Copyright by Prakash Singh 2009 The Thesis committee for Prakash Singh certifies that this is the approved version of the following thesis: Factors Affecting the Cost of Engineering for Transportation Projects APPROVED BY SUPERVISING COMMITTEE: Supervisors: William J O'Brien Khali R Persad Factors Affecting the Cost of Engineering for Transportation Projects by Prakash Singh, B.E. Thesis Presented to the Faculty of the Graduate School of the University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of Master of Science in Engineering The University of Texas at Austin December 2009 To my parents v ACKNOWLEDGEMENTS I would like to convey my thanks to Dr. William O’Brien and Dr. Khali Persad for guiding me throughout my research work. I would also like to thank to my fellow graduate student Krishnaprabha Krishnanivas Radhakrishnan for our collaboration as she developed her research on outsourcing procedures. PRAKASH SINGH The University of Texas at Austin December 2009 vi Factors Affecting the Cost of Engineering for Transportation Projects by Prakash Singh, MSE The University of Texas at Austin, 2009 SUPERVISOR: William J O’Brien State DOTs (department of transportation) spend billions of dollars on construction and maintenance of transportation projects every year. In addition, significant sums go to preliminary and construction engineering (PE and CE). For many projects, DOTs utilize engineering services from consultants, to supplement in-house engineering. The cost and quality of consultant’s engineering services compared to in-house, are important issues to justify the involvement of consultants. This report provides an analysis of those issues on Texas Department of Transportation (TXDOT) projects. Traditionally, the costs of PE and CE are calculated as a fixed percentage of total project construction cost, and the efficiency of engineering organizations is assessed by comparison of their gross percentages. However, the results presented here show that project scope and complexity are significant factors in PE and CE cost. Therefore, simplistic comparisons of PE and CE percentages can be misleading when applied across a mixed program of projects. vii TABLE OF CONTENTS ACKNOWLEDGEMENTS……………………………………………………………..v ABSTRACT……………………………………………………………………………...vi LIST OF TABLES………………………………………………………………………xi LIST OF FIGURES……………………………………………………………………xiii Chapter 1. Introduction…………………………………………………………………1 1.1Background………………………………………………………………………1 1.2 Research Scope………………………………………………………………….4 1.3 Benefits……………………………………………………………………….....4 1.4 Limitation………………………………………………………………………..5 1.5 Organization of Thesis…………………………………………………………..5 Chapter 2. Literature review…………………………………………………………....6 2.1 Project Development Process…………………………………………………...6 2.1.2 Planning and Programming………………………………………………..8 2.1.3 Preliminary Design………………………………………………………..9 2.1.4 Environmental……………………………………………………………10 2.1.5 Right of way utilities……………………………………………………..11 2.1.6 Plans, Specifications and Estimates (PS&E) Development……………… …………..………………………………...11 2.1.7 Letting……………………………………………………………………12 2.2 In-house and Consultant cost comparison……………………………………..12 viii 2.2.1 Texas State Department of Highways and Public Transportation (SDHPT),1987…………………………………………………..……….14 2.2.2 Ernst and Whinney, 1987………………………………………………...14 2.2.3 Center for Transportation Research, 1987……………………………….14 2.2.4 Texas Transportation Institute, 1987…………………………...………..15 2.2.5 Legislative Audit Bureau of Wisconsin, 1990…………………...………15 2.2.6 Professional Services Management Journal, 1992 ……………………....15 2.2.7 University of California at Berkeley, 1992……………………………....16 2.2.8 Alaska Legislative audit, 1994…………………………………………...16 2.2.9 Price Waterhouse Coopers, 1998……………………………………...…16 2.2.10 Louisiana State Department of Transportation study, 1999……………16 2.2.11 National Cooperative Highway Research Program, 2001…………..….17 2.2.12 California Department of Transportation study, 2007…………….……17 2.2.13 Department of Civil Engineering Polytechnic Institute of NYU, 2008.…………...………………………………………………….….…18 2.2.14 TxDOT survey of in-house and consultant cost, 2008………………....18 2.3 Summary…………………………………………………………………...…..20 Chapter 3. Data Description and Analysis Methodology ……………………………23 3.1 Data Summary…………………………………………………………………23 3.2 Function Codes……………………………………………………………..….24 3.3 PE Charges……………………………………………………………………..26 3.4 CE Charges…………………………………………………………………….29 ix 3.5 Data Checks…………………………………………………………………....32 3.6 Data Transforms……………………………………………………………..…35 3.7 Data Analysis Methodology……………………………………..…………….35 Chapter 4. Comparison of Costs for In-house and Mixed Projects…………………38 4.1 Initial Comparison of PE Costs………………………………………………...38 4.1.1 Difference between Mixed and In-house PE Costs Intercept……………39 4.1.2 Difference by Project Construction Cost Intercept………………………39 4.1.3 Difference by Project Construction Cost Slope………………………….40 4.2 Difference by Project Type…………………………………………………….40 4.2.1 Graphical Lines of Fit……………………………………………………42 4.2.2 Interpretation of Results……………………………………………….....46 Chapter 5. Direct Comparison of In-house and Consultant PE Costs……………...51 5.1 Function Codes with 100% Consultant Charges……………………………....51 5.2 Analysis and Results………………………………………...…………………52 Chapter 6. Quality of In-house and Mixed PE Projects……………………………...56 6.1 Change Order Analysis for Different Project Types…………………………...56 6.1.1 Change Order Analysis Results…………………………………….……58 Chapter 7. Differences in PE Costs across Districts………………………………….65 7.1 Summary of PE Costs by District……………………………………………...65 7.2 Simple Comparison of Districts……………………………………..…………69 7.3 Comparison of Districts Considering PE Provider…………………………….70 7.4 Comparison of Districts Considering PE Providers and Project Type………...76 x 7.5 Change Orders in Districts……………………………………………………..79 7.5.1 Change Order Rates………………………………………………...……79 7.5.2 Comparison of Change Orders across Districts………………………….80 Chapter 8. CE Results………………………………………………………………….88 8.1 Difference in CE cost by Project Types………………………………………..88 8.1.1 Graphical Lines of Fit……………………………………………………89 8.1.2 Interpretation of Results………………………………………………….92 8.2 Difference in CE Cost across Districts………………………………………..96 8.2.1 Difference in Districts Means……………………………………………98 8.2.2 Difference by CE Provider……………………………………………...105 8.3 Chapter Conclusion…………………………………………………………...112 Chapter 9. Conclusion………………………………………………………………...114 9.1 PE Costs……………………………………………………………………....115 9.2 Difference in Project Types, Districts, and PE Quality………………………116 9.3 CE Costs…………………………………………………………………..…..117 9.4 Recommendation……………………………………………………………...118 Appendix A…………………………………………………………………………….120 Bibliography………………………………………………………………….………..127 Vita…………………………………………………………………………………..…129 xi LIST OF TABLES 2.2 PE costs in various state DOTs as compiled by TxDOT…………………………….19 2.3 Summary of previous researches on in-house and consultant cost comparison……..20 3.1 Summary of TxDOT contracts let in FY 06-07……………………………………...23 3.2 TxDOT Function Codes for PE Charges…………………………………………….24 3.3 PE totals for TxDOT construction contracts let in FY 06-07………………………..26 3.4 CE totals for TxDOT construction contracts let in FY 06-07………………………..29 3.5 Projects Classified as Fully In-house or Mixed……………………………………...33 3.6 Summary of PE Charges by Function Code………………………………………....34 4.1 Observed Construction Cost and Estimated Percentage PE by Project Type………..49 5.1 Functions that were done 100% in-house or 100% consultant……………………....51 5. 2 Weights of Functions………………………………………………………………..52 5.3 Function Code SPSS Result………………………………………………………….53 5.4 Consultant/In-house Cost Ratios…………………………………………………….54 6.1 Absolute Value of Change Orders by Project Type………………………………….57 7.1 Summary of District Construction Contracts and PE Costs………………………....65 7.2 Summary of District Project Types and Numbers…………………………………...67 7.3 Summary of that data for projects let in FY 06-07…………………………………..68 7.4 Summary of District Change Orders………………………………………………....79 xii 8.1 Summary of construction totals and CE cost by districts……………………………96 8.2 CE spending on in-house and mixed projects………………………………………..97 xiii LIST OF FIGURES 1.1 Workload Fluctuations in a State DOT………………………………………….…….2 2.1 Project Development Process (TxDOT 2009)………………………………………...7 4.1 Total PE Costs for 1371 TxDOT Projects and Fitted Lines: Log-log Plot…………..43 4.2 Total PE Costs for 1371 TxDOT Projects and Fitted Line......................................…44 4.3 Total PE Costs for 1371 TxDOT Projects and Fitted Lines for In-house and Mixed Projects : Zoomed…………………………………………………………………….….45 4.4 Estimation of Percentage PE Costs for Mixed and In-house Projects based on 1371 TxDOT Projects………………………………………………………………………….46 4.5 Estimation of Percentage PE Costs for Mixed and In-house Projects based on 1371 TxDOT Projects: Zoomed………………………………………………………………..48 6.1 Total Change Orders and Fitted Lines: Log-log Plot...............................................…60 6.2 Total Change Orders and Fitted Lines: Normal Plot………………………………...61 6.3 Total Change Orders and Fitted Lines: Zoomed Plot………………………………..62 6.4 Estimated Change Order Percentage by Project Type…………………………….....63 6.5 Estimated Change Orders by Project Type: Zoomed Plot...........................................64 7.1 PE Cost Differences by District: Log-log Plot……………………………………....73 7.2 PE Cost Differences by District: Normal Plot……………………………………….74 7.3 PE Cost Differences by District: Zoomed Plot……………………………………....75 7.4 Summary of District Change Orders………………………………………………....83 7.5 Summary of District Change Orders………………………………………………....84 7.6 Absolute Change Orders by District: Zoomed Plot………………………………….85 xiv 7.7 Percentage Change Orders by District: All Projects…………………………………86 7.8 Percentage Change Orders by District: Projects Less Than $20m…………………..87 8.1 Total CE Costs for 731 TxDOT Projects- Log-log Plot……………………………..90 8.2 Total CE Costs for 731 TxDOT Projects: Normal Plot……………………………...91 8.3 Total CE Costs for 731 TxDOT Projects: Zoomed Plot………………………….….92 8.4 Estimation of Percentage CE Costs based on 731 TxDOT Projects: Normal Plot..…93 8.5 Estimation of Percentage CE Costs based on 731 TxDOT Projects: Zoomed Plot….94 8.6 Estimation of District CE Costs based on 731 TxDOT Projects: Log-Log Plot…...100 8.7 Estimation of District CE Costs based on 731 TxDOT Projects: Normal Plot…….101 8.8 Estimation of District CE Costs based on 731 TxDOT Projects: Zoomed Plot……102 8.9 Estimation of Percentage District CE Costs based on 731 TxDOT Projects: Normal Plot……………………………………………………………………………………...103 8.10 Estimation of Percentage District CE Costs based on 731 TxDOT Projects: Zoomed Plot……………………………………………………………………………………...104 8.11 Estimation of District CE Costs with consultant involvement: Log-Log Plot…….107 8.12 Estimation of District CE Costs with consultant involvement: Normal Plot…...…108 8.13 Estimation of District CE Costs with consultant involvement: Zoomed Plot….…109 8.14 Estimation of percentage District CE Costs with consultant involvement………..110 8.15 Estimation of percentage District CE Costs with consultant involvement: Zoomed Plot……………………………………………………………………………………...111 1 Chapter 1. Introduction This chapter provides background information, research scope, benefits, limitations and organization of the thesis. 1.1 Background There have been conflicting opinions about cost efficiency of organizations providing preliminary and construction engineering (PE and CE) services for transportation projects. State department of transportations (DOTs) need to hire consultants for various reasons, including variable in-house staff work load, lack of expertise in particular work, and legislated rules about outsourcing. A study conducted by National Cooperative Highway Research Program (NCHRP) in 1999, found that most of the DOTs say that staff shortage is the most important factor influencing the outsourcing decision (NCHRP 1999). Generally, engineering workload in DOTs fluctuates as illustrated in Figure 1.1. It is inefficient to staff up to maximum demand. Instead, there is some suitable level that satisfies needs most of the time, with the excess work being given to consultants. The optimal level of DOT’s internal capability should depend on the relative cost of in-house to outsource work. 2 Figure 1.1: Workload Fluctuations in a State DOT Source: (Hallowell, et al., 2006) According to a survey conducted by U.S. Government Accountability Office (GAO) in 2006, over the period 2001-2006 only twelve DOTs increased their in-house staffs while twenty one DOTs decreased staff even though total work done during the period increased (GAO 2006). To make up the staff deficiency, outsourcing of engineering works to consultants increased in most state DOTs. The National Association of State Highway and Transportation Unions (NASHTU) in 2007 study confirmed that use of consultants has increased in several DOTs. Various agencies, such as AASHTO, NCHRP, and TRB, and state DOTs, have conducted studies on the cost of engineering for in-house and outsourced work. However, most lack clarity in reporting and use different methods for cost estimation. Some use a 3 sample project as a basis for estimation. Such comparisons do not give a realistic picture, since each project is unique. Others use simplistic measures, such as average PE as a percentage of project cost. GAO 2006 study found seven factors that affect outsourcing decision for a project, namely, 1. Loss of in-house staff 2. Variation in workload 3. Specialized skills and equipments 4. Schedule constraint 5. Legal and policy requirements 6. Innovations 7. Cost savings States ranked these factors in order of importance in decision making, and it was found that cost saving is the least important (GAO 2006). Although DOTs say that cost saving is the least important factor, it is the one that is most often cited in arguments over how and what should be outsourced. Therefore a comparison of in-house and consultant cost is appropriate. It can not only provide a robust decision criterion but also affect other outsourcing decision factors in the long run. For example, if a particular project type is always very expensive, state policies can be changed. Moreover, cost of delivering the project needs to be assessed even if other factors are driving the outsourcing decision, especially in the current recession when DOT’s have had to cut budgets and “do more with less”. 4 1.2 Research Scope The primary aim of the work documented here was to compare in-house and consultant PE and CE costs on TXDOT projects. Data was obtained through a research contract between the Center for Transportation Research, The University of Texas at Austin, and TxDOT. Detailed description of data and research methodology are discussed in chapter 4. Four issues are addressed in this analysis: 1. The cost of engineering for projects done with in-house staff compared to using consultant forces. 2. The differences in engineering costs for different project types and across a range of work scopes. 3. The quality of engineering for projects done with in-house staff compared to using consultant forces. 4. The differences in engineering costs across TxDOT districts. 1.3 Benefits 1. Aid in decision making on outsourcing. 2. Identify difference in costs of in-house and consultant works. E.g. for certain district, project type, design task etc. 3. Determine if there are differences in quality of engineering services provided by in-house and consultants. 5 1.4 Limitations 1. Limited period (research was conducted on projects let in fiscal year 2006 and 2007 only). 2. The study was conducted only on TxDOT projects (not all state DOTs). 3. Data may be incomplete or under-recorded. 4. Not project specific but global statistical analysis. 1.5 Organization of the Thesis This thesis is organized in 9 chapters. This chapter provided the research background, scope, benefits and limitations. Chapter 2 is a compilation of various studies done for comparing in-house cost to consultant cost, along with the Project Development Process of TxDOT. Chapter 3 describes the data analyzed and the statistical methodology used. Chapter 4 provides a comparison of in-house PE costs to projects with consultant involvement. Chapter 5 goes in-depth into costs at the design function level. Chapter 6 gives an analysis of the costs of change orders for in-house and consultant projects. Chapter 7 contains the results of cross-district comparisons of PE costs. Chapter 8 provides a comparison of in-house CE costs to projects with consultant involvement. Chapter 9 provides conclusions and recommendations. 6 Chapter 2. Literature review This chapter presents a summary of project development process of TxDOT and various other studies that have been conducted to compare in-house cost to consultant cost for providing engineering services of transportation projects. 2.1 Project Development Process A Project Development Process manual, issued by TXDOT in June 2009, was reviewed to understand the steps involved in preliminary engineering for a transportation projects (TxDOT 2009). Main steps in the process are, 1. Planning and Programming 2. Preliminary Design 3. Environmental 4. Right of way and Utilities 5. PS&E Development 6. Letting A flowchart prepared by Texas Department of Transportation is attached in following page (Link for full size figure can be found in TxDOT’s Project Development Process manual, June 2009). 7 Figure 2.1: Project Development Process (TxDOT 2009) 8 2.1.2 Planning and Programming Need identification for the projects is the first step in planning and programming. A project can be identified in several ways, including recommendation from a state staff for maintenance of existing infrastructure or construction of a new road. After need identification, it is compared with other needed projects considering the funding constraint. Texas Transportation Commission (TTC) approves all the projects before development. Projects are authorized, either as one of three levels- Plan, Develop and Construct or as a feasibility study. A ‘PLAN’ authority for the project should be obtained by submitting a request to Transportation Planning and Programming division. During PLAN authority approval, scheduling and programming manager should obtain a TxDOT control-section-job (CSJ) number from Transportation Planning and Programming Division (TPP) for project development. Project authorization is done in three main steps in sequential order- Prepare cost estimate, Obtain approval of ‘PLAN’ authority and Obtain project specific minute order, if required. To receive sufficient funding, construction and ROW cost estimates are entered into Design and Construction Information System (DCIS). Different types of engineering related tasks are classified in DCIS classification and called as function codes. They are described in more details in chapter 4. The data should be correct in DCIS so that it will reflect correct project funding data in the Financial Information Management Systems (FIMS). After project authorization, study requirements for the project are determined after checking the compliance. 