Leading, Coincident, and Lagging Economic Indicators: How to Read the Cycle

Economists group indicators by their usual timing around changes in economic activity. A leading indicator tends to turn before a reference measure, a coincident indicator tends to turn at about the same time, and a lagging indicator tends to turn afterward. The Conference Board publishes composite indexes organized around these three roles. The labels are relative: they depend on the reference series, the country, the time period, and the method used to identify a turn. {source:conferenceBoardUsLeadingIndicators}

“Leading” does not mean that a series knows what will happen. A housing-permit count can move before construction, but a change may reflect financing costs, local supply, or policy rather than a broad downturn. “Coincident” does not mean the data are released at the same instant as the underlying activity. Payrolls and production measures arrive on different schedules and can be revised after publication. “Lagging” does not mean unimportant; a response that comes later can help show whether an earlier change has persisted.

This guide uses U.S. examples and the OECD’s international composite-indicator framework. Other countries select different series and define their reference cycles differently, so do not transfer an index threshold or component list from one country to another without checking its methodology.

The timing label only makes sense after the target is clear. A classical business-cycle measure looks at turning points in the level of broad economic activity. A growth-cycle measure asks whether activity is above or below an estimated long-run trend. A growth-rate cycle instead focuses on accelerations and slowdowns in the rate of change. Those turning points need not occur in the same month.

Since April 2012, the OECD has used GDP as the reference series for most economies; China is the exception, where the OECD FAQ identifies industrial production as the reference. The GDP-gap description therefore applies to most country examples, while China’s CLI follows its industrial-production reference cycle. This is a growth-cycle signal, not a promise to forecast a recession date or the next quarterly GDP growth rate. OECD guidance describes it as qualitative and aims to anticipate turning points six to nine months ahead. That is a design objective, not a guaranteed horizon for every country or cycle. {source:oecdInterpretingCompositeLeadingIndicators} {source:oecdCompositeLeadingIndicatorsFaq}

This distinction explains why a leading indicator can rise while output remains below trend, or fall while output remains above trend. The direction of the signal and the level relative to trend answer different questions. The OECD also warns that its CLI levels cannot be compared across economies to rank expected GDP growth, because each economy’s trend is estimated separately. A signal about a turning point is not a measured amount of future growth.

For an OECD CLI whose reference series is GDP, combine the level with its direction and read 100 as a signal about GDP levels six to nine months ahead, not a same-month report. A reading below 100 that is rising can signal that GDP levels are expected to remain below trend while the negative gap narrows; growth may be above its trend pace even though the level has not crossed the trend. A reading above 100 that is falling can signal a narrowing positive gap, with GDP levels expected to remain above trend while growth runs below its trend pace. For China, apply the same distinction to industrial production as the reference series rather than describing its CLI as a GDP-gap signal. The six-to-nine-month horizon is a design objective, not a guaranteed forecast, and slower growth is not automatically a contraction. {source:oecdInterpretingCompositeLeadingIndicators} {source:oecdCompositeLeadingIndicatorsFaq}

A leading series is selected or described because its turning points have tended to come earlier than those in a specified reference series. New orders can precede production because firms receive orders before they make and ship goods. Building permits can precede construction. Initial unemployment-insurance claims may react quickly to changes in layoffs; a composite index can invert the series so that rising claims count as a negative contribution. Financial prices and consumer expectations can also change before some slower economic data are published.

The reason for the lead varies by series. Orders are connected to production plans, permits to planned construction, and claims to new job loss. A stock-price move can combine expectations about profits, interest rates, and risk appetite. It is not a direct reading of future output. A leading series can also give a false signal, turn too late, or fail to anticipate a shock that arrives suddenly.

A lead is an empirical relationship, not a permanent property of a series. It can weaken when the economy changes, when a policy affects one sector unusually, or when a shock reaches many sectors at once. OECD guidance notes that a broad shock can make a composite CLI behave more like a coincident signal for a time. That is why a break in an established pattern deserves a tentative reading until more observations arrive. {source:oecdInterpretingCompositeLeadingIndicators}

A small hypothetical sequence makes the timing label concrete. Suppose a selected reference measure reaches a peak in month 0. Permits turn down in month −3, payroll employment peaks around month 0, and average unemployment duration continues to rise until month +2. In that one invented sequence, permits lead, payrolls are coincident, and duration lags. The three-month and two-month offsets are teaching examples, not typical lead times, observed U.S. data, or a forecast rule.

<!-- learn:illustration -->

Coincident indicators tend to move near the same turning points as their reference measure. U.S. examples used in the Conference Board’s coincident composite include nonfarm payroll employment, personal income less transfer payments, industrial production, and manufacturing and trade sales. Together these measures cover work, income, production, and sales rather than relying on one narrow market. {source:conferenceBoardUsLeiTechNotes2026}

A coincident measure is useful for describing current conditions, but “current” still has a publication delay. A monthly series may be released after the reference month, and its first estimate may later be revised. A researcher looking at the economy in real time sees the data vintage available then, not the value that will appear in a revised historical chart years later. A good comparison therefore records the observation period, release date, adjustment status, and data vintage.

