Testimony before the
Committee on Education and the Workforce
United States House of Representatives
April 14, 2011
Chairman Walberg, Ranking Member Woolsey, and members of the Subcommittee on Workforce Protections, thank you for inviting me to testify before you today. My name is James Sherk, and I am a senior policy analyst in labor economics at The Heritage Foundation. The views I express in this testimony are my own and should not be construed as representing any official position of The Heritage Foundation.
The GAO has identified many severe flaws in the process used to calculate Davis–Bacon prevailing wages. However, two aspects of the Department of Labor’s methodology are particularly problematic: the use of a non-representative sample and excessively small samples. These errors render Davis–Bacon wage estimates scientifically meaningless.
As a result of these flaws, Davis–Bacon wages vary wildly from market rates. In some states, such as South Dakota, Davis–Bacon rates are below market rates. In other states, such as California, Davis–Bacon rates are well above market wages. On average, the Davis–Bacon rates are 22 percent above market wages.
These errors hurt both workers and taxpayers. My estimates show that paying true prevailing wage rates—instead of inaccurate Davis–Bacon rates—would reduce government construction costs by $10.9 billion this year. Those savings could be used to either reduce the deficit or build more infrastructure at no additional cost to the public. The latter choice would mean jobs for an additional 155,000 construction workers.
Congress should insist that the Department of Labor produce scientific and accurate estimates of prevailing construction wages. The best way to do this is by transferring the resources and responsibility for conducting Davis–Bacon surveys to the Bureau of Labor Statistics. The Bureau of Labor Statistics has expertise in producing scientific wage estimates and could meet this responsibility by expanding its existing compensation surveys. The Department of Labor has no excuse for relying on unscientific and error-riddled prevailing wage estimates.
The Davis–Bacon Act
The Davis–Bacon Act (DBA) requires contractors on federally funded construction projects to pay their employees at least as much as other construction workers in the area earn—the “prevailing wage.” This prevents construction contractors from winning federal construction projects by bringing in outside workers earning below local wages.
Congress passed the Davis–Bacon Act in 1931 to prevent African–American workers from underbidding white union members on federal construction projects. During the Great Depression, many African Americans moved to the North to search for employment opportunities. In many cases, they won federal construction contracts that would have otherwise gone to white union members. The Davis–Bacon Act intentionally made it much more difficult for minorities to compete against white workers for these jobs.
Despite this origin, the Davis–Bacon Act remains on the books and applies to almost all federally funded construction projects. The Wage and Hour Division (WHD) of the Department of Labor estimates the local prevailing wages that federal contractors must pay.
Unscientific Survey Methodology
The Government Accountability Office (GAO) and the Office of Inspector General have frequently criticized the Wage and Hour Division’s survey methodology. A recent GAO report finds that serious flaws persist with Davis–Bacon surveys. Some of these problems can be solved by improving existing methods. These include processing delays and confusing surveys that lead to high error rates in returned forms.
However, the most significant problem with Davis–Bacon rates is the WHD methodology itself. The Wage and Hour Division uses unscientific methods to estimate construction wages. The GAO criticized WHD for not consulting with survey experts to design its survey, and this lack of expertise shows.
Two fundamental flaws render WHD wage estimates scientifically invalid.
First, WHD does not calculate Davis–Bacon wages using a representative sample. The importance of a representative sample is a fundamental statistical principle. A non-representative sample of wages reveals nothing about true prevailing wage rates.
Second, WHD bases the majority of its wage estimates on too few responses to be accurate. GAO reports that only one-quarter of Davis–Bacon wages are based on estimates of 29 or more workers. Fully 26 percent of Davis–Bacon estimates are based on the wages paid to six or fewer workers. Even if WHD properly randomized its surveys, these small sample sizes would make the results meaningless.
The WHD survey methodology is unscientific and incapable of accurately estimating construction wages. It will only approximate market pay by chance.
Professional statistical agencies estimate statistics by conducting surveys. The Bureau of Labor Statistics (BLS) does not have to interview every business every month to determine how many jobs the economy created. Instead, it surveys a representative sample of businesses. Statistical agencies achieve representative samples through random sampling. Using statistical principles, they can extrapolate from a randomly sampled survey to the overall economy.
Without a representative sample, surveys say nothing about the overall economy. As Nobel Prize-winning economist James Heckman has noted, “Wage or earnings functions estimated on selected samples do not, in general, estimate population wage functions.” Any introductory statistics text will make the same point.
