A Primer on Social Security Systems and Reforms
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
A Primer on Social Security Systems and Reforms Craig P. Aubuchon, Juan C. Conesa, and Carlos Garriga This article reviews the characteristics of different social security systems. Many configurations arise depending on the nature of a system’s funding and determination of benefits. Many reforms propose changing the social security systems. The authors focus their analysis of the transition from a pay-as-you-go to a fully funded system. They argue that the key component of any reform is the treatment of the implicit liabilities of a country’s social security system. The welfare gains accruing to some cohorts as a result of such reforms usually stem from either a partial or complete default on the implicit debt of the system, and in that sense the gains imply only a redistribution of welfare across agents. In contrast, the elimination of existing distortions in social security financ- ing can generate efficiency gains, allowing for welfare improvements for all agents. This result shifts the focus from the nature of the system itself and centers the debate on the distortions associated with social security. (JEL H2, E62, D31) Federal Reserve Bank of St. Louis Review, January/February 2011, 93(1), pp. 19-35. S ocial security,” by its simplest defini- based on individual contributions paid over tion, is a contract between a government time. In such systems, future obligations are and its constituents. Under this contract, fully funded by earlier contributions. citizens provide funding to a social Financial sustainability of the U.S. Social security system, and in exchange they receive Security system is an important policy concern benefits from the system during their nonwork- because of the aging population, particularly the ing years, generally during old age or prolonged baby boom cohorts. The recent recession, com- illness (disability). The U.S. Social Security sys- bined with the renewed political focus on tem was implemented in 1935 under President health care and long-term health costs, has led Franklin Roosevelt. This system, formally known to increased interest in the long-run financial as Old-Age, Survivors, and Disability Insurance stability of the U.S. Social Security system. While (OASDI), is a pay-as-you-go (PAYG) system in the Board of Trustees of OASDI has expressed which workers provide financing through a the need for long-term reform for several years, Social Security tax; these contributions provide their 2009 report described how lower gross benefits to the currently retired or disabled. The domestic product (GDP) and fewer covered system requires an implicit guarantee that future workers affect the long-term outlook (see Board generations will then provide the same support of Trustees, 2009). The Board moved forward its for them. In contrast, fully funded (FF) social projections for the year in which outlays will security systems require that benefits accrue exceed revenues (2016) and the year in which Craig P. Aubuchon was a research analyst at the Federal Reserve Bank of St. Louis. Juan C. Conesa is an associate professor in the department of economics at the Universitat Autònoma de Barcelona, Spain. Carlos Garriga is an economist at the Federal Reserve Bank of St. Louis. The authors thank Ricardo DiCecio, William Gavin, and Christopher Neely for their reviews and comments. © 2011, The Federal Reserve Bank of St. Louis. The views expressed in this article are those of the author(s) and do not necessarily reflect the views of the Federal Reserve System, the Board of Governors, or the regional Federal Reserve Banks. Articles may be reprinted, reproduced, published, distributed, displayed, and transmitted in their entirety if copyright notice, author name(s), and full citation are included. Abstracts, synopses, and other derivative works may be made only with prior written permission of the Federal Reserve Bank of St. Louis. F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W J A N UA RY / F E B R UA RY 2011 19
Aubuchon, Conesa, Garriga current trust funds will be exhausted (2037). The the social security system with future generations. 2010 Trustees Summary Report concludes that An immediate application of the equivalence these imbalances “demonstrate the need for result is that it would be straightforward to engi- timely and effective action. The sooner the solu- neer a Pareto-neutral social security transition. tions are adopted, the more varied and gradual The equivalent FF system produces the same level they can be” (see Social Security and Medicare of welfare, household decisions (i.e., labor supply Boards of Trustees, 2010). These solutions include and consumption), and output. If the reform higher taxes, lower benefits, or a combination of implies only the recognition of the implicit debt both to replenish the trust fund. However, some with no welfare effects, how can this be reconciled analysts advocate a transition to an FF Social with the sizable welfare gains noted in the litera- Security system. ture? In most simulations of reforms, pensions Building on the seminal work of Auerbach and social security contributions are eliminated and Kotlikoff (1987), several quantitative analy- over time in some arbitrary way. Generally, most ses simulate the transition from a PAYG to an FF such exercises imply some partial or complete system and find substantial efficiency and welfare default on promises (equivalent to a default on gains in the long run. However, the gains often the implicit debt of social security), which was come at the expense of the transition generation. the root of the large welfare losses of the transi- For example, Huang, Selahattin, and Sargent tion cohorts. When the government is allowed to (1997) show that partial or full privatization default on the implicit liabilities, the welfare of implies large short-run welfare losses that cannot either existing or future cohorts—but not both— be compensated by the long-run gains. Conesa can be improved. Therefore, the post-reform and Krueger (1999) show that in the presence of welfare gains by some individuals are the result uninsurable labor income uncertainty the welfare of not honoring these liabilities and not of the losses of the initial cohorts are large and constitute a political barrier to potential reforms. In contrast, reform itself. The equivalence result shows that a large empirical literature argues that the macro- when the implicit debt is honored, there is no economic effects of privatizations have been small, room for welfare improvements in the event of particularly with regard to aggregate saving rates.1 a privatization. The objective of this paper is to provide a Because intergenerational redistribution theoretical framework to illustrate the effects of alone cannot generate Pareto improvements in a reforming social security. We use a simple over- dynamically efficient economy, these improve- lapping generations model to demonstrate when ments are possible if and only if there are distor- a transition from a PAYG system to an FF system tions in social security financing or in the tax can (and cannot) be welfare improving for all system. In such circumstances, the presence of households in the economy. The model includes additional distortions, not the reform itself, is the the initial generations alive at the beginning of