Bea

con

Hill

Inst

itut

eProject Labor Agreements and

the Cost of School Constructionin Massachusetts

Paul Bachman, MSIEDarlene C. Chisholm, PhD

Jonathan Haughton, PhD David G. Tuerck, PhD

BHI Policy StudyS e p t e m b e r 2 0 0 3

Paul Bachman is a Research Assistant at BHI. He holds a MSIE from Suffolk University. DarleneC. Chisholm is an Associate Professor in Economics and a Senior Economist at BHI. She holds aDoctorate from the University of Washington. Jonathan Haughton, PhD, is an Associate Professorin the Economics Department at Suffolk University and a Senior Economist at the Beacon HillInstitute for Public Policy Research at Suffolk University. He holds a Doctorate from HavardUniversity. David G. Tuerck, PhD, is chairman of the Economics Department at SuffolkUniversity and Executive Director of the Beacon Hill Institute. He holds a doctorate fromthe University of Virginia. © 2003, Beacon Hill Institute at Suffolk University.

EXECUTIVE SUMMARY

Project Labor Agreements (PLAs) are agree-ments between construction clients (such astowns) and labor unions, which establish therules to be followed by firms that bid on con-struction projects. PLAs typically requirethat all workers be hired though union halls,that non-union workers pay dues for thelength of the project, and that union ruleson pensions, work conditions and disputeresolution be followed.

It is widely believed that Project LaborAgreements add to the cost of constructionprojects. In spring 2003, the Beacon Hill In-stitute broke new ground with a study thatwas the first to test this proposition with sta-tistical, rather than anecdotal, evidence.1 Inan analysis of 54 school construction projectsundertaken in the Greater Boston area since1995, the spring 2003 study found bid coststo be significantly higher when a school con-struction project was executed under a PLA.

This report updates and expands upon ourearlier study. We have been able to increasethe number of observations from 54 to 126by collecting data on more recent projectsand by actively pursuing missing informa-tion. This more complete data set has al-lowed us to test for the robustness of ourresults and to quantify the effect of PLAs onactual project costs (in addition to the effecton bid costs, which we used in our earlierstudy). Our main findings are as follows:

(i) PLA projects add an estimated$18.83 per square foot to the bidcost of construction (in 2001prices). We obtain this figure af-ter adjusting the data for inflation(using an index that includes thetrend in both construction wagesand in materials costs) and aftercontrolling both for the size ofprojects and for whether they in-volve new construction or reno-vations. Since the average costper square foot of construction is

2

$137.24, PLAs raise the cost ofbuilding schools by almost 14%.The data support this result at a99% confidence level.

(ii) Taken together, PLA projects ac-counted for 3.175 million squarefeet of construction with a com-bined bid cost of $481.4 million(in 2001 prices), based on theprojects that we were able to in-clude in our study. Our esti-mates show that this cost was$60 million higher than it wouldhave been if Project LaborAgreements had not been used.Our estimates show that the po-tential savings from not enteringa PLA on a school constructionproject range from $1.88 millionfor a 100,000-square-foot struc-ture to $5.6 million for a 300,000-square-foot structure.

(iii) PLA projects have higher actualcosts of construction, adding anestimated $16.51 per squarefoot, or 12%, to these costs (in2001 prices).2

(iv) The finding that PLA projectshave higher construction bidcosts is statistically highly sig-nificant and robust, in that:a. The effect persists even

when the data are subdi-vided, so that the effect isevident separately for largeand small projects, for el-ementary and secondaryschools, and for new con-struction and renovations;

b. The strong statistical resultspersist under a variety of es-timation techniques.

In short, the evidence that Project LaborAgreements have increased the cost ofschool construction in the Boston area since1995 is strong and the effect is substantial.On average, almost 70% of this additionalcost was borne by the Commonwealth ofMassachusetts, through the School BuildingAssistance Program; the remainder was

covered by the cities and towns that builtschools using Project Labor Agreements.Officials at both the state and local levelneed to be aware of the higher costsattributable to PLAs when they choose thebid requirements on future schoolconstruction projects.

I. INTRODUCTION

Project Labor Agreements (PLAs) discour-age non-union contractors from bidding onstate construction projects by requiringthem to conform to union rules and hirethrough union halls. It is widely believedthat construction projects are more expen-sive when a PLA is in effect. Until our re-cent study, however, the evidence for thishad been largely anecdotal.3 No one has, toour knowledge, attempted carefully to esti-mate the degree to which PLAs might in-crease project costs.

The current study updates and expandsupon our earlier study, by using a morecomplete data set. Once again it finds clearevidence on the differences in cost persquare foot between PLA and non-PLAprojects. This measure is based on an ex-amination of the cost of school constructionprojects in the greater Boston area since1995.

