Employment Cost Index Explained
Table of Contents
The Employment Cost Index (ECI) helps answer a question that an average wage cannot settle: are employers paying more for a given mix of jobs, or has the mix of workers changed? If lower-paid jobs disappear, average hourly earnings can rise even when nobody remaining receives a raise. That distinction matters when judging compensation pressure, business costs and monetary-policy news.
This guide explains the U.S. Bureau of Labor Statistics measure, separates wages from benefits, and works through a historical quarterly release. It focuses on compensation measurement; the timing of business cycle indicators is a separate question.
What Is the Employment Cost Index?
The ECI measures changes in hourly compensation costs to employers, controlling for shifts in employment among industries and occupations. BLS publishes indexes for total compensation, wages and salaries, and benefits. Its headline civilian coverage combines private industry with state and local government. [1]
The ECI asks how compensation costs would change with a fixed industry-and-occupation employment mix. It measures the cost of jobs rather than following the pay of the same individual workers.
“Civilian” does not mean every worker. Federal government and military workers are excluded, as are the self-employed, agricultural workers and workers in private households. The national measure covers the 50 states and the District of Columbia, rather than U.S. territories. Private-sector and state/local-government series should be identified separately when they are the population of interest. [1]
This is also different from counting jobs or estimating the unemployment rate and labor force participation rate. Employment can contract while compensation costs per hour continue to rise.
Wages, Benefits and Total Compensation
A salary increase captures only part of an employer’s labor bill. Health insurance, retirement contributions and paid leave also cost money. BLS separates those components so a change in one does not automatically become a claim about another.
| Component | What it measures | Interpretation caution |
|---|---|---|
| Wages and salaries | Employer wage and salary costs, including production bonuses and commissions. | Not after-tax take-home pay or the typical worker’s personal raise. |
| Benefits | Employer costs for paid leave, supplemental pay, insurance, retirement and legally required benefits. | A higher employer cost does not necessarily mean a more generous benefit for employees. |
| Total compensation | Wages and salaries plus benefits. | Its growth rate is not the sum or simple average of the two component growth rates. |
BLS classifications can differ from everyday usage. For example, a production bonus belongs in wages, while a nonproduction bonus and overtime premium belong in supplemental benefits. Read the component definitions before comparing a company’s pay announcement with a published ECI series. [1]
These are employer costs. An insurance premium can increase without an employee receiving a comparable improvement in coverage. Costs can also change when employers and employees change how they share the bill. [3]
How the ECI Holds the Employment Mix Fixed
BLS obtains compensation information through the establishment-based National Compensation Survey. It compares sampled jobs for which current- and previous-quarter information is available, groups them by ownership, industry and occupation, and estimates changes in hourly wage and benefit costs. Employment counts for these groups are held fixed between periodic reweightings. [2]
The design has two useful layers. Matching jobs reduces distortions from changes in the sampled jobs. Fixed employment weights prevent a shift toward a high-paying industry or occupation from being treated as an increase in the cost of the original employment mix. More expensive job groups contribute more to the compensation bill; this is not an equal-weight average of every group’s percentage change.
This expression teaches the fixed-employment principle. The official ECI is a modified Laspeyres index with matched-job cost relatives, survey weights, nonresponse adjustments, imputation and periodic reweighting. Reproducing the official index requires those additional procedures. It is not calculated by dividing this quarter’s total national payroll by this quarter’s headcount. [2]
Methodology as of November 15, 2026: BLS uses 2021 employment weights for December 2022 through September 2026 data. It has announced that the January 2027 release of December 2026 data will introduce 2025 weights and remove workers’ compensation costs from the ECI. “Fixed” therefore means fixed between reweightings, not frozen forever. [6]
Why Average Pay Can Rise Without Raises
Consider a hypothetical employer with two job groups: 100 workers earning $20 per hour and 100 earning $40. Everyone works the same number of hours, so worker counts and hours have the same relative weights. Next, 50 of the lower-paid jobs disappear. All remaining workers keep their original hourly pay; benefits are excluded from this wage-only illustration.
| Job group | Before | After | Hourly pay in both periods |
|---|---|---|---|
| Lower-paid jobs | 100 workers | 50 workers | $20 |
| Higher-paid jobs | 100 workers | 100 workers | $40 |
The initial average is (100 × $20 + 100 × $40) / 200 = $30 per hour. After the job losses, it is (50 × $20 + 100 × $40) / 150 = $33.33 per hour. Using unrounded values, average pay rises ($33.3333 / $30 − 1) × 100 = 11.1%, even though every remaining worker’s rate is unchanged.
