PART I THE CONCEPT OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEM VARIATIONS ON A THEME

Transcription

1 PART I THE CONCEPT OF NON-FINANCIAL DEFINED CONTRIBUTION SYSTEM VARIATIONS ON A THEME 15

2

3 Chapter 2 What Is NDC? Edward Palmer* IT IS ALWAYS A PRECARIOUS BUSINESS TO TRACE THE ORIGIN OF AN IDEA, but among the published works available to an international audience, Buchanan (1968) probably contains the first proposal that can be likened to what have emerged as notional or non-financial defined defined contribution (NDC) pension schemes. 1 A proposal in Boskin, Kotlikoff, and Shoven (1988) is also on this track, and the point systems of Germany and France certainly contain the embryo of an NDC system. Nevertheless, NDC as it is now known was not given a face until the arduous task of redesigning pension systems was taken on in a series of reforms in Europe in the 1990s. It is probably safe to say that NDC has emerged as a result of the interchange of ideas and discussions among pension experts and politicians working on pension reform issues in those countries where NDC has been implemented. Since the mid-1990s, the exchange of ideas about the possible contribution of NDC to the pension policy toolbox has been truly international, with both avid proponents and a healthy number of skeptics. To date, however, ten years after NDC legislation was passed in two countries Sweden and Italy no paper has attempted to bring all the strands together to provide a cohesive technical framework. That is the ambition of this study. In the general discussion in the next section, this study touches briefly on the differences in principle between NDC and alternatives for mandatory national pension schemes. The purpose of the study, however, is not to argue the comparative merits of alternative pension schemes, or to discuss what countries implementing NDC have done in practice. Instead, the aim of this study is to present what an NDC pension system is. The second section of the study briefly discusses efficiency and distribution with respect to NDC schemes. In the third section the generic NDC scheme is presented. The fourth section deals briefly with considerations in integrating social policy into the NDC framework and touches upon some other issues that can arise in constructing an NDC scheme. The study closes with a brief summary and some final remarks. * Edward Palmer is professor of social insurance economics at Uppsala University and head of the Division for Research at the Swedish Social Insurance Agency. The author is grateful for comments on this chapter generously provided by Robert Holzmann. 17

4 18 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES A General Overview of NDC A First Description of NDC An NDC scheme is a defined contribution, pay-as-you-go (PAYG) pension scheme. Contributions are defined in terms of a fixed contribution rate on individual earnings. These contributions are noted on an individual account. As opposed to a financial defined contribution (FDC) scheme, the contributions of participants noted on individual accounts are not funded. More specifically, individual account money is not invested in financial market instruments. Compared with an FDC scheme, where individual account money is invested in financial market assets, by definition, the pay-as-you-go individual account DC scheme is a notional DC scheme. It can be argued that it is more precise to use the terminology financial and nonfinancial that is, financial DC and non-financial DC to distinguish these two types of defined contribution schemes, stressing the fact that money in accounts is invested in market assets in the FDC scheme, but not in the NDC scheme. Note that in the pension literature financial DC schemes are generally referred to as individual account schemes. Given that an NDC scheme is also an individual account scheme, the use of the term individual account scheme to describe financial account schemes seems ambiguous. Góra and Palmer (2004) argue that financial individual account schemes should be called FDC schemes, and notional or non-financial account schemes should be called NDC schemes, which is the terminology used throughout in this study. Before describing NDC in greater detail, two brief statements about the difference between NDC and FDC individual account schemes are helpful. First, they differ with respect to their (potential) contribution to national saving. An NDC scheme provides no direct contribution to saving, except through the possible mechanism of demographic saving. An FDC scheme contributes to national saving during the build-up phase to maturity, but thereafter the potential effect on national saving will depend on the demography of the scheme participants. Furthermore, it is a well-discussed caveat in the economic literature on pensions that the net overall effect of mandatory pension schemes on national saving depends on private and government offsetting behavioral responses. Second, NDC and FDC differ with respect to the system rate of return. Participants in an FDC scheme earn a financial market rate of return, whereas participants in an NDC scheme earn an internal rate of return, which is determined by factors underlying the development of the economy. The next section discusses potential economic consequences of the differences between NDC and FDC schemes in greater depth. Now, what is an NDC scheme? A general description follows. In the next section, more precise criteria for a generic NDC scheme are developed. Basically, in an NDC scheme, participants or employers on their behalf pay contributions on earnings during their whole working career. Although there is a minimum age at which an annuity can be claimed, there is no pension age. As long as people continue to receive earnings from work, these earnings generate contributions to individual accounts. This is true even if the individual is already drawing on a partial or full annuity after the minimum age to claim the annuity. Once again, each period s contributions are noted on an individual account. This notional individual account grows with new contributions and is credited with a rate of return. In the generic NDC scheme, the rate of return is the internal rate of return. The NDC benefit is a life annuity. It can be claimed at any time from the minimum retirement age. The generic NDC annuity embodies a rate of return based on the same internal rate of return that is credited accounts during the accumulation phase and, importantly, cohort life expectancy at the time the annuity is claimed. Since newly granted annuities reflect life expectancy, in principle, NDC is an actuarially fair pension system.

