Design and build : solid ground or sinking sand? : the impact of delivery methods on financial risk and return of infrastructural projects

(1)

Design & Build: solid ground or sinking sand?

The impact of delivery methods on financial risk and

return of infrastructural projects

David Bogaerds | Student Number: 10409998 | January 22th, 2015 MSc. Accountancy & Control | Specialisation Control

Faculty of Economics and Business |UvA | dr. ir. S.P. van Triest Word count: 15314, 0

(2)

Abstract

Over the last decade, an increasing number of infrastructural projects in the Netherlands are contracted based on Design and Build delivery methods. This thesis investigates the financial impact for the contractor of the usage of a DB delivery method compared to the traditional DBB method by empirically comparing the financial results of 353 infrastructural projects. The impact of delivery method, client, working company, delivery year and project size were tested using both ANOVA’s and linear multiple regression on three variables: (1) pre-calculated versus final profit (2) most positive versus most negative forecasted profit during execution and (3) financial return as % of the revenue. This thesis shows that the financial return of DB project is higher than for DBB projects, but that the financial risk (difference between pre-calculation and final profit and forecasted profit bandwidth during execution) is also higher. It was also shown that both client and working company played an important role in the financial outcome, and that there was interplay between delivery method and these factors.

Acknowledgements

Special thanks to the large construction company that was so kind to collaborate in my research project. I would also like to acknowledge the assistance and valuable suggestions of dr. ir. S.P. van Triest of the University of Amsterdam. Also I would like to thank my sweet girlfriend Suzanne for all the moral support and motivational speeches.

Statement of Originality

This document is written by student David Bogaerds who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1 Introduction

The construction industry is one of the biggest industries in the Netherlands and has a major impact on the Dutch economy. With a yearly production of more than 50 billion Euros, the industry contributes approximately 6% to the Dutch GDP (Gross Domestic Product) and with more than 400,000 employees working in the construction industry, also 6% of the working population in the Netherlands works in construction (Bouwend Nederland, 2014).

An interesting study in the United Kingdom shows that the construction industry as a whole is underachieving in comparison to other industries (Egan, 1998). One of the biggest problems is the low and unreliable profit margin, but the industry also invests too little in capital, research, development and training. Furthermore, too many clients are dissatisfied with the industry’s overall performance. The same is stated in a Dutch report which also points at the fact that between 6% and 7% of the contract price of construction projects is being wasted on failures (Wamelink, Stoffele & van der Aalst, 2002). These are shocking figures, especially when compared to the small profit margins of only 1,9% for the year 2012 in the Netherlands (Bouwend Nederland, 2014).

Both studies point in the same direction and present the effectiveness of processes as one of the most important areas of improvement (Egan, 1998; Wamelink et al., 2002). Especially the integration of the designing phase and the construction phase is mentioned as being of major importance. In the classical delivery methods for construction projects, there was a clear separation between the designing phase and the execution phase. These kinds of contracts are called Design, Bid and Build contracts (DBB). In the Netherlands this delivery method is known as the UAV (Uniforme Administratieve Voorwaarden). Last two decades there has been a shift towards more integrated delivery methods, where the contractor is also responsible for the design. These new delivery methods are called Design and Build contracts (DB). The Dutch equivalent to this international phenomenon us called the UAV-GC (Uniforme Administratieve Voorwaarden – Geïntegreerde Contracten). Many studies emphasize the great advantage of this new integration of processes for the client, such as cost and time reduction in the realization of building projects (Roth, 1995; Konchar and Sanvido, 1998; Ojo, Aina and Yakeen, 2011).