9 Construction funding identification is the last task of this phase which is done after studying all the pertinent requirements. Outside sources, such as local government bodies (other than state and federal), other public entities and private sources are also considered. In summary, Planning and Programming section of project development process includes following task,  Need Identification  Project Authorization  Compliance with planning requirements  Study requirement determination  Construction funding identification 2.1.3 Preliminary Design Key individuals of the projects meet in Design Concept Conference for establishing fundamental aspects, concepts and preliminary engineering design criteria of a project. Design Summary Report (DSR) is used for preparing Design Concept Conference. Next task in preliminary design is obtaining the data necessary making project related decision. After data collection, public input is obtained by public meetings to incorporate those issues which are concerned to public. Some issues are not apparent to designer during data collection so it important to take inputs from public. To check, how well projects meet required performance, a value engineering study may be conducted after preliminary and geometric schematics. Before public 10 hearing, schematics are reviewed by district staff and stakeholders to make sure that design criteria, project needs and commitments are met. In summary, Preliminary Engineering section includes following tasks,  Design concept conference  Data collection/Preliminary Design preparation  Public meetings  Preliminary schematics  Geometric schematics  Value engineering  Geometric schematic approval 2.1.4 Environmental In the early stage of project development, identification of public and environmental concerns for the project is essential. After addressing environmental issues, documentation and public involvement, environmental clearance is required from Environmental Affairs Division or FHWA for the project to advance in the next step of project development. In summary, environmental section includes following tasks,  Preliminary environmental issues  Interagency coordination/permits  Environmental documentation  Public hearing  Environmental clearance 11 2.1.5 Right of way utilities Next step in project development process is finalization of Right of Way which involves data collection, map and property descriptions, right of way appraisals and acquisitions and utility adjustment. Right of way and utility data collection is done by performing preliminary right of way research, obtaining information and locating the existing utilities.Before acquiring property for a project, right of way maps and property descriptions are submitted. Project manager should maintain strong coordination with project engineers and surveyors. Appraisals and right of way acquisitions are done according to state and federal guidelines and utility adjustment is done for owners, affected by proposed work. 2.1.6 Plans, Specifications and Estimates (PS&E) Development Design conference is conducted in first phase of PS&E development. It provides an opportunity for stakeholders and key people to check design criteria, accept or change them and formally endorse decision. As a part of PS&E, following tasks are conducted  Design conference  Begin detailed design  Final alignments/profiles  Roadway design  Operational design 12  Bridge design  Drainage design  Retaining/noise walls & miscellaneous structures  Traffic control  PS&E assembly/design review 2.1.7 Letting PS&E packages including supporting documents are reviewed before TxDOT takes constructing bids. The process is suitable for design-bid-build project delivery. Primary responsibility from PS&E documents review to letting is taken by Design Division. Document review is generally conducted by checking to ensure that the project is authorized for letting, reviewing plans for consistency, checking for completeness of plans, checking for compliance with applicable rules and regulations, ensuring that funding is allocated, and ensuring that other legal requirements are met to satisfy the letting process. 2.2 In-house and Consultant cost comparison Lack of availability of data from DOTs is a major hurdle for most studies and this may be the reason why DOTs adopt cost comparisons based on a specific project. The United States General Accounting office concluded that it is difficult to compare costs of in- house and consultant projects. The challenges in obtaining information are difficulties in finding like projects to compare, using state DOT systems which do not have complete information, and assigning overhead costs to in-house projects (GAO, 2006). 13 Another hurdle is that many DOTs do not have a systematic accounting system. A statewide AASHTO survey conducted in 2006 revealed that states do not have a uniform system for estimating and tracking project costs (AASHTO 2006). Ohio DOT has developed estimates for engineering design cost control, based on the number of hours required to complete all project development activities included in project schedules. Completing a bridge design may require 150 hours of designer effort, 50 hours of drafter effort and 10 hours of checker effort. The cost of a specific activity can be estimated by applying these estimates to a fully loaded labor rate for the specific job classification (e.g. $45/hour for a designer). Likewise a time-staged budget to bring a project to bid can be estimated (AASHTO 2006). Many of the private sector-public sector comparisons fail due to the accounting standard related to the comparison of overheard cost. While the private sector is required to include building and utility expenses, certain states do not include it. Even the levels of management captured by agencies are different. In the calculation of in-house employee hours, administrative or training time is not accounted for in the rate used to compare to the private sector (Warne, et al., 2008). A standard metric is not followed in states for deciding overhead rates and it varies in the range 1.5 to 2.8 in addition to direct costs (Caltrans 2008). Most of the studies do not consider the hidden cost benefits of contracting out. For example seeking consultant’s help may reduce the project time and increase the quality of work. Another point usually missed is that pension-related obligations can be avoided if work is contracted out. Another report for a taxpayers group in California 14 concluded that state component of overhead costs is most often understated while doing comparisons with private sector (CEC White Paper 2008). Some of the studies done towards comparing in-house cost to consultant cost for various departments of transportation, are summarized below (Thai 2009), 2.2.1 Texas State Department of Highways and Public Transportation (SDHPT), 1987 Three studies were conducted by Texas State Department of Highways and Public Transportation (TxDOT’s former name), and all of them concluded that in-house is cheaper than consultant for providing engineering services. 2.2.2 Ernst and Whinney, 1987 Ernst and Whinney analyzed ten pairs of projects. Each pair consists of one consultant and one in-house project. The paired projects were of similar nature. Design cost were measured in three different way; design costs to construction costs ratio, design costs per plan sheet and design costs for every mile of roadway. Variations due to project types were controlled by these ratios. In-house projects were found to be cheaper. However, because of small number of data, statistical analysis could not be performed. 2.2.3 Center for Transportation Research, 1987 The Center for Transportation Research at The University of Texas at Austin found that overhead of consultant projects were 45% higher than in-house projects. Overheads were calculated as a ratio of indirect cost to direct cost. It was also commented that 15 consultant’s salary ranges from 5 to 22% higher than in-house staff. The research methodology was based on global projects that means, all projects were combined rather than paired. CTR concluded that in-house can provide engineering services at cheaper rate. 2.2.4 Texas Transportation Institute, 1987 Texas Transportation Institute (TTI) at Texas A&M University analyzed 18 pairs of projects and found that 15 pairs had less engineering cost for in-house projects. 2.2.5 Legislative Audit Bureau of Wisconsin, 1990 Legislative Audit Bureau of Wisconsin conducted a study for Wisconsin state to examine impact of increased number of consultant involvement in projects done during 1982 to 1989. It concluded that consultant involvements were no longer expensive, mainly due to projects given to consultants were less complex and in-house staffs were inefficient in managing projects. 2.2.6 Professional Services Management Journal, 1992 Federal Highway Administration (FHWA) data from 1979 to 1989 were used to analyze outsourcing effect by Fanning (1992). Data were taken from all fifty states. He stated that design costs were most for those states that outsourced less than 20%. And those who contracted out 50% to 70% of work showed least design to construction cost ratio. This implies that consultant involvement increased the engineering cost. 16 2.2.7 University of California at Berkeley, 1992 UC Berkeley studied 236 projects out of which 32 were consultant and rest in-house. The study was conducted to compare in-house engineering cost with consultant for California Department of Transportation. Average design cost was 15.46% of construction cost for projects done by consultant and 17.76% for in-house projects. That means, there was no significant difference. All projects were analyzed on aggregate basis therefore type, complexity and size factors were not considered. 2.2.8 Alaska Legislative audit, 1994 A legislative audit done in Alaska concluded that the cost of preliminary engineering is less when the PS&E work is completed by consultants (Alaska 1994). 2.2.9 Price Waterhouse Coopers, 1998 In Texas in 1998, Price Waterhouse Coopers came up with the finding that outsourcing was 62% more expensive for 8 of 13 different kinds of design work. 2.2.10 Louisiana State Department of Transportation study, 1999 Louisiana DOT conducted research on road and bridge design projects and found that consultants cost 20% more than in-house. But it was mainly due to extra dollars spent on contract management and supervision by in-house staff. The study also suggested some improvements for comparing in-house to consultant. These include, using same projects for comparison rather than similar and calculating detailed overhead rates which are comparable between in-house and consultant. 17 2.2.11 National Cooperative Highway Research Program, 2001 NCHRP found that, as a percentage of construction work program, the cost of using consultants is between 6.5-7.5% of the construction work program compared to in-house costs of 1.3-2.5% of construction work program, including the costs for managing consultants (NCHRP 2001). The counter argument is that DOTs underestimate in-house costs and report only a fraction of their costs. 2.2.12 California Department of Transportation study, 2007 LECG, LLC, a global expert services company, was contacted by California Department of Transportation for analyzing relative cost to tax payer for utilizing in-house staff and consultant. It was pointed that cost comparison should take several factors in to account such as indirect cost, productivity, capacity utilization and risks. Study found that for fiscal year 2006-2007, amount that state must pay to utilize in-house staff, ranges from $173,434 to $209,212 while for consultant it averages $193,000 (Hamm and Rodini 2007). It also stated that no allowances are made for four important factors, namely, o Less than 100% utilization for CALTRANS staff o The cost of idle capacity when demand is below capacity o Costs incurred by the state as the result of uninsured A/E errors and omissions o Cost resulting from project delay 18 While for consultant projects, contract administration is additional cost. However, mitigation of in-house training, gains from rapid project delivery and tax revenue on contractor profits are deductions from DOT’s cost (Hamm, Rodini, 2007). 2.2.13 Department of Civil Engineering Polytechnic Institute of NYU, 2008 Polytechnic institute of NYU conducted a study to compare in-house design engineers fees with consultant for New York State. It pointed out that even though it appears that design engineers are paid same, state engineers are offered numerous packages such as large paid leave, reduced work week than consultant engineers, etc. Therefore total cost to taxpayers is more for in-house design engineer. The study stated at 80% confidence that in-house design staffs are 14% more expensive than consultants (Griffis and Kwan 2008). 2.2.14 TxDOT survey of in-house and consultant cost, 2008 In late 2008, TxDOT staff conducted a survey of state DOTs on consultant and in-house PE costs as a percentage of construction cost. The next table shows the results of that survey. From this data it can be seen that costs in Texas are estimated to be relatively low compared to other states, and consultant costs are higher than in-house costs for all states. However, the basis for the estimates is not clear. 19 Table 2.2: PE costs in various state DOTs as compiled by TxDOT States 2005 2006 2007 Consult- ant Projects % In-House Projects % All Projects % Consult- ant Projects % In- House Projects % All Projects % Consult- ant Projects % In- House Projects % All Projects % Arkansas 5-8% 5-8% 5-8% California 15.70% 13.90% 16% Indiana 4-5% Kentucky 21.00% 21.00% 21.00% Maine 11.20% 7.29% 9.60% Massachusetts 6-8% 6-8% 6-8% Missouri 5.26% 5.26% Montana 22% 20% 16% Nevada 10.80% 6.50% New 10-15% 5-10% 8-10% 10-15% 5-10% 8-10% 10-15% 5-10% 8-10% New Jersey 11-22% 11-22% 13-23% New Mexico 6-12% 6-12% 6-12% North 5.40% 4.60% 4.90% Ohio 8.62% 5.46% 7.90% 8.62% 5.46% 7.90% 8.62% 5.46% 7.90% Pennsylvania 14-16% South Dakota 3.30% 3.00% 4.69% Tennessee 6.87% 6.79% 5.82% Texas 8.62% 3.43% 7.01% 9.30% 3.22% 6.31% 8.65% 3.18% 5.55% Utah 11.80% 12.80% 11.03% Virginia 10-15% 10-15% 10-15% Wisconsin 7.50% 5.06% 7.48% Wyoming 10% Above table shows that, consultants are more expensive for all the states. Most of the state DOT officials perceive that contracting out engineering services is more expensive than using in-house staff. However, some found opportunities for cost savings in some circumstances for specific activities. For example, the Utah DOT found that contracting out its pavement management data collection work is cheaper because it allowed the department to avoid having to invest in the expensive equipment required, 20 which tends to become obsolete quickly. A study done by a state DOT found that if the agency laid off consultant construction inspectors for at least 3 months yearly, the agency’s cost for the inspectors would equal that of in-house employees. But the state cannot take such a decision due to the concern that the department would not be able rehire them once their services were needed again (GAO 2006). 2.3 Summary The following table summarizes above findings along with some others found by Louisiana Transportation Research Center (Wilmot 1995), Table 2.3: Summary of previous researches on in-house and consultant cost comparison Study Findings No Difference Expensive By Consultants In-house Roy Jorgensen and Associates, 1977  Consultants are 100% more expensive Western Association of State Highway and Transportation officials, 1979  Consultants are more expensive for 11 states (83%) and same for 2 (17%) Maryland Department of Transportation, 1981  Consultants are 80% to 120% more expensive Vermont Department of Transportation, 1986  Consultants are 16% to 240% more expensive Texas State Department of Highways and Public Transportation (SDHPT), 1987  -- Ernst and Whinney, 1987  -- Center for Transportation Research, 1987  -- Texas Transportation Institute, 1987  -- Alabama Department of Transportation, 1989  Consultants are 69% to 100% more 21 expensive North Carolina Department of Transportation, 1990  -- Legislative Audit Bureau of Wisconsin, 1990  -- Michigan Department of Transportation, 1991  Consultants are 33% more expensive Professional Services Management Journal, 1992  -- University of California at Berkeley, 1992  -- Legislative Analyst, California, 1993  -- Missouri Highway and Transportation Department , 1993  Consultants are 31% more expensive Alaska Legislative Audit, 1994  -- Price Waterhouse Coopers, 1998  Consultants are 62% more expensive for 8 of 13 different kinds of design work Louisiana State Department of Transportation study, 1999  Consultants are 20% more expensive NCHRP, 2001  Consultants are more expensive by 4% to 6.2% California Department of Transportation study, 2007  Consultants are more expensive but in some cases they may be cheaper Department of Civil Engineering Polytechnic Institute of NYU, 2008  In-house is 14% more expensive TxDOT survey, 2008  -- Above table shows that, out of 23 studies done in the span of 30 years, 19 found that consultants are more expensive while 2 found that there is no significant difference and only 2 studies say that consultants are cheaper. As discussed previously, more detailed analyses, specific to design tasks for a given project type, are required to make a fair comparison. For this purpose we collected 22 data from TxDOT for fiscal years 2006 and 2007. Chapter 4 describes the data and analysis methodology. 