June 3, 2011

June 3, 2011 | Testimony on Labor

Response to the EPI: Davis–Bacon Rates Set Well Above Market Pay

Dear Chairman Walberg:

The Davis–Bacon Act (DBA) requires contractors on federally financed construction projects to pay at least the locally prevailing wage to their employees. These minimum Davis–Bacon rates are calculated by the Wage and Hour Division of the Department of Labor. The Government Accountability Office (GAO) has identified serious flaws in how the Department of Labor calculates Davis–Bacon rates. The GAO report shows that the Labor Department does not use a scientifically representative sample and bases wage rates on very few observations—in some cases, as few as three workers.[1]

Last month, your committee held a hearing to examine this report. At that time, I testified that these errors render DBA wage estimates inaccurate and scientifically meaningless.[2] I compared Davis–Bacon wage determinations with estimates of market pay calculated by the Bureau of Labor Statistics’ Occupational Employment Statistics (OES) survey. I explained that Davis–Bacon rates typically exceed market pay and that correcting the flaws in the DBA determinations would result in considerable savings for taxpayers.

Writing for the Economic Policy Institute (EPI), Ross Eisenbrey and Dr. Dale Belman responded to this testimony with a supplemental statement submitted to your office.[3] They argued that I made “serious errors” and that my testimony “grossly misrepresents” the relationship between DBA and market wages. In particular, the EPI strongly objects to comparing DBA rates and OES construction wages.

I am submitting this letter as supplemental testimony to respond to the EPI’s accusations and correct the record. The EPI makes four principal arguments: (1) OES data are incapable of meeting the statutory requirements for DBA prevailing wages and are not an appropriate benchmark for market wages; (2) my analysis fails to include DBA residential rates; (3) my analysis compares more recent DBA rates to older OES rates; and (4) I fail to adjust for the use of apprentices on the job. Adjusting for these factors, EPI contends that DBA rates are similar to OES figures.

These arguments are either completely inaccurate or highly misleading. As I will show, Davis–Bacon rates do in fact considerably exceed market pay. Congress could achieve considerable savings by either repealing the act or requiring the Department of Labor to use scientific methods to calculate DBA rates.

Appropriateness of OES Figures for Prevailing Wage Determinations

The EPI objects to comparing DBA rates with OES wage estimates. They write that:

As currently constructed, the OES is both significantly different and quite incapable of meeting the statutory requirements that the Davis–Bacon surveys have been designed to meet. The OES is not an appropriate benchmark for comparison and should not be idealized as the true “market rate.”[4]

The Department of Labor disagrees. The Bureau of Labor Statistics conducts the OES survey primarily so that Labor can enforce prevailing wage statutes. The Service Contract Act requires federal service contractors to pay their workers existing market wages. The Foreign Labor Certification program requires employers of high-skill immigrants to pay at least market wages. The Labor Department uses OES figures to enforce these programs.

This is because the OES survey contains none of the flaws the GAO identified in DBA determinations. Unlike DBA determinations, the OES uses scientific representative sampling techniques and makes appropriate adjustments for non-response. The survey has a very large sample size, and the BLS updates it annually. The OES is the most reliable occupational wage data that the federal government produces.

As I acknowledged in my testimony, the OES cannot—by itself—meet the statutory requirements for DBA enforcement. The DBA covers both wages and benefits, but OES does not include data on employee benefits. However, OES figures meet the statutory requirements for setting DBA wages and serve as an appropriate benchmark for market wages. In a letter to Congress explaining its decision not to use OES figures, the Clinton Administration nonetheless expressly acknowledged this fact.[5]

Davis–Bacon Act Covers Little Residential Housing

The EPI also objects to the specific comparisons between Davis–Bacon and OES/market wages that I made in my testimony. Their main objection is that I compared DBA “building” rates with OES figures that include both nonresidential and residential workers. They argue that an average of DBA “building” and “residential” rates is much closer to OES wages than the “building” wages are.[6] They argue that improving the accuracy of DBA determinations would therefore result in relatively little saving for taxpayers.

This argument is highly misleading. The EPI correctly points out that DBA residential wage rates can be significantly lower than building or heavy construction rates. These residential determinations are riddled with the same inaccuracies and methodological flaws the GAO identified in the building determinations. That they differ markedly from each other is not surprising.

However, the Davis–Bacon Act covers little residential housing. Less than 8 percent of direct federal construction spending goes to residential projects.[7] An even smaller portion of state and local government construction spending—government some of which the federal subsidizes and is covered by the DBA—goes to residential projects.[8] Most federally funded construction projects build government buildings or infrastructure. Lower DBA residential rates do little to reduce federal costs.[9] The relevant comparison is the difference between market wages and DBA rates on projects the government actually pays for.

