SP DISCUSSION PAPER NO. 9924 2085 1 Pension Plans and Retirement Incentives Richard Disney Edward Whitehouse August 1999 fILE GOPY J Pron~ttion LABOR MARKETS, PENSIONS, SOCIAL ASSISTANCE T H E W O R L D B A N K Pension plans and retirement incentives Richard Disney and Edward Whitehouse August 1999 6' ,, Pension plans and retirement incentives Richard Disney and Edward Whitehouse" May, 1999 Richard Disney is Professor of Economics at the University of Nottingham, a Research Fellow of the Institute for Fiscal Studies and a Director of Axia Economics. Edward Whitehouse is also a Director of Axia Economics. Comments welcomed at Axia Economics, 38 Concanon Road, London SW2 5TA; telephone +44 171 737 5815; e-mail info@axiaecon.com; web: http://www.axiaecon.com. We are grateful to Axel Borsch-Supan and participants at a Learning and Leadership Centre seminar at the World Bank in November 1998 - particularly Estelle James, Robert Palacios, Anita Schwarz and Pat Wiese - for helpful comments and advice. The usual disclaimer applies. 1 Table of Contents 1. A model of optimal retirement ............................................. 6 2. A simple retirement savings plan ............................................. 7 3. A simple defined- benefit plan ............................................. 9 4. The effect of the pension system on the return to working . . ......................... 10 4.1 The effect of a defined-contribution plan on the return to working .................................... 10 4.2 The effect of a defined-benefit plan on the return to working ............................................ 11 4.3 Defined-benefit and defined-contribution plans compared . ................................. 12 5. Measuring work incentives ................................................. 13 5.1 Effective tax rates..13 6. Extensions to the basic model: taxation . . .14 7. Extensions to the basic model: defined-contribution schemes . . . 15 8. Extensions to the basic model: defined-benefit schemes . . . 17 8.1 Non-linearities in pension accrual ..17 8.1.1 Effect on workincentives .18 8.2 Final, average and best salary schemes ..19 8.2.1 The impact of final salary schemes on work incentives .20 8.2.2 What do 'true' age-earnings profiles look like? .21 8.2.3 Age-earnings profiles and work incentives .23 8.3 'Actuarial' adjustments to defined-benefit pensions ..23 8.3.1 Actuarially neutral adjustments .25 8.3.2 The impact on work incentives .25 8.4 Ea rning while drawing pension ........................ 26 8.4.1 Partial retirement programmes .27 8.4.2 Combining work and pensions: reforms in the United Kingdom ............................... 28 8.4.3 Combining work and pensions: reforms in the United States ...................................... 29 8.4.4 Combining work and pensions: policy conclusions .............................................. 30 9. Conclusions, policy implications and future developments .............................................. 30 9.1 Recent policyinitiatives in OECD countries .................... .......................... 31 10. References...40 Table of Tables and Figures Figure 1. A static model of the retirement decision ...........................................................6 Figure 1. Earnings and defined-contribution pension by age ...........................................................8 Figure 2. Gross replacement rate by age, defined-contribution plan ..................................................8 Figure 3. Annuity rates by age and sex ..........................................................8 Figure 4. Earnings and defined-benefit pension by age ...........................................................9 Figure 5. Gross replacement rate by pension type and age .......................................................... 10 Figure 6. Earnings and change in pension wealth by age, defined-contribution plan ............ ....... 11 Figure 7. Earnings and change in pension wealth by age, defined-benefit plan ............................. 12 Figure 8. Change in pension wealth by age and type of plan .......................................................... 12 Figure 9. Adjusted replacement rate by age and type of plan .......................................................... 13 Figure 10. Effective tax rate on working by age and type of plan ...................................................... 14 Figure 11. The impact of a progressive personal income tax on replacement rates .... ................... 15 Figure 12. The impact of voluntary annuitisation on incentives in defined-contribution plans ... 16 Figure 13. Accrual rates by years of contributions .......................................................... 18 Table 1. Earnings measure in public, defined-benefit plans: OECD countnres ................. ............. 19 Table 2. Earnings measure used in public, defined-benefit plans: non-OECD countnres ............ 20 Figure 14. Defined-benefit pensions under average and final salary formulae with rising earnings .......................................................... 21 Figure 15. Age-earnings profiles by occupation, United Kingdom ................................................... 21 Figure 16. Relative earnings of older workers (ratio of 55-64 year olds' pay to 45-54 year olds) .22 Table 3. 'Actuarial' adjustments in defined-benefit plans .......................................................... 24 Figure 17. Neutral actuarial adjustment to defined-benefit pension by age .................................... 25 Figure 18. Impact of actuarial adjustments in defined-benefit schemes on replacement rates .... 26 Table 4. Earning while drawing pension, OECD countries .......................................................... 27 Table 5. Recent policy initiatives to promote employment of older workers .................................. 33 Annex 1. Accrual rates by years of contributions ....................... ................................... 34 3 Pension plans and retirement incentives Richard Disney and Edward Whitehouse The well-documented decline in the labour-force participation of older women and older men (in particular) is common to most industrialised countries. The proportion of men aged 55 to 64 in employment fell between 1980 and 1996 in all 17 OECD countries for which data are available, by an average of more than ten percentage points. The average employment rate of men in this age group in 1996 was a little under 60 per cent.' The reasons for this are complex, but probably involve both a demand effect - high and persistent unemployment, especially in Europe - and a supply effect - pension benefits and the value of other savings have increased. Many governments are worried about the decline in labour-force participation of older workers. Of 28 countries responding to the OECD's Ca?jg World questionnaire, 15 mentioned low effective retirement ages and poor work incentives in the pension system as a matter of policy concem.2 The population of most countries is ageing due to a mix of declining fertility and increasing longevity. It is desirable to encourage people to retire later to counterbalance the effect of ageing on the ratio of workers to dependants. Some, such as Chand and Jaeger (1996), have argued that this can be achieved with 'parametric reforms', tinkening with the rules of existing defined-benefit schemes. Many countries, however, have introduced or proposed more radical reforms emphasising the role of privately managed defined-contribution pensions. An obvious question is how these regimes are likely to effect retirement behaviour. We begin by modelling a simple retirement saving plan and looking at the optimal retirement date. This simple plan looks very sirnilar to a defined-contribution scheme. Optimal retirement depends' on prospective earnings and the evolution of the accumulated fund, which, in principle, are separable. We then move on to defined-benefit pension schemes, which are the norm in public and much private provision. Here there are significant interactions and complications. The pension formula is often non-linear, with accrual rates that vary with the number of years of contributions and formulae that depend on a lirnited number of 'best' or 'final' years of earnings. There are also 'spikes' when early retirement is first permitted, at the I OECD (1998b). 2 These were Belgium, the Czech Republic, Denmark, Finland, Germany, Greece, Hungary, Italy, Korea, the Netherlands, New Zealand, Norway, Poland, the Slovak Republic, Spain and Turkey. Kalisch and Aman (1998), Table 13. 4 standard retirement age etc. Pensions can be actuarially adjusted, depending on the year at which benefits are first drawn. We show that the incentives in a defined-benefit scheme are very different from the defined-contribution retirement saving plan. Most of the existing literature on retirement focuses on public-sector defined-benefit schemes.' Another strand looks at the effect of private defined-benefit schemes, commonly provided by employers.4 Defined-contribution schemes, however, are playing a bigger part in pension systems through out the world. In the United States, employer-based defined- contribution plans, known as 401(k)s, have tended to substitute for traditional defined-benefit schemes.5 The trend to defined contribution among employer plans is less pronounced in the United Kingdom, but many employers expect to change their provision in this direction in the future.6 Most of the growth of defined-contribution plans has been in individual pension accounts, known as personal pensions, which have mainly substituted for the public-sector, defined-benefit scheme. They now cover more than a quarter of employees. In addition, the new stakeholder pensions, announced in November 1998, will be group defined-contribution plans.7 In the new pension systems of Latin America and Eastern Europe, defined-contribution schemes are also a substitute for the public plan. Australia, Italy and Sweden have also increased the emphasis on privately managed defined-contribution schemes.8 Since all these schemes are recent, few people have retired with substantial defined- contribution pensions, so it is not surprising that the issue of retirement incentives in these plans has not yet been addressed. The object of this paper is to examnine the impact of type of pension scheme on retirement behaviour. Trends in labour-force participation of older workers and demand-side issues are addressed in a sister paper (Disney and Whitehouse, 1999a). The structure of the current paper is as follows. The following section describes a simple, theoretical model of optimal retirement. Section 2 introduces an empirical model of a simple retirement-savings plan, or defined-contribution pension scheme. Section 3 compares this with a defined-benefit scheme. Sections 4 and 5 examine the effect of these pension plans on work incentives. The following three sections extend the basic model to introduce 'real-world' features of pension plans. Section 9 concludes, examines the policy implications of our results and sets out an agenda for future research in this area. Recent intemational comparative studies include one of 11 OECD countnres coordinated by Gruber and Wise (1997, 1999). Bl6ndal and Scarpetta (1998) analyse a panel of 15 OECD countries, and draw on rmicroeconometric studies of five countries commissioned by the OECD. Other notable studies include Boskin and Hurd (1978), Burtless (1996), Hurd and Boskin (1984), Mitchell and Fields (1984) on the United States, and Meghir and Whitehouse (1997) on the United Kingdom. 4 Examples include Kotlikoff and Wise (1987), Lumsdaine, Stock and Wise (1990, 1994) on the United States; Disney, Meghir and Whitehouse (1994) on the United Kingdom; Palme and Svenson (1997) on Sweden; and Seike (1989, 1997) on Japan. 5 See Gustman and Steinmeier (1992), Ippolito (1995), and Kruse (1995) on the growth of defined- contribution schemes and the reasons for this trend. 6 Disney (1995). 7 Department of Social Security (1998). See also Disney, Emmerson and Tanner (1999). 8 See Queisser (1998) on Latin America, Palacios and Rocha (1998) on Hungary, Gora and Rutkowski (1998) on Poland, Flanagan (1999) on Australia and Tumbarello (1999) on Italy. 