Leading, lagging and coincident indicators: what turns first

Economic data don't all turn at the same time. Some series change direction months before the economy as a whole does, some move with it, and some only confirm a turn after it has happened. Economists sort them into three groups: leading, coincident and lagging indicators. Knowing which is which is the difference between reading the news and reading history.

The sorting is not arbitrary. It follows from how decisions are made in a real economy: a company stops hiring and cuts overtime before it lays anyone off; a builder stops pulling permits before construction falls; a household stops borrowing before its spending drops. This page explains the three groups, what goes into the Conference Board's composite indexes, and how a few key series behaved around the 2001 and 2007–09 recessions, including the times they gave false signals.

Three kinds of timing

An indicator's group is defined by when its turning points fall relative to the business-cycle peaks and troughs that the NBER publishes (see The business cycle: expansions, peaks, recessions and troughs).

• Leading indicators tend to turn before the cycle. They usually measure commitments and expectations: orders placed, permits filed, hours scheduled, credit extended, prices of assets that look ahead.
• Coincident indicators turn with the cycle. They measure activity itself: how many people are working, how much is produced and sold, how much income is earned.
• Lagging indicators turn after the cycle. They measure the consequences and the slow-moving stocks of things: the unemployment rate, inventories relative to sales, outstanding loans, inflation in services.

Two points follow. First, the groups describe tendencies, not laws; a series can lead at one turn and lag at the next. Second, coincident data are what the NBER's dating committee uses to decide when recessions began and ended (see What is a recession? How the NBER dates US recessions). Leading data are what investors and forecasters watch to get there first.

The Conference Board's composite indexes

The best-known way to combine indicators is the set of three monthly composite indexes published by The Conference Board, a private research organization. Each averages its components so that the noise in any one series is partly smoothed away.

The coincident index is close to the NBER's own checklist. The leading index mixes real activity (hours, orders, permits), labor (claims), financial markets (stocks, credit, the yield spread) and a survey (expectations). It is a convenience, not an oracle: its value depends on how its components happen to behave in a given cycle.

Why some data lead

The logic is easiest to see in the labor market.

1. A company that sees orders slowing first cuts overtime and hours. That shows up in average weekly hours.
2. If the slowdown continues, it starts laying people off. Each laid-off worker who applies for unemployment insurance becomes an initial claim, counted every week (see Jobless claims: reading the weekly unemployment insurance report).
3. As layoffs accumulate and hiring stops, payroll employment falls. That is a coincident indicator, measured monthly in the The jobs report: how to read nonfarm payrolls and the two surveys.
4. Finally, as people stay out of work longer and more enter the labor force looking, the unemployment rate keeps rising, often for months after output has started to recover. That makes it lagging.

The same layering appears elsewhere. In housing, permits come before starts, and starts before completed construction (Housing data: starts, permits and new home sales explained). In manufacturing, new orders come before production (PMI explained: how purchasing managers’ indexes work, Industrial production and capacity utilization: the Fed’s G.17). In finance, stock prices and the yield curve react to expected conditions, not current ones.

Claims versus unemployment: a worked example

The chart sets the leading series, initial claims smoothed over four weeks, above the lagging one, the unemployment rate. Two recessions fall in the window.

2001. The 4-week average of claims hit its low of 266,250 in the week ending April 15, 2000, and had risen 20% from that low, to 320,500, by the week ending November 18, 2000, four months before the March 2001 business-cycle peak. Claims peaked at 489,250 in October 2001. The unemployment rate, meanwhile, was at its low of 3.8% in April 2000, the same month as the claims low, so it did not lag at that turn. But it kept rising long after the November 2001 trough, to 6.3% in June 2003.

2007–09. Claims bottomed at 286,500 in the week ending February 4, 2006. This time their rise was slow: the 4-week average did not get 20% above the low until the week ending December 15, 2007, the month of the business-cycle peak, so on this measure claims were coincident at the top. At the bottom they led. The 4-week average peaked at 659,250 in the week ending March 28, 2009, three months before the June 2009 trough. Unemployment peaked at 10.0% in October 2009, seven months after claims.

