Reforming Social Security With Progressive Personal Accounts

1. Reforming Social Security With Progressive Personal Accounts John Geanakoplos Yale University Stephen P. ZeldesGraduate School of Business, Columbia University and NBER NBER Retirement Research Conference October 20, 2006 Woodstock, VT

2. President Bush has strongly advocated replacing part of Social Security with personal accounts • Massive effort in 2005, including 60-day, 60-city Presidential tour of US to promote this reform • Died in Congress due to strong Democratic opposition and some Republican hesitation • Funding for individual accounts nevertheless included in President’s 2007 budget request

3. Democrats versus Republicans • Democrats committed to keeping DB structure • Republicans committed to shifting to DC • Why? Which features of these structures are important to each side?

4. Core Goals • Democrats • Redistribution based on lifetime earnings • Risk-sharing across generations • Security via inflation indexed life-annuity • Republicans • Property rights via private ownership • Transparency regarding accrual of benefits • Market prices (facilitating financial planning) • Equity-like returns • Portfolio choice • Are these reconcilable? • If so, does that help balance the system?

5. Yes! Maybe? • Progressive Personal Accounts • Progressive via variable matching of contributions • Accounts hold new financial securities (PAAWs) that provide DB-like benefits but are priced. • Pricing PAAWs allows us to value incremental benefits and charge accordingly.

6. Outline (including work in progress) • Create individual account (DC) system with same payouts as current (DB) system • Create marketable pools of PAAWs • Calculate price of pooled PAAWs • Compute match / tax • Describe transition to progressive personal accounts • Use PAAW prices to design rules for self-balancing system

7. Related Literature • Implicit marginal tax rate / matching in current system • Feldstein and Samwick (1992) • Cushing (2005) • Feldstein and Liebman (2002) • Accrual of Social Security Benefits • Geanakoplos, Mitchell, and Zeldes (1999) • Jackson (2004) • Self-balancing / Notional DC systems • Valdes-Prieto (2000) • Borsch-Supan (2005) • Auerbach and Lee (2006) • New financial securities • Shiller (1993) (GDP bonds) • Bohn (2002) and Goetzmann (2005) (wage bonds) • Blake and Burrows (longevity bonds) • Valdes-Prieto (2005) pay as you go securities

8. Mechanics of Current OASDI SystemContributions / Taxes Contributions = tax rate xcovered earnings Covered earnings = min (earnings, cap) (2006 cap $94,200)

9. Mechanics of Current OASDI SystemCalculation of Benefits • Calculate relative earnings (ratio of individual covered earnings to average economy-wide earnings) • Average worker’s relative earnings across highest 35 years • Compute PIA (in average wage units) as concave function of average relative earnings • Benefit in first year = PIA (in average wage units) x average wage index Benefit in future years (each remaining year of life) indexed to CPI

10. Mechanics of Current OASDI SystemCalculation of Primary Insurance Amount (PIA) Concavity generates redistribution and individual risk sharingTying benefits to average wages generates aggregate risk sharing

11. Current System Provides Insurance Against • low lifetime earnings (35 year average) • longevity (annuity: inflation indexed payment for life) • falling behind wages of next generation • low wages of all current workers (because still get same retirement benefits)

12. Creating Individual Accounts that Mimic the Current System • Define new securities (PAAWs) • Compute accrual of balances • Trade PAAWs in financial markets

13. Personal accounts hold PAAWs: Personal Annuitized Average Wage Security • Each PAAW pays • Economy-wide average wage index in the year of retirement (conditional on living that long) • Same number of inflation-adjusted dollars for as long as the individual lives

14. PAAW is a composite of two securities • Average Wage Security • Pays average economy-wide earnings in a given year (like a futures contract) • Personal Annuity Unit (PANT) • Pays $1 (real) for as long as the individual lives • PAAW: composite security that pays off an uncertain number (security 1 above) of a one inflation-adjusted dollar life annuity (security 2 above)

15. PAAWs enhance property rights on accrued benefits • Current law • No legal property rights to benefits • No clear distinction between • benefits already accrued • benefits to be accrued in the future • So benefit cuts often applied across the board • PAAWs • Give owners legal right to collect accrued benefits • Clearly distinguish between accrued and other benefits • B/c PAAWs provide market value, if accrued benefits are cut, people would know how much is being “stolen”

