Reforming Social Security

Mark J. Warshawsky

Current Issue

Social Security reform is no longer in vogue these days, even on the right. Across the political spectrum, ignoring the program's fiscal challenges and design flaws is now taken to be politically astute. It isn't hard to see why: Turning a blind eye to problems that can only be addressed by expending political capital has short-term appeal. Being irresponsible is easy — until it comes time to pay the price.

While we ignore the program's ills, that price keeps growing. Social Security's finances continue to worsen, as has the overall fiscal capacity of the federal government. Labor markets and social and demographic conditions continue to diverge from the assumptions that originally underpinned the program, and policymakers continue to miss opportunities to build considerable savings for the poor through financial markets.

But the past few years were not wasted entirely. I served as deputy commissioner for retirement and disability policy at the Social Security Administration (SSA) from July 2017 through January 2021, where my colleagues and I were able to develop some ideas, data, and research to inform future reform efforts and strengthen Social Security. Our research reveals that Social Security plays a relatively smaller role in the financial security of older Americans than commonly thought, making reform easier from a policy perspective, if not politically.

No such reform is remotely likely in the Biden era. Would-be reformers should use this time to prepare the ground for essential arguments to come. It is worthwhile, therefore, to review the design and fiscal condition of Social Security, account for the changing character of retirement security in America, and consider a potential path for, and the nature of, reform.


Social Security taxes are high and apply to nearly all types of workers. Employees pay 6.2% of their income to earn Social Security benefits — 5.3% for retirement and survivor benefits and 0.9% for disability benefits — which is matched by identical amounts from their employers. A self-employed individual must pay the entire 12.4% payroll tax.

In 2021, the average worker paid $3,662 in Social Security payroll taxes; the maximum for employees was $8,853, or $17,707 for the self-employed. The only workers exempt from paying Social Security taxes are those covered by qualified state- and local-government pensions deemed sufficient. Currently, about a quarter of the state- and local-government workforce — primarily teachers, police, and firefighters — are exempt.

The maximum earnings subject to Social Security taxes in 2021 was $142,800, so no benefits are accrued on amounts above that threshold. Medicare payroll taxes, however, are not capped. One needed to earn a minimum of $1,470 in one quarter to receive one work credit, or $5,880 in a single year for four credits, which are put toward the 40 credits required to vest in — that is, qualify for — the program's retirement benefits. Fewer credits are needed to receive disability and survivor benefits, depending on age.

Social Security benefits are relatively generous, especially for low-wage workers. For those turning 62 in 2021 or who become disabled, the basic Social Security benefit (the so-called Primary Insurance Amount, or PIA) is 90% of the first $996 of average indexed monthly earnings (AIME), which are calculated for the individual based on the highest of up to 35 years of earnings, plus 32% of AIME between $996 and $6,002, plus 15% of AIME over $6,002 up to the cap on taxable earnings. The benefit formula is thus progressive, clearly favoring the lower paid.

The full retirement age — the age one is required to reach in order to claim full Social Security benefits — is 67 for those born in 1960 or later, up from 66 for those born from 1943 through 1954 and 65 for those born before 1938. The maximum monthly Social Security benefit for workers retiring at the full retirement age in 2021 was $3,148, or $37,776 a year. Benefits are increased if claimed past the full retirement age, up to age 70, and are reduced if claimed before, down to age 62. For workers who are disabled upon reaching retirement, their benefits automatically convert to retirement benefits at the full retirement age, with no reduction in amount. The average monthly benefit for new retirement awards was $1,636 in 2020 and $1,544 for all those currently receiving benefits. (For the disabled, comparable amounts are $1,454 and $1,277, respectively.) In 2022, benefits were raised by a cost-of-living adjustment of 5.9% — the largest increase in 40 years — owing to high price inflation.

In 2020, Social Security benefits represented 5.23% of GDP — 4.55% for retirement and survivors, and 0.69% for disability. Almost 5.8 million people received Social Security benefits for the first time that year. Of those, 58% were retired workers, 11% were disabled workers, and 31% were survivors and dependents of these and deceased workers. With the aging of the Baby Boom generation and the general increase in the size of the working population (thanks largely to immigration), the numbers of new awards to retired workers rose from 1.6 million in 1980 to 3.4 million in 2020. New awards to disabled workers also increased from roughly 0.4 million in 1980 to nearly 1 million in 2010, but have since subsided to 0.6 million in 2020 — likely due to a strong job market and adjudication reforms by SSA in the 2010s and, more recently, to pandemic-related stimulus payments and closings of Social Security field offices.

