Retirement income in America runs on Social Security, not IRAs, despite decades of advice telling workers to max out their individual retirement accounts. The 2026 EBRI/Greenwald Retirement Confidence Survey found more than nine in ten retirees rely on Social Security for income, while only about 54% draw from an IRA at all.
The Gap Between What Workers Expect and What Retirees Actually Do
Seventy one percent of workers surveyed expected an IRA to factor into their retirement income. Only 54% of actual retirees reported drawing from one. That 17 point gap between expectation and reality says something about how retirement planning advice diverges from lived outcomes. IRA usage as an income source has also been sliding: 59% of respondents cited it last year, versus 56% this year in earlier survey rounds, before settling near 54% in the 2026 data. Social Security requires nothing from the retiree beyond having worked and paid into the system. No allocation decisions, no rebalancing, no withdrawal rate calculations. The deposit or check simply arrives. That structural simplicity likely explains why it remains the near universal income source even as IRA participation softens.
$19.2 Trillion in IRA Assets, Concentrated in Relatively Few Households
U.S. retirement market assets totaled $49.1 trillion at the end of 2025, according to the 2026 ICI Investment Company Fact Book, and IRAs accounted for $19.2 trillion of that, the single largest category. Yet only about 42% of U.S. households owned an IRA in 2025. The math implies a skewed distribution: a large pool of assets held by a minority of households, many with balances well above the median. That median IRA balance stood at roughly $150,000 as of 2022, a figure that helps reconcile the trillion dollar headline number with the fact that fewer than half of households hold one of these accounts at all.
The ICI framework replaces the old three legged stool model (Social Security, pensions, personal savings) with a five layer pyramid: Social Security at the base, then homeownership, then workplace retirement plans, then IRAs, then other assets at the top. Social Security replaces 91% of lifetime earnings for very low earners but only 32% for the highest earners, meaning the upper layers of the pyramid matter disproportionately to higher income households who need less replacement from the base. IRAs, in that sense, matter most to the people who rely on Social Security the least.
What a $150,000 Balance Actually Produces in Annual Income
Applying the widely used 4% withdrawal rule to a $150,000 median IRA balance yields roughly $6,000 a year. Compare that with the average Social Security retirement benefit in 2025, about $1,976 a month, or roughly $23,712 annually. The IRA, for a typical account holder, supplements Social Security rather than approaching parity with it. Withdrawal behavior confirms that IRAs function as a secondary reserve rather than a primary income stream. At age 59, the first year distributions become penalty free, only 9.5% of IRA owners took a withdrawal. That participation rate climbs gradually, reaching 29.6% by age 72. Then at age 73, when required minimum distributions kick in for most current retirees, the share taking distributions jumps to 76.9%. Roth IRA owners, who face no RMDs during their lifetime, tend to leave balances untouched even longer than that.
Why the Withdrawal Timing Pattern Matters for Planning
The data points to a pattern rather than an anomaly: retirees hold IRA assets in reserve, often deferring withdrawals until the IRS effectively forces the issue through RMDs. That behavior is consistent with IRAs functioning as a supplemental cushion or late life liquidity source rather than a foundational income layer. For financial planning purposes, this means projections built primarily around IRA drawdown for early and mid retirement years may not match how most households actually behave. The practical implication is that Social Security timing decisions (claiming age, spousal benefit strategy) likely deserve more weight in retirement income planning than IRA drawdown modeling for a large share of retirees, given that IRA income remains modest and delayed for the median account holder. IRAs still matter, particularly for the roughly 42% of households that hold meaningful balances, but the aggregate data makes clear they arrive late, pay out relatively little on a median basis, and sit around the edges of a retirement budget still anchored by Social Security.
