Few minimum wage numbers have been quoted as often, or as selectively, as the ones produced by the Congressional Budget Office. Supporters of a $15 federal minimum point to the roughly 0.9 to 1.3 million people the CBO expected to rise above the poverty line. Opponents point to the million-plus jobs the same reports expected to disappear. Both sides are reading the same tables. They are just stopping at different rows.

This piece tries to read the whole table. It covers what the CBO actually estimated in its July 2019 and February 2021 analyses of the Raise the Wage Act, how the agency framed the tradeoff, what it said about the federal budget, and, maybe most usefully, how a careful reader should treat a "central estimate" that sits inside a very wide range.

Why the CBO's view carries weight

The CBO is not a think tank with a house view on labor markets. It is the office Congress relies on to put numbers on legislation, and it has every institutional reason to avoid the extremes. It does not run its own field experiments. Instead, it surveys the published research, settles on a set of assumptions about how responsive employment is to wage increases, and runs those assumptions through a model of the low-wage workforce.

That makes its estimates something like a weighted reading of the economics literature at a point in time. It is neither the last word nor a partisan talking point. It is a structured, transparent guess by people whose job is to be boring. That is precisely why both sides cite it, and why it is worth reading carefully rather than in headline form.

The 2019 report: a range, not a verdict

In July 2019, the CBO published The Effects on Employment and Family Income of Increasing the Federal Minimum Wage. It examined three options, the most prominent being the Raise the Wage Act of 2019, which would have lifted the federal minimum in steps to $15 an hour by 2025 and indexed it to median wages afterward.

The headline findings for the $15 option, in 2025, ran roughly as follows:

  • Pay: about 17 million workers who would otherwise earn less than $15 would see their wages rise, and millions more earning slightly above $15 might also see increases as employers adjusted pay scales.
  • Jobs: the CBO's median estimate was that about 1.3 million workers would be jobless who would otherwise have been employed. Its likely range ran from close to zero up to about 3.7 million.
  • Poverty: about 1.3 million people would be lifted out of poverty.
  • Income overall: the agency projected a small net decline in total real family income, because the gains to workers who got raises were offset by lost earnings, higher prices, and lower business income.

That last point is the one people tend to skip. In the CBO's accounting, a $15 minimum did not create new income so much as move it around. Low-wage workers who kept their jobs gained. Workers who lost jobs lost a great deal. Business owners and consumers paid through lower profits and higher prices. The effects on families also depended on where they sat: the agency expected families near the bottom to gain on balance, and families higher up to lose a little.

What "likely range" means The CBO defines its likely range as the band within which it believes there is roughly a two-thirds chance the true effect falls. That leaves about a one-in-three chance the outcome lands outside it, in either direction. A range from "about zero" to several million jobs is not a hedge; it is an honest statement that the research does not pin this down.

The February 2021 update

In February 2021, with Democrats considering whether to include a $15 minimum wage in the pandemic relief package passed through budget reconciliation, the CBO published The Budgetary Effects of the Raise the Wage Act of 2021. The bill would have raised the federal minimum to $15 by 2025 and phased out the lower subminimum wages for tipped workers, youth, and workers with disabilities.

The new estimates landed close to the old ones:

  • Jobs: employment would be lower by about 1.4 million workers once the increase was fully in effect, with a likely range from very little change to about 2.7 million.
  • Poverty: about 0.9 million people would move out of poverty.
  • Pay: again, roughly 17 million workers would see raises, with additional spillover to workers just above the new floor.
  • Deficit: the bill would increase federal deficits by about $54 billion over the 2021–2031 window.

Notice that the upper end of the job-loss range came down from 2019, and the CBO said it had revised its method to reflect more recent research, including work finding smaller employment effects than older studies. Yet the central estimate barely moved. That is worth sitting with. Even after giving more weight to the newer, more optimistic literature, the agency's best single guess was still more than a million jobs.

The poverty estimate also fell, from about 1.3 million to about 0.9 million. Part of the reason is that the baseline had changed: by 2021, many states and cities had already raised their own minimums, so a federal $15 floor would bind on fewer workers in fewer places than it would have a few years earlier.

The CBO's two reports do not say a $15 minimum is a disaster. They say it is a trade: raises for many, lost work for some, and a real chance that the losses are larger than the midpoint suggests.

How to read a central estimate with wide uncertainty

The single most common mistake in discussing these reports is treating "1.4 million" as a forecast. It is better understood as the middle of a distribution. A few practical rules help.

