Experimental Estimates of Compensation Levels and Trends for Workers in the 15 Largest Metropolitan Areas, 2004-05
by Michael K. Lettau and Christopher J. Guciardo
Bureau of Labor Statistics

Originally Posted: September 17, 2007

The BLS Employment Cost Index (ECI) is a principal Federal economic indicator for the United States. It measures the change in the cost of labor to employers over time. A related BLS program, the Employer Costs for Employee Compensation (ECEC), provides a comprehensive measure of the average cost in terms of dollars per hour of total compensation for U.S. workers, including the cost to employers for worker benefits, such as health insurance and retirement programs, along with wages and salaries. Since the inception of the two series, the only ECI and ECEC estimates available for geographic areas smaller than the entire United States have been for the broad groups of States that make up Census regions and divisions.

This article introduces experimental ECI and ECEC estimates for the 15 largest metropolitan areas in the United States, as ranked by total population in 2000.¹ The first section of the article reports the new ECI estimates, which show 12-month rates of change in compensation covering the period 2003 to 2005. The next section reports the ECEC estimates for the level of compensation in the 15 areas for March 2004 and March 2005. The third section discusses how these experimental estimates differ from the other compensation estimates currently available for metropolitan areas from BLS.

ECI for the 15 Largest Metropolitan Areas

Employment Cost Index (ECI) estimates for 15 U.S. metropolitan areas were calculated using the same index number formula used by the national ECI.² Employment for industry-occupation groups, which the index formula holds constant over time, has a reference period of May 2005 and refers to employment for the industry-occupation groups within the particular area.³

The upper half of table 1 shows 12-month changes in the ECI for total compensation, starting with the change from December 2003 to December 2004 and ending with the change from December 2004 to December 2005. Total compensation equals wages and salaries plus employer costs for 18 categories of benefits, so it is designed to be a comprehensive measure of the change in the cost to businesses of employing workers. Although the National Compensation Survey (NCS), which encompasses the ECI and ECEC programs, also includes workers in State and local government, the estimates in this article are restricted to workers in private industry. The far right-hand column of table 1 shows the rate of change from December 2003 to December 2005, converted to an average annual rate. The conversion is defined so that if the average annual rate were applied for 2 consecutive years, the net change over this time would equal the actual change from December 2003 to December 2005. The lower half of the table shows standard errors corresponding to the estimates in the top half. Data for the Nation (all areas), for private industry, are included for comparison.

For the period studied, the Seattle metropolitan area showed the highest estimate for the average annual rate of change (7.4 percent), and the Atlanta metropolitan area showed the lowest estimate (2.2 percent). Note that these are estimates of true population rates of change, and hence are subject to sampling error. To test the significance of these claims, we performed a multiple comparison test. The claim that Seattle has the largest rate among the 15 areas is statistically significant at the 10-percent confidence level, although the claim that Atlanta has the lowest rate is not statistically significant at the 10-percent level.⁴ For comparison, the estimate for the average annual rate of change in compensation for the entire United States over the same period is 3.3 percent. The standard errors associated with the 12-month rates of change for the areas range between 0.1 and 1.5 percentage points. The standard errors for the average annual rates of change over the 2-year period tend to be a bit lower, ranging from 0.2 to 0.6 percentage point.

When comparing the estimates of the growth rates between a single pair of areas, it is important to take into account their reliability. To help with this, table 2 shows the minimum difference in the growth rates required for statistical significance. It gives the value at which the difference becomes significant at the 10-percent level for the combination of standard errors defined by the row and the column. For example, the December 2003-to-December 2004 change for Washington DC equals 4.4 percent with a standard error of 0.4, and the December 2003-to-December 2004 change for San Francisco equals 2.5 percent with a standard error of 0.7. Because the difference, 1.9, in their growth rates is greater than 1.3, which is the significance threshold in table 2 associated with the pair of standard errors 0.4 and 0.7, it is statistically significant at the 10-percent level.⁵ Conversely, the March 2004-to-March 2005 changes for Chicago and Dallas are 2.5 percent and 2.8 percent, respectively. The standard errors for both areas are 0.5, so the difference, 0.3, in their growth rates is not statistically significant because it is less than 1.2 percentage points.

The upper half of table 3 shows 12-month percent changes in the ECI for wages and salaries, with the far right-hand column again showing the average annual rate of change over the 2-year period. Wages and salaries account for about 70 percent of total compensation, and employer costs for several of the benefits, such as paid leave, are tied directly to wage rates; thus, the growth in wages and salaries tends to be similar to the growth in total compensation. As with total compensation, the metropolitan area with the highest estimated average annual growth rate for wages and salaries from December 2003 to December 2005 was Seattle, at 4.2 percent, and the area with the lowest growth rate was Atlanta, at 1.1 percent. Yet, now both claims are not statistically significant at the 10-percent level. The lower half of the table shows the standard errors corresponding to the estimates in the upper half.

