What this covers

Mark Warshawsky of George Mason University's Mercatus Center joins Russ Roberts to examine how employer-paid health insurance reshapes the picture of wage inequality in America. The core argument turns on a simple structural fact: employers pay the same dollar amount for health coverage regardless of worker pay level, due to tax law and nondiscrimination rules. This means health insurance represents a much larger slice of compensation for low-wage workers than high-wage ones. When health costs surge—which they have, tripling between 1999 and 2014 according to Kaiser Foundation data—employers face pressure to suppress take-home pay growth. The result is that earnings inequality widens on paper even when total compensation remains flat or shrinks. Warshawsky uses previously unpublished Bureau of Labor Statistics data broken down by percentile to show this effect concretely.

The conversation ranges across the machinery that sustains this imbalance. The tax code's unlimited tax exclusion for employer health benefits, unlike the caps on 401k contributions, allows plans to balloon without triggering income taxation. Income inequality studies typically rely on tax records that exclude non-taxed benefits, creating what Warshawsky calls a systematic blindness to a large share of actual compensation. Between 1992 and 2010, health insurance grew from 7.5% to 11.6% of compensation for low-wage workers, while rising only from 4.5% to 5.7% for the highest earners—a disparity that explains roughly half the measured increase in earnings inequality. Roberts and Warshawsky also explore the harder terrain: whether mandating benefits harms worker freedom, whether Medicaid and charity care represent genuine alternatives, and whether the rising health spending actually delivers commensurate health gains. The discussion concludes that health policy and inequality policy cannot be disentangled, and that addressing health-cost inflation may be the real lever rather than tax-and-transfer responses to measured income gaps.

Sharpest takeaway

Warshawsky argues that because employer-paid health insurance costs the same in dollar terms for low- and high-paid workers and has risen dramatically, measuring inequality by earnings alone overstates rising inequality; total compensation reveals inequality is flat or even falling.

  • Health insurance is a fixed dollar cost per worker (by nondiscrimination rules and custom), so it is a larger share of compensation for low-paid workers
  • Rapidly rising health costs depress measured take-home pay more for low-paid workers, exaggerating earnings inequality
  • BLS employer-cost data show compensation inequality growing less or falling while earnings inequality rises

The argument · threads5 threads · 24 claims
0.85

Health insurance costs mechanically suppress take-home pay growth for lower-paid workers relative to higher-paid workers.

5 pointscentrality 5/5
  • Because health insurance costs the employer the same dollar amount regardless of pay and represents a larger share of compensation for low-paid workers, when health costs rise rapidly the employer must raise take-home pay less for low-paid workers, making it look like low-paid workers fall behind high-paid workers even though total compensation may be unchanged.

    “the employer will increase take-home pay for everybody; but they are going to have to increase it less for the low-paid worker than for the high-paid worker. Because the cost of having the worker has gone up... and that represents more of the total compensation for the low-paid worker”

  • BLS employer-survey data for 1992 show health insurance as a share of compensation was about 7.5% for low-paid workers (30th percentile, roughly minimum wage full-time) but only 4.5% for the highest-paid (99th percentile), because the employer's cost to provide health insurance is roughly the same regardless of pay level.

    “in 1992, the, at the low-income worker, their health insurance as a share of compensation was about 7-and-a-half percent... For the very highest-paid worker, in the 99th percentile... their health insurance as a share of compensation was only 4-and-a-half percent.”

  • By 2010, BLS data show health insurance as a share of compensation rose to 11.6% for low-paid (30th percentile) workers—a 4.1 percentage-point increase—versus 5.7% for the 99th percentile, only a 1.2 point increase, so the percentage-point rise was almost four times larger for low-paid than high-paid workers.

    “in 1992, the, at the low-income worker, their health insurance as a share of compensation was about 7-and-a-half percent... For the very highest-paid worker, in the 99th percentile... their health insurance as a share of compensation was only 4-and-a-half percent.”

  • Except in rare special circumstances such as a regulated monopoly with little competition, employers in a global economy cannot pass mandated benefit costs onto customers, so an increase in mandated benefit cost (x) forces take-home pay down because total compensation is bounded by the worker's marginal product (y).

    “that almost inevitably means lower take-home pay: lower salaries in terms of cash for that worker... benefits aren't free. They come out of the salaries of workers.”

  • Tax-code nondiscrimination requirements, plus custom, mean that low-paid and high-paid workers receive the same health insurance benefit in dollar terms from the employer, so health insurance costs the employer the same regardless of the worker's pay level.

    “in 1992, the, at the low-income worker, their health insurance as a share of compensation was about 7-and-a-half percent... For the very highest-paid worker, in the 99th percentile... their health insurance as a share of compensation was only 4-and-a-half percent.”

0.85

Measured earnings inequality overstates true compensation inequality because standard income data exclude employer-paid health benefits.

5 pointscentrality 5/5
  • Over the longer 1992-2010 period (comparable points in the business cycle), about half the increase in earnings inequality is explained by rising health care costs; 30th-percentile earnings grew 60% but compensation grew 70%, while 99th-percentile earnings grew 78% and compensation 82%, so compensation grew relatively more for lower-paid workers.

    “about half the increase in inequality is explained by the increase in health care costs... in the 30th percentile, earnings growth was 60%; but... compensation... the growth is 70%. Whereas if you look at the 99th percentile, earnings growth was 78% and compensation growth... was 82%.”

