Arguments about the unemployment rate have a peculiar structure. One side produces a figure; the other produces a larger one and calls it the real rate. The exchange proceeds as though the question were factual and one party had misapprehended the fact. It is not factual in that way. The headline rate is the output of a definition, and that definition was chosen because it answers a particular question effectively.
The standard definition classifies a person as unemployed if she is out of work, available to start, and has looked for work in the past four weeks. Each condition is performing a function. Availability excludes people who could not accept employment if it were offered, and consequently could not be matched to a vacancy by any policy. Recent search excludes people who have abandoned looking, on the ground that whatever is keeping them out is not a shortage of vacancies. Out of work excludes the underemployed. The question under examination concerns people without employment, not people whose employment is insufficient.
Every one of those exclusions constitutes a decision, and every one of them is contestable if the question changes. If what is wanted is a measure of slack -- how much additional work the economy could accommodate without pressure on wages -- then the underemployed manifestly belong in it, and so do people who would look if they thought looking would work. If what is wanted is a measure of hardship, availability is irrelevant and duration is central. If what is wanted is a leading indicator, the definition should be whichever one moves first, and that is an empirical matter with no dependence on hardship at all.
The statistical agencies have understood this for decades and publish a family of measures rather than one, lettered in a series from the narrowest to the broadest. The narrowest counts only those out of work for protracted periods; the broadest counts the underemployed and the discouraged as well, and comes out at approximately twice the headline figure. Both are published monthly, and neither is confidential.
The persistence of the argument therefore necessitates a different explanation from ignorance. What is happening is that a single number is being asked to serve as a summary of the labour market's whole condition, and no definition can accomplish that, because the market's condition is not one thing. The dispute survives because both parties want the authority of a fact for what is properly a choice of question -- and the choice of question is where the disagreement genuinely lives, and where it would have to be conducted if the numbers stopped being available to fight over. The agencies themselves are scrupulous about this and say so in their own notes, which nobody reads. What the notes say is that each measure answers a question, and that selecting among them is a matter for the person enquiring, not for the statistician.