Today, almost 65 million people receive Social Security benefits; 71% are retired workers, 13% are disabled workers, 9% are survivors of deceased workers, and 7% are the spouses and children of retired or disabled workers. About 15% of beneficiaries are younger than 62. Women have reached total parity with men, both as retired adult-worker and disabled-worker beneficiaries. Because women live longer than men on average, and because there are more widowed women than men receiving associated benefits, there are now more female than male beneficiaries of Social Security overall.

To what extent do America's seniors rely on Social Security benefits for their retirement income? For many years, policy analysts and the SSA itself relied on the Current Population Survey (CPS) alone to answer that question. According to these data, Social Security represented by far the largest component of seniors' incomes, and for nearly a quarter of that population, more than 90% of their income.

Doubts about the accuracy and completeness of this survey data, however, have mounted over the years. Under my direction, researchers at the SSA completed a lengthy, painstaking project of supplementing the CPS data with SSA and IRS administrative records. These new data, which represent the state of affairs in 2015, show a substantially lower share for Social Security in seniors' income and a much higher share for pensions and retirement accounts.

Using the CPS data only, Social Security accounted for nearly 35% of total senior income, while pensions and retirement accounts represented only 22%. With the new data, the reality is almost flipped: Social Security makes up only 30% of total senior income, while pensions and retirement accounts represent 36%. (The other main sources of income include earnings and asset income.) Similarly, using the CPS data alone, fewer than half of senior families had pension- and retirement-account income. But according to the new data, nearly 70% do — in range of the 85% to 90% of those with Social Security income. Finally, whereas according to the CPS data, more than 50% of the senior population relied on Social Security for at least half of their income (and 25% relied on it for 90% or more), the new, more accurate, and more inclusive data reduce those rates to 40% and less than 15% of the senior population, respectively.


Social Security is mostly a pay-as-you-go program — that is, current benefits are financed largely by current payroll taxes. Of the $1.1 trillion in annual program costs in 2020 (98% of which are paid in benefits), almost 90% was covered by payroll taxes, approximately 3% was covered by taxation of the Social Security benefits transferred from the general government account to the program, and over 7% was covered by interest earned on the bonds held in the combined retirement and disability trust funds paid by the federal government.

Yet according to the Social Security trustees, the trust funds began to shrink in 2021 — and by 2034, they will be exhausted. At that point, payroll taxes and taxes on benefits will only be able to pay 78% of program costs. The projected shortfall over the next 75 years is 3.54% of payroll, up from a 2.66% deficit in 2016. The cost of the program is also projected to increase, from the current 5% of GDP to more than 6% by 2080, while non-interest income remains essentially flat, at about 4.5% of GDP.

Although long and covering three or so generations, the 75-year horizon is arbitrary, and it cuts off additional years of deficit as well as future generations. The trustees, therefore, also calculate the shortfall over an infinite time horizon. On that basis, the deficit is now at 4.6% of payroll — up from 4% in 2016.

The program represents a considerable drain on the federal government's finances. Since 2010, Social Security has paid out more in benefits and expenses than it has collected in taxes and other non-interest income. This drain is projected to grow from about 2% of taxable payroll in the 2020s to about 3.5% in the year of program insolvency, and to almost 5% by around 2075 if scheduled benefits continue. The recent pandemic-induced recession hit Social Security especially hard, increasing the deficit and drawing the insolvency date forward by a year, to 2033.

The Concord Coalition recently produced some insightful analysis on the implications of Social Security's worsening finances for the federal government. According to the coalition, as the trust funds approach exhaustion and the federal government redeems its bonds to continue to pay Social Security benefits, publicly held debt will increase as a result — by almost $6 trillion by 2035. Even with today's unprecedented public-debt standards, that increase represents a large burden: about 15% of the year's projected GDP. In terms of interest cost, the added annual burden on public finances grows to almost 1% of GDP.

Although there are many reasons for this grim state of affairs, the demographic challenge is a primary driver. The number of retired workers is projected to double in about 50 years as members of the post-World War II Baby Boom generation continue to retire in increasing numbers. People are also living longer, while the birth rate has dropped in recent years. As a result, the trustees project that the ratio of 2.7 workers paying Social Security taxes to each person collecting benefits in 2020 will fall to 2.2 in 2039 — and slide further thereafter.

