Summary Statement
This study examines a large number of project labor agreements (PLAs) using a variety of techniques, including archival research, interviews, case studies and the statistical analysis of original data. PLAs are pre-hire collective bargaining agreements that establish the terms and conditions of employment on one or more construction projects. Interview evidence suggests that safety inputs are greater on PLA projects including language establishing labor/management safety committees.
2007
4. Bidding and Costs
The bidding research compares projects in the East Side Union High School district of San Jose, California with the San Jose Unified School district. The former used a PLA on a series of school construction projects while the latter did not. The research on costs examines 108 school construction projects in New England.
The results show the use of a PLA neither lowers the number of bidders nor increases costs when other important variables are taken into account.
The East Side Union High School district in San Jose is responsible for the education of 24,000 high school students. A neighboring district, the San Jose Unified School district, enrolls 32,000 students ranging from kindergarten through high school. In March 2002, voters in both districts approved bond issues for school construction, repair and renovation. The East Side vote allowed the district to borrow up to $300 million. In San Jose, the vote capped borrowing at $429 million. In 2004, the East Side district entered into a PLA with the Santa Clara and San Benito Building and Construction Trades Council. The San Jose district chose to build without a PLA.
The different decisions of the districts with regard to a PLA provided the perfect ingredients for a naturally occurring experiment.We can compare bidding behavior with the East Side district before and after the implementation of the PLA, and we can compare across districts.
There were 21 projects in the East Side district bid under the PLA and 35 projects bid during the same period without a PLA in the San Jose district. Also, there were 12 projects bid prior to the PLA agreement in the East Side district and 96 projects in the San Jose district during the same period. In sum, there were 164 projects, 21 of which were built under a PLA.
The East Side and San Jose districts are adjacent and, therefore, within the same construction market. The time is also the same. However, there are two potentially important differences. The East Side projects were, in dollar value, approximately two to three times larger than the San Jose projects both before and after the use of PLAs. Also, the two districts employ different bidding procedures. The East Side district favors hiring a single prime contractor, who then seeks its own subcontractors, while the San Jose district treats specialty contractors as individual prime contractors.
Statistics indicate that the East Side district received, on average, fewer bidders per bid opening than the San Jose district (approximately 4.5 versus approximately 4.0). This result would be consistent with the findings of those who argue that PLAs reduce the number of bids on a project, except that the result holds for both before and after the implementation of the PLA. In fact, the difference between the two districts decreases after the acceptance of the PLA. Further, there was a drop in the number of bidders across both districts over the time period. This decrease may be associated with an increase in construction activity in the area at the time. Bureau of the Labor Statistics data for the San Jose-Sunnyvale-Santa Clare area show more employment in construction during 2004 than in 2003. Assuming that this statistic reflects more construction activity, fewer contractors would be willing to bid the projects than if they were experiencing a slack period.
The small difference in the number of bidders both before and after the PLA across districts is likely tied to the differing methods of construction management. The San Jose district favors separate prime contracts on specialty work. Since there are more specialty than general contractors in most construction markets, that fact alone may account for more bidding activity.
One way to find out what the effects of all these possibilities are is to place a number of variables in a multiple regression model.23 In doing so, the only statistically significant variable that predicts bidding behavior is business cycle. In the period that construction activity increased, the number of bidders per bid opening decreased. Most notably, the results of the study indicate that the presence of a PLA has no statistically significant effect on the number of bidders per bid opening.
Whether PLAs increase or decrease the number of bidders is probably of little interest to those who ultimately pay for construction projects.What is of keen interest is whether PLAs increase, reduce or have no effect on project costs. In examining 108 school projects in New England, ten of which were built with PLAs, the presence of a PLA does not have a statistically significant effect on the final cost of a project. The research on costs is modeled closely after several studies done by the Beacon Hill Institute (BHI) at Suffolk University in Boston. In 2003 and 2004, BHI produced reports on the effects of PLAs on school construction costs in the Greater Boston area and in Connecticut. Their original study found that PLAs increased construction costs by 17.3% (or $31.74 per square foot) in the Boston area. A subsequent study, which corrected several problems in the first, lowered the estimate to about 12% (or $16.51). In extending the research to Connecticut, the researchers found a PLA premium of $30.00 per square foot.24
Similarly, the research includes a model, predicting costs on 108 school projects in New England. Studying schools has several advantages. First, there are more schools than, say, power plant projects in an area, which allows us to have enough observations within a relatively homogenous construction market. Further, while by no means identical, schools are enough alike to provide a basis for meaningful comparison. Finally, there are both public and private schools, which allows us to examine both private and public construction.
