from 1.2% to 1.5%). For those that were born since 1958 onwards, the loss on average is equal to -1.3% (from -1.1% to -1.6% in the different scenarios), see Graph 9. 34

37 3% Graph 9. Men Married. Differences between Relative Generational Accounts. New Pension Contract 2% 1% Loss because of the New Pension Contract Percentage 0% -1% Gain because of the New Pension Contract -2% -3% Birth year cohort SCENARIO1 SCENARIO2 SCENARIO3 SCENARIO4 Married women have an initial gain on average for birth cohorts 1940 to 1957 equal to 3.3% (from the range of 3.0% to 3.6% according to the scenario). For those that were born since 1958 onwards, the loss in average is equal to -0.9% (from -0.8% to - 1.0% in the different scenarios), see Graph 10. 8% Graph 10. Women Married. Differences between Relative Generational Accounts. New Pension Contract 6% Percentage 4% 2% Loss because of the New Pension Contract -1% -3% Gain because of the New Pension Contract -5% Birth year cohort SCENARIO1 SCENARIO2 SCENARIO3 SCENARIO4 35

38 Differences in the Generational Accounts with time For those that entered the system since 1940 to 2011, the difference in the value of the Generational Accounts of singles and married men has been positive, and has also increased with time. (G.A. Single Men G.A. Married Men). This indicates that the total GA of single men has increased with time in comparison with the one from married men. Although married men live longer they receive less pension benefits than singles. Taking into account the year they enter to the system, between the difference in the Generational Accounts of singles - married was 0.4; this difference increases up to 1.0 points between , see Table 5. This can be explained because of the increase in life expectancy for all men (married and single ones) in this way even when married live longer than single ones; the increase in benefits because of longer lives applies to both kinds of men. The GA of single ones increases more because they receive higher benefits than married ones. Table 5. GAP - Between married and singles. Generational Accounts: Men single-married Birth year (or year enter ( ) ( ) ( ) ( ) ( ) ( ) ( ) to the system) Sc Sc Sc Sc Single women receive higher pension benefits than married ones; and their Generational Accounts value is higher. Married women live on average more than singles, but this difference has decreased with time from 3.5 for those that enter the system between ; to a 0.3 on average for those that enter between , see Table Because of the low availability in the data for singles and married, this exercise does not explicitly take into account the fact that before 1985 married women did not have their own pension. For the married/single exercise the latest year s data were extrapolated according to the Wage Index increase. 36

39 Table 6. GAP. Between married and singles. Generational Accounts: Women single-married Birth year (or year enter ( ) ( ) ( ) ( ) ( ) ( ) ( ) to the system) Sc Sc Sc Sc One explanation of this lower effect for women in comparison with men is that women benefit more from marriage than men; married women live 4 years longer than single. In that way the difference between pension benefits received by single and married women decrease with time because they have a higher increase in life expectancy because of marriage. 6.3 Percentile analysis According to the distribution of income there are some differences between GA s of high income and low income individuals. High income affiliates make greater AOW contributions, and pension benefits are independent of income. Pension benefits only depend on civil status. Graph 11 shows GA s values per cohorts for high and low income men. GA s are higher for low income men, because they give lower contributions to the system and they get more benefits from the scheme (even when they have a lower life expectancy). For low income individuals there are two opposite effects on their GA s: i) a positive effect, because their contributions are lower than those for higher income individuals. This tends to increase low income individuals GA s vs. high income individuals, ii) a negative effect, because low income affiliates have a shorter life expectancy than high income ones. In this manner, low income men receive pension benefits for a shorter period than the others. According to Alessie et. Al (2009) life expectancy is 3 years shorter for men and women when they receive low income. 37

40 Graph 11. Relative Generational Accounts for Men. Low Income and High Income. Graph 12. Relative Generational Accounts for Women. Low Income and High Income. In the case of women, according to Alessie (2009) income has a positive effect on their life expectancy. Low income women live 3 years less than others, and women who are married to low income men have 2 less years of life expectancy in comparison to women who are married to high income men. The GA s for women with low and high income is presented in Graph