II. HISTORICAL BACKGROUND TOPROJECT LABOR AGREEMENTS

Project Labor Agreements are a form of“pre-hire” collective bargaining agreementbetween the construction clients (such astowns or school districts) and labor unionspertaining to a specific project, contract orwork location, and are unique to the con-struction industry. The terms of PLAs gen-erally recognize the participating unions asthe sole bargaining representatives for theworkers covered by the agreements, regard-less of current union membership status ofthese workers. A PLA requires all workersto be hired through the union hall referralsystem. Non-union workers must join thesignatory union of their respective craft and

3

pay dues for the length of the project. Theworkers’ wages, pension contributions andworking hours, along with the dispute reso-lution process and other work rules, are alsoprescribed in the agreement. PLAs super-sede all other collective bargaining agree-ments and prohibit strikes, slowdowns andlockouts for the duration of the project.4

Project Labor Agreements in the UnitedStates originated in the public works projectsof the Great Depression, which included theGrand Coulee Dam in Washington State in1938 and the Shasta Dam in California in1940. PLAs have continued to be used forlarge construction projects since World WarII, including the construction of CapeCanaveral in Florida, the current CentralArtery project (the “Big Dig”) in Boston, andeven private projects, such as the Alaskanpipeline and Disney World in Florida.

PLAs in the BalanceAs PLAs have become more common in pub-licly financed construction projects, and asthe number of non-union construction firmshas grown, PLAs have become controversial.Opponents of PLAs argue

(i) that PLA agreements raise thecost of undertaking projects, and

(ii) that non-union or open shop con-tractors are discouraged frombidding on jobs that have PLAs.

Opponents cite the PLA requirements that allemployees must be hired in union halls, payunion dues, contribute to union-sponsoredretirement plans, and follow union workrules. They argue that the use of a union hir-ing hall can force the contractors to hire unionworkers over their own work force. The con-tractors and their employees are required topay union wages, dues and contributionsinto union benefit plans even if they are cov-ered by their own plans. The work rules re-strict the contractors from using their ownmore flexible operating rules and procedures.These restrictive conditions cause costs to risefor a project that requires a PLA. It is worthnoting that whether or not a PLA is in effect,all contractors must adhere to any “prevail-ing wage” rules that may be in effect.

Furthermore, open-shop contractors con-tend that their competitive advantages arenullified by the PLA. The result is that inpractice, if not in principle, they are un-able to bid competitively on jobs that havea PLA requirement. In turn, the absenceof open-shop bidders for PLA projects re-sults in fewer bidders for the project, andwith fewer bidders, the lowest bids comein higher than if open-shop contractorshad participated. Therefore, the cost of theproject will be higher, with fewer biddersattempting to under-bid each other for thecontract. Some opponents also argue thatrequiring a PLA violates state competitivebidding laws that require a free and openbidding process. A number of critics evensee PLAs as a form of extortion, with animplicit threat that if a town does not agreeto a PLA, then there is more likely to bedisruption at the workplace.

Proponents of PLAs claim that the agree-ments provide for work conditions thatare harmonious, and that they guaranteewage costs for the life of the contract. Theycontend that the provisions that prohibitstrikes, slowdowns and lockouts keep theproject on time and prevent cost overrunsdue to delays. They argue, furthermore,that the wage stipulations allow firms ac-curately to estimate labor costs for the lifeof the project and thus have more accu-rate bids; and that the union rules allowfor a safer work environment, thereby re-ducing accidents and thus lowering thenumber of workman’s compensationclaims. In this view, workers’ union certi-fications ensure the quality of the workand save money by avoiding costly mis-takes.

The controversy over the use of PLAs in pub-lic construction projects has become more in-tense since the late 1980s. Open-shop (non-union) construction firms and industry orga-nizations have challenged PLAs in the courts.As discussed below, the executive and legisla-tive branches at the federal, state and local lev-els of government have at times taken posi-tions in favor of the use of PLAs.

4

PLAs at the Federal LevelThe executive branch of the federal gov-ernment has been involved in the PLA de-bate for over a decade. The administra-tion of George H. W. Bush issued an Ex-ecutive Order in 1992 forbidding the useof PLAs on federally funded projects.5

The Clinton Administration rescindedthat order in February 1993 and attemptedto go further in 1997, when it planned toissue an executive order requiring all fed-eral agencies to use PLAs on their con-struction projects. However, due to ex-tensive lobbying, the President insteadissued a memorandum encouraging theuse of PLAs on contracts over $5 millionfor construction projects, including reno-vation and repair work, for federallyowned facilities.6 President George W.Bush canceled the Clinton order on Feb-ruary 17, 2001 by issuing an ExecutiveOrder prohibiting PLAs on federallyfunded and assisted constructionprojects.7

PLAs in MassachusettsIn Massachusetts, PLAs appeared on thelegislative agendas of local and state gov-ernmental bodies as efforts were made torequire them on local constructionprojects. The City of Cambridge enacteda local ordinance that put in place manyof the same requirements that are foundin PLAs, for all public projects. The Mas-sachusetts legislature attempted to re-quire PLAs on a bond authorization forthe rebuilding and repair of courthousesthroughout the state. Under intense ne-gotiation between the legislature andGovernor Cellucci’s Office, a bill was pro-duced in 1998 that mandated PLAs forfunds allocated to courthouse construc-tion projects in Boston, Worcester, and FallRiver only. The legislation created a com-mission to recommend establishing cir-cumstances in which PLAs should beused. The legislation instructed the com-mission to consider the “appropriatenessand function and the size, complexity andduration of the public constructionprojects” when deciding whether or notto use PLAs.8

The litigation came to a head in a 1993 Su-preme Court case involving the cleanup ofthe Boston Harbor. In 1988, a federal courtdirected the Massachusetts Water ResourcesAuthority to clean up the pollution in Bos-ton Harbor. The Authority’s project man-agement firm, IFC Kaiser, negotiated a PLAwith the local construction unions for theproject. The precedent-setting aspect of thisPLA was that its use was mandated in theproject’s bid specifications.9 A non-uniontrade group filed a lawsuit contending thatrequiring the PLA as a part of the bid speci-fication violated the National Labor Rela-tions Act. The case was appealed to theUnited States Supreme Court, which, in1993, upheld the use of the PLA for theproject. The Supreme Court ruling openedthe door for the use of PLAs in other publicMassachusetts projects, including localschool construction.