Non-representative samples are not scientifically valid. They only provide information about those who respond to the survey. They provide no statistical information about wages or other aspects of the overall economy.
To see this, consider if Rush Limbaugh and Rachel Maddow hosted on-air polls about whether President Obama should be re-elected. Rush Limbaugh has a much more conservative audience than the country as a whole. He would probably find an overwhelming majority of respondents wanting to see Obama defeated. Rachel Maddow has a much more liberal audience than the country as a whole. Her viewers would probably say overwhelmingly that Obama deserves a second term. These straw polls might provide interesting information about the audience of the Rush Limbaugh and Rachel Maddow shows, but they would provide no useful information about President Obama’s actual re-election prospects. Concluding that President Obama was headed for a landslide defeat or landslide victory based on a non-representative survey would be unscientific and inaccurate.
Davis–Bacon Survey Is Self-Selected
A representative sample is unnecessary if the government knows the wages of every worker. Then the government could calculate average wages directly without generalizing from a sample. The Wage and Hour Division purports to have this information for construction workers. WHD sends surveys to every construction firm in a given region. WHD bases Davis–Bacon wages on the responses to this “census.” This will provide scientifically valid wage figures—if every business responds.
However, most businesses do not return Davis–Bacon wage surveys. Davis–Bacon surveys take considerable time and effort to complete, and many contractors do not expend staff resources to complete them. The surveys also ask for information in a form that many construction companies do not track. If contractors do not respond to the survey, WHD sends them a follow-up letter asking them to complete the forms. If that letter goes unanswered, they are ignored.
This methodology leads to very high non-response rates. Response rates are so low that WHD reduced its minimum data standards to wages of three workers from two companies. Too few employers responded to meet the old standard of data on six workers from at least three employers. Those employers who do respond tend to be those with large staffs. Unions also devote considerable effort to facilitate unionized employers completing and returning the surveys.
Consequently, Davis–Bacon rates are based on neither a representative sample nor a universal census of construction workers. They are based on a self-selected sample of large, unionized businesses. The GAO report confirms this. Nationwide, only 13.7 percent of construction workers are covered by union contracts. Nonetheless, 63 percent of Davis–Bacon rates are collectively bargained union wage rates. Union rates are more than four and a half times more common in the WHD survey than would occur in a representative sample. The Davis–Bacon survey is far from representative.
As a result, it is scientifically useless. Accurate estimates of prevailing construction wages cannot be made from a non-representative sample. Davis–Bacon rates will only approximate actual prevailing wages by chance.
Statistical Corrections Ignored
Professional statistical surveys do not suffer from these problems. The Bureau of Labor Statistics, for example, does not estimate job creation by conducting a census of all employers. Instead, BLS selects a smaller sample of businesses and takes several steps to make that sample representative.
First, the BLS strives to make its surveys as easy as possible to understand and complete. They test their surveys with employers before they put them in the field to ensure ease of use. The Wage and Hour Division does not do this.
Second, professional statistical agencies like the BLS follow up with employers who do not initially respond. This includes telephone calls and, in some cases, on-site visits to collect the required information.
As a result of these measures, BLS surveys have high response rates. For example, 78.4 percent of employers respond to the Occupational Employment Statistics survey. These high responses help make BLS surveys representative of the overall population.
Third, professional statistical agencies do not ignore employers that do not respond. Instead they make adjustments to correct for their absence. The two principal adjustments statistical agencies make are weighting and imputation.
Weighting involves adjusting the importance given to the respondents of the survey based on how likely they are to respond. Those groups who were more likely to respond count for less and vice versa. Pollsters do this on a regular basis. For example, a pollster might survey a state and get a sample with 60 percent men and 40 percent women. In fact, that state has equal numbers of men and women—women simply responded in lower numbers. The pollster would adjust the weight given to men’s and women’s responses so that both groups contributed equally to the final results. Statistical agencies weight responses by variables like firm size so that large businesses are not overrepresented.
Imputation involves substituting a missing response with a response from a similar respondent or respondents. For example, if a small construction firm does not return the Occupational Employment Statistics survey, the BLS does not assume that there are not any workers. Instead, the BLS would randomly select another nearby small construction firm that did respond and treat its response as the response of the missing firm. This introduces some error into the sample—but much less error than is introduced by completely ignoring non-responders.
The Wage and Hour Division does not weight Davis–Bacon survey responses or impute missing data. The Wage and Hour Division does not conduct any analysis at all of contractors who do not respond. WHD does not take basic statistical steps to obtain a representative sample. Their methodology has no scientific justification.
