Originally published on October 3, 2024, this analysis was republished on September 18, 2026, to reflect updated data.
By Benjamin Glasner
The tax-advantaged retirement savings system in the United States is one of the most effective wealth-building programs in the world. Too many working Americans, however, find themselves on the outside looking in. For many workers, access to employer-provided retirement plans is limited and sporadic. For those left without an employer-provided plan, the incentives in the retirement system fail to encourage adequate savings to prepare for the future.
Just how many Americans are left out of the tax-advantaged retirement savings system today? How many workers lack access to an employer-provided retirement savings plan? Who has plan access, but does not participate? And who misses out on employer contributions and matching benefits on the savings they are able to put toward their retirement?
These questions are at the core of the retirement savings debate in America today. Here we lay out the most direct answers to these questions that we can and offer policy ideas to help solve the retirement savings puzzle.[1]
Who lacks access?
About 76 million workers — nearly 52 percent of working Americans ages 18 to 64 — lack access to an employer-provided retirement plan. Among the 132 million Americans who work for an employer, about 61 million (46 percent) lack access to an employer-provided plan.[2] The other 15 million workers without access are self-employed.
Half of workers have no retirement plan at work
The interactive figure requires JavaScript. Please enable it to view the bars, squares, and tooltips.
Within the private sector, where the majority of workers are employed, 49.1 percent of workers lack access to employer-provided plans. For private-sector workers, the access gap is far wider for those engaging in part-time work (77 percent) than full-time work (42 percent).
Separately, 30.2 percent of government workers lack access to an employer-provided retirement plan. These public-sector workers are concentrated in state and local governments rather than federal agencies, and are disproportionately part-time, lower-wage, and younger employees.[3]
Who participates, and who gets a match?
Access to an employer-provided retirement plan is often the first step in setting up a financially-sound future for millions of American workers. Those who lack access, or cycle in and out of access across a variety of employers, face a serious hurdle to building toward a dignified retirement.
But not every plan is the same. A wide range of plan types, characteristics, and incentives exist across the world of work. One valuable plan feature that most American workers lack is an employer match or contribution. In the most recent available sample from the Survey of Income and Program Participation (SIPP), only 37 percent of all workers receive an employer contribution or match to an employer-provided retirement plan.
For workers who receive a contribution or match from their employer, the median amount received is about $3,000 per year. (The mean is about $5,800.)
Gaps by race, ethnicity, and education
Gaps in access, participation, and employer contributions reflect many of the key characteristics associated with Americans’ labor market outcomes and opportunities today, namely race, ethnicity, and education. Let’s look at how access gaps break out across private-sector employees on these dimensions, setting self-employed workers aside.
About 58 percent of Black non-Hispanic workers lack access to employer-provided plans. Among Hispanic workers, the access gap is even higher, at 66 percent. White non-Hispanic and Asian non-Hispanic workers see substantially lower shares lacking access, 42 percent and 37 percent respectively.
The education gradient is even steeper. While 31 percent of employees with a bachelor’s degree or more lack access to an employer-provided retirement plan, 64 percent of those with a high school diploma lack access. That portion grows to 82 percent among those with less than a high school diploma.
Thus, in today’s labor market, a bachelor’s degree is almost a prerequisite to having a better-than-average shot at accessing an employer-provided retirement plan and growing a sizable nest egg for the future.
These access gaps compound existing disparities in earnings and job quality, and feed into a difference in expectations of what retirement will look like.
Income, access, and employer contributions
Importantly, the federal government’s tax treatment of retirement savings is independent of any explicit racial, ethnic, or educational test. Where tax-advantaged retirement savings really differ among households is across measures of income.
One of the key benefits of SIPP over the BLS’s National Compensation Survey is that it provides rich detail on individual and household incomes. This allows us to assess how employer-provided plan access varies across the income distribution.
We can see clearly who is missing out. Unsurprisingly, the gap between high and low earners is stark.
In the lowest earnings decile, about 88 percent of private-sector employees lack access to an employer-provided retirement plan, 91 percent do not participate, and 91 percent receive no employer match or contribution. Among the top decile of earners, those figures fall to about 14, 18, and 24 percent respectively.[4]
At high income levels, meaningful portions of the workforce participate in employer-provided retirement savings plans even if they do not receive an employer contribution.