5 1. A model of optimal retirement Retirement is often perceived as an 'institutional' decision. Nevertheless, the decline in labour-force participation of 55-64 year olds, usually below the standard pensionable age, suggests at least some element of individual choice in the retirement process. The basic labour- supply model, with individuals mnaxirnising utility (defined over leisure and consumption) subject to a budget constraint can be extended to explore retirement. In this model, consumption is financed from income, which in tum comprises both earnings and the return on accumulated assets. In a lifetime model, we can also include the desire to bequeath some of the accumulated assets. However, this basic modeL does not in itself provide a rationale for retirement as we know it. If underlying wages and the utility of leisure are constant over the lifetime, there is no necessary reason for someone to work continuously until a particular age and then leave work for the rest of their life. The optimum would probably be reduced, but continuous, employrnent throughout the lifetime. We need further structural assumptions to identify an econormic motive for retirement. A straightforward explanation of retirement relies on systematic changes with age either in wages (assumed to reflect age-related changes in productivity) or in the utility of leisure or both.9 Figure 1 illustrates this story. Wages are assumed to follow an inverted-U pattern (see the discussion in section 8.2.2 below) and the marginal value of leisure to rise with age. In this model, individuals will work until the age, R, when the curves cross. R is a unique point. Note also that R is not the point at which either wages or the rnarginal utility of leisure are at their maximum. Figure 1. A static model of the retirement decision Wages/marginal value produ Marginal utility of leisure R age Note too the importance of the assumption that changes in the marginal product and utility of leisure are smooth, continuous and are characterised by simple functional forms. In more complex models there may be local equilibria, emphasising the need not for a static model but a cynx7dc model of the retirement decision. In particular, non-linearities may arise because 9 See Lazear (1986) and Disney (1996), pp. 201-203. 6 pension plans contain complex accrual structures. In this case, the individual should appraise the prospective gain in net utility (leisure less access to extra consumption) from retiring now relative to any point in the future. This is a more complex problem to analyse'° but intrinsically evaluates the continued 'return to working' at all points in time. To show how pension schemes differentially affect this return to continued working, the next sections examine a number of stylised cases. 2. A simple retirement savings plan We first consider a mandatory defined-contribution plan with a 10 per cent contribution rate. We assume the pension earns an investment return of 5 per cent a year. When the individual retires, he or she converts the fund to an annuity. We calculate the annuity rate using the 'riskless' interest rate, which we assume initially to be 2 per cent. To calculate pension benefits, we need the individual's lifetime earnings (and so contribution) profile. Initially, we assume a simple linear growth in earnings of 3 per cent a year and that the individual contributes from age 20.'1 Figure 2 shows the assumed earnings profile and the model's calculations of pension benefits. Figure 3 shows the gross replacement rate: the ratio of the annual pension to (current) earnings. The replacement rate at age 50 is very small: around 12 per cent. However, the pension grows rapidly, at around 12 per cent a year. The two lines move closer together in Figure 2, and the replacement rate in Figure 3 rises quickly. Each extra year of work adds to the accrued pension in three ways. First, an extra year's contributions are made. Secondly, the accumulated fund earns an additional year's investment returns, assumed to be 5 per cent. Thirdly, the year's delay in annuitising the pension means that the benefit increases in line with annuity rates. Figure 4 shows these annuity rates, calculated using mortality data from Thailand.12 These are actuarial, not market annuity rates."3 We have so far modelled pension benefits only for men since they are most affected by early retirement. ID Stock and Wise (1990, 1991). 1l The impact of more complex earnings profiles is examined below. 12 The annuity rate is the inverse of the discounted present value of the product of one rminus the mortality rate. 13 Market rates will tend to be lower, because of administrative charges and adverse selection. See Piggott and Doyle (1999), Friedman and Warshawsky (1988, 1990), Brugiavini (1993) and Dilnot et at (1994), pp. 148-151 for a discussion. 7 Figure 2. Earnings and defined-contribution pension by age 50000 - 40000 -./ 40000 earnings - 30000- a) ¢ 20000- 0~ 10000 penson 0 50 55 60 65 70 age Figure 3. Gross replacement rate by age, defined-contribution plan 70 - c D 60 -.6 40- c 30- E (D 20- (D 10 a) 50 55 60 65 70 age Figure 4. Annuity rates by age and sex 8 7- 7 c a) u 6 -//6 o men 4 3 - c ~~~~~women 2- c 1 5'0 55 60 65 70 age 8 3. A simple defined-benefit plan The initial defined-benefit scheme is based on average lifetime earnings."4 Early vears' earnings are uprated in line with prices. To allow direct, fair comparisons of retirement incentives between the defined-benefit and defined-contribution plans we equalise the resulting pension value at age 65. This gives an accrual rate of 1.7 per cent in the basic defined-benefit scheme. Again, for equivalence with the defined-contribution plan, we assume a 10 per cent contribution rate to this plan.'; Figure 5 shows the pension value by age (vhich is comparable to Figure 2). The defined benefit pension is much flatter across the lifecycle than the defined contributon scheme, increasing by around 5 per cent a vear compared with 12 per cent in the latter. Each additional year of work increases the pension in txvo ways. First, an extra vear's contribution adds to the number of vears in the defined-benefit formula. Secondly, the base for the defined benefit is increasing. Since we assume that real earnings grow continuously by 3 per cent a year, average lifetime earnings are about 11/2 per cent higher after each extra year of work. Figure 6 compares the gross replacement rate in the tWo baselne plans. As noted above, we have ensured that the pensions are equally generous at age 65, so differences between the curves reflect only intrinsic structural variation bet-ween the two types of plan. The curves thercfore intersect at age 65. The defined-benefit plan gives a much larger prospective replacement rate at early ages and is relatively flat. The defined-contribution pension increases close to exponentially with age. Figure 5 Earnings and defined-benefit pension by age 50000 - 40000 - earnings . 30000- , 20000- 10000- pension 50 55 60 65 70 age 14 Schemes based on fewer vears of earnings (e.g. final) are discussed below. l; This essentiall1 assumes that the real return on individuals' contributions to the defined-benefit plan is equal to the funded defined contribution plan. But revenues to the defined benefit scheme are the total wage bill multiplied by the contribution rate. If real returns exceed wage-bill growth, there xvill tend to be a deficit in the defined-benefit plan that must be financed from general revenues. 9 Figure 6. Gross replacement rate by pension type and age 70 - ° 60- ° 50 - defined benefit 40- c 30- a) E a) 20 - defined contribution 0 cl. 10 - 10 50 55 60 65 70 age 4. The effect of the pension system on the return to working The gross replacement rates in Figure 5 give some indication of retirement incentives. In a static framework, we would expect to see fewer people working the higher the replacement rate, both because of a substitution effect and an income effect (the higher the replacement rate, ce/ens pan bus, the higher is lifetime income). However, this ignores the dynamic nature of the retirement problem because it fails to capture all of the financial returns to continued work versus retirement. Working an additional year not only brings in earnings, but will also alter the value of the pension. The pension system can be thought of as an implicit tax or subsidy to continued working. Therefore, the change in pension wealth needs to be added to earnings to show the true, total reward for working."6 As noted above, working an additional year increases a defined-contribution pension in three different ways. We can ignore the contributions made in the year in question when measuring the reward to work, because they will be converted into an annuity next year. This is not the case in the defined-benefit plan, where contributions are not directly related to benefits. However, the other two sources of increased pension value - the additional year's investment returns and the higher annuity rate from delaying annuitisation for a vear - do increase pension wealth. However, the higher annuity rate in a year's time partly reflects the risk of dying within the year. This mortality risk (slightly) reduces the value of continuing work and deferring the pension, reflecting the odds of dying before receiving any pension.7 4.1 The effect of a defined-contribution plan on the return to working Figure 6 shows how the change in pension wealth vanres with age and, through the profile of earnings plus change in pension wealth, the impact on the return to working. At younger ages, the rate of growth of pension wealth is only slightly increasing or flat, because the value of the fund increases with each extra year's contributions. But at older ages, this is overtaken by the mortality risk (again, taken from the Thai data), and delaving retirement a year then reduces Lazear (1986). This effect was not included in Figure 5. See Gruber and \W'ise (1997. 1999) for a discussion of this issue. 10 pension wealth. The top line of the Figure shows the total reward to continued working: wages plus the increment to pension wealth. Figure 7. Earnings and change in pension wealth by age, defined-contribution plan 50000 - earnings + pension wealth 40000 - a) 30000 - :i ~~~~~earnings a 20000 - Cu 010000- change in pension wealth 0 0- -10000 - I , ,. , 50 55 60 65 70 age 4.2 The effect of a defined-benefit plan on the return to working Figure 7 gives a similar picture for the stylised defined-benefit pension. Pension wealth increases from an extra year's work through the additional year in the defined-benefit formula and through the increase in the earnings base (since real pay is assumed to grow each year). Working in the opposite direction, deferring drawing the benefit reduces pension wealth. Delaying the pension by one year also incurs mortalitv risk over the year, increasing the odds that the person might die without drawing any benefits. Finally, contributions are neutral in the defined-contribution scheme because the pension value equals contributions plus their associated investment return. In the defined-benefit plan, the pension benefit earned from a year's contributions can be greater or less than the value of the contributions paid (but is never equal). We therefore need to deduct contributions from the change in pension wealth and from earnings, to give the net return to working. These are the lower lines of each pair in Figure 7. 11 Figure 7. Earnings and change in pension wealth by age, defined-benefit plan 50000 - 40000 - earnings a 30000 - a 20000 - earnings - contributions + pension 10000 change in pension wealth 0- -10000- pension + contributions I I I, I 50 55 60 65 70 age 4.3 Defined-benefit and defined-contribution plans compared Figure 8 shows the change in pension wealth alone for the two types of scheme. The difference in the pattern between the two plans is much clearer in this Figure. The change in pension wealth is broadly liriear and downward sloping for the defined-benefit scheme"8, while the defined-contribution scheme is at first flat and then falls exponentially. This again implies a very different pattern of retirement incentives in the two plans. Figure 8. Change in pension wealth by age and type of plan L 2000- defined contribution c \ 0 0' -2000 - _C defined benefit c -4000- .C -6000 50 55 60 65 70 age Other studies, such as Kotlikoff and W ise (1985, 1989), havc similar findings. 