Keep the original 100 jobs of each type in the fixed basket instead. Its cost is 100 × $20 + 100 × $40 = $6,000 for one hour of work per job in either period. The simplified wage index remains 100, with a 0% change. This is the composition problem the ECI’s construction addresses; the illustration is not a reconstruction of the full BLS estimator.
The Historical Lesson From April 2020
The issue was especially visible during the pandemic. In its explanation of the April 2020 employment report, BLS warned that the disproportionate removal of lower-paid workers from payrolls put upward pressure on average hourly earnings. Some workers did receive increases, but the higher average could not be read as a general improvement in pay for the same workers. Federal Reserve staff subsequently used this episode to explain why different wage measures can diverge. [8] [7]
The fictional example isolates composition by holding every pay rate constant. The actual episode involved many simultaneous changes. Do not treat its 11.1% result as a historical statistic, or compare a monthly average-earnings movement directly with a quarterly ECI movement without aligning their periods.
ECI vs Average Hourly Earnings and ECEC
These BLS measures answer different questions. A useful comparison starts with their units, employment coverage and weighting, rather than deciding that one number must be wrong.
| Measure | Main question | Key distinction |
|---|---|---|
| Employment Cost Index | How are hourly employer costs changing for a fixed employment mix? | Quarterly indexes and growth rates; wages, benefits or total compensation; civilian or specified ownership coverage. |
| Average hourly earnings (AHE) | What are average hourly earnings on current private nonfarm payrolls? | Monthly CES earnings divided by hours; affected by changes in worker and industry mix; excludes employer benefit costs. |
| Employer Costs for Employee Compensation (ECEC) | How many dollars per hour do wages and benefits cost employers? | Cost levels using current employment weights; changes in those levels include composition effects. |
The ECI is especially useful for separating compensation-cost changes from broad employment shifts. AHE offers more frequent information. ECEC supplies dollar cost levels and component shares that an index cannot provide. Differences can reflect their designs, coverage and timing. [4] [7]
How to Read a Quarterly ECI Release
ECI observations refer to the pay period including the 12th of March, June, September or December. They are quarterly observations, not a monthly series or a continuous three-month average of everybody’s paycheck. Releases normally follow in January, April, July and October. [3] [5]
For a stable worked example, use the January 31, 2025 release covering December 2024. Its Table A reports the following civilian-worker figures. The quarterly columns are seasonally adjusted; the annual column is not seasonally adjusted. All are current-dollar changes, and these are the archived release’s figures rather than a claim about the latest data.
| Component | 3 months to Sep. 2024, SA | 3 months to Dec. 2024, SA | 12 months to Dec. 2024, NSA |
|---|---|---|---|
| Total compensation | 0.8% | 0.9% | 3.8% |
| Wages and salaries | 0.8% | 0.9% | 3.8% |
| Benefits | 0.8% | 0.8% | 3.6% |
Source: BLS archived release, Table A. SA means seasonally adjusted; NSA means not seasonally adjusted.
The total-compensation quarterly pace edged up, while its trailing annual rate was below the 4.2% reported for the year ending December 2023. That is a 4.2% − 3.8% = 0.4 percentage-point annual deceleration. It does not mean compensation costs fell: they were still rising. One faster quarter and a slower trailing year can coexist because the annual calculation includes earlier quarters. [4]
Index Points Are Not Percentage Changes
Table 1 in that release puts the seasonally adjusted civilian total-compensation index at 166.8 in September 2024 and 168.3 in December, with December 2005 equal to 100. The level is a relative cost index, not a dollar wage.
The increase is 168.3 − 166.8 = 1.5 index points, but the percentage change is (168.3 / 166.8 − 1) × 100 = 0.8993%, approximately 0.9%. Published rates use more precise underlying values, so rounded displayed levels need not reproduce every published change exactly. [3] [4]
Annualized Is Different From Year Over Year
A 0.9% quarterly change is not an annualized figure. If that exact pace repeated for four quarters, compounded growth would be [(1.009)4 − 1] × 100 = 3.6489%, or about 3.6%. That conditional calculation uses the rounded quarterly rate; it is neither a forecast nor the actual 3.8% trailing annual change. The release’s quarterly and annual columns also use different seasonal conventions.
What ECI Can Tell Investors and Policymakers
A broad rise in compensation costs can help an analyst understand employers’ cost pressures and competition for labor. The economics of labor demand explains why productivity and the demand for workers matter, while the Phillips curve provides a separate framework for discussing labor-market conditions and inflation.