Even this sober assessment may be too rosy, because the trustees assume that the birth rate will increase to 2.0 children per woman in their intermediate scenario. In reality, the rate has fallen from over 2.0 in the mid-2000s to 1.65 in 2020. While some of the decline may be temporary as women postpone childbearing, strong survey evidence and analysis by researchers at Boston College indicate that most, if not all, of the fertility slide is permanent. For example, surveys of the total births expected among women ages 20 to 24 show a decline from 2.46 in 1982 to 2.09 between 2017 and 2019. Further analysis finds a gap of about 0.45 children between projected completed fertility and expectations of fertility at younger ages. All this points to a more reasonable projection of a future birth rate closer to 1.65 (the current rate), rather than the trustees' projected rate of 2.0. As a result, the long-run finances of Social Security are much worse than the trustees project.

According to sensitivity analysis in the trustees' report, if the average fertility rate were 1.69, the actuarial balance over 75 years would be -4.26% of taxable payroll rather than -3.54%, and the annual balance for 2095 would be -6.69% instead of -4.34%, even as the year of trust-fund depletion does not change. Although an increase in immigration, legal and illegal, could make up some of the difference, the trustees have already incorporated an aggressive assumption for net immigration into their analysis.

Finally, although the trustees put on an impressive interagency effort in their report projections, they are not the only game in town. The Congressional Budget Office produces a more negative view: It projects that the trust funds will be exhausted in 2032 (when benefits would need to be cut by 25%), and that the 75-year actuarial deficit is 4.9% of taxable payroll.


Even setting aside the deteriorating finances of Social Security and the federal government at large, the program is in dire need of reform. Policymakers have not significantly altered Social Security's underlying structure since it was designed in 1935 for retirement, in 1939 for survivors' benefits, and in 1956 for disability. Since that time, the economy, the labor market, and American society in general have transformed in countless ways, rendering many of the underlying predicates of the program invalid.

To begin, many more women have entered the workforce in recent decades — in fact, their numbers are now on par with those of men. In January 1948, when the CPS first started collecting these data, the labor-force participation rate for adult women was 32%; in August 2021, it was 56.2%. For adult men, the numbers were 86.7% and 67.7%, respectively. Women are also increasingly becoming the primary breadwinners of their families: According to the CPS, in 2020, women outearned their husbands in about a third of couples — up from a sixth of couples 40 years earlier.

Lifespans have become much longer, too. According to the SSA, in 1940, only 54% of 21-year-old men and 61% of 21-year-old women survived to age 65, while the average additional life expectancy at age 65 was 12.7 years for men and 14.7 for women. In 2019, 81% of 21-year-old men and 89% of 21-year-old women survived to age 65, while the average additional life expectancy at age 65 was 18.1 years for men and 20.7 years for women.

In addition to more assured and extended lifespans, people are also working later in life, and for longer periods of time. According to the CPS, in 1948, the labor-force participation rate for people ages 20 to 24 was 64%; by 2020, it had risen to roughly 72%. For those aged 55 to 64, the rate rose from 57% to 65%. Looking at more recent trends, the labor-force participation rate for those age 65 and older has nearly doubled, from 10.7% at the end of 1985 (shortly after the last major Social Security reform in 1983) to 20.3% at the beginning of 2020. Although many factors have driven these changes, employment laws forbidding age and disability discrimination passed in the last 35 years, along with the disappearance of defined-benefit pension plans, have likely played a role in allowing and encouraging people to maintain longer careers.

Owing to these changes, Social Security now works against many populations and situations. It effectively subsidizes stay-at-home spouses — even those without children — who are disproportionately likely to live in high-income households. It also hurts working married women (and some men) who pay taxes on their earnings even though they would be eligible for spousal benefits of equal or greater value without working at all. And given that the program only considers the highest 35 years of inflation-adjusted wage earnings in the calculation of benefits while continuing to tax all years of earnings, it disfavors the less educated and the lower paid, who generally have to start working earlier in life and for longer than the highly educated.

The fact that Social Security represents a large portion of needed retirement income for the lowest-paid workers likely discourages them from saving for retirement on their own. According to the most recent Social Security trustees' report, those at the 25th percentile of the national lifetime-earnings distribution will have more than 53% of their pre-retirement income replaced by Social Security in 2021. To reach the generally recommended income-replacement-rate goal for retirement — 70% of prior earnings — the need for additional retirement savings is modest. Encouraging lower-paid workers to open retirement accounts would help them improve their financial security and standard of living later in life, and may even allow them to bequeath some money to their children and grandchildren. Yet according to the Federal Reserve, fewer than 40% of families in the lower half of the income distribution have retirement accounts and plans, compared to nearly all families in the upper decile. As a result, the lower paid are missing out on potential retirement savings. Social Security's structure should be enhancing, rather than discouraging, this population's financial security and independence.