12 5. Measuring work incentives We use two simple measures of work incentives: the replacement rate and the effective tax rate. Static studies, such as the OECD analvsis of tax and benefit systems and work incentives for prime-age workers (OECD, 1996a, 1997a), often use measures of replacement rates. We can adjust the static, gross replacement rates shown in Figure 5 to take account of the effects of working on pension contributions and the pension value. The replacement rate becomes: pension/(gross earnings + change in pension wealth) Figure 9 shows the results for the trvo different plan types. The pattern is similar to Figure 5, but the results are more pronounced. The replacement rate of the defined-benefit plan increases more rapid:lv when account is taken of the fall in pension wealth and contributions which (it is assumed) continue to be levied. Figure 9. Adjusted replacement rate by age and type of plan 80 - 70 o 6 0 defined benefit o 50 - 40- 30- E a) o 20- CL 20 defined contribution 50 55 60 65 70 age 5.1 Effective tax rates A second measure of work incentives is the effective tax rate. This was used in the international study, led by Gruber and Wise (1997, 1999), of the impact of social security programmes on retirement behaviour. In the simplest cases, this measure is the same as the replacement rate.'9 However, taking account of the effect of continued working on pension wealth, this becomes: 1 - (gross earnings + change in pension wealth - pension) /gross earnings These effective tax rates are the same as average effective tax rates in a static model. But when considering a dynamic labour-supply problem, such as retirement, they are best thought of as the marina/effective tax rate on an additional vear's work. l') Ignoring taxation and changes in pension wealth, for example, the effective tax rate is I - (earnings-pension)/eamings, which can be simplified to pension/earnings. See OECD (1997a), .\nnex 13 for a discussion. 13 Figure 10 shows the baseline effective tax rate results. The differences in incentives between the two different types of plan are very clear and are more pronounced than in the replacement rate comparisons in Figures 5 and 9. Defined-benefit schemes give a substantial incentive to retire earlier, and the effective tax rate on continued work from the pension system is only higher in the defined-contribution plan after age 68. This pattern explains whv governments need to impose minimum earlv retirement ages in defined-benefit plans, because workers have a sizeable incentive to retire at the earliest possible date. Defined-contribution schemes, in contrast, give a large incentive to continue working until quite advanced ages. Figure 10. Effective tax rate on working by age and type of plan 80 - o) 70- s) 60 - defined benefit 60. 50- 40- x X 30- defined contribution a,, 10 50 55 60 65 70 age 6. Extensions to the basic model: taxation Gruber and Wise stress the importance of other elements of the tax and benefit system on retirement incentives. For instance, progressive personal income tax systems imply a higher average tax rate on (higher) earnings than on (lower) pensions. Thus, net earnings are lower relative to net pensions than to gross pensions. The tax system can have a more complex effect. In some countries, pensioners are treated more generously than workers are. In the United Kingdom, for example, the tax-free allowance for single pensioners is L5,220-L5,440 (depending on age) or 29-34% greater than for workers.2 In addition, the tax credit for married pensioners is worth 76 per cent more than for married couples of working age. These allowances apply to all the income of those over state pensionable age, including earnings. In other countries, pensions are treated more generously by the tax system than labour income. For example, in Austria, 75 per cent of pension annuity income is tax-free, and in the United States, 15 per cent. Canada provides a pension income credit at the basic income tax rate on the first slice of annuity income. Account should also be taken of the impact of social security contributions for benefits other than pensions, such as disability, unemployment insurance et:. Again, these are usually 21) See Dilnot et a!. (1994) and Hemming and Kai (1981) for a discussion of the impact of this tax treatment and a proposal for reform. 21 See \Miitehouse (1999). 14 levied only on earnings and not on pension payments, although some countries, such as France and the Netherlands, have recently moved to broaden the base for social-security contributions. 22 We explore, first, the effect of a simple personal income tax system, with a zero-rate band of $5 000 and a single rate of 25 per cent thereafter. The pension is assumed to give a gross replacement rate of 50 per cent. The lowest line in Figure 11 shows the net replacement rate at different levels of earnings. Income tax is levied on gross earnings above $5 000, but the pension at this earnings level (of $2 500) would be tax-free. The net replacement rate peaks at gross earnings of $10 000. The pension of 50 per cent of earnings is taxed at higher income levels. At very high levels of earnings, the net replacement rate asvmptotes to the gross. Figure 11. The impact of a progressive personal income tax on replacement rates 70 - 65 - ith pension contributions 60 - wth higher pensioner allowance 55- baseline effect 5 0 - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 10o6oo 20600 30000 gross earnings Moving upwards in Figure 11, the next line shows a system which gives a larger allowance ($7 500) to the pensioner than the worker. Now the net replacement rate peaks later, and at a higher level (60 per cent). Finally, the highest line in Figure 11 adds in a 10 per cent social security contribution. This shifts the line upward at each point. Now, the net replacement rate can exceed 65 per cent, compared with the 50 per cent gross rate. 7. Extensions to the basic model: defined-contribution schemes Many countries allow drawdown from defined-contribution scheme rather than forcing conversion to an annuity.23 Drawdown potentially allows people to avoid 'timing risk', the risk that interest rates are loxv on the date that the pension fund is converted. It also permits those 22 These contributions deliver entitlement to these other benefits, but all social-security programmes involve a substantial degree of redistribution. In the absence of actuariallv calculated individual contribution rates, it is difficult not to treat these contributions as a tax. See OECD (1998a) for a discussion. 2; See Piggott and Dodle (1999) and Brugiavini (1993). 15 with a short life expectancy to avoid the losses from annuitisation, and either to enjoy a higher pension or leave a bequest.4 Under drawdown, the fund continues to earn the market investment return. In the model, we assume that this is higher than the riskless interest rate, which underlies the calculation of the ann'ity rate. The only way of continuing to earn the market return on accumulated pension in the basic model is to continue in work. However, if drawdown is allowed, the individual can retire and still earn this higher return. Figure 12 shows the impact of drawdown. Draw down allows a pension around 50 per cent higher than mandatonr annuitisation (assuming the individual follows a rational rule for taking out the funds). Figure 12. The impact of voluntary annuitisation on incentives in defined-contribution plans 100 - 90 - draw down 80 - mandatory annuitisation 70- 60- defined benefit 50 40- 30- 20 - 10 50 55 60 65 70 age 7.1 Retirement rules in defined-contribution schemes in practice Table 0. Retirement rules in mandatory defined-contribution systems Normal age Minimum Minimum replacement rate pension Argentina 58/63 - Bolivia 65 70 Chile 60/65 50 110 Colombia 55/60 - 110 El Salvador 55/60 70 160 Peru 65 50 110 24 Givren the correlation between income and longevity, compulsorn annuitisation is regressive. See Kothkoff (198x) for a discussion. 16 8. Extensions to the basic model: defined-benefit schemes The simple, stylised defined-benefit model ignores many of the complexities of 'real- world' plans. Many of these are likely to have substantial effects on retirement incentives. We have assumed that the plan's formula uses lifetime average earnings, and that each extra contribution delivers additional pension benefit. Many plans, however, have non-linear accrual structures, with floors and/or ceilings to contributions and/or benefits. This means schemes deliver different returns for people with different levels of earnings. Secondly, many plans have higher accrual factors for early years of contributions. As later years deliver a lower return to working in the form of extra pension entitlement, this can be a disincentive to continue working. Other schemes have maximurn pension levels or maximurn years of accrual in the plan. The additional pension accrual is zero above these limits, but often, contributions continue to be levied. Thirdly, only a limited number of years of earnings count in mots schemes, which use either 'final' or 'best' years in the plan. The effects of these rules are extremely complex. Assuming that earnings continue to increase (as in the basic model, see Figure 1 or 4) then plans based on either final or best years increase the return to working relative to an average-pay scheme. This is because the earnings base used in the pension formula is growing more quickly. If, however, earnings decline at older ages, then the pension value falls with each year of extra contributions in a final earnings plan. A best-years scheme has marginally lower returns to working than an average-salary scheme. Finally, the simple defined-benefit formula does not allow for adjustments to the pension depending on the age at which it is drawn. In some countries, pensions drawn at the earliest possible age are 'actuarially' reduced. In others, deferring drawing the pension beyond the normal age attracts an increment to the pension value when, eventually, it is drawn. 8.1 Non-linearities in pension accrual Figure 13 shows the structure of pension accrual in two public defined-benefit plans. A full 46 countries' profiles are in Annex 1. The horizontal axis shows the number of years of contributions, the vertical, the percentage of the relevant measure of earnings secured in pension for that year of contributions.25 We have chosen only the countries with non-linear accrual structures, and in the Annex ranked them from the most non-linear to the least.26 Iran offers a very high accrual rate of 3.3 per cent, but there is a maximum replacement rate of 100 per cent. This means that after 30 years of contribution, there can be no further increment to the pension. The 'spike' at 10 years indicates that the pension 'vests' at that point. After nine years of contributions the pension 25 These measures of earnings also vary significantly between countnres: see Section 7.2 below. Note that we have capped the annual accrual at 4 per cent to make the charts easier to compare. Some countries, e.g. Brazil, have higher accrual rates over some ranges. 26 The ranking is based on the coefficient of variation of the annual accrual from zero to 45 years of contribution. 17 entitlement is still zero. At ten years of contributions, the pension is 33 per cent of earnings. The spike indicates this change.27 Colombia's public system, shown in the second panel of Figure 13, is more complex than Iran's. A replacement rate of 65 per cent is given for 1 000 weeks of contributions. Each 50 weeks between 1 000 and 1 200 weeks earns an increment of 2 per cent up to a maximum of 73 per cent of earnings. Between 1 200 and 1 400 weeks, the increment is 3 per cent for each 50 weeks, to a maximum of 85 per cent. A minimum of 1 000 weeks contributions is required for the pension. Even these charts are a major simplification: they ignore 'sectoral' privileges (giving early retirement or reduced contribution requirements to particular occupations or industries) and credits granted, for example, for periods of unemployment, disability, education or child-rearing. Figure 13. Accrual rates by years of contributions Iran Colombia 4 4 3 3 2 2 j 1 1 0 5 1 1'5 20 25 30 35 4'0 45 0 10 15 20 25 30 35 4'0 45 8. 1.1 Effect on uork inceti T'he pension system acts as a disincentive to work at older ages, even below the normal pension age, in all the -countries shown in the Annex. For examnple, in 12 countries the accrual rate is zero after 40 years' or fewer of contributions (at an average of 34 years). In 40 countries, later years of contribution attract a lower rate than early years. 27 A 'legal-contract' approach (see Bulow, 1982 for an application to private defined-benefit schemes in the United States) would show a zero accrual for the first ni'ne years, and a 33 per cent accrual in the tenth year. This is also called the 'accumulated benefit obligation' (Bodie, 1991). The altemative, used here, is to show the accrual in the first ten years assumi'ng that the individual will make ten years of contributions (See, for example, Kotlikoff and Wise, 1985 on the Uniited States). A final approach is to compute the probability at any point before ten years that the individual will contribute for ten years, also called the 'projected benefit obligation'. This method is discussed in Disney and Wh~itehouse (1996) and compared with the other two measures for private defined-benefit schemes in the United Kingdom. See also Disney (1996), pp. 116-12 1. 