It is clear that this revolution in the way of working has had a major impact on client satisfaction and on the project execution. Still, there are few extensive studies that focus primarily on the financial impact for the contractor. It is sometimes assumed that the step towards DB delivery methods has shifted risks from the clients to the contractors, because the

(6)

responsibility for the design is no longer carried by the client, but by the contractor (Ernzen and Schexnyander, 2000, p 13.). The same study finds out that the profit margins are larger by using DB contracts. Construction companies were used to submit all deviations in the executions phase of a project as contract variation. The new delivery methods make construction companies responsible for the design themselves, and deviations in the execution phase cannot be submitted as contract variation anymore (CROW, 2005). The result is increased risks, but also increased opportunities to optimize the design (Ernzen et al., 2000).

Several big construction companies have presented large unexpected losses on their projects. Due to the late stage of observation of these losses, the impact for investors was enormous (De Financiële Telegraaf, 2014). One of the reasons for the major impact of these late discovered losses is the way revenue and profit are recognized by project-based organizations. For revenue recognition the percentage-of-completion method is commonly used by all the big construction companies. With this method, revenue and profit from projects in progress are recognised based on the stage of progress of the project (Škoda, 2008). A periodically forecast of the expected profit is used for profit recognition. Because revenue and profit are recognised before the end of the project, the risk level of construction projects is of primordial importance for investors. An increase in risk could have a devastating impact with the small profit margins the construction industry is used to in the last several years.

This thesis analyses the impact of the shift towards DB delivery methods on the financial forecasts and results. It is expected that these expected increased risks with new project delivery methods will be visible in the financial results of construction projects. For investors in the construction industry it is of primordial importance to be aware of the impact of delivery systems on the risk they are bearing. There has been a wide variety of research on the impact of new delivery methods from a client’s perspective, but few extensive research have been done on the impact of these methods for the contractors. Yet, a contractor-centered research perspective is of great importance for shareholders in the construction industry. That is why we consider an empirical study into this matter as being particularly significant. In this study the following research question will be answered:

“Does the shift towards DB delivery methods have a significant impact on financial risk and financial return of infrastructural projects?”

(7)

The next section will provide some theoretical background that is necessary for a good understanding of the topic. The first section presents a brief overview of the history, followed by some theoretical background about project delivery systems. Then we will consider some relevant (empirical) research literature and discuss the percentage-of-completion method in more detail. In the following chapters, the research framework with the research question and the hypotheses will be discussed, followed by an extensive description of the research methodology. Finally, we will present the results of our study and formulate an answer on the research question.

(8)

2 Literature review

2.1 Research on delivery method influence

Delivery methods are an interesting phenomenon that has also attracted the attention of the academic world. The last decade, a myriad of studies on delivery methods in the construction industry has emerged. Previous research mainly focuses on the impact of delivery methods on performance, costs and financial results. In this paragraph, the diachronic development of delivery methods will be briefly discussed to look at delivery methods in the right perspective and the implications of transitions in delivery methods will be discussed in terms of effects on the contractor.

2.1.1 A brief history

Project delivery systems have evolved over the years. They are rooted in the Middle Ages, where the medieval master builder was hired by an owner to design, engineer and construct an entire facility (Branca, 1987). But times have changed and new projects “required a new type of master builder who could accurately forecast costs for projects” (Konchar et al., 1998, p.435). Fast economic growth and technological developments in the early 20th_{century led to specialization}

and separation between designing and building in the construction industry. This resulted in the traditional “Design, Bid and Build (DBB)” approach, where the client contracted separate companies for designing and for building. For construction companies in the Netherlands, the traditional delivery method that comes with this approach is known as the UAV (Uniforme Administratieve Voorwaarden). This approach led to inefficiency, because knowledge of the designing phase was lost in the execution phase. As a consequence, the designing and the building companies could point to each other and play the blame game (Ojo et al., 2011).

The problems with traditional delivery methods led in the United States to the development of new forms of contracts, called “Design and Build (DB) (in the Netherlands known as UAV-GC). In this new form the contractor became also responsible for the design. This way the design and construction were integrated to lead to more efficient designs. In the end of the 90’s a quarter of the US construction revenue was executed based on these new delivery methods (Konchar et al., 1998).

