Issues

Work and Wages: What the Employment Numbers Measure

Employment statistics are among the most consequential numbers a statistical agency publishes and among the most frequently misread. Almost all of the confusion comes from a single source: the headline figures are produced by two different surveys, asking different questions, of different populations, with different error properties — and the news treats them as one number.

Two surveys, two answers

In the United States the monthly employment report draws on a household survey and an establishment survey, both published by the Bureau of Labor Statistics. Most countries run an equivalent pair.

The household survey asks people about their own situation. It produces the unemployment rate and the participation rate, and it counts a person once regardless of how many jobs they hold. It includes the self-employed and people working in businesses too small or too new to appear on any register.

The establishment survey asks employers how many people are on the payroll. It produces the "jobs added" figure and counts positions, so somebody holding two jobs is counted twice. It excludes the self-employed entirely and only picks up new firms through a statistical adjustment.

The two can move in opposite directions in the same month without either being wrong, and they routinely do. A month with strong payroll growth and a rising unemployment rate is not a paradox; it is two instruments measuring two things.

Why unemployment can fall for opposite reasons

To be counted as unemployed a person generally must be without work, available for work, and actively looking. Someone who stops looking leaves the labour force entirely and stops being unemployed — which lowers the rate. So the rate falls when people find jobs, and it also falls when people give up. These are opposite events with the same signature in the headline.

This is why the participation rate and the employment-to-population ratio belong beside the unemployment rate in any careful account, and why statistical agencies publish a family of broader measures that add discouraged workers and involuntary part-timers. Those alternatives are published every month in the same release; they are simply not the headline.

Gross flows and net change

The single most useful fact about labour markets is rarely reported: the net change is a small residual of two enormous gross flows. In a large economy, millions of jobs are created and millions destroyed every month through ordinary business formation, expansion, contraction and closure. A net figure in the low hundreds of thousands is the difference between two numbers each roughly twenty times its size.

Two things follow. First, a story about a large announced layoff at one firm tells you almost nothing about aggregate employment that month; it is a rounding error against the gross flows, even when the number is genuinely large. Second, "the economy created N jobs" is a net accounting statement, not a description of any identifiable set of new positions.

Revisions matter for the same reason. Early estimates are based on partial survey returns and are revised as more arrive; the revisions are frequently larger than the month-to-month changes being discussed. The special report on the jobs numbers works through the mechanics in detail.

Wages: nominal, real and composition

Average wage figures are subject to three separate traps.

• Nominal versus real. A pay rise below the inflation rate is a pay cut in purchasing power. Both series are published; only one is usually quoted.
• Composition effects. The average wage can rise because everyone got a raise, or because lower-paid jobs disappeared and the remaining workforce is differently composed. In sharp downturns the second effect can dominate and produce a rising average wage during a collapsing labour market.
• Mean versus median. Earnings distributions are strongly skewed, so the average sits well above the middle and moves for reasons the typical worker would not recognise.

Automation and offshoring: what the evidence supports

Both subjects generate confident coverage in both directions, and the research supports a more specific and less dramatic story. Technological change and trade exposure have both been found to displace particular workers in particular places, sometimes severely and for a long time, with effects that persist across a working life. Neither has been found to reduce aggregate employment over the long run in the economies studied; the labour force has repeatedly reallocated, though slowly and unevenly.

Holding both of those findings at once is the honest position. "Machines destroy jobs overall" is not supported; "the displaced are fine" is also not supported. The serious argument among labour economists is about the speed of adjustment and what, if anything, shortens it — an argument that is live and unresolved.

Where to look next

For how these figures interact with growth and prices, see the economy. For the trade side, see trade. The official glossary of labour terms published by the Bureau of Labor Statistics is short and settles most definitional disputes in a paragraph.