Changing Progressivity as a Means of Risk Protection in Investment-Based Social Security Andrew Samwick Dartmouth College and NBER October 21, 2006 That’s Quite a Mouthful What Does It Mean? Some proposals to restore solvency combine a scaled-back traditional benefit with a personal retirement account (PRA) invested in financial assets. Financial assets, particularly equities, introduce financial risk. To make reform more feasible, PRAs can be designed to minimize financial risk or mitigate its consequences. 2 Mechanisms To Minimize Financial Risk Don’t Invest in Equities Follow Life Cycle Investment Strategies Reduce exposure to equity risk as retirement approaches, by shifting steadily into bonds. Offer a Third-Party Guarantee At the cost of lower expected returns and higher required PRA contributions. Specify a minimum rate of return (hard) or a minimum benefit level (easier) that will be achieved by the PRA portfolio. Use Options To Protect Against Low Outcomes Either a put or a put plus a written call 3 Really, All of These Are Just a Version of “Don’t Invest in Equities.” Life Cycle funds shift to bonds with age. With Guarantees: The guarantor funds the guaranteed benefits with bonds, then lets the investor have the maximum of the bonds or the portfolio. With No-Loss Strategies: The more interesting question is whether the timing adds value, conditional on the average allocation to equities. The investor earmarks a portion of the contributions for bonds to return the nominal (or real) principal. With Pension Collars: The portfolio is (dynamically) equivalent to specified fractional ownership levels in the stock, a bond at the lower limit, and a bond at the higher limit. 4 What If Portfolio Restrictions Are Not Feasible or Desirable? Social Security already provides a benefit floor, and it would continue to do so to some degree in (almost) any reformed system. If Social Security were made more progressive, that benefit floor would increase (in relative terms). This, in turn, would allow us to be less concerned about exposure to equity in the PRAs and to allow them to be less tightly regulated. The paper quantifies how much equity risk we can shed based on how progressive we make the scaled-back traditional benefit. 5 Varying the Progressivity in the Scaled-Back Traditional Benefit Proportional: Floors at the 10th or 25th percentile: First move all benefits below the specified percentile up to that percentile’s benefit. Then scale all benefits down by whatever amount is needed to achieve a 40% aggregate reduction. Progressive: Reduce all benefits by 40% across the board. First reduce the AIME-to-PIA replacement rates down from {90, 32, 15} to {67.5, 16, 8} Then scale all benefits to achieve a 40% aggregate reduction. Uniform at Mean: Set all benefits equal to 40% of the original mean benefit. 6 Preview of Key Results Greater progressivity can substitute for higher equity allocations. Compared to a proportional cut in the traditional benefits: A commonly proposed progressive cut to the traditional benefit allows the worker to shed half the equity risk. A maximally progressive cut to a uniform benefit allows the worker to shed two thirds of the equity risk. Progressivity is more important when the investor is risk averse or the equity premium is lower. But progressivity does not change the desire to invest in equities much. We would observe similar amounts of financial risk in PRA portfolios regardless of how traditional benefits were cut. 7 Details of the Simulation Model One Cohort of Workers Start with the age-specific mean and quartiles of covered earnings in Kunkel (1996) for the years from 1980 – 1993. Scale them up to 2003 levels by the growth in the national average wage relative to the base year. Impute a lognormal cross-sectional distribution: Median = exp(m) Mean = exp(m + s2/2) Draw a 10,000 observation sample of wages for 30-year olds based on that distribution. 