Guide

1. What the Big Six are

In the United States a recession is not declared by a rule or a formula. It is declared by the Business Cycle Dating Committee of the National Bureau of Economic Research, a standing committee of economists who decide, after the fact, when the economy turned. Their dates are the official ones.

The committee is explicit about what it looks at. Alongside quarterly output it names six monthly indicators as the measures it tracks to judge the depth, breadth and duration of a downturn:

Nonfarm payroll employment
Household-survey employment
Real personal income less transfers
Real personal consumption expenditures
Real manufacturing and trade sales
Industrial production

This index is those six and nothing else. That is the whole idea. Rather than assemble a set of indicators that might predict what the committee will eventually decide, it reads the same instruments the committee reads, in one number, every month — so the question it answers is the committee’s own question rather than a proxy for it.

They are called coincident indicators because they turn with the economy rather than ahead of it. That is a limit and a strength at once, and both are stated plainly further down: this index cannot warn you, and it does not guess.

2. What this page is

A monthly reading of whether the United States economy is in recession, built from those six. The committee announces a peak many months after it happens, sometimes more than a year, because it waits for data to settle and for the turn to be unambiguous; this index has turned negative an average of 1.8 months after the peak, across all eight recessions since June 1968, without a false alarm. It is not a forecast of the committee’s decision — it is the same evidence, read sooner.

3. Reading the chart

The grey line is the six indicators’ growth, averaged, month by month. The blue line is a five-month average of the grey one, and it is the one to read — the track record below is measured on it.

Do not act on the grey line. It is there so you can see the underlying month-to-month movement, but it is far too noisy to trade a view on: it has crossed zero 146 times in fifty-eight years against the blue line’s 20.

The most recent month is usually built from five of the six. Real manufacturing and trade sales is published about a month behind the others, so the newest point on the chart is normally an average of five series rather than six — though this month all six have reported. A month short of one is marked provisional and cannot trigger a recession call on its own, which is also why the breadth reading is a percentage rather than a count out of six.

The moon or sun and the two As, at the right end of the scoreboard, set how the page looks and nothing it says. The moon puts the page on a dark ground and the sun puts it back on a light one; it opens dark unless you choose light. Every colour keeps its meaning — the blue line is still the one to read, red is still zero and amber the second trigger. The two As are the text size of the panel below the buttons: the lit A is the size you are reading at, a click makes the text one point larger and the next puts it back. The charts, the header and the scoreboard keep their size. Your browser remembers both choices.

Full width. The button beside the two As, when your browser window is at least 1,280 wide: the page across the whole window, over the site, for the long history and less scrolling. On a smaller window the button does not appear; if you do not see it, widen the browser window or zoom it out (Ctrl and −, or Cmd and − on a Mac). The charts are redrawn at the window's width — 1968 to today in about twice the detail on a 1,920 screen — and take its height in the same proportion as here; every font is two points larger, and the panels below run the full width. The two As still add their point to the panels. Click it again, or press Esc, to come back to this frame where you were; a window made narrower than 1,280 comes back by itself.

4. The ribbon, and the two ways the signal opens

Under the chart is a row of bars, one per episode: green marks a recession correctly identified, red a call that came to nothing. There are no red bars, and the reason is the whole record this page leads with — all eight recessions, no false alarms in fifty-eight years.

The signal opens two ways. It turns negative when the blue line crosses below zero — the six indicators collectively contracting — or when any single one of the six falls below −0.86, which is a collapse in one part of the economy deep enough to matter before the average has caught up. That second reading is taken on the blue line of the component chart, not the grey one — the same five-month average, applied to one indicator instead of six — and the difference is not a detail. On the grey component lines that level is reached in 56 of the 59 years in the record, which would make it meaningless. On the blue ones it is reached in 13, and 12 of those 13 contain a recession month.

It closes when the monthly rate turns back up — but only once the downturn has actually got somewhere. The grey line has to cross back up through zero and the call has to have reached −1.200 at some point along the way. A dip that never got that far has not earned an all-clear on one good month.

For the shallow ones there is a second route out: three of the last four months at or above zero, once the call is four months old. Two of the eight leave that way — their monthly rate never falls far enough for the first route — and the rest end on the turn. Three of four rather than a clean run because one bad print should not restart the clock: every one of the eight all-clears has a negative month inside its window, the smallest −0.05 in September 1982.

Then the signal sleeps for six months. Once a call has ended it cannot start again for half a year, and that is not a way of flattering the record — it is what stops the page announcing a recovery and then taking it back. The monthly rate turns up before the five-month average does, so without it four of the eight calls would end and re-open a month later. No recession has ever followed another within twelve months — the tightest gap in the record is a year, in July 1980 to July 1981 — so a six-month sleep cannot hide a second recession the record has ever seen.

So in REPORT’s table you will find months where the monthly rate is positive and the signal is still calling — 11 of them across the eight recessions. Those are the months the bar under the chart is still green.

Nothing on this page is dated with hindsight. The month a call opens and the month it ends are both decided from data already published when they happen, so no month is ever reclassified afterwards: the bars under the chart, the state at the top of the page and REPORT’s table all say what this page said at the time. That was not true of the construction this one replaced, whose all-clear could only be confirmed five months after the fact.

The second condition exists because an average can be held up by its own members. In January 1970 industrial production had fallen through the rule, and in February 1974 real personal income less transfers — in both cases with the blue line still above zero, held up by the components that had not turned yet, and taking months to follow. Reading the weakest of the six as well as their average pulls those two calls forward: on the blue line alone the index turned negative an average of 2.4 months after a recession began, with a worst case of 5; with the collapse condition it is 1.8 months and a worst case of 3.