key to any welfare improvement. Therefore, siz- the reform as well as future generations. We first able welfare gains are possible when substantial show that a PAYG system and an FF system can economic distortions occur, but welfare gains are be equivalent by using a simple government trans- negligible if the social security contributions and fer mechanism that makes explicit the implicit the tax code are close to optimal. These results debt of the PAYG system. This is nothing more might explain why the empirical literature has than the recognition of the implicit liability of found small macroeconomic effects resulting from many reforms in actual economies. Our exercise 1 Coronado (2002) finds that after the 1981 Chilean social security suggests that no macroeconomic and welfare reform (covered in the next section), wealthy families increased aggregate savings by approximately 7 percent. In contrast, Butelman effects should be expected unless distortions and Gallego (2000) find that low-income Chilean households are minimized or eliminated. Successful reform increased their level of debt. Disney, Emmerson, and Smith (2003) and Granville and Mallick (2002) find no effect on aggregate house- requires some substance, not just relabeling of hold saving rates after the 1986 U.K. reforms. government debt. 20 J A N UA RY / F E B R UA RY 2011 F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Aubuchon, Conesa, Garriga Table 1 Configurations of Social Security Systems Type of plan Funded Unfunded Defined benefit Traditional employer pension United States (example: Switzerland) Australia United Kingdom Defined contribution U.S. Roth IRA or 401(k) Chile Latin America Australia United Kingdom Notional defined contribution Italy Sweden OVERVIEW OF SOCIAL SECURITY Policies to Protect the Old and Promote Growth, SYSTEMS and advocated a three-pillar model to social secu- rity. The three pillars have been broadly inter- Not all social security systems are designed preted as the World Bank model and consist of the same. Obviously, different countries have (i) a publicly managed, unfunded, defined-benefit defined their social security contracts according pillar; (ii) a privately managed, funded, defined- to the principles that shape their cultures and contribution pillar; and (iii) a voluntary savings economies. In general, there are four broad types pillar. Orszag and Stiglitz (1999) point out that of social security systems. These systems com- bine two elements. First, a system can either be the World Bank did not explicitly argue for a unfunded (such as a PAYG system) or FF (based privately managed second pillar but that many on accumulated assets). Second, a system can scholars have interpreted it as such.3 This three- provide payments based on either defined benefits pillar approach should sound familiar to most or defined contributions. Note that an FF system Americans: Those who participate in Social is not the same as a privatized system. An FF Security through payroll taxes, save through a system is simply a model for savings and usually 401(k) plan or individual retirement account represents a switch from a defined-benefit to a (IRA), and manage their own private assets engage defined-contribution system.2 Diamond (1996) in the three pillars advocated by the World Bank. states, “I think that the distinction between con- Many workers are inadequately prepared for tribution and benefit base is more illuminating retirement because they do not participate in the than the distinction between privatization and voluntary second and third pillars. According government-run systems, for various pieces of to the 2007 Survey of Consumer Finances (SCF ), either type of system can be privatized” (p. 75). only 60.9 percent of U.S. households 55 to 64 Table 1 summarizes the different configurations years of age have a retirement account outside of social security systems. Social Security and only 41.6 percent of house- In 1994, the World Bank released a compre- hensive review, Averting the Old Age Crisis: holds 35 years of age and younger have a retire- ment account in place (see Federal Reserve Board, 2 Diamond (2004) gives an alternative definition of “defined contri- 3 bution” and notes that the heart of a defined-contribution system For a recent example, see the February 2009 testimony of Alicia is the fact that the risk to future outcomes is on the side of benefits, Munnell and Peter Drucker before the U.S. House of Representa- which are a function of the realized returns on funded contribu- tives; they articulate that a strong second pillar of an FF defined- tions. This is in contrast to the risk of an unfunded system, which contribution plan will help diversify risk and improve retirement generally falls on future taxes. portfolios. F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W J A N UA RY / F E B R UA RY 2011 21
Aubuchon, Conesa, Garriga 2007).4 Across all ages, the SCF finds that 57.7 cal and demographic risk since retirement benefits percent of families have some rights to a defined- are funded and cannot be reduced through taxes. benefit pension or account plan through a current Several other countries, including most of or past employer. Thus, for any combination of Latin America, have a funded defined-contribution reasons, a large majority of families have chosen pillar that follows Chile’s example. Valdés-Prieto not to participate in retirement savings outside (1998) presents a summary of the reforms in Peru, Social Security, even though Social Security is Colombia, Argentina, Bolivia, Mexico, El Salvador, designed to replace only 40 percent of pre- and Uruguay. He offers five reasons why Chile’s retirement income for the average worker retiring model is so successful, including low levels of in 2007 (Social Security Administration, 2008). private-sector corruption, little political pressure The differences across systems (see Table 1) are on investment options, and successful implemen- explained next. tation of a redistributive means-tested benefit to workers not covered by the Administradora de Funded Defined-Contribution System Fondos de Pensiones. Bateman and Piggot (1998) summarize the Current private pensions, such as 401(k) plans pension reforms in Australia and the success of and Roth IRAs, are a type of funded defined- its funded defined-contribution pillar, the super- contribution system. Workers contribute a per- annuation guarantee. The Australian system oper- centage of their salaries during working years, ates under a model similar to Chile’s with the often with a matching employer contribution up exceptions that fund choices for each pension to an established limit. Workers are free to choose are governed by a board of trustees and assets and the investment of their funds and are eligible to allocation are unrestricted. Australia’s system is withdraw their savings during their retirement also unique in that taxes are levied at all three years, with a total sum equal to their defined possible points—contributions, investment earn- contribution plus investment earnings. Chile’s ings, and benefits. Finally, Australia offers the social security reform in 1981 remains the best- choice of either a lump-sum payment or an annu- known international example of an FF defined- ity at retirement. The lump-sum payment is contribution system; Diamond (1996) presents a unique among funded defined-contribution plans; survey of