8 Details of the Simulation Method Time-Series Earnings Process A deterministic component that mimics the low-education income profile from HSZ (1995). An AR(1) stochastic component to log earnings with r = 0.95 and s drawn from a uniform distribution on [0.05, 0.20] Backcast to 21 and forecast to 67. Even for a single cohort, this is a very stylized model. 9 Details of the Simulation Method PRA Investment Returns For each observation, at each age, assign a randomly drawn “year” from the Ibbotson (2006) data of asset returns from 1926 – 2005. Make a few additional assumptions: Equity is 75-25 large vs small stocks Govt bonds are equally long-, medium-, and short-term Parameter that varies is share in bonds (assumed 50-50 corporate-government) relative to equity. Keep the variation, but reset the means: Follow SSA’s assumptions when it scores plans: 6.2% equity, 3.2% corp bonds, 2.7% govt bonds (net of 30 basis points in administrative costs) Consider alternative equity means of 5.2% and 4.2% 10 Details of the Simulation Model Benefit Calculations Traditional Benefit PRA Benefit Project the national average wage based on the average wage growth for this cohort over its working career. Use this series and the highest 35 years of earnings to compute the AIME. Use this series to update the bendpoints in the PIA-toAIME formula and compute benefits. Modify as appropriate to increase progressivity. Accumulate a 2- or 3-percent contribution on each year of covered earnings. Convert accumulations to a real annuity benefit based on the period life table for 2002. Combine 40% of the first with all of the second. 11 Figure 1: Changing Progressivity Impact of PAYGO Reductions on Total Income 80 100 50% Bond Portfolios 20 40 60 These differences are relatively unimportant. 0 These differences are very important. 10000 20000 30000 40000 Annual Benefits Proportional Floor at 25th Percentile Progressive Uniform at Mean 12 Figure 2: Shifting from Bonds to Equity Impact of Portfolio Choice on Total Income 40 60 80 100 Proportional Benefit Reductions 0 20 With SSA’s equity premium, high equity allocations aren’t the problem. 10000 20000 30000 40000 Annual Benefits 100% Bonds 50% Bonds Life Cycle 0% Bonds 13 And with 2% off the equity premium Impact of Portfolio Choice on Total Income 40 60 80 100 Proportional Benefit Reductions 0 20 Even these differences are not particularly large. 10000 20000 30000 40000 Annual Benefits 100% Bonds 50% Bonds Life Cycle 0% Bonds 14 Bond Share in Portfolio Table 1 Benefit Distributions under SSA Equity Return Assumptions Method of Reducing Traditional Benefits to Restore Solvency Expected Proportional Floor at 10th Floor at 25th Progressive Uniform at Benefits Reductions Percentile Percentile Reductions Mean Benefit Scheduled PIA Reduced PIA 21785 13071 100 90 80 70 60 50 40 30 20 10 0 17914 18367 18634 19435 20064 20766 21552 22432 23417 24519 25752 Standard Deviations 3289 3150 2797 1644 Cut by 40% 1973 1890 1678 987 0 0 4980 4910 4722 Increase Progressivity, Decreasing4102 Variation 3286 5278 5500 6308 7160 8349 9986 12209 15184 19101 24183 5208 5022 4407 5432 5248 4643 6244 6070 5501 Decrease Bond Share 7100 6939 6411 Increasing Equity Share 8294 8147 7669 9938 9807 9382 Raising Expected Benefits 12167 12053 11682 Raising Variation 15148 15049 14728 19071 18986 18711 24157 24085 23850 3596 3848 4763 5733 7061 8847 11216 14327 18366 23553 15 Calibrating Risk Aversion How Much Would You Pay to Avoid a 50-50 Chance of Gaining or Losing 25%? Coefficient of Expected Certainty Relative Risk Aversion Utility