1 8 8.2 Final, average and best salary schemes Tables 1 and 2 show the formulae used in 80 different countries' public, defined-benefit plans. The Tables rank countries with final and best salary formulae inversely: from the longest to the shortest averaging periods. OECD countries, shown in Table 1, divide evenly between the three different groups. First, those that average earnings across all or most of the working life. Secondly, those that take a measure of earnings in the final few years of the working life and, thirdly, those that use a number of' 'best' years. Table 1. Earnings measure in public, defined-benefit plans: OECD countries Average Best Final Belgium Norway (20) Czech Republic (average since 1985) Germany Austria (15) Portugal (best 10 of 15) United Kingdom Sweden (15) Turkey (5-7) United States (ex worst 5 years) France (11) Greece (5) Canada (ex 15% worse years) Mexico (5) Hungary (best 4 of 5) Source: Department of Health and Human Services (1997) In other countries, listed in Table 2, final salary schemes dominate. Only 14 per cent of countries use average pay and 18 per cent use a measure of best earnings. Schemes outside the OECD also tend to consider rather fewer years' earnings. In final pay plans, the average in OECD countries is around 7 years, compared with fewer than 4 years in lower-income countries. There is also rather less variation among OECD countries. Outside the OECD, three countnres use only the final month's pay in the pension formula, while six countries average over the last ten years. In best earnings schemes, the OECD average is a 15-year formula, compared with 6 years outside the OECD. There are two main reasons for adopting short averaging periods in defined-benefit plans. First, they are a simple way to correct the effects of high and volatile inflation. Secondly, they are administratively simpler than tracking work and contribution records right across the working life. However, they can lead to high costs, strategic manipulation of earnings profiles and disproportionately higher benefits going to higher-income workers, because they tend to have more steeply rising age-earnings profiles (see below and World Bank, 1994, Box 4.8). 19 Table 2. Earnings measure used in public, defined-benefit plans: non-OECD countries Average Final Final cont. Albania Argentina (10) Niger (best of 3 or 5) C6te d'lvoire Colombia (10) Rwanda (best of 3 or 5) Congo (Kinshasa) El Salvador (10) Tunisia (best of 3 or 5) Cyprus Haiti (10) Dominica (best 3 of final 10) Jamaica Madagascar (10) Malta (best cons. of 10) Liberia Uruguay (10) Dominican Republic (2) Mauritius Guatemala (5) Egypt (2) Philippines Mali (5) Iran (2) Trinidad and Tobago Romania (best cons. 5 of 10) Jordan (2) Cuba (best 5 of 10) Oman (2) Best Slovakia (best 5 of 10) Saudi Arabia (2) Poland (6 of final 15) Nicaragua (3,4,5) Costa Rica (best 2 of 5) Croatia (cons. 10) Peru (3,4,5) Syria (best of 2 or best cons. 5 in 10) Serbia (cons. 10) Brazil (3) Georgia (best of 1 or 5) Slovenia (cons. 10) Ethiopia (3) Moldova (1) Panama (7) Iraq (3) Pakistan (1) Belarus (cons. 5 of final 15) Libya (3) Kuwait (final month) Kyrgyzstan (cons. 5 of 15) Paraguay (3) . Lebanon (final month) Turkmenistan (cons. 5 of 15) Cameroon (best of 3 or 5) Nigeria (final month) Russia (cons. 5 or final 2) Central African R. (best of 3 or 5) Ukraine (cons. 5 or final 2) Congo-Brazzaville (best of 3 or 5) Ecuador (5) Gabon (best of 3 or 5) Bulgaria (cons. 3 of final 15) Mauritania (best of 3 or 5) -Algeria (3) Morocco (best of 3 or 5) Note: cons. = consecutive Source: Department of Health and Human Services (1997) 8.2.1 The anpaa offial saIry schenw on tork incr Our baseline model assumes that earnings grow linearly with age. Final pay, under this assumption, is higher than average pay, and is growing more quickly. Figure 14 shows the pattern of work incentives by age for a final- and an average-salary plan using the measure of replacement rates, adjusted for changes in pension wealth. To equalise the pension value at age 65, the accrual rate is around 1 per cent of final-pay in that scheme, compared with 1.7 per cent of average earnings. The differences in the incentive structure are not large. But the fact that final pay increases more rapidly with age than average earnings reduces adjusted replacement rates when younger and increases them when older.28 This result depends critically on the structure of age-earnings profiles, and this issue is considered next. 28 Lazear (1979) argues that this explains why employers impose mandatory retirement in defined- benefit schemes. A mix of higher, seniority pay and backloaded pension benefits mean that the employer's costs of continuing to employ older workers exceed productivity, meaning it is in their interest for them to leave. See also Burkhauser and Quinn (1983) and Hutchens (1986). 20 Figure 14. Defined-benefit pensions under average and final salary formulae with rising earnings 80 - 60 - average earnings 40- 20- defined contribution 50 55 60 65 70 age 8.2.2 WhVal do 'tne 'age-earningssprojiles look like? Simple cross-section analysis of age-earnings profiles generally shows an inverted-U shape, xvith real earnings falling at older ages. Figure 15 shows this pattern using data on hourly earnings for the United Kingdom.29 Separate wage equations, including a quadratic term in age, were estimated for each occupational group. Figure 15. Age-earnings profiles by occupation, United Kingdom 3 - professional 2- managerial skilled manual I semi-/unskilled 0 _______________ _T_------___ 20 30 40 50 60 65 age Source: Disney and Whitehouse (1991), Tables 2, 5 and 7 The pay of professional, and to a lesser extent, managenal workers rises steeply with age initially. Professional earnings flatten when workers reach their mid-50s, with a rather earlier peak for managers. In contrast, the profiles for manual workers are much flatter and peak earlier, in the early to mid-40s. TI'he decline in earnings after their peak is also relatively larger, so that 29 Disnev and \Xhitehouse (1991). The data are drawvn from a pool of Family Expenditure Sunrey cross-sections for 1978 to 1986. 21 workers from their late 50s onwards earn the same or less than workers in their 20s. This shows that both schemes based on both final ad best earnings are regressive. Professional workers final pay is much higher relative to their average pay than the same ration of manual workers. Figure 16 shows a simple measure of the slope of age-earnings profiles for a range of OECD countries: the ratio of earnings of 55-64 year olds to that of 45-54 year olds. In Germany, Italy and Switzerland the difference is very small, while in France, the oldest workers earn more than their irnmediate juniors do. At the other end of the spectrum, wages for the oldest workers in Australia, Canada, Mexico and the United Kingdom are 17 per cent below people aged 45-54. These economies are conventionally classified as those with more flexible labour markets. The decline in Japan, in contrast, reflects the lifetime employment system and the occupational downgrading of older workers within the economic group to which their employer belongs. Interestingly, the oldest workers in the United States (the archetypal flexible labour market) earn only 10 per cent less than their. juniors do. Figure 16. Relative earnings of older workers (ratio of 55-64 year olds' pay to 45-54 year olds) France Switzerland Italy Germany New Zealand Spain Denmark Sweden Finland Czech Republic United States Portugal Japan Canada Australia Mexico Unted Kingdom 0.6 07 08 09 1 1 1 Source: OECD (1 998c) However, these cross-section analyses conflate age and cohort effects. Lower earnings of older workers in cross-section also reflect cohort differences in education, training, labour-market experience etc. that cannot be disentangled from the pure age effect. Following the same cohort over time, other studies have found that age-earnings profiles are close to linear, with pay rising even at older ages.30 However, earnings are endogenous to the retirement decision, so there is a sample selection problem as the people working at older ages are not representative of the whole cohort. Even cohort-based studies are therefore a biased measure of age-earnings profiles. Attempting to control for this selection raises a problem of simultaneity. We can only know the true nature of this selection process if we know why people retire early. And we can only know why people retire early if we know what true, individual age- earnings profiles look like. 30 See, for example, Meghir and Whitehouse (1996) and Gosing, Machin and Meghir (1998). This issue is also explored in Freeman (1979, 1989), Berger (1983), Disney (1996), Chapter 6 and Kotlikoff and Gokhale (1992) 22 Another important field of literature looks at the effect of job displacement on earnings. This is particularly relevant for the incentives of workers who leave their jobs involuntarily and so face a choice of either retirement or starting a new job. In the United States, for example, earnings of older workers who lose their job and start a new one are 39 per cent lower (for a period of two years) than people who remain in their jobs.3 8.2.3 Age-eamrvigs p/es xumd k irncnte Age-earnings profiles that are flatter or declining at older ages have a number of effects on the measures of incentives compared with our baseline results. First, the denominator of the measures of incentives is lower because of lower pay, increasing replacement rates and effective tax rates for both defined-contribution and defined-benefit schemes. Secondly, the decline in final salary at older ages means the relative effects of final- and average-earnings defined-benefit schemes is reversed from the pattern in Figure 14. Final-salary schemes will give higher replacement rates at younger ages. Overall, the effect of final pay formulae relative to average- pay varies with the relationship between earnings and age. Given the discussion of the previous paragraph, this will be particularly relevant for people who leave a job involuntarily. Final-salary systems will give a powerful disincentive to take another job with lower pay. This, along with the problems raised at the beginning of section 8.2, suggests that defined-benefit schemes should preferably be based on average-salary. 8.3 'Actuarial' adjustments to defined-benefitpensions Some countnres adjust defined-benefit pensions to reflect the fact that they are drawn either early or late. Table 3 shows the size of these adjustments in the 32 countries that have them. It lists countries in inverse order of the size of the adjustment. Actuarial adjustments are more common in OECD countries than outside: 14 out of the total of 29 OECD members use them. There is little difference between adjustments to early or late pensions. But there is a big difference between OECD and non-OECD countries, with averages of 6½h per cent in the former and a little under 3½h per cent in the latter for each year the pension is drawn early or late. These adjustments apply over very different age ranges in different countries. In 11, actuarial changes cover only early pension claims (before the 'normal' pension age). In many countries, there is evidence that the majority draw their pension at this 'early' age, despite the reduction in their pension benefit. In 12 countries, increments are only given for deferring the pension, while in six, increments are decrements are applied both to early and late pensions. Notional-accounts pension systems also adjust the pension according to the age at which it is drawn. The implied adjustment will be 8-9 per cent a year in Poland and around 7 per cent m Latvia." 31 Couch (1998). See also Jacobson, LaLonde and Sullivan (1993) and Ruhm (1990). 32 Chlon, Gora and Rutkowski (1999) on Poland and Fox and Palmer (1999) on Latvia. See Disney and Whitehouse (1999b) for a discussion of retirement incentives in notional-accounts systems. 23 Table 3. 'Actuarial' adjustments in defined-benefit plans Age adjustments apply Size of adjustment Japan 60-70 early & late 12 France?? 