For example, let's posit a worker earning $30,000 a year contributing 3% of his earnings to a personal retirement account beginning in 2006, the year President George W. Bush proposed that such accounts start operation. Assuming a 3% increase in earnings every year and that he invested in a portfolio of 60% low-cost indexed domestic-equity funds and 40% government bonds, continually rebalanced, his account would have reached $32,325 by 2020. This amount would be 71% of his then annual income; he would have realized an annual return of 8.36%.

The existing structure also incentivizes disability claims in the late- career years. As mentioned earlier, if a worker claims retirement benefits before the full retirement age, his benefits are reduced. But if that worker claims disability at the tail end of his career (as is more common when wage levels are high), his disability benefits convert automatically to retirement benefits without the attendant reduction. This gives him an advantage over workers who claim retirement benefits before the full retirement age, even though many do so for health reasons that are insufficient to render them eligible for disability benefits.

The cutoff for full spousal benefits — 10 years of marriage — is not only arbitrary, but differs from pension law, where the division of property (including retirement accounts) is subject to negotiation but not necessarily connected to a particular length of marriage. On the other hand, multiple marriages of 10 years or longer can result in increased spousal benefits.

As mentioned earlier, more than a quarter of state- and local-government workers are exempt from paying Social Security taxes, as their pension plans are qualified as replacement plans. But this proportion is not evenly spread across states and occupations. Nearly all Ohio and Massachusetts government employees are exempt, for instance, but few from New York and Vermont are. Teachers, meanwhile, are more likely to be exempt than government accountants. Because many of these exempt workers earn Social Security benefits anyway (albeit at a somewhat reduced level according to complicated offset formulas arising from either part-time or summer work, or early- or late-career bridge jobs in the private sector), the exemption is unfair to private-sector workers, who will inevitably bear the full cost of Social Security's shortfall.

The longer we delay putting Social Security on a sound financial basis, the larger the burden today's younger workers will share. It will also fall on future generations in the form of increased taxes or reductions in benefits. Not a single scenario or stochastic result contained in the trustees' report shows the trust funds remaining financially solvent — and most show insolvency within about 10 years. Uncertainty, therefore, offers no excuse for further delay.

Some have suggested addressing Social Security's financial troubles by removing the cap on taxable earnings and massively importing more funds into the program without making any structural changes. As I have argued in the past, this solution is inefficient, ineffective, and unfair. Given that the rich have already been tapped as new revenue sources for the Affordable Care Act and Medicare (and are being eyed for more), putting them on the hook for Social Security's deficit could spur further tax-avoidance behavior, which would reduce the revenues raised from the tax-cap increase considerably below actuarial calculations. And because the employer share of the payroll tax ultimately falls on workers, such a move would constitute a massive tax increase on the middle class. Most Americans earning more than $142,000 are not wealthy, especially if they live on the coasts, have large families, or have high educational or health expenses. Many will have trouble bearing the additional tax burden if the cap were removed.

The volatility of personal income over time also renders such a large tax increase unfair. Many workers only experience high earnings temporarily, or at the close of their careers. According to the SSA, almost 20% of current and future covered workers are projected to earn above the taxable maximum during at least one year, and about a third of the future aged beneficiaries projected to earn over the taxable maximum in 2050 will be below the highest shared-earnings quintile. Taxing above the current maximum, therefore, does not mean one is capturing the earnings of only workers with very high lifetime incomes.

Finally, as the SSA actuary observes, removing the taxable-earnings cap fills a declining part of the shortfall over time. If any benefits are credited for the increase in the level of earnings covered, the amount covered is even less.

Eliminating the taxable-earnings cap would do little to address Social Security's shortcomings — it may even further undermine the program by weakening national productivity growth, savings, and labor-force participation. To succeed on this front, policymakers will need to pursue more meaningful, structural change.


The basic elements of program reform flow naturally from the financial and social developments just described.

Consistent with the historical rate of increasing longevity and rising labor-force participation at older ages, the normal retirement age should be raised by one month every two years. This rate of change should also apply to the early retirement age and the maximum claim age (currently 70), as well as the eligibility ages for old- and disabled-widow benefits. As I have explained more fully elsewhere, the adjustment factors for early and late claiming of retirement benefits — which depend on the age when benefits are claimed — are seriously out of date, as both interest and mortality rates have declined since the rules were designed and put into place.