Yet Project labor agreements have remainedcontroversial. The city of Lynn, Massachu-setts agreed to PLAs for a series of newschool construction projects in 1997. Ac-cording to the Lynn Building Department,the projects were bid and construction be-gan that year. However, several non-unionconstruction firms challenged the PLA incourt on the grounds that it violatedMassachusetts’s competitive bidding laws.The Court ruled that the plaintiffs suffered“irreparable harm” because “they would berequired to conform to a variety of unionpractices and would be limited in their au-tonomy to negotiate employment with non-union workers.”10 The Court allowed thatthe city had the authority to enter into a PLAbut that it “may not exercise its authorityarbitrarily or capriciously” and added, “aPLA must be evaluated in the light of aproject’s size, complexity, and duration.”11

The Court then found that the Lynn schoolsfailed to meet these criteria, and granted apreliminary injunction preventing the cityfrom requiring bidders to sign a PLA in or-der to work on the project.12 The City ofLynn subsequently opened the bidding forthe projects without requiring firms to signa PLA.

5

The outcome was different in the case of thecity of Malden, which in 1996 began a five-year $100-million series of projects to replaceits schools serving kindergarten througheighth-grade, and to remodel Malden HighSchool. The projects were to be accom-plished by closing nine existing schools, re-placing five schools, and demolishing three.

On the recommendation of its constructionproject management firm, O’Brien-Kreitzberg, Inc., the city negotiated a PLAwith the Building and Construction TradesCouncil of the Metropolitan District, AFL-CIO and the New Council of Carpenters,AFL-CIO. The agreement included many ofthe PLA provisions discussed in Section II,including: the recognition of unions as thesole and exclusive bargaining representa-tives of all project employees; hiring throughthe union referral process; the requirementof contractors to contribute to union em-ployee benefit plans; uniform work rules anddispute resolution; and prohibiting strikes,picketing, work stoppages, slowdowns, andlockouts.13 The PLA was approved by a voteof the City of Malden municipal buildingcommittee in May of 1997; union approvalfollowed.

In the initial phase of the project, the city bidthe construction of the Beebe and Rooseveltschools as one project, with the stipulationthat the project was subject to the PLA re-quirement. When the bids were reviewedby the city, the lowest exceeded the projectbudget and all bids were subsequently re-jected. The project was modified and the cityoffered each school for bid separately. OnNovember 7, 1997, seven open-shop (non-union) contractors with public sector build-ing experience filed for a motion of prelimi-nary injunction against the use of a PLA inthe bidding process. The plaintiffs arguedthat the PLA violated the state’s competitivebidding laws, and that they would have bidfor both projects if the PLA were not in-cluded. The court denied the request for apreliminary injunction, and when the plain-tiffs filed an appeal, the Massachusetts Su-preme Judicial Court chose to hear the case.

The Supreme Judicial Court reaffirmed thelower court’s denial of the preliminary in-junction. The court majority argued thatthe objectives of the state’s competitive bid-ding laws were to “obtain the lowest pricefor its work that the competition among re-sponsible contractors [could] secure” andto create an “honest and open procedurefor competition for public contracts.”14 TheCourt accepted the plaintiffs’ assertion that“they were inhibited from bidding, andthat this inhibition could have anti-com-petitive effects.”15 However, the Court con-cluded, “that PLAs on public projects arenot absolutely prohibited.”16 In echoing thedecision of the Lynn case, and that of a NewYork case involving the restoration of theTappan Zee Bridge, the Court stated that“the project is of such size, duration, tim-ing, and complexity that the goals of thecompetitive bidding statute can not other-wise be achieved and the record demon-strates that the awarding authority under-took a careful, reasoned process to con-clude that the adoption of a PLA furtheredthe statutory goals.”17 The Court went onto state, “it may be that in certain cases,sheer size of a project warrants the adop-tion of a PLA. In most circumstances, thebuilding of a school will not, in and of it-self, justify the use of a PLA.” This firstphase of the construction project came inon budget and on time, with no labor in-terruptions, according to city officials.18

School Construction Financing inMassachusettsThe School Building Assistance Programin Massachusetts has aided public schoolconstruction for more than half a century.The program began in 1948 as a three-yeareffort to provide resources to local commu-nities for the building of schools for the“Baby Boom” generation, with a 25% per-cent reimbursement rate for the localschool districts.19

The program has since grown substantially,and has widespread political support.20