Inappropriately Small Samples
The Davis–Bacon methodology suffers from a second fundamental scientific flaw. Even with a proper representative sample, the Wage and Hour Division surveys too few workers to make statistically accurate estimates.
Averages in a representative sample are unlikely to exactly match the average in the overall economy. The power of statistical inference is that it allows researchers to estimate their margin of error. The sample may not exactly match the overall population, but researchers can determine how far off they are likely to be.
As sample size decreases, surveys become less accurate, and their margin of error increases. For example, a representative poll of 1,000 Americans has a margin of error of ± 3.1 percent while a poll of 100 Americans has a margin of error ± 10.0 percent.
If sample sizes become too small, however, estimating even the margin of error becomes impossible. Statistical inference is based on the central limit theorem. The central limit theorem only applies to samples of sufficiently large size, in most cases requiring a sample of at least 30 observations. Researchers cannot estimate how inaccurate the results of smaller samples are.
The Wage and Hour Division routinely uses samples of less than 30 workers. The GAO found that only 25 percent of Davis–Bacon rates are based on data from 29 or more workers. A greater proportion of wage rates (26 percent) are based on data from six or fewer workers.
Even a properly randomized representative sample of six workers would be too small from which to make statistical inferences. No professional pollster would conduct a survey of six voters.
The WHD minimum data standards are observations on three workers from two employers. That minimum standard should be data on at least 30 randomly selected workers. The Wage and Hour Division’s existing methodology lacks statistical validity.
Inaccurate Wage Determinations
The Wage and Hour Division uses unscientific methods and unrepresentative data to estimate prevailing wages. Unsurprisingly, Davis–Bacon rates typically bear little relation to actual prevailing wages. The table below shows Davis–Bacon and market wages (estimated by the Bureau of Labor Statistics) for several U.S. cities. The appendix to this testimony explains the methodology for these comparisons. Davis–Bacon rates vary wildly from actual market pay.
For most cities, Davis–Bacon rates are well above market wages. Plumbers in Jackson, Michigan, earn $28.23 an hour, but their Davis–Bacon rates are $32.79 an hour—a 16 percent premium. Carpenters in the Twin City region in Minnesota earn $23.92 an hour, but the Wage and Hour Division requires federal contractors to pay $31.77 an hour—a 33 percent premium. Electricians in Sonoma County, California, earn $28.55 an hour, but Davis–Bacon rates are 54 percent higher at $44.00 an hour.
In some cities, however, the Wage and Hour Division’s flawed methodology reports Davis–Bacon rates below prevailing market wages. Davis–Bacon rates for plumbers in Sioux Falls, South Dakota, are 17 percent below market wages. The Wage and Hour Division contends that prevailing wages for electricians in Spartanburg, South Carolina, are only $7.85 an hour—55 percent below their actual level of $17.47 an hour. Davis–Bacon rates for carpenters in Spartanburg are even worse—the federal minimum wage of $7.25 an hour.
Nationwide, the Wage and Hour Division reports Davis–Bacon wages that average 22 percent above actual market pay. These inaccurate rates inflate the cost of federal construction projects by 9.9 percent.
These inaccurate Davis–Bacon rates harm both workers and taxpayers. In most cities, Davis–Bacon rates unnecessarily raise construction costs. In essence, the government hires four construction workers for the price of five. The construction workers fortunate enough to work on a federal project no doubt appreciate this premium. However, these inaccuracies will inflate the cost of federally funded construction projects by $10.9 billion this year.
In other cities, the Davis–Bacon inaccuracies depress market pay. Davis–Bacon rates are minimum wages, so below-market determinations do not force contractors to pay substandard wages. They do, however, encourage contractors to reduce their bids—putting downward pressure on wages.
If the Department of Labor used accurate wage determinations, Congress could build the same amount of infrastructure at substantially lower cost. The savings from paying market wages would reduce the deficit.
Alternatively, accurate wage determinations would allow Congress to build more infrastructure at no extra cost to taxpayers. This would enable the government to provide more public services and employ an additional 155,000 construction workers in 2011. This is not a minor consideration when unemployment in the construction industry is above 20 percent.
If Congress is going to keep the Davis–Bacon Act on the books, it should require the Department of Labor to estimate prevailing wages scientifically. Taxpayers receive no value from overpaying some workers and underpaying others.