2020 shows the other extreme. Weekly claims were 208,000 in the week ending March 7, 2020, and 6,137,000 four weeks later, the highest weekly figure in the data since 2000. In a sudden shock, a "leading" series and the event itself arrive in the same month.

More turning points from the data

The same exercise across other series gives a sense of the spread.

Industrial production hit both NBER dates exactly. Payrolls matched the peak but kept falling for eight months after the trough. In 2001 they did the same thing at greater length: payrolls peaked in February 2001, a month before the business-cycle peak, and did not bottom until August 2003, 21 months after the trough. Even "coincident" data lag at troughs when hiring recovers slowly.

Permits gave the longest warning, 27 months, which is typical of a housing-led downturn and not a general rule. The Yield curve inversion: what it signals and its recession record page discusses the yield-curve signal in full.

False signals and limits

• Leading means early, not right. Monthly 10-year minus 3-month spreads were negative in every month from November 2022 to November 2024, and as of the NBER's published dates no recession has followed. The The Sahm rule: a recession signal from the unemployment rate, an unemployment-based trigger, crossed its threshold in August 2024, also without a recession.
• Stock prices are a component, not an outside check. The S&P 500 is in the LEI, so a falling market pulls the index down. Using the LEI to forecast stocks is partly circular. SPY's monthly close fell 25% from December 2021 to September 2022 without a recession.
• Lead times vary. The table above shows leads of zero to 27 months. A signal without a reliable lead time is hard to act on.
• Structural change. Manufacturing hours and goods orders matter less in a service-heavy economy than they did when composite indexes were first built.
• Noise and revisions. Weekly claims swing with holidays, weather and strikes, which is why the 4-week average exists. Monthly series are revised (see Seasonal adjustment and data revisions: why the numbers change).
• Composite changes. Index providers change components and methods from time to time, so long histories of a composite index are not always like-for-like.

The three groups compared

How Finabase does it

Finabase does not publish economic indicators or a leading index; there is no economic calendar and no claims, permits or payrolls data in the app. The chart and turning-point dates on this page are ours, computed from public US-government series via FRED.

Two of the Conference Board's leading components are market prices, and those the app does show. SPY's chart and the indexes page cover the S&P 500, and the Treasury table lists yields from 1 month to 30 years, so you can read the shape of the curve and see whether short yields sit above long ones. For the releases themselves, the economy topic in Discover collects news coverage.

Frequently asked questions

What is a leading economic indicator?

A data series that tends to change direction before the economy as a whole does. Common examples are weekly initial jobless claims, building permits, new manufacturing orders, the spread between long and short Treasury yields, and stock prices.

Is unemployment a leading or lagging indicator?

The unemployment rate is lagging. It usually keeps rising after a recession ends: it peaked in October 2009, four months after the June 2009 trough, and in June 2003, 19 months after the November 2001 trough. Weekly jobless claims are the leading labor measure.

What is the Conference Board Leading Economic Index?

A monthly composite of ten indicators, including jobless claims, building permits, manufacturing hours and orders, stock prices, a credit index, the 10-year Treasury minus fed funds spread and consumer expectations. It is designed to turn before the economy, but like its components it gives false signals.

Which indicators does the NBER use to date recessions?

Coincident ones. The committee looks mainly at real personal income less transfers, payroll and household-survey employment, real consumer spending, real manufacturing and trade sales and industrial production, plus quarterly GDP and GDI.

Summary

Economic indicators fall into three timing groups. Leading indicators, such as claims, permits, orders, the yield curve and stock prices, turn before the cycle because they measure plans and expectations. Coincident ones, such as payrolls, real income, sales and industrial production, turn with it and are what the NBER uses to date recessions. Lagging ones, such as the unemployment rate and inflation, confirm turns after the fact. In the 2007–09 recession, permits led by more than two years, industrial production matched the NBER dates exactly, and unemployment peaked seven months after claims. The groups are tendencies, not rules: leads vary from zero to more than two years, and leading indicators, including the yield curve from 2022 to 2024, give false alarms.