Relative importance: putting the Employment Cost Index (ECI) into perspective

The Employment Cost Index (ECI) is a Principal Federal Economic Indicator (PFEI) published by the U.S. Bureau of Labor Statistics (BLS) that measures the change in the hourly labor cost to employers over time.¹ More specifically, it estimates fluctuations in wages, salaries, and benefits (for example, paid leave, supplemental and premium pay, insurances, retirement and savings, and legally required benefits) paid by employers for work performed by employees. One key feature of the ECI is that it holds fixed the distribution of employment across ownership, industry, and occupation groups so these measures reflect pure changes in the price of labor.² It offers a measure of how much more or how much less it costs to employ a given workforce from one period to another. Since these weights are periodically updated, the ECI is a type of modified Laspeyres index.³ In addition to giving a snapshot for the aggregate costs of employment, BLS also publishes ECI subindexes, which track these changes by industry, occupation group, geographic location, bargaining status (union and nonunion), and ownership (civilian, private industry, and state and local government).⁴

The rest of the article is organized as follows. We discuss the benefits of the ECI and comment on some challenges to understanding how the index estimates for subgroups (for example, by industry group or occupation group) are reflected in the aggregate, headline figures. Then, we introduce relative importance (RI) percentages that put these movements into context by showing how compensation costs are distributed across job types. Using these percentages, we create relative importance–employment distribution ratios (RI–ED), which reveal a given group’s labor costs relative to a representative worker. Finally, we discuss the dynamic behavior of RI values over time and across index reweight periods.

Using and interpreting ECI estimates

The ECI can be a useful tool to determine changes in the costs of employing labor by certain job and establishment characteristics. For example, consider a firm looking to assess compensation costs associated with its workers. The firm uses the appropriate ECI subindex and finds that costs for wages and benefits of workers like those it employs increased by 2 percent. The firm can decide if it wants to similarly increase wages and benefits by 2 percent or perhaps exceed the published estimate to attract or retain workers. In a slightly more involved example, the firm may be constructing a business plan and must forecast its labor costs into the future. The firm can use historical ECI estimates to conduct these slightly more advanced statistical analyses.⁵ Jobseekers can use the ECI to see how compensation is changing by industry, occupation, and location. Policymakers can use it to understand the changing U.S. economy or to assess the efficacy of their legislation.

While the ECI has many potential uses—only a few of which were highlighted above—it does not directly contextualize how aggregate trends and fluctuations are influenced by different groups of jobs.⁶ Suppose an economy is made up of two occupations, A and B. Suppose also that ECI estimates suggest compensation growth for those employed in occupation A is 10 percent, while growth for those in occupation B is only 1 percent. If the overall ECI estimates suggest compensation costs increase by 2 percent, we should conclude that occupation A is a lower proportion of total compensation expenditures than occupation B in the index’s population. In the ECI, expenditure shares are proportions of cost weights, which are the product of the fixed employment weights and average compensation costs. Recognizing this pattern becomes more complex when we consider many groups.

Relative importance percentages

This article introduces a diagnostic tool, currently being considered for publication by BLS, to measure how relatively important compensation costs are for a domain (a group of jobs defined by ownership, industry, occupation, or some other identifier) within a published ECI series. These RI tables report the wage and benefit cost weight shares of different published domains. While cost weights represent wage (or benefit) bill expenditures, their RI percentages describe how these expenditures are distributed across job types in the ECI’s weight structure.

Before we formalize the definition of a series’ RI value, we note that an ECI is constructed as follows:

$W_{0,i}={\overline{Y}}_{0,i}{E}_i$

$R_{t,i}=\frac{{\overline{Y}}_{t,i}}{{\overline{Y}}_{t-1,i}}$

$M_{t,i}=\prod _{s=1}^t{R}_{s,i}$$I_t=\frac{\sum _i{W}_{0,i}{M}_{t,i}}{\sum _i{W}_{0,i}}\times 100$

$I_t=\frac{\sum _i{W}_{0,i}{M}_{t,i}}{\sum _i{W}_{0,i}}\times 100$

where

• i is the cell integer index,

• t is the reference period integer index,

• $W_{0,i}$ is the cost weight (wage or benefit bill) for cell i in the base period,

• ${\overline{Y}}_{t,i}$ is the weighted average cost in reference period t for cell i,

• $E_i$ is the base period employment (weight) for cell i,

• $R_{t,i}$ is the matched-quote relative of cell i in reference period t,

• $M_{t,i}$ are multiplicatively accumulated relatives from the base period through t, and

• $I_t$ is the ECI in reference period t.