Moreover, the rules are inconsistent among categories of beneficiaries — workers, spouses, and widows and widowers — in terms of age ranges and fair adjustment factors. Under current interest and mortality rates, the early retirement reduction for workers should be about 16% lower than current factors. Reducing the early retirement penalty would generally favor poorer and minority populations, who typically claim earlier and have higher than average mortality rates. Delayed retirement credits should be about 6% lower, disfavoring high-income and educated workers who generally claim late. The appropriate dynamic system for shifting the adjustment factors annually is a five- or 10-year moving average for continuous changes to the adjustment factors, which would remove unusual volatility. Policy logic also supports consistency in the factors across beneficiary categories as well as age ranges: Given higher life expectancies and longer working lives, 62 to 72 is now an appropriate initial age range for benefit claims.

The basis for the cost-of-living adjustment needs to change as well. The current consumer price index (CPI) for urban wage earners is based on a shrinking data set, and does not reflect the substitution of lower-priced goods and services for higher-priced ones quickly enough. Policy should instead employ the chained CPI for all urban consumers. Social Security benefits should be taxed in a manner consistent with employer-provided pension income, and the existing lower-income thresholds should be phased out over the next 20 years. The resulting increase in income-tax revenue should be transferred to the trust funds.

To achieve more fairness among different types of workers, those with at least 45 years of earnings and still working should be exempt from the payroll tax (both employee and employer shares, as well as those of the self-employed). If a worker achieves the 45-year mark before reaching full retirement age, his disabled-worker insurance status should be maintained. To measure a full working career more accurately and fairly in modern times, the number of earning years used to calculate benefits should be raised from 35 to 40 over the course of 10 years. The confusing retirement-earnings tax imposed on earnings above minimal amounts before the full retirement age should be eliminated, as it discourages work. (It was adopted during the Great Depression, when the government wanted to open up positions for younger workers.) Finally, all new state- and local-government workers should be covered by Social Security. As I will explain below, workers currently exempt should be subject to new, more accurate, and more administrable offset-adjustment factors for any Social Security benefits they may have earned.

While maintaining survivors' benefits as is, policymakers should eliminate spousal benefits over time. Earnings sharing should be instituted gradually, whereby the earnings records of married individuals are combined and split equally for each year of marriage for the purpose of calculating each individual's Social Security benefit. In the case of multiple marriages and divorces, the sharing should occur with different spouses over the individual's lifetime during each period of marriage. Because the SSA does not have contemporaneous marital data, this change will be administratively challenging for the agency for the time being; however, it does have plans to obtain such data from the states in order to adjudicate Supplemental Security Income (SSI) claims more effectively. To account for increases in earnings inequality, the earnings above the taxable maximum should be removed from the calculation of the average wage index used to adjust the worker's historical earnings record to account for annual wage inflation. Alternatively, the agency can use the median amount.

To make Social Security sustainably solvent without levying new payroll taxes, an additional benefit-element adjustment will likely be necessary. Scheduled increases in benefits should be slowed for those under age 61 by applying an annual multiplication factor to their scheduled benefits for as long as necessary to achieve solvency.

To make up for the necessary reduction in scheduled Social Security retirement benefits, low-wage and disabled workers should be able to access a national system of personal retirement accounts, supported by government matches of employee contributions. Contributions for disabled workers will be entirely government funded, as I will describe in more detail below.


To determine eligibility for disability-insurance benefits, the SSA uses a five-step evaluation process. First, the agency determines whether the claimant had sufficient years of coverage to be insured, is still working, and is earning above the substantial gainful-activity level. Second, it determines whether the claimant's disability is of sufficient significance and duration (actual or expected) to be considered further. Third, the SSA determines whether a pure medical determination can be made as to whether the disability is sufficiently severe to meet the agency's body-system listings. If not, it determines whether the claimant can still perform his previous work despite the disability (step four), or if other work is reasonably available in the economy for him to perform (step five), again given the person's residual functional capacity (RFC) and taking into account his age, education, and work experience.

The latter two steps constitute the medical-vocational rules, and they have not changed since 1978. They rely largely on job-requirements data from the Dictionary of Occupational Titles, which was created during the Great Depression and last (partially) updated by the Department of Labor in 1991. Given their age, it shouldn't come as a surprise that the rules refer to occupations long extinct and omit newer ones: Among its 14,000 occupations are phonograph-cartridge assembler and web-press operator (print), but no website designer. They have never included the mental requirements of work — an increasingly important factor in modern society. And the regulation adopted in 1978 includes a fairly prescriptive medical-vocational grid for step five, which is outdated and largely inadequate.