(9)

In the Netherlands, the developments in the United States were followed, and in the year 2005 the new delivery method UAV-GC 2005 was introduced. After that, a strong increase in use of integrated delivery methods has been seen in the Netherlands. For the year 2010 in the United States, the DB delivery method is already used for 40% of the non-residential construction projects, which is a 10% increase since 2005 (Design and Build Institute of America, 2014). Both clients and contractors are still busy getting used to this new way of working and this new way of thinking.

2.1.2 Project delivery systems: DB versus DBB

‘A project delivery method is a process of designing and constructing any facility’ (Shrestha, O’Connor & Gibson, 2012, p.1). It is used by owners and clients as a method to deliver and finance constructed facilities. Shestra (2007) further specifies that project delivery method can be seen as a process in which the components of design and construction are combined to deliver a project. Factors that come into play in this process are the roles and responsibilities of all parties in question, the cost of materials, the activities and the labor that are needed.

The crucial difference between design and build (DB) and design-bid-build (DBB) contracts is that when using a DB contract, the designing and construction services are contracted by one single entity. It relies on a single point of responsibility contract. This is not the case for DBB where “owners procure, design and construct separately” (Shresta, 2007, p.2). The DB approach is used to minimize risk for the client and wants to reduce the delivery schedule by eliminating the strong separation between the designing and the construction phase known from DBB contracts. Risks are reduced for the client because there is one point of responsibility: when something goes wrong, there are not two parties who can point at each other (Murdoch and Hughes, 2007).

The figure below shows this basic difference between the newer delivery system (DB) and the traditional delivery system (DBB) in a well-organized way (figure 1). In the traditional model, the owner contracts two parties: a designer and a contractor. Usually in the first stage of the process, the owner contracts a designing party, which works out a very detailed design. After the designing phase, the owner contracts a contractor for the construction phase. The contractor gets a very detailed description of everything that has to be build. The result of this is that when things work out differently in the construction phase, all the variations can be submitted as being contract variations to the client.

(10)

Let us consider a fictitious example. Suppose that the design prescribes ten roadblocks for an infrastructural project. During the construction phase, it is discovered that not ten, but eleven roadblocks are needed. In that case, the eleventh road block can be submitted as contract variations. For the owner/client, this can lead to many unexpected costs. In the DB project delivery method, the contractor is also responsible for the design. When in our imaginary infrastructural project ten roadblocks are prescribed in the designing phase and in the construction phase it seems to be necessary to use an eleventh road block, this eleventh road block cannot be submitted as contract variation. This is the case because the contractor himself is also responsible for only designing ten roadblocks; a costly miscalculation.

Figure 1: Graphic representation of traditional versus design-build project delivery (Design-Build Institute of America, 2014)

Usually design-build projects are contracted by a general contractor, but most contactors also hire designing professionals. In many cases there is a partnership between a construction company and a designing company like engineering companies. This also holds true for the company that is part of this research.

2.1.3 Research on DB versus DBB contracts

The role and responsibility of the construction companies has changed revolutionary. The shift toward DB contracts is a worldwide development and has inspired a wide variety of academic work (Konchar et al., 1998; Ojo et al., 2011). Many studies try to assess the impact of this development on performance outcomes. The majority of these studies are empirical researches based on large project samples. They usually point in the same direction: the design and build approach is a good thing, for the client.

(11)

The first empirical study that focused on the benefits of DB contracts was conducted by Roth (1995). In his study, he compares the construction of twelve Navy childcare facilities: six were built using a DB delivery system, and six using DBB. His conclusion is that “the use of DB contracts within a selected sample of NAVFAC's MILCON program is significantly reducing combined design and construction costs” (p. 44). It is important to note that the client perspective is chosen here: the costs for designing and construction were lower for the client.

