A recession is a significant decline in economic activity spread across the economy and lasting more than a few months. In the United States it is not declared by the government, the Federal Reserve or a statistical agency. It is dated by the Business Cycle Dating Committee of the National Bureau of Economic Research, a private non-profit, in months rather than quarters, and usually well after the fact. The widely repeated two-negative-quarters rule is not the definition the committee uses.

Where does the two-quarters rule come from, and why is it wrong for the US?

It is a rule of thumb that became a definition through repetition, and it is a reasonable one in some countries.

The UK Treasury says the commonly accepted definition of a recession in the United Kingdom is two or more consecutive quarters of contraction in national gross domestic product. Several other statistical agencies use something similar, because it is simple, mechanical and needs no committee.

The US does not work that way. The NBER's own guidance says most recessions its procedures identify do consist of two or more quarters of declining real GDP, but not all of them. The 2001 recession, which it dates from March to November 2001, is the standard counterexample: there was never more than one quarter of contraction in a row.

So in the United States the rule gets the right answer most of the time and gets it by accident.

What does the NBER committee actually look at?

Three criteria it calls depth, diffusion and duration.

Depth is how far activity falls. Diffusion is how widely the decline is spread across the economy rather than concentrated in one sector. Duration is how long it lasts.

The important subtlety is that these are not three hurdles each of which must be cleared. The committee treats them as somewhat interchangeable: each must be met individually to some degree, but an extreme reading on one can partly offset a weaker reading on another.

February 2020 is the clean illustration. The fall in activity was so deep and so widely diffused that the committee classified the downturn as a recession even though it was obviously going to be short. It turned out to be the shortest on record. Duration was weak; depth and diffusion carried it.

Which indicators matter besides GDP?

More than most coverage suggests. The committee says explicitly that it does not identify economic activity solely with real GDP and real gross domestic income, and that it places considerable emphasis on monthly indicators, which requires looking beyond quarterly national accounts.

Employment is central, because payrolls are monthly, broad and revised less dramatically than GDP. Real personal income less transfers, household survey employment, real consumer spending, industrial production and real wholesale and retail sales all feature in its discussions.

That mix is also why the dating is in months. A quarterly series cannot locate a turning point to a month, and the committee's chronology is a list of months. If you want to follow the same inputs between committee statements, the monthly jobs report is the place to start.

How long after a recession starts is it declared?

Long enough that the announcement is rarely useful for decisions.

The committee waits for data that is complete and revised enough to be confident, and it accepts a lag rather than risk reversing a call. The 2020 recession was declared in June 2020, a few months after the February peak, which was unusually fast because the collapse was unambiguous. The end of that recession, dated to April 2020, was not announced until July 2021, more than a year after it had happened.

The practical consequence: by the time the NBER confirms a recession, markets have usually priced it, the Federal Reserve has usually responded, and the question for a household has moved on from whether one started to whether it has ended.

Is there a difference between a recession and a depression?

There is no official threshold, and that is the honest answer.

A depression is understood to mean a decline that is much deeper and much longer than a recession, and the only universally accepted example in the US is the 1930s. No committee dates depressions, no agency defines one, and the word is used descriptively.

Anyone offering a precise numerical cutoff, such as a 10 per cent fall in output or three years of contraction, is quoting a convention rather than a standard.

What about a technical recession, a growth recession or a soft landing?

Three pieces of vocabulary worth separating.

A technical recession is the two-negative-quarters outcome. It is a statistical description, not an NBER call, and it is the phrase careful reporting uses when GDP has fallen twice without any committee having spoken.

A growth recession, sometimes called a slowdown, is growth that is positive but below the rate needed to keep unemployment steady. Activity is rising and the labour market is still loosening. No recession is dated.

A soft landing is the policy outcome in which inflation falls back toward target without a recession. It is a forecast or a judgement, not a measurement, and whether one has been achieved is usually argued about for years afterwards. What the Fed is trying to do, and the tools it has, are covered in how the Federal Reserve works.

How can I tell in real time, if the official call comes late?

You cannot know, but you can watch the same monthly series the committee watches, which is better than waiting for a quarterly GDP print.

Payroll employment is the single most-followed one, because a sustained fall in jobs is hard to reconcile with an expansion. The unemployment rate matters, and so does the pattern of its rise rather than its level. Real personal income less transfers removes the effect of government payments. Industrial production covers the goods economy, which often turns first.

What none of these gives you is a date. Turning points are visible in hindsight, which is precisely why the committee operates in hindsight.

What actually happens to households in a recession?

The channel that matters most is the labour market, not the stock market.

A recession shows up as hiring slowing before firing starts. Job openings fall, the quit rate drops because people stop moving voluntarily, hours get cut, and only then do layoffs broaden. For anyone in work, the first signal is usually that it has become harder to leave rather than that the job has gone.

Borrowing gets more expensive in a different way than headline rates suggest. Policy rates often fall during a downturn, but lenders tighten standards at the same time, so the advertised rate drops while the chance of being approved drops with it. Credit availability and the price of credit move in opposite directions.

Income from savings also moves. If the central bank cuts, deposit rates follow downward faster than mortgage rates do, which is a recurring complaint and a straightforward consequence of how banks make money on the spread between the two.

None of this is a forecast of any particular downturn. It is the pattern that has repeated, and the part of it that reaches a household first is almost always hiring.

Does a recession mean falling prices?

No, and the 1970s are the reason this matters.

Recession is about activity, not prices. Inflation can stay high during a downturn, which is the condition called stagflation, and it is harder for policymakers to treat because the usual remedy for one half worsens the other. Falling prices across the economy is deflation, a separate phenomenon with its own risks. Inflation and deflation are separate measurements from activity, and should be read that way.

The short version

For the United States, a recession is whatever the NBER's Business Cycle Dating Committee says it is, judged on depth, diffusion and duration across monthly data, dated to a month, and announced late.

For everywhere else, check which definition the local statistical agency uses before comparing countries, because a US recession and a UK recession are not measured by the same rule.

This article explains how recessions are defined and dated. It is not investment advice.