Date of Award

Spring 1-1-2015

Document Type


Degree Name

Doctor of Philosophy (PhD)

First Advisor

Paul M. Goodrum

Second Advisor

Keith Molenaar

Third Advisor

Mathew R. Hallowell

Fourth Advisor

James Diekmann

Fifth Advisor

Ray Littlejohn


The goal of finding a reliable productivity metric for the construction industry has remained elusive for decades. Indeed, the goal of developing such an index for the US construction industry was described as a significant industry need by governmental agencies, industry practitioners, and academics in a 1983 Business Roundtable industry report (BRT 1983). This need was later echoed as a high priority in Advancing the Competitiveness and Efficiency of the U.S. Construction Industry, a report by the U.S. National Research Council (NRC 2009).

Difficulty in measuring real output in the industry is a challenge that has prevented reliable productivity metrics. An alternative approach would be to consistently measure activity productivity across multiple construction projects through the United States and develop an aggregate measure accordingly. However, activity measures are inconsistent across both construction projects and even projects within the same company. Identifying an industry standard measure of productivity via accounting code would be a critical first step towards improving industry performance.

This dissertation uses multiple sources of industry data to develop several proposed master code of accounts for mechanical piping systems, which represent scopes of work that cut across multiple construction sectors that exhibit some of the most complex construction systems on most jobsites. This effort builds on the work of Dadi et al (2014), but it significantly extends on this previous work using wide-ranging and detailed data sources. In addition, the dissertation utilizes Classification and Regression Trees (CART) as a more objective and scientific approach of identifying a more natural master code structure.

The dissertation’s primary contributions occur in three parts. First, the dissertation develops a methodological approach to the development of a system of productivity measurement though a post hoc analysis of leading factors of influence exposed through data mining. Second, the characteristics of large scale of estimating source data are assessed as a virtual census of industry rates. Third, in an unrestrained, non-parametric, system of primary factors, natural forms of interaction occur between dimensional domains that can be illustrated through the formation of Euler diagrams under first order logic and set theory.