After several extensions, today “the school

6

building assistance program is the largestcapital grant program operated by theCommonwealth…and the costs of theschool building assistance program are in-creasing at an unsustainable rate.”21 In1999, the program offered, on average, a69% reimbursement rate for the construc-tion and financing costs of school projects.Over the period 1991-1999 the Common-wealth of Massachusetts made total con-tributions to the program of more than $1.7billion.22

The financial commitment for the state roseconsistently over the 1990s. In fiscal year(FY) 1999, the annual payment for schoolconstruction projects was $201 million, a58% increase from the $127 million appro-priated in 1991.23 By FY 2003 school con-struction appropriations had jumped to$362 million, a remarkable 80% increaseover the FY 1999 level.24 According to theSchool Building Assistance Programwebsite, for FY 2003, 283 constructionprojects appeared on the Priority List, with19 new projects receiving authorization.The rapid growth of the program hasprompted increased attention to the issue.A report entitled Reconstructing the SchoolBuilding Assistance Program Policy Report,published in 2000, predicted that by FY2002 “this program will achieve ‘budgetbuster’ status.”25 It is within this fiscal en-vironment that school construction costshave become an important concern in thebuilding of public schools in Massachu-setts.

III. THE EVIDENCE ON PLAs

Although there is substantial anecdotalevidence that PLAs raise constructioncosts, until recently there has been littleformal statistical evidence of such an ef-fect. To compare PLA with non-PLA costsit would be necessary to compare construc-tion projects of a similar nature – for in-stance road repairs – where some projectsare done with a PLA in place, and othersare not. Situations such as these are rare,and even when they occur, the relevant in-formation is difficult to obtain.

We have, however, found one suitable“natural experiment” that allows us for-mally to compare the bid costs of PLA andnon-PLA projects. Driven by an increase inthe student population, and encouraged byfinancial support from the state, many ofthe roughly one hundred towns and citiesin the greater Boston area have financedschool construction over the past severalyears. Some towns had PLAs in effect dur-ing the construction bidding process whileothers did not. Using data on constructionbid costs, adjusted for inflation with an ap-propriate construction cost index, we esti-mated the difference in bid cost per squarefoot of construction between schools witha PLA in effect and schools with no suchagreement. Before reporting the results, wepresent the sources of the data that we used,then explain how we adjusted for the risein construction bid costs over time.

Data SourcesSurprisingly, the Commonwealth of Mas-sachusetts does not keep adequate or de-tailed information on the schools that arebuilt largely at its expense. We started byobtaining data on bid costs and other vari-ables from F.W. Dodge, McGraw-Hill Con-struction Information Group, a division ofthe McGraw-Hill Companies, in Lexington,Massachusetts. Dodge provided us with in-formation on school construction projectsin the greater Boston area for the period1995 through 2003, including contact infor-mation for town and school district officials,construction companies, and architecturalfirms. Using this (and other) contact infor-mation – for town and city officials, and insome cases architects and contractors - wecollected data for each school-constructionproject listed in Dodge, including the baseconstruction bid, the size of the project mea-sured in square feet, whether there was aPLA requirement on the project, and the na-ture of the construction (new or additionversus renovation). Every observation onbid or actual costs provided by Dodge wasverified using at least one other source, usu-ally in writing. This care was taken to en-

7

sure that erroneous numbers did not findtheir way into the data set.26

We then excluded all projects with a valua-tion below $5 million, on the grounds thatprojects of this size are typically too small tobe of interest to union contractors. We fur-ther focused our study on school construc-tion projects between 40,000 and 400,000square feet in size, in order to exclude ab-normally small or large projects. Our samplecomprises the 126 projects for which we haddata, 17% of which involved PLAs, the re-mainder of which did not.27 Several towns,attempting to realize economies of scale sav-ings, included construction at multipleschool sites in a single bid as one largeproject. We had no choice but to treat thesemultiple school cases as one constructionproject and therefore as one observation inour statistical analysis.28 For these projects,we used the base construction bid for theproject and divided it by the sum of the newand renovated square footage for all theschools within the project to determine thecost per square foot.29

Adjusting for InflationOur sample of schools covers the period 1995to the present. In order to compare the con-struction bid costs of PLA with non-PLAschools, it was first necessary to correct forthe fact that construction costs rose duringthis period, so that all costs could be ex-pressed in 2001 prices. Specifically, we con-structed a cost index that included both thetrend in construction wages and the trend inmaterials costs between 1995 and 2001. Us-ing 2001 as the base year, we first constructeda wage index, which was based on totalwages and salaries for construction workersin Massachusetts (BEA table SA05) dividedby the total number of construction workersin that sector (BEA table SA25).30

In order to account for the changes in mate-rials costs, we constructed a price indexbased on the producer price index for the“other” subcomponent of Intermediate Mate-rials, Supplies, and Components, as reported inThe Economic Report of the President, February

2003.31 To construct the final cost indexused in our analysis, we weighted thewage index and the adjusted producerprice index equally, to reflect the relativeimportance of wages and materials costsin a typical construction project.

Comparing PLA to Non-PLA ProjectsA comparison of the key characteristics ofthe school construction projects in townswith a PLA (“PLA projects”) with thosewhere there was no such agreement (“non-PLA projects”) is shown in Table 1. A no-table pattern in the data is that PLAprojects, on average, cost $18.26 ($152.46minus $134.20) more per square foot (in2001 prices) than non-PLA projects.

