The Prevailing Theory

A closer look at whether the state’s prevailing wage law actually delivers on its stated legislative purpose.

Among all the proverbial “third rails” that exist in politics, state prevailing wage laws feature some of highest voltage.  These controversial state-level versions of the Depression-era federal Davis-Bacon Act mandate pay levels for public construction projects that meet certain thresholds to ensure that government projects offer compensation that corresponds to local market rates for various occupations.

This on-again, off-again topic is back in the news.  In a preemptive effort to inoculate Minnesota policy makers against what the authors themselves describe as the “legislative virus” of prevailing wage repeal infecting the Midwest, researchers from the Midwest Economic Policy Institute and Colorado State University-Pueblo released a report that studied 600 Minnesota school district construction bids and found prevailing wage had no statistically significant effect on school construction costs.[1]

It might seem very strange that so much political drama can arise out of a law requiring something so seemingly normal and unremarkable as paying the levels of compensation that already exist in a local economy.  The hint that something else may be going on here can be found in the language of Minnesota’s own prevailing wage statute:

It is in the public interest that public buildings and other public works be constructed and maintained by the best means and highest quality of labor reasonably available and that persons working on public works be compensated according to the real value of the services they perform.  It is therefore the policy of this state that wages of laborers, workers, and mechanics on projects financed in whole or part by state funds should be comparable to wages paid for similar work in the community as a whole.”

Minnesota Statutes, Chapter 177, Section 41

Few would argue with the statute’s statement of public interest or its intent.  But determining and quantifying the “real value” of someone’s labor is an inherently challenging and arguably subjective task.  That’s especially true when it involves the use of imprecise definitions and vague concepts such as “comparable wages,” “similar work,” and “in the community as a whole.”  While most prevailing wage debates center on whether these laws should exist, an overlooked and underappreciated policy issue is the process states that have implemented these laws use to actually determine their prevailing wage rates.

The OLA Chimes In…

In Minnesota, the Department of Labor and Industry (DLI) administers the prevailing wage rate-setting process.  DLI uses a survey to collect information from contractors and other interested parties on the wages paid to employees working on non-residential construction projects, along with cost of their fringe benefits.  The survey data is then used to set prevailing wage rates for a variety of “job groups” (electrician, carpenter, etc.) in each of Minnesota’s 87 counties.  The survey design includes three features we believe are important to understand.

First, the survey is project-based and not employer- or firm-based.  The survey does not gather information comprehensively on a county-by-county basis about the wages and benefits employers located in the county pay to employees who work on nonresidential construction projects.  Instead, it only gathers information on the wages and benefits paid to employees working at individual project sites in each county.  Second, any individual construction site can have multiple projects, since for prevailing wage purposes the term “project” is defined by specific contracts, and not the finished construction.  The state has designed the survey process so that multiple contractors working at a site may each submit a survey reflecting their own work.  Moreover, an individual contractor with multiple contracts for different stages of work at a construction site may report employee wage and benefit rates associated with each individual contract as a separate “project.”  Finally, the survey is voluntary.  It does not use random sampling with follow-up of non-respondents, as is generally the case for government survey projects.  Participation in the state’s prevailing wage survey is completely voluntary.  (And since the survey is based on projects which would need to be identified in real time, without a statewide database of public construction projects DLI would be hard-pressed to contact non-respondents even if state law mandated that it try to.)

In 2006 legislators directed the Office of the Legislative Auditor (OLA) to examine Minnesota’s prevailing wage requirement, and as part of that examination required an evaluation of the method used to set prevailing wage rates in Minnesota.  The OLA’s evaluation report, published in 2007, concluded that DLI, “uses reasonable methods to collect wage and benefit information for the purpose of setting prevailing wage rates.”  However, this conclusion did not come without some important qualifications and caveats:

  • The OLA concluded it was not possible to determine if DLI survey results were actually representative of the non-residential construction industry
  • The OLA found that the survey response rate appeared “to be low” based on the roughly 14,000 potential respondents it estimated could have responded to the survey
  • The OLA concluded the method used to calculate prevailing wage rates may sometimes result in rates that are not representative of wages and benefits paid for non-residential construction work

The OLA’s conclusion about the continuing efficacy of the rate-setting method employed was essentially compelled by the lack of any better approach.  Or as the OLA stated, “Although we have some concerns about the wage data collected by the survey, there is not an alternative source of construction wage data that is clearly better than the data collected by the Department of Labor and Industry.”

That’s a highly pragmatic and practical conclusion.  But if the issues and concerns flagged by the OLA persist or have worsened, at some point that conclusion about how “reasonable” the rate setting methods are can be called into question.

…And So Does the MCFE

Last year, the Construction Education Foundation of Minnesota contracted with the MCFE to revisit the analysis and findings of the OLA’s evaluation to see what, if anything, had changed.[2]  We examined prevailing wage survey data from projects completed in 2015-2016 which constituted over 50,000 individual records related to non-residential construction projects.  We explored the degree to which rate setting relies on importing rates from other areas.  And we used survey data to estimate how prevailing wage rates would change under different prevailing wage calculation scenarios.  Our findings and conclusions centered on how well the current rate setting process actually delivers on the stated legislative intent of the state’s prevailing wage law and the extent to which issues the OLA flagged 10 years ago continue to exist.