The grid used in carrying out the medical-vocational inquiry is made up of four charts — one each for sedentary, light, medium, and heavy work — which are plotted against age, education, and skill level. The grid was designed to allow officials to determine, without further analysis, whether an individual is vocationally disabled, thereby making claim adjudication simpler and more consistent among different agency actors and across levels of adjudication (initial, reconsideration, and administrative-law judge). However, it is now used directly in only about 10% of step-five cases. For the rest, it serves as a guide that requires detailed but often semi-informed analysis of job requirements and numbers of available jobs in the labor market, given the individual's RFC. If a claim involves any assertion of mental impairment or a non-exertional factor (like pain or fatigue) in any aspect of work — and even if these assertions are combined with physical impairments — the case goes off the grid and into an often subjective judgment of how many fewer jobs are available due to the additional impairments. As a result, the need for vocational experts and hearings — which add months and sometimes years to the benefit-application process — has risen.

In cases where the grid is applied or used as a framework, it can produce arbitrary and inequitable results, often related to age. For example, a 32-year-old veteran with a consistent work record who has a traumatic brain injury but can work at a sedentary job would be deemed not disabled, but a 51-year-old with no past relevant work experience and a relatively minor musculoskeletal impairment would be found disabled. A 49-year-old with consistent record of heavy work who suffers a neurological injury and can only do unskilled sedentary work would be judged not disabled, but a similarly situated 50-year-old would be found disabled. In fact, at the initial level of adjudication, those age 55 and older are three times more likely to be awarded disability than those 49 and under. There are particularly noticeable jumps at the 50- ("closely approaching advanced age"), 55- ("advanced age"), and 60-year demarcations. The natural increase in disability awards with age would be much more gradual.

Changes in American labor-force trends since the 1970s only add weight to the case for reform. The current rules presume that people ages 50, 55, and 60 warrant automatic determination of vocational disability in many circumstances due to presumed age-related job discrimination, social difficulty, or employer expectations. But before the Covid-19 pandemic, the labor-force participation rate was higher overall — especially among older and disabled people — than in years past. Advances in technology and changes in the economy have reduced the physical aspect of many jobs, including some that were considered "heavy effort" long ago. Light or sedentary work is more common now, even for those with less education or limited experience. Workers are also more educated, many with post-high-school training. Work hours and conditions have become more flexible for many jobs, even before the pandemic-induced Zoom revolution. And as the overall health and functioning of the population — especially among the older workers — has improved, research indicates that more individuals at higher ages have the mental and, for some, the physical capacity to adjust to new work. At the same time, studies by labor economists show that the extent of physical work associated with routine manual tasks has declined dramatically, while work associated with abstract, non-routine cognitive tasks has increased. Policymakers should seek out opportunities to update the disability rules in light of these changes.

In the early 2000s, SSA attempted to do just that, but its efforts foundered on political opposition and a lack of data. Today, we can address the latter problem with new data. After six years of preparation, testing, and surveying of a statistically representative sample of employers, the Bureau of Labor Statistics, in conjunction with the SSA, has produced a first wave of occupational-requirements survey (ORS) information; this includes requirements related to exertional elements (e.g., climbing stairs), additional physical elements (e.g., fine manipulation), sensory elements (e.g., hearing), and environmental elements (e.g., extreme cold). Because the ORS categorizes occupations according to about 1,000 standard occupational classifications (SOCs) — the same standards that the federal government uses — data from ORS can be matched with other federal survey data and sorted by or restricted based on education, skill, strength, work schedule, prior work experience, or training requirements as dictated by law or policy.

Though the first wave of ORS did not include mental requirements (memory, adapting, etc.), these data are available through a validated Delphi study that Abt Associates conducted for the SSA in 2019. ORS's second wave includes these requirements, and it will be interesting to see how the Abt results compare.

The first wave of ORS was completed in 2019 and analyzed in 2020; the second wave — which uses a larger sample of survey respondents, relies more extensively on new statistical techniques to overcome disclosure restrictions, and includes the Covid-19 experience — will be available in 2024. Thus far, ORS has cost the SSA about $220 million, or roughly $35 million a year.

With this new, nationally representative data on work requirements, policymakers can make sensible policy changes to existing disability programs. For instance, if only entry-level work is considered at step five, the complex transferability-of-skills analysis performed at older ages can be omitted entirely — a significant simplification. Similarly, considering only 10 years of past relevant work at step four (rather than 15) would streamline the adjudication process. The educational levels can also be simplified and extended to those that are vocationally relevant. To the extent that ORS data are sufficiently "thick" at the SOC level, or that some SOCs can be combined, college education should be added as a factor, as even its sedentary work may differ from that of jobs requiring only a high-school education. Finally, vocational and economic research indicates that age 60 is now a natural break point for disability policy, as individuals over that age are more prone to injury and take longer to heal. The rules, therefore, should not take heavy work into consideration for these older workers as possible alternative employment.