41 Table 7. The best scenario is for women, of Low income and High income. Year of enter to the system Men Low Income Women Low Income Men High Income Women High Income As we can see in Table 7, low income individuals have higher values for their Generational Accounts, when they enter to the system in 1960 and also in New Pension Contract, NPC In this case the effects of the NPC that are considered are: i) increase in the retirement age, and ii) extra increase in the pension benefits. Regarding the first category, the decreasing impact ceteris paribus on benefits is higher for low income individuals than for high income ones. As can be seen in Graph 13, low income individuals will have an even shorter retirement period than high income ones because of the reform (and because their lower life expectancy), and consequently their relative GA (benefits/contributions) will be more negatively affected than in the high income case because of the NPC. Thus, one year less in the retirement period is relatively more expensive for low income individuals than for high income ones. Graph 13. Increase in the retirement age for low income and high income individuals. 39

42 Men According to Graphs 14 and 15, low income men perceive a bigger gain than high income men when pension benefits rise between Average gain for low income men is 3.1% while 1.4% for high income men, the difference is thus 1.7%. In the same way, after birth cohort 1955 low income men are affected more than high income ones because of the reform. This can be analysed in two periods. For the first period to birth cohorts 1955 to 1957, high income individuals receive a positive average gain of 0.7% in comparison with low income individuals that perceive a loss of 0.7% in the same period. In the second period, after birth cohorts 1958 both individuals perceive a loss, but this is bigger for low income men. With the NPC, loss for low income men is on average 5.2% and 1.5% for the high income ones. This result can be explained by two reasons: first, the differences between contributions by high income and low income individuals (according to their salary), and second, the differences in life expectancy for both types of individuals. Assuming that low income individuals have a lower life expectancy in comparison with high income ones, if they have to work one or two more years then they will still need to contribute for more years despite having shorter lives than high income ones. 40

43 Graph 14. Men Low Income Percentage 8% 6% 4% 2% 0% -2% -4% -6% -8% -10% -12% -14% Birth year Scenario1 Scenario2 Scenario3 Scenario4 Graph 15. Men High Income Percentage 4% 3% 2% 1% 0% -1% -2% -3% -4% Birth year Scenario1 Scenario2 Scenario3 Scenario4 Women In the case of women and according to Graphs 16 and 17, low income women benefit more from NPC. On average low income women have a gain 10.3% vs. 4.2% for high income women. 41

44 Graph 16. Women Low Income 20% 10% Percentage 0% -10% -20% -30% -40% Birth year Scenario1 Scenario2 Scenario3 Scenario4 Graph 17. Women High Income 10% 8% 6% Percentage 4% 2% 0% -2% -4% -6% -8% Birth year Scenario1 Scenario2 Scenario3 Scenario4 6.4 Interest rate changes One part of this exercise is to check the sensibility of the results to changes in the variables, such as the interest rate. In the basement circumstance it is consider that the interest rate is equal to 3%, when the interest decreases the value of the Generational Accounts decreases, this can be explained because of the higher discount imposed to the future values (pension benefits) and implicitly less ponderation of the benefits. The results concerning the differences between men and women remain the same in the general case, see Graphs 18 and 19. According to the Aaron condition, 42

45 for example with changes in r from a 3% to a 4% in the period , we have the following: Aaron condition with r = 3% the economy is dynamically inefficient (the economy is dynamically inefficient in the long run with r=2% also) With an increase in the interest rate now the economy becomes efficient. In this way with r = 4% a funded plan appears to be even more attractive than a PAYGO. Graph 18. Men Relative Generational Accounts. Sensibility analysis, changes in the interest rate Graph 19. Women Relative Generational Accounts. Sensibility analysis, changes in the interest rate 43