23 Chapter 3. Data Description and Analysis Methodology This chapter provides a description of the data obtained from TxDOT and the methodology used for data analysis. Actual charges for PE and CE for projects let in fiscal years 2006 and 2007 (FY 06-07: September 2005- August 2007) were obtained from TxDOT’s Construction Division and Finance Division. 3.1 Data Summary The next table is a summary of the projects by project type. Construction cost was computed as the sum of contract letting amount plus net change orders. Table 3.1: Summary of TxDOT contracts let in FY 06-07 Project Type Project Type Description No of (Contract Amount + BCF Border Crossing Facility 1 $4,345,638.04 BR Bridge Replacement 236 $608,983,868.40 BWR Bridge Widening Or 55 $222,272,397.38 CNF Convert Non-Freeway To Freeway 7 $301,692,143.84 CTM Corridor Traffic Management 14 $62,471,234.99 FBO Ferry Boat 1 $22,512,000.00 HES Hazard Elimination & Safety 4 $5,240,528.55 INC Interchange (New or 28 $787,298,018.28 LSE Landscape and Scenic 83 $41,463,949.04 MSC Miscellaneous Construction 349 $818,837,999.76 NLF New Location Freeway 1 $67,466,929.41 NNF New Location Non-Freeway 12 $193,373,350.63 OV Overlay 184 $611,568,634.47 RER Rehabilitation of Existing Road 192 $1,013,188,529.29 RES Restoration 50 $167,257,222.79 ROW Right of Way 2 $146,173,826.42 SC Seal Coat 85 $460,855,529.66 SFT Safety Project 311 $1,064,450,294.13 24 SKP SKIP (Exempt from sealing – 6 $8,488,995.93 SRA Safety Rest Area 3 $42,035,563.16 TC Tunnel Construction 1 $165,509.87 TDP Traffic Protection Devices 4 $8,214,080.41 TS Traffic Signal 57 $31,839,098.29 UGN Upgrade to Standards Non- 13 $68,956,309.65 UPG Upgrade to Standards Freeway 12 $186,878,396.43 WF Widen Freeway 14 $825,697,696.07 WNF Widen Non-Freeway 70 $1,049,760,200.79 Total 1795 $8,821,487,945.68 In terms of frequency of project types, the top ten in order are MSC, SFT, BR, RER, OV, SC, LSE, WNF, TS, and BWR. In terms of dollar volume, the top ten in order are SFT, WNF, RER, WF, MSC, INC, OV, BR, SC, and CNF. The analysis will pay particular attention to these project types. Apart from the above set of 1795 projects, TxDOT provided another list of 65 contracts that were tagged as Exceptions. Upon review, it was found that 28 of the Exceptions were also included in the first list. After removing the repeats, there was data on (1795+65- 28=1832) projects. 3.2 Function Codes TxDOT PE charges are collected at the function code level for each design job (Control Section Job (CSJ)). Following are the function codes used by TxDOT for PE accounting. Table 3.2: TxDOT Function Codes for PE Charges Function Code Function Description 102 Feasibility Studies 25 110 Route and Design Studies 120 Social, Economic and Environmental Studies and Public Involvement 126 Donated Items or Services 130 Right -of-Way Data (State or Contract Provided) 145 Managing Contracted or Donated Advance PE Services. Also includes all costs to acquire the consultant contract(s) and services Applicable to advance PE, Function Codes 102 -150. Advance PE are activities in Function Codes 102 through 150. 146 Rework by TxDOT of complete consultant plans on advance PE projects. Advance PE are activities in function codes 102 through 150. 150 Field Surveying and Photogrammetry 160 Roadway Design Controls (Computations and Drafting) 161 Drainage 162 Signing, Pavement Markings, Signalization (Permanent) 163 Miscellaneous (Roadway) 164 Managing Contracted or donated PS & E PE Services. Also includes all costs to acquire the Consultants Contract(s) and Services applicable to PS & E, Function Codes 160 - 190. PS & E PE are activities in function code 160 through 190. 165 Traffic Management Systems (Permanent) 166 Rework By TxDOT Of Completed Consultant Plans on PE & E projects. PS & E PE are activities in function codes 160 through 190. Rework Segment 76 FCs 160-190 for metric conversion. For reworking existing PS&E to metric units on projects already into plan preparation. 169 Donated Items or Services 170 Bridge Design 180 District Design Review and Processing 181 Austin Office Processing (State Prepared P.S. & E.) 182 Austin Office Processing (Consultant Prepared P.S. & E.) 190 Other Pre-letting date Charges, Not Otherwise Classified. 191 Toll Feasibility Studies 192 Comprehensive Development Agreement Procurement 193 Toll Collection Planning 3.3 PE Charges A TxDOT construction contract (Contract Control Section Job (CCSJ)) includes one or more design jobs (CSJ). TxDOT categorizes PE charges as Consultant PE, Indirect PE, or 26 In-house PE costs. The total PE cost for a construction contract was computed as the sum of all PE charges (functions codes 100-199) in all design CSJs that had been combined into the CCSJ for letting. TxDOT provided a status for each contract, namely, Closed (account finalized), Closing (account not finalized), Inactive (pending resolution), and Open (accounts still being charged). The next table is a summary of the PE totals for all 1832 projects. Table 3.3: PE totals for TxDOT construction contracts let in FY 06-07 Project Type Status Total PE Life- to-Date Consultant PE Costs Indirect PE Costs In-house PE Costs Bridge Replacement Closed $10,322,188.38 $6,250,806.64 $598,306.81 $3,473,074.93 Closing $12,759.18 $455.53 $589.94 $11,713.71 Inactive $18,821,527.61 $12,410,483.36 $979,069.73 $5,431,974.52 Open $24,800,985.62 $16,460,358.73 $1,314,432.88 $7,026,194.01 Total $53,957,460.79 $35,122,104.26 $2,892,399.36 $15,942,957.17 Ferry Open $1,708,164.41 $1,649,330.17 $56,294.91 $2,539.33 Total $1,708,164.41 $1,649,330.17 $56,294.91 $2,539.33 Landscape/ Scenic Enhancement Closed $554,047.46 $69,779.42 $26,463.52 $457,804.52 Inactive $131,870.80 $0.00 $8,346.06 $123,524.74 Open $554,111.28 $0.00 $23,825.99 $530,285.29 Total $1,240,029.54 $69,779.42 $58,635.57 $1,111,614.55 Border Crossing Facility Open $173,263.78 $149,377.13 $10,123.55 $13,763.10 Total $173,263.78 $149,377.13 $10,123.55 $13,763.10 ROW Total 0 0 0 0 Seal Coat Closed $1,550,173.83 $30,897.64 $76,515.79 $1,442,760.40 Inactive $90,239.55 $51,891.68 $3,521.63 $34,826.24 Open $357,106.90 $7,109.50 $17,089.29 $332,908.11 Total $1,997,520.28 $89,898.82 $97,126.71 $1,810,494.75 Tunnel Construction Closed $117,895.41 $107,494.68 $3,824.87 $6,575.86 Total $117,895.41 $107,494.68 $3,824.87 $6,575.86 Traffic Protection Devices Inactive $302,362.60 $258,756.60 $10,238.99 $33,367.01 Open $124,283.94 $46,281.62 $5,863.08 $72,139.24 Total $426,646.54 $305,038.22 $16,102.07 $105,506.25 27 Project Type Status Total PE Life- to-Date Consultant PE Costs Indirect PE Costs In-house PE Costs Upgrade to Standards Freeway Closed $1,941,866.88 $248,420.64 $91,545.40 $1,601,900.84 Inactive $462,046.45 $97,572.41 $25,364.97 $339,109.07 Open $2,327,521.29 $540,632.87 $94,157.29 $1,692,731.13 Total $4,731,434.62 $886,625.92 $211,067.66 $3,633,741.04 Bridge Widening or Rehabilitate Closed $1,168,951.09 $499,892.61 $64,010.83 $605,047.65 Inactive $4,479,512.77 $2,943,916.78 $230,447.65 $1,305,148.34 Open $8,460,435.81 $5,602,488.28 $411,675.12 $2,446,272.41 Total $14,108,899.67 $9,046,297.67 $706,133.60 $4,356,468.40 Convert Non- Freeway to Freeway Inactive $2,720,716.36 $1,110,662.82 $135,646.52 $1,474,407.02 Open $16,616,384.64 $12,632,079.53 $1,058,817.58 $2,925,487.53 Total $19,337,101.00 $13,742,742.35 $1,194,464.10 $4,399,894.55 Hazard Elimination & Safety Inactive $72,864.49 $0.00 $3,153.92 $69,710.57 Open $476,817.92 $359,389.82 $21,420.59 $96,007.51 Total $549,682.41 $359,389.82 $24,574.51 $165,718.08 Interchange (New or Reconstruct) Closed $44,215.77 $2,050.00 $1,635.66 $40,530.11 Inactive $372,351.44 $313,569.03 $17,890.23 $40,892.18 Open $40,631,770.11 $24,824,184.80 $2,183,660.95 $13,623,924.36 Total $41,048,337.32 $25,139,803.83 $2,203,186.84 $13,705,346.65 New Location Freeway Open $13,849,319.31 $11,473,259.35 $678,202.81 $1,697,857.15 Total $13,849,319.31 $11,473,259.35 $678,202.81 $1,697,857.15 New Location Non-Freeway Inactive $247,336.48 $68,231.45 $11,270.25 $167,834.78 Open $7,498,294.75 $4,309,574.10 $435,445.36 $2,753,275.29 Total $7,745,631.23 $4,377,805.55 $446,715.61 $2,921,110.07 Overlay Closed $4,195,393.95 $1,156,833.65 $238,627.62 $2,799,932.68 Inactive $2,230,735.86 $259,666.03 $104,886.17 $1,866,183.66 Open $4,478,210.37 $1,026,054.55 $199,302.38 $3,252,853.44 Total $10,904,340.18 $2,442,554.23 $542,816.17 $7,918,969.78 Rehabilitate Existing Roads Closed $6,824,111.73 $3,588,677.13 $367,113.04 $2,868,321.56 Inactive $14,933,771.11 $7,583,143.32 $864,915.64 $6,485,712.15 Open $33,772,199.74 $23,096,092.75 $1,793,997.20 $8,882,109.79 Total $55,530,082.58 $34,267,913.20 $3,026,025.88 $18,236,143.50 All Safety Bond Program Closed $16,792,631.71 $9,228,494.07 $832,079.22 $6,732,058.42 Inactive $5,626,212.63 $2,948,471.92 $281,299.91 $2,396,440.80 Open $28,195,286.22 $18,519,345.25 $1,362,250.32 $8,313,690.65 Total $50,614,130.56 $30,696,311.24 $2,475,629.45 $17,442,189.87 Safety Rest Area Open $3,485,582.31 $1,202,273.24 $204,650.45 $2,078,658.62 Total $3,485,582.31 $1,202,273.24 $204,650.45 $2,078,658.62 Traffic Signal Closed $863,419.79 $409,229.38 $44,193.11 $409,997.30 28 Project Type Status Total PE Life- to-Date Consultant PE Costs Indirect PE Costs In-house PE Costs Inactive $693,931.12 $300,885.33 $33,819.10 $359,226.69 Open $1,762,387.86 $1,098,558.48 $84,993.49 $578,835.89 Total $3,319,738.77 $1,808,673.19 $163,005.70 $1,348,059.88 Upgrade to Standards Non- Freeway Closed $89,962.15 $0.00 $7,327.66 $82,634.49 Inactive $1,436,926.85 $824,652.06 $102,369.70 $509,905.09 Open $5,953,036.71 $2,482,276.20 $379,764.13 $3,090,996.38 Total $7,479,925.71 $3,306,928.26 $489,461.49 $3,683,535.96 Widening Freeway Closed $26,805.55 $0.00 $1,100.39 $25,705.16 Inactive $47,958.75 $0.00 $2,222.35 $45,736.40 Open $38,005,017.58 $21,187,580.12 $1,913,864.64 $14,903,572.82 Total $38,079,781.88 $21,187,580.12 $1,917,187.38 $14,975,014.38 Widening Non- Freeway Closed $1,502,658.65 $739,441.74 $80,767.36 $682,449.55 Inactive $1,670,287.15 $1,159,906.59 $98,296.56 $412,084.00 Open $72,062,639.22 $46,628,380.54 $4,028,239.57 $21,406,019.11 Total $75,235,585.02 $48,527,728.87 $4,207,303.49 $22,500,552.66 Corridor Traffic Management Closed $151,455.26 $102,701.80 $7,012.55 $41,740.91 Inactive $74,702.16 $0.00 $2,586.66 $72,115.50 Open $2,043,176.27 $609,676.76 $92,227.01 $1,341,272.50 Total $2,269,333.69 $712,378.56 $101,826.22 $1,455,128.91 Utility Adjustments Total 0 0 0 0 SKIP (Transp. Enh. Program) Inactive $129,369.22 $91,339.10 $5,546.28 $32,483.84 Total $129,369.22 $91,339.10 $5,546.28 $32,483.84 Restoration Closed $2,664,983.40 $1,140,746.49 $140,681.42 $1,383,555.49 Closing $108,203.59 $49,839.65 $5,695.74 $52,668.20 Inactive $1,955,687.73 $873,914.30 $96,297.37 $985,476.06 Open $3,012,753.23 $1,514,975.48 $149,452.18 $1,348,325.57 Total $7,741,627.95 $3,579,475.92 $392,126.71 $3,770,025.32 Bridge Preventive Mnt Total 0 0 0 0 Bridge Preventive Mnt - Sealed Open $523.49 $0.00 $33.18 $490.31 Total $523.49 $0.00 $33.18 $490.31 Misc Construction Closed $8,893,208.92 $3,669,638.51 $443,042.28 $4,780,528.13 Closing $127,122.82 $0.00 $5,257.56 $121,865.26 Inactive $5,623,854.53 $2,089,295.54 $263,298.34 $3,271,260.65 Open $40,176,737.92 $23,144,170.17 $1,973,505.99 $15,059,061.76 Total $54,820,924.19 $28,903,104.22 $2,685,104.17 $23,232,715.80 Grand Total $470,602,331.86 $279,245,207.34 $24,809,568.74 $166,547,555.78 29 Project Type Status Total PE Life- to-Date Consultant PE Costs Indirect PE Costs In-house PE Costs Total Closed $57,703,969.93 $27,245,104.40 $3,024,247.53 $27,434,618.00 Total Closing $248,085.59 $50,295.18 $11,543.24 $186,247.17 Total Inactive $62,124,265.66 $33,386,358.32 $3,280,488.03 $25,457,419.31 Total Open $350,526,010.68 $218,563,449.44 $18,493,289.94 $113,469,271.30 Out of a total of about $471 million spent on PE for these 1832 projects, about 60% was consultant charges, 35% was in-house charges, and 5% was indirect costs. 3.4 CE Charges The total CE cost for a project was computed as the sum of consultant, in-house, and indirect charges for that CCSJ. The next table is a summary of the CE charges for 1832 projects. Table 3.4: CE totals for TxDOT construction contracts let in FY 06-07 Project Type Status Total CE Life- to-Date Consultant CE Costs Indirect CE Costs In-house CE Costs Bridge Replacement Closed $4,745,065.46 $275,317.28 $226,995.46 $4,242,752.72 Closing $36,323.33 $1,685.22 $1,747.96 $32,890.15 Inactive $8,888,027.21 $244,742.38 $440,348.75 $8,202,936.08 Open $12,765,883.37 $395,417.76 $594,139.99 $11,776,325.62 Total $26,435,299.37 $917,162.64 $1,263,232.16 $24,254,904.57 Ferry Open $57,244.41 $0.00 $1,567.87 $55,676.54 Total $57,244.41 $0.00 $1,567.87 $55,676.54 Landscape/Sce nic Enhancement Closed $1,131,479.29 $6,063.25 $56,423.47 $1,068,992.57 Inactive $348,315.33 $1,162.03 $25,113.09 $322,040.21 Open $1,515,993.20 $736.52 $56,727.41 $1,458,529.27 Total $2,995,787.82 $7,961.80 $138,263.97 $2,849,562.05 Border Crossing Fac Open $37,743.17 $0.00 $1,141.97 $36,601.20 Total $37,743.17 $0.00 $1,141.97 $36,601.20 ROW Open $468,612.53 $75,564.41 $18,197.85 $374,850.27 Total $468,612.53 $75,564.41 $18,197.85 $374,850.27 Seal Coat Closed $8,037,501.96 $117,968.69 $409,581.38 $7,509,951.89 30 Project Type Status Total CE Life- to-Date Consultant CE Costs Indirect CE Costs In-house CE Costs Inactive $151,034.61 $229.41 $7,302.15 $143,503.05 Open $2,778,337.45 $343,305.81 $142,473.11 $2,292,558.53 Total $10,966,874.02 $461,503.91 $559,356.64 $9,946,013.47 Tunnel Construction Closed $8,741.45 $0.00 $308.57 $8,432.88 Total $8,741.45 $0.00 $308.57 $8,432.88 Traffic Protection Devices Inactive $120,394.11 $0.00 $4,561.19 $115,832.92 Open $261,585.20 $0.00 $9,806.03 $251,779.17 Total $381,979.31 $0.00 $14,367.22 $367,612.09 Upgrade to Standards Freeway Closed $1,010,162.66 $5,318.50 $54,493.10 $950,351.06 Inactive $1,268,516.15 $1,035.00 $70,236.57 $1,197,244.58 Open $2,241,948.53 $40,137.06 $97,857.97 $2,103,953.50 Total $4,520,627.34 $46,490.56 $222,587.64 $4,251,549.14 Bridge Widening or Rehabilitate Closed $1,169,123.74 $44,854.99 $56,672.65 $1,067,596.10 Inactive $2,712,701.90 $105,924.62 $127,250.00 $2,479,527.28 Open $4,613,620.55 $223,149.71 $209,413.17 $4,181,057.67 Total $8,495,446.19 $373,929.32 $393,335.82 $7,728,181.05 Convert Non- Freeway to Freeway Inactive $180,796.43 $20,895.95 $6,809.86 $153,090.62 Open $7,174,798.01 $386,528.50 $342,963.47 $6,445,306.04 Total $7,355,594.44 $407,424.45 $349,773.33 $6,598,396.66 Hazard Elimination & Safety Inactive $80,760.59 $0.00 $3,617.90 $77,142.69 Open $268,088.17 $0.00 $11,371.85 $256,716.32 Total $348,848.76 $0.00 $14,989.75 $333,859.01 Interchange (New or Reconstruct) Closed $194,476.66 $1,586.00 $9,502.00 $183,388.66 Inactive $255,978.02 $39,439.28 $13,361.01 $203,177.73 Open $20,560,535.20 $2,074,298.14 $842,496.20 $17,643,740.86 Total $21,010,989.88 $2,115,323.42 $865,359.21 $18,030,307.25 New Location Freeway Open $5,623,182.01 $414,829.32 $262,318.60 $4,946,034.09 Total $5,623,182.01 $414,829.32 $262,318.60 $4,946,034.09 New Location Non-Freeway Inactive $203,890.20 $347.00 $8,233.85 $195,309.35 Open $3,807,571.02 $361,501.85 $174,348.53 $3,271,720.64 Total $4,011,461.22 $361,848.85 $182,582.38 $3,467,029.99 Overlay Closed $9,879,016.40 $931,126.77 $486,782.27 $8,461,107.36 Inactive $6,698,447.10 $251,202.52 $332,630.43 $6,114,614.15 Open $5,472,153.29 $497,338.00 $240,682.83 $4,734,132.46 Total $22,049,616.79 $1,679,667.29 $1,060,095.53 $19,309,853.97 Rehabilitate Existing Roads Closed $7,600,429.12 $216,679.37 $393,435.93 $6,990,313.82 Inactive $13,084,555.98 $765,758.07 $689,500.38 $11,629,297.53 Open $23,452,632.16 $1,473,596.06 $1,162,767.16 $20,816,268.94 Total $44,137,617.26 $2,456,033.50 $2,245,703.47 $39,435,880.29 All Safety Bond Program Closed $12,445,244.99 $333,281.41 $597,429.28 $11,514,534.30 Inactive $5,653,161.47 $97,372.32 $267,891.19 $5,287,897.96 Open $17,969,331.52 $608,947.89 $827,432.69 $16,532,950.94 Total $36,067,737.98 $1,039,601.62 $1,692,753.16 $33,335,383.20 31 Project Type Status Total CE Life- to-Date Consultant CE Costs Indirect CE Costs In-house CE Costs Safety Rest Area Open $1,241,866.30 $15,567.60 $57,997.78 $1,168,300.92 Total $1,241,866.30 $15,567.60 $57,997.78 $1,168,300.92 Traffic Signal Closed $1,150,624.74 $12,684.16 $55,557.54 $1,082,383.04 Inactive $712,329.22 $1,944.72 $33,255.13 $677,129.37 Open $2,232,291.21 $2,120.32 $96,066.58 $2,134,104.31 Total $4,095,245.17 $16,749.20 $184,879.25 $3,893,616.72 Upgrade to Standards Non- Freeway Closed $97,497.67 $0.00 $7,287.62 $90,210.05 Inactive $642,475.84 $5,017.00 $42,084.61 $595,374.23 Open $2,752,876.41 $104,734.26 $142,682.12 $2,505,460.03 Total $3,492,849.92 $109,751.26 $192,054.35 $3,191,044.31 Widening Freeway Closed $312,976.53 $775.54 $10,737.66 $301,463.33 Inactive $509,504.92 $24,373.04 $26,335.23 $458,796.65 Open $15,732,895.63 $1,422,233.59 $613,150.75 $13,697,511.29 Total $16,555,377.08 $1,447,382.17 $650,223.64 $14,457,771.27 Widening Non- Freeway Closed $933,913.07 $65,762.26 $40,109.57 $828,041.24 Inactive $1,296,815.71 $65,381.63 $57,926.03 $1,173,508.05 Open $31,153,631.20 $2,642,633.10 $1,361,795.73 $27,149,202.37 Total $33,384,359.98 $2,773,776.99 $1,459,831.33 $29,150,751.66 Corridor Traffic Management Closed $102,023.43 $0.00 $5,198.88 $96,824.55 Inactive $20,153.40 $5,848.23 $663.96 $13,641.21 Open $1,725,314.80 $33,496.69 $61,602.52 $1,630,215.59 Total $1,847,491.63 $39,344.92 $67,465.36 $1,740,681.35 Utility Adjustments Inactive $26,856.84 $2,995.15 $1,013.82 $22,847.87 Open $534,715.83 $115,077.45 $19,437.30 $400,201.08 Total $561,572.67 $118,072.60 $20,451.12 $423,048.95 SKIP (Exempt from sealing – Transp. Enh. Program Inactive $416,920.51 $133,631.57 $23,475.08 $259,813.86 Total $416,920.51 $133,631.57 $23,475.08 $259,813.86 Restoration Closed $2,785,642.61 $38,127.25 $127,337.56 $2,620,177.80 Closing $202,937.49 $0.00 $8,850.85 $194,086.64 Inactive $1,935,977.43 $97,077.78 $87,046.35 $1,751,853.30 Open $3,831,781.81 $134,155.04 $193,736.05 $3,503,890.72 Total $8,756,339.34 $269,360.07 $416,970.81 $8,070,008.46 Bridge Preventive Mnt - Not Sealed Open $547.24 $0.00 $24.61 $522.63 Total $547.24 $0.00 $24.61 $522.63 Bridge Preventive Mnt - Sealed Open $127,468.34 $0.00 $6,754.77 $120,713.57 Total $127,468.34 $0.00 $6,754.77 $120,713.57 Misc Construction Closed $7,271,676.02 $159,250.29 $359,317.36 $6,753,108.37 Closing $81,419.54 $0.00 $3,507.07 $77,912.47 Inactive $5,702,777.24 $191,265.60 $260,721.55 $5,250,790.09 Open $24,260,747.78 $1,285,612.69 $1,020,487.15 $21,954,647.94 Total $37,316,620.58 $1,636,128.58 $1,644,033.13 $34,036,458.87 32 Project Type Status Total CE Life- to-Date Consultant CE Costs Indirect CE Costs In-house CE Costs Grand Total $302,770,062.71 $16,917,106.05 $14,010,096.37 $271,842,860.29 Total Closed $58,875,595.80 $2,208,795.76 $2,897,170.30 $53,769,629.74 Total Closing $320,680.36 $1,685.22 $14,105.88 $304,889.26 Total Inactive $50,910,390.21 $2,055,643.30 $2,529,378.13 $46,325,368.78 Total Open $192,663,396.34 $12,650,981.77 $8,569,442.06 $171,442,972.51 Out of a total of about $300 million spent on CE for these 1832 projects, about 5% was consultant charges, 90% was in-house charges, and 5% was indirect costs. 