Konchar et al. (1998) investigated 351 US construction projects to study the impact of delivery methods on performance. The outcome of their study was that DB contracts outperform the traditional DBB contracts in terms of unit costs, construction speed, delivery speed, cost growth and schedule growth. In their conclusions chapter they state: “This research puts an empirical face on the fact that projects administered using design and build project delivery can achieve significantly improved cost and schedule advantages” (p. 444). The outcomes of their study seems to be positive for both client and constructor. An empirical study by Bennett, Pohecary and Robinson (1996) showed comparable results.

The same picture emerges for the semi-public sector, that was studied by Molenaar et al. (1999). They studied the performance of DB delivery systems in the semi-public sector and compared the performance of 104 different construction projects on four outcomes: budget, schedule, administrative burden and owner satisfaction. They concluded that clients were generally satisfied using a DB delivery system. Again, the outcome is interpreted from a client -centered perspective. Yet, it is interesting to consider the impact of this delivery system for the contractor.

One of the few studies that concentrates on the financial impact of delivery methods for the contractor is done by Ernzen et al. (2000). Their results show that the average profit margin for construction projects using DB delivery methods is 3,5% higher than when DBB delivery methods are used. They came to this conclusion based on a sample of projects from a big American construction company from the years 1991 to 1997. Besides this quantitative research they also compared the performance of two comparable projects: one with a DB delivery method and one under DBB. They studied especially the comparison between pre-calculated and realized labor costs in the execution phase. This research also shows that the risk for the contactor is significantly higher using DB contracts, especially the risk of labor cost overruns for the contractor. They state that by using DB delivery methods, the scope of the project is not always clear at the moment of bidding and calculations. The project is contracted for a fixed price while the design is still partially complete. In some situations this leads to later cost

(12)

overruns. On the other hand there is also the opportunity for optimization by the contractor. In their conclusions chapter they state: “The contractor has more opportunity to decrease labor costs during the design process with the DB contracting method; however, the lack of a clear project scope definition makes it difficult to estimate labor cost. Work item quantity and labor productivity fluctuations are, therefore, to be expected” (p. 14). Using DB contracts, risk has shifted from the client to the contractor, with a higher profit margin for the contactor in return. Likewise, Öztaş and Ökmen (2004) emphasize the increased risks under DB contracts, but both for the contractor and the client.

The interest in DBB versus DB delivery methods has also inspired a couple of African scholars. Ojo et al. (2011) also opt for the client’s perspective, concluding that DB contracts led to earlier delivery of the project, less cost overruns for the client and a higher satisfaction rate. In their sample of Nigerian building projects (DBB, n=53; DB, n=15), the researchers focused on the influence of delivery method on cost and time overruns. Although the design-build contracting projects did better and had less overruns than the design-bid-build projects in terms of money (21.4% versus 42.6%) and time (36.8% versus 135.6%), the authors state that still both contracting methods involve cost and time overruns.

Although their main focus varies slightly from study to study, the majority of researchers opt for an empirical method on a sample of projects. Interestingly, the results seem to be quite comparable: the DB delivery method seems to work favorably for the client. However, almost all of these studies seem to focus on the position of the client, and not the contractor. Popular targets are a comparison for delivery speed, quality and cost overruns. Again, they seem to focus on the cost overruns for the clients only, while “one of the major differences between Design-Build projects and others is that most of the risks associated with Design-Design-Build projects are borne by the contractors” (Ikediashi, Mendie, Achuenu & Olodokun, 2012, p. 37). So although the increased risk for contractors is highly relevant, the focus of most studies is client-centered, while the increased risk for contractors is only mentioned a few times. The studies make clear that something has changed fundamentally in one of the biggest worldwide industries.

Much has been written about the shift towards DB contracts, but these studies still do not clarify what the (financial) impact is for the contractor. The only study that focuses on the impact for the contractor is done by Ernzen et al. (2000) as mentioned earlier. This makes it difficult to predict a theoretical outcome for this research. The findings of Ernzen and Schexnyander will be the starting point for our empirical analysis.