ECEC for the 15 Largest Metropolitan Areas

Employer Costs for Employee Compensation estimates for the 15 largest U.S. metropolitan areas were calculated using procedures that are similar to what is used currently for the national ECEC.⁶

Table 4 shows ECEC estimates for March 2004 and for March 2005, along with the corresponding standard errors and percent relative standard errors. The percent relative standard error is the standard error expressed as a percent of the average cost, so it equals 100 times the standard error divided by the average cost. In comparison to the ECEC estimates for the metropolitan areas, the ECEC for total compensation for all U.S. workers in private industries was $23.29 in March 2004 and $24.17 in March 2005. The standard errors for almost all the estimates are less than $1.00, resulting in percent relative standard errors primarily in the range of 1.5 to 3.0 percent.

Whether the difference between the ECEC estimates for two areas is statistically significant depends on the magnitude of the two estimates and their relative standard errors. As a very rough guideline, if the difference between the estimates for any two areas from table 4 exceeds $2.00, it is generally statistically significant at the 10-percent level.⁷ For example, the March 2004 estimate for Dallas equals $24.63, and the March 2004 estimate for Miami equals $20.36. Their estimated difference in average compensation of $4.27 is statistically significant. The March 2004 estimate for Phoenix equals $24.22. Its estimated difference of $3.86 from the estimate for Miami is also statistically significant at the 10-percent level, but its difference of $0.41 from the estimate for Dallas is not.

The ECEC estimate for a metropolitan area reflects the local composition of the workforce, so differences in the composition of workers among the areas will affect any comparison of their average total compensation. The National Compensation Survey reported estimates of "pay relatives" for about 80 metropolitan areas for July 2005. Pay relatives compare average hourly earnings among the areas after adjusting for differences in their establishment and occupational characteristics and their occupation composition.⁸ For some purposes, they may provide more appropriate comparisons of the average level of pay among metropolitan areas.⁹

Table 5 shows employer costs for total compensation broken down into the cost for wages and salaries and the total cost for benefits. The total cost for benefits equals the sum of the costs for paid leave, premium pay for overtime, insurance, retirement and savings, and such legally-required benefits as Social Security, Medicare, and workers compensation. Metropolitan areas with an above-average ECEC estimate for wages and salaries tend to have above-average ECEC estimates for total benefits.

San Francisco has the highest estimate for average wages and salaries, at $22.64, although this cost is not statistically different (at the 10-percent significance level) from the second-highest estimate, that for Boston, at $22.33. Miami had the lowest estimate for average wages and salaries. San Francisco and Boston also had relatively high average benefits costs, although Detroit had the highest; Miami had the lowest cost of benefits.

ECEC cost levels are not the only interesting estimates; also of interest is how employer compensation is allocated between wages versus benefits. Table 5 also displays estimates for the average costs as a percent of total compensation.

Other Measures of Compensation for Metropolitan Areas

Because BLS currently publishes other estimates of average pay for metropolitan areas in the United States, it is important to understand how the new experimental ECI and ECEC estimates differ from them. The Occupational Employment Statistics (OES) program reported average hourly wages in approximately 500 areas for May 2005, and the National Compensation Survey program reported average hourly earnings in about 80 areas during the calendar year 2005. Also, as mentioned previously, the NCS published pay relatives for about 80 metropolitan areas for July 2005 that compare average hourly earnings among the areas after adjusting for differences in workforce composition between the areas.¹⁰

With other estimates for average pay in metropolitan areas already available from BLS, one might ask, Why are additional estimates from the ECI and ECEC programs warranted? Table 6 shows 2005 estimates of compensation for the entire United States and for the Los Angeles metropolitan area from various BLS programs. The ECEC and ECI estimates for Los Angeles (as well as other metropolitan areas) are available for the first time and show compensation broken into its wage component and its benefits component.