60-70 late 10 Finland 60-70 early & late 6-12 Liechtenstein 65-70 late 8.4 Spain 60-65 early 8 United Kingdom 65-70 late 7.5 Sweden 60-70 early & late 6-8.4 United States 62-70 early & late 6-6.7 Canada 60-70 early 6 Dominica 60- late 6 Germany 63-67 early & late 6 Greece 60-65 early 6 Pakistan 55-60 early 6 C6te d'lvoire 50-55 early 5 Guinea 55- late 5 Israel 65-70 late 5 Mali 50-55 early 5 Mexico 60-65 early 5 Senegal 53-55 early 5 Albania 60- late 4 Czech Republic 60-70 late 4 Cuba 60-65 late 1.5-4 Italy 57-65 early & late 3.5 El Salvador 60- late 3 Honduras 65- late 3 Hungary 60-70 late 3 Panama 62- late 2 Sudan 45-60 early 1.2-1.9 Costa Rica 62-65 late 1.5 Croatia -60 early 1.33 Algeria -60 early 1 Nicaragua 60-65 late 1 Turkey 55- late 1 Note: Australia is considering introducing a deferred retirement bonus plan for men working from age 65 to 70 (and 61-66 for women). Ages where adjustments apply are for men where these differ from women. Costa Rica gives 1.5 per cent for first year of deferral, 2 per cent for second year and 2.5 per cent for the third. In Croatia, the reductions only apply for pensions up to normal retirement age (60). New system for Italy applies to contributions since December 1995. Early pensions in Mexico only available if the individual is involuntarily unemployed. Adjustment is based on age coefficient of 4.72 at age 57 and 6.136 at age 65. Sweden is planning to remove the upper age limit for deferral in 1999. The 6 per cent rate applies before 65, 8.4 per cent after 65. The United Kingdom will increase the deferral rate to 10 per cent after 2010 and remove the age 70 ceiling for deferral. The United States will increase its rate to 8 per cent by 2001: see section 8.4.3 below. The Hungarian reform introduced a penalty for retirement below age 62 equivalent to 3.6% per annum in the long run. The Polish reform introduces an increment for delayed retirement equivalent based on the notional interest rate and higher annuity level through the new system of notional accounts for contributors aged 50 or under in 1999. Source: Department of Health and Human Services (1997), Kalisch and Aman (1998); Palacios and Rocha (1999); Chlon, Gora and Rutkowski (1999). 24 8.3. 1 A cuarzid nmral adusnerts How do these adjustments compare with the actuarially neutral increment or decrement? We define the actuarially neutral increment or decrement as the change that would keep the net present value of the pension constant. This has three elements. First, the change in the annuity rate between the two relevant ages captures the effect of the one-year delay in claiming the pension. Secondly, account must also be taken of the risk of dying during the year. Thirdly, the net value of the delayed pension must be discounted back to the present. Figure 17 shows the results for Thai mortality data. The rate increases with age because of the increase in mortality and because, at older ages, one year's delay is a larger proportion of the total expected duration of pension payment. For men, the rate increases from 7 per cent at age 50 to 12 per cent at age 70. For women, the rate will be hig} -r because of their longer life expectancy. Comparing Figure 17 with the rates in Table 3 shows that the average adjustment is below the actuarially neutral level in most countries, with the (probable33) exceptions of Japan, Finland Liechtenstein Spain, the United Kingdom and Sweden.34 Figure 17. Neutral actuarial adjustment to defined-benefit pension by age 12 10 8 -: 5 5 4 50 55 60 65 70 age 8.3.2 The impaaonuork xues Figure 18 examines the effect of actuarial adjustments on work incentives. We have assumed a 5 per cent change in the pension for each year the pension is drawn before or after age 65, which is approximately the average adjustment in Table 3. The adjusted replacement rate for the defined-benefit scheme now has a more pronounced upward slope, but not as strong as the defined-contribution plan. We do not have life tables for all these countries to be able accurately to assess this claim. 34 Studies have also found that private plans tend to have less than fair actuarial decerements for early retirement. See, for example, Kotlikoff and Smnith (1983) and the discussions in Quinn, Burkhauser and Myers (1990) and Kotlikoff and Wise (1985). 25 Figure 18. Impact of actuarial adjustments in defined-benefit schemes on replacement rates 80 - 60- 40 -actuarially adjusted defined benef 20 = defined contribution 50 55 60 65 70 age 8.4 Earning while drawingpension Our basic model of retirement incentives assumed that people had to give up work when they claimed their pension. In many countries, working while drawing pension is possible. This complicates the analysis of pension plans and retirement incentives enormously, as most models can only be identified by equating the point of retirement (i.e. withdrawing from the labour force) 15 with the point of first drawing the pension. Table 4 shows the position in OECD countries. Again, policies differ. In Ireland, Portugal and Spain, the pension is paid conditional on withdrawal from paid work. People in France must retire definitively from their usual job, although it is possible to take another job. In the majority of countries, however, a limited amount of work is compatible with drawing the pension. Canada has the most liberal of these regimes. People can earn up to 160 per cent of average earnings with full pension, and then the pension is withdrawn at 15 cents for each dollar of earnings above that point. In Greece, pensioners may earn up to 116 per cent of average earnings, but then the whole pension is withdrawn. Italy, Japan and the United States also have earnings limits. In Japan, for example, 20 per cent of the pension is withdrawn when earnings are between 17 and 90 per cent of average earnings. Above 90 per cent of average earnings, the whole pension is withdrawn. However, in both these countries, deferral is possible, allowing people to get round the effect of the earnings test. (Section 8.4.3 below discusses the United States' system in more detail.) At the foot of the Table are eight countries that irnpose little or no restrictions on combining work with pension payments. 55 Sec in/eralia Gustman and Steinmeier (1984). 26 Table 4. Earning while drawing pension, OECD countries Disregard Withdrawal rate (%) (% of average earnings) Deferral not possible Canada 160 15 Greece 116 full Iceland 59 25 Denmark 50 60 Austria 30 full Belgium 33 100 Norway 18 50 Australia 8 50 France none full Ireland none full Portugal none full Spain none full Deferral possible Italy 23 100 Japan 17-90 20 90 full United States 38 33-50 No restrictions Finland Germany Netherlands New Zealand Poland Sweden Switzerland United Kingdom Note: Pension in Ireland, Portugal and Spain conditional on withdrawal from work. Pension withdrawn at a 100 per cent rate between 29 and 33 per cent of average earnings in Belgium. Italy gives a higher disregard for self-employment incomes (which are an important income source). Australia has a means-tested system and all income sources (including private pensions) are withdrawn against the public pension. Portugal has recently allowed people to work and claim full pension (Kalisch and Aman, 1999, Table 16). The United States has a lower disregard (18 per cent of average earnings) for people aged 62-64; Poland refers to reformed system. Source: Blondal and Scarpetta (1998), based on Department of Health and Human Services (1997), European Commission (1996) and OECD (1996b); Kalisch and Aman (1998); Chlon, Gora and Rutkowski (1999). 8.4.1 Patial reaertproga=s A small number of countries have partial or phased retirement schemes.36 These allow people to cut their hours of work while claiming part of their pension. In Denmark, people who reduce their working hours between age 60 and the normal pensionable age of 67 can receive part 36 See Casey (1998), Laczko (1988) and Naegele (1996) for a discussion of gradual retirement schemes. Gustman and Steinmeier (1984) and Ruhm (1990) discuss 'bridge' jobs as a route to partial retirement in the United States. 27 of their pension.37 People must have been employed full-time for 10 of the previous 20 years and cut their hours to between 12 and 30 per week. Germany reimburses employers the cost of paying a 20-per-cent earnings supplement and maintaining 90 per cent of previous pension contributions to people who reduce their hours by half or more. Luxembourg allows people to combine half of their earnings with a one-half pension. Japan pays lower pensions to people aged between 60 and 64 if they reduce work attachment. If the combined pension and lower wage is below $24 000 the pension is cut by 20 per cent. Between $24 000 and $36 000, the pension is reduced at a marginal rate of 50 per cent, with 100 per cent withdrawal against earnings over $36 000. Canada is also considering a partial retirement scheme. Partial retirement programmes are attractive in theory. They might encourage people to remain in work longer than they would given a binary choice between full-time work or retirement (although they might reduce the labour supply of some workers who otherwise would have remained in full-time work). In particular, they might allow any age-related decline in stamina or capabilties to be accommodated. In practice, however, the take-up of partial retirement schemes has been very limited. In Gernany, for exarnple, only 2,000 of an elgible population of 1.2 million availed themselves of this option. In some countries, such as France and Germany, early retirement (on only marginally lower incomes) proved more attractive than partial retirement. In Germany and the United Kingdom, employer-run defined-benefit programmes dependent on final pay were not co- ordinated with the public programme, so gradual retirement would resuk in a much lower private pension. Many employers were also unable or reluctant to re-organise work around partial retirees. 8.4.2 Canhnviguork x ndpiw2.: refn in the UnaJKinglorm The United Kingdom used to have an earnings limit of around a third of average earnings. Above this level, the pension was withdrawn, at 50 per cent over a short range and then at 100 per cent. However, most people who were working and earning above the limit deferred their pension, earning increments in their eventual pension of 71/2 per cent per year of deferral gIable 3). Only 2,500 people had their pension capped by the earnings test while 200,000 deferred. A further 200,000 were claiming the pension and working, but did not have their pension reduced because they were below the limit. The rule could have affected them if they would have chosen to work more in the absence of the earnings test. The United Kingdom abolished the earnings rule in 1989. The effects of this policy are complex.38 For those who are currently working and either deferring their pension or have deferred, the change is a pure income effect, suggesting reduced labour-supply incentives. For people not working or earning below the earnings lirnit, there will be a positive substitution effect. Both are likely to be small. The fiscal effect of abolition will also be tiny. Since the actuarial adjustments for deferrals are fairly close to neutral on average, there will be an up-front cost as people claim their pension earlier, with a saving on the adjustments in the future. Current regulations require members of employer-provided schemes to retire fully from their job before they can draw their pension. People could draw their occupational pension and Ministry of Social Affairs (1995). See VAhitehouse (1990) for a detailed analysis. 28 work, but would have to take a different job. The government last year9 proposed a more flexible regime, including the possibility of partial retirement. People will be able to draw their occupational pension at any age between 50 and 75 irrespective of whether they actually retire. People will be able to draw the benefits from additional voluntary contributions into employer schemes at any age between 50 and 75, irrespective of when the main occupational pension is drawn. These two reforms are designed to allow a more flexible transition from full-time work to retirement - as in the partial retirement programmes described in the previous section - rather than the current 'all-or-nothing' choice. 