(13)

Previous studies have shown the advantages of DB contracts for the client: fewer cost overruns and less time overruns for the client. The next step is to answer the question whether these advantages also count financially for the contractor. The fact that less cost overruns occur for the client could be because the contractor is not able to submit additional costs as contract variation. In this case, the advantage for the client could be a disadvantage for the contractor. On the other hand, it could also be the case that the integration of designing and building leads to a more efficient design and execution of the project. If so, then it could be that the advantage for the client is also an advantage for the contractor.

Since the shift towards DB contracts leads to more fixed-price contracts with less contract variations, the expectation for this research is that both the risks as the opportunities have increased for the contractor. If something goes wrong in the designing or building phase, the risks of cost overruns are totally for the contractor. On the other hand, if there are opportunities to make the design more efficient, there are also greater opportunities when smart solutions are found. The expectation is that the financial uncertainty and risk for the contractor will increase by using DB delivery methods. The main reason for this expectation is that in the bidding and contracting phase the scope of the project is less clear than when DBB contracts are used.

Researchers Sample

size

Perspective Project type Main findings

Roth (1995) 12 client US Naval facliities DB outperf orms DBB: less costs and earlier delivery than DBB.

Konchar et al. (1998) 351 client US building projects DB outperforms DBB: 6,1% less costs, 12% faster delivery, 5,2% less cost grow th than DBB.

Molenaar et al (1999) 104 client US Public sector DB leads to higher client satisfaction than DBB.

Ernzen et al. (2000) 2 / 24 contractor US building projects DB show s more risk in labor cost fluctuation but also a 3,5 percentage points higher profit margin than DBB.

Ojo et al. (2011) 68 client Nigerian building projects

DB show s less time overrun (36,8% vs. 135,6%), less cost overrun (21,4% vs. 42,6%) and higher client satisfaction (78% vs. 51%) than DBB.

Hale et al (2009) 77 client US Naval facilities DB projects take less time to complete and show less time & cost grow th than DBB.

(14)

2.2 Percentage-of-completion

Some earlier studies indicate that the usage of DB delivery methods has an impact on the uncertainty (financial risks) and on the profit (financial return) of construction projects. This is very interesting for investors in the construction industry, because the percentage-of-completion method is used by most construction companies for revenue and profit recognition. Based on the progress of the project, revenue and profit are recognised. For recognition of profits and losses, a profit forecast of the project is composed. Reservations are made for expected losses in the year they are discovered, and profits are recognised based on the progress of completion (Škoda, 2008). From the moment the execution of the project begins, the project starts generating revenue and profit. When a DB delivery method is used, the scope of the project is sometimes still unclear (Ernzen et al., 2000). This makes this generated revenue and profit less reliable.

By using the DB delivery methods it could be more difficult to make a trustworthy forecast of the expected profit. Internationally, IAS 11 is for construction contracts the salient standard addressing the accounting for construction contracts (Dobler, 2008). The Dutch standard is described in RJ 221:301/309, and a more detailed description of usage of this method is offered in RJ 221:414c.

The big importance of these forecasts for the management of construction companies, but also for investors in the industry, makes this a very relevant and interesting research (Larson & Brown, 2004). Both the expected profit in the bidding and contracting phase as the forecasted profits during execution are part of this thesis.

(15)

3 Research framework

3.1 Objective

The main objective of this research is to investigate the impact of delivery methods (DB vs. DBB) on financial risk and financial return of infrastructure projects. This impact is studied on three different prognostic outcomes, namely on pre-calculation and final profit, the bandwidth of profit forecasts and the financial return of projects.