In fact, among private-sector employees who receive no employer contribution, those in the top half of the earnings distribution are about five times more likely to participate in an employer-provided retirement plan than those in the bottom half (19.9 percent versus 3.9 percent). The top decile is 27 times more likely than the lowest earnings decile to participate.[5]
High-earning workers do not face the same liquidity constraints to saving that lower-earning workers do. That means that higher-earning workers can also more easily access the main federal tax benefit associated with retirement saving: deferring one’s income tax bill for retirement contributions until old age, when income will likely fall into a lower tax bracket.
Expanding access to retirement plans for workers at the bottom of the income distribution is a necessary first step in addressing the future of retirement, but access alone will have a limited impact if it is not paired with a match or additional contribution of some variety. When we consider workers under the tightest liquidity constraints, we need to find ways to significantly improve the value of saving today to support living standards tomorrow.
What the tax code pays for
These differences in access, participation, and employer contributions are a barrier to the long-run health of retirement savings across the country, particularly for the lowest-income workers and their families.
Some may look at this as a natural element of labor-market differentiation. Both businesses and workers are making decisions about where they work, how they spend, and how they save, so gaps that result may reflect a range of preferences.
The problem is that the government is not neutral on this.
The tax code does not treat retirement savings as a cash-equivalent source of compensation — it incentivizes retirement savings. In effect, taxpayers subsidize the retirement savings of high earners through foregone collection of tax revenue. Using the most recent SIPP data, we estimate that tax-advantaged retirement savings through income and payroll taxes cost between $256 billion and $356 billion in forgone federal revenue, measured on a present-value basis for savings made in 2024.[6]
When foregone tax revenue is understood as spending, subsidized retirement savings ranks among the largest expenditures in the federal budget.
That spending flows predominantly through accounts that half of workers lack access to via an employer-provided plan. The 48 percent of workers who are fortunate to have access to an employer-provided retirement account capture 97.5 percent of the tax benefit in our estimates,[7] meaning that the 52 percent of workers without access capture just 2.5 percent. The average worker with access collected $3,394 in 2024 while the average worker without access to an employer-provided retirement account saw a benefit of just $81. Only 3.7 percent of workers lacking access to employer-provided plans receive any tax benefit at all. This small portion of workers does so primarily through private Individual Retirement Accounts.
The mean benefit captured by bottom-decile earners from forgone tax revenue is $92 per year. The mean benefit captured by top-decile earners is $5,084. The bottom half of workers combined receives only about 10 percent of all dollars spent on tax-advantaged retirement accounts.
If we break this benefit out by worker characteristics, we see that workers with a bachelor’s degree or more capture 73 percent. We can also see the gap play out along racial and ethnic boundaries, similar to the access, participation, and employer contribution gaps: $2,763 for Asian (non-Hispanic), $1,923 for White (non-Hispanic), $1,260 for Black (non-Hispanic), and $907 for Hispanic workers.[8]
The tax code pays workers to save, but for many that requires (1) an easy on-ramp to the system or (2) sufficient knowledge to use the system, plus enough liquidity to capture the benefits.
In part, these disparities in subsidized saving help motivate the Saver’s Match design, which can redirect federal support toward the workers the current design misses.
The Saver’s Match
Beginning in 2027, the Saver’s Match (enacted in the SECURE 2.0 Act) replaces the nonrefundable Saver’s Credit with a refundable government match of up to $1,000 (half of the first $2,000 a worker contributes) deposited directly into a retirement account, and phased out by income and filing status.[9]
About 30 million workers (20.1 percent) are income-eligible for a Saver’s Match — but roughly 21 million of them, about seven in ten, lack a qualifying account (a 401(k)-type account or traditional IRA) to use and facilitate the receipt of the Saver’s Match. This puts the Saver’s Match, as currently designed, in a difficult place. The wedge between eligibility and access is not something it is currently built to overcome, and will likely hinder the effectiveness of the policy for building wealth among the lowest-income working Americans.