Bureau of Labor Statistics
The Wage and Hour Division estimates prevailing wages so poorly because it is not a professional statistical agency. The Wage and Hour Division is an enforcement agency. WHD enforces federal laws regulating wages and many working conditions, such as minimum wages, prevailing wages, child labor, overtime, and the Family and Medical Leave Act. WHD has no expertise in conducting scientific wage surveys.
The Bureau of Labor Statistics does. The BLS has extensive experience in conducting scientific wage surveys. Bureau of Labor Statistics methodology, accuracy, and data quality are internationally respected. They have the expertise in scientifically estimating wages that the Wage and Hour Division lacks.
The Bureau of Labor Statistics already conducts two nationwide wage surveys that scientifically estimate occupational wages: the National Compensation Survey (NCS) and the Occupational Employment Statistics (OES). Unlike the WHD survey, these surveys have high response rates, and BLS corrects for non-response with weighting and imputation. Both surveys have large sample sizes, are conducted in a timely manner, and are updated annually. The Department of Labor uses OES data to enforce prevailing wages for the Foreign Labor Certification program and the Service Contract Act. If Congress wants accurate Davis–Bacon rates, it should require the Department of Labor to use BLS data.
Better Geographic Coverage
The Department of Labor previously rejected the idea of using BLS data. One of the reasons they gave for doing so was concerns about BLS’s geographic coverage. While the Wage and Hour Division issues Davis–Bacon rates for individual counties, the Bureau of Labor Statistics reports wages for Metropolitan Statistical Areas (MSAs). Some large counties are their own MSA, but most MSAs are agglomerations of multiple economically linked counties.
The Davis–Bacon Act states:
The minimum wages shall be based on the wages the Secretary of Labor determines to be prevailing for the corresponding classes of laborers and mechanics employed on projects of a character similar to the contract work in the civil subdivision of the State in which the work is to be performed, or in the District of Columbia if the work is to be performed there.
The GAO argues that this provision prevents the Department of Labor from estimating prevailing wages at the MSA level. The Wage and Hour Division disagrees with this legal analysis. In response to a 2004 Inspector General report, the Wage and Hour Division stated that “the Davis–Bacon Act does not prohibit issuing wage determinations for broader geographic areas such as an MSA, and we routinely issue such wage determinations when sufficient data are not available on a county basis.”
The GAO report reveals just how routine those broader geographic determinations are. Only 11 percent of Davis–Bacon rates are based on data from a single county. Forty-two percent of Davis–Bacon rates are based on groupings of counties analogous to an MSA, while 40 percent of job classifications are based on statewide data.
Switching to BLS data at the MSA level would eliminate wage determinations based on statewide data. This would much more closely approximate prevailing local wages than the WHD currently does.
Congress should transfer responsibility for collecting Davis–Bacon prevailing wage data to the Bureau of Labor Statistics. The OES already provides annual wage data for most construction jobs across the country. WHD could currently use OES wage data to set Davis–Bacon wage rates. The chief obstacle to using OES data is calculating hourly fringe benefit rates as required by the Davis–Bacon Act—the OES does not cover employee benefits.
The National Compensation Survey covers benefits, and the WHD determined that the NCS provides the information necessary to enforce the Davis–Bacon Act in the areas that it surveys. However, the NCS provides local wage information for only 154 metropolitan and non-metropolitan statistical areas. These MSAs cover just half of the U.S. population. Consequently, neither the OES nor the NCS directly provides all of the information necessary to enforce the Davis–Bacon Act.
These problems are solvable. To calculate prevailing construction benefits, the Bureau of Labor Statistics could:
Expand the National Compensation Survey. The BLS could expand the construction portion of the NCS to provide nationwide coverage of construction workers. The Inspector General suggested this approach in 2004.
Collect Construction Benefits with the OES. The BLS could collect benefits data from construction employers through the OES. This would require overhauling the OES survey and would take some time to set up and train staff to conduct properly.
Econometrically Model Benefits. A third approach involves using NCS data to econometrically model the relationship between wages and benefits in the construction industry. That model could be applied to the existing OES data to estimate fringe benefits for different construction occupations.
These solutions are not trivial undertakings. They would require Congress to transfer the resources for conducting Davis–Bacon surveys from WHD to the BLS. However, if Congress did so, the BLS could do what the WHD does not: scientifically and accurately estimate prevailing construction wages.