3.2 Research question

The refined main research question of our study is as follows: “Do different delivery methods

(design-bid-build versus design-build) have a significant impact on financial risk and financial return of infrastructural projects?” This research question contains two components: the financial risk and the financial

return. It is expected that the new delivery methods have increased risks for contractors, as indicated in several previous studies (see for instance Ikediashi et al., 2012; Ernzen et al., 2000). Also it is expected that the shift towards DB delivery methods has increased the financial return on infrastructural projects (Ernzen et al., 2000).

3.2.1 Sub-questions on financial risks

Two different sub-questions will be examined to measure the level of financial risk for both DB and DBB delivery methods. Firstly, the difference between pre-calculated result (the expected profit before the execution phase) and the final profit will be tested. The following sub-question is the main focus for the first part of our analysis: “Is there a difference in financial risk (difference

between pre-calculated and final profit) between design and build delivery methods on one side and design, bid and build delivery methods on the other side?”

Besides that, it is relevant to examine the profit forecast during the execution phase of the project, and the impact of delivery methods on the uncertainty (fluctuation or bandwidth) of profit forecasts. The sub-question that we will try to answer for this part of our analysis is therefore: “Is there a difference in financial risk (most optimistic versus most pessimistic profit forecast) between

(16)

3.2.2 Sub-question on financial return

Besides the impact on financial risks and uncertainty, it is expected that the shift towards DB contracts also influences the financial return for infrastructural projects. Only few studies are mention the impact of the DB delivery method on the financial risks for the contractor, but even less investigate the influence of the shift towards DB on financial return. One empirical research that compared the financial return for DB versus DBB projects is done by Ernzen et al. (2000) as mentioned earlier. Their study shows convincingly that the financial return of DB projects is higher. The expectation for our study will be in line with this research: the expectation is that the financial return on DB projects will be higher. Hence, we will try to find an answer to the following sub-question: “Is there a difference in financial return (financial return as % of the revenue) between

(17)

4 Research methodology

4.1 Data collection

In order to answer our research question, a large sample of project results of a construction company was needed. For our study, the project results were delivered by the infrastructural division of one of the largest construction companies in the Netherlands. The infrastructure division consists of several operating companies, such as constructing and maintaining, roads, railroads, bridges, tunnels and waterways. The project results of the following working companies are used for our analysis:

 Concrete and hydraulics: building bridges, viaducts, tunnels, parking garages and stations;  Infrastructure: constructing roads and highways (asphalt). Three different regions are

part of this research;

 Environment: soil investigation and remediation;  Railroads: construction and maintaining railways;

After accepting the condition of confidentiality, a series of Excel files containing information on many different construction projects (n=353) from 2009 to 2013 was made available for our study. The criteria for inclusion in our research project were that (1) the projects were finished between 2009 and 2013, (2) the delivery method of the project is well reported, and (3) the project has a contract price.. Only the projects executed under DBB and DB are included. Projects with other or unclear delivery methods are excluded from this research. The final profit for each project is mostly the same as the last profit forecast. To make sure that the right final profit number is used, these numbers are compared to a separate file with official final profit numbers. Four client types used in this research, while six were available in the data. Municipals and provinces were combined because only a few projects with provinces were part of the sample. For the same reason the projects with developers as a client are combined with companies in general. For the regression analysis, project size (revenue) is included as a continuous variable. To include project size in the ANOVA analysis, a categorical variable is made with four project size categories ( < 0,5 mln.; 0,5 – 1 mln.; 1 – 5 mln.; 5 – 10 mln.; > 10 mln.)

For the second research question, which looks specifically at profit forecasts, only projects are included with a duration of at least half a year. Due to several violations of these inclusion criteria or to incomplete information, a small number of projects were excluded from our

(18)

analysis. The information of the remaining projects was collected and combined in a final large data set which eventually contained the results of 353 construction projects. This final data set was converted into a SPSS file so that further statistical analyses could be executed.

4.2 Sample information

The data that was used for this study came from different sources, but is all delivered by one construction company in the Netherlands. Firstly, there was a series of files containing basic information about each project. This information included the project number, project name, delivery year, client, type of work, revenue and final profit. Also for most projects the delivery method was known, although for some projects the delivery method had to be retrieved from other sources. Some were reported in the ERP system, others in separate Excel files.