In April 2026, a White House executive order established TrumpIRA.gov, with the aim of connecting workers without an employer plan to low-cost individual retirement accounts modeled after the federal Thrift Savings Plan (TSP) that are eligible for the Saver’s Match.[10] While these plans will not be able to automatically enroll individuals without further action from Congress, it is possible that TrumpIRA.gov will meaningfully help workers currently struggling to save for retirement access much-needed, user-friendly retirement savings vehicles that can also facilitate access to the Saver’s Match.
So what are the facts?
America’s tax-advantaged retirement savings system is a powerful wealth-builder, but it is not one to which all workers have access. If we can find ways to close the access gap, we can help tens of millions of workers reach a dignified retirement as a just reward for years of work. How many could be helped?
55 million private-sector employees — nearly half — are left without access to an employer-provided retirement plan.
41 million private-sector employees without a bachelor’s degree — about six in ten — have no employer-provided retirement plan.
21 million low- and moderate-income workers lack a retirement savings plan that would allow them to access the Saver’s Match that they are eligible for — about seven in ten.
Of the hundreds of billions of dollars we spend in the form of forgone tax revenue to subsidize retirement, 72 percent flows to households in the top 30 percent of the income distribution, 55 percent to the top 20 percent, and 33 percent to the top 10 percent.
The problem is clear. The task now is to decide how we solve it.
Appendix: Additional figures and tables
Appendix: A technical walk-through
This portion of the draft is intended for those looking to assess the details of what we measured, the Survey of Income and Program Participation questions behind each statistic we used, and what our particular choices and their limits are.
In an attempt to ease the reading of this more technical section, it is organized as a set of questions I believe a reader is likely to ask. The full code, data documentation, and reproducible outputs can be found in the project’s replication repository on GitHub.[11]
Why do we use SIPP rather than another survey?
A number of major surveys measure retirement-plan coverage. For example: the National Compensation Survey (NCS), the Current Population Survey (CPS), the Survey of Consumer Finances (SCF), and the Survey of Income and Program Participation (SIPP). We use SIPP because it best combines worker-level detail on all three of our pillars (access, participation, and employer contributions). SIPP also oversamples low-income households while still capturing detailed income information across different sources, as well as information on program participation. This is essential for describing the workers the retirement system leaves out.
The estimates in this analysis rest on about 12,300 sample respondents who, after weighting, represent 147.3 million workers.
Why are these access gaps larger than the ones I see from the National Compensation Survey?
Two things are the most likely explanations.
First, SIPP is self-reported and relies on an individual’s ability to answer questions correctly about their own circumstances. Within SIPP, workers are asked to describe their own plan coverage, and this can lead to some under-reporting of eligibility, particularly around plans they may not know they have access to.
Second, the specific questions asked of SIPP respondents differ from what is asked through the NCS, which surveys establishments. The NCS puts the private-sector lack-of-access rate closer to 28 percent, against the SIPP’s 49 percent.[12] The NCS records an offer-based definition of access (whether an employer makes a plan available to a group of employees). Since it is sampled at the establishment level, that is also how access is measured. If you are employed at an establishment that reports offering a plan, you are said to have access.
That differs from our definition in the SIPP, which focuses on whether a plan exists, if an individual worker is personally eligible for a plan, and if they are aware of it.
The NCS definition sits closer to whether a plan exists at the firm at all than to our headline test of a worker being both offered a plan and included in it.
The NCS is also narrower in coverage and in detail: its private-sector series is built around jobs at surveyed establishments and excludes the self-employed, so it does not describe the full all-worker population this piece is concerned with. Also, the NCS is reported by employers rather than by workers. That means it cannot break access down by the worker-level characteristics like income, race, and education.
This also gets particularly difficult to parse when we dig into the sampling and design details. The NCS samples at the establishment level, often a single physical location, and each establishment is counted as a separate entity even if it belongs to a larger firm. Firms can be composed of multiple physical locations, and therefore multiple establishments.[13] A retirement plan, by contrast, is sponsored, and its minimum-coverage rules are tested, at the level of the employer as a whole, i.e., the firm.
Retirement plan policies generally are applied across the firm and its affiliated entities under common control.[14] Because a firm-wide plan may lawfully exclude certain classes of workers and still be offered, an establishment can report a plan as available even to workers who are not themselves eligible.
