The 4 Percent Mirage: What an Unemployment Rate Can—and Cannot—Tell You
The subject of this series is not the unemployment statistic by itself, but the relationship between that observed rate and the economy’s sustainable capacity. This chapter asks why a number such as 4 percent cannot serve as a permanent full-employment line, and it gives the reader a practical way to judge what such a number does—and does not—establish. The key distinction is simple: the unemployment rate is observed, while the benchmark used to interpret it is inferred.
A number is not a benchmark
An unemployment rate is a compact description of what is happening in the labor market at a particular time. It tells us the share of the labor force that is not working but is available for work and seeking work, or expecting to be recalled from a temporary layoff. That is useful information. It is not, however, a verdict on whether the economy has reached the limit of sustainable employment.
To make that verdict, economists need a comparison point. They may call it the natural rate of unemployment, the noncyclical rate, or an unemployment “star.” The terminology varies with the model and institution, but the basic idea is similar: this is the rate expected to remain after temporary cyclical disturbances have run their course, given the economy’s underlying supply conditions. It is not directly printed in the same release as the observed unemployment rate. It is estimated from a model that tries to separate persistent forces from temporary ones.
That difference changes the meaning of a sentence such as “unemployment is 4 percent.” If the relevant benchmark is 5 percent, the reading could indicate that the labor market is operating below its sustainable unemployment rate, with demand pressing it unusually hard. If the benchmark is 3.5 percent, the same observation could indicate more room to expand. If the benchmark has moved because the economy’s productive capacity has changed, comparing today’s 4 percent with yesterday’s benchmark may be misleading even when the statistic itself is measured in exactly the same way.
The number alone cannot tell us which interpretation is right. It cannot distinguish a temporarily strong demand environment from a change in labor supply, participation, productivity, or the capacity of firms and workers to match with one another. Nor can it tell us whether a low rate is already generating inflationary pressure. Those judgments require a reference frame and other evidence, including the behavior of prices and wages.
This is why “full employment” is better understood as a relationship than as a line on a chart. The relationship connects an observed outcome to a model of what the economy can sustain. The benchmark is important precisely because it is hidden: it helps policymakers assess slack, but it must be inferred from imperfect signals.
The diagram captures the first discipline of reading a full-employment claim. Start with the observed rate, but do not stop there. Ask which benchmark is being used, what forces determine it, and whether nominal evidence supports the implied interpretation of slack.
A moving benchmark appears even in one forecast
The Congressional Budget Office makes the same point in a different vocabulary. It defines real potential GDP as the amount of output that could be produced if labor and capital were employed at their maximum sustainable rates. It defines the potential labor force in similarly conditional terms: the size of the labor force if output and other important variables were at their maximum sustainable amounts. The CBO’s noncyclical unemployment rate is the portion of unemployment arising from sources other than changes in aggregate demand.
These definitions do not describe a fixed social or economic constant. “Maximum sustainable” is a condition that depends on the structure of the economy. The potential labor force depends on who is of working age, who participates, and how population changes. Potential output also depends on capital and productivity. A noncyclical unemployment rate depends on factors affecting workers’ and firms’ decisions and on how much labor demand the economy can sustain without treating every fluctuation in demand as a permanent change in capacity.
The CBO’s own forecast illustrates the point without requiring a debate over any one model. In its outlook, the unemployment rate reaches 4.6 percent in 2026, declines thereafter, and moves toward about 4.2 percent from 2032 onward, which the agency describes as roughly in line with its estimate of the noncyclical rate. The destination in that forecast is not 4 percent, and the path is not flat. The rate first reflects a recovery in economic activity and later reflects an economy growing near its projected potential, with labor-market conditions settling toward a different benchmark.
The reason for that movement is not a mysterious adjustment to a target. The CBO links it to changes in the supply of labor and the demand for it. Population aging and lower net immigration slow labor-force growth. The overall labor-force participation rate is projected to decline from 62.5 percent in 2025 to 61.9 percent in 2036, while participation among prime-age workers is partly lifted by policy incentives. The agency projects potential labor-force growth of about 0.4 percent annually from 2026 to 2036, slower than in the decade before the pandemic.
Those changes affect more than the number of people available to work. A smaller labor force can reduce potential output and weaken labor demand, which can slightly raise the noncyclical component of unemployment even as worker shortages in some sectors put upward pressure on wages. Capital accumulation and productivity pull in another direction. The CBO expects the diffusion of generative artificial intelligence to raise productivity gradually, while higher tariffs reduce efficiency and output relative to what would otherwise occur. Policy can change incentives, demand, investment, and labor supply at the same time.
This is the practical reason a benchmark must have a time horizon. A number that seems plausible in a mature expansion may not be appropriate after a demographic shift, a major policy change, a productivity acceleration, or a recession that changes how workers and firms behave. Even an institutional forecast that uses a stable definition does not promise a stable value.
The real question is what happened to capacity
There is a second reason a 4 percent reading cannot answer the full-employment question on its own: economists may disagree about where persistence belongs. A conventional trend-cycle model treats the trend as a slow-moving, supply-driven estimate of the star and assigns temporary movements to the cycle. That separation is analytically useful, but it assumes that the trend and cycle are independent.
The Federal Reserve research examined here relaxes that assumption. Its framework allows cyclical disturbances to leave persistent, though in the preferred specification ultimately temporary, effects on productive and labor-market capacity. A recession can therefore alter the path against which later observations are judged. The supply-driven star remains a separate object, but broader capacity can carry a history of earlier cyclical conditions.
That is not a license to call every weak period permanent damage. It is a warning about the accounting of persistence. One model may explain a long, hump-shaped movement in output and unemployment by making the cycle highly persistent. Another may attribute more of that movement to capacity that responds to past shocks and leave a smaller cycle. Both models start with the same observed data; they differ in how they infer the hidden benchmark and the gap around it.
The difference matters for policy. The gap that best explains current inflation may be measured relative to a history-dependent productive or labor-market capacity. The gap that matters for judging whether the economy is away from its supply-driven star can be different. A large movement in output or unemployment accompanied by relatively little change in inflation and wages may be evidence for a supply-side interpretation rather than proof that demand has pushed the economy beyond sustainable employment.
The research paper’s historical examples make the issue concrete. During the Global Financial Crisis, its scarring specification estimated the natural rate of unemployment at about 7 percent, compared with about 5 percent in the benchmark specification. In the 2018–2019 expansion, the scarring interpretation reconciled low unemployment and solid output with below-target inflation by treating the labor market as roughly in balance, while the benchmark implied a sizable negative unemployment gap. These are not two different unemployment releases. They are different readings of the same broad history.
The responsible conclusion is therefore narrower and more useful than “4 percent is full employment” or “4 percent is not full employment.” Four percent is an observation. To interpret it, ask four questions: compared with which benchmark; shaped by which supply and demand forces; checked against which inflation and wage evidence; and measured over what horizon? The answers may support a judgment of slack, balance, or overheating, but the judgment comes from the reference frame and the evidence—not from the number alone.
That leaves the central question for the next stage of the analysis: if the benchmark cannot be observed directly, how should we understand the stars, potential output, and gaps that economists use to construct it?