Review on Nordic and international literature on pension systems and fiscal sustainability

Transcription

1 Review on Nordic and international literature on pension systems and fiscal sustainability By Erling Holmøy and Nils Martin Stølen, Research Department, Statistics Norway 1. Introduction The main objectives of the Norwegian pension reform from 2011 are to Improve fiscal sustainability in the long run Increase supply of labour These objectives are linked together because higher employment strengthens government budgets, especially if higher employment is achieved through later retirement. Although our review is based on literature from a broad spectre of countries, we emphasize analyses of pension reforms in the other main Nordic countries Sweden, Denmark and Finland. We have also concentrated our survey to analyses of pension reforms which share the same objectives as the Norwegian, i.e. to improve long run fiscal sustainability and raise employment. Our survey is confined to relevant and quantitative analyses of high professional quality. Earlier analyses of effects from the Norwegian reform in e.g. Christensen et al (2012), Fredriksen and Stølen (2011) and Holmøy and Stensnes (2008) are just mentioned without going into details. Although these analyses may be extended and improved, the aim has been to analyse effects from the Norwegian pension reform fulfilling the criteria above. In analysing effects from pension reforms it seems relevant to separate between the following kinds of effects: 1. Direct (mechanical) effects on government pension expenditures. These effects do neither include behavioural effects nor general equilibrium effects. Even though important effects are excluded, it is of great value to show the direct effects because i) they contain budget effects following directly from shifts in the rules without any uncertainty; ii) they are important components in more complete analyses; iii) neglecting of behaviour and general equilibrium effects open for an even more detailed and correct treatment of detailed rules and heterogeneity in the population. The analyses made in Christensen et al. (2012) and Fredriksen and Stølen (2011) exemplify this kind of analysis. 1

2 2. Behavioural effects, especially on labour supply. A review of studies of these effects is the main purpose of the survey ordered from the Frisch Centre, and outside the main priority of our survey. However, because changes in employment have strong fiscal effects, they are mentioned where relevant. 3. General equilibrium effects. These effects work via shifts in relative prices and incomes in addition to adjustments in tax rates and/or government spending required neutralize government budget effects. Holmøy and Stensnes (2008) include and identify both direct, behavioural and general equilibrium effects. The rest of this review is organized as follows: Section 2 reviews analyses of fiscal effects of reforms or significant changes in the public pension system in Sweden, Finland and Denmark. Section 3 reviews the same kind of analyses confined to actual reforms, proposed or implemented, in non- Nordic countries. Section 4 presents relevant analyses of more stylized reforms. Section 5 points out some lessons to be learned from the reviewed literature for similar analyses of the Norwegian pension reform. 2. Analyses of reforms in the Nordic countries Like most other OECD-countries, the Nordic countries have for many years been aware that they are facing population ageing and a rapid increase in future government expenditures on pensions and health care services. According to Kangas, Lundberg, Ploug and Ploug (2006) Finland, Sweden and Denmark have followed quite different routes their new pension systems. The Swedish reform in the late 1990s was an abrupt overhaul of the whole system. The interest organizations played a minor role in the planning. The reform process in Finland was gradual and piecemeal starting in the early 1990s. Interest organizations took active part in the process. Kangas et al. characterize the Danish route as close to policy drift and changes without reforms, and without any great master plan (pp 8). As a response to the lack of a earnings- related legislated old-age pension scheme, the occupational and individual pension programs were expanded, whereas the national scheme became increasingly means-tested Finland According to a report prepared for Finland Ministry of Social Affairs and Health (2005) the old-age dependency ratio (the ratio of over 65 to the working age population) is expected to double from 2005 to 2030 and remain high thereafter. A number of steps have been taken to prepare for the change in the age structure of the population. The most important is the pension reform which took effect at the beginning of