5 WHAT IS NDC? 19 Thus an NDC scheme distributes individual resources over the life cycle, but within the framework of a countrywide (universal) PAYG insurance scheme. In this sense it can be likened to an illiquid, individual cash balance scheme, with the important difference that it is an insurance scheme: that is, it redistributes the capital of the deceased to the survivors in the scheme. It fulfills the function of insuring against the individual risk of outliving the average participant. Because the NDC benefit is based on individual contributions from individual earnings, it may not provide sufficient coverage in old age for everyone because, for one reason or another, some individuals will not have had sufficient earnings before retirement to provide a sufficient benefit in old age. This is not a problem inherent to NDC; rather, it is a characteristic of any earnings-related benefit. As opposed to many possible defined benefit (DB) formulas, the pension formula in an NDC scheme contains no built-in redistribution of the system s revenues. Thus an NDC scheme must be supplemented with some form of low-income support scheme, in the form of minimum income or minimum benefit guarantee. NDC accounts can also be supplemented with noncontributory rights, e.g., for childcare years. What is essential is that these be financed with revenues exogenous to the NDC scheme, i.e., general tax revenue. Efficiency and Distribution Both NDC and defined benefit pay-as-you-go that is, non-financial defined benefit schemes differ from financial schemes in terms of economic efficiency. The relevant discount rate for money set aside today to pay for consumption tomorrow is the financial market rate of return. In terms of this metric, a pension scheme that gives a risk-adjusted rate of return below the risk-adjusted financial market rate creates a tax wedge, and this means participants would attain a higher level of lifetime consumption with the counterfactual financial market rate of return. Rated in terms of efficiency, non-financial pension schemes, among these NDC schemes, are only potentially at least as efficient as financial pension schemes. They would be as efficient as financial schemes in a steady-state (closed) Golden Rule economy 2 but otherwise not necessarily so. If the rate of return on assets, r, traded in the financial market dominates the growth rate of the economy, λ + g, where λ is population (labor force) growth and g is productivity growth or, more exactly, the internal rate of return (IRR) of the NDC scheme (see the next section), 3 then financial schemes dominate non-financial schemes in terms of efficiency. Nevertheless, although empirical evidence from the past century suggests this is the case, at least if the account portfolio includes equities, 4 it need not always be the case in specific country settings and specific periods. Although r > IRR means, ceteris paribus, that an FDC scheme dominates an NDC scheme, by linking individual benefits solely to individual contributions, an NDC scheme does not have the tax effect inherent in a DB-PAYG scheme. However, in the final analysis one should take into account the effects of the tax regime. 5 For example, a country s taxes on earnings and capital may differ. In addition, although by definition the IRR of an NDC scheme depends on the development of the country s economy, in a comparison with an FDC counterfactual, it is important that two other considerations are kept in mind. First, an FDC scheme with a portfolio dominated or exclusively invested in home country debt is exposed to home country risk, just as NDC is. Second, an FDC scheme counterfactual, also holding only or mainly government bonds, can be likened to an NDC scheme fully monetized with Buchanan bonds. However, if the government bond rate is greater than the IRR, this advantage will be paid for through taxes levied on the same countrywide insurance collective, that is, the