the literature examining the pros and the authors point out that from a societal perspec- cons of this model. Under the Chilean system, tive, the policy is inefficient because individuals all workers are required to contribute 10 percent who spend their lump sums must then rely on of their salary into a savings plan of their choice, the state. which is administered and regulated by the The U.K. system also offers a privatized, Administradora de Fondos de Pensiones. As in funded defined-contribution system but a unique the United States, eligibility for retirement is based one, in that it allows workers to opt out of their on age and early retirement is available to those public, unfunded defined-benefit system. Indeed, with sufficient accumulated savings. At retire- as Johnson (1998) reports, between 1988 and 1992, ment, workers can choose monthly withdrawals the United Kingdom offered an additional 2 per- or purchase an annuity. Furthermore, workers cent “incentive” rebate to workers switching to a are guaranteed a minimum pension paid from the private pension. Thus, an unfunded future benefit general revenue fund. In Chile, the benefits of is replaced with a currently funded contribution. such a system include reduced exposure to politi- Johnson notes that younger workers benefit more from switching and, as expected, have done so 4 The 2007 SCF does not include annuities in this measure and notes in large numbers. However, studies have also that some “families may have used funds from previous employ- shown that many workers have opted out of an ment to purchase an annuity at retirement” (p. A23). The survey also notes that among older age brackets (55 to 64 years), workers can occupational pension program to join the private withdraw funds from some retirement accounts as early as age 59½. pension, in essence giving up any company con- 22 J A N UA RY / F E B R UA RY 2011 F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Aubuchon, Conesa, Garriga tribution to their plans. Thus, the issue of choice benefits only if they are below an income thresh- has been effective for most, but some workers old or as a universal benefit given to all workers, have made second-best choices. often calculated as a percentage of the earnings average over a set number of working years. Funded Defined-Benefit System Benefits are also linked to either wage growth or price growth so that benefits stay roughly in line More traditional pensions, similar to those awarded to older U.S. workers during previous with the cost of living. The U.S. Social Security decades, are good examples of funded defined- system is generally considered an unfunded benefit systems. Workers pay into the pension defined-benefit system. Workers pay into the system, and the corporation manages how these system through a tax, which is then transferred contributions are invested. Workers then receive to the current retired generation in the form of a a defined benefit at retirement, which is usually defined benefit. In the United States, the universal based on years of service or some other related benefit is based on a worker’s average earnings measure. over a 35-year period, up to a certain income level. Switzerland currently offers a hybrid system: Furthermore, under the U.S. system, benefits a funded defined-contribution system with a are also provided to spouses and dependents. guaranteed return. The burden of the Swiss com- Diamond (2004) asserts that the unfunded nature pulsory occupational pension scheme (overseen of U.S. Social Security makes sense given the by a trustee board) is placed directly on individ- early decision to redistribute wealth and provide ual corporations. The board chooses the pension full benefits to the initial generation—which did insurance and the amount and is responsible for not pay into the system—because the incomplete- enrolling workers. Workers contribute 17 percent funding risk is shared across future cohorts. An of their salary, half of which is matched by the upcoming section outlines the model for a PAYG corporation. Under these requirements, the system system that generates this type of redistribution. is purely a defined-contribution plan. However, Unfunded defined-benefit plans are generally each pension plan is required to return a minimum useful for a redistribution of wealth—that is, to of 4 percent nominal interest each year.5 If a help guarantee a minimum level of income for pension plan is ever underfunded, the firm or any worker who participates in the labor force corporation must make up the difference. This for an agreed-upon number of years. Unfunded legal, and explicit, guarantee on returns makes defined-benefit plans often favor workers with the Swiss system more of a defined-benefit plan lower incomes or with noncontinuous work his- since workers know with some certainty the tories. Often, the benefit is set at a basic subsis- value of their future annuity. tence level and is intended to be supplemented with other retirement savings. For example, Unfunded Defined-Benefit System Switzerland provides a guaranteed minimum pen- A publicly operated, unfunded defined-benefit sion for its entire population, with the pension plan constitutes the first pillar of social security paid from general revenues. Hence, the minimum among most countries. These systems are often pension is independent of a worker’s salary or described as PAYG because current workers pay time in the labor force. This universal benefit taxes to provide a benefit to the current retired redistributes income to poorer workers during generation. An individual’s benefits are offered their retirement years. either as means-tested, such that a worker receives Australia and the United Kingdom, among other countries, have a public, unfunded, defined- 5 The 4 percent rate was chosen to be slightly below the long-run benefit system as a first pillar for social security. return on Swiss government bonds of 4.5 percent. Hepp (1998) In particular, the United Kingdom has a two-tier summarizes, “The return guarantee was deemed unambitious enough to avoid frequent funding shortfalls...but explicit enough state pension scheme. First, workers are provided to enhance the credibility of the system” (p. 536). a basic state pension, paid at a single flat rate that F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W J A N UA RY / F E B R UA RY 2011 23