The SSA has been working hard on a new disability-eligibility regulation (Vocational Regulations Modernization, or VRM) for the last 10 years. Within the federal government, there have been extensive policy discussions and analyses, data investigations, literature reviews, field and administrative-law judge consultations, legal consultations, actuarial- and economic-impact estimates, distributional analyses, cost estimates, automation-software preparation, and thorough documentation. My rough guess of the additional manpower and resource cost for this effort is $100 million, resulting in a total cost to the taxpayer for VRM of at least $320 million. Of course, change is politically difficult, and advocates will try to define the population down to almost a tautology in order to show people experiencing losses. But the SSA, as a non-political agency, should rise to the need for prudent stewardship of the program and issue a disability-reform regulation.

A few further reforms will reinforce these necessary changes. All disabled workers should be converted to retired-worker status when they reach early retirement age — currently 62. To remove the incentive to claim disability late in one's career, the benefit reduction for retirement at that age should be applied accordingly, with a transition for disabled beneficiaries who already qualify as disabled. Medicare eligibility should be extended to age 65 on the basis of disability, but new disability applications should not be accepted after the early retirement age. To simplify the program and encourage the return to work for those disabled beneficiaries able to do so, the trial work periods and other complex post-disability, work-related provisions should be eliminated and replaced with extended disability eligibility, including Medicare coverage, until medical improvement occurs or the beneficiary begins work. Like SSI, disability benefits should be reduced for earnings above certain minimum thresholds, perhaps according to a step function with proportionally increasing reductions. To reduce gaming of the system and searching for easy adjudicators, the rules should prohibit individuals from applying for disability benefits more than once in a three-year period, except in unusual circumstances.

Finally, administrative-law judges should not advocate for the claimant in the majority of cases where there is a third-party advocate, usually an attorney, representing the claimant. Instead, officials should field a demonstration project to test whether more fully compensated attorney representation during the initial stage of adjudication at the state-agency level — where representation is now relatively uncommon — could speed up the process without a loss of accuracy or change in ultimate award rates. Some preliminary econometric evidence suggests that this change could be successful; the demonstration project should test the hypothesis further by paying for attorney representation of a randomly selected set of applicants beyond the lower rate that they would otherwise pay for a successful but quick award.


As mentioned, about a quarter of state- and local-government workers are not covered by Social Security — that is, they do not have payroll taxes taken from their earnings, nor do they earn benefits on those earnings. Instead, they are covered by government-employee pensions that have to meet certain retirement-benefit adequacy rules. Yet despite these rules, some workers still fall between the cracks.

For example, if a worker does not work long enough to vest in his government pension (often five but sometimes as long as 10 years), he will receive no retirement benefit, and maybe no Social Security benefits unless he can piece together 10 years of additional jobs over his lifetime for which he paid taxes. For this and the reasons outlined above, gradually switching to a system in which all new government workers are enrolled in Social Security would be a wise policy move.

Of course, during the transition, we would still have to consider existing exempt workers. One issue policymakers will have to grapple with on this front is the fact that many government workers receive generous pensions for non-covered careers of, say, 25 years with their governmental employer, retire early, and still receive Social Security benefits after working 10 years of second careers, summer jobs, youth employment, and so on.

The problem stems from the fact that the Social Security benefit formula is weighted so that benefits replace a greater percentage of the wages of lower-wage workers than they do of higher-wage workers. Prior to 1983, employees who weren't covered by Social Security (and were often highly paid) had their Social Security benefits calculated as if they were low-wage workers for the length of their career. This was clearly unfair, so Congress passed the Windfall Elimination Provision (WEP) to remove that advantage by reducing a worker's retirement or disability benefit if he received a pension from work not covered by Social Security. A comparable provision designed for similar fairness reasons — the Government Pension Offset (GPO) — reduces spouse and widow benefits and serves as a partial proxy for the dual-entitlement rule.

The current WEP and GPO formulas are only approximately correct, and actually disfavor lower-wage workers. The deeper problem, however, is that government-worker beneficiaries don't always declare their pensions, and SSA field offices often overlook these omissions (especially at the initial claim). Correcting such oversights imposes a heavy administrative burden on the SSA, in part due to the need to collect overpayments from beneficiaries when non-Social Security pensions are discovered. Surprisingly, SSA and IRS lawyers have declared that tax-disclosure rules prevent the SSA from using IRS 1099-R data to correctly calculate Social Security benefits, even though accurate benefits directly affect the proper amount of income taxes on those benefits. This compliance problem is thus adversely affecting Social Security's finances.