Besides the files containing basic information for each project, all management reports on financial results were received for all projects. These files show the pre-calculated profit and a quarterly forecasted profit for each project. Each of the aforementioned types of basic information served as input for the statistical analysis, either as independent nominal/categorical variables (client, delivery method, working unit, delivery year) or as numerical variables (revenue, pre-calculated profit, quarterly forecasted profit, final profit). Since there were many different clients, we decided to categorize them in four different categories in a way that would result in enough projects for each category, while preserving the main characteristics of the original client. The four categories that were created for client are displayed in the table below, together with the coding of the other relevant categorical variables in this data set (Table 1: Coding categories for delivery method, client and working unit).

Variable Categories Code

Delivery method DBB (in the Netherlands better know n as UAV) DB (in the Netherlands better know n as UAV -GC)

1 2 Client Municipalities / Provinces

Companies

Rijksw aterstaat (Water Board)

Prorail (governmental railroad building/maintaining company

1 2 3 4 Working unit Infrastructure Mid-West (Infra-MW)

Infrastructure North-West (Infra-NW) Infrastructure South-West (Infra-SW) Environment

Railroads

Concrete and hydraulics (C & H)

1 2 4 5 7 8 Table 1: Coding categories for delivery method, client and working unit

This numerical information could be combined with the basic data set of project information. There was enough information available on the independent, dependent and control variables in our study. Since the reports are very similar in nature over the five years that were selected for

(19)

our research (2009-2013), the information is comparable over an extended period of time and therefore well suited for our analysis.

4.3 Research method and intended analyses

The final data set was converted into an SPSS file for further statistical analyses. After an exploration of the data with descriptive measures of central tendency, normality and spread, some inferential statistical methods were used, both one-way and two-way ANOVA’s but also a regression analysis. For all tests, the level of significance is set at the conventional level of α = 0.05, which means that the null hypothesis will only be rejected if the p-value associated with the test statistic is lower than this chosen level of significance. Furthermore, several regression analyses were executed.

For each project in our final sample, the following information is combined in the data set: the name, number, working company, year of completion, revenue, pre-calculated result, final result, and all the quarterly profit forecasts delivered. The focus of this study is the impact of the delivery method (DBB or DB) on three different measures:

1. Pre-calculation and final profit, i.e. the difference between pre-calculated and final result. Therefore, the absolute difference between pre-calculation and final profit was calculated. In order to make the projects comparable, this absolute difference was corrected for the project size (namely by dividing the absolute difference by the total revenue for each projects);

2. The bandwidth of profit forecasts, i.e. the variation and uncertainty in the profit forecasts. This measure is calculated as the difference between the most optimistic (highest) and most pessimistic (lowest) forecasted profit which could also be called the profit bandwidth, corrected for the size of the project. Again, this was done by dividing this range by the total revenue, so that the profit bandwidth of all projects are one-to-one comparable. 3. The financial return of projects, i.e. the final profit of a construction project divided by the toe

total contract price (revenue)

Three different dependent variables were created to test the three different hypotheses. In our study, we made use of one-way ANOVA’s to assess the impact of delivery method (either DBB or DB) on these three outcome variables. Besides that, also the impact of the working company, client, delivery year and profit size were part of this research. They were also included in two-way

(20)

ANOVA’s to test the combined impact on the outcome variable and the interplay between these predictive variables. After that, separate variables were made for every working company and client, so a regression model could be build. This way the single impact of every working company and client on the outcome variable could be tested, combined with the impact of delivery method and revenue. Also the predictive power of the model as a whole could be tested.