Aubuchon, Conesa, Garriga offers a subsistence-level income.6 Most workers the amount of contributions made and return on are covered by occupational pension schemes, the invested capital. The income pension system which, in one form or another, predate the state is also a so-called PAYG system, which means scheme; for workers not covered by the occupa- that the pension contributions paid in every tional program, the state also offers the State month are used to pay current income pension Earning Related Pension Scheme (SERPS). A benefits to those who have already retired.”7 unique feature of SERPS is that workers can opt However, because of the notional interest rate, out of the plan and the associated National Sundén (1998) states that the Swedish plan is Insurance contribution if their private pension “more similar to a defined benefit plan...since the government has to cover its pension liability or occupational scheme provides a guaranteed through annual contributions” (p. 582). minimum pension equivalent to SERPS. In con- trast, Australia offers a means-tested benefit, such that the full-rate pension is equal to 25 per- cent of male average earnings (40 percent for CURRENT STATE OF U.S. SOCIAL couples), but this payment is phased out as retire- SECURITY ment income and other assets accumulated under The unfunded defined-benefit plan of the the other two pillars increase. Similarly, Sweden United States, known as OASDI, contains two offers a means-tested pension for workers with separate parts. The first, and the focus of this no or low income. This guaranteed pension is paper, is the Old-Age and Survivors Insurance financed through general revenue taxes and is (OASI), which pays monthly benefits to retired independent of workers’ notional defined- workers and their families. The second compo- contribution (NDC) plans. nent, Disability Insurance (DI), pays monthly benefits to disabled workers and their families. Unfunded Defined-Contribution System The 2009 annual report of the Board of Trustees presents the most-current picture of Social Sweden and Italy are concrete examples of Security in the United States and notes that, in countries with an unfunded defined-contribution 2008, almost 35 million retired workers and their social security system. In recent years, both coun- dependents and another 6 million survivors of tries have switched to an NDC plan. The govern- deceased workers received benefits. An estimated ment credits each worker for the taxes he or she 162 million workers paid Social Security and and the employer contribute, and then pays a payroll taxes. Total benefits paid in 2008 were benefit equal to the worker’s contributions plus $615 billion, and the Social Security Trust Fund a notional interest rate. Första AP-fonden (AP1, collected $805 billion, prompting the Trustees to one of the pension funds managing the Swedish note that “the combined OASI and DI Trust Funds system) describes the Swedish income pension are adequately financed over the next ten years” system as follows: “The income pension system (see Board of Trustees, 2009, p. 2). However, the is of the defined contribution type, meaning that Trustees also state very clearly: “The financial the size of future pension benefits depends on condition of the Social Security and Medicare programs remains challenging” and that the cur- 6 Johnson and Stears (1996) point out that the basic state pension is rent PAYG system “does not satisfy the short range technically a defined-contribution scheme, since the pension is paid only to workers who contribute taxes during 90 percent of their test of financial adequacy.” working life and have Class 1 contributions on earnings of at least These long-run problems arise as the baby 52 times the weekly lower earnings limit during each working year. Thus, to qualify for the pension at age 65, a worker must have boom generation begins to retire and reduces the contributed for 44 years (male) or 39 years (female). However, the number of covered workers per beneficiary from authors note that “there are so many special provisions that virtu- ally everyone becomes eligible, no matter what their employment 7 and contribution history...” (p. 1111). These exceptions include a More information on the Swedish system is available at the Första deduction on working years for home responsibility protection, AP-fonden website which helps guarantee coverage for stay-at-home caregivers. (www.ap1.se/en/Our-mission/The-Swedish-pension-system/). 24 J A N UA RY / F E B R UA RY 2011 F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Aubuchon, Conesa, Garriga a historical average of three workers per benefi- a decrease in the number of hours worked (say, ciary to just two workers per beneficiary. As is from an individual declining overtime hours) is well documented elsewhere, the Board of Trustees the deadweight loss of the tax. Feldstein and (2009) currently predicts that income received Liebman (2002) note that under standard theory, will exceed benefits payments in 2016. Because the deadweight loss of a tax system increases with of the surplus accrued and interest generated on the square of the marginal tax rate. Over time, the these savings, the payment of benefits will not increased deadweight loss makes continued tax be reduced until 2037, when the surplus is pre- increases to fund a demographic shift less desir- dicted to run out. At this point, the Trustees esti- able since each subsequent tax increase results mate that under their intermediate assumptions, in a larger deadweight loss. payable benefits will be 76 percent of scheduled In the early 1980s, the United States faced a benefits and by 2038, tax income will cover 74 similar Social Security dilemma and pursued a percent of scheduled benefits.8 strategy of tax increases and delayed benefits in The Trustees conclude that for the trust fund the form of an increase in the normal retirement to remain solvent throughout the 75-year projec- age (NRA). The Social Security Amendments of tion period and pay scheduled benefits at 100 1983 extended the NRA from age 65 to age 67 percent, one of three things must happen under for the cohort of workers that turns 62 in 2022. business as usual.9 First, the combined payroll tax Diamond (1996) provides a summary of the could be immediately and permanently increased changes, noting that the law did not change the from 12.4 percent to 14.4 percent. Second, sched- minimum age (62 years) to claim retirement bene- uled benefits could be reduced by an amount fits, nor did it extend the age to obtain benefits equal to an immediate and permanent reduction independent of earnings (70 years). Rather, it of 13 percent. Or third, a general revenue transfer simply changed the level of benefits as a function equivalent to $5.3 trillion in present value terms of the age at which they are first claimed. With could be made to the trust fund. an NRA of 65, workers can receive 80 percent of These measures all imply a welfare loss of their benefits starting at the minimum age, 62. some type for workers and retirees during the 75- Under the new NRA of 67, workers receive only year projection period. An important economic 70 percent of their scheduled benefit starting at consideration to the Trustees’ conclusions is the age 62. Thus, extending the NRA by two years is deadweight loss associated with an increase in the equivalent of cutting benefits by one-eighth. the payroll tax. With a higher tax rate the govern- Diamond also notes that extending the NRA might ment would collect more revenue unless the have unintended consequences since there higher taxes distort an individual’s decision on were no corresponding benefit cuts for early how much to work. The revenue lost because of withdrawals for DI. This provides an incentive 8 to apply for DI benefits at the earliest date. The Under intermediate assumptions, the Board estimates a total fertil- ity rate of 2 children per woman, an annual percentage change in loss from future income of working years and productivity for the total U.S. economy of 1.7 percent, an unem- revenue savings from the reduced benefit repre- ployment rate of 5.5 percent, and an inflation rate (measured by the consumer price index) of 2.8 percent. Jeske (2003) notes that slight sents another source of potential welfare loss. changes to the Board’s assumptions can lead to drastic changes in Gramlich (1998) provides a brief overview of the long-term forecast of Social Security. It is important to note that these changes run both ways: With slightly higher growth the recommendations from the 1994-96 Social Social Security will face no funding problems, but with slightly Security Advisory Council. This group, which lower growth Social Security will be even less likely to meet existing obligations. Jeske (2003) concludes that a PAYG system Gramlich chaired, offered three options to address “therefore implies a substantial amount of risk, contrary to the the long-run actuarial soundness of Social amount that proponents of social security would admit” (p. 16). Security.10 Options included 9 The Trustees also note the impact of the 2007-09 recession on the Social Security system and the “business as usual” scenario, includ- 10 ing raising estimates for the projected deficit and modeling lower See Pecchenino and Pollard (1998) for a formal theoretical treat- GDP growth in the upcoming years. ment of the three proposals put forth by Gramlich (1998). F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W J A N UA RY / F E B R UA RY 2011 25