The table shows that the cost per squarefoot it higher for PLA than for non-PLAprojects. A formal test shows this differ-ence to be highly statistically significant,so the difference does not appear to be dueto chance.32 However, this test is not con-clusive, because it is possible that PLAprojects are systematically different – forinstance larger, or concentrated on newbuildings rather than renovations.

One way to determine whether or not thedifference in PLA versus non-PLA projectsis robust to differences in project size andtype is with a formal regression analysis.The dependent variable is the cost persquare foot of construction (in 2001 prices).The independent variable of most inter-est to us is a dummy variable that is setequal to 1 for PLA projects and to 0 other-wise. We control for whether the projectinvolves new construction or a renovationby including a dummy variable set equalto 1 for new projects and to 0 otherwise.We also control for the impact of a project’sscope on the cost per square foot by con-trolling explicitly for square footage, andfor square footage squared. This is desir-able because there may be economies ofscale (within reason) in school construc-tion, so that larger schools may have lowercosts per square foot. The ordinary least

8

squares regression results are presented inTable 2.

Our regression results show that PLAprojects add an estimated $18.83 per squarefoot (in 2001 prices) to the bid cost, control-ling for whether or not the project involvesnew construction, and controlling for theproject’s square footage. A formal (one-tailed) test of the statistical significance ofthis coefficient gives a p-value of 0.000,which means that there is less than a 0.1%chance that we have accidentally found thatPLA projects are more expensive than non-PLA projects. Put another way, there is atleast a 99.9% probability that PLA projectsreally are more expensive than non-PLAprojects, holding other measurable aspectsof a project constant. The equation alsoshows that projects involving new construc-tion, rather than renovations, experiencesignificantly higher costs per square foot,as one would expect.

With an adjusted R2 = 0.31, the equation “ex-plains” a respectable 31% of the variationin construction bid costs across towns.Clearly, other factors also influence the costof construction – the exact nature of the site,the materials used for flooring and roofing,the outside finish, and the like. As a practi-cal matter, collecting viable information at

this level of detail, for all 126 projects,would be impossible. Thus our equationnecessarily excludes these unobservablevariables. However, this does not under-mine our finding of a substantial PLA ef-fect. For the PLA effect shown here to beoverstated, it would have to be the case thatPLA projects systematically use more ex-pensive materials, or add more enhance-ments and “bells and whistles,” than non-PLA projects. Our conversations withbuilders, town officials and architects sug-gest that PLA projects are not systemati-cally more upscale. This gives us confi-dence that the PLA effect shown here isreal.

RobustnessIt is helpful to explore the robustness of ourresults. In other words, is there still a PLAeffect if we only look at elementary schoolconstruction, or new projects, or mid-sizeprojects, or if we use actual costs ratherthan bid costs. The results of this exerciseare summarized in Table 3.

The first column indicates the sample, orsub-sample, used in estimating the regres-sion equation. The first four rows use thelargest possible sample, but vary in whichother variables are included in the equa-

TABLE 1: SUMMARY STATISTICS FOR CONSTRUCTION PROJECTS BY PLA STATUS

Variable

Winningconstruction bid

in millions of 2001dollars

Size of project(square feet)

Construction bidcost/square footin 2001 dollars*

Number ofstories

Mean

PLANon-PLA

$22.92$16.95

151,213131,440

$152.46$134.20

3.112.39

Standard Deviation

PLANon-PLA

$ 10.71$ 7.77

69,43267,656

$ 19.99$ 24.44

0.760.78

MinimumPLA

Non-PLA

$7.37$6.30

45,19045,000

$128.56 $72.72

11

Maximum

PLANon-PLA

$42.31$40.89

286,650383,000

$202.93$199.26

44

Total sample size is 126, with 21 PLA projects and 105 non-PLA projects. Costs are measured in 2001dollars; see text for details.

9

tion. Our analysis proceeded by runningseparate regressions for

1. elementary and non-elementaryschools;

2. new construction projects and reno-vations;

3. mid-size projects (100,000 to300,000 square feet) only;

4. small projects (defined as below themedian of 118,500 square feet) andlarge projects;

5. the largest available sample, usingfinal costs (rather than bid costs);and

6. a smaller sample, using final costs,but excluding those cases where re-ported final costs equaled reportedbid costs.

The “PLA effect” column shows the esti-mate of the effect of having a PLA on thecost of construction (in dollars per squarefoot, in 2001 prices), and the adjoining “p-value” column measures the statistical sig-nificance of these coefficients. In every casethe PLA effect is statistically significant atthe 10% level or better. The size of the PLAeffect differs somewhat, depending on thesample examined and the other variablesthat are included. The results of the“baseline” regression analysis presented inTable 2 are reproduced here in the first row;this equation has the virtue of including asmany observations as possible, while be-ing parsimonious in the use of variables.