A series can also be coincident with one target and less so with another. The industrial production and capacity utilization guide explains how the two measures differ, but the U.S. economy includes a much larger service sector. Payroll employment captures jobs rather than output. No one series is a complete substitute for broad activity, and a coincident label does not make two measures interchangeable.

Lagging indicators tend to turn after the reference series. In the Conference Board’s U.S. lagging composite, examples include average duration of unemployment, commercial and industrial loans, the average prime rate, and inventory-to-sales measures. Unemployment duration may keep rising after output has begun to recover because it takes time for hiring to absorb people who lost work. Lending and inventories can adjust only after firms and households respond to the earlier change. {source:conferenceBoardUsLeiTechNotes2026}

A lagging indicator can help confirm that a change was broad or persistent, or show how the effects spread through credit, labor, and inventory decisions. But confirmation arrives after the initial turn. It is a poor choice if the question is whether to detect a turning point as early as possible. The category is still empirical: the timing can vary, and the same series may not lag every measure in every episode.

Do not read “lagging” as “caused by” the earlier activity measure. A sequence in time alone does not establish why a series moved. Credit conditions, policy, demographics, and sector-specific shocks can affect a lagging series too.

A composite index combines components so that a broad signal is less dependent on the noise in any one series. The Conference Board’s September 2026 technical note for its August U.S. release describes a Leading Economic Index with ten components, a Coincident Economic Index with four, and a Lagging Economic Index with seven. Its component groups and weights belong to that specific index methodology; they are not a universal definition of the three categories. The note also identifies the data vintage used and says that some inputs are estimated when source data are not yet available. {source:conferenceBoardUsLeiTechNotes2026}

The OECD selects CLI components separately for each country using criteria such as economic relevance, cyclical behavior, data quality, timeliness, and availability. Combining several series can make the signal more resilient to a disturbance affecting only one component and can produce more stable lead times than individual inputs. It cannot remove all missed turns or false alarms. Component selection, trend estimation, seasonal adjustment, missing-data treatment, and later source-data revisions all affect the published path. {source:oecdCompositeLeadingIndicatorsFaq} {source:oecdInterpretingCompositeLeadingIndicators}

A composite reading also needs its scale and transformation. The OECD CLI is normalized around 100 as a trend reference in its system. The Conference Board’s U.S. indexes use a stated base such as 2016=100 in the cited 2026 release. Those uses of 100 are not the same rule. A Conference Board value below its base of 100 does not, by itself, mean recession; a value above the OECD trend reference does not report a GDP growth rate of that size.

A leading index is one input to analysis. A recession chronology is a retrospective judgment about a broad decline, while a forecast estimates what may happen over a stated horizon. The National Bureau of Economic Research’s U.S. Business Cycle Dating Committee considers a range of monthly measures of aggregate activity. It says there is no fixed rule for which measures contribute information or how they are weighted, and it allows time for data revisions before dating peaks and troughs. {source:nberBusinessCycleDatingFAQ}

That process differs from reading a leading composite. The OECD CLI targets turning points relative to trend; NBER dates peaks and troughs in broad activity after reviewing multiple data series. Neither a single LEI decline nor a pair of quarterly GDP contractions automatically determines an NBER recession date. For a separate explanation of the U.S. Sahm Rule, see the Sahm Rule guide; for the gap between output and estimated capacity, see the output-gap guide.

A forecast is different again. It may assign a numerical growth rate to a specific quarter and incorporate assumptions about policy, prices, and international conditions. A leading composite usually communicates the direction or phase of a cycle under its own construction, not a precise GDP number. The OECD explicitly distinguishes its CLI from its numerical GDP projections. A rising or falling index should not be converted into a point forecast unless the publisher’s method actually supports that conversion.

Before interpreting a headline signal, write down what it refers to:

1. Target: Which country, aggregate measure, and cycle concept are used—activity level, deviation from trend, or growth rate?
2. Series: Is this one component or a composite? Which components are included, and are any inverted so that their signs run opposite to the original series?
3. Movement: Is the release reporting an index level, a monthly change, a six-month trend, or a diffusion measure? These quantities answer different questions.
4. Timing: What is the reference month, release date, and data vintage? Are some inputs estimated or revised?
5. Breadth: Do several components and independent measures point in the same direction, or is one volatile input driving the headline?
6. Claim: Does the source describe an early signal, a historical classification, or a quantitative forecast? Do not substitute one for another.

For example, a yield-curve inversion is a financial indicator that can provide recession information, but its interpretation depends on the spread, sample, and forecast horizon. The yield-curve guide explains those limits. Broad indicator sets help organize evidence; they do not turn uncertainty into a countdown.