First, look at the shape of the range, not just the midpoint. In both reports the lower bound sat near zero and the upper bound sat well above the central figure. That asymmetry matters. It means the CBO saw a plausible scenario in which job losses were negligible, and another plausible scenario in which they were roughly double the headline number. A policymaker who cares about downside risk should weigh the upper end, not just the middle.

Second, remember what the range is measuring. The spread mostly reflects disagreement in the research about how sensitive employment is to wage floors, plus uncertainty about how wages would grow anyway. It is not a statistical confidence interval from a single study. It is the CBO's judgment about how much the economics profession genuinely disagrees.

Third, a net job number hides who loses. A loss of 1.4 million jobs in an economy of well over 150 million workers sounds small in percentage terms. But those losses do not spread evenly. They concentrate among the least experienced, least skilled, and lowest-paid workers, and in lower-wage regions where $15 sits far above the local median wage. The CBO itself noted that effects would be larger where wages are lower.

Fourth, "jobs lost" and "people lifted from poverty" are not the same people. The workers who gain raises and the workers who lose jobs overlap in demographics but not in outcomes. Netting one against the other as if they cancel out is a mistake. A family that crosses the poverty line and a teenager who never gets hired are both real, and neither number tells you about the other.

Where the evidence is genuinely contested Serious economists disagree about the CBO's numbers in both directions. Researchers in the tradition of Card and Krueger, and more recent work such as Cengiz, Dube, Lindner, and Zipperer (2019), find employment effects of past minimum wage increases that are small or hard to detect, and some argued the CBO still leaned too heavily on older studies. Others, including David Neumark and coauthors, read the full body of research as pointing more consistently toward job losses for low-skilled workers. It is also true that most of the evidence comes from increases far smaller, relative to local wages, than a nationwide $15 floor would have been in many states. That cuts both ways: past studies may not tell us much about a jump that large.

The deficit side: why a wage law has a budget score

It surprises some readers that a minimum wage, which costs the government nothing directly, would add to the deficit at all. The CBO's reasoning ran through several channels. Higher wages raise the prices of goods and services, including some the federal government pays for, notably health care purchased through programs like Medicare and Medicaid. The agency also expected changes in interest rates and in spending on programs such as unemployment compensation. On the other side of the ledger, some workers would earn more and rely less on certain income-tested benefits.

On net, the CBO found the costs outweighed the savings by roughly $54 billion over ten years. In the context of the federal budget, that is a modest figure. Its larger significance was procedural: because the bill was being considered through reconciliation, its budget effects had to be more than incidental to its purpose under the Senate's Byrd rule. The Senate parliamentarian advised that the provision did not meet that test, and the $15 minimum was dropped from what became the American Rescue Plan Act.

For our purposes, the budget score is a reminder that wage mandates are not free. Their costs simply arrive through prices, hours, and hiring decisions instead of an appropriations line.

What it means for the argument

We argue on this site that wage floors are a blunt instrument, and that direct income support for low-wage workers, such as an expanded Earned Income Tax Credit, is a better way to raise living standards. The CBO reports neither prove nor disprove that position. But they sharpen it in three ways.

The poverty effect is real, but small relative to the cost. Under the 2021 estimate, about 0.9 million people would leave poverty, against a central estimate of 1.4 million lost jobs and a net drag on total real income. A policy explicitly aimed at poverty can usually do better than that ratio. The EITC is targeted by household income and family size; a federal minimum wage is targeted by hourly pay, and a large share of its benefits go to workers in households that are not poor.

Uniform floors ignore geography. Much of the risk in the upper end of the CBO's range comes from imposing one number on very different labor markets. Fifteen dollars in Seattle and fifteen dollars in rural Mississippi are not the same policy. Our state-by-state guide shows how wide those gaps already are.

The people most at risk are the least visible. The workers who never get a first job do not show up as layoffs or protests. They show up, if at all, years later as thinner résumés and lower lifetime earnings. That is the core of the first-job-ladder concern, and it is exactly the kind of cost a headline midpoint tends to understate.

None of this requires pretending the CBO said something it didn't. Its reports are cautious, and the lower end of its range includes scenarios in which a $15 minimum costs very few jobs. A fair reading is that the agency saw meaningful benefits, meaningful risks, and a genuine scientific dispute about which would dominate. Our view is that when the downside lands on the most vulnerable workers, and when better-targeted tools exist, that uncertainty is a reason to prefer the tools that don't gamble with anyone's first job. You can test your own assumptions in the simulator.

Sources and further reading