In analyzing the robustness of our results,four points are worth making. First, thereappears to be a significantly larger PLA ef-fect for junior high and high schools

($34.60/sq.ft.) than for elementary schools($12.49/sq.ft.); possibly the secondary-levelschools are more complex to build. Second,the PLA effect for new construction ($14.90/sq.ft.) is smaller than for renovations($25.67/sq.ft.); perhaps renovations areharder to predict accurately. Third, the PLAeffect for mid-sized projects – defined asthose between 100,000 and 300,000 squarefeet – is, at $19.92/sq.ft., similar to that forthe sample as a whole ($18.83/sq.ft.).

Fourth, and most interestingly, the PLA ef-fect is essentially the same whether one usesbid costs or actual costs of construction.33

This is important, because a construction in-dustry newsletter criticized our earlier studyfor using bid costs only, on the grounds that“what’s sorely needed is data that comparesfinal construction costs on structures that aresimilar.”34

Of the 126 projects, information on actualconstruction costs was reported in only 62cases; for this sub-sample, the PLA effect was$16.51/sq.ft. for actual costs (a 12% increase)and $16.92/sq.ft. for bid costs. For twelvecases, the project was reported to be “on bud-get,” which we took to mean that reportedactual cost was the same as the reported bidcost. While this is certainly plausible, we didexperiment by removing these cases and es-timating the PLA effect with the remaining50 cases. For the restricted sub-sample thePLA effect was $11.80/sq.ft. for actual costs,which is very similar to the effect for bidcosts ($11.52/sq.ft.).

TABLE 2: ORDINARY LEAST SQUARES ESTIMATION OF REAL CONSTRUCTION BID PERSQUARE FOOT

Variable Coefficient Standard Error p-value (one-tailed test)

Constant 138.69 4.96 0.00

PLA 18.83 3.93 0.00

New 17.89 2.72 0.00

Square Feet -12.36 4.97 0.00

Adjusted R2 is 0.31. Sample size is 126. Square footage is measured in 100,000s.

10

An examination of the residuals for ourpreferred equation (row 2 in Table 3)showed no evidence ofheteroskedasticity. A robust regressionestimate of this equation, using theHuber-White estimator, also found thep-value to be 0.000 (to three decimalplaces).35

Following standard practice, our regres-sions use ordinary least squares, whichmeans that each observation (here, aschool building project) carries equalweight in the regression. However, wealso estimated our preferred equation us-ing weights, where each project is givena weight that is in proportion to thesquare footage that it represents. Thismeans that a project of 150,000 squarefeet, for instance, would have twice asmuch weight in the equation as a projectof 75,000 square feet. The weighted re-gression shows a PLA effect of $20.51/sq.ft., again highly statistically signifi-cant.

IV. CONCLUSION

It is widely believed that Project LaborAgreements add to the cost of construc-tion projects. However, there has until re-cently been no statistically robust studyof whether PLAs add to construction costsin practice, and if they do, how large theeffects are. By carefully constructing a da-tabase with information on the bid costsof school construction projects under-taken in the greater Boston area since1995, and comparing the costs in townswith, and without, PLAs, we found thefollowing:

(v) PLA projects have higher con-struction-bid costs; we aremore than 99% confident ofthis assertion, based on theavailable data.

(vi) PLA projects have higher ac-tual costs; again we are morethan 99% confident of thisfinding, based on the availabledata.

(vii) The finding that PLA projectshave higher construction costsis robust, in that:

TABLE 3. REGRESSION ESTIMATES OF THE “PLA EFFECT” FOR DIFFERENT SUB-SAMPLES AND MODELSPECIFICATIONS

Mean cost/sq ftPLA effect($/sq ft)

p-value Other variablesincludedk

Sample size(# of PLAprojects)

AdjustedR

2

Non-PLA projects PLA projects

Bid cost/sq ftAll observations 18.83 0.000 New, sqft 126 (21) 0.31 134.2 152.5

All observations 19.09 0.000 New, sqft, sqft2 126 (21) 0.31 134.2 152.5All observations (weighted)

a20.51 0.000 New, sqft 126 (21) 0.39 134.2 152.5

All observationsb

17.86 0.004 New, sqft, sqft2, floors,element, distance

117 (18) 0.29 134.4 152.5

All observationsb

12.91 0.036 New, floors, element,distance

117 (18) 0.23 134.4 152.5

Elementary schools only 12.49 0.053 New, sqft, sqft2 76 (15) 0.12 140.4 149.7Jr. Hi & Hi schools only 34.60 0.000 New, sqft, sqft2 50 (6) 0.51 125.5 159.5Memo: p value

c0.04

New construction only 14.90 0.003 Sqft, sqft2 85 (16) 0.12 141.7 151.9Renovations only 25.67 0.056 Sqft, sqft2 41 (5) 0.23 119.9 151.0Mid-size projects only

d19.92 0.001 New, sqft, sqft2 74 (16) 0.31 128.1 152.4

Small projects onlye

14.41 0.095 New, sqft, sqft2 64 (7) 0.19 141.2 156.2Large projects only

f20.01 0.003 New, sqft, sqft2 62 (14) 0.32 125.9 150.6

Actual costs/sq ftSample 1

g16.51 0.012 New, sqft, sqft2 62 (14) 0.40 133.6 153.1

Memo: bid costsh

16.92 0.009 New, sqft, sqft2 62 (14) 0.45 128.8 149.4Memo: p value

c0.77

Sample 2i

11.80 0.094 New, sqft, sqft2 50 (10) 0.44 133.0 146.2Memo: bid costs

j11.52 0.093 New, sqft, sqft2 50 (10) 0.50 127.3 141.1

Memo: p valuec

0.86

Notes: Maximum sample size: 126. The baseline regression is in first row (boldface) and reproduces the results shown in Table 2.aWeighted regression, where observations were weighted by the size (in square feet) of each project.