3 The life expectancy adjustment is the most important element of the reform aiming to postpone average effective retirement age and/or reduce pension benefits if life expectancy increases. The lifeexpectancy coefficient adjusts the pensions upon retirement to the changes in longevity as of The life expectancy coefficient is the difference of the remaining expected lifetime at age 62 in a particular year compared to the base year 2009, based on population statistics. It cuts the initial pension benefit accordingly. It is possible to counteract the effect of the life expectancy coefficient by postponing retirement. The reform also strengthens the linkage between contributions and benefits and is aiming at an optimally stable development in pension contributions over time. The Finnish Government is also carrying out a more general employment policy programme with the aim to create favourable preconditions for significant improvement of the employment rate. The first pillar of the Finnish public pension system is made up of two statutory pension schemes. One is the national pension scheme guaranteeing a minimum pension to all residents, whereas the other is an employment-based, earnings-related pension scheme. Increases in the earnings-related pension reduce the benefits from the national pension by 50 per cent. If the earnings-related pension is above a defined level, the national pension is not paid at all. The national pensions are financed based on a pay-as-you-go system. Approximately ¾ of the earnings-related pensions are covered pay-as-you-go, while the rest is based on funding with contributions from both employers and employees. Despite the partially funding the earnings-related pension scheme is entirely of the defined-benefit type. Retirement age is flexible between the ages 62-68, and pension entitlements accrue without any cap at the rate of 1.5 per cent of wages a year from the age of 18 to 52, 1.9 per from 53 to 62 and 4.5 per cent a year from 63 to 68. Entitlements in the earnings-related system are indexed by putting 80 per cent weight on wages and 20 per cent on changes in prices. Pensions under payment are indexed by 80 per cent weighting on prices and 20 per cent on wages. The life expectancy coefficients adjust the pensions to be paid to the changes in longevity compared to For a given retirement age longer life expectancy causes lower pension benefits per year. However, employees may compensate this reduction by postponing retirement. To prevent the increase in pension contributions further, funding for old-age pensions has been increased since 2003 with the target to reach an optimally stable trend in earnings-related pension contributions. Because of the pension reform, increases in both the earnings-related pension expenditures and pension contributions are expected to remain far below those predicted with the old system. Nevertheless, the Working Group analysing Finland s national strategy for Finland Ministry of Social Affairs and Health (2005) predicted the average contribution rate to increase by 6 percentage points from 2005 to Without the pension reform the growth of the contributions would have almost doubled. Because of the life expectancy adjustments and a reduction of early retirement pension schemes, the 3

4 reform is expected to increase the employment rate towards 2030 particularly in older age groups as their retirement ages rise. Achievement the goals set for the pension reform and the effectiveness of these goals are monitored, and a set of specific indicators have been established for that purpose. In order to get a fresh international view of the Finnish pension system, the Finnish Centre for Pensions decided in 2011 to commission an independent evaluation study of the Finnish pension scheme. The evaluation focused on the following issues: 1. The adequacy of pensions and the financial sustainability of the system 2. The policy design of the Finnish earnings-related pension scheme 3. Governance issues in the earnings-related pension scheme Professor Nicholas Barr from the London School of Economics has evaluated the two first sets of issues, and the evaluation is published in a report from The Finnish Centre for Pensions; Barr (2011). Professor Keith Ambachtsheer, Director of the Rotman International Centre for Pension Management, has evaluated the governance issues, see Ambachtsheer (2011). The effects of the Finnish pension reform and a strategy for a socially sustainable Finland are also discussed in white papers from the Finnish Ministry of Social Affairs, e.g. (2011). Barr denotes his approach in analysing the two first sets of issues for systemic. It is important to evaluate the pension system as a whole because the system may have multiple objectives and face multiple risks. Loosely he divides the risk into systemic risks (including macroeconomic risk, demographic risk and political risk), market risks and risks connected with individual behaviour. Because of market imperfections such as imperfect information, non-rational behaviour, incomplete markets and progressive taxation, formulating policy within a first-best framework may be a useful benchmark, but a bad guide to pension design. Analyses should then be framed in second-best terms, and there is no single best pension system. Barr s evaluation is based on economic considerations linked to statistical information and demographic and economic projections documented by the Finnish Centre for Pensions (2011) and Finland Ministry of Social Affairs and Health (2011). Demographic projections together with predictions for old-age pension expenditures and macroeconomic development provide a realistic picture of the effect from the pension reform on fiscal sustainability. Effects on adequacy may also be analysed in a relevant way by using simple models for typical individuals. But this approach is far from being as comprehensive, precise and consistent as the evaluation of the pension reform in Norway made by the MOSART-model regarding pension expenditures, fiscal sustainability and income distribution documented in Christensen et al. (2012) in combination with a consistent macroeconomic analysis documented in Holmøy and Stensnes (2008). 4

5 In his report Barr discusses strengths and weaknesses of the Finnish pension system. Even though the system has existed for several years, there is still a large spike in retirements at age 63. If retirement age stays constant in a system with life expectancy adjustments, adequacy will diminish. More and more persons will then only receive the national pension, and fewer will receive earnings-related pensions. On the other hand in spite of the life expectancy adjustment, average compensation rates necessary to finance the system have to increase because of larger cohorts born after The Second World War replacing smaller cohorts born before the War. According to Barr more flexible retirement in the meaning of later retirement than present is desirable. The adjustment for delayed start to pension seems to be faulty in several ways. Adjustments to earnings-related pensions for a delayed start should be applied to the pension a person has accumulated till the age of 63 rather than a higher accrual rate of 4.5 per cent on the flow of earnings after age 63. It would also be desirable if the pension system allowed partial deferral, e.g. the option to draw 25, 50 or 75 per cent of a person s pension. Policy should also gradually increase the earliest eligibility age. High incidence of poverty among single pensioners suggests a need to review benefits for that group, including the relative size of the national pension and guarantee pension for single people and couples Sweden The need for a Swedish pension reform was triggered by the crises in the early 1990s, which probably also created greater acceptance for quite radical and typically unpopular policy changes. Prior to the reform the defined benefit (DB) system required only 30 years of contributions to give a full benefit. The benefit was based on an average of the 15 years of earnings, unfavourable for individuals with long working careers. The reform was approved by Parliament in Under the new Notional Defined Contribution (NDC) System, it is optional to retire after 61 years of age, as in the Norwegian system, under actuarially fair conditions. Every year of contributions enters in the calculation of pensions. A person with an average wage will increase his yearly pension benefit by nearly 60 per cent if he/she postpones retirement until 67 years of age compared with leaving at 61. A yearly statement of account informs workers of the costs and benefits of retirement. The new system is phased in gradually for generations born between 1938 and 1953, while fully affecting those born after According to Dominczak, Franco and Palmer (2012), the de facto retirement age has increased continuously after the introduction of the NDC. However, as Dominczak et al. (2012) are more cautious than Gruber and Wise (2007) (see below) when they claim that there is no evidence so far for causality. Dominczak et al discuss the experiences with the first wave of NDC reforms by discussing the circumstances under which these systems were established, and their pros and cons with reference 5