3.5 Data Checks Of the 1832 CCSJs summarized, 732 were classified as Closed, with a total construction cost of about $1.5 billion. Another 643 CCSJs are Closing or still Open, with a total construction cost of about $6.2 billion. The researchers made an assumption that all the CCSJs that are Open, Closing, or Closed are valid projects that have already gone to letting, and that the PE charges on those projects are final amounts. Data from Inactive projects was discarded. Thus, there was PE data on (732+643=1375) individual CCSJs. However, 4 of them had zero PE charges, so those were discarded, leaving 1371 projects. When the PE charges on each contract were totaled, it was found that 623 of the projects had no consultant charges associated with them. For this study such projects were classified as Fully In-house projects. The remaining 749 projects have a combination of in-house and consultant charges, and for this study were classified as Mixed projects. There were no projects with zero in-house charges, so there is not a category for Fully Consultant projects (see Chapter 4). Of the 1371 construction contracts 33 selected for analysis of PE charges, 731 were classified as “Closed”, meaning construction was complete. Only Closed projects were selected for analysis of CE charges. Of these 731 projects, 286 had consultant charges and are classified as Mixed, while 446 are fully In-house. The next table is a summary of the projects selected for analysis. Table 3.5: Projects Classified as Fully In-house or Mixed Project Type Projects Selected for CE Analysis Open and Closing Contracts Projects Selected for PE Analysis Closed Contracts Fully In- house CE Mixed CE Total Contracts Fully In- house PE Mixed PE BR 70 36 34 77 147 10 137 FBO 0 0 0 1 1 0 1 LSE 43 41 2 33 76 72 4 BCF 0 0 0 1 1 0 1 SC 65 45 20 13 78 73 5 TC 1 1 0 0 1 0 1 TPD 0 0 0 3 3 2 1 UPG 5 4 1 4 9 5 4 BWR 12 8 4 24 36 5 31 CNF 0 0 0 5 5 0 5 HES 0 0 0 2 2 1 1 INC 1 0 1 23 24 1 23 NLF 0 0 0 3 3 0 3 NNF 0 0 0 6 6 0 6 OV 111 47 64 28 139 116 23 RER 50 25 25 70 120 39 81 SB 160 82 78 99 259 98 161 SRA 0 0 0 3 3 0 3 TS 21 18 3 22 43 27 16 UGN 1 1 0 8 9 2 7 WF 2 1 1 11 13 1 12 WNF 6 1 5 62 68 3 65 CTM 2 2 0 11 13 6 7 RES 26 14 12 15 41 17 24 MSC 155 120 35 117 272 145 127 34 Totals 731 446 285 641 1372 623 749 For in-house work, dollar charges as well as PE hours were provided. The next table is a summary of that data by function code. Table 3.6: Summary of PE Charges by Function Code Func- tion Total PE Life- to-Date Indirect PE Charges Consultant PE Charges In-house PE Charges In-house PE Hours 102 $1,056,099.07 $72,325.32 $582,595.14 $401,178.61 8637 110 $32,268,964.61 $1,888,478.07 $19,682,670.97 $10,697,815.57 227699 111 $0.00 $0.00 $0.00 $0.00 261 117 $14,424.66 $1,115.59 $4,036.39 $9,272.68 288 119 $0.00 $0.00 $0.00 $0.00 25 120 $21,668,916.60 $1,172,520.06 $11,960,191.20 $8,536,205.34 143570 130 $34,220,439.19 $1,939,444.44 $27,517,521.39 $4,763,473.36 105526 140 $0.00 $0.00 $0.00 $0.00 66 145 $4,446,376.34 $255,667.65 $638,710.56 $3,551,998.13 64678 146 $128,382.32 $8,478.05 $0.00 $119,904.27 2857 150 $53,751,613.18 $2,983,834.16 $42,923,276.59 $7,844,502.43 161520 160 $54,414,967.90 $2,887,533.37 $31,425,140.65 $20,102,293.88 447008 161 $32,873,018.57 $1,679,967.52 $24,274,696.84 $6,918,354.21 163878 162 $19,369,025.86 $935,767.85 $11,884,560.74 $6,548,697.27 130515 163 $75,122,028.10 $3,775,165.35 $37,923,406.48 $33,423,456.27 763170 164 $23,511,987.55 $1,179,784.53 $11,124,012.04 $11,208,190.98 182446 165 $4,762,558.08 $224,561.48 $1,579,193.53 $2,958,803.07 50106 166 $204,722.40 $10,167.43 $0.00 $194,554.97 3649 167 $2,179.72 $137.07 $0.00 $2,042.65 70 170 $32,796,772.08 $1,663,301.81 $21,097,046.14 $10,036,424.13 207807 180 $6,501,889.83 $301,190.45 $0.00 $6,200,699.38 118183 181 $2,084,362.49 $97,201.98 $0.00 $1,987,160.51 50409 182 $983,200.64 $43,323.12 $0.00 $939,877.52 23122 183 $1,168.01 $84.17 $0.00 $1,083.84 20 190 $6,293,481.45 $313,383.60 $2,029,565.06 $3,950,532.79 40567 191 $559,474.26 $21,089.70 $538,384.56 $0.00 0 192 $3,537.79 $200.25 $3,337.54 $0.00 0 193 $129,840.07 $6,980.42 $122,859.65 $0.00 0 195 $0.00 $0.00 $0.00 $0.00 2 35 3.6 Data Transforms During analysis, it was noted that the data exhibits log-normal distributions, i.e., a large number of projects have low values of PE and construction costs, and few projects have high values. To reduce modeling error, log transforms were used, i.e., the continuous variables were converted to their base 10 logarithm values. Where a value (e.g., PE cost) was found to be less than 1, it was changed to 1 to get a logarithm value of 0. This technique is commonly used to transform continuous variables. At worst, if the transform is not valid, the statistical relationship would return a coefficient close to 1, indicating there is no log-normal behavior. PE and CE costs were converted to LogPECost and LogCECost. Project construction cost was converted to LogConstructionCost. Project types were designated as binary or switch variables, i.e., a project type is present (value = 1) or absent (value = 0). Districts were similarly designated as binary variables. Multiplicative interaction terms were also introduced to find model relationships that have different slopes for specific binary variables. 3.7 Data Analysis Methodology The objective of the analyses was to determine if there are differences in PE and CE costs for different groups, namely, between in-house and consultant, across project types, according to project cost, or across districts. The statistical technique chosen was stepwise regression. 36 Stepwise regression is a particular type of regression analysis that also yields analysis of variance (ANOVA). After formulation of a general relationship between the dependent variable (PE or CE costs) and a provided set of independent variables, the independent variables are tested iteratively and automatically added to or removed from the model. Variables can be categorical (giving ANOVA), continuous (giving a regression equation), or interaction terms (which are products of other variables). Criteria for adding or removing variables are defined by the F-test. For this analysis, Fin was set at 3.84 and Fout at 2.71, equivalent to a statistical significance above 95% for entry and below 90% for removal. Variables are added iteratively and the partial F values are re-computed. If the significance of an ‘in’ variable falls below Fout it is removed. The process continues until there is no provided variable that can be added or removed. The analysis starts by identifying the provided independent variable with the highest F-value. If none are found, the analysis ends, giving the population mean of the dependent variable as the model estimate. A statistically significant categorical variable indicates that the presence of that variable divides the population, giving the same result as ANOVA. A statistically significant continuous variable indicates a linear relationship, and the intercept and slope of the relationship are calculated. The final model may contain categorical and continuous variables, as well as any of the interaction variables postulated. The coefficients of the variables in the model indicate their relative effect on the dependent variable. 37 This method is able to find the best combination of provided independent variables to estimate the dependent variable, and was used in all the analyses presented later in this report. The SPSS statistical analysis program was used for the computations. 38 Chapter 4. Comparison of Costs for In-house and Mixed Projects This chapter describes the results of a comparison of the cost of PE for projects done entirely in-house by TxDOT to the cost for projects done with consultant involvement. As discussed in the data description earlier, it was found that 623 PE projects had no consultant charges associated with them. For this study such projects were classified as Fully In-house projects. The remaining 749 projects have a combination of in-house and consultant charges, and for this study were classified as Mixed projects. There were no projects with zero in-house charges, so there is not a category for Fully Consultant projects (see Chapter 4). Therefore, this analysis compares 749 Mixed projects to 623 Fully In-house projects, a sufficient sample to determine if there are statistical differences between the groups. 4.1 Initial Comparison of PE Costs The initial model tested was a linear relationship of the form: logPE Cost = (Mixed + In-house Constants) + logConstruction Cost*(Mixed + In-house Coefficients) Mixed was treated as the reference variable. The SPSS results for stepwise regression are: Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 5.357 .019 286.458 .000 In-house -1.096 .028 -.730 -39.466 .000 2 (Constant) 2.321 .120 19.382 .000 39 In-house -.878 .024 -.585 -36.016 .000 LogConstrCost .469 .018 .415 25.566 .000 3 (Constant) 1.239 .154 8.062 .000 In-house 1.467 .222 .977 6.611 .000 LogProjCost .637 .024 .563 26.920 .000 In-H*LogConstrCost -.378 .036 -1.524 -10.628 .000 Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .730 .532 .532 .511833641 .532 1557.571 1 1369 .000 2 .827 .683 .683 .421192951 .151 653.617 1 1368 .000 3 .841 .708 .707 .404949652 .024 112.947 1 1367 .000 4.1.1 Difference between Mixed and In-house PE Costs Intercept At the first step (Model 1), SPSS automatically selected the variable “In-house”, indicating that, above anything else, In-house projects have significantly different PE costs from Mixed projects. For In-house projects the median value of PE cost is 10^(5.357-1.096) = $18,239, with a 95% confidence range of $16,032 to $20,749. For Mixed projects the value is 10^5.357 =$227,510, with a 95% confidence range of $208,449 to $248,313. Thus, PE for the median Mixed project is estimated to be 12.47 times as expensive as the median In-house project. The model has an R-squared value of 0.532 and a high value of significance (p-value = 0.000). 4.1.2 Difference by Project Construction Cost Intercept With stepwise regression, at the second step (Model 2) project construction cost is found to be significant, and the “In-house” coefficient is commensurately changed. For In- house projects PE cost is estimated to be 10^((2.321-0.878)+0.469*LogConstrCost). PE 40 cost for Mixed projects is 10^(2.321+0.469*LogConstrCost). Thus, PE cost increases with increasing project size with a power factor of 0.469, confirming the log-normal distribution. For two projects of identical construction cost, the PE cost of a Mixed project is estimated to be 7.55 times the cost of the In-house project. The model’s adjusted R-squared value increased to 0.683 (p-value = 0.000). 4.1.3 Difference by Project Construction Cost Slope Model 3 finds that the interaction term between provider and project size is significant (“In-H*LogConstrCost”), and the other coefficients are commensurately changed. In other words, the relationship between PE cost and project size is different for each provider. PE cost increases with increasing project size, but with different slopes and intercepts for In- house and Mixed projects. For In-house projects PE cost is 10^((1.239+1.467)+((0.637- 0.378*LogConstrCost). PE cost for Mixed projects is 10^(1.239+0.637*LogConstrCost) For example, for a $1 million project, Mixed PE Cost is estimated at $115,080 compared to In-house PE Cost of $18,197, a factor of 6.32 times. The model’s adjusted R-squared value increased to 0.707 with a p-value of 0.000 (high significance). 4.2 Difference by Project Type The next test was to determine if the PE cost-Construction Cost relationship differs by project type. Project type was treated as a binary variable. The model tested was a linear relationship of the form: 41 logPE Cost = (Mixed + In-house + ProjectType + Interaction Constants) + logConstruction Cost*(Mixed + In-house + ProjectType + Interaction Coefficients) In the SPSS stepwise regression, Project Types were automatically entered in order of significance, the other project types remaining in a pool if their PE costs are not different from each other. The SPSS results are below, with the variables listed in the order they automatically entered the model: Variable Unstandardized Coefficients Standardized Coefficients B Std. Error Beta t Sig. (Constant) 1.193 0.168 7.097 0.000 IH= INHOUSE 1.12 0.257 0.746 4.359 0.000 CONT= LOG (Total Construction Cost) 0.631 0.026 0.558 24.408 0.000 IH*CONT -0.275 0.041 -1.111 -6.66 0.000 SC -1.03 0.166 -0.321 -6.218 0.000 IH*OV -0.384 0.044 -0.143 -8.683 0.000 BR 0.22 0.037 0.091 5.954 0.000 LSE -0.256 0.053 -0.078 -4.853 0.000 IH*MSC*CONT -0.278 0.082 -0.649 -3.407 0.001 IH*SC*CONT 0.086 0.026 0.173 3.278 0.001 IH*MSC 1.291 0.471 0.529 2.743 0.006 MSC 0.124 0.039 0.066 3.157 0.002 INC 0.23 0.078 0.043 2.943 0.003 WF 0.273 0.106 0.037 2.582 0.010 BWR 0.158 0.066 0.033 2.394 0.017 IH*WNF -0.577 0.226 -0.036 -2.561 0.011 WNF 0.125 0.055 0.035 2.265 0.024 UPG 0.237 0.12 0.027 1.972 0.049 42 Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 17 .867 .752 .749 .375045757 .001 3.888 1 1353 .049 IH is the binary variable distinguishing Fully In-house projects from Mixed projects. SC, OV, BR, LSE, MSC, INC, WF, BWR, WNF and UPG are binary variables representing the presence of specific project types as listed earlier. The project types not listed are found to be statistically similar, and will be called “Other Projects”. The multiplicative variables listed are interaction terms found to be statistically significant. For example, the negative value of the IH* CONT coefficient means that the slope of the PE cost-Construction cost relationship for In-house projects is less than the slope for Mixed projects. The t-values and significance of the coefficients are as listed. The Adjusted R- square of the model is 0.749, and the standard error of the estimate is 0.375. The F- significance of the model is 0.000. These numbers indicate that the model is statistically sound and explains almost 75% of the observed variance in PE charges. Therefore, it is seen that there are statistically significant differences in PE costs between Fully In-house projects and Mixed projects, and PE costs for some project types are different from the costs for others. The differences are best illustrated graphically. 4.2.1 Graphical Lines of Fit The next figure shows the fitted lines estimated by the model overlaid on the actual PE data for all projects, validating the postulated log-log model. The labeled lines are for the 43 project types as listed earlier, with the lines for In-house projects labeled with an ‘In’ prefix, and those for Mixed projects with an “Mx” prefix. Each project type line is plotted only for the observed range of project construction cost for that project type and PE provider. Figure 4.1: Total PE Costs for 1371 TxDOT Projects and Fitted Lines: Log-log Plot 44 To give a better sense of the numbers, the fitted lines are shown in the next figure on a standard scale. Each line is plotted only for the observed range of project construction cost for that project type. Figure 4.2: Total PE Costs for 1371 TxDOT Projects and Fitted Lines Because all the In-house projects are comparatively smaller in construction cost than Mixed project, the lines for In-house projects are not easily seen in this plot. The next plot shows the same data zoomed in to the $20 million construction cost range. 45 Figure 4.3: Total PE Costs for 1371 TxDOT Projects and Fitted Lines for In-house and Mixed Projects : Zoomed The graphs indicate that, as project construction cost increases, PE cost also increases, but by a diminishing amount- confirming economies of scale. If a letting program includes multiple small-dollar projects, it will have a higher PE cost rate than one with large projects of comparable total value. Viewed another way, PE output (dollars let per dollar PE cost) must vary depending on size and complexity of the projects being designed. 46 4.2.2 Interpretation of Results These results show that project construction cost, PE provider, and project type account for about 75% of the variance in PE costs. The differences in PE costs among project types can best be seen when the fitted lines are transformed to estimate the percentage PE, i.e., the estimated PE cost from the fitted lines are divided by actual project construction cost and expressed as a percentage. The next figure shows the plots for Mixed and In-house projects. Figure 4.4: Estimation of Percentage PE Costs for Mixed and In-house Projects based on 1371 TxDOT Projects 47 For all project types the percentage PE decreases as project cost increases. WF is the highest line, indicating that Widening Freeways are the most costly PE work. This project type may also include NLF- New Location Freeways and CNF- Converting Non Freeway to Freeway. Next down is UPG- Upgrading Freeways to Standards, followed by INC- Interchanges and BR- Bridge Replacement. Fairly close next are BWR- Bridge Widen/Rehab and WNF- Widen Non Freeway, with MSC- Miscellaneous Construction essentially on the same line. The next line is the pool group, labeled as “Other Mixed”, for project types not identified as statistically different from each other. For example, OV-Overlays are in the pool group. This chart can be interpreted as an indicator of the relative complexity of the various project types, with the more complex ones higher up and the less complex lower down. To see in-house projects more clearly, the same data is plotted for lower contract values only. Note that the scale is different, to show more detail. 