46 7. Conclusions and final considerations Because of the higher life expectancy of women, they seem to have higher returns from the public pension system AOW in The Netherlands. The value of Generational Accounts is higher for women than for men. These values have decreased for those cohorts born since the creation of the AOW until In this way, an implicit transfer from men to women is confirmed. Those transfers have decreased until now, according to a decreasing gap between Generational Accounts of men and women. The New Pension Contract has a positive impact on men and women that were born before and during This is because of the extra increase in the pension benefits. The size of the gain is bigger for women than for men, which can be explained also by the higher life expectancy of women. For the period , the results differ slightly between men and women insofar as men that retire between those periods will have a welfare loss under the New Contract. In the meantime, women that retire in these years will get a positive gain, which nevertheless will be lower than the positive effect perceived by women that retired in the previous years. After 2024 men and women will perceive a decreasing loss in GA s that will depend on the economic growth scenario considered for the Dutch economy. Scenario 1 (yearly economic growth equal to 2.05%) will bring higher increases in contributions and higher benefits, but since individuals contribute for more years (around 40 years) than they perceive benefits (around 20 years), then they perceive a GA s decrease in the long run. For example, with the high growth scenario the economy is dynamically inefficient for the period (and r=3%), but in the long run (after 2040) the economy becomes dynamically efficient and individuals will find the Fully Funded scheme more attractive in comparison with the PAYGO. Thanks to the analysis according to civil status and taking into account the restriction in the availability of the data, it is possible to check that the value of Generational Accounts of single individuals is slightly higher than for married ones; this result is obtained extrapolating the data of the recent years to 1957, when the AOW was created. In addition to the fact that married individuals live longer on average than single ones (positive effect), they also get less pension benefits when they retire (negative effect). The results obtained in this exercise suggest that there is a transfer from married to single ones. This result is different in relation to previous exercises presented in the literature, and can be explained by the assumptions on the 44

47 effect of marriage over life expectancy and the assumptions that individuals remain married or single for their entire life (divorce or marriage probabilities are not considered). In addition, increases in interest rates affect the values of the Generational Accounts negatively for all the cases evaluated, because if there are fewer discounts of the future pension benefits for the individuals, those discounts will have less weight in the Generational Accounts. As a result, the effect of changes in the interest rate is higher for women than for men, and decreases for the younger cohorts. There is a proven transfer in the AOW system from younger to older generations, because of the lower values of the Generational Accounts for younger cohorts in comparison to older cohorts. This is initially one of the main purposes of the system. It is also important to distinguish the windfall gain effects in GA s for specific cohorts according to the changes in regulation (for example, the extra increase in benefits of 0.6% from 2013 to 2028), and it is important to take into account that in the long run the GA s of young individuals decrease in comparison to the old (when the PAYGO scheme started). Within generations, there are transfers from men to women, which are coherent to higher life expectancies of women and lower salaries. Besides that, those transfers present a decreasing pattern over time in part because of the increasing participation of women in the labour market and higher salaries (contributions to the system). For further studies, there is a possibility to formulate a future collaboration between SVB and The Ministry of Social Affairs in order to apply their microsimulation model specifically to the AOW case. This can take the analysis of this study to a micro level and can be realized through an integrated work between the two entities. 45