If the stated policy intent continues to be “ensuring comparable wages for similar work in the community,” our findings suggested that ten-plus years later, the process DLI uses to determine prevailing wages leaves a lot to be desired.  Our study period included 5,133 total job groups for which prevailing wages could be set.[3]  Of these, only 1,419 prevailing wages (28% of the total) were set using survey data generated from projects within the relevant county.  The majority of prevailing wages were set using data imported from adjacent counties (2,607, or 51% of the total) or using other methods – primarily historical information (314, or 6% of the total).[4]

Moreover, using wage data from adjacent counties to set prevailing wage rates frequently resulted in wage importation from locations featuring different economic and demographic characteristics.  The accompanying table shows how prevailing wage rates were exported between counties based on whether they are in a Metropolitan Statistical Area, a Micropolitan Statistical Area, or outside of either of the two (“rural”).[5]  Notably, 872 (61%) of the prevailing wage rates imported into rural counties used data that came either primarily or exclusively from non-rural counties.

What the rate determination method does appear to do quite well is establish collectively bargained union rates as prevailing wage rates.  The OLA recognized that potential outcome in its 2007 report, which stated that then-existing structural features of the wage determination process “could lead to the over-representation of union wages among the reported wages.”  That report further identified over two-thirds of commercial construction prevailing wage rates set using current own-county or adjacent-county data as union rates.  Roughly ten years later, union rates appear to be even more prevalent.  We found that 75% of the nearly 4,000 commercial construction prevailing wage rates set using own-county or adjacent county data from this survey[6] are union rates.

Two structural features of the rate determination process play an influential role in this outcome.  First, since the survey is based on projects and not employers, an individual who works on multiple projects can (and does) have his or her compensation data submitted multiple times for calculation purposes.  We found 233 instances of one employee’s compensation data being submitted on at least 10 different projects for the same job group in the same county for the same employer.  Second, the state sets the prevailing wage using the mathematical mode – the most frequently reported rate, down to the penny – of all rates reported for each job class in each county.  Put them together, it shouldn’t be surprising a very small number of individuals often determine the prevailing wage rate even in populous counties.

Our analysis of prevailing wage determinations under alternative calculation methods offers evidence that union rates would continue to have a disproportionate influence in setting the rates.  Alternative calculation methods (mean, median, etc.) should capture a broader diversity of wages and benefits that exist among all commercial construction contractors in the state and therefore yield different results.  Yet we found prevailing wage rates calculated under these alternative methods resulted very little if any change to established rates in the majority of circumstances.

Two realities underlie this finding.  First, 18% of the 1,419 prevailing wage rates set using current, own-county survey data had one data point (i.e., one person working on one project).  In another 29% of cases, multiple records were used to generate a prevailing wage rate but all of those records had identical compensation rates to the penny.  When all the data points are the same, alternative calculation methods (average, median, mode, etc.) will all generate identical outcomes.

Keeping the Intent But Ending the Pretense

Minnesota’s prevailing wage laws are used to set compensation costs for billions of dollars of public construction projects across the state annually.  If we believe it’s important to keep these provisions in place, we should have confidence that they set prevailing wage rates in a rigorous way that ensures statutory intent and reliably reflects labor compensation rates for similar work in the local economy.  Given the self-selection bias and other design flaws inherent in the current methodology – which resembles an on-line newspaper poll far more closely than any rigorous survey method we can think of – confidence in the system is almost certainly misplaced.

To fix the process we recommended a shift to employer-based surveying combined with switch to a “majority-average” method to calculate the prevailing wage (in which the mode would be used if it represents a majority of responses and the average would be used otherwise).  Hybrid approaches like this are used by a plurality of states where prevailing wage laws still exist to ensure that competing notions – the most frequently reported wage versus the average wage – both factor into the calculations creating some linkage between reporting frequency and central tendency.  Such reforms would better capture and reflect the law’s stated intent by making certain rates are based on wage and benefit structures found among local employers.

Even though such a proposal would leave prevailing wage provisions in law, we suspect it would still face considerable political obstacles.  That’s because it seems to us the primary value of the current prevailing wage determination process is the appearance it gives.  It offers the look of rigorous, evidence-based policy implementation that generates an outcome many would find highly unfair and discriminatory if instead the state simply mandated it.


[1] Manzo and Duncan, An Examination of Minnesota’s Prevailing Wage Law: Effects on Costs, Training, and Economic Development, July, 2018

[2] Minnesota's Prevailing Wage: An Evaluation of the Rate-Setting Process, February, 2018.  By way of full disclosure, the Foundation is the training and education arm of Associated Builders and Contractors, Inc.  MCFE is solely responsible for the report’s content.  The Midwest Economic Policy Institute / Colorado State University report contains no information on origins of the study or acknowledgements regarding financial or other support for the effort.  It is left to the reader to speculate why non-Minnesota scholars would study the effects of Minnesota’s prevailing wage laws on school construction costs.

[3] Prevailing wage rates could be set for 59 individual job classes in each of Minnesota’s 87 counties.  The study did not examine job groups that are specific to highway and heavy construction; mostly because of the small number of survey records (292 in total, from 27 counties).

[4] The remaining 15% of potential prevailing wage rates were not set at all.  If adjacent county information is not sufficient and no rate was certified in the prior year, DLI does not certify a rate unless requested to do so.

[5] Metropolitan and micropolitan statistical areas are geographic regions located around an urban center with a high degree of social and economic integration as measured by commuting ties.

[6] 1,419 rates generated from own-county survey data plus 2,607 rates generated from survey data imported from other counties.