Aubuchon, Conesa, Garriga • a Maintenance of Benefits plan, with mini- ity to compensate the initial generation, Kotlikoff mal changes in benefit schedules or tax finds long-run efficiency and welfare gains above rates but a large portion of trust fund assets the baseline scenario but below the uncompen- invested in equities, with the goal of a sated transition. Kotlikoff concludes: “[T]he higher rate of return to restore actuarial extent to which privatization results in efficiency balance; gains depends on the ability of future generations • a Publicly Held Individual Accounts (IAs) to compensate current workers for the loss of con- plan to replace the defined-benefit system sumption as a result of financing the transition” with a large-scale defined-contribution (p. 37). Conesa and Krueger (1999) propose an system, with OASDI as a weaker first pillar environment augmented to include uninsurable that provides a poverty-line flat benefit; labor income risk, hence giving a PAYG social • a Two-Tiered System with Privately Held security system an additional role as a partial Individual Accounts plan, which Gramlich insurance device, and show that the transition termed the “kind and gentle” benefit cut offers similar conclusions. plan. With IAs, high-wage workers would Birkeland and Prescott (2007) consider an experience slight benefit cuts and workers overlapping generations model with no popula- would be required to contribute to cen- tion growth in both a PAYG system and an FF trally managed investment accounts that system. They compare the models under four convert to real annuities upon retirement. demographic scenarios calibrated to match the current United States and a future United States The next subsection reviews some of the theo- with lower population growth and longer retire- retical contributions addressing the issue of a ments. They note that the PAYG system has little transition from a PAYG to an FF system, hence or no explicit debt but does not fully maximize providing a quantitative evaluation of many of welfare for the society. Early in their paper, the the reforms discussed so far. authors challenge the notion of debt, stating “The government debt that a country owes to its citizens Effects of a Transition from a is not debt in the usual sense…[it] is a mechanism Pay-as-You-Go to a Fully Funded System that facilitates intergenerational borrowing and The differences between a PAYG and an FF lending, and is an integral part of a welfare- social security system have been studied exten- improving saving-for-retirement system” (p. 2). sively in the economic literature. Here we review Birkeland and Prescott (2007) also note that several recent papers that study the welfare the Congressional Budget Office estimates that implications of the transition. For a more com- the current implicit guarantees of the PAYG system prehensive and nuanced review of the existing represent liabilities four times gross national literature on social security reform, see Feldstein income. These implicit guarantees represent a and Liebman (2002) and Diamond (2004). form of debt for the unborn generation, which Kotlikoff (1998) considers intergenerational lowers lifetime welfare. The authors find that welfare and efficiency in U.S. Social Security with current U.S. demographic assumptions, reform and advocates a consumption tax to finance welfare under an FF system is 9.2 percent higher the transition from a PAYG system to a defined- than a PAYG system. The FF system has a higher contribution personal security system. Under his explicit debt-to-gross national income ratio, and model, an uncompensated welfare transition individuals work more under the FF system results in significant increases to capital stock because workers have a higher take-home wage (36.7 percent), aggregate labor supply (3.7 percent), without a Social Security tax.11 Lifetime con- output (11.2 percent), and real wages (7.1 percent) 11 compared with the baseline model. However, this Their model considers a population growth rate of 1 percent, an NRA of age 65, and a 20-year retirement. Their future U.S. model scenario leads to short-run decreases in aggregate considers a country with no population growth, an NRA of age welfare. Using a lump-sum redistribution author- 65, and a 30-year retirement. 26 J A N UA RY / F E B R UA RY 2011 F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Aubuchon, Conesa, Garriga sumption also remains higher for a future United social security contract between a government States with an FF system, by 5.5 percent. While and its constituents. Our goal is to show the basic the authors find that aggregate welfare is better mechanisms by which a shift to an FF system can, than under an FF system and note that an increase in some cases, be welfare improving, even for in explicit government debt is not a burden to the transition generation. As Orszag and Stiglitz future generations, they do not consider the wel- (1999) point out, initial conditions matter and as fare of the transition generation. such, it is an “issue of whether a shift to individ- Similarly, Jeske (2003) considers the transition ual accounts would be socially beneficial” and between social security systems and also finds not an issue of whether or not “in a tabula rasa that in the long run every generation benefits sense, an individual account system would have more from an FF system. In that system, individ- been preferable to a public defined benefit system ual savings are higher, which in turn increase in the first place” (p. 5). the aggregate capital stock. The higher capital stock acts as a buffer to the economy in the event of large aggregate economic shocks, which disrupt A MODEL ECONOMY WITH A the PAYG social security system. Jeske (2003) PAY-AS-YOU-GO SYSTEM argues that “private savings are more desirable and affordable if both benefits and contributions We follow the framework of Samuelson (1958, [to PAYG] are lower” (p. 16). As do Birkeland and 1975) and Diamond (1965) and consider the sim- Prescott (2007), Jeske similarly acknowledges that plest scenario of a two-period overlapping gener- social security reform has beneficial long-term ations model. Individuals work during the first effects but that, in the short run, “a large portion period of their life and are retired during the sec- of the population will be worse off...