Research by The Pew Charitable Trusts on individual retirement accounts (IRAs) with automatic enrollment, known as auto-IRAs or Secure Choice accounts, illustrates how retirees could use their auto-IRA accounts to delay claiming Social Security benefits, thereby boosting their monthly and annual payment amounts for life. Pew found that workers who invested 6% of their wages in auto-IRAs could delay claiming Social Security benefits for a year or longer. This deferral has the potential to increase their monthly and annual benefits by 7%-8% a year for life.
About 42% of American private sector workers lack access to a workplace retirement plan such as a traditional defined benefit plan, a 401(k), or an IRA, while 49% participate in one.1 Black and Hispanic workers report lower access rates than White workers, and larger companies are more likely than smaller ones to offer a plan. Older and more highly educated workers are more likely to have access to a workplace retirement savings plan.
In California, Colorado, Connecticut, Illinois, Maryland, New Jersey, and Oregon, private sector workers who don’t have access to workplace retirement savings plans are, or will be, automatically enrolled in auto-IRAs to which they contribute a preset percentage of their wages or salaries. Workers can change the contribution percentage to these auto-IRAs or opt out entirely. Since 2012, at least 20 other states, among them Pennsylvania, Virginia, and Wisconsin, have considered auto-IRA legislation.
Because IRAs help grow savings that can be withdrawn without penalty beginning at age 59½, offering state auto-IRAs to employees who wouldn’t otherwise have access to these accounts is an important savings tool, giving workers options as they reach retirement. For instance, they could convert their savings to annuities, or use the balances as rainy day savings to cover large repairs, medical bills, or other purposes.
Or IRA account owners who retire at age 62 could use their accounts as a short-term income source—funding the first months or years of retirement with monthly withdrawals from their IRAs in amounts equal to the Social Security benefit that they otherwise would have begun taking at age 62. Such an approach would allow them to delay claiming Social Security, significantly increasing their Social Security monthly benefits. On the flip side, claiming Social Security benefits as early as age 62, when they first become available, leads to a large and lifelong reduction in benefits.2
This brief updates earlier research by The Pew Charitable Trusts on auto-IRAs and Social Security.3 The earlier study, which assumed a 3% default contribution rate, showed that, by 2050, 39% of workers participating in state auto-IRA programs could delay claiming Social Security by a year or more, and 20% of retirees could delay claiming by at least two years. In the past few years, California, Illinois, and Oregon have begun using a 5% default contribution rate, and other states are considering using a 6% default rate. (Some proposals, from AARP and others, would even extend the auto-IRA model nationwide and potentially to all workers, including employees at smaller firms and independent workers. 4) This brief uses a 6% default contribution rate and, in another change from the earlier report, uses the income and demographic characteristics of an actual state, Pennsylvania, which is considering establishing an auto-IRA program.
Among the brief’s key findings:
By keeping total lifetime benefits relatively constant for different claiming ages, the Social Security formula is structured to eliminate any financial advantage from claiming early. Those who start collecting benefits at an earlier age receive lower monthly and annual payments because they will receive benefits over a longer period than those who start at the normal retirement age—67 years for most workers today7—or later. These reductions are adjusted using actuarial life expectancy tables to account for average lifespans, and they are permanent.
|Age at claiming||Actuarial reduction to Social Security benefits||Full monthly benefit of $1,000 after actuarial reduction for early claiming||Percent benefit increase after delaying claiming by one year|
Sources: Pew calculations and actuarial reduction amounts from Social Security Administration, “Retirement Planner: Full Retirement Age,” at https://www.ssa.gov/planners/retire/retirechart.html
© 2021 The Pew Charitable Trusts
The actuarial reductions mean that somebody who claims Social Security benefits at age 62 would receive a monthly benefit 30% lower than if he or she waited until age 67 to claim: In Table 1, the worker who starts collecting at 62 would receive $700 a month, compared with $1,000 for the worker who waits until 67. In general, if a worker can delay claiming benefits for just a year, monthly benefit payments will increase by 7% to more than 8% (fourth column) for the worker’s lifetime.8
Delaying the start of Social Security payments, and thus increasing the amount of each payment, can be especially advantageous to married couples, because when a recipient dies, the surviving spouse continues to receive the higher of the two spouses’ benefits—whether or not he or she was the primary earner.9
At Pew’s request, the Social Security Administration (SSA) used its microsimulation model to explore how many retirees in Pennsylvania and throughout the country could use their auto-IRA accounts to delay claiming Social Security by one or more years.10 The following assumptions were built into the microsimulation:
Results of the microsimulation were tabulated for workers who will have reached at least age 6013 by 2050 (the farthest projection year possible for these simulation specifications) and who will have retired and claimed their accounts and Social Security by that year.14 By the end of 2050, the hypothetical program will have been effective for 32 years, so some workers would have 32 years of potential contributions to their auto-IRAs.15 But under the simulation, actual contribution years are much lower for most workers—mainly because many workers move between jobs during their careers and they are eligible to contribute to auto-IRAs only if their employers meet program requirements such as having 10 or more employees and not already offering a retirement savings plan.