8.4.3 Cxmbiguwrk dpss: refms md2ie UnitedStates The United States has also altered the rules for earning while drawing social security.40 The original test, introduced in the 1930s, took away the whole pension once earnings exceeded a fifth of average earnings. There has been a gradual liberalisation in the rules. First, the exempt amount was increased periodically during the 1960s and then indexed to inflation from 1972. Each year from 1978 to 1982, there were above-inflation increases. The limit for 65-69 year olds will increase from $12,500 to $30,000 between 1996 and 2002. Secondly, there have been reductions in the withdrawal rate. In 1960, a schedule of 50 and 100 per cent withdrawal rates replaced the immediate loss of all benefits one the limit was reached. The 100 per cent band was abolished in 1972. The withdrawal rate for 65-69 year olds was cut to 33 per cent in 1990. Finally, the earnings test was abolished completely for people aged 72 and over in 1954 and for 70 and 71 year olds in 1983. The eventual pension is higher to reflect benefits withdrawn under the earnings test, as in the deferral system in the United Kingdom. This increase, called the delayed retirement credit, was introduced in 1973 at a rate of 1 per cent for each year of deferral between 64 and 69. In 1982, it was increased to 3 per cent. Since 1990, it has increased at half a percentage point a year, and will continue until it reaches 8 per cent. When it reaches this final target level, it will be approximately actuarially neutral. In 1989, 27 per cent of Americans aged 65-69 were working. Of these, 38 per cent had some benefits withdrawn under the earnings test and 29 per cent were working and not claiming pension benefits.4' Further evidence of the impact of the earnings test is that 9 per cent had earnings very close to the earnings-test limit. Most empirical studies of the earnings test's effect on labour supply have found it unimportant.42 More recent analysis, however, is able to use the 'natural experiment' of changes to the earnings test that affect different age groups in different ways to obtain more accurate estimates of its effect. Friedberg (1998a) finds that eliminating the test would increase aggregate hours worked of those currently affected by the rule by 5.3 per cent (taking account of the decline in hours of those with the highest earnings from the income effect). Her study, however, treats the earnings test as a pure tax, despite the eventual increase in the pension benefit when it 39 Department of Social Security (1998). 40 See Friedberg (1998). 41 Bondar (1993). 42 See inter aliza, Bondar (1993), Honig and Reimers (1989), Leonesio (1990, 1993), Packard (1990). Robbins and Robbins (1989) found a substantial effect, but Leonesio (1990) dismnisses this work as having 'serious theoretical and methodological shortcomings'. 29 is finally claimed.43 She gives four reasons for this treatment. First, people are aware of the earnings test - 73 per cent of pensioners under age 72 told the New Benficiary Surrey of 1982 that they knew of it44 - but are not aware of the link between the test and subsequent pension credits. She cites newspaper and financial magazine descriptions of the earnings test that ignore the credits.45 Secondly, people are myopic or, thirdly, face borrowing constraints and so under- value the increase in future income. Finally, the current credits are below the actuarially neutral level, especially for men and those with short life expectancy. However, these last three arguments suggest that only part of the earnings test reduction is a tax. 8.4.4 am( ninguzrk adp si.ispo icyaindsio Our discussion of the United Kingdom and the United States shows that earnings tests have extremely complex interactions with rules for actuarial adjustments. There is a sizeable dead-weight cost to abolishing earnings tests, as people draw the pension at the earliest possible age and continue to work as before. Poland, for example, has introduced an earnings test in its recent reform. The aim is to prevent people, particularly in occupations with special early retirement provisions, from drawing pension and continuing to work.46 The government expects this to cut spending by 0.3 per cent of GDP in 2003, around 15 per cent of the total reduction in spending anticipated from the recent reform. The Slovak Republic has drafted similar legislation to prevent people combining work and pensions. Italy is also moving to limit these opportunities. 9. Conclusions, policy implications and future developments We have presented a simple model of how different types of pension plan affect retirement incentives. The results show a powerful incentive to leave work at the earliest possible age in defined-benefit plans. Defined-contribution plans, in contrast, encourage people to remain in work longer. Examining 'real-world' defined-benefit schemes, we found many have features that provide more profound disincentives to work, such as * early retirement programmes that enable people to leave work with no reduction in pension or a decrement that is less than actuarially fair a non-linear accrual structures or maximum pensions that give a low or zero increment to pension for working at older ages 43 There is some empirical evidence to support this claim. Reimers and Honig (1993, 1996) compare the response of labour-market re-entry behaviour of men before and after age 65 to changes in the earnings limit. These are expected to differ, because of the variance in the actuarial adjustments. However, they found very similar effects for both age groups. Friedberg could not find an effect from the credit when she included this in the modelling. 44 Leonesio (1990). 45 Simon (1996) and Kristhof (1997). 46 See Gora and Rutkowski (1998). 30 * final-salary formulae that encourage people to leave the labour force once earnings reach their peak * pension systems that still levy employer and/or employee contributions even when no additional pension is earned * earnings tests that prevent people from combining work and pensions and do not actuarially fair increments for defernrng a pension claim 9.1 Policy implications Addressing these problems should be a central part of any pension reform, and these reforms should be informed, at the minimum, by the types of analysis we have carried out here. The problems in existing public pensions we outlined previously suggest a number of useful reforms. First early retirement schemes should be curtailed. This might involve moving towards actuarial reductions in early-retirement pensions, increases in the age at which they can be claimed or tighter conditions for entitlement, such as duration of employment or participation in training programmes. We recognise that many of these schemes were a response to genuine labour-market problems faced by older workers. However, problems such as age discrimination" and the lack of appropriate skills are best addressed directly. Secondly, many of the distortions to labour-supply incentives caused by defined-benefit pensions could be mitigated by moving to pensions based on average salary across the working life rather than a limited number of best or final years. Maximum pensions and limits to the number of years of contributions that earn pensions should be removed to give people an incentive to work beyond these limits. Contributions should not be levied at ages or in years when people do not earn a pension entitlement. Thirdly, while early-retirement schemes do need to enforce earnings tests to avoid abuse, it should be possible to allow people to combine pensions at the standard age with some work. Alternatively, people should be able defer drawing their pension at actuarially fair rates. We briefly considered partial retirement schemes. These might encourage people who would otherwise leave the workforce retire gradually, by moving to part-time work. However, these programnes have rarely been taken up by a significant proportion of the eligible population. There are reasonable explanations for this failure, which suggests that partial retirement is unlikely to be much of a panacea. 9.2 Recent policy initiatives in OECD countnies We noted in the introduction that more than half of OECD countries mentioned low effective retirement ages and poor work incentives in the pension system as a matter of policy concern. Only problems with the financial viability, mentioned by three-quarters of governrnents, were the subject of more widespread worry. Table 5 shows that more than 80 per cent of OECD countries have recently introduced reforms designed to promote the employment of older workers. Nine countries will increase the standard age for state pensions by an average of three years. In addition, Australia will raise the 47 See, inter alia, Hutchens (1986, 1988), Hutchens (1986, 1988), Johnson and Neumark (1996), Neumark and Stock (1997), OECD (1998c). 31 age at which people can draw private pensions. Eight other countries will increase the age at which women can draw pensions. In seven, this will equalise women's pensionable age with men's. The average increase is over four years. In practice, most people retire well before the standard pension age in the vast majority of countries. The effective retirement age is on average five years younger than the 'normal' retirement age. Twelve countries are therefore aiming to restrict eligibility for early retirement, which may be a more potent policy for raising effective retirement ages than changing the standard pension age. Belgium, France, Greece, Hungary, Italy and Portugal will increase the number of contribution years required to qualify for a pension. Finland, Germany, Hungary, Italy and Poland will increase the minimum age to qualify for pensions. Denmark requires local authorities to offer training and labour-market re-integration programmes before an early retirement pension can be granted. Finally, five countries will introduce or have proposed schemes to encourage people to work after normal pension age, as discussed above in sections 8.3 and 8.4. 9.3 Future developments The model we have developed is a useful tool for examnining the financial rewards to working that generates a number of practical policy conclusions.48 However, it is at the moment simplistic and, although we have examined a number of features of different countries' pension systems, we have not studied any of these programmes as a whole. Our future work on retirement therefore has two main aims. First, to explore how. the incentives generated by pension schemes affect labour-supply behaviour. Secondly, to assess the impact on incentives of a number of countnres' pension systems and how they might be reformed to reduce labour- market distortions and promote work. 48 We will provide the model, implemented in the Stata statistics and data programning language, on request. 32 Table 5. Recent policy initiatives to promote employment of older workers Increase in pension age Equalising pension age Discouraging early Encouraging work after retirement pension age Australia: from 55 to 60 in Australia: to 65 (by 2013) Austria: reduced access Australia: deferred pension age for private pensions (by bonus plan proposed (men 2025) 65-70 and women 61-66) Czech Republic: from 60 to Belgium: to 65 (by 2009) Belgium: contribution years Norway: smaller reduction 62 for men and 53-57 to 57- for retirement at 60 from 20 for working while drawing 61 for women (by 2007) to 35 (by 2005) pension for ages 67-70 Finland: from 63 to 65 Germany: from 60 to 65 (by Denmark: local authorities Sweden: allow actuarial 2004) to provide training etc. increases for deferral after before pension can be age 70 Hungary: to 62 for men (by Greece: from 60 to 65 (post- MS&hd: lower benefit and United Kingdom: increased 2001) and for women (by 1993 labour-market minimum age from 55 to 58 deferral increment; increm- 2009) entrants) ents for deferral after age 70 Italy: from 63 to 65 for men Poland: from 60 to 65 to France: contribution years United States: higher and 58 to 60 for women (by equalise with men from 371/% to 40 earnings limit; increased 2000); 57-65 in new scheme deferral increment Japan: from 60 to 65 for Portugal: to 65 (by 1999) Gernany: minimum age to men (by 20013); from 59 to 62 (from 2012) 65 for women (by 2018) Korea: from 60 to 65 (by Switzerland: from 62 to 64 Greece: minimum 2033) (by 2005), remains below contribution years from 13%2 men's age of 65 to 15 New Zealand: from 62 to 65 United Kingdom: from 60 Hungary: higher age and (by 2001) to 65 (by 2020) contribution years Spain: from 60 to 65 Ireland: special scheme for (people entering the labour civil servants ceased market after 1967) United States: from 65 to Italy: minimum age 52 (from 67 (by 2027) 1997); contribution years increased from 35 to 40 years (from 2008) Poland: increase in minimum age from 60 to 62 planned Portugal: increase in minimum contribution years from 10 to 15 Source: Kalisch and Aman (1998), OECD (1997b) 33 Annex 1. Accrual rates by years of contributions Iran Colombia 4 4 3 3 0 2 2 0 _ _ _ _ _ _ _ _ _ _ _0 _ _ _ _ _ _ _ _ _ _ 