6 20 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES recipient of the higher return; although there are likely to be tax-distributional effects that distinguish these two approaches. The expected value of an NDC pension in any period t is exactly the amount in an individual account at that time. This amount, in turn, has, been determined by the individual s contributions and the rate of return on his or her account. This is just another way of saying that in a generic NDC scheme there is no built-in mechanism either explicit or implicit to redistribute income, as opposed to a DB scheme. A DB scheme involves at least some within-scheme redistribution, by definition. Furthermore, an NDC scheme differs from defined benefit pay-as-you-go (DB-PAYG) schemes in how it deals with the effects of changing demographic and economic factors. In an NDC scheme, the effects of economic and demographic fluctuations are accommodated endogenously, as the NDC internal rate of return adjusts account values during both the accumulation and the payout phases, and as annuities adjust to changes in birth cohort life expectancy at retirement. In a DB-PAYG scheme, benefits are fixed by contract at least in principle and system adjustments to exogenous economic and demographic shocks are accommodated through changes in the contribution rate. Obviously, this approach to dealing with exogenous economic and demographic risks can generate very different distributional outcomes from NDC, which adapts by adjusting accounts. Because an NDC scheme has a constant contribution rate over time, successive generations of participants can expect to pay the same fixed percentage of their earnings into the scheme. If this property is chosen as a definition of intergenerational fairness, then an NDC scheme is intergenerationally fair. Each generation will pay a fixed percent of lifetime earnings into the scheme and can expect to receive a stream of benefits, determined by these individual payments and the system rate of return. NDC schemes can nevertheless give varying outcomes across cohorts. This is because the value of individual benefits depends on the intertemporal rate of return, which will vary from period to period. This is true of any pension scheme, however. Individual outcomes are a function of either the market rate of return (financial schemes) or some form of indexation (non-financial schemes), and the distribution of individual outcomes over time will be a function of either the development of the financial market or the economy, respectively, with the exact outcome depending on how the system rules link the scheme to either of these. Variation in the NDC intertemporal rate of return means that outcomes for given amounts of real valued contributions can vary over cohorts, and even over persons in the same birth cohort, depending on when they enter and exit the labor force. A generic NDC scheme distributes the resources of the scheme through the time dimension of the IRR, to be defined more precisely in the next section. In addition, it is important to note that, to date, all country designs of NDC have included a transfer mechanism through the actual annuity formula employed, through the application of unisex life expectancy. Furthermore, taxes and transfers reenter the picture indirectly through the external (to the NDC scheme) tax and transfer system, since the government can transfer general tax revenues to the accounts of the NDC participants. It follows then that what the NDC scheme does in practice is to provide a setup for transparent accounting of the flows of money into the overall system and the sources of finance for these flows. NDC does not eliminate the political risk in this sense, but it provides a transparent framework for weighing the pros and cons of political decisions on taxes and transfers regarding a nation s pension system. In sum, an NDC scheme is dominated by or is at least as good as an FDC scheme on the basis of efficiency. Whether a particular FDC counterfactual dominates a generic NDC

7 WHAT IS NDC? 21 scheme in practice will depend on the investment portfolio of the FDC scheme and the taxregime backdrop. In addition, dominance is not only dependent on these institutional considerations, it is also time-dependent, as the relation between the financial market rate of return and the IRR is dependent on the home economy and indirectly, on the international economy, and the domestic and international financial markets. By definition, payas-you-go DB schemes include intra- and intergenerational tax distortions through internal redistribution not embedded in a generic NDC scheme but NDC is not necessarily free of distributional effects, either. Also, by maintaining a constant contribution rate, an NDC scheme means all generations pay the same percent of earnings into the scheme. The value of these contributions depends, however, on the economy-determined rate of return, but this is also true of DB-PAYG schemes. Finally, for the same intragenerational distributional goals, an NDC scheme also dominates defined benefit schemes by making any form of distribution and the financial sources for this transparent. Generic NDC Design The central component of the NDC scheme is the individual lifetime account. This section describes how account values arise and how annuities are calculated. It begins with the concept of notional capital. Note that the entire discussion is in terms of real values. In practice, however, the system is inflation-neutral since it is set up in terms of nominal values. NOTIONAL CAPITAL According to the NDC formula, for an individual, i, paying contributions on wages (or earnings), w, during an accounting period, t, with contribution rate, c, notional capital at the end of any period, T, is T K = cw I it, it, t= 1, (2.1) where cw i,t denotes the individual s contributions in period t. Note that usually a ceiling will be set on earnings giving rights in the system. I t is an index, calculated from an internal rate of return, α t : t I I t T T 1 1 = ( + α ) t+ 1 = 1. t (2.2) The internal rate of return is discussed in greater detail below, but for now it is sufficient to say that it is the rate of return that maintains financial balance over time, for a given contribution rate. Some participants will not live long enough to claim a benefit, and this creates inheritance capital, which in a closed system will be distributed to the account values of the survivors in the insurance collective. In practice, this can be done on a birth cohort basis, and in accordance with the individual s share of capital in the cohort s total capital. The inheritance capital is distributed in conjunction with some set age, such as the minimum age at which an annuity can be claimed. This adds another component to individual capital at retirement, increasing the return on capital.