That can inflate the numbers as reported by the NCS. Like with many things, the “truth” is likely somewhere between the NCS and SIPP.
Who counts as a “worker” in these numbers?
The civilian working population is everyone ages 18 to 64 who held a job in the December 2024 reference month.[15] We define full-time work as 35 or more usual weekly hours on the primary job.
For our analysis, workers are sorted into private-sector, government, and self-employed classes from the SIPP class-of-worker item.[16] We exclude active-duty military from this analysis, and due to the household-level effects on tax filing and tax-advantaged retirement savings, we drop every household that contains an active-duty member. Our sample sums to a total of 147.3 million workers, all weighted by the SIPP final person weight.
What point in time do the numbers describe?
Our measures of access, participation, and matching are all derived from questions asked in the December reference month. This restriction means we are unable to assess seasonal patterns and miss information on those employed during other points in the year. Income and earnings, on the other hand, are full calendar-year 2024 totals. SIPP collects each person’s monthly records, allowing us to simply sum across months.
Months in which a person held no job count as zero labor income. But when someone is present in the survey for fewer than twelve months (because they entered or left the sample mid-year) their observed monthly income values are scaled up to a full year.
Why are the self-employed counted as lacking access?
By definition, the self-employed have no employer to sponsor a plan, so counting them among those “lacking employer-provided access” is literally correct but can inflate an all-worker gap. We therefore report the number as a whole, but focus the analysis on the employed subset.
Of the 76.2 million workers who lack access to an employer-provided retirement plan, 61.1 million are employees whose employer offers nothing or excludes them. This places the employee subset employer-access gap at 46.2 percent of the 132 million who work for an employer.
Access, in this analysis, means a worker’s employer sponsors a retirement plan for which the worker is eligible, or the worker already holds a plan through that employer. Gathering that information using SIPP requires routing each worker down one of two mutually exclusive branches.
The retirement module asks every in-universe respondent which types of retirement plan they have, from any source (the current job, a former job, or on their own): a 401(k)-type account (EOWN_THR401), an IRA or Keogh (EOWN_IRAKEO), or a defined-benefit or cash-balance pension (EOWN_PENSION). Holding one of these plans is the gate that decides which branch a worker is sent down.
Branch A — A worker who owns an account type and holds a December job is asked whether that account is provided through the main employer:
“Any [401k, 403b, 503b, or Thrift Savings Plan / IRA or Keogh] account(s) — or [defined-benefit or cash balance] plan(s) — provided through main employer or business during the reference period.” (EMJOB_401, EMJOB_IRA, EMJOB_PEN)
Universe: only workers who own the corresponding account type (EOWN_* equal to yes) and who held a job in the December reference month. A “yes” here establishes access (the worker holds a plan through the current employer) and that worker is then skipped past the offer questions below, so the two branches never overlap.
Branch B — A worker who does not report a plan through the main employer, because they own no account or own only a personal or former-employer account, is asked whether the employer offers a plan at all, and, if so, whether they are included in it:
“Did … main employer or business have any kind of pension or retirement plans for anyone in the company or organization?” (EPENSNYN)
Universe: workers age 15 and older who held a December job and did not report a plan through their main employer — that is, who answered “no” to, or were not asked, all three of the EMJOB_* questions above.
“Was … included in the pension or retirement plan(s) offered by … main employer or business?” (EINCPENS)
Universe: only workers whose employer offers a plan (those who answered “yes” to EPENSNYN).
A worker has access if either branch is satisfied. If they hold an employer plan (Branch A), or their employer offers one and they are included in it (Branch B — a “yes” to EINCPENS), they are said to have access. A worker lacking access is one whose employer sponsors nothing (a “no” to EPENSNYN) or sponsors a plan but excludes them (a “no” to EINCPENS), possibly due to hours or tenure rules, for example.
What does “participation” mean?