[T]he prob- ond period. Consider the problem faced by an lem of privatization is the unfunded liability to individual born in period t (who will be retired pay for current retirees” (p. 22). in period t +1). During the working period, indi- Conesa and Garriga (2008) study the optimal viduals provide labor, denoted as lt , and are paid financing of the transition from a PAYG to an FF in return a wage rate per unit of labor, wt . They system. By maximizing over the entire policy consume some goods in the first period, c1,t ; they space and following an optimal fiscal policy pay social security taxes on their wage income, approach, the authors find it is possible to finance denoted by the tax rate, τt ; and they can save for such a transition in a Pareto-improving manner the next period, at +1, where savings today are for all generations. In their model, the fiscal author- assets that pay interest in the next period. Hence ity changes the labor income tax over time: first by the budget constraint for these individuals is lowering the labor income tax during the transi- tion generation, issuing government debt to fund (1) c1,t + at +1 ≤ w t t − τ t w t t . existing obligations, and then raising taxes over During the retirement period—which is the next time to repay debt. Measured as equivalent vari- period, t +1—individuals do not produce any ation in consumption, the authors find that in the goods; instead, they merely consume the principal transition from a PAYF to an FF system future new- and interest on their private savings and their born generations experience a welfare increase pension payments. Therefore, their budget con- between 3 percent and 8 percent. Such a scheme straint is given by allows for welfare gains for both actual and future generations, and the key aspect is the reduction (2) c2,t +1 ≤ (1 + rt +1 )at +1 + pt +1, of the distortions introduced by the tax system. None of the above-mentioned papers consid- where c2,t +1 denotes the consumption during ers the political ramifications of social security retirement (in period t +1) of the individuals born reform, nor do they address the social justice in period t, rt +1 denotes the interest payments issues underlying the need for and extent of a collected on savings, and pt +1 denotes the social F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W J A N UA RY / F E B R UA RY 2011 27
Aubuchon, Conesa, Garriga security payments collected by the retired in program.13 In the model, workers save throughout period t +1. their lifetimes. Whether this savings program is If there is no exogenous restriction on the mandatory is irrelevant, since workers could also sign and magnitude of savings,12 both budget save (or borrow) privately if they chose to do so. constraints can be combined as Hence, given an optimal consumption allocation if the government increases compulsory savings, c2,t +1 p individuals would respond by decreasing their (3) c1,t + ≤ (1 − τ t )w t t + t +1 . 1 + rt +1 1 + rt +1 private savings by the same amount. This expression simply states that the net present This can be easily seen by consolidating the value of consumption over the life cycle of indi- budget constraints as follows: viduals cannot be larger than the net present value (5) c1,t + at +1 + atm+1 ≤ w t t , of after-tax payments (labor income when indi- viduals are young and pensions when old). Given where atm+1 = τtwt lt denotes mandatory contribu- this constraint, individuals would choose 共c1,t , tions to an IRA, computed as a fraction of current c2,t+1,lt 兲 given 共τt,wt ,pt+1,rt+1兲 and their preferences labor income. These contributions are then capi- (i.e., how much they value consumption in the talized at the market rate of return and constitute present with respect to consumption in the future the funding of the future retirement pension. and how much they value consumption today Hence, next period’s budget constraint will with respect to how much they dislike working). include a pension, denoted Finally, in this world the government operates (6) c2,t +1 ≤ (1 + rt +1 )at +1 + pt +1, the social security system in a standard PAYG fashion: Social security contributions of the cur- where now pt+1 = 共1+rt+1兲atm+1 and therefore the rent working-age population finance the pension pensions are funded by defined contributions. payments of the currently retired population. In Notice, though, that the net present value particular, if there are 共1+n兲 workers per retired budget constraint is person (think of n as a constant growth rate of c2,t +1 p population), the social security system would be (7) c1,t + ≤ (1 − τ t )w t t + t +1 = w t t . 1 + rt +1 1 + rt +1 balanced when In other words, whether savings are compulsory (4) (1 + n)τ tw t t = pt . does not matter; in the end, the net present value Notice it is an unfunded system in the following of consumption is independent of the level of sense. The individuals born in period t will compulsory contributions to social security. Effectively, then, in this simple model environ- contribute to the system τtwt lt . However, when ment an FF system is equivalent to private savings. they retire they will collect pensions pt+1 = 共1+n兲τt+1wt+1lt+1 that are not related to their own past contributions. COMPARISON OF THE TWO SOCIAL SECURITY SYSTEMS A FULLY FUNDED SYSTEM We now compare the two social security sys- An alternative to the implicit guarantee of tems. For simplicity, we consider a stationary the PAYG structure is an FF system of social secu- 13 It is important to make the distinction that it is not a necessary rity. Under our previous notation, an FF system condition for FF system funds to be invested in equities. As Orszag is a defined-contribution plan similar to a 401(k) and Stiglitz (1999) state, “...prefunding and privatization are dis- tinct concepts, and conflating them confuses rather than informs the debate” (p. 9). Indeed, an FF system could be fully invested in 12 The presence, for example, of borrowing constraints might com- government securities, with low risk and lower (but sometimes plicate the analysis. guaranteed) returns. 28 J A N UA RY / F E B R UA RY 2011 F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Aubuchon, Conesa, Garriga world where social security contributions are con- in turn receive a pension benefit when they retire stant over time, τt+1 = τt ; wages grow at some during the next period. Their pension benefit will constant rate g—that is, wt+1 = 共1+g兲wt ; and hours then be paid by the young generation of the next worked by each generation of workers are also period, and so on. constant, lt+1 = lt . Consider we are now in period T. Imagine Under this scenario, the PAYG social security that the government decides to switch immedi- pensions are given by ately to a funded system during this period without honoring the implicit debt of the current (8) pt +1 = (1 + n )τ t +1w t +1 t +1 = (1 + n ) (1 + g )τ t w t t retirees (the workers of the previous period), that and therefore the return of a PAYG social secu- is, pT = 0. The current old generation would incur rity system is equal to 共1+n兲共1+g兲. Clearly, if a welfare loss equal to the sum of the pension 共1+rt+1兲 > 共1+n兲共1+g兲, then each individual born obligations to the current retirees. If private sav- in this stationary world will benefit more from a ings earn a higher return than the implicit return funded system, since the right-hand side of