Among the new experimental estimates, the locality ECEC estimates for wages are the most like other wage estimates for metropolitan areas currently published by BLS. They are designed to measure wages using a concept similar to that used to produce the NCS occupational wages and the OES hourly wages, although differences in the scope and calculation of the estimates make them unlikely to align exactly.¹¹ The other new statistics provide information about compensation in 15 metropolitan areas that had not been available previously. In particular, the locality ECEC estimates for benefits provide information on the costs to employers for nonwage compensation, and the locality ECI estimates provide a measure of the change in compensation costs over time for the area using a fixed-weight index.¹²

Future Plans for the Statistics

The estimates in this article were calculated as part of a research project to determine the feasibility of publishing ECI and ECEC estimates for U.S. metropolitan areas and therefore are considered experimental--that is, they are not official BLS estimates. BLS plans to begin regular publication of ECI and ECEC estimates for metropolitan areas in the coming years.

Michael K. Lettau
Research Economist, Compensation Research and Program Development Group, Office of Compensation and Working Conditions, Bureau of Labor Statistics.
Telephone: (202) 691-7396; E-mail: Lettau.Michael@bls.gov

Christopher J. Guciardo
Mathematical Statistician, National Compensation Survey Sampling and Collection Branch, Office of Compensation and Working Conditions, Bureau of Labor Statistics.
Telephone: (202) 691-6151; E-mail: Guciardo.Chris@bls.gov

Notes

Acknowledgement: The authors would like to thank Thomas G. Moerhle for his help in calculating the experimental series for this article.

1 The definitions of the metropolitan areas for this article are based on the Office of Management and Budget (OMB) area definitions from the 1990 decennial census. The National Compensation Survey, under which the data for the ECI and ECEC are collected, has begun to switch to area definitions based on the 2000 decennial census. See Jason Tehonica, "New Area Sample Selected for the National Compensation Survey," Compensation and Working Conditions Online, March 30, 2005, on the Internet at http://www.bls.gov/opub/cwc/cm20050318ar01p1.htm.

2 See Donald G. Wood, "Estimation procedures for the Employment Cost Index," Monthly Labor Review, May 1982, pp. 40-42, on the Internet at http://www.bls.gov/opub/mlr/1982/05/rpt3full.pdf.

3 The national ECI uses a Laspeyres formula, which holds industry-occupation employment constant at values during an initial period to isolate the change in compensation not caused by a change in the composition of workers. For example, since March 2006, the national ECI has used industry-occupation employment with a reference period of 2002. The ECI estimates in this article use a reference period for employment that is either during or after the period of the change, so they are not strictly consistent with the Laspeyres formula of using a reference period prior to the change. Historically, however, the growth rate in the ECI has not been sensitive to the choice of reference period for the industry-occupation employment. See Michael K. Lettau, Mark A. Loewenstein, and Steve P. Paben, "Is the ECI sensitive to the method of aggregation? an update," Monthly Labor Review, December 2002, pp. 23-28, on the Internet at http://www.bls.gov/opub/mlr/2002/12/art3abs.htm.

4 Each test computes the probability p that an area has the maximum (or minimum) estimate merely as a result of sampling error, rather than because the area’s true population rate is large (or small). Therefore, if p is less than 10 percent, then we are "more than 90 percent confident" that the area’s population rate is indeed the true population maximum (or minimum)--that is, our claim is statistically significant at the 10-percent level.

5 The calculation of the minimum difference required for statistical significance at the 10-percent level equals , which equals approximately 1.3 percentage points.

6 See Albert E. Schwenk, "Measuring Trends in the Structure and Levels of Employer Costs for Employee Compensation," Compensation and Working Conditions, summer 1997, pp. 3-14, on the Internet at: http://www.bls.gov/opub/cwc/archive/summer1997art1.pdf.

7 The calculation of the minimum difference required for statistical significant at the 10-percent level equals , where ECEC1 is the ECEC estimate and percent RSE1 is the percent relative standard error for the first area, and ECEC2 is the ECEC estimate and percent RSE2 is the percent relative standard error for the second area.

8 See Maury B. Gittleman, "Pay relatives for metropolitan areas in the NCS," Monthly Labor Review, March 2005, pp. 46-53, on the Internet at http://www.bls.gov/opub/mlr/2005/03/art4abs.htm.

9 The pay relatives only cover wages and salaries. BLS does not report estimates for total compensation relatives.

10 See Gittleman, "Pay Relatives for Metropolitan Areas in the NCS."

11 See Chapter 3 of the BLS Handbook of Methods for a description of the procedures used for the OES estimates, available on the Internet at http://www.bls.gov/opub/hom/homch3_a.htm. See Chapter 8 of the Handbook for a description of the procedures used for the NCS estimates, available on the Internet at http://www.bls.gov/opub/hom/homch8_a.htm.

12 See John W. Ruser, "Employment Cost Index: What is it?," Monthly Labor Review, September 2001, pp. 3-16, available on the Internet at http://www.bls.gov/opub/mlr/2001/09/art1abs.htm.