6 to the economic performance in Sweden, Italy, Latvia and Poland. Here, the paper overlaps with other papers by Edward Palmer. There is little doubt that the authors think that NDC has been a success in many respects. Does this imply that the also the Norwegian pension system should be a more complete NDC system with an autonomous budget? Such a question is relevant in an evaluation of the Norwegian reform, especially if the majority of experts are NDC enthusiasts. Well accepted criteria for evaluations are obviously important, also in such a discussion. They may be hard to formulate, especially since issues of political economy are clearly involved, a point that becomes very clear in this paper. Karlström, Palme and Svensson (2011) represents a completely different kind of analysis of pension reform in Sweden. Their paper study the welfare effects of a hypothetical reform of the Swedish public pension system, where eligibility to pension benefits is delayed by 3 years for all pension schemes. This means that all economic incentives to exit from the labour market are delayed by 3 years. Though stylized and hypothetical, such a change is in line with a main element in most proposed or implemented pension reforms. It is also identical to the stylized reform analyzed for different countries in Gruber and Wise (2004, 2007), but the focus on welfare effects make it different from the papers in Gruber and Wise (2007) reviewed below. The welfare effects are simulated by using an option value model, which has been estimated on Swedish panel data covering the cohorts. The welfare effects are confined to these cohorts. The study offers two methodological contributions: 1) an improved measure of compensating variation, and 2) a test of the binary against the multinominal option value model. The paper concludes that the binary model can not be statistically rejected. The paper finds that the changes in economic incentives delays retirement significantly. This, in turn, reduces the welfare loss of the households compared with a mechanical calculation, which ignores behavioural effects. With retirement response the average individual welfare loss equals 90,5 SEK, whereas the public budget increases by 150,2 SEK per individual. Thus, the behavioural response implies an average welfare gain of 40 percent of the budget change. The methodological approach taken in this paper is obviously relevant and promising, especially with respect to the labour supply effects. It also clearly demonstrates the importance of including behavioural effects when assessing fiscal effects and welfare effects. However, there are some reasons to question the simulated fiscal effects. Whereas, individual taxes and government transfers seem to be taken very accurately into account through the private budget constraint, the model description says nothing about revenues from indirect taxes and taxes collected from firms/companies. When the additional labour supply is employed which is assumed in the analysis most tax bases in the 6

7 Swedish economy will expand. Most obviously, this is true for payroll taxes paid by firms. However, increased employment will also raise output and consumption, and thereby the indirect tax bases. The same will be true for the base for profit taxes and indirect taxes paid by the business sector. As a matter of fact these effects are, more or less completely, incorporated in the Gruber and Wise (2007) approach. Hence, it is not easy to see why they have been ignored here. If they really are ignored, the analysis underrates the true tax wedges between leisure and consumption. If all tax bases were taken into account, as they would have been in a proper CGE model, a given change in incentives which increases labour supply, would generate a stronger positive budget effect, a greater reduction of the individual welfare losses, and a higher social welfare gain than the actual results Denmark The "Welfare Agreement" of 2006 included substantial changes, perhaps breaking with the judgement by Kangas et al. quoted above. It reversed the 2004 decision to lower retirement age from 67 to 65. It decided to increase eligibility age of early retirement from 60 to 62 between 2019 and 2022, and to increase the minimum contribution period for receiving this benefit from 25 to 30 years The normal retirement age is to be increased from age 65 to 67 between 2024 and Retirement ages became indexed to the average life expectancy of 60-years old from 2025 onwards. This reform package affects mainly people younger than 48 years of age at the end of We are not aware of specific empirical analyses of the accumulated changes in the Danish pension system, which estimates the likely development under the actual new schemes with hypothetical scenarios consistent with other pension systems, including prolongation of the former system. However, the Welfare Commission (2006) discusses and suggested important pension reforms: Firstly, the retirement ages should be indexed to the average life expectancy. Specifically, the commission adviced that the eligibility ages should be adjusted monthly in accordance with relevant life expectancy estimates. Consistent with this, it also recommended phasing out gradually the generous early retirement scheme Etterlønnsordningen (Post-employment Wage (PEW)). The reasons for this advice are well known, and similar policy changes were included in the Welfare Agreement of The Welfare Commission used the overlapping CGE model DREAM to make many of the interesting policy discussions concrete and realistic, and DREAM included a model of the pension system. But, as far as we know DREAM was not used to highlight the effects of isolated changes in the pension system. Rather, DREAM was used to show how the proposed changes in the pension system, combined with several other policy changes, could work to re-establish long run fiscal sustainability in Denmark. 7

