The Most Quoted Number

Three kinds of unemployment, and the distinction matters because the treatments are different.

Frictional unemployment is people between jobs. It is unavoidable and mostly benign (an economy in which nobody was ever searching is one in which nobody ever moves to a better match) and it is reduced by better information rather than by more demand.

Structural unemployment is a mismatch: the skills people have are not the skills being demanded, or the people are in the wrong place. It responds to training, to housing policy that lets people move, and to time.

Cyclical unemployment is a shortfall of aggregate demand. There is nothing wrong with the workers; there is not enough spending to employ them. It responds to monetary and fiscal policy.

Getting the diagnosis wrong is a recurring and expensive error, and it has a direction. In every serious recession, a chorus explains that the unemployed lack the skills employers need, and governments fund retraining. If the problem is a demand shortfall, this trains people for jobs that do not exist and blames them for the delay.

The way to tell them apart is worth knowing. If vacancies are low and unemployment is high, demand is the constraint. If vacancies are high and unemployment is also high, the people and the jobs exist and are not finding each other, which is a matching problem. Plotting the two against each other gives the Beveridge curve, and a shift outward, more unemployment at any level of vacancies, is the signature of matching getting worse.

Which brings up why the models that won the 2010 Nobel exist at all.

Standard supply and demand cannot explain how a market can contain both unfilled vacancies and unemployed people. If the market clears, one or the other should be zero. Diamond, Mortensen and Pissarides modelled matching as a process that takes time and effort on both sides: workers search, firms post vacancies, and meetings happen at a rate determined by both. Unemployment and vacancies coexist because finding the right match is costly, not because anybody is being unreasonable.

Now the number itself, which is quoted more than any other economic statistic and measures something narrower than most people assume.

To be counted as unemployed you must be out of work, available to work, and actively searching. Each condition excludes somebody.

A person who has looked for two years, concluded there is nothing, and stopped is not unemployed. They have left the labour force, and the unemployment rate improves when they do. Someone working six hours a week who wants forty is employed. Someone in a training scheme is neither.

So an unemployment rate falling can mean that people found jobs, or that people gave up. Those are opposite events and the headline number does not distinguish them. This is why the participation rate matters (what fraction of the working-age population is in the labour market at all) and why broader measures exist, such as the US U-6, which includes discouraged and involuntarily part-time workers and typically runs several points higher.

Two further concepts, both consequential.

Full employment does not mean zero unemployment. The usual definition is the rate consistent with stable inflation, which is not observed but estimated, and has been revised downward repeatedly over the past three decades as economies ran hotter than expected without inflation appearing. It is quoted with far more confidence than its estimation deserves, and policy made on a wrong estimate keeps unemployment unnecessarily high.

And hysteresis, borrowed from physics, where a system's current state depends on its history. Long-term unemployment erodes skills, networks and habits, and employers treat a long gap on a CV as a signal about the applicant. So a temporary demand shock, if allowed to run long enough, converts people from cyclically unemployed to structurally unemployed, and the damage persists after the recession ends.

If hysteresis is real, and the European experience of the 1980s is the standard evidence, then acting slowly in a downturn is not merely a delay in recovery. It permanently raises the level of unemployment an economy settles at, which is a strong argument for responding early and forcefully rather than waiting for confirmation.