48 Figure 4.5: Estimation of Percentage PE Costs for Mixed and In-house Projects based on 1371 TxDOT Projects: Zoomed As before, for all project types the percentage PE decreases as project cost increases. Of Fully In-house project types, UPG- Upgrade Freeway to Standards, is the most costly, followed closed by BR- Bridge Replacement, and BWR- Bridge Widen/Rehab. Below that group is LSE- Landscape projects. Next down are MSC- Miscellaneous Construction, OV-Overlays, and WNF- Widen Non Freeways. Both In- house and Mixed SC-Seal coats are fairly close at the bottom, indicating that this is the 49 cheapest project type. Note that there are no “In-house” lines for the most costly Mixed PE project types, namely WF, NLF, CNF, and INC, because hardly any are done Fully In-house. The following table summarizes the observed construction cost and estimated (fitted line) percentage PE by project type for Mixed and Fully In-house projects. Generally, for Fully In-house projects, the median construction cost and estimated PE percentage are lower. Table 4.1: Observed Construction Cost and Estimated Percentage PE by Project Type Projects Ranges Medians Type No. Construction Cost Est. % PE Constr. Cost Est. % PE In-house BR 10 $123k-$1.748m 18.0-3.3% $472k 7.7% Mixed BR 136 $182k-$144.041m 29.7-2.5% $1.133m 15.1% In-house BWR 5 $276k-$1.849m 9.3-2.7% $384k 7.5% Mixed BWR 30 $372k-$76.821m 19.7-2.8% $2.308m 10.1% Mixed CNF 7 $22.089m-$99.785m 3.0-1.7% $38.311m 2.5% In-house INC 1 - - $18.555m 0.7% Mixed INC 26 $2.411m-$69.908m 11.7-3.4% $23.971m 5.0% In-house LSE 72 $40k-$2.826m 12.4-0.8% $250k 3.8% Mixed LSE 4 $134k-$1.126m 11.1-5.1% $208k 9.5% In-house MSC 144 $49k-$14.492m 25.2-0.1% $455k 3.2% Mixed MSC 124 $60k-$74.904m 35.8-2.6% $1.508m 10.9% Mixed NLF 1 - - $67.467m 2.0% In-house OV 116 $160k-$11.275m 3.8-0.2% $2.022m 0.7% Mixed OV 20 $134k-$9.789m 20.0-4.1% $3.136m 6.3% In-house SC 74 $396k-$18.483m 1.4-0.2% $4.790m 0.4% Mixed SC 5 $1.092m-$8.045m 0.9-0.4% $6.984m 0.4% In-house UPG 5 $718k-$8.331m 6.0-1.2% $5.700m 1.6% Mixed UPG 5 $3.489m-$62.416m 10.4-3.6% $14.774m 6.1% In-house WF 1 - - $394k 9.6% Mixed WF 13 $4.144m-$176.140m 10.6-2.7% $59.365m 4.0% 50 In-house WNF 3 $2.395m-$8.023m 0.6-0.3% $2.704m 0.5% Mixed WNF 59 $1.552m-$82.910m 10.8-2.5% $13.668m 4.8% Other In-house 285 $29k-$22.425m 27.6-0.4% $776m 2.7% Other Mixed 327 $58k-$154.257m 27.2-1.5% $3.390m 6.1% If project construction cost is a proxy for project scope, then Fully In-house projects are smaller in scope than Mixed projects. As seen earlier, the more complex project types are rarely done In-house. Thus, the portfolio of Mixed projects is different from Fully In-house in scope and complexity. In that case, a gross PE percentage comparison is simplistic and misleading, and caution should be exercised in interpreting such numbers from any DOT. Clearly, gross percentage PE depends on the mix of project types and project costs, aside from PE provider. However, in this dataset, for those project types done Fully In-house or Mixed, the statistically estimated percentage PE is always less for In-house projects than for Mixed projects, as illustrated in the graphs. This finding must be qualified with some caveats. Project type and construction cost are not the only measures of PE needs: two projects of the same type and equal construction cost may have entirely different PE requirements. The fact that a project required consultant PE suggests that the in-house staff, for whatever reason, could not do the work. Finally, this analysis uses PE cost data recorded by TxDOT. The timing of this work did not allow for examination of the accuracy of the PE charges recorded for projects. 51 Chapter 5. Direct Comparison of In-house and Consultant PE Costs This chapter presents a comparison of PE costs for In-house and Consultant work at the function code level. In the previously presented Project Type analyses, it was found that In-house projects have lower PE costs than Mixed projects. There are no 100% Consultant projects, so it is not possible to do a direct comparison at the project level. However, since PE costs are tracked at the function code level, it is possible to find projects in which specific functions are recorded as having been done 100% in-house PE or 100% consultant PE, and to do a statistical comparison of those. 5.1 Function Codes with 100% Consultant Charges The next table lists those functions which were found to be done 100% In-house or 100% Consultant, for a number of projects. Table 5.1: Functions that were done 100% in-house or 100% consultant Function Description Total Projects 102 Feasibility Studies 89 110 Route and Design Studies 344 120 Social, Economic and Environmental Studies and Public Involvement 346 130 Right -of-Way Data (State or Contract Provided) 335 150 Field Surveying and Photogrammetry 442 160 Roadway Design Controls (Computations and Drafting) 631 161 Drainage 454 162 Signing, Pavement Markings, Signalization (Permanent) 467 163 Miscellaneous (Roadway) 560 170 Bridge Design 206 52 Functions with a small number of projects were not included, to ensure that the statistical analysis would be valid. The next table shows the total PE cost expended on these functions for the entire dataset, and the computed ‘weight’ of each function out of total PE expenditures. Table 5.2: Weights of Functions Function Total PE $ Weight Construction Cost of Projects 102 1,056,099 0.26% 2,722,515,830 110 32,268,965 7.93% 6,406,203,153 120 21,668,917 5.32% 6,594,209,647 130 34,220,439 8.40% 5,728,101,105 150 53,751,613 13.20% 6,519,670,707 160 54,414,968 13.36% 6,814,693,936 161 32,873,019 8.07% 5,605,781,825 162 19,369,026 4.76% 5,745,882,565 163 75,122,028 18.45% 7,475,603,564 170 32,796,772 8.05% 4,890,816,757 Total 87.81% Of the 29 PE functions tracked by TxDOT, the above 10 functions make up 87.81% of total preliminary engineering cost. These functions cover most of the PE cost and are thus sufficient to draw conclusions about their effects on overall PE costs. 5.2 Analysis and Results As before, to compare In-house PE Costs to the other group (in this case, 100% Consultant functions), stepwise regression was done. ‘In-house’ was used as the switch variable. Project Construction Cost (‘Total contract’) was also submitted to see if it had an effect on PE costs at the function level, as it was found to do at the project level. Each 53 function was analyzed separately, and the results are summarized in the next table. The detailed SPSS outputs are in Appendix A. Table 5.3: Function Code SPSS Results Step 1: Difference of Means Step 2: Difference when Project Cost Effect Taken Into Account Func- tion Cons- tant In- house Consultant/ IH Ratio Cons- tant In- house Total Contract Consultant/ IH Ratio 102 3.870 -1.180 15.14 3.870 -1.180 15.14 110 3.979 -1.057 11.40 1.908 -1.062 0.334 11.53 120 1.309 0 1.00 1.756 -0.498 0.350 3.15 130 3.704 -0.922 8.36 2.231 -0.937 0.237 8.65 150 4.356 -1.027 10.64 2.536 -1.025 0.292 10.59 160 0.389 0 1.00 0.933 -0.660 0.573 4.57 161 4.519 -0.788 6.14 1.022 -0.750 0.539 5.62 162 0.848 0 1.00 1.178 -0.260 0.414 1.82 163 0.537 0 1.00 1.018 -0.512 0.559 3.25 170 4.148 -0.623 4.20 1.420 -0.583 0.415 3.83 The Step 1 portion of the table computed the difference in mean PE costs between 100% In-house work and 100% Consultant work by function code. In every case the Consultant to In-house PE cost ratio was estimated as greater than or equal to 1. Step 2 computed the difference in mean PE costs when project size (Construction Cost) is taken into account. This time, in every case the Consultant to In-house PE cost ratio was estimated as greater than 1. Construction cost is significant in every case except Function 102, i.e., PE costs at the function code level are correlated with project size, increasing at a tapering rate as construction cost increases. It is clear from these results that for the same project size and the same PE function, in-house cost is less than consultant cost, by a factor that ranges from 1.82 for Function 162 (Signing) up to 15.14 for Function 102 (Feasibility Studies). Of course, 54 project construction cost is not a true measure of project size or complexity, and it is also clear that TxDOT hired the consultants because in-house staff was not able to do the work. To compute an overall Consultant to In-house ratio of PE costs, there are multiple approaches. Here one approach is presented, for illustration purposes only. The Consultant/In-house ratios computed at Step 2 above are weighted by the percentages computed for each function in Table 4.2, to arrive at an estimated overall Consultant/In- house ratio. The computation is shown in the next table. Table 5.4: Consultant/In-house Cost Ratios Function Weight Consultant/ In- house Ratio Effect 102 0.26% 15.14 0.0394 110 7.93% 11.53 0.9143 120 5.32% 3.15 0.1676 130 8.40% 8.65 0.7266 150 13.20% 10.59 1.3979 160 13.36% 4.57 0.6106 161 8.07% 5.62 0.4535 162 4.76% 1.82 0.0866 163 18.45% 3.25 0.5996 170 8.05% 3.83 0.3083 Other Functions 12.19% 1.00 (assumed) 0.1219 Total 100.00% 5.4265 These numbers indicate that Consultant PE is about 5.4 times as costly as In- house PE when project size (cost) is controlled for, with the caveats previously discussed. This ratio can be compared to the previously presented results of Mixed to In-house PE, namely: 55  PE for the median Mixed project is 12.47 times as expensive as the median In-house project.  PE cost increases with increasing project size, and for two projects of identical construction cost, the PE cost of a Mixed project is 7.55 times the cost of the In-house project.  PE cost increases with increasing project size, but with different slopes and intercepts for In-house and Mixed projects. For a $1 million project, Mixed PE Cost is estimated at $115,080 compared to In-house PE Cost of $18,197, a factor of 6.32 times. These results differ because of the different assumptions used in the statistical modeling and computations, but they are consistent in detecting a difference in magnitude of PE costs for consultant projects compared to in-house work. The analysis cannot determine if the accounting or record-keeping of in-house costs is accurate. 56 Chapter 6. Quality of In-house and Mixed PE Projects In this chapter, the value of change orders approved during project construction is analyzed to determine if there are any differences in the quality of PE for In-house and Mixed projects as reflected in change orders. A change order can be of positive or negative sign. A positive sign means the client spends more than planned for construction, while a negative sign means the client spends less than planned. Any change order is undesirable, since it affects the client’s ability to manage his larger work program. Positive change orders create deficits or delays, while negative change orders mean money is left over that might have been utilized to build another project. Change orders are generally caused by changes in project scope or design errors. If the project is perfectly scoped during PE, there should not be any change orders during construction due to re-scoping. Similarly, an error-free design should result in zero change orders. Thus, the absolute value of change orders is one indication of the quality of the PE work. Admittedly, there are multiple causes for change orders, so this analysis is at best only indicative of PE quality. 6.1 Change Order Analysis for Different Project Types The change orders in each project were summed first (i.e., negatives and positives could cancel each other). Of the 1371 construction projects studied, 1370 had non-zero change orders. The net value (positive or negative) was called total change orders, and the sign was then deleted to assign an absolute value of change order total for each project, as summarized by project type in the next table. 57 Traffic Signals have the highest change order percentage at 7.28%, followed by Landscaping (6.03%), RER (5.66%), and MSC (5.36%). The variation suggests that project type and project cost may be factors in change orders. Stepwise regression was run using value of change orders as the dependent variable, project cost as an independent variable, and project types as switch variables. Table 6.1: Absolute Value of Change Orders by Project Type Project Type Total PE Total Construction Cost Absolute Change Order Totals Change Orders as % of Constr. BCF $173,264 $4,340,402 $5,236 0.12% BR $36,560,781 $492,979,632 $9,092,843 1.84% BWR $13,486,429 $180,245,906 $5,099,409 2.83% CNF $17,424,537 $302,483,400 $2,825,932 0.93% CTM $2,009,864 $60,663,608 $592,463 0.98% FBO $1,708,164 $22,512,000 $0 0.00% HES $473,632 $3,966,527 $102,876 2.59% INC $46,566,058 $767,225,316 $18,903,151 2.46% LSE $1,066,817 $32,958,958 $1,986,869 6.03% MSC $53,083,978 $664,266,148 $35,583,165 5.36% NLF $4,747,721 $67,299,167 $167,762 0.25% NNF $10,052,820 $185,533,411 $2,872,430 1.55% OV $6,482,558 $399,070,320 $17,264,115 4.33% RER $35,405,718 $681,251,748 $38,536,959 5.66% RES $4,727,968 $125,500,828 $4,819,650 3.84% ROW $7,673,516 $144,225,877 $1,947,949 1.35% SC $1,741,610 $427,983,182 $11,583,735 2.71% SFT $49,055,353 $906,037,924 $30,974,791 3.42% SRA $3,556,839 $42,035,563 $0 0.00% TC $117,895 $165,510 $0 0.00% TPD $70,125 $7,156,452 $236,737 3.31% TS $2,682,635 $25,316,440 $1,842,385 7.28% UGN $3,556,812 $53,526,170 $2,498,003 4.67% UPG $5,976,360 $135,998,603 $1,633,240 1.20% WF $37,795,068 $818,646,592 $8,291,332 1.01% WNF $60,794,086 $1,001,017,297 $22,529,231 2.25% 58 6.1.1 Change Order Analysis Results As before, log transformation of the data was done. Change orders which were less than $1 were changed to $1 for this purpose. Following are the SPSS results at the final step: Model Summary Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 9 .488 .238 .233 1.543355970 .002 4.162 1 1361 .042 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 9 (Constant) -3.446 .433 -7.961 .000 Contract Amount 1.203 .068 .455 17.620 .000 SRA -5.117 .895 -.136 -5.719 .000 FBO -5.398 1.546 -.083 -3.491 .000 BR -.406 .141 -.071 -2.874 .004 SC -.539 .185 -.071 -2.917 .004 I*LSE .542 .200 .069 2.705 .007 I*SFT -.613 .199 -.089 -3.078 .002 SFT .350 .133 .078 2.629 .009 RER .311 .152 .051 2.040 .042 a. Dependent Variable: Absolute Change Order The final model has an R-squared value is 0.238, a relatively low number, meaning the model may not adequately estimate change orders. There is no general coefficient for In-house projects, meaning there is no significant difference in change 59 orders between In-house and Mixed projects. However, some project types are found to differ significantly from the norm. The B coefficients indicate the effect of the significant variables. The 1.203 coefficient (>1) for Contract Amount indicates that, as project construction cost increases, change orders increase at a faster rate. This suggests that larger or more complex projects are likely to see a higher rate of change orders. The negative coefficients for some project types indicate that those are likely to have fewer change orders. Among projects with Mixed PE, Ferry Boat (FBO) and Safety Rest Area (SRA) are likely to have the least change orders. Seal Coat (SC) and Bridge Replacement (BR) with Mixed PE also have lower than average change orders, while Mixed PE Rehabilitation of Existing Roads (RER) and Safety Projects (SFT) have higher than average change orders. Two In-house PE project types differ from the norm: Landscape and Scenic Project (LSE) have higher change order rates, while Safety Projects (SFT) are lower. Essentially, change orders relate more to project type than to PE provider. The following charts illustrate the results. The first chart is absolute change order versus contract amount on a log-log scale with raw data and fitted lines of the project types found to be significantly different from the norm. The second chart is the same data on a normal scale, and the third one is a normal scale chart of the same data but only for smaller contract amounts. 60 Figure 6.1: Total Change Orders and Fitted Lines: Log-log Plot 0 1 2 3 4 5 6 7 8 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 A bs ol ut e C ha ng e O rd er Contract Amount Change Order for different Project Types (log‐Log) Change Order BR SC In‐LSE In‐SFT SFT RER Other 61 Figure 6.2: Total Change Orders and Fitted Lines: Normal Plot 62 Figure 6.3: Total Change Orders and Fitted Lines: Zoomed Plot The charts confirm that change order rates for in-house projects span the same spectrum as those for mixed projects. Project type is the most significant factor in change order rate. Overall, consultant involvement seems to make no difference in the rate of change orders, and by extension, the quality and completeness of PE for different project types. The percentage change order versus total contract amount for selected project types is plotted in the next two figures. 63 Figure 6.4: Estimated Change Order Percentage by Project Type 64 Figure 6.5: Estimated Change Orders by Project Type: Zoomed Plot The first figure is the percentage value of change orders (absolute) versus total contract amount and fitted lines of estimated change order percentage by project types. The second graph is the same plot for low contract amounts (up to $20 million) so the pattern for smaller contracts can be observed. Change order percentage is low for smaller contract amount and increases with increase in contract amount but tends to level off at larger contract values. Based on the assumption stated earlier in this analysis, it is seen that the quality of PE done by In-house teams and those involving consultants (Mixed) as reflected in change orders is not significantly different at 95% confidence level. 65 Chapter 7. Differences in PE Costs across Districts In this chapter, differences in PE costs across TxDOT districts are examined. Every district has unique challenges such as existing infrastructure, traffic demand, staffing, availability of PE consultants, etc., which may influence the types of projects required, construction costs, and PE costs. 