8 However, the Commission used the model to estimate the budget effect of early retirement. Using DREAM, it was possible to estimate the trade-off between increasing the effective age of old-age retirement and early retirement. If long run fiscal sustainability should be achieved with a stronger increase in the retirement age instead of cutting the PEW, then the retirement age would have to be increased by 2 ½ month instead of 1 month per year, when all other exogenous varaiables are kept constant. Andersen (2008) provides a theoretical discussion of indexing retirement ages to longevity. He shows that such an indexing is not sufficient for maintaining a given social security system. The paper finds that, given a political reform process based on the so-called legislative procedure, the retirement age should be raised more than proportionally to the increase in longevity, but in addition the young generations should also make larger transfers to the old ones. Denmark has also been a part of the large internationally comparative research project Social Security Around the World organized by the professors Jonathan Gruber at the Massachusetts Institute of Technology and David A. Wise at the Harvard Kennedy School. We will review this project and some of its country studies below, and we will comment briefly on the Danish country study in that context. The same applies to the results concerning Denmark in the pension analyses presented Economic Policy Committee (EPC) (2011). 3. Analyses of actual reforms outside the Nordic countries 3.1. OECD assessments of pension challenges and pension reforms Evaluation of macroeconomic policy, including pension challenges, among the member countries is an important task for the Organisation for Economic Co-operation and Development (OECD). OECDreports discussing pension challenges and pension reforms are produced almost every year, cf. OECD (2012). The pension experts in the OECD Social Policy Division are also allowed to publish signed articles regarding this topic, cf. Whiteford and Whitehouse (2006), Whitehouse et al. (2009) and Whitehouse (2012). The challenges posed by demographic changes in almost all OECD-countries are well recognized. Due to the baby boom generation approaching retirement age, falling fertility rates and increasing life expectancy, all these countries are experiencing a significant population ageing. The share of the population aged 65 years and over is projected nearly to double from 2000 to 2050 in the OECDcountries as an average. Caused by the baby-boom generation the sharpest effects will generally be felt 8

9 over the next two to three decades. These developments have major implications for public policy. OECD-projections show that expenditures for old-age pensions relative to gross domestic product (GDP) may increase by 3-4 percentage points, even after pension reforms still bring phased in are taken into account. In addition aging could increase health-care expenditures by 3.5 percentage points of GDP and spending on long-term care by 1 percentage point. Either this development has to cause higher taxes, or growth in expenditures has to be reduced. Many OECD countries have already undertaken a wide range of pension reforms, including changes in benefit formulas, changing the indexation of pensions in payments, linking pensions to higher life expectancy, increasing the role of private provision, as well as reforms designed to increase incentives for later retirement. The reports by Whiteford and Whitehouse (2006) and Whitehouse et al. (2009) provide a survey over the pension systems in the OECD countries and recent pension reforms since respectively 1970 and In the reports it is discussed what has been achieved, and what remains to be done. Whitehouse et al. (2009) link the reforms to six main objectives of pension systems: Coverage by both mandatory and voluntary schemes Most OECD and EU countries have broad coverage of their pension systems, and only a few countries have introduced changes designed to expand the coverage of the mandatory system. Most of the changes regarding coverage have mainly been aimed at private pensions Adequacy of retirement benefits Many countries have moved to strengthen retirement income safety nets to improve pension adequacy. Financial sustainability and affordability of pensions to taxpayers and contributors The main motivation for the reforms in most countries has been to strengthen the financial sustainability of the public pension systems. Although measures have been adopted to reduce pension expenditures, particularly in the long run, direct cuts in benefits are rare. Many countries have changed the earnings measure to calculate pensions from being based on only a few years of best earnings or final earnings towards the use of lifetime earnings. Change in valorisation of past earnings from economy-wide wage growth towards more weight on prices has also taken place in many countries. For pension under payment there has been a change in indexation towards full or partial indexation to prices. Adjustments of pensions in payments have also in practice been at variance with policies of automatic indexation. Another important change in pension systems designed to ensure financial sustainability has been to link pensions to changes in life expectancy. This is done in four different ways: 9