16. How can individual accounts (with irrevocable annual accrual) mimic redistribution based on lifetime income? • Current system: redistribution based on lifetime rather than current income • Typical individual account plans exclude redistribution or base it on current income • We replicate redistribution based on lifetime income by using a variable government match • Match formula is relatively simple: declines with PAAW balance (PBAL)

17. How are PAAWs accrued? • PBALit = number of units of PAAWs accrued by worker i as of year t = benefits a worker would be entitled to under the current system if all future earnings = 0 • PBAL can rise, but can never fall • Other definitions of accrual exist , but our definition accumulates balances most rapidly • PAAW accrual is a function of • New contributions • Accumulated balances (PBAL)

18. Additional PAAWs Per Additional Contribution (measured in average wage units) High when PBAL is low Using this schedule for every contribution mimics the lifetime redistribution of the current system And low when PBAL is high

19. Variable Match Proposal • A historically high wage worker will be getting worse allocation of PAAWs per contribution than a historically low wage worker of same age. • Could be added to other, more traditional, individual account proposals (e.g. MacGuineas, Liebman, and Samwick, 2005) as a way of enhancing progressivity

20. Notes about Variable Government Matching • Formula slightly different for late-in-life contributions • only excess of contribution over 35th highest relative contribution to date counts toward PAAW accrual • Match so far defined in units of PAAWs • Match “rate” (i.e. dollar of match per dollar of contribution) requires market valuation of PAAWs (later)

21. Simple Numerical Examples • Worker 1: Relative earnings = 1 (earnings = average economy-wide earnings in every year) • Worker 2: Relative earnings = average relative earnings for cohort born in 1938 • Worker 3: Earnings = ½ earnings of Worker 2 • Worker 4: Earnings = 1.5 earnings of Worker 2 • Future work: examine realizations of stochastic earnings process

22. Additional PAAWs Per Additional Contribution (measured in average wage units)

23. PAAW Balances (PBAL)

24. Extensions (future work) • Spousal benefits • Separate spousal account, with vesting after 10 years of marriage • Survivors benefits (children/spouses) • Can be handled as well

25. Turning PAAWs into Marketable Securities • Require individuals to sell small percentage of their PAAWs (e.g. 10%) to the market • All individuals in same cohort receive same price • SSA (or private firms) package together pools of PAAWs and sell them • Similar to pools of mortgages • Over time, pools of PAAWs turn into pools of PANTS, which are identical to longevity or survivor bonds

26. Pricing PAAWs • Allows individuals to observe • value of account • value of contribution – value of additional PAAWs • Useful later for • valuing aggregate S. Sec. assets and liabilities • making system self-balancing • allowing for the possibility of trade within accounts • providing market signal about aggregate survival probabilities

27. PAAWs (especially PANTS) market useful for private sector • Would help private insurance companies price annuities. • Same for reverse mortgages. • Could be held by these companies as hedge or collateral for their promises.

28. Approaches to pricing PAAWs • Assume risk neutrality • Allow for risk aversion

29. Pricing PAAWs under risk neutrality Based on assumptions in 2005 SSA Trustees Report

30. Projected market price of one PAAW(under risk neutrality, measured in average wage units) • Price rises with age because: • r > growth of average wages • probability of survival to retirement rises with age

31. Projected Market Value of Accrued PAAWs(measured in average wage units)

32. Defining the match • Match (tax/subsidy) = value of extra PAAWs - value of contribution • Average match rate • [P PAAW * (Δ PBAL) / Annual contribution] -1 • Marginal match rate • [P PAAW * (increment to PBAL per additional dollar of contribution)] -1

33. Properties of the Match • Match rate can be + or – • Match rate cannot be < -100% • balances cannot be taken away • all redistribution occurs on the way in, none on the way out • Unlike simple DC plans, match rate not constant across people or time • Depends on PBAL, price of PAAWs, and fraction of contribution that “counts”

34. Marginal Match Rate

35. On net, match rate is positive for young and negative for old • Match lower for old because • Progressivity means that a given relative wage contribution generates more PAAWs when young (when PBAL is low ) • 35-year averaging formula means that earnings in 36th year and beyond only increase PBAL by the amount they exceed the 35th highest relative wage to date • Although this is partially offset by • Price of PAAW rises with age [and holding PBAL constant, increment to PBAL due to additional contribution (measured in average wage units) is constant with age] • Match rate concept differs from incentive to work (i.e. implicit marginal tax rate computed by Feldstein/Samwick and others), because working today also affects future match rates