7.1 Summary of PE Costs by District As discussed earlier, PE costs for 1371 projects were analyzed. The next table is a summary of number of construction contracts by district, construction totals, and PE costs for the study period (fiscal years 2006-07). Table 7.1: Summary of District Construction Contracts and PE Costs District Number of Contracts Total Construction $ Total PE $ Gross PE % Abilene 37 $149,846,011 $7,005,173 4.67% Amarillo 40 $178,804,155 $3,259,286 1.82% Atlanta 49 $386,631,367 $15,466,124 4.00% Austin 109 $431,660,038 $25,127,681 5.82% Beaumont 63 $319,127,112 $8,831,399 2.77% Brownwood 26 $70,072,105 $3,587,950 5.12% Bryan 47 $197,849,364 $7,960,397 4.02% Childress 22 $75,548,620 $2,741,639 3.63% Corpus Christi 34 $226,263,299 $12,999,692 5.75% Dallas 133 $819,933,768 $39,300,785 4.79% El Paso 27 $136,225,382 $12,597,949 9.25% Fort Worth 93 $468,270,332 $21,916,960 4.68% Houston 141 $1,304,692,034 $59,505,654 4.56% Laredo 26 $133,632,362 $12,371,894 9.26% Lubbock 22 $161,825,224 $7,107,303 4.39% 66 Lufkin 65 $264,661,989 $19,364,324 7.32% Odessa 34 $104,874,569 $3,396,488 3.24% Paris 42 $146,575,270 $8,117,279 5.54% Pharr 46 $318,791,402 $18,792,697 5.89% San Angelo 22 $123,540,484 $5,281,175 4.27% San Antonio 127 $767,737,373 $69,694,296 9.08% Tyler 43 $283,599,356 $12,690,606 4.47% Waco 42 $326,470,207 $17,379,655 5.32% Wichita Falls 45 $156,403,438 $7,133,144 4.56% Yoakum 36 $170,923,109 $5,539,881 3.24% Totals 1371 $7,723,958,370 $407,169,431 5.27% Houston District has the largest number of projects (141) and the largest construction volume for FY 06-07, with a total contract amount of $1.3 billion. Dallas, San Antonio and Austin follow with 133, 127 and 109 construction contracts valued at $819 Million, $768 Million and $432 Million, respectively. Childress, Lubbock and San Angelo did the least number of projects (22 each), while Brownwood had the least construction volume ($70 Million in 26 projects) during FY 06-07. For the projects analyzed, Laredo, El Paso, and San Antonio have the highest % PE, at 9.26%, 9.25%, and 9.08% respectively. Amarillo has the lowest at 1.82%, with Beaumont at 2.77%, and Yoakum and Odessa at 3.24%. This high variation in percentage PE across districts is of concern. However, as was seen in earlier analyses, project type, construction cost, and PE provider could be the reason for this variation. The number and types of projects being done by each district are summarized in the next table. 67 Table 7.2: Summary of District Project Types and Numbers Most districts did BR, LSE, MSC, OV, RER, SC and SFT projects. Very few did CNF, CTM, FBO, HES, NLF, NNF, ROW, SRA, TC, TPD, UGN, UPG and WF projects. Certain types of projects may be driving up some district PE costs. However, there is no obvious indication from the mix of projects as to why Laredo, El Paso, and San Antonio have high PE %. D is tr ic t A bi le ne A m ar ill o A tla nt a A us tin B ea um on t B ro w nw oo d B ry an C hi ld re ss Co rp us C hr is ti D al la s El P as o Fo rt W or th H ou st on La re do Lu bb oc k Lu fk in O de ss a Pa ri s Ph ar r Sa n A ng el o Sa n A nt on io Ty le r W ac o W ic hi ta F al ls Y oa ku m BCF 1 BR 6 1 9 14 9 2 5 3 4 13 20 7 3 8 1 4 3 1 7 4 4 12 7 BWR 5 3 1 1 1 5 2 1 1 1 8 2 1 2 CNF 1 1 1 1 2 1 CTM 1 1 6 5 FBO 1 HES 2 INC 1 1 4 1 2 9 2 1 1 3 1 1 LSE 4 2 1 3 1 1 16 1 2 15 1 3 2 1 3 1 9 2 2 4 2 MSC 3 13 6 24 12 7 6 5 2 37 13 19 45 6 5 5 6 3 3 4 23 10 5 5 1 NLF 1 NNF 1 1 2 2 1 1 1 OV 7 12 2 19 12 2 1 2 4 4 1 3 20 3 1 4 3 2 2 2 13 5 1 9 2 RER 5 6 2 3 7 6 4 6 4 9 1 3 8 3 4 6 3 4 1 6 19 3 4 5 2 RES 2 9 5 2 1 2 2 1 2 3 3 7 ROW 1 1 SC 4 2 3 6 2 2 3 2 3 5 1 2 2 5 4 3 7 2 3 5 2 5 2 4 SFT 2 2 22 34 14 16 3 8 14 2 23 8 5 2 27 6 13 9 21 9 12 5 6 SRA 1 1 1 TC 1 TPD 2 TS 2 2 12 7 2 1 4 3 1 2 5 1 UGN 1 1 2 1 1 1 1 UPG 4 4 1 2 1 WF 1 3 2 5 1 1 1 WNF 2 1 1 1 1 1 7 3 8 3 16 7 5 1 1 2 Totals 37 40 49 109 63 26 47 22 34 133 27 93 141 26 22 65 34 42 46 22 127 43 42 45 36 68 The mix of in-house and consultant work in each district is also of interest. The next table is a summary of that data for projects let in FY 06-07. Table 7.3: Summary of that data for projects let in FY 06-07 District In-house Projects Mixed Projects No. of Pro- jects Construction $ PE Spending Mean PE% No. of Pro- jects Construction $ PE Spending Mean PE% Abilene 20 54,364,324 751,259 1.4% 17 95,481,687 6,253,914 6.5% Amarillo 30 87,647,360 591,523 0.7% 10 91,156,796 2,667,763 2.9% Atlanta 25 75,847,022 793,376 1.0% 24 310,784,345 14,672,749 4.7% Austin 41 67,089,846 694,465 1.0% 68 364,570,192 24,433,216 6.7% Beaumont 32 69,418,805 764,986 1.1% 31 249,708,306 8,066,413 3.2% Brownwood 18 27,104,825 712,428 2.6% 8 42,967,280 2,875,522 6.7% Bryan 31 52,288,738 1,001,074 1.9% 16 145,560,626 6,959,323 4.8% Childress 10 26,678,658 163,137 0.6% 12 48,869,962 2,578,502 5.3% Corpus Christi 13 40,065,345 589,317 1.5% 21 186,197,954 12,410,375 6.7% Dallas 44 77,212,112 1,005,438 1.3% 89 742,721,657 38,295,347 5.2% El Paso 10 29,142,435 362,178 1.2% 17 107,082,947 12,235,771 11.4% Fort Worth 32 68,877,569 869,514 1.3% 61 399,392,763 21,047,446 5.3% Houston 77 164,717,748 3,267,078 2.0% 64 1,139,974,285 56,238,576 4.9% Laredo 5 7,012,167 41,847 0.6% 21 126,620,195 12,330,047 9.7% Lubbock 14 59,586,810 502,768 0.8% 8 102,238,414 6,604,535 6.5% Lufkin 26 36,473,261 530,097 1.5% 39 228,188,728 18,834,228 8.3% Odessa 31 75,867,340 1,154,977 1.5% 3 29,007,229 2,241,511 7.7% Paris 20 46,093,842 652,336 1.4% 22 100,481,428 7,464,944 7.4% Pharr 18 39,190,907 547,991 1.4% 28 279,600,495 18,244,706 6.5% San Angelo 15 47,215,288 713,667 1.5% 7 76,325,196 4,567,508 6.0% San Antonio 34 61,778,687 1,081,211 1.8% 93 705,958,686 68,613,085 9.7% Tyler 23 106,394,558 751,559 0.7% 20 177,204,798 11,939,047 6.7% Waco 17 38,587,663 542,559 1.4% 25 287,882,544 16,837,096 5.8% Wichita Falls 24 67,271,141 472,907 0.7% 21 89,132,297 6,660,237 7.5% Yoakum 12 35,942,957 236,590 0.7% 24 134,980,152 5,303,290 3.9% Odessa (91%) had the highest percentage of In-house projects followed by Amarillo (75%) and Brownwood (69%) while Laredo (81%) has done the highest percentage of Mixed projects followed by San Antonio (27%) and Dallas (33%). In every case the mean PE % for Mixed projects is far higher than that for In-house projects. Looking 69 specifically at Laredo, El Paso, and San Antonio, it is seen that these districts have the highest Mixed PE %, but their In-house PE % are normal. 7.2 Simple Comparison of Districts In this analysis at the district level, PE costs are estimated as functions of project construction cost. District is introduced as a switch variable to allow comparisons across districts. The following is the final SPSS result: Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 10 (Constant) .451 .148 3.052 .002 Total Contract Amount .691 .023 .611 29.469 .000 San Antonio .379 .055 .147 6.929 .000 Amarillo -.359 .092 -.081 -3.888 .000 Fort Worth .235 .063 .079 3.744 .000 El Paso .316 .112 .059 2.836 .005 Dallas .162 .053 .064 3.029 .003 Laredo .305 .114 .056 2.687 .007 Pharr .203 .087 .049 2.345 .019 Lufkin .154 .074 .044 2.097 .036 Waco .188 .090 .043 2.079 .038 Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df 1 df2 Sig. F Change 10 .651 .424 .420 .5697475 .002 4.323 1 1360 .038 The model shown has an adjusted R-squared value is 0.420, a reasonable number, meaning the model adequately estimates differences across districts. The B coefficients 70 indicate the effect of the significant variables. The 0.691 coefficient for Contract Amount indicates that, as project construction cost increases, PE cost increases at a tapering rate. The districts not shown are the statistically selected pool (not necessarily the average districts). Eight districts have PE costs above the pool. When the effect of Contract Amount (project size) is separated, San Antonio has the highest PE costs at 2.39 times the pool, followed by El Paso at 2.07, Laredo at 2.02, Fort Worth at 1.72, Pharr at 1.60, Waco at 1.54, Dallas at 1.45, and Lufkin at 1.43. Amarillo is lower than the pool, at 0.44 times the remaining districts. 7.3 Comparison of Districts Considering PE Provider Previous analysis found that PE costs are a function of PE provider (In-house or Mixed). In this section, PE provider and district are introduced to determine whether the differences in PE costs across districts are due to the choice of PE provider. As before, PE cost is the dependent variable while project construction cost and the interaction variable for provider and project cost were selected as continuous independent variables. Switch variables were District, Provider, and the interaction term for district and provider. Following are the SPSS results of this analysis. Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 13 (Constant) 1.211 .149 8.140 .000 In-house 1.458 .218 .971 6.699 .000 Total Contract Amount .639 .023 .565 27.919 .000 IH*CONTRACT -.365 .035 -1.474 -10.489 .000 San Antonio .143 .037 .056 3.877 .000 71 In*Austin*Cont -.051 .011 -.068 -4.708 .000 Amarillo -.238 .064 -.054 -3.749 .000 In*Childress*Cont -.066 .021 -.045 -3.129 .002 Yoakum -.190 .066 -.041 -2.857 .004 In*Wichita Falls*Cont -.038 .013 -.041 -2.832 .005 In*Lufkin*Cont -.501 .136 -.533 -3.685 .000 In*Lufkin 2.716 .793 .495 3.424 .001 In*Laredo*Cont -.080 .029 -.039 -2.762 .006 El Paso .183 .076 .034 2.396 .017 Model Summary Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df 1 df2 Sig. F Change 13 .854 .729 .726 .3915863 .001 5.739 1 1357 .017 In the table, an ‘In’ or ‘IH’ prefix indicates in-house projects for that district (different intercepts) and the ‘Cont’ suffix indicates interaction with total contract amount (different PE-construction cost slopes). The final model has an adjusted R-squared value is 0.726, a large number, meaning the model strongly estimates PE costs. However, comparing this figure to the R-square value of 0.420 for districts only, it is clear that PE provider is a major factor in PE costs across all districts. The B coefficients are listed in order of entry into the model. For example, the switch variable In-house entered first, indicating that PE provider is the strongest indicator of differences in PE costs. Next, Total Contract Amount (project construction cost) entered, saying that project size is the next strongest factor in PE costs. Thirdly, the interaction variable IH*Contract entered, indicating that there are major differences in the sizes of projects done In-house versus Mixed. Then District variables began to enter. 72 A positive B coefficient for a district indicates that that district’s PE costs are higher even when project size is accounted for. Now it is seen that El Paso has the highest factor for Mixed projects at 0.183 (equivalent to multiplier of 1.52 or 52% greater than the pool districts’ Mixed projects). San Antonio’s Mixed projects factor is 0.143 (=1.39 or 39% greater than the pool. Amarillo’s Mixed projects come in lowest at -0.238, or 58% of the pool, with Yoakum’s at 65%. Austin, Childress, Wichita Falls, and Laredo show a lower slope for In-house projects compared to the pool. It means that, as project size increases, their in-house PE costs increase more slowly than the pool districts. For Lufkin, In-house PE costs decrease as project size increases, an aberration. The results are best illustrated graphically, in the next 3 charts. 73 Figure 7.1: PE Cost Differences by District: Log-log Plot 74 Figure 7.2: PE Cost Differences by District: Normal Plot 75 Figure 7.3: PE Cost Differences by District: Zoomed Plot In every district in-house projects have less PE cost than mixed projects for the usual range of project size, consistent with the results of previous analyses. All districts have fairly similar in-house PE costs that increase with project size. However, there are large differences in the costs of Mixed projects across districts, with El Paso, San Antonio, and Bryan being higher than average, and Amarillo, Yoakum, and Beaumont being lower than average. 76 7.4 Comparison of Districts Considering PE Provider and Project Type Earlier analyses showed that project type is an important factor in predicting PE costs both for In-house and Mixed projects. In this section, project type is introduced along with district and PE provider as switch variables. As before, PE cost is the dependent variable, project construction cost in the continuous independent variable, and interaction terms among the predictor variables are also tested. The SPSS results are: Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 28 (Constant) 1.234 .150 8.239 .000 In-house 1.417 .246 .944 5.769 .000 Total Contract Amount .642 .022 .567 28.768 .000 IH*CONTRACT -.314 .039 -1.265 -7.957 .000 SC -1.204 .160 -.375 -7.540 .000 OV -.485 .045 -.194 -10.841 .000 BR .095 .038 .039 2.526 .012 San Antonio .209 .034 .081 6.074 .000 LSE -.473 .058 -.145 -8.219 .000 I*MSC*Cont -.270 .079 -.631 -3.427 .001 I*SC*Cont .095 .025 .191 3.770 .000 In*Houston .113 .046 .035 2.443 .015 El Paso .194 .071 .036 2.745 .006 I*MSC 1.192 .457 .489 2.609 .009 Amarillo -.181 .060 -.041 -3.042 .002 In*Austin -.186 .060 -.042 -3.092 .002 In*Lufkin*Cont -.464 .125 -.493 -3.711 .000 In*Lufkin 2.360 .733 .430 3.220 .001 Childress -.182 .078 -.031 -2.349 .019 Yoakum -.130 .062 -.028 -2.094 .036 I*TS*Cont -.058 .015 -.059 -3.953 .000 RER -.162 .039 -.062 -4.157 .000 SFT -.145 .032 -.077 -4.513 .000 RES -.211 .063 -.047 -3.346 .001 I*WNF -.618 .211 -.039 -2.932 .003 Lufkin .169 .060 .048 2.811 .005 In*Wichita Falls*Cont -.028 .012 -.030 -2.222 .026 77 In*Laredo*Cont -.093 .030 -.045 -3.128 .002 Laredo .220 .080 .040 2.758 .006 Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 28 .881 .776 .772 .3573719 .001 7.605 1 1342 .006 In the table, an ‘In’ or ‘IH’ prefix indicates in-house projects for that district (different intercepts) and the ‘Cont’ suffix indicates interaction with total contract amount (different PE-construction cost slopes). The final model has an adjusted R-squared value is 0.772, an improvement over the district-PE provider analysis figure of 0.728, meaning that there is a measurable difference among districts even after project type and size is accounted for. The B coefficients are listed in order of entry into the model. As before, the switch variable In-house entered first, indicating that PE provider is the strongest indicator of differences in PE costs. Next, Total Contract Amount (project construction cost) entered, saying that project size is the next strongest factor in PE costs. Thirdly, the interaction variable IH*Contract entered, indicating that there are major differences in the sizes of projects done In-house versus Mixed. Then three Mixed project types (SC, OV, and BR) entered, indicating that these project types have different costs compared to other project types, across all districts. A positive B coefficient for a district indicates that its PE costs are higher even when PE provider, project type and size are accounted for. Now it is seen that Laredo has the highest factor for Mixed projects at 0.22 (equivalent to multiplier of 1.66 or 66% 78 greater than the pool districts’ Mixed projects). San Antonio’s Mixed projects factor is 0.209 (=1.62 or 62% greater than the pool), and El Paso’s Mixed projects factor is 0.194 (1.56 or 56% greater than the pool). Childress and Amarillo’s Mixed projects come in lowest at -0.182, or 66% of the pool, with Yoakum’s at 74%. Houston has an intercept higher than the pool for in-house projects, meaning that its in-house costs are somewhat above typical (130% of pool), while Austin has an intercept lower (65% of pool). Wichita Falls and Laredo have slopes slightly lower than the pool for in-house work, meaning their in-house PE costs do not increase as quickly as other districts when project size increases. Lufkin displays unusual behavior, with PE costs decreasing as project size increases, perhaps due to the influence of a few unusual projects. This analysis verifies that in-house projects have less PE cost than mixed projects for the usual range of project size across all districts and all project types. Most districts have fairly similar in-house PE costs that increase with project size. However, there are large differences in the costs of Mixed projects across districts, with Laredo, San Antonio, and El Paso being higher than average, and Childress, Amarillo, and Yoakum being lower than average. The reasons for the differences across districts are not clear. Perhaps they have a higher involvement of historically underutilized consultants, but that data was not available for this analysis. 79 7.5 Change Orders in Districts In this section, differences in change order rates across districts are analyzed to determine if any districts have unusual levels of change orders. 7.5.1 Change Order Rates As before, the net value of change orders was computed for each project, and the sign was deleted to create an absolute value. The totals of absolute change orders for each district are shown in the following table. Laredo has the highest change order rate at 7.43%, followed by Childress and Austin. Yoakum has the lowest rate at 1.28%, followed by Wichita Falls and Paris. Table 7.4: Summary of District Change Orders District Contract Amount Absolute Change Orders Percentage Change Orders Abilene $146,674,578 $3,782,458 2.58% Amarillo $174,104,715 $4,967,967 2.85% Atlanta $382,089,574 $8,290,990 2.17% Austin $409,878,723 $22,612,635 5.52% Beaumont $314,138,802 $10,865,128 3.46% Brownwood $68,716,323 $1,390,456 2.02% Bryan $196,901,400 $3,518,140 1.79% Childress $71,580,027 $3,968,594 5.54% Corpus Christi $222,331,836 $4,148,136 1.87% Dallas $781,314,898 $40,099,985 5.13% El Paso $136,180,179 $4,304,108 3.16% Fort Worth $459,070,345 $15,231,286 3.32% Houston $1,288,364,624 $21,065,924 1.64% Laredo $125,403,926 $9,321,739 7.43% Lubbock $158,702,539 $3,657,547 2.30% Lufkin $258,520,074 $7,495,187 2.90% Odessa $104,301,316 $2,567,504 2.46% Paris $145,063,504 $2,133,260 1.47% Pharr $314,257,079 $5,370,686 1.71% 80 San Angelo $123,859,762 $4,464,287 3.60% San Antonio $746,772,641 $23,912,983 3.20% Tyler $279,450,741 $5,025,190 1.80% Waco $321,804,259 $6,750,265 2.10% Wichita Falls $155,142,641 $2,280,269 1.47% Yoakum $169,595,813 $2,165,540 1.28% Totals $7,554,220,319 $219,390,264 2.90% 7.5.2 Comparison of Change Orders across Districts For analysis, absolute values of change orders for each project were chosen as the dependent variable. ‘In-house’ and ‘District’ were used as switch variables. Project construction cost (‘contract amount’) and interactions with in-house and district are the continuous independent variables. Stepwise regression was run as before, and the SPSS results are shown next. Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 15 (Constant) -3.284 .405 -8.107 .000 Contract Amount 1.198 .064 .453 18.713 .000 Wichita Falls -1.755 .333 -.177 -5.267 .000 In*Paris*Cont -.259 .056 -.108 -4.630 .000 Yoakum -1.064 .258 -.097 -4.124 .000 Pharr -.913 .230 -.093 -3.971 .000 Corpus Christi -.960 .265 -.085 -3.619 .000 In*Childress -1.484 .483 -.072 -3.069 .002 In*Dallas .711 .236 .071 3.016 .003 Laredo .817 .302 .063 2.705 .007 Lubbock -.815 .328 -.058 -2.486 .013 In*Houston 4.536 2.112 .593 2.148 .032 81 In*Wichita Falls*Cont .165 .074 .075 2.228 .026 Odessa -.569 .266 -.050 -2.138 .033 Bryan -.455 .227 -.047 -1.999 .046 In*Houston*Cont -.693 .351 -.545 -1.976 .048 Model Summary Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 15 .515 .265 .257 1.51877468 .002 3.904 1 1355 .048 In the table, an ‘In’ prefix indicates in-house projects for that district (different intercepts) and the ‘Cont’ suffix indicates interaction with total contract amount (different PE-construction cost slopes). The final model has an R-squared value is 0.257, a small number, meaning the model is not a strong estimator of PE costs. However, it does find that there are statistically significant differences in change orders for some districts compared to others. The B coefficients are listed in order of entry into the model. For example, Contract Amount (project construction cost) entered first, saying that project size is the strongest factor in change orders. Then District variables began to enter. The B coefficient for Contract Amount is 1.198, indicating that, as project size increases, change orders increase at a faster rate. There is no B coefficient for In-house, indicating that there is no statewide difference in change order rates for in-house projects versus Mixed projects. However, there are differences among districts. 82 For Mixed projects, Laredo has a rate of 6.56 times the pool (i.e., districts not named here). Several districts are far below the pool: Wichita Falls being the lowest at 0.018 times the pool, Yoakum at 0.086, Corpus Christi at 0.11, Pharr at 0.122, Lubbock at 0.153, Odessa at 0.27, and Bryan at 0.351. For in-house projects, the picture is more muddled. Only Childress stands out on the low side, being at 0.033 times the pool. Dallas stands out on the high side, being at 5.14 times the pool, and with a higher slope even than for Mixed projects. As Dallas’ project size increases, change orders for in-house projects increase at a faster rate than for Mixed projects, and for other districts. Wichita Falls ‘compensates’ for its Mixed projects by having a similar steeper rate as Dallas for in-house projects. Paris and Houston have flatter slopes than the pool for in-house projects. The following figures show these results graphically. 83 Figure 7.4: Absolute Change Orders by District: Log-log Plot 84 Figure 7.5: Absolute Change Orders by District: Normal Plot 85 Figure 7.6: Absolute Change Orders by District: Zoomed Plot Percentage change orders by district are shown in the next two charts, 86 Figure 7.7: Percentage Change Orders by District: All Projects 87 Figure 7.8: Percentage Change Orders by District: Projects Less Than $20m This analysis shows that most of the districts (18 out of 25) have similar change order rates for In-house and Mixed projects. The remaining 7 districts have different results, perhaps due to unique project conditions. Overall, it is seen that quality of preliminary engineering done by in-house and mixed teams, as measured by absolute value of change orders, is not significantly different across districts. 88 Chapter 8. CE Results 8.1 Difference in CE cost by Project Types This chapter presents the results of a statistical analysis of Construction Engineering (CE) costs in TxDOT. A stepwise regression analysis in the SPSS statistical analysis program for CE charges on 731 projects produced the following equation: Log (Total CE Cost) = 0.314 + 0.737 Log (CONT) + 0.269 TS + 0.134 BR + 0.112 LSE – 0.157 OV – 0.214 SC Coefficient table and model summary of SPSS results are shown below, Coefficients 7 (Constant) .314 .108 2.899 .004 Total Contract .737 .018 .943 40.786 .000 OV -.157 .026 -.127 -6.150 .000 SC -.214 .033 -.138 -6.448 .000 TS .269 .055 .099 4.903 .000 BR .134 .030 .088 4.393 .000 LSE .112 .040 .060 2.840 .005 IBWRCONT .036 .014 .049 2.515 .012 Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .823 .677 .677 .252544708 .677 1528.961 1 729 .000 89 2 .831 .690 .690 .247483054 .013 31.125 1 728 .000 3 .842 .708 .707 .240330625 .018 44.977 1 727 .000 4 .846 .715 .714 .237707459 .007 17.134 1 726 .000 5 .849 .721 .719 .235403239 .006 15.282 1 725 .000 6 .851 .724 .722 .234367373 .003 7.423 1 724 .007 7 .852 .726 .724 .233509923 .002 6.327 1 723 .012 CONT is the total construction cost plus change orders, and Log is the base 10 logarithm of the numbers. TS, BR, LSE, OV, SC and in-house BWR are binary variables representing the presence of specific project types, namely Traffic Signals, Bridge Replacement, Landscaping, Overlays, Sealcoats and in-house Bridge Widening or Rehabilitation. The coefficients are all significant at the 0.012 level or better. The Adjusted R-square is 0.724, and the standard error of the estimate is 0.234. These numbers indicate that there is no significant difference between CE costs for fully in- house projects compared to mixed projects, but there are differences in CE costs for specific project types. 8.1.1 Graphical Lines of Fit The next figure shows the fitted lines overlaid on the actual data for all projects. A log- log plot is used to show the linear relationships and the estimation of the lines of fit. 90 Figure 8.1: Total CE Costs for 731 TxDOT Projects- Log-log Plot The same data is shown in the next figure on a standard scale. 91 Figure 8.2: Total CE Costs for 731 TxDOT Projects: Normal Plot 92 Figure 8.3: Total CE Costs for 731 TxDOT Projects: Zoomed Plot 8.1.2 Interpretation of Results These results show that project construction cost and project type account for about 72% of the variance in CE costs. The difference among project types can best be seen when the fitted lines are transformed to estimate the percentage CE, i.e., the estimated CE cost from the fitted lines are divided by project construction cost and expressed as a percentage. The next figure shows the plot. 93 Figure 8.4: Estimation of Percentage CE Costs based on 731 TxDOT Projects: Normal Plot 94 Figure 8.5: Estimation of Percentage CE Costs based on 731 TxDOT Projects: Zoomed Plot The line for “Other” represents all project types except those named. The line for Traffic Signals is highest, with Bridge Replacement and Landscape also higher but nearer to “Other”. Sealcoats are lowest, with Overlays slightly closer to “Other”. The line for each project type is plotted only for the range of project costs observed for that project type. For all project types the percentage CE decreases as project cost increases. Traffic Signals are statistically the most expensive in CE costs, with CE costs ranging from over 20% for low-dollar projects, to 10% for projects in the $1 million 95 range. Bridge Replacements are the next highest, ranging from around 15% for very small projects, to around 5% for $3 million projects. Landscaping projects are very close in cost to BR, ranging from around 16% for small projects to around 6% as projects approach $2 million. Sealcoats are statistically the least expensive in CE costs, with CE costs ranging from around 4% for minor sealcoats to 1.5% for those around $16 million. Overlays are the next lowest, ranging from around 7% for very small projects, to around 2% for $10 million projects. The “Other Projects” category is a single line because no statistical difference in CE costs was found among the remaining project types. For those projects let and completed in FY 06-07, the fitted line for percentage CE costs ranges from around 12% for the smallest projects, to around 3% for those in the $15 million range. It must be stressed that these results are based only on those project completed in the two year period. Naturally, those are smaller and simpler projects. There was insufficient data on larger and more complex project types to determine if they are statistically more or less expensive, so they are lumped in the “Other” category. Also, there were insufficient completed projects with consultant CE involvement (445 out of 731 projects were in-house projects) to draw any statistical conclusions about differences in their CE costs. 96 8.2 Difference in CE Cost across Districts In this section, differences in CE costs across TxDOT districts are examined. Every district has unique challenges such as terrain, traffic, staffing, availability of CE consultants, etc., which may influence construction costs, project types, and CE costs. As discussed earlier, 731 projects were analyzed. The following table is a summary of number of construction contracts by district, construction totals, and CE costs. Table 8.1: Summary of construction totals and CE cost by districts District Number of Projects Total Contract Amount Total CE Abilene 22 49,607,988.71 2,284,617.25 Amarillo 31 88,720,482.55 2,617,282.27 Atlanta 22 45,890,514.20 1,547,026.85 Austin 64 91,137,500.20 4,619,818.14 Beaumont 43 72,763,298.05 2,417,592.86 Brownwood 15 21,686,885.93 1,380,752.00 Bryan 35 56,253,705.03 2,246,730.76 Childress 16 42,879,638.63 1,402,575.53 Corpus Christi 21 45,829,118.56 1,770,253.88 Dallas 53 80,683,274.85 4,305,284.49 El Paso 11 31,216,789.01 1,475,724.30 Fort Worth 45 77,436,539.33 3,600,534.28 Houston 58 74,655,806.83 3,670,291.17 Laredo 18 37,449,943.17 1,699,311.66 Lubbock 10 28,176,321.76 830,350.38 Lufkin 39 84,552,059.04 3,077,905.13 Odessa 25 57,422,884.22 2,336,138.97 Paris 25 48,455,128.99 2,463,658.49 Pharr 17 52,506,402.63 1,654,971.64 San Angelo 9 23,154,298.26 1,066,970.56 San Antonio 52 97,023,148.14 3,899,516.46 Tyler 26 64,938,970.62 2,006,032.93 Waco 13 21,379,318.56 813,596.20 Wichita Falls 32 63,891,687.16 2,144,900.76 Yoakum 29 92,868,296.42 3,152,817.93 97 Austin district has highest number of projects (64) while San Antonio has highest construction volume ($97 Million) spent FY 2006-2007 followed by Yoakum ($92 Million), Austin ($91 Million), Amarillo ($88 Million) and Lufkin ($84 Million). San Angelo district did least amount of projects (9) during these period. We also note that, Houston district did 58 projects but total contract amount is less than $75 Million. However many other district’s such as Yoakum (29 projects), San Antonio (52 projects), Lufkin (39 projects), Amarillo (31 projects) etc. total spending on construction is more than Houston, even though number of projects done, are less. This is because projects completed in 2 years time period is usually small ($28k - $16M) and Houston district most of small projects FY 06-07. Following table shows CE spending on In-house and Mixed Projects, and CE percentage of total construction contract amount. Table 8.2: CE spending on in-house and mixed projects District In-house Projects Mixed Projects No. of Proj- ects Total Construc- tion CE Spending Mean CE% No. of Proj- ects Total Construc- tion CE Spending Mean CE% Abilene 20 $45,206,244 $2,187,765 8.2% 2 $4,401,745 $96,852 2.2% Amarillo 28 $71,258,736 $1,980,950 6.2% 3 $17,461,747 $636,332 3.7% Atlanta 21 $44,872,332 $1,373,966 4.2% 1 $1,018,182 $173,061 17.0% Austin 9 $4,896,479 $358,503 7.0% 55 $86,241,022 $4,261,315 6.2% Beau- mont 19 $22,357,283 $789,540 5.0% 24 $50,406,015 $1,628,053 4.0% Brown- wood 11 $10,403,756 $726,738 15.9% 4 $11,283,130 $654,014 9.2% Bryan 14 $17,682,706 $829,404 8.7% 21 $38,570,999 $1,417,327 5.0% Childress 15 $30,849,103 $1,209,468 4.7% 1 $12,030,536 $193,108 1.6% Corpus Christi 8 $15,434,979 $695,709 7.8% 13 $30,394,140 $1,074,545 5.5% Dallas 41 $34,041,244 $2,226,384 8.5% 12 $46,642,031 $2,078,901 7.1% 98 El Paso 7 $13,472,320 $652,639 6.0% 4 $17,744,469 $823,085 6.7% Fort Worth 29 $45,445,542 $1,775,810 15.6% 16 $31,990,997 $1,824,724 10.3% Houston 23 $12,605,223 $765,968 7.0% 35 $62,050,583 $2,904,323 6.5% Laredo 7 $15,253,090 $659,619 5.8% 11 $22,196,853 $1,039,692 7.3% Lubbock 10 $28,176,322 $830,350 5.3% 0 --- --- --- Lufkin 31 $54,210,888 $2,014,075 7.3% 8 $30,341,171 $1,063,830 3.7% Odessa 24 $48,236,522 $2,115,598 14.0% 1 $9,186,362 $220,541 2.4% Paris 25 $48,455,129 $2,463,658 9.6% 0 --- --- --- Pharr 6 $5,391,603 $144,255 7.7% 11 $47,114,800 $1,510,716 4.6% San Angelo 7 $16,733,568 $896,320 10.1% 2 $6,420,730 $170,651 3.0% San Antonio 13 $14,227,568 $524,000 5.5% 39 $82,795,580 $3,375,517 5.0% Tyler 26 $64,938,971 $2,006,033 5.8% 0 --- --- --- Waco 10 $6,103,594 $277,433 6.3% 3 $15,275,725 $536,164 3.4% Wichita Falls 26 $48,536,727 $1,493,676 7.3% 6 $15,354,960 $651,225 5.9% Yoakum 15 $20,536,596 $918,172 10.1% 14 $72,331,701 $2,234,646 3.4% The above table shows that most of the smaller districts did more in-house CE than mixed CE. However some bigger districts such as Houston, Austin and San Antonio used significant consultant help for construction engineering. Lubbock, Paris and Tyler districts did all CE in-house. Average CE % across all districts for in-house projects was 8%, while for mixed projects it was 5.6%. 8.2.1 Difference in District Means The analysis was done in two stages. In the first stage, only the difference in the district means was calculated. In the second stage, district means and slopes of the CE cost- construction cost relationship were computed. The following is the SPSS result for the district means: 99 Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 8 (Constant) .744 .097 7.705 .000 Total Contract .659 .016 .842 41.269 .000 Beaumont -.136 .038 -.072 -3.549 .000 Fort Worth .172 .038 .093 4.526 .000 Brownwood .247 .063 .079 3.894 .000 Odessa .170 .050 .070 3.438 .001 Paris .143 .049 .059 2.897 .004 Dallas .092 .035 .054 2.641 .008 San Angelo .164 .081 .041 2.024 .043 Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .823 .677 .677 .25254470 .677 1528.961 1 729 .000 2 .828 .686 .685 .24940317 .008 19.481 1 728 .000 3 .831 .691 .690 .24728081 .006 13.550 1 727 .000 4 .834 .696 .695 .24541573 .005 12.092 1 726 .001 5 .837 .700 .698 .24399396 .004 9.486 1 725 .002 6 .839 .703 .701 .24297298 .003 7.106 1 724 .008 7 .840 .706 .703 .24204531 .003 6.560 1 723 .011 8 .841 .708 .704 .24152864 .002 4.097 1 722 .043 These results show that 7 districts are different from the pool when a fixed CE cost-construction cost slope is assumed. The coefficients for the districts indicate that Brownwood district has the highest CE cost followed by Fort Worth and Odessa. The following charts show fitted lines for CE cost for various districts. 100 Figure 8.6: Estimation of District CE Costs based on 731 TxDOT Projects: Log- Log Plot 101 Figure 8.7: Estimation of District CE Costs based on 731 TxDOT Projects: Normal Plot 102 Figure 8.8: Estimation of District CE Costs based on 731 TxDOT Projects: Zoomed Plot Beaumont district has a lower CE percentage than average while all other significantly different districts have higher CE percentage. Brownwood has the highest CE percentage. Fort Worth, Odessa, Paris and San Angelo have very close CE percentage. The following figure shows the pattern of CE percentage for different contract amounts. 103 Figure 8.9: Estimation of Percentage District CE Costs based on 731 TxDOT Projects: Normal Plot 104 Figure 8.10: Estimation of Percentage District CE Costs based on 731 TxDOT Projects: Zoomed Plot The charts show that CE percentage is high for small contract amount and it decreases as the contract amount increases. This district result is similar to the project type analysis. Brownwood has the highest CE percentage while Beaumont has the lowest. According to the project type analysis, TS and BR projects have the highest CE percentage but the Brownwood district did not do those projects in significant number (only 1 BR project). There is no clear explanation for Brownwood’s high CE percentage. 105 8.2.2 Difference by CE Provider Consultant involvement can affect CE percentage of any district. To find this, interaction of total contract amount with Provider as a switch variable was submitted. Total construction engineering cost was selected as dependent variable while in- house, districts and interaction of in-house and districts were selected as switch variables. Total contract amount and interaction of in-house and contract amount were selected as independent variables. Stepwise regression was run to see how preliminary engineering cost is affected after considering interaction of districts with total contract amount and CE provider. The SPSS result of this analysis is shown below. Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 10 (Constant) .827 .099 8.380 .000 Total Contract .650 .016 .832 40.762 .000 Beaumont -.139 .038 -.074 -3.670 .000 Fort Worth .178 .037 .096 4.751 .000 Brownwood .261 .063 .083 4.164 .000 In*Odessa .210 .051 .084 4.155 .000 In*Paris .182 .050 .074 3.648 .000 IH*CONTRACT -.012 .003 -.077 -3.647 .000 Dallas .111 .035 .065 3.187 .001 In*San Angelo .274 .091 .060 3.002 .003 In*Abilene*Cont .024 .009 .053 2.603 .009 Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 10 .846 .716 .712 .23854392 .003 6.775 1 720 .009 106 In the table, ‘In’ and ‘IH’ prefix implies in-house projects for that district (different intercept) and ‘Cont’ suffix implies interaction with total contract amount (different slopes. In the coefficient table, the in-house variable (binary variable for in-house project) is not significant, however, interaction of in-house project with total contract amount (IH*Contract) is significant. Because it has negative coefficient, in-house projects are found to have less CE percentage than mixed projects. However, for 4 districts, namely, Odessa, Paris, San Angelo and Abilene, in-house projects have higher CE percentage than average. In-house projects of San Angelo district have the highest coefficient, followed by Brownwood district, while Beaumont has the lowest. The following are the graphical representations of the results. 