Participation is treated as a worker-level decision conditional on access. It is defined as contributing to a plan held through a worker’s current employer. If a worker lacks access to an employer-provided account, they cannot participate in it. The three SIPP contribution questions are asked only of workers who hold that plan type through their main employer (the Branch A group from the access question) and cover all three plan types:
“During the reference period, respondent contributed to the [401k, 403b, 503b, or Thrift Savings Plan / IRA or Keogh] account(s) — or [defined-benefit or cash balance] plan(s) — provided through their main employer or business.” (ESCNTYN_401, ESCNTYN_IRA, ESCNTYN_PEN)
Universe: for each item, only workers who hold the corresponding plan through the main employer — EMJOB_401 = 1 for the 401(k)-type question, EMJOB_IRA = 1 for the IRA or Keogh question, and EMJOB_PEN = 1 for the pension question. A worker’s own saving in a personal IRA outside of work is out of scope and not counted as participation.
On this basis, about 92 percent of workers with observed access participate, and roughly 43 percent of all workers participate.[17]
What does an employer contribution or match mean?
The receipt of an employer contribution into the worker’s account can take many forms. It could be a fixed contribution amount or share of income. It could also be a designated match on employee-side contributions. The exact details of the policy are not defined in the SIPP question. The question on employer contributions was also only asked of workers who hold that account type through their main employer:
“Main employer or business contributed to respondent’s [401k, 403b, 503b, or Thrift Savings Plan / IRA or Keogh] account(s) during the reference period.” (EECNTYN_401, EECNTYN_IRA)
Universe: EMJOB_401 = 1 for the 401(k)-type question and EMJOB_IRA = 1 for the IRA or Keogh question.
SIPP asks no employer-contribution question for defined-benefit or cash-balance pensions. There is no pension counterpart to these two items, so employer contributions to traditional pensions are not observed here. We focus on defined-contribution and IRA-type concepts.
SIPP also records the dollar amount of the employer contribution:
“Total amount main employer or business contributed to respondent’s retirement plan(s) during the reference period.” (TECNTAMT)
Universe: workers who hold a 401(k)-type or IRA or Keogh account through the main employer (EMJOB_401 = 1 or EMJOB_IRA = 1)
This amount is what lets us report a median employer contribution of about $3,000 per year among those who receive one.
Does “plan” mean a 401(k)-style account, or something broader?
In the context of this piece, plan means something broader.
The offer-and-inclusion questions ask about a “pension or retirement plan” of any kind, including defined-benefit and cash-balance pensions (EMJOB_PEN).
Eligibility for a traditional pension counts as having access, because that is still a type of employer-provided retirement plan. This mixture of DB and DC plans is the main reason the government access gap looks small. Many public workers are covered by defined-benefit pensions. If we only want to look at savings account plans, like 401(k)s or IRAs, then we would need to use the account-ownership variables.
How do defined-benefit pensions fit across the three pillars?
A defined-benefit pension is counted in the access question. The offer-and-inclusion questions ask about a “pension or retirement plan” of any kind, allowing us to “catch” both DB and DC plans. One of the SIPP variables (EMJOB_PEN) is specifically for defined-benefit and cash-balance plans. These plans are included in the share with access.
Participation and employer contribution concepts are instead built around defined-contribution plans.
Participation is defined as contributing to a plan. Traditional pensions tend to be employer-funded, meaning that fairly few covered workers “contribute.”
In our analysis, a pension-covered worker who does not “contribute” is counted as having access, but not participating.
Employer contributions and matching are even narrower. SIPP records an employer contribution only for 401(k)-type and IRA or Keogh accounts, so the employer funding of a defined-benefit pension is not observed.
Due to the prevalence of defined-benefit pensions in the public sector, this analysis focuses on the private sector of employed workers.
How precise are these estimates?
All estimates are weighted to the December 2024 population with the SIPP final person weight. We show 95 percent confidence intervals on a number of estimates, which are built from SIPP’s 240 replicate weights and method, using the 2025 SIPP Users’ Guide.[18]
What do we mean by the “cost” of the retirement tax breaks, and how did we estimate it?
This analysis attempts to replicate the Congressional Budget Office’s (CBO) present-value method as best as possible given data constraints and available documentation.
For each worker, we
- Take the retirement contributions observed in calendar-year 2024 (both the worker’s own and the employer’s), and
- Compute the present value of the federal income and payroll taxes that are foregone on those contributions.
This requires us to build a tax unit for each worker and simulate their federal tax liability twice, once with the contributions and once without, and take the difference. The tax calculations run through the PolicyEngine-US microsimulation model under 2024 tax law.[19]
Where does the $256 billion to $356 billion range come from?