Now, however, the SSA databases include nearly four decades of non-covered earnings. Given these new data, analysts and policymakers have been thinking about how to more accurately account for non-covered earnings and adjust Social Security benefits directly, without knowledge of pension amounts. I propose to do so by replacing the current WEP with a sliding-scale formula, using the current law's PIA and AIME calculations:

1/3 / (25 * 45 / (covered years + non-covered years)) * (35 – covered years).

In addition, the rule for qualification as a sufficient government pension exempt from Social Security would be amended to require five-year vesting, consistent with IRS rules for private pension plans.

The sliding scale takes into account the fact that retired workers have different extents of covered and non-covered employment. Those with 10 years of covered earnings and 35 years of non-covered earnings would receive the maximum WEP reduction (one-third of their covered PIA), while those with 35 years covered and 10 or fewer years non-covered would receive no WEP reduction. Proportional reductions would apply to workers between these two extremes, while a similar but adjusted formula would apply to disabled workers.

I would also replace the current GPO reduction with a sliding-scale formula:

1 – 1/3 / (25 * 45 / (covered years + non-covered years)) * (35 – non-covered years).

Spouses and survivors with 10 or fewer years of covered earnings and 35 years of non-covered earnings would receive the maximum GPO reduction (100% of the amount of a hypothetical PIA based on the spouse's or survivor's own non-covered earnings), while those with 35 years of covered earnings and 10 years of non-covered earnings would receive the minimum GPO reduction (two-thirds of the amount of a hypothetical PIA based on the spouse's or survivor's own non-covered earnings). Again, proportional reductions would apply between these two extremes.


Finally, to make up for the necessary reduction in scheduled Social Security benefits, low-wage and disabled workers should be able to access a national system of personal retirement accounts. These voluntary accounts should include automatic enrollment (with an opt-out option) for all low-wage workers — including the self-employed — earning up to $40,000 annually. The low-wage worker accounts should be funded by workers' contributions of up to 3% of earnings. The accounts should receive the same tax treatment as Roth IRAs, meaning that while contributions would not be tax deductible, retirees would not have to pay taxes on the savings they use from these accounts.

Drawing from general revenues, the federal government should match worker contributions as follows: $1.50 for each dollar contributed from up to $15,000 in earnings, $1.25 on the contributions arising from the next $5,000 in earnings, and so on, until the match is reduced to $0.25 on the dollar for contributions from earnings between $35,000 and $40,000. Under this scheme, a worker earning $15,000 would contribute $450 annually and receive a $675 match, a worker earning $20,000 would contribute $600 annually and receive an $862.50 match, and so on. The government contribution should vest after three years of continuous earnings. For disabled workers younger than the early retirement age, the government should make all contributions, both worker and matching, from general revenues, in amounts that would occur if the annualized AIME for the beneficiary, price indexed to the current year, was assumed to be the worker's current annual earnings.

The account system used in this process should include a few simple and well-diversified investment vehicles, such as target-date and balanced funds. For retirement, an indexed joint-and-survivor life annuity should be a mandatory requirement for the distribution of funds accumulated through the government match. But for funds accumulated through the worker's contributions, systematic withdrawals for a period of five years beyond the couple's life expectancy should be allowed, which will enable the couple to fund various needs on variable paths during retirement as well as leave bequests. A small expense charge (around 0.3%) based on assets held in account should be levied to pay for ongoing administrative costs, while start-up costs for the system would be paid out of general revenues. The account system should be administered by the SSA.


Any enduring Social Security reform must incorporate conservative and liberal concerns and goals. But it must also ensure sustainable solvency, treat all workers fairly, and adapt to a 21st-century economy, society, and workforce. For reform to be truly effective, it must apply to both the retirement and disability parts of the program.

The proposals laid out here attempt to offer such a framework — one that leaves plenty of room for policymakers to negotiate the details. By pursuing an approach along these lines, Social Security can continue to provide fair and dependable basic support that all Americans expect and deserve.

Mark J. Warshawsky is a senior fellow at the American Enterprise Institute.


from the


A weekly newsletter with free essays from past issues of National Affairs and The Public Interest that shed light on the week's pressing issues.


to your National Affairs subscriber account.

Already a subscriber? Activate your account.


Unlimited access to intelligent essays on the nation’s affairs.

Subscribe to National Affairs.