of the social security system, then the generation equation (7) is larger than the right-hand side of working in period T is better off (as are all subse- equation (3). quent generations). However, the contributions Usually both empirical data and economic of the current workers are invested to finance their theory tend to confirm that, on average, the return own future pensions, aT+1m = τt wt lt . Effectively, on private investment is larger than the growth the government has defaulted on its obligations rate of a stationary economy, 共1+n兲共1+g兲. Notice to the currently retired. With this transition that a systematic violation of this condition would scheme the initial cohorts of retired individuals imply that an economy is inefficiently overaccu- (or those close to retirement) bear the cost at the mulating capital (see Samuelson, 1975). expense of current young and future generations. Clearly, if the return on private investment is larger than the growth rate of the economy, one Transition II: Default on Future would fare better born in a world with an FF sys- Generations tem than a PAYG system. Nevertheless, as previ- ously discussed, this is the correct answer to the Consider now that we decide to keep the wrong question. The relevant issue relates to the promise to the currently retired, and as such, we following question: Given that the current world still need to pay their pensions. We finance these has a PAYG social security system, is anything pensions by issuing debt that must be repaid gained by switching to an FF system? Answering during the next period. The government budget this question requires consideration of the events constraint is now BT +1 = pT and the currently that occur during the transition, which in our young are the only ones who absorb the new debt theoretical model exists for just one period. (in a closed economy) and their budget constraint becomes TRANSITION BETWEEN SOCIAL BT +1 (9) c1,T + aT +1 + ≤ w T T . SECURITY SYSTEMS 1+ n Notice than per capita debt of the young must be Transition I: Default on the Currently Retired (10) BT +1 p = T = (1 + n)τ T wT T = τ w . T T T 1+ n 1+ n 1+ n Let us consider the scenario in which our economy is operating under a PAYG structure. Clearly, the per capita debt of the young now is Workers in this period contribute to the social equal to the social security contributions under security fund by paying a social security tax, the original PAYG system. In the next period, the under the assumed social contract that they will budget constraint of the retired will be F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W J A N UA RY / F E B R UA RY 2011 29
Aubuchon, Conesa, Garriga BT +1 In the PAYG regime, total taxes net of discounted c2,T +1 ≤ (1 + rT +1 ) sT +1 + (1 + rT +1 ) pensions paid by households are (11) 1 +n = (1 + rT +1 ) sT +1 + (1 + rT +1 )τ T wT T . p p p (17) τ w − = − , 1 + r (1 + n ) (1 + g ) 1 + r Since the rate of return on private savings (or in government debt) is larger, the transition cohort where the equality comes from using expression will benefit more than it would with the PAYG (8). social security system. Simple algebra shows that expression (16) is However, future generations will experience larger than expression (17) and, as such, all future a welfare loss, which can be seen by examining generations now fare worse than under the origi- the budget constraint of the workers born in nal PAYG system. period T +1: Welfare-Neutral Transition: No-Default B Case (12) c1,T +1 + aT +2 + T +2 ≤ w T +1T +1 − tT +1 . 1+ n We now consider an economy that honors These households will absorb the outstanding the implicit debt of the PAYG system but does not debt and will have to pay taxes, denoted by tT +1. benefit one generation at the expense of others, The reason can be seen from the government as was the case in the previous two examples. budget constraint: We follow the approach of Conesa and Garriga (2008) and allow the government to issue recog- (13) BT +2 + TT +1 = (1 + rT +1 ) BT +1, nition bonds equal to the value of the govern- ment’s implicit pension obligations to current where total taxes collected equal the taxes per workers.14 worker multiplied by the number of workers; The no-default plan could proceed as follows: that is, TT +1 = 共1+n兲tT +1. Notice that in the absence Since current workers are still paying their social of this new tax, the outstanding debt would security contributions (to honor the benefits of explode to infinity, creating a Ponzi scheme that the current retired generation), the government cannot be in equilibrium. will issue these workers a direct monetary trans- To maintain a constant level of debt, BT +2 = fer financed by government debt (equivalently, BT +1, the new tax should be enough to cover the recognition bonds could be issued) equal to the interest payments on the initial debt issued: net present value of their (lost) pension in the next period. The transfer received by the current TT +1 = rT +1 BT +1 workers is then equal to pT +1/共1 + rT +1兲. By con- (14) (1 + n) tT +1 = rT +1pT . struction, current workers are indifferent between this arrangement and the previous PAYG system. Hence households now must pay more in net The budget constraints of the transition gen- present value to the tax authority than under the eration are now defined as PAYG framework: (18) c1,T + aT +1 + bT +1 ≤ w T T − τ T wT T + trT (15) c2,T +2 NOW: c1,T +1 + ≤ w T +1 T +1 − tT +1 (19) c2,T +1 ≤ (1 + rT +1 )aT +1 + (1 + rT +1 ) bT +1 . 1 + rT +2 c p PAYG: c1,T +1 + 2,T +2 ≤ wT +1T +1 − τ T +1wT +1T +1 + T + 2 . Notice, however, that these are the same budget 1 + rT +2 1 + rT +2 constraints as in the PAYG model, once we under- Notice that taxes paid now equal the interest 14 payments on the government debt initially issued: This idea is not restricted to the particular strategy explored in Conesa and Garriga (2008); the term “recognition bond” has been rp used elsewhere, including by Feldstein and Liebman (2002). (16) t = . Diamond (1996) also highlights this example from the experience 1+ n of Chile in 1981. 30 J A N UA RY / F E B R UA RY 2011 F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
Aubuchon, Conesa, Garriga stand that (i) the transfer collected by workers is acknowledges that a surplus is sufficient, but not equal to debt issued trT = bT +1 and (ii) principal a necessary condition, to reform. During a review and interest on the debt (or the recognition bonds) of the reforms in several Latin American countries are equal to the pensions 共1 + rT +1兲bT +1 = pT +1. in the early and mid-1990s, he finds that even In the next period, the total amount of the countries emerging from hyperinflation have recognition bonds (or government bonds to pay successfully managed a transition by issuing debt for the transfer) must be paid with interest. By after the inflationary period. construction, though, this quantity is exactly equal to the amount of the pensions, so the new Welfare-Improving Transition: generation of workers born in period T +1 must No-Default and Lower Labor Distortions pay taxes and again be compensated by a trans- fer in the exact manner as the previous generation, We now use the intuition and fiscal policy and so on until infinity. approach from Conesa and Garriga (2008, 2009) The introduction of recognition bonds does to present the case in which a switch to an FF not increase the level of debt for our model’s