US Unemployment Rate

The definition is narrower than it sounds

The unemployment rate is the number of people without work, available to start a job, and having actively looked for one in the past four weeks, divided by the labour force. The labour force is everyone employed plus everyone unemployed. People who have stopped looking are in neither group, which is the single most important thing to understand about this number.

That exclusion is why the rate can fall for two opposite reasons. It falls when unemployed people find jobs, which is good. It also falls when unemployed people give up and stop looking, which is not. The two are indistinguishable in the headline number. During 2020 the participation rate dropped by more than three percentage points in a single month, and some of the apparent improvement in later months came from people leaving the labour force rather than finding work.

The unemployment rate is also deliberately not a measure of hardship. It says nothing about hours, wages or whether a part-time job is being counted as full employment. The U-6 measure, which includes discouraged workers and people working part-time who want full-time hours, typically runs four to five percentage points above the headline. Both are published; the headline is the one that gets quoted.

Why it lags

Hiring and firing are expensive, so firms adjust slowly. They stop posting vacancies first, then stop backfilling attrition, then lay off. The unemployment rate only moves when the last of those happens, which puts it several months behind the shock that caused it. In 2008 the rate bottomed in the spring of that year, well after credit markets had seized, and peaked in late 2009, well after output had started to recover.

That lag is why the Sahm rule exists. It compares the three-month average rate with its low over the previous twelve months and triggers when the gap reaches half a percentage point. Using a three-month average removes single-month noise; using a twelve-month low anchors the comparison to the cycle rather than to a fixed threshold. The rule has a good record and a known weakness: it can trigger in expansions when the labour force grows quickly enough to raise the rate without any layoffs.

The record

The participation question deserves separating out. The prime-age employment-population ratio counts everyone aged 25 to 54 who holds a job, whether or not they are looking for one, so it removes both the definitional distortion and the demographic drift of an ageing population. It is not a substitute for the unemployment rate, but when the two disagree, the ratio usually describes the labour market more honestly. A falling unemployment rate alongside a falling prime-age ratio is a warning dressed as a recovery.

The series starts in 1948. The minimum is 2.5 percent in May 1953, at the end of the Korean War mobilisation, and the modern equivalent was 3.5 percent in 2019 and again in early 2020. The maximum is 14.8 percent in April 2020, which needs a caveat: millions of people were classified as unemployed while on temporary layoff and expecting to return, and the measurement machinery was strained. It was a real number describing a real rupture, but it was not comparable with the 10.8 percent of 1982, which was composed of permanent job losses.

Between those extremes, the post-war median is about 5.5 percent. The top tenth of the distribution begins around 7.8 percent. A reading near four percent, where the series has sat for most of the past two years, is in the bottom fifth of the record: tight by post-war standards, though not as tight as 2019.

The shape of the recoveries has changed. After 1982 unemployment took six years to return to five percent; after 2009 it took seven. After 2020 it took roughly two, which was unprecedented and is the main reason the subsequent inflation episode was as broad as it was.

How we calculate it

We take UNRATE from FRED, the seasonally adjusted civilian unemployment rate published by the Bureau of Labor Statistics, and show it unmodified. No smoothing, no annualisation: the number is the monthly rate as published.

The percentile is the share of monthly observations since January 1948 that sit below the current reading. The peak in the summary line is the maximum of the full series.

The source is a Bureau of Labor Statistics release in the public domain, distributed through FRED.

Questions

Related

• Sahm rule
• Industrial production
• CPI inflation
• Bank credit impulse
• Plans and full-data access