More information on the calculation details of the ECI, including cells and matched-quotes, is available in the Employment Cost Index chapter of the Handbook of Methods.⁷

With this notation established, the RI of group g from the set of mutually exclusive and exhaustive groups G is defined as follows:

${\mathrm{\theta }}_t^g=\frac{\sum _{i\in g}{W}_{0,i}{M}_{t,i}}{\sum _i{W}_{0,i}{M}_{t,i}}\times 100$

G may be the set of ownership groups, occupation groups, or industry groups, among others. The numerator of the above equation is the sum of cost weights for cells in group g, while the denominator is the total cost weight of all cells. It is easy to verify:

$\sum _{g\in G}{\mathrm{\theta }}_t^g=100$

Table 1 presents RI values for private industry workers by occupation groups in December 2024, augmented with ECI 3-month percent change estimates for March 2025 and employment distribution shares for comparison. This example will explain how to interpret RI estimates and explain how RI estimates are related to ECI data. The employment share calculations use Occupational Employment and Wage Statistics (OEWS) employment counts by ownership, industry, and occupation group, which are then benchmarked to match employment counts by ownership and industry group from the Quarterly Census of Employment and Wages (QCEW) in March 2021.⁸

For this period, management, professional, and related occupations made up 49.5 percent of total wage cost weights and 52.4 percent of total benefit cost weights. Sales and office occupations, in contrast, were only 18.7 and 16.6 percent of wage and benefit cost weights, respectively. How do these RI values relate to the ECI estimates themselves? A quick calculation reveals that the all occupations’ estimate is approximately the arithmetic average of 3-month percent changes across subgroups with the prior RI values as weights. For benefits, we have the following:
 

1.7 ≈ (1.9 x 0.524) + (1.9 x 0.166) + (1.1 x 0.092) + (1.4 x 0.144) + (1.7 x 0.074)

where the approximate qualifier is an artifact of the rounding of published estimates. Indeed, with some algebraic manipulation, we can show this relationship is exact. That is, aggregated ECI 3-month percent change estimates are weighted arithmetic averages of its subgroup estimates, with last period’s RI values as the weights:

$\left(\frac{I_t}{I_{t-1}}-1\right)=\frac{1}{100}\sum _{gϵG}{\theta }_{t-1}^g\left(\frac{I_t^g}{I_{t-1}^g}-1\right)$

The relative importance–employment distribution ratio

Table 1 also allows for comparison between RIs and shares of the employment distribution. For management, professional, and related occupations, the 49.5 percent (wages) and 52.4 percent (benefits) of total cost weights are contrasted with only a 30.5-percent share of the employed workforce, meaning these workers cost relatively more than their proportion in the aggregate economy. On the other side, service occupations account for 19.9 percent of employment but are only 10.8 percent of wage cost weights and 7.4 percent of benefit cost weights. To create a single measure that highlights the relationship between ECI cost weight shares and employment distribution shares, we next calculate a relative importance–employment distribution ratio (RI–ED). Table 2 contains the RI–ED, along with December 2024 Employer Costs for Employee Compensation (ECEC) mean cost estimates.⁹

If wages and benefits were the same across groups, RI–ED ratios would equal 1.00 for each. Deviations from this benchmark indicate how much more (if greater than 1.00) or less (if less than 1.00) a given group’s labor costs are relative to those of the index’s representative (average) worker. For example, natural resources, construction, and maintenance workers cost 2 percent less in wages than the average among all private workers, and production, transportation, and material moving workers cost 26 percent less. This is a powerful result, because it allows data users another avenue to evaluate relative differences in levels (of mean wage costs, in this example) of the index population. Analytically, we can show the following:

RI–ED =$\left(\frac{\sum _{i\in g}{\overline{Y}}_{0,i}{M}_{t,i}{E}_i}{\sum _{i\in g}{E}_i}\right)\times {\left(\frac{\sum _i{\overline{Y}}_{0,i}{M}_{t,i}{E}_i}{\sum _i{E}_i}\right)}^{-1}$

Note that ${\overline{Y}}_{0,i}{M}_{t,i}$ is the average cost of cell i in reference period t.Then, the first term in the above equation is the weighted average cost of group g. Since the second term is the weighted average cost of all workers, the RI–ED is a measure of a group’s cost relative to the index’s representative worker. We can validate this interpretation using ECEC estimates. Dividing an occupation group’s mean cost by the all occupation mean cost roughly corresponds to the RI–ED value. For instance, using the wages values in Table 2, the mean hourly wage cost for all occupations is $31.47. The mean hourly wage cost for management, professional, and related occupations is $51.52. Dividing the occupation group’s mean cost by the overall mean cost gives 1.64, closely approximating the RI–ED ratio of 1.62. Note that this calculation will not be exact because the weighting in the ECI (fixed) differs from the ECEC (variable). Nevertheless, it offers an easy check. 
 

RI values over time

To get a sense of RI values over time, we report occupation group data for several reference periods in table 3. Both wage and benefit RI values were relatively stable from December 2019 through September 2022. One might wonder why there is such little change in the RI of labor costs after the huge economic shock caused by COVID-19. For example, we might expect service occupations to present a large negative effect, knowing that such workers were hit particularly hard. However, the ECI measures the cost of employing a particular type of labor while holding occupation and industry composition fixed. As workers are laid off, total expenditures on wages decrease, but the cost per hour worked to employ a worker may not change. The fixed-weight nature of the ECI masks these decreases. In other words, while service worker employment was disproportionately affected by the COVID-19 shock, the ECI considers changes in the cost of labor except for these distributional shifts.¹⁰

The only noticeable changes observed in table 3 are seen between September and December 2022 for management, professional, and related occupations (an increase) and sales and office occupations (a decrease). These do not stem directly from the pandemic but rather a reweight in December 2022 using employment data for 2021.

Over long periods, fixed-weight indexes may imply a basket that differs greatly from realized purchases. Contrast this with chained indexes, which update weights each period (or, at least, very frequently). These differences highlight the distinction between pure price and cost-of-living indexes. Pure price indexes measure price change holding composition fixed, and cost-of-living indexes allow for substitution. Both are useful in their own contexts. Pure price changes are important in discussions about inflation, while cost-of-living changes are valuable when evolving consumer (or establishment) preferences are front and center. In any case, the ECI belongs to the pure price variety, but to keep the index relevant for contemporary data users, reweighting takes place infrequently (about every 10 years).¹¹ 

The design of the ECI and its reweighting schedule contribute to a dynamic pattern wherein RI values are relatively flat between reweight dates but may jump across them. These jumps reflect shifts in employment, which have accumulated since the base period (or last reweight). We show this for occupation groups in chart 1 from the ECI’s current base period (December 2005) through December 2024. The RI percentages for the five occupation groups are stacked so that they sum to 100 in each reference period. Vertical black lines are included to indicate the timing of reweights. From chart 1, we can see the change in RI values when the ECI is reweighted. Management, professional, and related occupations have grown in their relative importance over time, while sales and office occupations have declined. The other groups have seen changes, but not as large.

Conclusion

In summary, the ECI is an essential measure to assess changes in the cost of labor. The relative importance (RI) values introduced in this article help put into perspective how these costs fit into overall trends in the economy. They represent the wage and benefit cost weight shares of different published ECI series (such as occupation groups or industry groups), and describe how compensation expenditures are distributed across various domains. The exercise constructing relative importance–employment distribution ratios shows how these diagnostics can be combined with other information (namely employment shares) to reveal relative differences in the cost levels of different segments of the employed labor force. Finally, observing RI values over time reinforces a core feature of the ECI as a pure price index. This is because the ECI uses a fixed basket and does not capture changes in compensation driven by shifts in industry and occupation composition.