Cross Industry and Nonresidential Building Rates Similar

Contrary to the EPI’s assertions, nonresidential building rates do not differ significantly from the overall construction rates. Looking only at them would not change the finding that DBA rates substantially exceed market wages.

The OES survey collects wage rates by occupation and by detailed industry. These detailed industries include residential building, nonresidential building, heavy civil and engineering construction, and specialty trade contractors. Unfortunately, the OES does not publicly release these industry-level wages for Metropolitan Statistical Areas. The MSA-level data only report occupational wages across all industries.

I used cross-industry MSA data in my testimony to compare local Davis–Bacon and OES rates. Although national OES rates for just nonresidential building construction do differ from the cross-industry figures, they do not differ greatly. For example, nationwide, plumbers and pipefitters earn an average hourly wage of $24.10. In the nonresidential building sector, they earn $25.42 an hour.

Hourly Wages by Construction Sector

Table 1 shows national hourly wages for several construction occupations. The first line reports the overall cross-industry wage—the local-level figures I reported in my testimony. The next two lines report the wages for residential and nonresidential building construction. Nonresidential building wages are usually somewhat higher than the cross-industry construction wages. However, the difference is modest: only about 5 percent.[10]

The flawed DBA methodology significantly inflates the cost of taxpayer-funded construction projects. Using nonresidential building rates instead of the cross-industry figures does not change this finding.

Timeliness of Comparisons

The EPI further objects to “comparing Davis–Bacon wage determinations which were in effect in 2011 with OES data from May 2009.” They argue that “[w]hen Davis–Bacon rates which were in effect in May 2009 are used, the apparent differences between the Davis–Bacon rates and OES rates are substantially lower.”[11]

This argument is completely mistaken. I reported the most recent Davis–Bacon and OES rates at the time of the hearing.[12] This is the appropriate comparison. Neither survey is, of course, immediately up-to-date. No agency can conduct a survey instantaneously. However, OES survey data are processed more rapidly and updated more frequently than Davis–Bacon determinations. This—along with its scientifically valid methodology—is one of the principal reasons to prefer OES figures.

The Bureau of Labor Statistics updates OES figures each May with data from the previous year. Consequently, the OES always reports figures from the previous 12 to 24 months. DBA surveys are much less current. The Labor Department currently takes three years to process DBA survey results. The Labor Department wants to reduce that time to 17 months, but—as the GAO reported—it is not close to meeting this goal.[13]

Once it releases determinations, the Labor Department can take years, even decades, to update them.[14] For example, the Davis–Bacon building rates in Kent County, Michigan, date to 1987.[15] The residential survey for Hillsdale, Michigan, has not been updated since 1979.[16]

The EPI suggests comparing OES surveys conducted in 2009 and released in May 2010 with Davis–Bacon surveys primarily conducted before 2006 and released by May 2009. This comparison makes little sense and reveals nothing about the accuracy of either survey.

The EPI’s suggested comparison also reveals nothing about taxpayer savings from using BLS data. If Labor used OES figures, it would use the most recent data available. These are the correct figures to compare to current Davis–Bacon rates.

Helpers and Apprentices

The EPI finally complains that I do “not take into account that the surveyed Davis Bacon rate does not include apprentice rates and will therefore be higher than the average rates actually paid on a project.”[17] This charge reflects a misunderstanding of how the OES classifies low-skill construction workers.

OES occupation categories follow the Standard Occupational Classification (SOC) system. As the EPI explains, construction unions use a formal apprenticeship system, and apprentices earn lower wages as they develop their skills. Merit shop (non-union) construction firms typically use craft training programs and also pay entry-level construction workers lower wages.

The SOC classifies these less-skilled workers separately. The OES reports the wages of “helper” workers separately from the rates for fully experienced employees. Table 2 displays the difference in these wages for several occupations. Helpers typically earn about 40 percent less than experienced workers.

Wages for Helper and Regular Construction Workers

In my testimony, I compared the journey-level DBA rates with the OES rates for fully experienced employees. The EPI objects that this does not include apprentices’ lower wages. My figures do not include helpers’ lower wages either. I reported an apples-to-apples comparison of the rates paid to fully experienced employees using Davis–Bacon rates and OES data.

The EPI suggests comparing the OES rate for experienced workers to a weighted average of apprentice and journey-level DBA rates. This would deflate DBA rates to account for low-skill workers without similarly adjusting OES figures. This would present an inaccurate and misleading comparison.


The flawed Davis–Bacon determination methodology reports wage rates that differ significantly from those actually paid to construction workers. The EPI’s arguments to the contrary do not withstand scrutiny. The Occupational Employment Statistics survey uses a scientifically representative sample, is updated far more frequently than Davis–Bacon rates, and is an appropriate benchmark for market wages. Looking at non-residential building wages instead of the cross-industry wages only slightly changes the national averages. Comparing an average of apprentice and journey-level wages to the rates paid to experienced non-union construction workers would be misleading.