10 - Some countries have introduced mandatory defined contribution (DC) plans. When people retire, the accumulated contributions and investment returns are turned into a pension or annuity. As life expectancy increases, a given amount of capital will buy a smaller annuity. - Italy, Latvia, Poland and Sweden have introduced notional accounts schemes (Notional Defined Contributions). These are public pensions primarily financed pay-as-you-go, but they mimic some features of defined contribution schemes in that an annuity calculation is made at the time of retirement. - Finland, Germany and Portugal have introduced a link between the value of public pension benefits and life expectancy in their defined benefit (DB) systems. - Some counties, like Denmark, will automatically link pension eligibility age to life expectancy. Economic efficiency Many of the potential improvements to financial sustainability of pension systems come from measures to improve retirement incentives. A higher proportion of population in work will have a double impact because it simultaneously will increase the number of contributors to finance social protection and reduce the number to be supported. Whiteford and Whitehouse (2006) show that the trends in participation rates for people aged differ significantly between the sexes. Participation rates for older men have fallen substantially since 1970 in nearly all OECD countries, but at very different rates. However, in most countries this is offset by increasing participation rates for older women. Most OECD countries have a normal pension eligibility age of 65. Participation rates at different ages, however, imply very different effective retirement ages, and the effective retirement age for men has been declining. Thus, there is probably a significant scope to raise employment rates for older people in many countries. Since 1990 many countries have increased normal pension ages, and many countries have also tightened the qualifying conditions for early retirement. Administrative efficiency Means to reduce administrative costs, improve compliance and reduce the problems of unmatched record have been introduced in several countries. Security of benefits in the face of different risks and uncertainties All types of pension schemes are subject to a variety of risks, which may be reduced by diversification. Reforms in many countries have increased diversity in pension provisions through the introduction of mandatory defined contributions schemes. However, a disadvantage with these schemes is that they are more exposed for investment risk than public pay-as-you-go pensions. This is highlighted by the recent financial crises and is also discussed by OECD (2012). The introduction of 10

11 automatic adjustment mechanisms in public pension systems improve their sustainability, but may also raise adequacy problems as the risk of higher life expectancy is shifted towards future pensioners. Whitehouse, D addio, Chomik and Reilly (2009) describe the main changes in the pension systems in 38 industrialized countries over the last decades, and assess to what extent targets have been achieved, including improving long run fiscal sustainability. Despite many differences, there has been a general trends in favour of less generous pension promises. The description of the Norwegian system and reform seems misleading in some respects. The paper uses the OECD pension models for 20 countries that have had major pension reforms to illustrate the magnitude of the change in various transformations of individual pension entitlements, including the pension wealth and compensation ratios. The simulations capture mechanical effects only, i.e. they neglect behavioural responses and equilibrium repercusions. The model does not simulate representative or average individual behaviour. For the sake of comparability, they compute the changes facing stylized type individuals, all of which are full career workers entering the labour market in 2006 at the age of 20 and working full time in all years until the standard pension eligibility age. Results are provided for there earnings levels: 50,100 and 150 per cent of the economy wide average. As replacement rates do not capture the impact of many changes to pension systems, the paper also calculates pension wealth, i.e. the lifetime value of benefits at the time of retirement. These calculations take into account the differences in pension age, life expectancy and indexation procedure between the countries. Two stylized scenarios are compared: 1) Pre-reform calculates the benefits contingent on the prereform rules; 2) Post-reform does the same, but uses rules applying in As a simplification, all new rules are fully in place in 2006 even if they actually are phased in gradually. The calculations capture all mandatory private and public pension schemes. The assumptions about relevant economic variables are the same for all countries. Pension reforms will cut lifetime pension benefits in 18 of 20 countries for which entitlements were modelled, on average by 22 (23) percent for men (women). Assuming an average economy wide earnings profile, the net pension wealth in Norway increases from 7.7 to 8.4 times annual earnings. The increase is because private pensions have become mandatory. In Sweden the corresponding change is from 8.7 to 7.1, and in Finland from 8 to 6.6. On average, pension spending falls with 4.4 percent in (However, this estimate is taken from the European Commission (2009).) The referred spending reduction averages half of the increase in 11