36. Pricing PAAWs under Risk Aversion(work in progress) • Recall that PAAW is a composite security, based on • Average Wage Security • PANT • First, price each separately, then price composite (QUANTOS analogy) • Difficulty: these securities are not currently traded, nor are they spanned by currently traded securities

37. Pricing Average Wage Securities • Payout • Average economy-wide earnings in specified year (e.g. 2030) • Models (at least two possibilities) • Find portfolio of traded securities that best matches payout, and assume that residual risk has price = 0 OR • Assume model such as consumption CAPM • Complications • Return on average wage security ≠ growth of wages • Correlation of wage growth and stock returns low over short horizons (qtrly/annual), but likely to be higher over longer horizons (cointegration) • Implications • Social Security liabilities may be less than typical estimates

38. Pricing PANTs • Similar methodology can be used to price pools of PANTS • Equivalent to survivor or longevity bonds

39. The current system is not balanced • Accrued Liabilities. • Social Security taxes have no connection to benefits, either as a flow or in present value. Issuing PAAWS according to old rules does not automatically balance, even if steps taken to make it balance now. • Benefits cannot be market-valued • Pricing PAAWS gives us a way to address balance.

40. Transition from the current system • Geanakoplos, Mitchell, Zeldes (1999) describe • how to compute accrued benefits • how to implement “cold-turkey” conversion to individual account system, by issuing “transition bonds” or “legacy bonds” = PV of accrued benefits • Possible choices for bonds • Treasury Bonds (nominal or TIPS) • Claims on % of future payroll tax (Valdes-Prieto, 2005) • PAAWs • Issuing PAAWs as transition bonds leaves accrued future benefits unchanged (in all states of the world)

41. Balancing Social Security • Give PAAWs in lieu of accrued benefits • Legacy Tax to pay for accrued benefits • Buy PAAWS at market value, but with redistribution. • Government hedges.

42. Creating Initial Balance • Impose a “legacy tax” • On all payroll (including above the cap), or on all income • GMZ estimate: 3% of taxable payroll • Lower if tax above cap. Would be 1-2%. • Legacy debt created by giving retirees in 1940s – 1970s more benefits greater in PV than contributions • So why impose this just on lower earner workers? • Might also want to reduce legacy PAAWs awarded in transition (i.e. cut benefits before locking in property rights)

43. Buying at Market Value • After legacy tax takes care of accrued benefits, social security tax could be used to buy PAAWS at market value. The system is then by definition balanced on the way in. (So for example, system balance is protected against predictable increases in longevity.) • However, this abandons many of the insurance virtues of the current system.

44. Self-Balancing = Allocating Aggregate Risk • Need theory of who can best bear the risks.

45. Adding back some insurance within cohort • In each year, scale all new PAAW accruals of each cohort (according to current SS rules) by an aggregate factor λ • λ = aggregate contributions collected / market value of new PAAWs without adjustment • Implies that annual increment is fully funded

46. Adding insurance across all people contributing at the same time • Same as before, but now involves all people contributing in a year, not just the cohort contributing that year.

47. Hedging is Difficult • If live expectancy increases faster than expected, new PAAW prices reflect that, and government is protected. • But old PAAWS now become more valuable, exposing government. • If earnings temporarily higher than expected, then PAAWS pay more, exposing government. (Under old system, that would be partly offset by higher price for new PAAWS. That could be built in.)

48. If can’t hedge, mutual fund paygo • Change units for PAAWs payout to claims on taxable payroll in a given year • Claims for a given year will be issued to many different generations in varying amounts • In a given year, each claim pays 1/N of total revenue (N=total claims issued) • Alternative: create different tranches • Total payout in each year = revenue • Very old get less risky tranche • Younger old get more risky tranche

49. Conclusions • Translate DB into the language of DC, to facilitate debate over individual accounts • Preserves redistribution and risk sharing of the current system • Clarifies the link between contributions and benefits • Enhances property rights of the system • Might lead to a political compromise between Democrats and Republicans

50. Conclusions • It should be possible to create and trade pools of PAAWs • providing an estimate of the market value of each individual’s account • providing an estimate of market value of system liabilities • opening up the possibility of allowing (limited) trade in these accounts