8 22 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES THE ANNUITY Total capital at retirement, beginning in period τ, is the sum of capital accredited the individual s account prior to the period of retirement. Total capital at the end of the period prior to retirement, K τ 1, divided by an annuity factor, G, gives the initial value of the annuity, P τ, for pensioner j, from birth cohort κ P j, τ = K j, τ 1 ( ) GLE κ, α LE κ (2.3) The annuity is a function of two factors. The first is cohort-based life expectancy, LE, for cohort κ at retirement. The second is the internal rate of return, α, computed over the same period. The latter becomes an indexation formula in practice. The life expectancy factor. In principle, men and women would have separate LE factors. However, in countries where NDC schemes have been introduced, the annuity is calculated using unisex life expectancy. The use of unisex life expectancy in calculating the annuity introduces an explicit redistribution from those who live a shorter life to those who live a longer life, for this reason, given current longevity patterns and a unisex LE factor that entails a transfer from men to women. The rate of return and the life annuity. An NDC scheme provides a life annuity that encompasses a real rate of return determined by the internal rate of return, discussed below. The rate of return for the period over which the annuity is to be paid out is only known ex post, whereas the annuity is calculated ex ante. This means that in practice a computational rule is required to determine how α is to be entered into the annuity formula. One procedure is to include an ex ante value of α in the formula. In practice this could be any value reasonably close to the expected internal rate of return. This approach requires a second rule for dealing with the deviation of the actual outcomes from the α included ex ante in the formula used to compute the annuity. A convenient method is to adjust (for example, annually) for the difference between the rate of return included in the annuity, which can be called the norm, and the actual rate of return determined on an ex post basis (annually). If the actual return is less than the norm used in the annuity, a negative correction is called for, and vice versa. 6 In practice, use of formula 2.3 frontloads the annuity and, in economic terms, creates a higher annuity in the beginning of the total payout period compared with straight-forward yearly indexation, but at the expense of a lower annuity (compared with straight-forward annual indexation) in the latter segment of the payout period. This is consistent with assuming that the time preference of individuals is weighed in favor of more consumption now than in the future, and creating the annuity to reflect this assumption. The alternative to including the rate of return in the annuity is to index the annuity annually, still with the internal rate of return. This form of real indexation gives higher consumption for each year into the future as long as there is real growth but at the expense of a lower initial value. This is because the individual lifetime pension-benefit stream available to distribute is fixed. This in turn means that if the annuity is higher from the beginning, by including a norm in the calculation of the initial benefit, then benefit-based resources available for consumption when young are also higher. However, since there is a fixed sum of money to distribute over the lifetime for the individual, resources will be lower (relatively to a contemporaneous wage earner s earnings) when older.

9 WHAT IS NDC? 23 Note also that the individual account is an entity that exists until death. This is a convenient property, for two reasons. First, in practice, it is always possible to continue to work and pay contributions on earnings, even after a full annuity has been claimed. Additional work and contributions always enhance account values. Second, an annuity can always be converted into an account value and vice versa, using formula 2.3, at any time after retirement, using the expected payout period to make the conversion. Note also that it is possible to grant partial annuities of any percentage since it is technically easy to convert annuities back into account values and vice versa. This feature of the NDC scheme makes it very suitable for combining work and pensions in any proportion, enabling gradual retirement from the labor force. In sum, the lifetime capital of an individual j from retirement is determined by his or her account at retirement, cohort life expectancy at retirement, and the internal rate of return, expressed as a norm based on the internal rate of return (with the deviation correction factor) or in the form of annual indexation with the internal rate of return. What s more, total (available) individual lifetime benefit payments in the NDC scheme can be distributed over the life cycle of the retiree, either weighting them more heavily toward the beginning of the retirement period or allowing them to increase gradually as the country s prosperity increases during the life annuity period. THE INTERNAL RATE OF RETURN AND FINANCIAL BALANCE The internal rate of return is the rate of return required to keep the scheme in financial balance, where financial balance is defined as the state where the present value of overall system assets PV(A t ) equals the present value of total system liabilities PV(L t ): that is, PV(L t ) = PV(A t ). (2.4) System liabilities at any time t are the sum of all NDC commitments to all living participants, both workers and pensioners. Where the liability to worker i is K i,t and the liability to pensioner j is P j,τ (where, for simplicity, τ denotes the date when members of cohort κ became pensioners) 7 total liabilities are PV(L t ) = Σ K i,t + Σ P j,τ,t (2.5) at time t. The present value of assets is the present value of the stream of future contributions for all workers in period t plus funded (technical) reserves. This can be expressed as ψ t i, t t i= 1 PV( A ) = TD cw + Fund (2.6) where TD denotes the turnover duration of contribution assets. TD is a density function, developed and described in Settergren and Mikula (2006). In discrete terms, the density function in period t is the product of the earnings-weighted average number of years participants have worked based on the age-earnings profile in period t and the payment weighted average number of years payments have to be made in year t based on the payment profile in period t. The turnover duration of the contribution asset is the average time a unit of money (such as a dollar or a euro) is in the system. In a discrete time frame, TD represents a static time cross-section density of money flows in the system, as seen at time t. With each new period t there is potentially a new density of money flows, thus generating a new value of TD.