Equivalent 1 -0.0323 0.9682 2 -1.0667 0.9375 3 -0.6044 0.9095 4 -0.4804 0.8853 5 -0.4463 0.8651 Risk Premium 3.18% 6.25% 9.05% 11.47% 13.49% 16 Bond Share in Portfolio Table 1 Benefit Distributions under SSA Equity Return Assumptions Method of Reducing Traditional Benefits to Restore Solvency Expected Proportional Floor at 10th Floor at 25th Progressive Uniform at Benefits Reductions Percentile Percentile Reductions Mean Benefit 100 17914 16202 90 18367 16513 80 18634 16675 70 19435 17130 60 20064 17412 50 20766 17661 40 21552 17867 30 22432 18023 20 23417 18124 10 24519 18170 0 25752 18161 Bond Share Equivalent to Proportional: Certainty Equivalents with CRRA = 3 16302 16491 16804 16617 16814 17132 16781 16981 17302 17240 17449 17777 17525 17738 18070 17776 17993 18327 17983 18202 18538 18140 18360 18696 18240 18461 18796 18285 18504 18838 18275 18491 18823 27 42 56 How Much Equity Risk Can Be Avoided? (56 – 10)/90 = 51% 17230 17577 17756 18252 18555 18818 19031 19188 19284 19318 19293 72 (72 – 10)/90 = 69% 17 Table 2 Benefit Distributions under SSA Equity Return Assumptions and Lower Risk Aversion Method of Reducing Traditional Benefits to Restore Solvency Bond Share Expected Proportional Floor at 10th Floor at 25th Progressive Uniform at in Portfolio Benefits Reductions Percentile Percentile Reductions Mean Benefit 100 17914 17289 90 18367 17687 80 18634 17911 70 19435 18563 60 20064 19031 50 20766 19512 40 21552 19995 30 22432 20470 20 23417 20925 10 24519 21348 0 25752 21728 Bond Share Equivalent to Proportional: Certainty Equivalents with CRRA = 1 17315 17373 17500 17714 17774 17905 17938 18000 18132 18591 18655 18793 19061 19126 19267 19542 19609 19752 20025 20094 20239 20501 20571 20717 20956 21027 21174 21380 21451 21599 21760 21832 21980 1 3 7 17655 18066 18297 18967 19447 19937 20427 20908 21367 21793 22174 12 With less risk aversion, the all-equity portfolio dominates, and greater progressivity would not enable investors to shed much equity risk. 18 Table 3 Benefit Distributions under SSA Equity Return Assumptions and Higher Risk Aversion Method of Reducing Traditional Benefits to Restore Solvency Bond Share Expected Proportional Floor at 10th Floor at 25th Progressive Uniform at in Portfolio Benefits Reductions Percentile Percentile Reductions Mean Benefit 100 17914 15318 90 18367 15569 80 18634 15691 70 19435 16029 60 20064 16215 50 20766 16359 40 21552 16457 30 22432 16506 20 23417 16508 10 24519 16468 0 25752 16391 Bond Share Equivalent to Proportional: Certainty Equivalents with CRRA = 5 15513 15833 16247 15772 16103 16521 15897 16233 16654 16243 16592 17018 16433 16787 17215 16579 16937 17365 16677 17035 17464 16725 17081 17510 16725 17078 17505 16681 17029 17455 16600 16941 17364 55 72 90 16898 17199 17343 17733 17939 18091 18186 18223 18204 18136 18025 100+ With more risk aversion, the optimal equity portfolio shares fall from 90% to 80% or 70% as progressivity increases, and greater progressivity could enable investors to 19 shed all or almost all equity risk. Table 4 Benefit Distributions under Lower Equity Returns and Risk Aversion = 3 Method of Reducing Traditional Benefits to Restore Solvency Bond Share Expected Proportional Floor at 10th Floor at 25th Progressive Uniform at in Portfolio Benefits Reductions Percentile Percentile Reductions Mean Benefit Certainty Equivalents with Equity Return = 5.2% 100 17914 16202 16302 16491 16804 90 18228 16418 16521 16715 17032 80 18478 16570 16675 16873 17192 70 18933 16806 16913 17115 17438 60 19329 16964 17071 17276 17600 50 19758 17087 17195 17401 17726 40 20223 17173 17281 17487 17811 30 20728 17218 17326 17531 17854 20 21274 17222 17329 17533 17853 10 21865 17187 17292 17494 17811 0 22504 17116 17219 17418 17732 Bond Share Equivalent to Proportional: 47 63 79 Knocking 100 basis points off the equity premium has analogous effects as increasing risk aversion: slightly lower equity allocations are optimal, and all or almost all equity risk could be shed with higher progressivity. 