Returning to the BHI studies, there were a number of problems with the research. But the main complaint is with the presumption stated in the following paragraph:
- Clearly, other factors also influence the cost of construction—the exact nature of the site, the materials used for flooring and roofing, the outside finish, and the like. As a practical matter, collecting viable information at this level of detail for all 126 projects, would be impossible. Thus, our equation necessarily excludes these unobservable variables. However, this does not undermine our finding of a substantial PLA effect. For the PLA effect shown here to be overstated, it would have to be the case that PLA projects systematically use more expensive materials or add more enhancements and “bells and whistles” than non-PLA projects. Our conversations with builders, town officials and architects suggest that PLA projects are not systematically more upscale.25
The BHI researchers dismiss the possibility that PLA projects have more amenities or are more complex than non-PLA projects. Such factors, however, determine why projects are built with PLAs in the first place. To hold otherwise is to ignore prevailing public policy. In many states—particularly in New England—court decisions require public owners to establish the need for a PLA before using one. The size of a project, its complexity and the need for timely completion are all variables that must be considered.
Since the BHI researchers do not believe that PLA projects are “systematically more upscale” they included very few variables in their models that could affect construction costs. Other than whether a PLA had been used, they controlled for little more than the size of the project in square feet, whether a project was new construction or a renovation and, in the Connecticut study, the number of stories and if the project involved an elementary or high school. The methodological problem with such a lean specification is that effects are attributed to the presence of a PLA when they actually result from some unobserved variable or variables.
Finding detailed information for a large number of construction projects is very difficult work. However, we were able to find information— through speaking with architects, construction managers, school department officials, etc.—on thirty variables across the 108 projects in New England.
The descriptive statistics alone tell us that PLA covered projects are inherently different than non- PLA projects. For example, the average square footage for a PLA school is approximately 157,000 while a non-PLA school is close to 118,000. PLA schools average more than three stories while non-PLA schools average fewer than three. All the PLA projects required prior demolition work, while less than half of the non-PLA schools required such work.
Using the data we assembled, we created a multiple regression model.26 The dependent variable is the logarithm of the final cost of a project. Using the logarithm of final cost rather than final cost itself allows us to interpret the effects of the independent variables in percentage terms.
When we enter all the variables in a regression equation, we find that significant positive effects are associated with the size of a project (i.e. square footage), whether the building is an elementary school, the construction of an auditorium, cafeteria or kitchen, whether the roof includes both low and steep pitches, and whether the project was located in an urban area. While our model suggests that a PLA adds 7.8% to project costs, the result is not statistically significant. In fact, the PLA variable is so weakly predictive, that the actual effect could range anywhere from -14.4% to 29.9%.
The inherent difficulties in this type of research—identifying the labor relations practices on projects, gathering information on building amenities, materials and aspects of design, etc.— make it unlikely that large samples can ever be used. But small samples, such as the ones by BHI and this one, have a number of problems. Perhaps the main problem is that they can be very sensitive to outlying values. One or two projects that are very different from the majority can skew results. Therefore, results need to be interpreted with caution.
Nonetheless, our conclusion is that the additional costs observed on PLA projects by previous researchers likely have little to do with the PLA itself, but result from the additional amenities or requirements that are inherent in large, complex jobs, which are more likely to be covered by PLAs. We find no strong evidence that PLAs affect final costs either positively or negatively.
To conclude, if PLAs are, in fact, cost neutral, then more attention must be paid for other outcomes that can be achieved with PLAs, such as timely completion, better safety outcomes, training opportunities and industry recruitment. The next chapter investigates some of these issues through case studies of four projects, each of which had distinctive requirements.