10 24 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES In the generic NDC model, both the accounts of workers during the accumulation period and the accounts of pensioners converted into annuities earn the same rate of return in a given period t. In a steady state, defined by a fixed workforce-age-wage-distribution and fixed agemortality rates, an NDC rate of return based on the (instantaneous) rate of growth of the contribution base is sufficient to maintain financial equilibrium. The rate of return is determined by the rate of growth of productivity, g, and the population, or more specifically in the actual pension context, the labor force, λ. In a steady state, the latter is determined by the rate of growth of the working-age population, and is what Samuelson (1958) called the biological rate of return. The rate of return of λ + g is not sufficient to maintain financial equilibrium 8 under various circumstances. Given any rate of economic growth that is, λ + g different distributions of contribution and benefit payment flows will affect the NDC scheme s ability to maintain temporal balance between assets and liabilities. In insurance terms, the time money remains in the system before it has to be paid out affects the instantaneous liquidity of the system. This is shown in Settergren and Mikula (2006) in the context of a morethan-two generation model, since two-generation models are not sufficient to bring out this feature. If it were known in period t that a unit of money were to remain in the system some duration of time so that TD t is greater or less than TD t-1, then the value of assets according to expression 2.6 would change solely as a result of this, and this change would constitute a component of the internal rate of return ρ = [PV(A t )/PV(L t )] 1. (2.7) In the generic NDC scheme, this gives an internal real rate of return, α = g + λ + ρ, (2.8) also derived in Settergren and Mikula (2006). The component ρ constitutes an adjustment of the well-known real growth criterion, g + λ, which can arise in nonsteady state. In practice, an operative procedure has to be developed to determine the value of ρ, based on an estimate of liabilities and assets (see below). Also in practice, ρ will consist of a pure payment timing component and a component that picks up what can be called system noise : that is, design features that can lead to a less than ideal outcome. There is a market-based financial alternative to determining the internal rate of return. This would be for the government to issue Buchanan bonds to cover the liabilities of the system. Full market monetiziation of an NDC scheme through the issuance of Buchanan bonds can be criticized on two counts, however. First, if the wage sum, which using the terminology here is Σ w i,t (from equation 2.1) is proportionate to national income, or approximately so over the long run which would be the case if the share of profits in income were constant over time then contributions, which constitute new account assets, would also be a constant proportion of national income and the tax base. This means that new Buchanan bonds earn a government decreed rate of return, equal to growth of income, λ t + g t, and that the ratio of government Buchanan bond liabilities to national income could grow at the rate λ t + g t for any given contribution rate, c, (and any given profit ratio). However, if the government were to set a rate of return on the Buchanan bonds of λ t + g t, this would be tantamount to running a nonmonetized NDC scheme with internal rate of return λ t + g t but with a deadweight loss required to cover

11 WHAT IS NDC? 25 market issuance and transaction costs. This makes the nonmonetized NDC scheme more efficient. 9 In addition, this procedure would not account for the possible effects on system finances of the factor ρ. Second, if the bond (which over time would have to be a series of bonds) were to be a normal government bond(s), the rate of return would depend on the government s status as a borrower on the market. In such a set-up, if the rate of return on bonds is greater than the rate of growth, the money transferred to NDC accounts through this mechanism would be financed by a tax on the same persons who are the workers and pensioners in the NDC scheme. This not only introduces transaction costs, but also rate of return and a time-related distributional profile that depend on the government s creditworthiness and market evaluation of government policy. Properties of NDC NDC is identified through a set of properties, which are satisfied by setting up the system as described above. These properties are: Property 1. At any time the present value of an individual s lifetime benefit equals the individual s account balance. For each participant and at all times, the amount in the account, K, is the present or expected value of his or her benefit. The value of the account is determined by the individual s own contributions and the system s internal rate of return. This is a first property of an NDC scheme. It can be stated (where E denotes the expected value) as E(P t ) = K t. (2.9) This is a property of a defined contribution system. Note that it has an important economic implication: Contributions constitute a payment for the individual s own pension, similar to premium payments to a financial account system. Fulfillment of this property is a necessary condition for a scheme to be called an NDC scheme. Property 2. To maintain a fixed contribution rate, total NDC system assets must equal or be greater than total liabilities. In the generic system, set out above, assets always equal liabilities. This gives a second property defining an NDC scheme A t L t. (2.10) Fulfillment of this property is necessary to maintain a constant long-term contribution rate. This is also a property that is consistent with a financial defined contribution scheme, where A = L. The process of individual account valuation distributes the returns among the participants in the generic NDC system, fulfilling the equality In practice, design issues impose various constraints. As long as A > L, NDC schemes other than pure generic NDC schemes can also be called NDC schemes according to this definition. Property 3. The NDC benefit is constructed as a life annuity reflecting life expectancy at retirement. A key feature of an NDC scheme is that the annuity is based on life expectancy. A correct valuation of life expectancy leads to a correct valuation of the liabilities associated with annuity payments. Even if a lump-sum payment were theoretically possible, it is not a