12 pension spending due to demographic changes until The authors therefore conclude that the pension reform agenda is not finished. Changes since the paper was published (written) in 2009 have probably reduced the relevance of some of the following policy recommendations, or made them more necessary: A fundamental overhaul is needed in Greece, Luxembourg, Slovenia and Spain. The paper claims that the reform process has stalled or gone into reverse in some countries including Austria, Ireland, Norway, the US and Italy. In many countries private savings have not increased sufficiently to compensate for the cuts in public benefits. This may cause an increasing problem of inadequate income among the elderly. A poverty problem may also emerge among individuals with low earnings as the new systems include a stronger pension earning link. Too many pension systems still encourages early retirement Assements of pension reform effects in the projections from European Policy Committee (EPC) In EPC (2011) the so-called Ageing Working Group (AWG) presents and discuss long run macroeconomic projections for the EU Member States. In addition to the main macroeconomic aggregates, the projections include pension expenditures. The projections are derived as a joint work between AWG and specialists in each country. The development of pension expenditures are constructed from assessing the development of the following components: 1) Population by age and gender, and the demographic dependency ratio, 2) the coverage ratio, 3) the participation and employment rates, 4) the benefit ratio. The relevance in this review of these projections is that they try to identify the contribution from recent pension reforms. However, these pension reform effects are most explicit for the employment projections. However, the employment rate is highly relevant for fiscal sustainability evaluations. One may question if such work is research, but if the work is properly done, the relevance can hardly be questioned. The participation rates should capture country-specific information about the relationship between retirement behaviour and the parameters of the pension system, together with cross-country econometric evidence of the impact of changes in the implicit tax rate on continuing work and retirement decisions. The typical distribution of the retirement age tends to have spikes/modes at both the minimum age for early retirement and the statutory retirement ages. The exit age from the labour force increased steadily in most EU countries between 2001 and The average exit age was 61.4 years in 2009, 1½ years higher than in Projections 12

13 show an average increase of 1.9 in the effective retirement rate for men. In most of the 22 EU Member States that have recently legislated pension reforms, pension reforms are projected to raise participation rates of older workers (aged 55 to 64) significantly. Specifically, averaging over all countries in the EU27, the participation rate of workers aged is projected to increase from the present level by 7.7 %p in 2020, 13.2 %p in 2040, and 13.8 %p in 2060 as pure result of pension reforms. In the euro area, the impact is estimated to be even larger, reaching 15.4 %p in The strongest effects are expected in Germany, France, Hungary, Italy, Slovenia and Slovakia. One should keep in mind that these effects will have rather modest impacts on the total participation rate, since the initial participation rates are rather small for old workers. The downward impact on pension expenditure of the coverage ratio has become stronger in the recent projections (-2.9 p.p. vs p.p. of GDP). This reflects changes in pension policies that have aimed at increasing the effective retirement age either through increases in the statutory retirement age and/or through increases in the career requirements for full pension requirements and/or tightened access to early and disability pension schemes. In comparison with the 2009 projection results, especially Luxembourg, Greece, Italy and the Czech Republic record a substantially higher downward impact of the coverage ratio on pension expenditure. 1 On the opposite, a lower impact is projected for Malta and Cyprus. A rather small employment effect contributes to offset the dependency effect on public pension expenditure in most EU countries. In most EU countries the a negative benefit ratio effect has become stronger between the projection rounds I 2009 and On the EU27 level, the effect has changed from -2.6 p.p. of GDP in 2009 to p.p. of GDP in 2012 in the 2012 projections. It is claimed that this reflects pension reforms in many countries. Figure 1. Effects of implemented and proposed pension reforms on average exit age from the labour force in As cross-border workers in Luxembourg are not covered in the labour force projections for the pension projection exercise, a deeper analysis of the employment effect contribution as well as the coverage ratio contribution is not meaningful. 13

14 SE NL CY PT BG FI RO UK EE ES LT DK PL SK AT EU27 DE GR EA17 MT HU FR CZ SI IT Country Average exit age from the labour force in 2060 IT DK UK SE DE CZ CY ES PT EE BG PL LT GR NL MT FI RO HU SI FR AT SK Kilde: Commission services, EPC (2011). MEN - avg exit age (no reform) WOMEN - avg exit age (no reform) MEN - impact of pensions reforms WOMEN - impact of pensions reforms Age Figure 2. Effects of pension reforms on labour market participation rates in EU. Deviations from a scenario with no pension reforms. Percentage points Kilde: Commission services, EPC (2011). 14

15 3.3. Reforms in Latinamerica Cerda (2008) discusses the effects on aggregate growth, government finances and main distributional patterns of the pension reform in Chile of Chile was one of the first countries to implement a reform which transformed a Pay-as-you-go (PAYG) system into an individual account system (IA). In the new system individuals save a constant fraction of their labour income in an individual account managed by private firms. The author could base the 24 ex-post reform evaluation on 24 years of observations. The counterfactual scenarios are based on the parametric rules of the old PAYG system before and after the revisions in Thus, the analysis compares two versions of the old PAYG system and the new IA system. The scenarios are simulated from 1981 until 2050 by a calibrated Computable General Equilibrium (CGE) model with Overlapping Generations (OLG). The model specifies 10 groups of representative agents within each cohort, differing with respect to initial human capital endowments. Thus, the analysis captures some aspects of income redistribution. Each agent maximizes a discounted sum of utility obtained in each year over their life cycles. They decide the time paths of consumption, labour supply until retirement, and the time of retirement. The agents can also choose to work in the informal rather than the formal sector. Informal work implies no taxes, as well as no pension entitlements beyond the minimum benefit. Flexible prices ensure that the aggregate labour supply, consumption and savings are consistent with rational producer behaviour and the economy wide constraints. The simulations show that maintaining the PAYG system would have generated much weaker fiscal development than actually observed both historically and between 2005 and With the PAYG system the author concludes that Chile would have faced a cricis in government finances by the end of the 20th century. The parametric reform of the PAYG system in 1979 delays the crisis until The IA-reform appears to have been a sufficient step to avoid such financial crises. Interestingly, the analysis is able to discuss quantitatively the well known transition problems of moving from PAYG-system to a funded system, in this case with IA. In the transition period, pension benefits must be paid to individuals who have retired prior to the reform, and bonds equal to the accumulated entitlements are paid to individuals who have switched to the IA system but contributed to the PAYG system. The transition costs are considerable. However, they are somewhat mitigated by stronger growth in employment, capital, output and tax bases caused by the reform. Pension benefits would have been smaller in the PAYG scenarios than in the IA scenario, and a larger fraction of the retirees would receive only the minimum benefit. 15