The range reflects the rate of return on retirement balances. Following CBO, the base estimate assumes balances earn a 3.5 percent return and are drawn down in equal installments between ages 65 and 85, taxed at the same marginal rate that applied in the year of contribution; that yields $256 billion. A 6 percent return, reflecting the higher-rate environment since 2022, raises the cost to $356 billion. The Treasury’s Office of Tax Analysis publishes its own tax-expenditure estimates on this same present-value basis.[20]
What had to be estimated rather than directly observed?
SIPP does not observe three key elements.
First, employer contributions to defined-benefit pensions. SIPP does not record these directly, so this exercise requires an allocated national normal-cost total from the BEA National Income and Product Accounts, specifically across the workers who report a defined-benefit plan in proportion to their earnings.
Second, we need some information on the share of 401(k) plans that are Roth. We draw data from Vanguard and the IRS to answer this.
Third, personal IRA contributions made outside of work. We need a sense of the total picture of savings as this exercise is inclusive of non-employer-provided retirement account savings. To do this, we use data from the IRS Statistics of Income totals by income class and assign them to a population that matches IRS counts of actual contributors.
All other observed employee contributions are treated as pre-tax deferrals.[21]
How do these numbers compare with official government estimates?
We got fairly close to the most recent CBO estimates of the combined income and payroll tax expenditure for 2019, $276 billion. Our base estimate for 2024 was $256 billion.[22]
There are a number of reasons why we may expect the level to differ when using SIPP data for this, but importantly, the distributional component of the analysis aligns very well. The top income quintile receives 60 to 62 percent of the income-tax expenditure in our estimate, against 63 percent in CBO’s.
How did we decide who is eligible for the Saver’s Match?
The Saver’s Match matches 50 percent of the first $2,000 a worker contributes, up to $1,000, phasing out with income by filing status. In order to assess eligibility for the Saver’s Match, we need to make some projections based on income data in the year 2024.
There are five reasons that we are working with an approximation.
First, we use a SIPP income measure as a proxy for adjusted gross income, which the survey does not report directly, and we apply the statutory thresholds unindexed.
Second, filing status comes from what SIPP records for the prior tax year, which means that we are assuming a consistency in filing status; inconsistency could shift eligibility for the Saver’s Match.
Third, following the statute as written, we exclude full-time students and workers observed as claimed dependents. Unfortunately, SIPP observes dependency for tax filing status only within a limited age range. This leaves some noise in the analysis. For joint filers, we combine the couple’s income through SIPP’s spouse pointer, but it is possible that not every spouse gets matched. When a spouse cannot be matched, which occurs for about 4 percent of joint filers, only the worker’s own income is used. This moderately inflates eligibility.
Fourth, workers with no recorded prior-year filing status (about 9 percent of workers) cannot be placed on the phase-out schedule and are counted as not eligible.
Fifth, workers with negative annual income are counted as not eligible even though a negative AGI would qualify under the statute, and the full-time-student exclusion is observable only for post-secondary enrollment.
What is the “eligible but locked out” wedge?
It is the gap between who qualifies for the match on income and who can actually receive it. About 29.6 million workers (20.1 percent) are income-eligible, but roughly 21.0 million of them — about seven in ten — hold no qualifying account to receive the match into. That leaves about 8.7 million who are both eligible and positioned to claim it. This wedge is the central Saver’s Match finding: eligibility does not equal access.