12 26 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES desirable feature of a mandatory system that aspires to protect people in old age (and taxpayers) from the temptations of short-sightedness or the results of otherwise poor planning or management of money. Property 4. Financial balance in the generic system requires that accounts be valued at the internal rate g + λ + ρ. If in practice the rate of return is set lower, for example, by not factoring full real growth into the annuity, which may be the result of a policy restriction during the implementation phase 10 than compared with the counterfactual of generic NDC, a surplus arises and if it is used for purposes other than accrediting accounts, then it should be viewed as a tax. If the system is monetized with Buchanan bonds, then the monetization process gives rise to transaction costs, which is also a tax, albeit a tax whose financial source is exogenous to the NDC system. In this case, it does not impede upon the function of the internal rate of return. RESERVE FUND All other things equal, if the NDC scheme is not monetized with Buchanan bonds, the occurrence of large and small fluctuating birth cohorts means that, ceteris paribus, contributions of large cohorts should be funded and paid out when the large cohorts become pensioners. If this is not done, there will be a tendency for the system to be out of financial equilibrium. For example, in a two-or-more-cohort world with a large and small cohorts, this would result in a situation where L > A when large cohorts are working, with the difference between this outcome and A = L being the amount of the missing fund. With automatic balancing (see discussion below) the automatic balance would be triggered to restore the condition A = L. Also, the rate of return on reserves must be at least equal to the internal rate of return. The reserve fund serves the purpose of collecting and disbursing the demographic balance. It would be difficult, if not impossible, to do this without the institution of the reserve fund. More generally, the reserve fund is a general buffer fund. Its sources of funds and withdrawals will depend on the exact design of the system. A number of examples can be given of how design affects the reserves. One is that accounts can be augmented with noncontributory rights. These must have a source of financing, such as general tax revenues or a special contribution rate earmarked for this purpose, and these revenues should be transferred to the reserve fund. The financial source is the asset counterpart of the liability to the account holder. This money should also earn at least the rate of return accredited individual accounts of workers. A second example is a tax overhang from the old system, to be elaborated upon below, which should be quantified and funded in the sense that it also needs to earn the internal rate of return. A third example of the use of the reserve fund is the Swedish reform principle, which uses g to index accounts and benefits, based on the principle that the value of benefits should increase at the rate of growth of the per capita covered wage. A liquidity band is defined between the value A/L = 1 and some value greater than unity. As long as A/L remains within the band, the surplus remains as liquidity in the system, and the system is insulated from changes in λ + ρ in expression 2.8. Liquidity in excess of the band ceiling is distributed according to a rule, and the automatic balancing mechanism (see below) balances the system for A < L. Within the band, no reserves are disbursed. In practice, reserves are needed for other reasons, as will be discussed in the next section.

13 WHAT IS NDC? 27 AUTONOMY An NDC system is self-adjusting financially and can run without external intervention. The generic system is autonomous by construction, which is a result of fulfillling properties 1 through 4, and maintaining a reserve fund. Of course, whether politicians will let the system run free from intervention is another question. Finally, no country has introduced the generic model as it is presented here. Design in implementation varies considerably among countries that have implemented varieties of NDC schemes. Where they generally differ is in the indexation rules chosen, and to date, only Sweden has legislated an ABM. There are many reasons for the deviations from the generic model. These include the transitional considerations of phasing in the indexation of new NDC benefits with the already existing stock of pension benefits at the time of implementation (Palmer 2006), political goals, and data and information requirements. Automatic Balancing Because system designs deviate from the generic model, in practice, equation 2.7 will encompass all design deviations from the generic model. Hence in practice, equation 2.7 yields b t = [PV(A t )/PV(L t )] 1 (2.11) where b t = ρ t if the model is generic, but otherwise not. The value b t is the factor with which the accounts of workers and annuities of pensioners needs to be adjusted to maintain balance. This is the automatic balancing mechanism (ABM). The ABM mechanism is described as follows. If PV(A t ) < PV(L t ), regardless of the cause, account values of workers and annuities of pensioners will have to be given a lower rate of return to bring the system back into financial equilibrium. On the other hand, the system can distribute a higher rate of return to participants if PV(A t ) > PV(L t ). The ABM is used in the country in which it was conceived, Sweden. 11 Other countries adopting NDC have relied on system design that can be expected to give A > L primarily as a result of not distributing all possible internal returns. This is tantamount to imposing a tax on returns and using the revenues for other purposes. 12 Some examples of the most common sources of generic design deviations follow. One is that ex post mortality rates can deviate from the ex ante values used to compute life expectancy for newly granted annuities. Another is that annuities may cost more (or less) than thought ex ante. For example, persons with higher than average incomes (and contributions) may live longer, and the annuity factor may not be designed to account for this, or if it is, it might not do the whole job. Another form of generic deviation would be the choice to use the per capita rate, g, as the rate of return, rather than a measure closer to the internal rate of return. This choice can be motivated by the goal of maintaining a constant ratio of an average pension to an average wage. Yet another form of generic deviation is the choice of the wage sum rate g + λ for indexation, instead of g + λ + ρ. Note also that, ceteris paribus, the choice of g + λ for indexation would be the same as using g together with a balancing mechanism without the proposed Swedish liquidity interval. There are more examples of how financial imbalance can arise. A rate of return on reserves that deviates from the rate accredited individual accounts also creates imbalance. Potential imbalance is also caused by the fact that in practice indices used to compute the rate of return need to be based on historical data, and probably a smoothing mechanism,