It is a rare condition, not a second signal. In fifty-eight years it has opened an episode on its own two times, in January 1970 and February 1974 — both of them recessions. The other six calls were the blue line crossing zero, and on those the two conditions agree exactly. There is one signal and one ribbon; the second condition changes when it opens, never whether it was right.

You can watch it coming. The WEAKEST pane below the chart draws the lowest of the six each month against that same −0.86 line, and each of the six component charts carries it too, so the one responsible is always identifiable. The trigger is drawn in amber and zero in red, on that pane and on all six, which is the same red as the chart above. It sits at 0.20 today (Household-survey employment), 1.06 away from the trigger. It is a demanding line: outside recessions the weakest of the six has been below it in only 13 of 606 months.

Neither condition is an early warning. Neither has ever fired before a recession began. Both recognise a downturn already under way; neither sees one coming.

The blue line is calculated from the grey one, so the two can never tell you different things. The same relationship holds all the way down: the six component charts below average out to the chart above, exactly, so you can use them to see which part of the economy is driving a reading.

Those first two keys in the legend are switches. Click either to hide that line and give the other the whole chart — useful on the long windows, where a monthly line across fifty-eight years is dense to read underneath a smooth one. One of them is always on: clicking the last visible key does nothing, and the red zero rule never goes away.

Zero is the line that matters and it means one thing: the six indicators are collectively flat. Above it they are growing, below it they are contracting. It is not a below-trend line and not a fitted level — it is where the arithmetic puts zero growth, and nothing downstream can move it.

5. The REPORT panel

REPORT is the only panel on this page that is written rather than computed. Everything else — every chart, every figure in every panel — is calculated from the six series, so it cannot say anything the data does not. The first tab is a person writing about what the index is doing, and it should be read against the charts rather than instead of them. It has four tabs.

ANALYSIS is that read, rewritten each month when the data is refreshed; the last six are kept, newest first. Opening REPORT sets the chart to the last two years and the pane below it to BREADTH, because that is the window and the view the read is written against. It does that once. Move either control yourself and this panel will not touch it again.

VINTAGES is the honest answer to a question this page otherwise dodges: how much does a published reading move afterwards? Every rebuild recomputes the whole history from data the statistical agencies may since have revised. Two charts, side by side: the last 12 rebuilds of the monthly rate on the left and of the signal on the right, over the same 13 months, one line each. Either can borrow the other’s newest line with the key beneath it, and running the mouse across either says what a month reads now, what it first read, and how far it has moved. The largest disagreement between any two rebuilds about one month is 0.18 index points on the signal against 0.38 on the monthly rate — the averaging absorbing most of the revision. The tab also says which of those lines are reconstructions rather than rebuilds this page has made, which for now is most of them.

DATA is the published series as numbers over the same window — the signal, the monthly rate, the weakest of the six, the level, the drawdown, breadth, the probability and each of the six components. Every number there is a number a chart above is drawn from.

PATHS is what a recession has done to these lines before. All eight since 1968 are aligned on the month each began, drawn from 12 months before that peak to 12 after, and shown twice — the monthly rate on the left and the signal on the right.

The shaded band is the middle six of the eight. At each month the highest and lowest readings are dropped and the rest shaded, because the untrimmed range is largely the gap between two unusual episodes — 2007 at the bottom and 1973 at the top — and a range that wide holds almost every ordinary month as well, which tells you nothing. The blackwhite line through it is the median, and it is the line to read: on the signal it bottoms at −0.43 six months after the peak. The deepest single episode was −2.37 in 2020, and the shallowest −0.43 in 2001; both are named under the charts, so what the trim drops is not lost.

The dashed line is the current stretch, with the latest month at zero. That placement is an alignment, not a forecast. It is the only way to put the run-ups side by side, and this index cannot warn in advance — the last section of this guide says why, and it applies to that chart more than to any other on the page.

6. The track record

Every US recession since June 1968, and how long this index took to call it. All eight were identified and there were no false alarms — no occasion in fifty-eight years where the blue line went below zero and no recession followed.

Recession began	ended	signal late	signal clear	opened on
1969-12	1970-11	+1	+0	a component collapsed
1973-11	1975-03	+3	+0	a component collapsed
1980-01	1980-07	+2	−1	the index crossed zero
1981-07	1982-11	+2	−1	the index crossed zero
1990-07	1991-03	+2	−1	the index crossed zero
2001-03	2001-11	+1	+5	the index crossed zero
2007-12	2009-06	+2	+0	the index crossed zero
2020-02	2020-04	+1	+0	the index crossed zero

On average it turns negative 1.8 months after a recession has begun and gives the all-clear 0.2 months after it has ended. The slowest call was 3 months (1973-11); the slowest all-clear was 5 months (2001-03).

Three of the all-clears come a month before the official trough, which is why there are minus signs in that column. The committee dates a trough in retrospect, and a month’s data is not published until the following month, so a call ending one month early arrives about when the trough itself does. What the index is held to is the distance either way, not the direction: no all-clear in the record, or in any rebuild of it, comes more than one month early.

For comparison, the committee that dates these recessions officially takes considerably longer than either figure.

7. What it will not do

It will not warn you in advance. These six series measure the economy as it currently is, so there is no advance information in them to extract, and this is a limit of the data rather than of the method: every attempt to make this index fire earlier also made it fire during slowdowns that never became recessions. An earlier signal built from these six is not a better version of this page. It is a less reliable one.

If you want advance warning you need indicators that lead the economy rather than move with it — credit spreads, jobless claims, equity prices — which is what the WLEI is for. Use the WLEI for whether a recession is coming, and this page for whether one is here.

NBER Big-6 Coincident Index Ver 1.6 (PUBLIC)