Notes
- Given the upcoming implementation of the Saver’s Match — and the growing attention on the state of retirement savings from the White House — we focus on all workers here: private-sector, government, and self-employed. We use data from the Survey of Income and Program Participation (SIPP) published by the Census Bureau to get as up-to-date an analysis as possible (U.S. Census Bureau, Survey of Income and Program Participation, 2025 release (pooled 2022–2025 panels; December 2024 reference month, reference year 2024). https://www.census.gov/programs-surveys/sipp.html. Estimates are weighted using the final person weight). A detailed explanation for why we use SIPP over other sources of retirement data can be found in the appendix at the end of this article.[↩]
- Employed workers ages 18 to 64 in the December reference month of the Survey of Income and Program Participation, 2025 release (pooled 2022–2025 panels). Access is defined as the worker’s employer sponsoring a plan and the worker being eligible to participate in it.[↩]
- Government workers number 19.9 million of the 147.3 million-worker population in the sample; 6.0 million (30.2 percent) lack access, about 8 percent of all workers without access. The gap is most pronounced among state and local government employees, who make up 85 percent of government employees lacking access. Whereas 34 percent of local government employees and 30 percent of state government employees lack access, only 23 percent of federal employees lack access. That access gap is higher than anticipated and is likely a mixture of misreporting of the class of worker question, the inclusion of USPS workers, and real incorrect reporting among federal employees. Within government, the gap concentrates on the same margins seen elsewhere: part-time workers lack access at 66 percent versus 24 percent for full-time; the rate falls from about 81 percent in the lowest all-worker earnings decile to about one in ten near the top; and it is highest for the youngest (68 percent for ages 18–24) and least-educated (50 percent for those without a high school diploma) workers. By headcount, most government workers without access are nonetheless full-time (66 percent), because the public workforce is overwhelmingly full-time. The overall government gap stays small relative to the private sector (30 versus 49 percent) largely because SIPP counts defined-benefit pensions as access (see appendix).[↩]
- Deciles here are tenths of individual annual earnings among all workers, applied to the private-sector employees shown.[↩]
- The underlying rates are 34.9 percent (top decile) and 1.3 percent (bottom decile). The bottom-decile rate is calculated from a very small sample of 11 unweighted participants. That means the ratio’s replicate-weight 95 percent confidence interval spans from 7 to 47.[↩]
- EIG analysis of the SIPP 2025 release (pooled 2022–2025 panels), applying the Congressional Budget Office’s present-value method to calendar-year 2024 contributions (federal income plus payroll tax); see the appendix for method. Under CBO’s parameters (a 3.5 percent return), the combined expenditure is $256 billion ($255.7 billion, standard error $6.9 billion); under a 6 percent return consistent with the post-2022 rate environment, it rises to $356 billion (standard error $10.4 billion). For comparison, CBO estimated $276 billion for 2019 ($202 billion income tax and $74 billion payroll tax): Congressional Budget Office, “The Distribution of Major Tax Expenditures in 2019,” October 2021 (report no. 57413). https://www.cbo.gov/publication/57413. The distribution replicates CBO’s: the top income quintile receives 60 to 62 percent of the income-tax expenditure here, versus 63 percent in CBO’s 2019 estimate.[↩]
- This result is built on workers aged 18 to 64. These workers account for 97 percent of the total estimated expenditure and capture the bulk of the relationship.[↩]
- The gradients extend across job types: government workers average $4,275 per year (34.4 percent of the dollars on 13.5 percent of workers), compared with $1,396 for private-sector and $377 for self-employed workers; full-time workers average $2,080, against $301 for part-time workers. The government figure is dominated by the imputed value of employer defined-benefit accruals: SIPP does not observe employer defined-benefit contributions, so national normal-cost totals were allocated across reported defined-benefit participants (see the appendix). Replicate-weight standard errors for the headline cross-cuts: mean benefit with employer access $3,394 (SE $87) versus $81 (SE $11) without; the with-access share of subsidy dollars 97.5 percent (SE 0.3 points); share benefiting without access 3.7 percent (SE 0.3 points); mean benefit for match recipients $3,465 (SE $91). Personal IRA contributions made outside work are not observed in SIPP and are imputed: dollars are assigned to a population matching IRS counts of actual IRA contributors (about 15 million taxpayers in 2023), prioritizing owners with the largest balances, because administrative data show most IRA owners contribute nothing in a given year. The assumption-free alternative — spreading the same dollars across all IRA owners — would raise the share of access-lacking workers receiving any benefit to about 14.5 percent while leaving aggregate totals and decile shares essentially unchanged.[↩]