14 28 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES which means they are always one or more periods behind. This may be no or only a small problem for random positive and negative fluctuations, but it can become a serious problem for negative trends, especially in λ, if g is the chosen system rate of return. Finally, there may be a liability overhang after conversion from the old DB to the new NDC scheme, with unfinanced liabilities. How a tax overhang should be dealt with is discussed below. Valdés-Prieto (2006) presents an alternative to the ABM. He envisages market determination of the balancing mechanism through the issuance of a limited number of bonds with property rights, which he labels integration to financial markets, or IFM. He argues that full coverage of NDC debt is not essential to its valuation, and that a limited issuance of debt would be sufficient to enable the NDC rate of return to be determined by the market. It is not evident, however, why this rate of return would be preferable to λ + g + ρ. Computing Life Expectancy Some procedure must be adopted to compute life expectancy at retirement. At least three procedures can be employed in practice. These are described and discussed briefly in this section. One approach is to form a committee of demographic experts, charged with the responsibility of performing analyses that lead to official cohort projections. Their proceedings, analytical report, and minutes from deliberations would be published to assure that the process is as transparent as possible. Revisions can be made annually, but with relatively accurate long-term projections from the outset, only small, infrequent revisions would be needed. Revisions would always apply to nonretired cohorts and would be greater the farther from retirement age the cohort is. The transparent process would support system autonomy from political interventions in the projection process. A second approach is to base the estimates of birth-cohort life expectancy on known ex post cross-sectional survival data. This provides stronger protection from political intervention since it involves no process of judgment. However, inherent in this procedure is the almost certain risk that projections will lag behind reality, systematically creating higher than warranted annuities for older workers, ceteris paribus, shifting costs to future cohorts. This financial strain on the system is eventually rectified if the system design includes automatic balancing, but a distributional effect remains. Furthermore, all other things equal, this procedure yields an unfinanced deficit. A third alternative is to use current (ex post) cross-sectional data and regularly adjust the benefits of all retirees in accordance with new life expectancy tables as they are revised. Compared with trying to get the projection right from the beginning, as in the first alternative, or burying the problem in the general error-correction mechanism, as in the second, the naked perpetual change model, although right on target, is nevertheless the least appealing for the policy maker. All other things equal, it means that the administrator will be constantly decreasing the value of the benefits of retirees of all ages as new life expectancy statistics become available. For some longer-lived pensioners, this process of benefit deflation could go on for over a quarter of a century. This method shares the ceteris paribus disadvantage of the second method, where the automatic balancing mechanism performs the task of balancing all possible causes of a deviation between assets and liabilities, including errors in projecting life expectancy. On balance, the first and third approaches explicitly fulfill the criterion of financial balance, whereas the second and third provide greater autonomy, at least in principle, since they are based on actual outcomes, which presumably cannot be influenced by politicians. The first approach has a good chance of being autonomous, however, if its operation and proceedings are made transparent for the public.

