Why do governments deliberately tolerate some unemployment?

A certain kind of economic news lands like a rebuke. The recovery is under way, the reports say; the labour market is strong, even running hot, perhaps a little too hot, and something may have to be done to cool it. Meanwhile the applications you send still go unanswered, or the recovery that lifted the headline numbers has not reached the part of the job market you are standing in. The obvious question is why anyone in charge would describe a labour market as too hot while people are still looking for work, and why cooling it, on purpose, could be treated as the responsible thing to do.

The answer is the frame behind nearly every downturn a worker lives through, and it is the reason the previous chapter handed off to this one. A certain amount of unemployment is treated, deliberately, as the price of something the same policymakers are also charged with protecting: stable prices. The claim of this chapter is that the trade is real but the number governing it is soft. When you hear the job market called too hot, policy is acting on an estimate of how low unemployment can go before prices start to climb faster. That estimate has wide error bars, it moves over the years, and when it is set wrong the bill does not fall on the officials who set it. It falls on the people who stay out of work.

The floor that is not zero

Start with a fact that sounds like a paradox: full employment does not mean nobody is unemployed. At any moment a share of workers are between jobs, moving from one to the next, returning to the labour force, or newly out of school and still searching. That churn never stops, and an economy running as hot as it possibly can would still record several percent unemployed simply because people take time to match with work. No policy aims to abolish that, and none could.

The deliberate part is not a second layer of joblessness stacked on top of that churn; it is a posture toward it, and the posture rests on a mechanism. When unemployment falls low enough, workers become hard to find and hard to keep, which is the leverage the opening chapter of this volume described, now operating across a whole economy at once. Employers competing for scarce workers bid wages up faster; firms facing higher wage bills raise prices to cover them; and if that continues, prices across the economy climb faster and faster rather than settling. Economists give the turning point a name: the natural rate of unemployment, or the non-accelerating-inflation rate, the rate below which inflation tends to pick up speed. That rate is not the churn plus a chosen cushion; it is an estimate of how much unemployment the churn and the mismatch between workers and jobs produce on their own, which makes it a guess at where scarcity begins. A government that has made stable prices a formal goal, as most rich countries did over the past half-century (why every country ended up with a central bank aiming at roughly two percent inflation), treats that estimated rate as a boundary it does not want unemployment to fall durably below, and it will cool the labour market, on purpose, to defend the boundary. The deliberate part is that posture. The tolerated band of joblessness runs from zero up to the estimated boundary, most of it churn and mismatch no policy could abolish, and it is tolerated because the alternative, in this framework, is prices that will not hold still.

The floor is an estimate, and it moves

Here is the difficulty at the centre of the whole arrangement. The rate that governs it cannot be observed. Nobody can point to a labour market and read off the number below which inflation will accelerate; it has to be inferred from how wages and prices behaved the last time unemployment was high or low, and that inference is a model, revised whenever the data give a reason to revise it. The floor policy aims at is an estimate, and estimates carry error bars. To see how wide, it helps to watch the same official body estimate the same past and arrive, years apart, at very different numbers.

The picture is the argument. Two readings of the same natural rate, made by the same office thirteen years apart, disagree by about two-thirds of a percentage point for the end of 2013, and the gap widens further along the rest of the earlier projection. In a labour force of more than a hundred and sixty million, two-thirds of a point is more than a million jobs, which is the difference between a policy that judges the economy to have room to run and one that judges it to be overheating and in need of cooling. And the disagreement is not a quirk of one revision. The office re-estimates the entire path each time it publishes, reaching back decades, because the natural rate is a quantity the data only ever reveal in hindsight and imperfectly even then.

The revision here ran in a particular direction, and the reason is instructive. Through the late 2010s, unemployment kept falling past the level economists had marked as the floor, down toward and then below 4 percent, and the accelerating inflation the framework expected did not arrive. Faced with an estimate the world kept undershooting without consequence, the estimators concluded the floor had been too high and marked it down. That is what a model is supposed to do. It is also an admission that for years the earlier number had been wrong in a specific, costly direction: it had counselled treating the labour market as fully recovered while there was in fact more room to hire. The doctrine has since moved with the data: the Federal Reserve’s current guidance says it weighs a wide range of labour-market information rather than relying on any single estimate of the sustainable rate, and that employment can at times run above real-time estimates of maximum employment without necessarily putting price stability at risk.

Who is billed for the error

The trade at the heart of the policy is defensible on its own terms. Runaway inflation is itself a tax that falls hardest on people with savings in cash and wages that lag prices, so a central bank that keeps prices stable is protecting something real. The problem is not that a trade-off exists; it is where the cost of getting the number wrong lands. When the estimated floor is set too high, policy stays tighter than it needed to be, hiring is held back that could have gone ahead, and the people left waiting are not a random draw from the workforce. They are disproportionately the last hired and first cut: younger workers, those with fewer credentials, those entering the labour market for the first time, and workers in communities where employment recovers last. This is the mechanism behind the recovery that skips you. An economy can be genuinely healing in the aggregate, the reported rate falling month by month, while the queue you are standing in has barely moved, because the recovery reaches the back of the line last and a floor set too high stops it before it arrives.

The asymmetry is the point worth holding onto. The officials who set the estimate too high do not bear the cost of the miss; the workers who stay jobless do, in lost earnings, lost experience, and the lasting scar that a long spell of unemployment leaves on a working life. What that scar does to a person, and why a bad recession can mark someone for years after the charts recover, is taken up in a later volume (Volume V, on what a downturn does to the people it puts out of work). Here the point is narrower and about the machinery: the band of unemployment policy tolerates is not a precise dial set to a known value. It is a wide guess, and the guess is paid for by whoever is standing on the wrong side of it.

None of this makes the framework a conspiracy, and it is worth being exact about what it is not. No one draws up a list of people to keep out of work, and the same institutions that treat the floor as a limit are also charged, under the mandate, with pushing employment as high as that limit allows. The tolerated band is a by-product of aiming at stable prices with an imprecise instrument, not a goal in itself. But a by-product can still be a cost, and a cost that reliably falls on the same people is worth naming plainly, which the framework’s own language, with its talk of a natural rate and a hot market to be cooled, tends to keep out of view.

The frame behind the flashpoints

This chapter has been about the whole labour market at once, the aggregate that policy watches and steers. It is the backdrop for the more personal questions the rest of this volume takes up, because the amount of slack in the labour market is what sets the leverage every individual worker brings to a bargain. When the market runs cool by design, replaceable workers have less room to push for more, which is the mechanism the opening chapters traced from the other direction. The tolerated band is the macro setting of that dial.

It also leads straight into the volume’s flashpoints. If the government will not let the labour market run hot enough to lift the wages of those with the least leverage, the obvious alternative is to legislate a floor under their pay directly. Whether that floor costs them their jobs, as its opponents have long insisted, or leaves employment roughly intact while lifting pay, is a question labour economics has tested more than almost any other, and it is (whether raising the minimum wage destroys jobs) where the next chapter turns.