The Department of Labor’s methods for calculating prevailing construction wages are scientifically unsound. The Government Accountability Office report demonstrates that the Wage and Hour Division calculates Davis–Bacon rates with a self-selected sample instead of a representative sample. Non-representative samples do not provide reliable information. WHD does not use basic statistical techniques, such as measuring non-response and weighting their data to mitigate this bias. Even if WHD did use a representative sample, they have too few responses to be accurate.
Unsurprisingly, Davis–Bacon rates bear little correlation to market wages. In some cities they are below market rates, while in others they are well above market rates. On average, Davis–Bacon rates are inflated 22 percent above market pay. These inaccuracies hurt both workers and taxpayers.
Congress already spends $600 million a year on another agency with professional expertise in calculating labor market statistics: the Bureau of Labor Statistics. BLS surveys do not suffer from the methodological shortfalls that plague WHD prevailing wage estimates. The BLS is internationally respected for conducting scientific and accurate surveys. If Congress wants accurate Davis–Bacon surveys, it should direct the Bureau of Labor Statistics to conduct them.
Bureau of Labor Statistics (BLS) and Wage and Hour Division (WHD) wage estimates are not directly comparable. To report comparable wage rates, The Heritage Foundation was guided by the methodology outlined by the Beacon Hill Institute in their comprehensive report comparing market and Davis–Bacon wages.
Market wage data come from the Occupational Employment Statistics program within the BLS. These data can be found online at http://www.bls.gov/oes/. Data on Davis–Bacon wages came from the U.S. Government Printing Office, “Davis–Bacon Wage Determinations,” at http://www.gpo.gov/davisbacon.
Three job categories were selected for comparison: carpenters, electricians, and plumbers/pipefitters. The Davis–Bacon rate for each category was determined as follows. The Davis–Bacon rates for “Building” construction were identified from the online postings. Davis–Bacon rates often specify wages for general and specific tasks within an occupation. There may be wages for general “electricians,” but also separate rates for electricians who perform specialized tasks. In these cases, the wages of the most general category were selected.
The BLS and WHD estimate wages for different geographic areas. The WHD issues wage rates at the county level, while the OES estimates wages for metropolitan statistical areas. The Heritage Foundation used county-level Davis–Bacon wages to create MSA-level Davis–Bacon wage rates. In MSAs with only one county, Davis–Bacon rates were calculated as explained above and directly compared to BLS data. In MSAs with multiple counties, Davis–Bacon rates were calculated separately for each county. A weighted average of Davis–Bacon rates was constructed, using as weights the relative population of each county according to Census Bureau estimates from the year 2009, which can be found online at http://quickfacts.census.gov/qfd/index.html. This weighted average was the final Davis–Bacon rate compared to BLS data.
In a few cases, the Davis Bacon rate is not the same for the entire county—for example, a certain occupation’s wage rate may vary for different geographic regions within a single county. In these cases, The Heritage Foundation used the rate from the most populous part of the county.
MSAs Examined and Their Constituent Counties:
MSA: Jackson, MI MSA
Counties: Jackson County
MSA: Minneapolis–St. Paul–Bloomington, MN–WI MSA
Counties: Anoka County, MN; Carver County, MN; Chisago County, MN; Dakota County, MN; Hennepin County, MN; Isanti County, MN; Ramsey County, MN; Scott County, MN; Sherburne County, MN; Washington County, MN; Wright County, MN; Pierce County, WI; St. Croix County, WI
MSA: Sioux Falls, SD MSA
Counties: Lincoln County, McCook County, Minnehaha County, Turner County
MSA: Erie, PA MSA
County: Erie County
MSA: Santa Rosa–Petaluma, CA MSA
County: Sonoma County
MSA: Lafayette, IN MSA
Counties: Benton County, Carroll County, Tippecanoe County
MSA: Terre Haute, IN MSA
Counties: Clay County, Sullivan County, Vermillion County, Vigo County
MSA: Spartanburg, SC MSA
County: Spartanburg County
MSA: Lakeland–Winter Haven, FL MSA
County: Polk County
MSA: Oakland–Fremont–Hayward, CA MSA
Counties: Alameda County, Contra Costa County
MSA: Newark–Union, NJ–PA MSA
Counties: Essex County, NJ; Hunterdon County, NJ; Morris County, NJ; Sussex County, NJ; Union County, NJ; Pike County, PA
MSA: Cleveland–Elyria–Mentor, OH MSA
Counties: Cuyahoga County, Geauga County, Lake County, Lorain County, Medina County
MSA: Nassau–Suffolk, NY MSA
Counties: Nassau County, Suffolk County
MSA: Honolulu, HI MSA
County: Honolulu County