15 WHAT IS NDC? 29 Converting to NDC with a Tax Overhang from the Old DB Framework A country may wish to introduce NDC but is confronted with what we will call a tax overhang from the previous system. A tax overhang can be defined vis-à-vis the counterfactual of having had an NDC scheme from the very outset, including a contribution rate fixed at some specific level. This is obviously a controversial statement, but the logic in it is that if the policy maker has decided that NDC is the best option, then it is consistent to view old pay-as-you-go commitments that cannot be financed by the contribution rate chosen for the NDC scheme as a tax overhang. An example of a tax overhang would be the failure to create a fund for a large birth cohort of persons for whom most of their working careers have lapsed at the time of conversion to NDC. Another example would be old-system commitments that are more generous than those the NDC formula provides, and which the government decides to honor as a part of the transition to NDC. In principle, the tax overhang can be estimated and the government s commitment to honor it can be made transparent by including the debt to the system as a fund asset in equation 2.11, which defines total assets. To finance the overhang, the government could transfer money or issue bond(s) to be held by the pension fund, yielding the internal rate of return, and which would be monetized at a pace needed to finance tax-overhang commitments. Alternatively, the amount owed would be kept on the books, and the government would finance the amount owed (with the internal rate of return) as payment commitments occur, through a tax levied as the liabilities have to be honored. In practice, there is no difference between these two methods, although issuing a bond has the formal advantage of being a more steadfast commitment. NDC and Social Policy Considerations The Scale of the System The introduction of NDC does not mark the downfall of social policy. On the contrary, it provides a means for separating the goals of social policy from the goal of smoothing out individual consumption over the life cycle with the aid of a mandatory insurance scheme. A first consideration is that the scale of the NDC scheme like any other mandatory pension scheme should be set so as not to impose unjustifiably on individual preferences regarding the time preference of consumption. For example, in poorer societies, a large pension in the future paid at the expense of current consumption can easily conflict with the obvious time preference of persons living in poverty who would prefer higher consumption in the present. This suggests a lower scale for the system and a high minimum pension age. On the other hand, in richer societies, many people but never all can prefer a lower scale to leave room for individual choices for work and saving over the life cycle. A reasonable scale goal in a developed market economy is to provide an adequately dimensioned NDC benefit in old age for the normal worker retiring at or close to the minimum retirement age, but leaving room for individual choice regarding consumption and leisure. The minimum retirement age needs to be set so that people cannot claim a benefit at an age that by definition gives an annuity that on average is too low, given that benefits are based on life expectancy from retirement. Generally speaking, the size of an annuity is determined by individual preferences between work and leisure. In an NDC scheme, individuals themselves determine the timing and scope of their withdrawal from the labor market. If the tax rules provide equal treatment of pensions and earnings, an NDC scheme is neutral in the work-leisure decisions of older workers (at least above the guarantee level).

16 30 PENSION REFORM: ISSUES AND PROSPECTS FOR NDC SCHEMES Also, NDC accounts are infinitely devisable. This means that only practical considerations determine what percent of a full annuity an individual can claim at any time after reaching the minimum pension age and it is possible to combine work and a pension in any proportions and at any age. Finally, NDC (like FDC), by basing benefits on individual accounts, facilitates labor mobility between occupations, branches, sectors of the economy, and regions of the country, or for that matter, between countries. Minimum Pension Income for the Elderly By definition, although an NDC scheme provides some benefit for all who work and contribute to the scheme, it will not necessarily provide a benefit that gives sufficient income for all to live on. There will always be some in the population whose genetic and human capital or other individual characteristics or circumstances will lead to an insufficient NDC benefit. Continuing with the discussion in the previous section, generally speaking, it is important to emphasize that the sufficiency of the NDC benefit depends on the minimum pension age. Data on national pension systems reveal that people tend to exit the labor force at the minimum possible age, even though this may give them a low lifelong benefit. Part of the essence of the mandatory scheme is protect people from their own bad judgment in this sense. Also, the practice of setting lower pension ages for women, whose exit from the labor force may be timed to coincide with the exit of her (often older) spouse, itself can lead to poverty in old age for surviving women. Finally, it is reasonable to adjust (index) this age, once thoughtfully set, to changing longevity. Given a reasonably set minimum pension age, some form of minimum income floor (for example, a guarantee or flat rate) is unavoidable. For various reasons, there will always be a certain part of the older population that will not have a sufficient contribution-based pension to live on. Therefore an NDC scheme will always need to be supported at the zero level by some form of external transfer paid by general tax revenues. Finally, given some form of minimum income transfer floor for the worst-off in society, a threshold effect will occur at this floor level that may mean that additional contributions to the NDC scheme in old age will not have a counterpart in a benefit, in the sense that they do not bring the contributor above the guarantee level. Remaining in the formal labor force may nevertheless be worthwhile given individual preferences on work and leisure or necessary, because of dire circumstances. In countries where the informal market is an important feature of the economic landscape, people may take their minimum benefit and (continue to) work informally. This could be regarded as an informal counterpart to flexible retirement in a formal market economy. Putting Distributional Policy into NDC It is possible to supplement NDC (and FDC) schemes with noncontributory rights. What is essential, to maintain financial stability, is to finance these as they are granted, since these represent liabilities that need to have an asset counterpart in the financial balance. Since all contributions to the NDC scheme on earnings give rights directly to the contributing participant, there is no room to finance noncontributory rights from individual NDC contributions. Instead, these will have to be paid from other sources of public revenues. The financing of noncontributory rights in the same accounting period in which they are granted keeps order in the system. The alternative would be for the accounting system to record a claim on the state budget (including the internal rate of interest), with the state making partial payments on this debt sometime in the future when it becomes time to honor the commitments. This creates a different payment distributional profile. In the worst case, a future government can decide not to honor these rights.