The committee should also note what the Economic Policy Institute did not say. Not once did Mr. Eisenbrey and Dr. Belman defend the existing Davis–Bacon methodology as scientific or accurate. They kept silent because the methodology is indefensible. No professional economist could suggest using a self-selected sample to estimate wages or drawing statistical inference from a sample of five workers. They did not dispute the main point I made to the committee: The Labor Department uses a deeply flawed methodology to determine Davis–Bacon rates.

This flawed methodology leads to Davis–Bacon rates that significantly exceed market pay. The savings from more accurate DBA wage determinations are real and substantial. Congress should seriously examine requiring the Labor Department to use scientific methods to estimate Davis–Bacon wages.

Thank you again for the opportunity to testify about this important matter.

James Sherk
Senior Policy Analyst in Labor Economics
The Heritage Foundation

About the Author

James Sherk Research Fellow, Labor Economics
Center for Data Analysis

Show references in this report

[1] U.S. Government Accountability Office, Davis–Bacon Act: Methodological Changes Needed to Improve Wage Survey, GAO-11-152, March 2011, at http://www.gao.gov/new.items/d11152.pdf.

[2] James Sherk, “Examining the Department of Labor’s Implementation of the Davis–Bacon Act,” testimony before the Committee on Education and the Workforce, United States House of Representatives, April 14, 2011, at http://www.heritage.org/research/testimony/2011/04/examining-the-department-of-labors-implementation-of-the-davis-bacon-act.

[3] Ross Eisenbrey and Dale Belman, “Supplemental Testimony: The Department of Labor’s Implementation of the Davis–Bacon Act,” Economic Policy Institute, May 4, 2011, at http://www.epi.org/analysis_and_opinion/entry/supplemental_testimony_the_department_of_labors_implementation_of_the_davis/. Ross Eisenbrey is Vice President of the Economic Policy Institute; Dr. Belman teaches at Michigan State University.

[4] Ibid., p. 2.

[5] “BLS’s Occupational Employment Statistics (OES) survey is a feasible/viable approach for wage rates in that it produces: wage rate estimates by occupation for individual Metropolitan Statistical Areas (MSAs) and for nonmetropolitan areas; and, data by SIC code which can be differentiated between building, residential, and heavy/highway general contractors.” Letter from Bernard Anderson, Assistant Secretary for Employment Standards, to Senator Arlen Specter and Senator Tom Harkin, January 17, 2001.

[6] Eisenbrey and Belman, “Supplemental Testimony: The Department of Labor’s Implementation of the Davis–Bacon Act,” p. 2.

[7] U.S. Census Bureau, “Federal Construction,” at http://www.census.gov/const/www/fedpage.html. Figures seasonally adjusted.

[8] U.S. Census Bureau, “State and Local Construction,” at http://www.census.gov/const/www/statepage.html. Figures seasonally adjusted. Just 2.7 percent of state and local construction spending in 2010 was on residential projects.

[9] Additionally, since DBA determinations are minimum rates, lower rates do not necessarily imply lower construction costs. The government would need enough market power in a locality to drive wages below market rates for an inaccurately low DBA determination to reduce costs.

[10] The results in particular localities will differ—above and below—from this national figure. By definition, this does not affect the average national Davis–Bacon premium.

[11] Eisenbrey and Belman, “Supplemental Testimony: The Department of Labor’s Implementation of the Davis–Bacon Act,” p. 2.

[12] Since then, the BLS released the May 2010 occupational wage estimates.

[13] U.S. Government Accountability Office, Davis–Bacon Act: Methodological Changes Needed to Improve Wage Survey, pp. 12–18.

[14] One component of DBA surveys is regularly kept up-to-date: union wage determinations. The Labor Department updates union-prevailing DBA rates without conducting a new survey when unions negotiate a new collective bargaining agreement. Since union wages are typically higher than non-union wages, this raises DBA rates. However, the flawed DBA survey reports that union wages prevail in almost two-thirds of sampled jurisdictions. Actual union density stands at 13 percent in the construction industry, and union wages prevail in only a handful of urban areas. This misreporting of the prevalence of union wages introduces a further inaccuracy into the flawed Davis–Bacon determinations.

[15] U.S. Government Printing Office, Davis–Bacon Wage Determinations by State, General Decision MI20100006, at http://frwebgate.access.gpo.gov/cgi-bin/getdoc.cgi?dbname=Davis-Bacon&docid=MI20100006.

[16] U.S. Government Printing Office, Davis–Bacon Wage Determinations by State, General Decision MI20100043, at http://frwebgate.access.gpo.gov/cgi-bin/getdoc.cgi?dbname=Davis-Bacon&docid=MI20100043.

[17] Eisenbrey and Belman, “Supplemental Testimony: The Department of Labor’s Implementation of the Davis–Bacon Act,” p. 3.