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 Spain Equatorial Guinea 4 4 3 3 2 2 0 _ _ _ _ _ _ _ _ _ _ _0 _ _ _ _ _ _ _ _ _ _ 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 Tunisia Guatemala 4 4 3 3 2 2 0- 0- L__ _ _ __ _ _ _ 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 Morocco Panama 4 4 3 3 2 2 1- 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 34 Dominica Brazil 4 4 3 3 2 2 1 1 0 ~ 0 5 1115 20 25 30 35 45 0 5 101 t5 20 25 3o 35 4o 45 Hungary Costa Rica 4 4 3 3 2 2 1 1 __ 0 _ _ _ _ _ _ _ _ _ _ _0 _ _ _ _ _ _ _ _ _ 0 5 1052025303~540O45 0 5 O 1015202530354!045 Nigeria Guyana 4 4 3 3 2 2 1 1 0 5 10 15 20 2O 5 30 35 40 45 6 5 10 15 20 25 3O 35 40 45 Syria Libya 4 4 3 3 2 2 1 1 0 _ _ _ _ _ _ _ _ _ _ _ _0 _ _ _ _ _ _ _ _ _ 0 5 ' 10152'025303 5 4!045 0 5 O 101522'0 30~35 40 45 35 Turkey Ethiopia 4 4 3 3 2 2 1- 1 L 0 5 10 1 5 20 25 30 35 40 45 0 5 10 1 5 20 25 30 35 40 45 Honduras Latvia 4 4 3 3 2 2 1 1 0- _ _ _ _ _ _ _ _ _ _ _ 04 -_ _ _ _ _ _ _ _ _ 0 5 10 1 5 20 25 30 35 40 45 0 5 10 1S 20 25 30 35 40 45 Haiti Israel 4 4 3 3 2 2 1 1 O _ _ _ _ _ _ _ _ _ _ 0 _ _ _ _ _ _ _ _ _ _ _ _ -0 5 1'0 1'5 i 2'5 30 35 40 45 0 5 10 15 20 25 30 35 40 45 Ghana Jordan 4 4 3 3 2 2 1 1 0 _ _ _ _ _ _ _ _ _ _ _ _0 _ _ _ _ _ _ _ _ _ 0 5 10 15 20 25 30 35 40 45 5 1015 20 25 30 35 40 45 36 Madagascar Slovenia, Croatia, Yugoslavia 4 4 3 3 2 - ------- 2 0 5 10 15 20 25 30 35 40 45 0 5 1015 20 25 30 35 40 45 Dominican Republic Nicaragua 4 4 3 3 2 2 1 I 0 _ _ _ _ _ _ _ _ _ _ _0 _ _ _ _ _ _ _ _ _ 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 Kazakhstan Liberia 4 4 3 3 2 2 1 1 0 O- 0 0 5 101t 5 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 El Salvador Paraguay 4 4 3 3 2 2 1 1 O -_ 0 - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 0 5 10 1'5 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 37 Former Soviet Union1 Ecuador 4 4 3 3 2 2 0 _ 0 l_ l _O- l 0 l _ l _ l _ l _ l _ l _ l _l 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 Cuba Czech and Slovak Republics 4 4 3 3 2 2 1 1 6 5 1O 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 Gabon Poland 4 4 3 3 2 2 1 1 0 _ _ _ _ _ _ _ _ _ _ _ _0 _ _ _ _ _ _ _ _ _ 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 5 30 35 40 45 Uruguay Rwanda 4 4 3 3 2 2 1 I 0 _ _ _ _ _ _ _ _ _ _0 _ _ _ _ _ _ _ _ _ _ 0 5 10 15 20 25 30 35 40 45 0 5 10 1t5 20 25 30 5 40 45 38 Cameroon, Chad, Venezuela Argentina 4 4 3 3 , 2 2 1 1 0 5 10 15 20 25 30 35 40 4 5 0 5 1 10 5 20 25 30 35 40 45 Central African Republic Zimbabwe 4- 4 3- 3 2 ~~~~~~~~~~~~~2 1 1 0 4, 0 , . . . . , , . , 0 5 10 15 2025 30 35 4045 0 5 10 15 20 25 30 35 40 45 Bulgaria Austria 4 4 3 3 2 2 _ _ _ 1 1 0- ,,,,,0 -__l_l_l_l_l_l _l _l _l 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 354 45 1. Belarus, Georgia, Kyrgyzstan, Moldova, Russia, Turkmenistan, Ukraine, Uzbekistan 39 10. Bibliography Anderson, P, Gustman, A.L. and Steimeier, T.L. (1997), 'Trends in male labour force participation and retirement: some evidence on the role of pensions and social security in the 1970s and 1980s', Working Paper no. 6208, National Bureau of Economic Research, Cambridge, Mass. Bacon, B.R. and Gallagher, P. (1995), 'Early retirees - trends and their use of superannuation benefits and social security payments', Retirement Income Modelling Taskforce, Department of Social Security, Canberra. Berger, M.C. (1983), 'Changes in labour force composition and male earnings: a production approach', Joumal of Hm an Resowrces, vol. 17, no. 2, pp. 177-196. Blanchet, D.C., Brousse, C. and Okba, M. (1996), 'Retraite, preretraite, neutralite actuanielle et courverture du risque de chomage en fin carriere', Ecorcm.que at Statist ique, pp. 291-292. - and Pele, L-P. (1999), 'Social security and retirement in France', in Gruber, J. and Wise, D.A. (eds), Socil Seurity andRetimnent arcwd the World, University of Chicago Press for National Bureau of Economic Research. (Also Working Paper no. 6214, National Bureau of Economic Research, Cambridge, Mass., 1997) Blinder, A.S, Gordon, R.H. and Wise, D.E. (1980), 'Reconsidering the disincentive effects of social security', National Tax Journal, vol. 33, pp. 431-442. Bl6ndal, S. and Scarpetta, S. (1998), 'The retirement decision in OECD countries', Ageing Working Paper 1.4, OECD Paris. (Also Economics Department Working Paper no. 202, OECD) Bodie, Z. (1991), 'Shortfall risk and pension fund asset management', FinanciaAnalysts'Joural, May/June. Boldrin, M., Jimenez-Martini, S. and Peracchi, F. (1999), 'Social security and retirement in Spain', in Gruber, J. and Wise, DA. (eds), Social Seavity andRetinant anxnd dX Word, University of Chicago Press for National Bureau of Economic Research. (Also Working Paper no. 6136, National Bureau of Economic Research, Cambridge, Mass.) Bondar, J. (1993), 'Beneficiaries affected by the annual earnings test, 1989', Social Seamcy Bultin, vol. 56, pp. 20-28. B6rsch-Supan, A. (1992), Population ageing, social security design and early retirement', Jounal of Instiutional ad 7Thwtical Ecorncs, vol. 48. - and Schnabel, R. (1999), 'Social security and retirement in Germany', in Gruber, J. and Wise, D.A. (eds), Social S-uty andRetimnit aund tbe World, University of Chicago Press for National Bureau of Economic Research. (Also Working Paper no. 6153, National Bureau of Economic Research, Cambridge, Mass., 1997) Boskin, Mj. (1977), 'Social security and retirement decisions', EconanicInquiry, vol. 15, no. 1, pp. 1-25. - and Hurd, M.D. (1978), 'The effect of social security on early retirement', Jounal qfPublic Econanics, vol. 10, pp. 361-377. Brugiavini, A. (1993), 'Uncertainty resolution and the tirming of individual annuity purchases', Jownal of Public Econarics, vol. 50, pp. 31-62. - (1999), 'Social security and retirement in Italy', in Gruber, J. and Wise, D.A. (eds), Social Sewrity and Reti7nt arond the Word, University of Chicago Press for National Bureau of Economic Research. (Also Working Paper no. 6155, National Bureau of Economic Research, Cambridge, Mass.) Bulow,J.I. (1982), 'What are corporate pension liabilities?' QuxneyyJounalofEconxns, vol. 97, pp. 435- 450. Burkhauser, R.V. (1980), 'The early acceptance of social security: an asset maximisation approach', Indusn and Labour Relations Review, vol. 33, pp. 484-492. - and Quinn, J.F. (1983), 'Is mandatory retirement overrated? Evidence from the 1970s', Journal of HwnanResowres, vol. 18, no. 3, pp. 337-358. 40 Burtless, G. (1986), 'Social security, unanticipated benefit increases and the timing of retirement', ReZwWof Ecornonc Studies, vol. 53, pp. 781-805. Cain, G. (1986), 'The economic analysis of labour-market discrimination: a survey', in Ashenfelter, 0. and Layard, P.R.G (eds), Handbook of Labour Econanis, North Holland, Amsterdam. Casey, B. (1998), 'Incentives and disincentives to early and late retirement', Ageing Working Paper no. 3.3, OECD, Paris. Chan, S. and Stevens, A.H. (1999), job loss and retirement behaviour of older men', Working Paper, no. 6920, National Bureau of Economic Research, Cambridge, Mass. Chand, S.K. and Jaeger, A. (1996), 'Ageing populations and public pension schemes', Occasional Paper no. 147, International Monetary Fund, Washington, D.C. Chlion, A., Gora, M. and Rutkowski, M. (1999), 'Shaping pension reform in Poland: Seaeity thnxj DkziEty', Pension Reform Primer country case study, Social Protection Discussion Paper no. 9815, World Bank, Washington, D.C. Commission of the European Communities (1996), MISSOC- Social Proection tn the Memihr States ofthe C : Situation on 1 July 1995 and Eolution, Brussels. Couch, K. (1998), 'Late life job displacement', The Gerniokgist, vol. 38, pp. 7-17. Denmark, Ministry of Social Affairs (1995), 'Social policy in Denmark- social pensions and semi- retirement pensions', Copenhagen. Diamond, P. and Gruber, J. (1999), 'Social security and retitement in the United States', in Gruber, J. and Wise, DA. (eds), Social Seacty andRetznrunt amund the World, University of Chicago Press for National Bureau of Economic Research. (Also Working Paper no. 6097, National Bureau of Economnic Research, Cambridge, Mass., 1997) - and Hausman, J. (1984), 'Individual retirement and savings behaviour', Joumal ofPuliic Econagics, vol. 8, pp. 275-298. Dilnot, A.W., Disney, R.F., Johnson, P.G. and Whitehouse, E.R. (1994), Peaions Poliy in the UK: An EconanicAnaysis, Institute for Fiscal Studies, London. Disney, R.F. (1995), 'Occupational pension schemes: prospects and reform in the UK', Fiscal Studies, vol. 16, pp. 19-39. - (1996), Cmn We Affondto GrozvOker? A Pertseateon thx Ecancmcs ofAgemg, MIT Press, Cambridge, Mass. -, Emmerson, C. and Tanner, S. (1999), Partnership in Pensions: An Assessment, Institute for Fiscal Studies, London. Meghir, CH.D. and Whitehouse, E.R. (1994), Retirement behaviour in Britain', Fiscal Studies, vol. 15, no. 1, pp. 24-43. - and Whitehouse, E.R. (1991), 'Occupational and industrial earnings over time: the use of pooled cross- section data', Working Paper no. 91/7, Institute for Fiscal Studies, London. - and - (1996), 'How much are occupational pension entitlements worth in Britain?', Ecanxnmc, vol. 63, pp. 213-238. - and - (1999a), Retirement: the demand side', Social Protection Discussion Paper, World Bank, Washington, D.C., forthcoming. - and - (1999b), 'Notional accounts pensions: microeconomic and macroeconomic aspects', Social Protection Discussion Paper, World Bank, Washington, D.C., forthcoming. Flanagan, K. (1999), 'Pension reform in Australia', Pension Reform Primer country case study, Social Protection Discussion Paper, World Bank, Washington, D.C., forthcoming. Freeman, R.B. (1979), 'The effect of demographic factors on age-earnings profiles', Joumal ofHwnan Resources, vol. 14, pp. 289-319. -(1989), Labour Markets in Actioz Essays i Empiical Ecorcs, Harvester Wheatsheaf. 41 Friedberg, L. (1998a), 'The social security earnings test and labour supply of older men, in Poterba, JM (ed.), Tax Policy and the Econormy, vol. 12, MIT Press for National Bureau of Economic Research, Cambridge, Mass. - (1998b), 'The effect of old-age assistance on retirement', Working Paper no. 6548, National Bureau of Economic Research, Cambridge, Mass. Friedman, B. and Warshawsky, M. (1988), 'Annuity prices and savings behaviour in the US', in Bodie, Z., Shoven, J.B. and Wise, D.A. (eds), Pensions t the US Economy, University of Chicago Press for National Bureau of Economic Research. - and - (1990), 'The cost of annuities: implications for savings behaviour and bequests', QuareyJournal of Econciics, vol. 105, pp. 135-154. Gosling, A. Machin, S. and Meghir, C.H.D. (1998), 'The changing distribution of male wages in the UK', Working Paper no. 98/9, Institute for Fiscal Studies, London. Gruber, J. (1999), 'Social security and retirement in Canada', in Gruber, J. and Wise, D-A. (eds), So"al SecwmriyandRetranautarow7d the Wod, University of Chicago Press fdr National Bureau of Economic Research. (Also Working Paper no. 6308, National Bureau of Economic Research, Cambridge, Mass., 1997) - and Wise, DA. (1997), 'Social security programs and retirement around the world', Working Paper no. 6134, National Bureau of Economic Research, Cambridge, Mass. - and - (1999), Social S wo-* Prqmr nsandRetintarowund the World, University of Chicago Press for National Bureau of Economic Research. Gustman, A.L and Steinmeier, T.L. (1984), 'Partial retirement and the analysis of retirement behaviour', Indrti and LabourRelatians Review, vol. 37, no. 3, pp. 403-415. - and - (1986), 'A structural retirement model', Econc,na,?c, vol. 54, no. 3, pp. 555-584. - and - (1992), 'The stampede towards defined-contribution pension plans: fact or fiction?', Indu7ial Relations, vol. 31, no. 2, pp. 361-369. Honig, M. and Reimers, C. (1989), 'Is it worth elirninating the earnings test?' American EconancReuzew, Papers and T-ings, vol.79, pp. 103-107. Hurd, M.D. and Boskin, M.J. (1984), The effect of social security on retirement in the early 1970s', C)art4Joural of Econacs, vol. 94. Hutchens, R. (1986), 'Delayed payments contracts and a firn's propensity to hire older workers', Joonal of Labour Econcnics, vol. 4, pp. 439-475. - (1988), 'Do job opportunities decline with age?' Industrial and labo2rRelations Revew, vol. 41, pp. 89-99. Ippolito, RA. (1990), Towards explaining earlier retirement after 1970', Indwstrial and LabourRelations Revew, vol. 43, no. 5, pp. 556-569. - (1995), Toward explaining the growth of defined-contribution plans', IndwstrialRelatons, 1. 