16 4. Analyses of stylized reforms outside the Nordic countries 4.1. Social Security Around the World: Gruber and Wise (2007) For several years the professors Jonathan Gruber at the Massachusets Institute of Technology and David A. Wise at the Harvard Kennedy School have organized an international comparative project which study the relationship between social security program provisions, retirement and government finances. The project relies on country specific analysis for 11 countries, each one conducted by analysts in that country. To make the country study as comparable as possible, they are based on comparable descriptive data and analytic calculations. The first phase (Gruber and Wise 1999) described the retirement incentives inherent in existing pension systems, as well as the proportion of older persons out of the labour force. Gruber and Wise conclude that these studies reveal a strong correlation between these variables, and that the relationship is largely causal. In the second phase (Gruber and Wise 2004) the country studies estimated the effects of the retirement incentives on retirement using the comparable micro data sets and the same econometric models. Gruber and Wise conclude that these studies show that the retirement incentives have a strong causal effect on retirement. The estimated effects vary across countries, but they are all large and similar in countries with very different cultural histories, labour market institutions, and other social characteristics. The third phase of the project studied the issue which is most relevant for this survey. Here, all country studies report the fiscal implications of the following 3 stylized but relevant reforms of the public oldage pension system: (1) Three-year increment in eligibility ages: All eligibility ages are increased by three years, including the early retirement age, the normal retirement age, and the eligibility ages of receipt of other related to disability, unemployment, or other retirement pathways. (2) Actuarially fair: Benefits are reduced (increased) actuarially fair if taken before (after) the normal retirement age. (3) Common reform (or new system): In all countries, the early retirement age is set at age 60 and the normal retirement age at 65. Benefits taken before (after) age 65 are reduced (increased) actuarially, by 6 percent for each year before (after) age 65. The replacement rate at age 65 is set at 60 percent of (projected) age 60 earnings. 16

17 In each country study include the following simulations: 1) Use the retirement models estimated in Phase II to predict the distribution of retirement ages, under current law (the base case). 2) Compute the fiscal position of the cohort, i.e. the total expected benefits paid to the cohort and total expected taxes paid by the cohort, given the computed base distribution of retirement ages. 3) Use the retirement models to predict the distribution of retirement ages, as well as the corresponding fiscal position of the cohort under each of the 3 reforms. 4) Calculate the fiscal implication of reform as the difference between fiscal positions under the base and the reform systems. 5) Divide the fiscal implication into two components: i) The mechanical effect (effect ignoring behavioural changes in retirement ages); ii) the behavioural effect (the additional incremental effect due to retirement response). Reforms were simulated for each country for 6 different combinations of incentive measures and simulation approaches. Retirement incentives are measured by the option value and the peak value of the Social Security Wealth (SSW). SSW is the expected present value of future pension benefits. The SSA is the difference in SSW by postponing retirement by one year. The peak value of SWW is the maximum difference in SSW between retiring at future ages and retiring at the current age. The option value and the peak value are proportional to each other. In the option value model the individual compare the value in discounted utility terms - of retiring now to the maximum of the expected values of retiring at all future ages. The value of retiring at future ages includes both possible pension additions and future earnings. The simulation methods correspond to different approaches to the econometric identification problems, especially: how should one distinguish the effect on retirement of the first eligibility age from the effect of the incentive measure, given eligibility? Retirement jumps from almost zero at the first eligibility age in almost all countries. To address this and other identification issues each country study estimated two different model specifications with respect to age: i) linear age trend; ii) agespecific dummy variables, which represent taste for leisure and do not change when the pension system changes; iii) age-specific dummy variables, and the deviation between these and the linear age trend is assumed to represent effects of changes in the retirement in the pension system. The results were robust to this choice of specification. Gruber and Wise (2007) prefer to present results based on the linear age trend approach. Gruber and Wise (2007) conclude that the simulations show that changes in retirement incentives as in the 3 stylized reforms, can have very large fiscal implications, both for pension expenditures and for government revenues. 17