gov- defined-contribution model can be welfare ernment; rather it makes the debt explicit. Pakko improving. The existing literature has shown (2009) provides a brief overview of the U.S. fed- that in a dynamically efficient economy, it is not eral deficit and explains the distinction between possible to raise aggregate welfare by redistribut- “on-budget” and “off-budget” items. Social ing resources across generations (a result that goes Security is an off-budget item and is reported back to Diamond, 1965), which is basically our approach in the previous reform scenarios. The only as part of the unified budget. Currently, the first two scenarios were situations in which one Social Security Trust Fund in the United States cohort might benefit at the expense of others; is counted as a surplus since tax receipts are larger the last one was a Pareto-neutral privatization than benefit payments. This surplus appears in (nobody won or lost; we simply made the implicit reported figures of the combined budget. For debt explicit). However, if the economy is ineffi- example, in 2008 the official measure reported cient because of distortions, then we can increase by the government—the unified budget deficit— aggregate welfare by removing the distortions. was $455 billion. The on-budget deficit was Furthermore, we have seen that in the case $638 billion, with an off-budget surplus of $183 of a decline in the labor force (or a corresponding billion, funded primarily from Social Security. increase in the dependency ratio), the two policy Pakko (2009) questions which deficit measure— options are increasing the payroll tax or cutting on-budget or unified budget—the government benefits. Increasing the payroll tax worsens the will report starting in 2017 when Social Security distortion on labor. outlays exceed revenues. The implicit debt guar- In the baseline scenario the distortion in the antee to future generations is currently not economy comes from the tax on labor, which is reported, even though it is politically unlikely used to finance the PAYG system. Pensions are that the federal government will default on these viewed as a pure transfer, while contributions future obligations. are viewed as a pure tax. Notice that actual pen- Chile approached a similar transition in sion systems do have some connection between 1981 by incurring no debt; instead, they began labor income and pension entitlements so that, in building a fiscal surplus three years before the reality, individuals may realize the link between reform started. Chilean GDP grew at an average their individual contributions and their pension of 8 percent per year during this period, and entitlements, thus reducing the distortion. How- the high growth fueled increased tax revenue. ever, insofar as the connection between contri- Diamond (1996) states that “it may be that a sur- butions and pensions is not one to one (because plus is a contributing condition for a successful of redistributive considerations usually present privatization” (p. 80). Valdés-Prieto (1998) in most systems), there will still be a distortion. F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W J A N UA RY / F E B R UA RY 2011 31
Aubuchon, Conesa, Garriga Another way to view the distortion is by social security entitlements) financed with debt looking at the equivalent economy in which the and lowering labor income taxes on impact to implicit liabilities of the PAYG system have been increase them later. made explicit. From equation (18), we see that The introduction of capital income taxes in workers are paying taxes as a function of their the analysis allows for the generation of additional labor income, τTwT lT, and at the same time they welfare gains since it drastically reduces the need are receiving a compensatory transfer, trT, inde- for compensatory transfers for the initial genera- pendent of their labor supply decisions. Basic tions alive. On average, capital income taxes economic principles imply a deadweight loss translate into very large subsidies, especially for because of this scheme. Moreover, this distortion the oldest cohorts. Effectively, changing the fiscal would increase in the labor supply elasticity. treatment of capital income can become a close Reducing both the tax and the transfer would substitute for compensatory lump-sum transfers result in an efficiency gain. to the initial old generation. Most countries have additional distortions that are important in studying social security reforms. For example, mandatory retirement rules CONCLUSION AND POLICY could be eliminated during a reform, as suggested IMPLICATIONS by Conesa and Garriga (2003). Without this restric- tion, the transition to an FF system requires a We have examined some of the “myths” sur- lower level of compensating transfers and ensures rounding social security as presented by Orsagz a faster convergence to the new steady state. In and Stiglitz (1999), particularly by focusing on addition, the government can change the tax treat- comparing the transition between systems as ment of capital income of retirees as an alterna- opposed to tabula rasa comparisons between a tive compensation mechanism. In considering PAYG and an FF system. We have presented the these different distortions, the relevant set of findings of other researchers who have docu- budget constraints becomes mented the welfare gain under an FF defined- contribution system without considering the (20) c1,T + aT +1 + bT +1 ≤ w T T − τ 1,T wT T + trT welfare cost for the transition generation. We build on the notion that a PAYG social security (21) ( ) ( ) c2,T +1 ≤ 1 − τ 2,T +1 w T + 1T +1 1 + rT +1 (1 − θT +1 ) aT +1 system is just an implicit liability for the tax + (1 + rT +1 ) bT +1 . authority, and as such it could be converted into an explicit liability (i.e., government debt) with- In addition to the implicit debt of the social out cost. After such conversion the government security system, a larger set of distortions has can focus on designing reforms without inevitably been made explicit. Distortions on labor income generating welfare losses for some generations. of the young are now denoted by τ1,T , while the The key insight is that this scenario is possible distortion on labor income of the old is τ2,T+1 only if the distortions (introduced either by the (with compulsory retirement this is equal to 100 financing of social security or other types) are percent), and θT+1 denotes distortions on invest- reduced. ment decisions. Hence, the focus should be shifted from the Given this scenario, it is possible for a govern- nature of the social security system itself and the ment to reduce the distortions and generate wel- debate centered instead on the distortions intro- fare improvements, implying a lower level of duced by all tax-transfer schemes currently pres- compensatory transfers or a lower level of recog- ent in the economy. The debate on social security nition bonds. Such a strategy is illustrated in reform then becomes a debate on how to allocate Conesa and Garriga (2008, 2009). They show that to different cohorts the efficiency gains generated the optimal social security reform consists of by the reduction of these distortions. providing compensatory transfers to the initial old generation (transfers almost as large as their 32 J A N UA RY / F E B R UA RY 2011 F E D E R A L R E S E R V E B A N K O F S T. LO U I S R E V I E W
You can also read