34, no. 1, pp. 1-20. Jacobson, L., LaLonde, R. and Sullivan, D. (1993), 'Earnings losses of displaced workers', American EconcmnReew, September, pp. 685-709. Johnson, R.W. and Neumark,D. (1996), 'Age discrimination, job separation and employment status of older workers: evidence from self reports', Working Paper no. 5619, National Bureau of Economic Research, Cambridge, Mass. Kalisch, D.W. and Aman, T. (1998), 'Retirement income systems: the reform process across OECD countnres', Ageing Working Paper no. 3.4, OECD, Paris. Kapteyn, A. and de Vos, K. (1999), 'Social security and retirement in the Netherlands', in Gruber, J. and Wise, D.A. (eds), Socia Sealty andRetirea arvud the Word, University of Chicago Press for National Bureau of Economic Research. (Also Working Paper no. 6135, National Bureau of Economic Research, Cambridge, Mass.) 42 Kotlikoff, L.J. and Gokhale, J. (1992), 'Estimating a firm's age-productivity profile using the present value of workers' earnings', QmarterwJoumal of Econanis, vol. 107, pp. 1215-1242. - and Smnith, D.E. (1983), Pensions in the A"necan Econnmy, University of Chicago Press for the National Bureau of Economic Research. - and Wise, D.A. (1985), 'Labour compensation and the structure of private pension plans: evidence for contractual versus spot labour markets', in Wise, D.A. (ed.), Pensions, Labour ad Individual Choice, University of Chicago Press for the National Bureau of Economic Research. - and - (1987), 'The incentive effects of private pension plans', in Bodie, Z., Shoven, J.B. and Wise, D.A. (eds), Issues i Pension Eanan2cs, University of Chicago Press for the National Bureau of Economnic Research. - and - (1989), 'Employee retirement and a firm's pension plan', in Wise, D.A. (ed.), The Econan2cs of Ageing, University of Chicago Press for the National Bureau of Economic Research. Kristhof, K. (1997), 'Personal finance, Los Angles Thnes, September 5. Kruse, D.L. (1995), 'Pension substitution in the 1980s: why the shift to defined contribution?' Inotrinal Relations, vol. 34, no. 2, pp. 218-240. Lazear, E.P. (1979), Why is there mandatory retirement?' Jonal ofPolital Econany, vol. 87, no. 6, pp. 1261-84. - (1986), 'Retirement from the labour force', in Ashenfeher, 0. and Layard, P.R.G. (eds), HarSok of Labour Ecorcs, North Holland, Amsterdam. Leonesio, M.V. (1990), 'Effects of the social security earnings test on the labour-market activity of elderly Americans: a review of the evidence', Social Secuity Bulletin, vol. 53, pp. 2-21. - (1993), 'Social security and older workers', in Mitchell, O.S. (ed.), As the WorkforreAges, ILR Press. Lindeboom, M. (1998), Microeconometric analysis of the retirement decision: the Netherlands', Ageing Working Paper no. xx, OECD, Paris. (Also Working Paper no. 207, Economics Department, OECD) Lumsdaine, R.L. (1995), 'Factors affecting labour supply decisions and retirement income' in Hanushek, E. and Mariato, N. (eds), Assess'gKY oui fRetiznent Beaviour, National Academy Press, Washington, D.C. ,Stock, J.H. and Wise, D.A. (1990), 'Efficient windows and labour force reduction',JoI nalofPu!ic Econcncs, vol. 43, pp. 131-159. and - (1992), Three models of retirement: computational complexity versus predictive validity', in Wise, DA. (ed.), Topics in the Econcmns ofAgeing, University of Chicago Press for National Bureau of Economic Research. and - (1994), 'Pension plan provisions and retirement: men andwomen, Medicare and models', in Wise, D-A. (ed.), Studies m2 the Ecornacs ofAgeing, University of Chicago Press for National Bureau of Economic Research. Meghir, C.H.D. and Whitehouse, E.R. (1996), The evolution of wages in the UK: evidence from micro data', Joumal of Labur Ecowos. - and - (1997), 'Labour-market transitions and retirement of men in the UK', JacnalofEconcmencs, vol. 79, pp. 327-354. Miniaci, R. (1998), 'Microeconometric analysis of the retirement decision: Italy', Ageing Working Paper no. xx, OECD, Paris. (Also Working Paper no. 205, Economics Department, OECD) Miniaci, R. and Strancelli, E. (1998), 'Microeconometric analysis of the retirement decision: the United Kingdom, Ageing Working Paper no. xx, OECD, Paris. (Also Working Paper no. 206, Economics Department, OECD) Mitchell, O.S. and Fields, G. (1984), 'The economics of retirement behaviour',JoarnlofalayurEconanis, vol. 2, pp. 84-105. Morris, C.N. (1981), The age allowance', Fiscal Str!ies, vol. 2., no. 3. 43 Naegele, G. (1996), 'New trends in gradual retirement in Germany', Gei2i Papers on Risk and Insurance, no. 81, pp. 524-533. Nemark, D. and Stock, W.A. (1997), 'Age discrimination laws and labour market efficiency', Working Paper no. 6088, National Bureau of Economic Research, Cambridge, Mass. OECD (1995a), The Transitionfirn Work to Retivmwnt, Social Policy Studies no. 16, Paris. - (1995b), The Labour Market and Older Workers, Social Policy Studies no. 17, Paris. - (1996a), 'Making work pay', Emplat Outlook, July. - (1996b), Econcmic Suvy - Canada, Paris. -(1997a), The OECDJobs Strategr Making Work Pay: Taxation, Benefits, Emplaynt and Unanpljmet, Paris. - (1997b), The OECDJobs Strategy: Mm7lrr Cuntries' Experience, Panrs. - (1998a), T1x Tax-Benefit Positi of Emplows: 1998 Edition, Paris. - (1998b), Laour Force Statistics, 1975-96, Paris. -(1998c), Work-force ageing in OECD countries', Empkcnent Oulook, pp. 123-151, Paris. (Also Ageing Working Paper no. 4.1) Oshio, T. and Yashiro, N. (1999), 'Social security and retirement in Japan', in Gruber, J. and Wise, D.A. (eds), Social Secrity and Retinr-t arund the World University of Chicago Press for National Bureau of Economic Research. (Also Working Paper no. 6156, National Bureau of Economic Research, Cambridge, Mass., 1997) Packard, M.D. (1990), The earnings test and the short-nm work response to its elimination', Soa Seurity Bulletin, vol. 53, pp. 2-16. Palacios, R. and Rocha, R. (1998), 'The Hungarian pension system in transition', Pension Reform Primer working paper series, Social Protection Discussion paper no. 9805, World Bank, Washington, D.C. Palme, M. and Svenson, I. (1999), 'Social security, occupational pensions and retirement in Sweden', in Gruber,J. and Wise, DA. (eds), Social SecuT lyramsandRetanentanxidthe WorAl University of Chicago Press for National Bureau of Economic Research. (Also Working Paper no. 184, Stockholm School of Econornics and Working Paper no. 6137, National Bureau of Economic Research, Cambridge, Mass., 1997) Pestieau, P. and Stijns, J-P. (1999), 'Social security and retirement in Belgium', in Gruber, J. and Wise, DA. (eds), Social Secw yandRetientarnawdthe Workl, University of Chicago Press for National Bureau of Economic Research. (Also Working Paper no. 6169, National Bureau of Economic Research, Cambridge, Mass., 1997) Piggott, J. and Doyle, S. (1999), 'Annuitising mandated retirement accumulations: a primer', Social Protection Discussion Paper, World Bank, forthcoming. Podgursky, M. and Swaim, P. (1987), 'Job displacement earnings loss: evidence from the displaced worker survey', Indstrial and Labour Relatios Reuew, vol. 40, pp. 17-29. Queisser, M. (1998), The second-generation pension refomns in Latin America', Ageing Working Paper no. 5.4, OECD, Paris. Quinn, J.F., Burkhauser, R.V. and Myers, DA. (1990), Passing the Ton*: The Influren of Econcni' Incntizes on Work andRetimnar, W.E. Upjohn Institute for Employment Research, Kalamazoo, Mich. Cahill, K. and Weather, R. (1998), Miniaci, R (1998), 'Microeconometric analysis of the retirement decision: the United States', Ageing Working Paper no. xx, OECD, Paris. (Also Working Paper no. 203, Economics Department, OECD) Reimers, C. and Honig, M. (1993), 'The perceived budget constraint under social security: evidence from re-entry behaviour',Jouma ofLabour Econanics, vol. 11, no. 1, pp. 184-204. - and - (1996), 'Responses to social security by men and women: myopic and far-sighted behaviour?', Jounal ofHwnanResoue, vol.31, pp. 359-382. 44 Robbins, A. and Robbins, G. (1989), 'Paying people not to work: the economic costs of the social security retirement earnings limit', Report no. 142, National Centre for Policy Analysis. Ruhm, C.J. (1990), 'Bridge jobs and partial retirement',Joun&l ofLabourEconanics, vol. 8, pp. 482-501. - (1995), 'Secular changes in the work and retirement patterns of older men', Jounl of Hmn Resowws, vol. 30, pp. 362-385. - (1996), 'Do pensions increase the labour supply of older men?' Joanl ofPubic Econonirs, vol. 59, no. 2, pp. 157.175. Samwick, AA. (1998), 'New evidence on pension, social security and the tining of retirement', Working Paper no. 6534, National Bureau of Economic Research, Cambridge, Mass. Seike, A. (1989), The effects of the employee pension on the labour supply of the Japanese elderly, Note no. 2862, Rand Corporation, Santa Monica, Cal. - (1997), 'Labour-market inplications of social security: company pension plans, public pensions and retirement behaviour of the elderly in Japan', in Hurd, M.D. and Yashiro, N. (eds), 7e Econanc Effects of Agezng in the United States andJapan, University of Chicago Press. Simon, R. (1996), 'How to be sure you never go broke', Money, October. Stock, J.H. and Wise, D.A. (1990a), 'Pensions, the option value of work and retirement', Eccnntnca, vol. 58,pp. 1151-1180. - and - (1990b), 'The pension inducement to retire: an option value analysis', in Wise, D.A. (ed), Isw.es in the Econoracs ofAgezing, University of Chicago Press for National Bureau of Economic Research. Tumbarello, P. (1999), 'Pension reformn in Italy', Social Protection Discussion Paper, World Bank, Washington, D.C., forthcoming. United Kingdom, Department of Social Security (1998), A New C0nractfor Welfae.: Paroersho zn Pensions, London. United States, Department of Health and Human Services (1997), Soial Seawity Pmgrams th7nogjut the Wordd, Washington, D.C. Whitehouse, E.R. (1990), 'The abolition of the pensions "earnings rule"', Fiscal Studies, vol. 11, no. 3, pp. 55-70. - (1999), 'The tax treatment of funded pensions', Social Protection Discussion Paper, World Bank, Washington, D.C., forthconiing. World Bank (1994), Awrt* the Old-Age Crisis: Polxies to Protat th Oldari Pnote GnywtA Oxford University Press for the World Bank. 45 Social Protection Discussion Paper Series continued No. Title 9905 The Effects of Legislative Change on Female Labour Supply: Marriage and Divorce, Child and Spousal Support, Property Division and Pension Splitting 9904 Social Protection as Social Risk Management: Conceptual Underpinnings for the Social Protection Sector Strategy Paper 9903 A Bundle of Joy or an Expensive Luxury: A Comparative Analysis of the Economic Environment for Family Formation in Western Europe 9902 World Bank Lending for Labor Markets: 1991 to 1998 9901 Active Labor Market Programs: A Review of the Evidence from Evaluations 9818 Child Labor and School Enrollment in Thailand in the 1 990s 9817 Supervising Mandatory Funded Pension Systems: Issues and Challenges 9816 Getting an Earful: A Review of Beneficiary Assessments of Social Funds 9815 The Quest for Pension Reform: Poland's Security through Diversity 9814 Family Allowances 9813 Unemployment Benefits 9812 The Role of Choice in the Transition to a Funded Pension System 9811 An Alternative Technical Education System: A Case Study of Mexico 9810 Pension Reform in Britain 9809 Financing the Transition to Multipillar 9808 Women and Labor Market Changes in the Global Economy: Growth Helps, Inequalities Hurt and Public Policy Matters 9807 A World Bank Perspective on Pension Reform 9806 Government Guarantees on Pension Fund Returns 9805 The Hungarian Pension System in Transition 9804 Risks in Pensions and Annuities: Efficient Designs Social Protection Discussion Paper Series continued No. Title 9803 Building an Environment for Pension Reform in Developing Countries 9802 Export Processing Zones: A Review in Need of Update 9801 World Bank Lending for Labor Markets: 1991 to 1996 Summary Findings The value of defined-contribution pensions, where the benefit depends on contributions and investment returns, shows a very different pattern with age of retirement from defined-benefit pensions. DB schemes, which are the norm in most public and much private provision around the world, provide an incentive to retire at the earliest possible age. DC schemes, in contrast, encourage people to remain in the labor force. The paper also assesses the features of DB plans which provide the greatest disincentive to continued employment at older ages, such as their accural structure with age and formulae based on final rather than average earnings. Reforms to DB systems, such as pension increments for later retirement, are examined. But the incentives are not as powerful as those in a DC plan. 3bout thi rHUMAN DEVELOPMENT NETWORK ... ?,e W_d ,_an peso ,_o,~ie~i spoie~ opeesv, up4 - _ai reoreft poI Xe~nn and rimpee ttges,kLS refom Wa'