  • Compensation inequality may be falling, or at least has not increased nearly as much as earnings and income data show; the broader robust point across time periods and data sources is that the increase in inequality is highly exaggerated when measured by earnings or income rather than total compensation.

    “the increase in inequality is highly exaggerated when we look at just earnings or income as opposed to total compensation.”

  • For most workers, compensation is a package of pay plus benefits, and a major part—especially expensive health insurance—is paid mostly by the employer, but because take-home pay is easy to measure and benefits are not, researchers and workers underestimate true compensation while employers track these costs carefully.

    “for most workers, most of the cost is paid for by the employer. And as we'll discuss, health insurance is very expensive.”

  • Scholars who study income inequality typically focus on income tax data, but that captures only take-home pay and excludes non-taxed compensation like employer health benefits, which essentially do not appear on the tax return, so such studies miss a large share of total compensation.

    “a lot of scholars that look at income inequality basically focus on income tax data. But again, that's take-home pay. It does not include these non-taxed sources of compensation.”

  • Income measurement and tax systems designed in the 1940s-50s, when health insurance was cheap and almost a rounding error, have become increasingly outmoded as ways of understanding people's wellbeing because health costs have grown so rapidly that excluding them now seriously distorts measures of what workers really earn.

    “when you have these costs increasing as rapidly as they have, those systems of measurement become increasingly outmoded as ways of looking at ways of people's wellbeing”

0.65

Health insurance costs have exploded faster than economic growth, creating a fundamental mismatch in employer compensation budgets.

4 pointscentrality 4/5
  • According to Kaiser Foundation data, the average cost to an employer of health insurance tripled between 1999 and 2014—just 15 years—even though incomes and the economy did not triple, creating a massive dislocation in the pay package.

    “Just from 1999 to 2014--so, 15 years--the cost to an employer... of health insurance, so this is what the employer is paying... tripled. That's just in 15 years.”

  • Health care costs are exploding, and there is a strong debate—with some reason to believe—that the rapidly rising spending is not providing a corresponding marginal increase in health and longevity, making it questionable whether the added benefits are worth their rapidly increasing cost at the individual, company, and societal levels.

    “there's some reason to think that they are not providing that marginal benefit, again, particularly when you consider the very rapid increase in cost.”

  • Despite costing nearly 12% of compensation for low-paid workers, employer health insurance persists because it covers a real risk workers value, it is likely cheaper than buying individually (avoiding underwriting), it carries a large tax advantage, the alternatives (Medicaid, charity care) are second-bests, and the cost rose gradually rather than suddenly, muting the reaction.

    “it is still likely to be cheaper to be covered by the employer than it is to go out in the marketplace and buy insurance on your own... it is a tax advantage to get it through your employer as opposed to paying for it on your own.”

  • Because health care is now nearly $1 of every $5 in the economy, inequality and health care policy are so intertwined that ACA repeal and replacement should be discussed together; reform must aggressively and directly address the root cause of measured income inequality—the rising cost of health care—rather than treating it indirectly.

    “as we go forward in design or replacement for the Affordable Care Act, I think we really need to address the root cause of the problem--how the income inequality, which is the rising cost of health care”

0.40

Tax exclusions and non-discrimination rules allow employer health insurance to grow without the limits that apply to other benefits.

2 pointscentrality 3/5
  • Because employer-provided health insurance is excluded from taxable income with no dollar limits (unlike 401k contributions), even very generous plans with no deductibles or copays are entirely tax-free, which is a contributing factor that lets health plans grow seemingly without limit.

    “there are no such limits for health insurance. So, even very, very generous plans--no deductibles, no copays, no employee contribution for the insurance premium... Totally tax free. So, that obviously is a contributing factor to, you know, having these plans and sometimes I think many people argue for them to just grow seemingly without limit.”

  • Because people have different preferences for benefits depending on age, risk attitudes, and willingness to seek care, a one-size-fits-all benefit package is probably not the best solution, even before considering broad societal questions.

    “people have different preferences for these things; and one size fits all is probably not the best solution.”

0.32

Direct BLS data by compensation level provides stronger empirical support than prior indirect analyses of compensation inequality.

3 pointscentrality 2/5
  • Other analysts at the CBO and Brookings reached similar conclusions but with weaker empirical support because they relied on indirect data requiring imputations, smoothing, and inconsistent sources prone to recall error, whereas Warshawsky obtained the same well-respected BLS employer-survey data broken down by compensation level—detail not previously published—giving a direct, high-quality measure.

    “they are using data sources which are... indirect. They have to do some imputations; they have to do some smoothing... those data sources are inferior to what I have, which is the same data source that the Bureau of Labor Statistics uses”

  • If a large part of measured earnings inequality is caused by rising health care costs rather than unfairness or power imbalances in the labor market, then the appropriate policy is to address the cause—the cost of health care—rather than the symptom, which leads to a different set of prescriptions than raising taxes on the high-paid.

    “if a large part of the explanation is the increase in the cost of health care, well, if that's the cause, then you presumably should address the cause and not the symptoms.”

  • The finely-cut BLS data Warshawsky used are not publicly available because cutting the data by compensation level raises the risk of revealing confidential employer information (e.g., identifying a specific company), and the BLS has limited staff—so replication requires others to negotiate access, a balance between transparency and confidentiality.

    “when you cut the data more finely--as you need to, in this exercise, to look at levels of compensation--you have a little bit more risk of revealing the confidential information.”