17230 17471 17640 17895 18061 18187 18270 18307 18298 18246 18154 100+ 20 Table 5 Benefit Distributions under SSA Equity Return Assumptions and 3 Percent PRAs Method of Reducing Traditional Benefits to Restore Solvency Bond Share Expected Proportional Floor at 10th Floor at 25th Progressive Uniform at in Portfolio Benefits Reductions Percentile Percentile Reductions Mean Benefit 100 20336 17834 90 21016 18281 80 21416 18506 70 22618 19131 60 23560 19496 50 24614 19800 40 25793 20029 30 27112 20178 20 28589 20245 10 30243 20231 0 32092 20144 Bond Share Equivalent to Proportional: Certainty Equivalents with CRRA = 3 17955 18179 18525 18407 18641 18994 18634 18872 19228 19265 19514 19877 19633 19888 20254 19939 20197 20566 20170 20430 20800 20319 20579 20950 20384 20643 21014 20369 20625 20994 20279 20532 20898 35 48 60 19049 19540 19784 20457 20845 21162 21397 21543 21600 21569 21459 73 With larger PRAs, optimal equity allocations fall from 90% to 80%, and greater progressivity facilitates shedding about the same proportions of equity risk (half and 21 two thirds). Life Cycle Strategies Start with low bond allocations at young ages, shifting over time to high bond allocations. Compared to a uniform 50% allocation in bonds, these strategies have lower return and lower risk. 5 to 95 in increments of 2% per year 27.5 to 72.5 increments of 1% per year. The average PRA balance grows with age, so equity allocations above 50 multiply smaller balances. In general, these strategies don’t outperform uniform portfolio allocations. With 200 basis points off the equity premium, the second strategy can (mildly) dominate the uniform 5050 portfolio, which was previously optimal. 22 Table 6 Benefit Distributions under Life Cycle Portfolio Allocations, Equity Returns = 4.2% Method of Reducing Traditional Benefits to Restore Solvency Bond Share Expected Proportional Floor at 10th Floor at 25th Progressive Uniform at in Portfolio Benefits Reductions Percentile Percentile Reductions Mean Benefit Standard Deviations 5 to 95 18651 6148 6087 5923 5388 4713 27.5 to 72.5 18772 6271 6211 6048 5519 4850 50 18898 6591 6532 6376 5869 5233 Certainty Equivalents with CRRA = 3 5 to 95 18651 16550 16653 16848 17165 17605 27.5 to 72.5 18772 16592 16695 16892 17209 17649 55 54 53 52 51 50 49 48 47 46 45 18789 18810 18832 18854 18876 18898 18921 18943 18966 18988 19011 16577 16579 16581 16582 16584 16585 16585 16586 16586 16585 16585 16680 16682 16683 16685 16686 16687 16688 16688 16688 16688 16687 16876 16878 16880 16881 16882 16883 16883 16884 16883 16883 16882 17191 17193 17195 17196 17197 17198 17198 17198 17198 17197 17196 17628 17630 17631 17632 17632 17632 17632 17631 17630 17629 17627 23 Conclusions Higher progressivity, in this framework, makes workers better off. It also allows them to maintain expected utility with lower exposure to equity risk. In the baseline, half to two-thirds of this risk can be eliminated. More at higher risk aversion or lower equity premiums, despite lower optimal equity allocations. However, greater progressivity does not reduce their desire to invest in equities much. Life Cycle strategies are of some, but limited, use in improving welfare. 24 Possibilities for Further Research Simulating the portfolio returns Should I be assigning everyone the same sequence of “years” and bootstrapping the results? More sensitivity tests More variety in (deterministic) wage profiles Couples versus single households Multiple cohorts Actual versus hypothetical workers 25