bSmaller sample size

because values were missing for some variables. cTests difference in PLA effect between the previous two rows.

dOnly projects

between 100,000 and 300,000 square feet. eLess than 118,500 square feet (median project size in sample).

fGreater than 118,500

square feet (median project size in sample). gLargest available sample for which actual costs were reported.

hExcludes

observations where reported actual cost equaled reported bid cost. iActual costs were, on average, $4.48 higher than bid costs, for

this sample. jActual costs were, on average, $5.56 higher than bid costs, for this sample. kNew = 1 if new construction, 0 if

renovation. Sqft = number of square feet in project. Sqft2 = number of square feet squared. Floors = number of stories. Element= 1 if elementary school, 0 otherwise. Distance = miles from Boston.

11

a. Robust estimation tech-niques still show a strongstatistically significant ef-fect;

b. A regression that weightsobservations by project sizealso shows the effect;

c. The effect persists evenwhen the data are subdi-vided, so that the effect isevident separately for largeprojects and for small, forelementary schools and forsecondary schools, for newconstruction and for reno-vations.

(viii) PLA projects add an estimated$18.83 per square foot to the bidcost of construction (in 2001prices), representing an almost14% increase in costs over theaverage non-PLA project. Thelow estimates find that actualproject costs are raised by8.4%;36 the high estimates findthat bid costs are raised by14.9%.37

(ix) PLA projects add an estimated$16.51 per square foot to theactual cost of construction (in2001 prices). This may be anunderestimate, since it is basedon a subsample for which thePLA effect for bid costs ($16.92/sq.ft.) is somewhat below thefull-sample PLA effect ($18.83/sq.ft.).

In sum, the evidence that Project LaborAgreements have increased the cost ofschool construction in the Boston area since1995 is strong. The effect is also substantial.Taken together, PLA projects accounted for3.175 million square feet of constructionwith a combined cost of $481.4 million (in2001 prices), based on the projects that wewere able to include in our study. Ourestimates show that this cost was $60million higher than it would have been ifProject Labor Agreements had not beenused.38 On average, almost 70% of this

additional cost was borne by theCommonwealth of Massachusetts,through the School Building AssistanceProgram; the remainder was covered bythe cities and towns that built schoolsusing Project Labor Agreements.Officials at both the state and local levelneed to be aware of the higher costsimplied by PLAs when they choose thebid requirements on future schoolconstruction projects. �

12

ENDNOTES

1 Jonathan Haughton, Darlene C. Chisholm and Paul Bachman, The Effects of ProjectLabor Agreements on Publicly Funded Construction Projects in Massachusetts, (Boston:Beacon Hill Institute, March 2003).2 This may be an underestimate, since it is based on a subsample for which the PLA effectfor bid costs ($16.92/sq.ft.) is somewhat below the full-sample PLA effect ($18.83/sq.ft.).3 Haughton, et al., Effects of Project Labor Agreements.4 U.S. General Accounting Office, Project Labor Agreements: The Extent of Their Use

and Related Information, (Washington D.C.: 1998), Publication No. GAO/GGD-98-82.

Retrieved September 25, 2002 from the General Accounting Office web site: http://

www.gao.gov/archives/1998/gg98082.pdf.5 Ibid., p. 46 Ibid., p. 47 Worcester Municipal Research Bureau, “Project Labor Agreements on PublicConstruction Projects: The Case For and Against,” Report No. 01-4, (May 21, 2001).p. 7.8 Herbert R. Northrup and Linda E. Alario, “Government-Mandated Project LaborAgreements in Construction, The Institutional Facts and Issues and Key Litigation:Moving Toward Union Monopoly on Federal and State Financed Projects,”Government Union Review 19, no 3, (2000): 91.9 Ibid., pp. 12-13.10 Amanti & Sons, Inc., and Others v. City of Lynn, Supr. Ct. Memorandum of Decisionand Order on Plantiffs for Preliminary Injunction (Civil Action No.907-2780-C, June 9,1997): 2, as quoted in Northrup and Alario, p. 91.11 Ibid., p.91.12 Worcester Municipal Research Bureau, “Project Labor Agreements,” p. 9.13 John T. Callahan & Sons, Inc., & Others v. City of Malden & Another. SJC-07959,Lexis-Nexis 493, 22 July 1999 p.2.14 Ibid., p. 2.15 Ibid., p. 2.16 Ibid., p. 2.17 Ibid., p. 6.18 Worcester Municipal Research Bureau, “Project Labor Agreements,” p. 9.19 Massachusetts Executive Office of Administration and Finance, Reconstructing theSchool Building Assistance Program, Policy Report Series No. 3 (January, 2000): p. 1.20 Rick Klein, “Eyes Turn to State’s $726m Reserve,” Boston Globe, September 4, 2003, p.A1.21 Massachusetts G.L. Chapter 70B: Section 1 School building assistance program; [on-line] http://www.state.ma.us/legis/laws/mgl/70B-1.htm.22 Massachusetts Executive Office of Administration and Finance, Reconstructing theSchool Building Assistance Program Policy Report, p. 2.23 FY 1999 refers to the period July 1, 1998 through June 30, 1999.24 http://finance1.doe.mass.edu/doe_budget/1_doe.html.25 Massachusetts Executive Office of Administration and Finance, Reconstructing theSchool Building Assistance Program Policy Report, p.2.26 Our earlier study incorrectly identified schools in Wilmington and Melrose as havingbeen built under PLAs; this was based on erroneous information that was supplied tous. This is why we have taken care to double check every bit of information, in a goodfaith effort to ensure that no further errors remain.27 PLA contracts were in effect in the following towns: Boston, Lawrence, Lynn, Malden,Medford, Milton, and Waltham (for two of the four schools in the data set). The ClassicalHigh School project in Lynn is considered a PLA project since our construction bid