The Federal Reserve’s February 2022 Monetary Policy Report discussed ECI alongside other wage measures and labor-market indicators. It also stressed that the inflation implications of wage gains depend on productivity growth. A higher hourly labor cost can be partly offset by more output per hour; ECI alone does not measure labor cost per unit of output. [9]
Nor does a compensation increase pass mechanically into consumer prices. Firms’ output, margins, nonlabor costs and demand conditions matter. Comparing ECI with consumer price inflation can help distinguish nominal and real changes, but those indexes measure different things. There is no universal ECI cutoff that by itself determines a rate decision, a recession or an investment trade.
For each release, work through these checks:
- Identify the population and component. Civilian total compensation is different from private-industry wages or state/local benefits.
- Record the observation and release dates. Separate the quarter being measured from when the information became available.
- Match the horizon and seasonal treatment. Use comparable quarterly rates for recent momentum and comparable annual rates for the longer view.
- Read the components and relevant sectors. Ask whether the movement is broad or concentrated, while respecting each subgroup’s construction.
- Keep the vintage. Preserve the original release when studying what investors knew then; identify revisions when using a current historical series.
- Cross-check the interpretation. Review employment, productivity and price data before drawing conclusions about business conditions or the phase of the business cycle.
Common Mistakes When Reading ECI
- Calling total compensation “wages.” Benefits can behave differently; select the wages-and-salaries series when that is the question.
- Reading a higher average as a universal raise. AHE can change with composition, and ECI measures jobs rather than the same people’s pay.
- Adding component growth rates. Wage and benefit costs combine in dollars; their percentage changes require appropriate cost weights.
- Mixing units or horizons. Index points, quarterly percent changes, annualized rates and trailing annual growth are distinct quantities.
- Equating slower growth with falling costs. A positive growth rate still increases the compensation-cost level.
- Turning the release into a standalone forecast. A compensation observation does not by itself establish future prices, policy or market returns.
Limitations of the Employment Cost Index
ECI deliberately holds broad employment composition fixed. It therefore does not describe the current workforce’s average dollar compensation, total payroll, each person’s purchasing power or worker welfare.
Fixed industry-and-occupation weights do not eliminate every composition issue. Workers’ experience or productivity can change within a job. ECI is not a same-person wage tracker, and changes in who fills a job can matter in ways a job-cost measure does not fully capture. [7]
Detailed series also need care. BLS uses varying sample distributions for regional, union/nonunion and excluding-incentive-worker indexes; those cuts are not strictly comparable to the main fixed-weight aggregate, industry and occupational indexes. The ECI does not provide a separate index for every state or employer. [2] [5]
Finally, ECI is a sample estimate with nonresponse, imputation and sampling uncertainty. A small difference between rounded quarterly rates does not establish a statistically significant change. BLS publishes standard errors for unadjusted changes; seasonally adjusted histories are revised through the annual seasonal process for the most recent five years. Unadjusted series are normally final on publication, subject to correction of errors. [4] [5]
Frequently Asked Questions
Sources
- BLS Handbook of Methods: Employment Cost Index, Concepts. Coverage and wage/benefit definitions.
- BLS Handbook of Methods: Employment Cost Index, Calculation. Matched jobs, fixed employment weights, aggregation and subgroup qualifications.
- BLS: Employment Cost Index Questions and Answers. Reference periods, index changes and constant-dollar estimates.
- BLS: Employment Cost Index—December 2024, January 31, 2025 release. Archived Table A, Table 1 and technical note.
- BLS Handbook of Methods: Employment Cost Index, Presentation. Publication schedule, uses and revision policy.
- BLS: Introducing 2025 fixed employment weights and removing workers’ compensation costs. February 11, 2026 notice announcing changes effective with December 2026 data.
- Ahn, Moyon and Villar: What do various wage measures tell us about underlying wage growth?. Federal Reserve staff FEDS Notes, January 8, 2025; measurement comparisons and limitations.
- BLS: COVID-19 and the April 2020 Employment Situation. Historical explanation of composition effects in average earnings.
- Federal Reserve: February 2022 Monetary Policy Report, Part 1. Compensation measures and productivity in policy analysis.
Sources accessed November 15, 2026. Historical release figures retain their stated publication vintage; the workforce-mix calculation is hypothetical.
Disclaimer
This article is for educational and informational purposes only and does not constitute investment, employment or policy advice. Economic estimates, methodologies and seasonal histories can change. Use the relevant official series and release when making decisions.