The microsimulation model tries to account for workers who enter and leave the auto-IRA program as their work situation changes. Additionally, older workers who entered the auto-IRA system just a few years before retiring would have fewer contribution years. As a result, about two-thirds of workers who would retire and claim their accounts by 2050 would have contributed to their auto-IRAs for 10 years or fewer. Of course, many workers might be able to contribute to other retirement plans such as a 401(k) at different points over their careers, but the goal of this study was to isolate the effects of a hypothetical state-facilitated auto-IRA program on retirement security. Rollovers into auto-IRAs from other retirement plans, while they may increase account balances, were not captured by the microsimulation.
Still, the results indicate that a sizable percentage of auto-IRA participants would be able to use their auto-IRA account balances to delay claiming Social Security by at least a year. By 2050, more than half of workers in these programs could delay claiming Social Security by a year or more. More than a third of retirees could delay claiming for at least two years.
|Percentage who would claim one or more years later||Percentage who would claim two or more years later||Median months a plan participant could delay claiming|
|Household income quintile|
|< High school||64.2%||44.2%||20|
Source: Social Security Administration, results from Modeling Income in the Near Term (MINT) microsimulation model, May 2019. * Results for “other” race and ethnicity, which combines Asian Americans and Native Americans, are not shown.
© 2021 The Pew Charitable Trusts
The microsimulation results show benefits for some groups that often are not able to, or do not, save in traditional retirement systems. For example, the modeling indicates that Black and Hispanic workers would be able to use auto-IRA accounts to delay claiming Social Security benefits for somewhat greater lengths of time than White workers because they’re less likely to work for an employer that offers a traditional defined-benefit or defined-contribution pension plan.16 Thus, Black and Hispanic workers on average would participate in auto-IRA plans for longer periods of time than Whites—and therefore accumulate relatively higher auto-IRA balances. That money could then be used to delay Social Security benefits longer.
And employees with average and higher-than-average incomes—as well as those with higher-than-average education levels—also reap large benefits from auto-IRA participation and the opportunity to use IRA account balances to delay claiming. The middle (second to fourth) income quintiles had the highest share of workers who would be able to use their accounts to delay Social Security benefits by one year. These findings appear to be driven by the number of years workers are eligible for, and contribute to, their auto-IRAs. Workers in the second and third income quintiles had a median of seven years of contributions to their auto-IRA accounts; those in the first and fourth quintiles had a median of six years of contributions; and workers in the highest (5th) income quintile had a median three years of contributions. Higher earners are more likely to work for an employer that offers a defined benefit or defined contribution retirement plan;17 consequently, they are less likely to participate in an auto-IRA, but when they do participate, they contribute 6% of a higher income.18
When it comes to education, workers with college or graduate degrees may be able to delay taking Social Security for longer periods than those with an associate degree. There are two parts to this result. First, more education is associated with higher earnings,19 and higher earners can contribute more to auto-IRAs each year that they participate—subject to contribution limits—and therefore build up larger IRA balances.20 Second, Social Security benefits make up a larger share of preretirement wages for lower earners than they do for higher earners.21 So higher earners reach retirement with relatively high IRA balances, on one hand, and the Social Security benefits they need to replace, on the other hand, are proportionately less as a share of earnings than those of lower-wage earners—enabling them to delay claiming benefits longer than lower earners.
Of course, delaying Social Security may not be appropriate for everyone. Some people may decide not to delay because they prefer to hold some of their savings, such as an IRA, in reserve for long-term care or as a cushion for emergencies, such as house or car repairs, health crises not covered by Medicare, or other expenses.
And many people—such as the unemployed, people with health problems that make work difficult or impossible, or those with a terminal illness or a lower-than-average life expectancy—might find claiming Social Security as early as age 62 to be advantageous. Although retirees with lower-than-average life expectancies, for instance, would still be able to increase their monthly Social Security benefits by withdrawing money from their IRAs to help delay claiming Social Security, doing so could have other implications: Higher monthly Social Security benefits over a shorter-than-average lifespan could mean lower total lifetime benefits.22
Social Security’s financial problems may affect how retirees approach deciding when to begin claiming benefits. Unless Congress acts, Social Security will face a financial shortfall in the coming decades, with the impact likely to be felt starting in 2035—when, under the current structure, the program will be able to pay only about 79% of scheduled benefits, leaving many of those claiming benefits potentially in limbo.23 It is not clear how or when Congress will address this potential shortfall; possible solutions include benefit cuts, payroll tax increases, or infusions of other revenue into the program. But in any case, the prospect of a Social Security reform package that could include benefit cuts makes more acute the need to help workers build retirement resources through their own savings.