- SECURE 2.0 Act of 2022, § 103, codified at 26 U.S.C. § 6433. Effective for taxable years beginning after 2026.[↩]
- Retirement Savings for Americans Act of 2025 (S. 1526 / H.R. 2696, 119th Congress); White House Presidential Action establishing TrumpIRA.gov, April 30, 2026. https://www.whitehouse.gov/presidential-actions/2026/04/promoting-retirement-savings-access-for-american-workers-by-establishing-trumpira-gov/ (accessed July 10, 2026).[↩]
- EIG-Retirement-Fast-Facts repository (link).[↩]
- U.S. Bureau of Labor Statistics, National Compensation Survey: Employee Benefits in the United States, March 2024. https://www.bls.gov/ncs/ebs/. Cited for comparison; not computed in this analysis.[↩]
- U.S. Bureau of Labor Statistics, “National Compensation Measures: Concepts,” Handbook of Methods, last modified December 15, 2017. https://www.bls.gov/opub/hom/ncs/concepts.htm (accessed July 16, 2026). The NCS defines an establishment as “a single economic unit … usually at a single physical location” and, when it is owned by a larger entity, counts only the sampled location; it defines access as a benefit plan that is “available for [employees’] use.” The civilian-worker scope excludes the self-employed, federal workers, and the military.[↩]
- A tax-advantaged retirement plan’s minimum-coverage requirement is applied at the level of the employer, and entities under common control or in an affiliated service group are treated as a single employer for that purpose. 26 U.S.C. § 410(b); 26 U.S.C. § 414(b), (c), (m).[↩]
- SIPP monthly employment recode RMESR, employed categories; a work arrangement on any of the six December job lines, EJB1-6_JBORSE.[↩]
- EJB1-6_CLWRK, taken from the first job line that reports one.[↩]
- If we only gather a worker’s access via the offer-and-inclusion questions then we lack information on contribution items. If we count all of them as non-participants then that sets a conservative floor of 86 percent.[↩]
- Standard errors use the SIPP 2025 replicate weights via Fay’s modified balanced repeated replication (perturbation factor k = 0.5), per the 2025 SIPP Users’ Guide, section 7.2.3.[↩]
- Daniel Feenberg and Elisabeth Coutts, “An Introduction to the TAXSIM Model,” Journal of Policy Analysis and Management 12, no. 1, 1993 (TAXSIM version 35). The tax-expenditure estimates in this analysis are computed with PolicyEngine-US version 1.772.0, an open-source microsimulation model of U.S. federal and state tax-benefit law (https://github.com/PolicyEngine/policyengine-us); it replaces TAXSIM, which supports tax years only through 2023. TAXSIM is retained as a regression cross-check: on a dual run of the same tax units under 2023 law (the latest year both engines cover), the two engines agree on the aggregate tax-expenditure legs within about 2.3 percent, with a per-unit correlation of 0.999. This analysis simulates federal tax only.[↩]
- U.S. Department of the Treasury, Office of Tax Analysis, “Tax Expenditures FY2027,” December 16, 2025. https://home.treasury.gov/system/files/131/Tax-Expenditures-FY2027.pdf (accessed July 11, 2026).[↩]
- Beyond these three imputations, we also make a few other simplifications for the simulation. We do not model married-filing-separately returns explicitly. Also, qualifying-relative dependents are not modeled. For the payroll-tax leg, we exclude the 0.9 percent Additional Medicare Tax. SIPP dividend income is treated as qualified and pension income as fully taxable. Our estimates are likely biased downward relative to CBO’s and Treasury’s figures since the SIPP surveys understate top incomes relative to tax returns.[↩]
- EIG analysis of the SIPP 2025 release (pooled 2022–2025 panels), applying the Congressional Budget Office’s present-value method to calendar-year 2024 contributions (federal income plus payroll tax); see the appendix for method. Under CBO’s parameters (a 3.5 percent return), the combined expenditure is $256 billion ($255.7 billion, standard error $6.9 billion); under a 6 percent return consistent with the post-2022 rate environment, it rises to $356 billion (standard error $10.4 billion). For comparison, CBO estimated $276 billion for 2019 ($202 billion income tax and $74 billion payroll tax): Congressional Budget Office, “The Distribution of Major Tax Expenditures in 2019,” October 2021 (report no. 57413). https://www.cbo.gov/publication/57413. The distribution replicates CBO’s: the top income quintile receives 60 to 62 percent of the income-tax expenditure here, versus 63 percent in CBO’s 2019 estimate.[↩]