48 8. References Alessie, et. Al (2009). Individual Income and Remaining Life Expectancy at the Statutory Retirement Age of 65 in The Netherlands. NETSPAR. Discussion Paper 09/ September, Auerbach, a. et. al. (1991) Generational Accounts A Meaningful Alternative to Deficit Accounting. NBER, Working Paper. No Barr, N. Diamond, P. (2008). Reforming Pensions: Principles and Policy Choices. Oxford University Press. Bonenkamp, J. (2007) Measuring lifetime redistribution in Dutch occupational pensions. CPB Discussion Paper. No. 81. Borsch-Supan, A. (2001) How Much is Transfer and How Much is Insurance in a Pay-as-you-go System? The German Case. Scand. J. of Economics 103(3), 505±524. Bovenberg, L (2007). The Life-Course Perspective and Social Policies: an Overview of the Issues. Netspar, discussion paper ( ), p.p.21. Bovenberg, L.(2003) Financing Retirement in the European Union. International Tax and Public Finance. Vol 10 (6). Pp Castellino, O. (1995) Redistribution between and within generations in the Italian social security system. Richerche Economiche Vol. 49, Dekkers, G. Nelissen, J. Verbon, H. (1995) Intergenerational Equity and Pension Reform: The Case of The Netherlands. Public Finance / Finances Publiques. Vol 50 (2). pp Draper et. al (2010). Calibration of Gamma CPB Memorandum No Fenge, R and Werding, M (2003) Ageing and the tax implied in public pension schemes: simulations for selected OECD countries, Futagami, R (2002) Income Redistribution effect of public pensions between dynastic families. Review of Income and Wealth, Series 48, Number 2. Harding, A. (1993) Lifetime Income Distribution and Redistribution: Applications of a Microsimulation Model. Amsterdam: North Holland. Harmsen, C.N. (1999). Cross-cultural marriages. Monthly Bulletin of Population Statistics, August-September. CBS. Harry ter Rele (1997) Generational Accounts for the Dutch Public Sector. De Economist. Vol. 146, Number 4,

49 Haveman, R (1994) Should Generational Accounts Replace Public Budgets and Deficits? Journal of Economic Perspectives, Vol 8. N. 1. Pp Herten, L. van, K. Oudshoorn, R. Perenboom, Y. Mulder, N. Hoeymans and D. Deeg, (2002), Gezonde levensverwachting naar sociaal-economische status, TNO rapport , Leiden Hoevenaars, J. & Ponds, E.H.M., (2008). Valuation of intergenerational transfers in collective funded pension schemes Open Access publications from Tilburg University, , Tilburg University. Kotlikoff, Laurence J., (2002). "Generational policy," Handbook of Public Economics, in: A. J. Auerbach & M. Feldstein (ed.), Handbook of Public Economics, edition 1, volume 4, chapter 27, pages Klotz (2007), Soziale Unterschiede in der Sterblichkeit, Bildungsspezifische Sterbetafeln 2001/2002, Statistiche Nachrichten. Kunzler (2002), Arme sterben fruher; Soziale Schicht, Mortalitat und Rentenalterpolitik in der Schweiz, Caritas Verlag. Mahieu, R (2007). The Changing role of premiums in the Dutch first pillar pension, AOW. Tilburg University Netspar SVB Thesis. Mazzaferro, C. (2009) Measuring intragenerational and intergenerational redistribution in the reformed Italian social security system Mc.Carthy, D. et al. (2011). Generational Accounts for the United Kingdom. National Institute of Economic and Social Research, NIESR. Discussion Paper No.377. Medijainen, M. Generational Accounting as a Tool to Evaluate the Fiscal Sustainability of Estonia (2010). University of Tartu Faculty of Economics and Business Administration, Working Paper Series, No. 74. Morrison, R. (2007) Rates of Return in the Canada Pension Plan: Sub-Populations of Special Policy Interest. DYNACAN Team. Government of Canada. Nelissen, Jan. H.M. (1995) Lifetime income redistribution by the old-age state pension in The Netherlands. Journal of Public Economics 58, Ponds, E. (2010). Lecture notes Macroeconomics of pensions and Aging. Reader 1: Topics in Public Pensions and Reform (p.p ). SVB, Sociale Verzekeringsbank, Amstelveen, (2008). The Dutch State Pension. Past, Present and Future Wolfson, M.C. (1998) Historical Generational accounting with heterogeneous Populations. Pp Government Finances and Generational Equity. Statistics Canada. Chapter 8. 47

50 Appendix 1. Principal regulation changes AOW 48