13

About the Authors

Paul Bachman, MSIE. Mr. Bachman is Research Economist at the Beacon Hill Institute for Public PolicyResearch at Suffolk University. He holds a Master of Science in International Economics from SuffolkUniversity.

Darlene C. Chisholm, PhD. Dr. Chisholm is Senior Economist at the Beacon Hill Institute for PublicPolicy Research at Suffolk University and Associate Professor in the Economics Department at SuffolkUniversity. She holds a Doctorate in Economics from the University of Washington.

Jonathan Haughton, PhD. Dr. Haughton is Senior Economist at the Beacon Hill Institute for Public PolicyResearch at Suffolk University and Associate Professor in the Economics Department at Suffolk Univer-sity. He holds a Doctorate in Economics from Harvard University.

David G. Tuerck, PhD. Dr. Tuerck is Director of the Beacon Hill Institute for Public Policy Research atSuffolk University and Chairman of the Economics Department at Suffolk University. He holds aDoctorate in Economics from the University of Virginia. His dissertation director was James M.Buchanan, Nobel Laureate in Economics.

The authors would like to thank Dali Jing, Corina Murg and Hatesh Radia for their contributions to thisstudy.

information predates the lawsuit that overturned the PLA requirement and forced theproject to be re-bid without a PLA requirement.28 These include projects in the towns of Andover, Beverly, Brockton, Haverhill, FallRiver, Lancaster, Medford, Taunton, Walpole and Weston.29 Industry professionals including architects, construction project managers, andestimators concurred with our treatment of the data.30 The main source of wage and salary data is the Bureau of Economic Analysis web site,(http: //www.bea.gov, accessed September 10, 2003). The series used the StandardIndustrial Classification (SIC) through 2001. We used the North American IndustrialClassification Series (NAICS) for wages and salaries for 2001 and 2002, and employmentfor 2001. To get the percentage increase in construction employment in Massachusettsbetween 2001 and 2002 we used data from the Massachusetts Department ofEmployment and Training (http://www.detma.org/lmilmi.htm#790, accessedSeptember 10, 2003). We also used this source to measure the growth in constructionwages between 2002 (first 7 months) and 2003 (first 7 months).31 The source of the producer price index is Table B-66, “Producer Price Indexes by Stageof Processing, Special Groups, 1974-2001,” Economic Report of the President, February 2003,p. 353. We assumed no growth in this index between 2002 and 2003, in line with recenthistorical experience.32 The null hypothesis is that PLA projects do not cost more than non-PLA projects. Atwo-sample t-test with unequal variances rejected this null hypothesis decisively (p =0.0004).33 Note that final base construction costs, like bid costs, contain site work and, for someprojects, demolition costs.34 “Do PLA projects raise costs? New study that supports ABC’s arguments attacked as‘inflated and unreliable’,” Cockshaw’s Construction Labor News + Opinion 33, no. 5, (2003): 8.35 The t-statistic for the PLA dummy variable in the equation reported in Table 2 and linetwo of Table 3 was 3.80; for the robust regression the t-statistic was 3.77.36 PLA effect from sample 2 for actual costs ($11.80/sq.ft.) divided by average actual costsfor this group ($135.65/sq.ft.)37 PLA effect from weighted regression for bid costs on full sample ($20.51/sq.ft.) divided byaverage bid costs for this group ($137.24/sq.ft.).38 $60 million = 3.175 million sq.ft. times $18.83 per sq.ft.

ENDNOTES (Cont.)

The Beacon Hill Institute for Public Policy ResearchSuffolk University

8 Ashburton Place

Boston, MA 02108

Phone: 617-573-8750 Fax: 617-720-4272

bhi@beaconhill.org

http://www.beaconhill.org

ISBN 1-886320-18-7

The Beacon Hill Institute at Suffolk University in Boston focuses on federal, state and localeconomic policies as they affect citizens and businesses. The institute conducts research and

educational programs to provide timely, concise and readable analyses that help voters,policymakers and opinion leaders understand today’s leading public policy issues.

©September 2003 by the Beacon Hill Institute at Suffolk UniversityISBN 1-886320-18-7