18 Three-year increment increment in eligibility Fiscal net improvement is measured by changes in expenditures minus changes in tax revenues. Measured by simulations on the option value model allowing changes in the age dummies when the retirement rules changes, the fiscal net improvement caused by the Three-year increment in eligibility ages averages 27 % of the pre-reform benefit costs (1% of GDP) over all 12 countries. Gruber and Wise believe that these averages reflect the most likely long-run effect of the illustrative reforms. UK (authors: R. Blundell and C. Emerson) would experience the strongest fiscal gain, 42%, (83% of GDP). 1/3 of this gain can be attributed to the mechanical effect. The contribution from behavioural effects is twice as large. The authors note that part of the gain will be offset by increased spending on means-tested income support and disability benefits. For Germany the corresponding effect is 36 % (0,60% of GDP). Also the US and the Netherlands would experience a fiscal gain above the average. The nature and the relative importance of the different contributions to the total fiscal effect can be illustrated by the study for Canada (authors: M. Baker, J. Gruber, and K. Milligan). The Three-year increment reform reduces the average individual present discounted value (PDV) reduces future benefits by 20, choosing thousand Euros as unit. Total taxes, including payroll taxes, income taxes and consumption taxes increase by 5. Consumption tax revenues are imputed based on the income associated with each policy. Payroll tax revenues include the share of general revenues that are associated with Social Security programs, as imputed in each country. The fiscal implication is the change in benefits minus the change in taxes, i.e = -25, corresponding to 22.3 percent of the base benefit costs and 0,40 % of GDP. The mechanical effect contributes with 19 units of the total benefit reduction of 20. The reduction in government benefits payments minus revenues ranges from about.30 to.45 percent of GDP, depending on the estimation method. The behavioural effect, i.e. the increase in the typical retirement age, is slightly negative due to some peculiarities in the Canadian system. Turning to taxes, the mechanical effect reduces total taxes by 5, whereas the behavioural effect raises taxes by 10 reflecting delayed retirement. Actuarial Fair reform On average across all countries the fiscal implication of the Actuarial Fair reform is a net fiscal gain equivalent to about 26 percent of the base benefit cost. However, there are large variations between countries, depending on the extent of actuarial fairness in the pre-reform schemes. The Actuarial Fair reform has no effect in Canada because benefits are already adjusted actuarially under the pre-reform base system. For the same reason the fiscal implication is small in the US (authors: C. Coile, J. Gruber and P. Diamond). On the other hand, the degree of actuarial fairness was relatively low in the prereform pension systems in most European countries. 18

19 Germany (Country study authors: A. Børsch-Supan, S. Kohnz, G. Mastrobuoni, and R. Schnabel) is found to have the strongest fiscal effect of the Actuarially Fair reform. The pre-reform system in Germany is considered as generous. There are no actuarial adjustments in benefits taken before or after the normal retirement age of 65, once benefits are available. (This has changed somewhat after the study was carried out.) The incentives to retire early are very strong. In addition to the social security program per se, a large fraction of workers in Germany retire through disability and unemployment programs, which essentially provide early retirement benefits before the age 60. Replacing this system with the Actuarially Fair system would increase the mean retirement age for men by about 3 years for both men and women. In terms of thousands Euros, the fiscal implication of this reform is a net budget gain of 83 resulting from a benefit reduction of 17 and an increase in total taxes of 49. The net gain equals 18 % of the base benefit costs and 1,2 % of GDP. The contributions to the net gain from the mechanical and the behavioural effects, equal 35 and 48, respectively. The increase in employment due to delayed retirement implies a behavioural effect 51 on total taxes. Increased earnings imply a positive behavioural effect on benefits equal to 20. The increase in contributions are only slightly less than the increase in benefits in Actuarial Fair system. In France (Country study authors: E. Walraet, R. Mahieu, D. Blanchet and L-P Pele), the pre-reform early retirement age in France is 60, and the normal retirement age depends on the number of "validated" participation quarters. There is no actuarial adjustment of benefits if they are taken after the age of first eligibility. If benefits are taken before the normal retirement, they are reduced more than the actuarial adjustment would be. The Actuarial Fair reform would prolong labour market participation and would increase benefits for many retirees. Thus, both the behavioural and the mechanical effects of the actuarial reform increase the cost of benefits. The total fiscal effect of this reform is net budget loss equal to 10% of the pre-reform benefit costs. Common reform The fiscal effect of the Common reform varies greatly. In Canada, the US and the UK the common system would be more generous than current system, whereas the opposite is the case for the 9 European countries. The UK would experience a decrease in the net government revenues equivalent to 80 % of the pre-reform benefit costs. The corresponding figures in Canada and the US are approximately 40 and 60 %. Denmark and Sweden are the only representatives of the Nordic countries in Gruber and Wise (2007). The Swedish study is carried out by M. Palme and I. Svensson. The fiscal improvement in Sweden of Three-year increment 19