107 Figure 8.11: Estimation of District CE Costs with consultant involvement: Log- Log Plot 108 Figure 8.12: Estimation of District CE Costs with consultant involvement: Normal Plot 109 Figure 8.13: Estimation of District CE Costs with consultant involvement: Zoomed Plot The above charts show that Brownwood district has the highest CE percentage even though its coefficient is lower than that of in-house San Angelo. This is because total contract amount has positive coefficient and average contract amount of Brownwood district is more than San Angelo. We also note that, in-house projects have less CE percentage than mixed projects for all the districts. The following charts are the graphical representation of CE percentage for various districts. 110 Figure 8.14: Estimation of percentage District CE Costs with consultant involvement 111 Figure 8.15: Estimation of percentage District CE Costs with consultant involvement: Zoomed Plot The charts show that CE percentage decreases with increase in contract amount. In-house and mixed projects both decrease at about the same rate. Similar to previous results, in-house projects exhibit slightly less CE percentage than mixed ones. This analysis shows that in-house projects have less construction engineering cost than mixed projects, for a given district and contract amount. 112 8.3 Chapter Conclusion This chapter presented the results of a statistical analysis of PE and CE costs for TxDOT projects let for construction in FY06 and 07. The time span of the data was limited in order to reduce the effect of inflation. Data on 1832 construction projects was obtained from the Construction Division and the Finance Division of TxDOT. Of these 1832 projects, CE charges on 731 completed projects were analyzed separately. Stepwise regression analysis was performed in the SPSS statistical analysis program. The independent variables introduced were project construction cost, project type (binary variable), and PE provider (binary variable: Fully In-house or Mixed), as well as interaction terms for project type*construction cost and provider type*construction cost. The CE cost analysis found that project construction cost and project type account for about 72% of the variance in CE costs. For all project types the percentage CE decreases as project cost increases. In terms of relative CE cost, the project types were found to rank as follows, from most to least costly: Traffic Signals, Bridge Replacement, Landscape, Other, Overlays, and Sealcoats. Analysis found that there was no significant difference between in-house and mixed projects. In district level analysis it was found that in-house projects are significantly different from mixed projects in CE cost. In-house projects have lower CE percentage than mixed projects for any district and contract amount. There were some anomalies in the results. For example, average percentage CE of in-house was 8% while for mixed, it 113 was 5.6% but regression analysis at 95% confidence showed that, district wise in-house CE percentage is lesser; while for project types, it is not significantly different. It must be stressed that the CE results are based only on those project let and completed in FY 06-07. Naturally, those are smaller and simpler projects. There was insufficient data on larger and more complex project types to determine if they are statistically more or less expensive in CE. 114 Chapter 9. Conclusions Department of Transportations’ (DOTs) face big challenge in determining optimum level of outsourcing for engineering services within the state. In-house capacity, work load, lack of expertise and legislated rules are the important factors that drive the outsourcing decision, however, cost of providing these services is the one which is most often cited in arguments. Various studies have been conducted but lack of availability of data from DOTs is a major hurdle for most studies and this may be the reason why DOTs adopt cost comparisons based on a specific project. To address this issue, data for large number of projects with function code level details, was collected from TxDOT, through a research contract with the Center for Transportation Research at the University of Texas at Austin. This report presented the results of a statistical analysis of PE and CE costs for TxDOT projects let for construction in FY06 and 07. The time span of the data was limited in order to reduce the effect of inflation. Four issues are addressed in this analysis: 1. The cost of engineering for projects done with in-house staff compared to using consultant forces. 2. The differences in engineering costs for different project types and across a range of work scopes. 3. The quality of engineering for projects done with in-house staff compared to using consultant forces. 4. The differences in engineering costs across TxDOT districts. 115 Data on 1832 construction projects was obtained from the Construction Division and the Finance Division of TxDOT. Of these 1832 projects, PE charges on 1371 projects in construction or completed were selected for analysis, and CE charges on 731 completed projects were analyzed separately. There were 26 different project types. Each project type’s preliminary engineering data is available at the function code level. However, even though there are many projects done entirely with in-house staff, no projects done entirely by consultants were found, so PE provider was designated as Fully In-house or Mixed. Stepwise regression analysis was performed in the SPSS statistical analysis program. The independent variables introduced were project construction cost, project type (switch variable), PE provider (binary variable: Fully In-house or Mixed), District (switch variable for 25 TxDOT districts), as well as interaction terms for project type*construction cost and provider type*construction cost. 9.1 PE Costs The PE cost analysis found that project construction cost, PE provider, and project type account for about 75% of the variance in PE costs. For all project types the percentage PE decreases as project cost increases- confirming economies of scale. If a letting program includes multiple small-dollar projects, it will have a higher PE cost than one with large projects. Viewed another way, PE output (dollars let per dollar PE cost) must vary depending on size and complexity of the projects being designed. 116 The analysis found that there are statistically significant differences in PE costs between Fully In-house projects and Mixed projects. For all project types, the statistically estimated PE percentage for Fully In-house projects is lower than for Mixed projects. These results were presented in graphs. In gross terms, PE percentages for In-house and Mixed projects are 1.29% and 6.20% respectively for the full set of projects studied. However, in most cases the construction cost of the median In-house project is smaller than that of the median Mixed project. If construction cost is a proxy for project scope, then Fully In-house projects are smaller in scope than Mixed projects. To test a direct comparison between in-house and consultant costs, the data was analyzed at the function code level. The ten most used functions, which make up to 88% of the total preliminary engineering cost, were analyzed. It was found that, by one methodology, consultant PE costs were 5.2 times as high as in-house PE charges. 9.2 Differences in Project Types, Districts, and PE Quality In terms of relative costliness, the project types were found to rank in the following order from most to least costly: Widen Freeway (including New Location Freeway and Convert Non-Freeway to Freeway), Upgrade Freeway to Standards, Interchange, Bridge Replacement, Bridge Widen/Rehab, Widen Non-Freeway, Miscellaneous Construction, Other, Landscape, Overlays and Sealcoats. When PE costs across TxDOT districts were compared, it was found PE provider (in-house or mixed) is still the largest factor in PE cost differences. The next most important factor is project size as measured by construction cost. After these two are 117 taken into account, differences among project types emerge. But ultimately it was found that Mixed projects in Laredo, El Paso, San Antonio had higher PE percentage than average while Childress, Yoakum and Amarillo had lower. It is speculated that the high cost districts may have higher involvement of historically underutilized businesses, but that data was not available for this analysis. To compare the quality of PE on mixed projects to in-house projects, the absolute value of change orders on each project was analyzed. It was found that there is no significant difference in change order rates for in-house and mixed projects. 9.3 CE Costs The CE cost analysis found that project construction cost and project type account for about 72% of the variance in CE costs. For all project types the percentage CE decreases as project cost increases. In terms of relative CE cost, the project types were found to rank as follows, from most to least costly: Traffic Signals, Bridge Widening and Rehabilitation, Bridge Replacement, Landscape, Other, Overlays, and Sealcoats. Analysis found that there was no significant difference between in-house and mixed projects. Similar to PE analyses, district level analysis was conducted for CE charges. It was found there was significant difference between in-house and mixed projects across districts. Average in-house projects exhibit less CE percentage than average mixed projects. However, in-house projects of San Angelo, Odessa and Fort Worth showed more CE percentage than mixed projects of those districts respectively. Percentage CE in all the districts and for all the project types decreases with increase in contract amount. It 118 ranged from 2% to 23 % of the total contract amount but more than 90% projects had CE percentage less than 15%. CE results are based only on those project let and completed in FY 06-07. Naturally, those are smaller and simpler projects. There was insufficient data on larger and more complex project types to determine if they are statistically more or less expensive in CE. 9.4 Recommendations Project type and size are not the only measures of PE needs: two projects of the same type and equal construction cost may have entirely different PE requirements. The fact that a project required consultant PE suggests that the in-house staff, for whatever reason, could not do the work. Moreover, the more complex project types are rarely done In- house. Thus, the portfolio of Mixed projects is different from Fully In-house in scope and complexity. In that case, a gross PE percentage for each class is a simplistic and misleading measure, and caution should be exercised in interpreting such numbers from any DOT. Even when project type, size and PE provider are taken into account, El Paso, Laredo, and San Antonio districts were found to have higher than typical PE costs for Mixed projects. It is not clear why this is the case, and further research is recommended to clarify whether this was a one-time phenomenon for the specific data studied, or whether there are unique conditions contributing to higher costs in these districts. 119 Throughout the analyses, the differences found in costs between in-house and mixed projects were consistent and large, so much so as to raise speculation for the reasons. While this statistical analysis cannot uncover the reasons, it does bring into question the accuracy of the in-house charges. A further line of inquiry would be to compare the total PE charges in each district to the number of full-time staff working on PE, to determine how much of their time is actually charged to projects. 120 Appendix A Regression results of Function Code PE Cost analysis Function 102, Model Summary Mod el R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .517 .268 .259 .71900595 .268 31.800 1 87 .000 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 3.870 .192 20.141 .000 In-house -1.180 .209 -.517 -5.639 .000 Function 110, Model Summary Mod el R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .457a .209 .207 .80765490 .209 90.316 1 342 .000 2 .493b .243 .238 .79134983 .034 15.238 1 341 .000 121 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 3.979 .100 39.723 .000 In-house -1.057 .111 -.457 -9.503 .000 2 (Constant) 1.908 .540 3.537 .000 In-house -1.062 .109 -.459 -9.744 .000 Total (Contract) .334 .085 .184 3.904 .000 Function 120, Model Summary Mod el R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .272 .074 .071 .64483414 .074 27.417 1 344 .000 2 .302 .091 .086 .63978397 .017 6.452 1 343 .012 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 1.309 .401 3.268 .001 Total Contract .344 .066 .272 5.236 .000 2 (Constant) 1.756 .435 4.040 .000 Total Contract .350 .065 .276 5.365 .000 In-house -.498 .196 -.131 -2.540 .012 122 Function 130, Model Summary Mod el R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .517a .268 .266 .755368435 .268 121.761 1 333 .000 2 .535b .286 .282 .747067292 .018 8.441 1 332 .004 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 3.704 .063 58.426 .000 In-house -.922 .084 -.517 -11.035 .000 2 (Constant) 2.231 .511 4.366 .000 In-house -.937 .083 -.526 -11.317 .000 Total Contract (Log) .237 .082 .135 2.905 .004 Function 150, Model Summary Mode l R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .584 .341 .340 .697860361 .341 228.008 1 440 .000 123 2 .614 .377 .374 .679735412 .035 24.778 1 439 .000 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 4.356 .053 82.095 .000 In-house -1.027 .068 -.584 -15.100 .000 2 (Constant) 2.536 .369 6.867 .000 In-house -1.025 .066 -.583 -15.466 .000 Total Contract (Log) .292 .059 .188 4.978 .000 Function 160, Model Summary Mode l R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .360 .130 .128 .798278848 .130 93.615 1 629 .000 2 .475 .226 .224 .753367843 .096 78.229 1 628 .000 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) .389 .374 1.041 .298 Total Contract (Log) .576 .060 .360 9.675 .000 2 (Constant) .933 .358 2.604 .009 Total Contract (Log) .573 .056 .358 10.201 .000 In-house -.660 .075 -.311 -8.845 .000 124 Function 161, Model Summary Mode l R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .494 .244 .242 .694201199 .244 145.955 1 452 .000 2 .632 .399 .396 .619698107 .155 116.216 1 451 .000 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 4.519 .047 95.446 .000 In-house -.788 .065 -.494 -12.081 .000 2 (Constant) 1.022 .327 3.126 .002 In-house -.750 .058 -.470 -12.847 .000 Total Contract (Log) .539 .050 .394 10.780 .000 Function 162, Model Summary Mode l R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .342 .117 .115 .744578923 .117 61.773 1 465 .000 2 .377 .142 .138 .734884002 .025 13.350 1 464 .000 125 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) .848 .358 2.368 .018 Total Contract (Log) .440 .056 .342 7.860 .000 2 (Constant) 1.178 .365 3.228 .001 Total Contract (Log) .414 .056 .322 7.437 .000 In-house -.260 .071 -.158 -3.654 .000 Function 163, Model Summary Mode l R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .419 .175 .174 .714769871 .175 118.501 1 558 .000 2 .469 .220 .218 .695515544 .045 32.322 1 557 .000 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) .537 .320 1.681 .093 Total Contract (Log) .565 .052 .419 10.886 .000 2 (Constant) 1.018 .322 3.159 .002 Total Contract (Log) .559 .050 .415 11.083 .000 In-house -.512 .090 -.213 -5.685 .000 126 Function 170, Model Summary Mode l R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .306 .093 .089 .916318803 .093 21.032 1 204 .000 2 .394 .156 .147 .886560629 .062 14.925 1 203 .000 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) 4.148 .111 37.331 .000 In-house -.623 .136 -.306 -4.586 .000 2 (Constant) 1.420 .714 1.987 .048 In-house -.583 .132 -.286 -4.426 .000 Total Contact (Log) .415 .107 .250 3.863 .000 127 Bibliography NCHRP. April 1999. NCHRP synthesis 277. National Cooperative Highway Research Program. Hallowell, and David Rogge. December 2006. Evaluation of Outsourcing in the Oregon Department of Transportation-Final Report. GAO. 2006. Increased Reliance on Contractors Can Pose Oversight Challenges for Federal and State Officials Source, U.S. General Accounting Office. URL http://www.gao.gov/new.items/d08198.pdf TxDOT. June 2009. Project Development Process Manual of Texas Department of Transportation from Mark Marek, P.E. Revised June 2009. AASHTO. 2006. Managing Project Cost and Quality. AASHTO - SCOQ National Survey. AASHTO. 2007. AASHTO subcommittee document, Strategies for Coping with Construction Project Staffing Demands. Warne, Tom, et al. 2008. A National Assessment of Transportation Strategies and Practices: Lessons for California. Tom Warne and Associates report. URL http://www.celsoc.org/userdocuments/File/warne-report.pdf CEC White Paper, 2008. Steve Stagner. Texas Council of Engineering Companies (CEC) Thai, Khi. 2009. International Handbook of Public Procurement. Public Administration and Public Policy/146 Alaska 1994. Alaska Division of Legislative audit, Highway Design Cost and Quality Comparison. URL http://www.legaudit.state.ak.us/pages/audits/1995/4472.htm#summary NCHRP. 2001. NCHRP web document 39. Innovations in Private Involvement in Project Delivery. National Cooperative Highway Research Program. William, Hamm and Mark, Rodini. April 9, 2007. Cost to Tax Payers of Obtaining Architectural and Engineering Services: State Employees vs. Private Consulting Firms. 128 Griffis, F.H. and Kwan, Chi Ming. October 30, 2008. NYSDOT Engineering Design Cost: In-house versus Outsourced Design. URL http://www.acecny.org/PDF/PolyStudyFinal.pdf Wilmot, Chester. June 1995. Investigation into the cost effectiveness of using consultants versus in-house staff in providing engineering services for Louisiana’s Department of Transportation and Development. Technical Assistant Report No. 3. Louisiana Transportation Research Center. 129 VITA Prakash Singh was born in Charaiya village, Uttar Pradesh, India on January 14, 1983 to his parents Harihar Prasad Singh and Neha Singh. After completing his high school at Government Inter College, Pratapgarh, he entered National Institute of Technology, Durgapur, India in July, 2002 to pursue undergraduate degree in Civil Engineering. He received his Bachelor of Engineering Degree in June, 2006. Before coming to graduate school of The University of Texas at Austin in August 2008, Prakash worked as a senior engineer in Larsen & Toubro Limited, India from July, 2006 to July, 2008. During his graduate study at The University of Texas at Austin, he also worked for Center for Transportation Research as a graduate research assistant. Permanent Address: HN 20, Rani Sati Lane, Indira Nagar Mumbai, India 400097 This thesis was typed by the author.