Although many workers associate retiring from the workforce with the start of claiming Social Security benefits, the two decisions are actually independent of one another. As a result of delaying their benefits for a year, workers could stay in the workforce if they choose or are able to do so, while building up a permanent and meaningful increase in monthly benefits. Or, workers could retire from the workforce but enjoy the equivalent of their benefit amount by taking withdrawals from their IRAs, all the while delaying claiming Social Security to increase final benefits.
State-sponsored auto-IRA accounts with a 6% default contribution rate would allow more than half of participating workers to delay claiming their Social Security benefits for a year or more. This approach may not be appropriate for all retirees, but for those who can take advantage of it, the strategy could raise Social Security payments permanently during their retirement years.
Pew worked with the Social Security Administration (SSA), which used its Modeling Income in the Near Term (MINT) microsimulation model. To forecast retirement income, MINT projects work histories, mortality, marital status, and disability status for all individuals in the simulation on a national level, based on historical probabilities that are derived from publicly available national datasets matched to Social Security administrative data.
The microsimulation used specifications that are common to many auto-IRA programs. Specifically, workers would be eligible for auto-IRAs if they are employed by businesses that had 10 or more employees, are age 18 or older, and if their current employer did not already offer a traditional defined benefit or defined contribution plan, such as a 401(k) plan.24 The simulation model also captures the demographic characteristics of Pennsylvanians who are likely to take part in the proposed state-sponsored auto-IRA.
Workers are assumed to start participating in 2019 with a default contribution rate of 6%. The microsimulation took into account workers’ ability to opt out of the program, as well as the probability that some workers would contribute more than (or less than) the 6% percent default. Rates of participation in auto-IRA plans are based on the results of a 2016 Pew survey of employees after all the program’s details had been presented (Table A1).25
|Auto-IRA participation rate assumptions, 6% contribution rate|
|Stay in the program||72.18%|
|Lower than default rate||6.85%|
|Higher than default rate||4.11%|
|Opt out of the program||16.86%|
At the time that SSA ran the MINT model for this study, actual participation rates based on a 6% default contribution rate were not yet available, although more recently we have begun to see results from states such as Oregon. Studies have suggested, however, that inertia or “default stickiness” causes employees to be slow to move away from default contribution rates, even when defaults are relatively high, and particularly in the case of low-income employees.26 However, a study of participants by Vanguard, one of the world’s largest investment management companies, found that participants with default contribution rates above 3% were 22 percentage points more likely to decrease their contributions to an amount below the default rate relative to participants with a default rate of 3%. This finding was despite the fact that many of the plans in the Vanguard database included an employer match—something that is not currently included in auto-IRA plans—that was higher than the default contribution rate.27
Income quintiles used in the analysis are as follows:
|Household income quintile||2050|
As in the previous Pew report, the new microsimulation does not attempt to measure a retiree’s total income— including spousal income, pensions, or income from drawing down savings—because the study is concerned only with the increase to retirement income that can result from using auto-IRA savings to defer Social Security benefits for a year or more. Although recent evidence suggests that auto-enrollment may lead some workers to take on additional debt, the modeling here does not address whole-household effects on income or assets. For similar reasons, the simulation did not attempt to estimate tax impacts—for example, a comparison of after-tax Social Security benefits to nontaxable Roth IRA withdrawals is not shown.
Annual real rates of return are assumed to be a constant 2.9% for government bonds, 3.4% for corporate bonds, and 6.4% for stocks. Annual administrative fees are assumed to be 0.75% of assets.
Workers’ auto-IRA account balances at retirement are primarily driven by the number of years they contributed to their accounts. Years of contributions, in turn, reflect modeling assumptions about how many years a worker met eligibility requirements for participation (i.e., working for an employer that did not already offer a retirement plan and that had 10 or more employees) and assumptions about how many years the worker chose to stay in the program instead of opting out. Thus, while many simulated participants could potentially have contributed to auto-IRAs for the program’s full life, for most participants the actual number of contribution years is lower. Nearly two-thirds of auto-IRA holders are projected to contribute for 10 or fewer years.
Workers are assumed to access their accounts at retirement, which in the MINT model is interpreted as working less than 20 hours a week. Beneficiaries are assumed to make monthly drawdowns from their auto-IRAs equal to the monthly Social Security benefit they could have received at the earliest claiming age of 62. Workers do not access their accounts for cash-outs before retirement, after a job change, or for other qualified purposes such as a health or education expense; this is likely to be an unrealistic assumption. Results are calculated on a national basis and are shown for retirees through 2050, which is the furthest into the future that the model can project for these types of simulations.
Resources for federal, state, and local decision-makers
Data-driven state policy innovations across America