(21)

5 Results

5.1 Sample description

The total sample contains the project results of 353 construction projects that were completed between the years 2009 and 2013. The most important predictive variable is delivery method, but besides that also the impact of the client, the working company, the year of completion and the project size will be examined. The spread over the two different delivery methods (DBB and DB) is shown in the table below (Table 2: frequency of projects for both delivery methods). The majority of projects in the sample (71%) used the more traditional DBB delivery method.

Frequency Percentage DBB 251 71%

DB 102 29%

Total 353 100%

Table 2: frequency of projects for both delivery methods

The projects that are part of this research are spread over several working companies of the infrastructural division of a big construction company in the Netherlands. The figure below shows the frequency of projects in this research for each working company (Table 3: frequencies for both delivery methods for each working company). Because the delivery method is the most important variable for this research, the frequency for both delivery methods for each working company is shown. All working companies work with both DBB and DB delivery methods. More than half of the projects are executed by working company Infra-ZW, the working company with not only the highest number for the DBB contracts, but also for the DB.

DBB DB Total Infra-MW 10 1 11 Infra-NW 69 7 76 Infra-ZW 119 63 182 Environemt 5 5 10 Railroads 30 24 54 C & H 18 2 20 Total 251 102 353

(22)

The different clients of the projects are clustered in four different categories, as described in the previous chapter (see 4.2). The table below shows the distribution of DBB and DB delivery methods, clustered by client (Table 4: frequencies for both delivery methods for each client category). More than half of the projects in our sample have municipalities and provinces as client (51%), but companies are also well represented (34%). For municipalities and provinces, the distribution between DBB and DB is less evenly distributed than for the other clients.

DBB DB Total Municipalities / provinces 160 20 180

Com panies 56 64 120

Rijksw aterstaat (Water Board) 16 4 20

Prorail 19 14 33

Total 251 102 353

Table 4: frequencies for both delivery methods for each client category

All projects that are part of this research were completed between 2009 and 2013. The table below presents the spread over the years and over the two delivery methods (Table 5: frequencies for both delivery methods for each year). The majority of projects in our sample were completed between 2010 and 2011. There is no clear increase of either of the contract forms over the years, with χ²(4)=4.475, p=0.346.

DBB DB Total 2009 9 2 11 2010 64 37 101 2011 92 32 124 2012 25 10 35 2013 60 21 81 Total 250 102 352

Table 5: frequencies for both delivery methods for each year

The projects that are part of this research differ strongly in project size based on the total revenue (contract price + contract variations). The smallest contract price is only 18,000 euros while the biggest project has a contract price of more than 20 million euros. The table below (Table 6: Frequencies for five project revenue categories for both delivery methods) shows the spread over five revenue categories. The amounts are quite comparable for the four lowest categories, but all five projects above 10 million euros are DB projects.

(23)

DBB DB Total 0 - 0,5 m ln. 135 58 193 0,5 - 1 m ln. 58 20 78 1 - 5 m ln. 48 17 65 5 - 10 m ln. 9 2 11 > 10 m ln. 0 5 5 250 102 352

Table 6: Frequencies for five project revenue categories for both delivery methods

5.2 Analysis of pre-calculated vs. final profit

The aim of our first analysis is to compare the pre-calculated result for each project with the final profit at the end of the project execution. A variable is created of the absolute difference between the pre-calculated result and the final profit. This number is corrected for the project size (total revenue for each project). This variable will be used to test the first hypothesis for this research.

5.2.1 Descriptive statistics and variance analysis

The expectation is that delivery method, working company, client and project size will also have influence on the difference between pre-calculated and final profit. Firstly, we will start describing the means and standard deviations of this outcome variable for each of the predictive variables used and test the influence on the outcome variable using one-way and two-way ANOVA’s. In the second part of this paragraph, a regression analysis will be used to test the impact of all variables on the outcome variable. The table below shows the means and standard deviations for all projects for each DBB and DB delivery methods (Table 7: mean and standard deviation for both delivery methods on pre-calculated versus final profit). A remarkable observation at first sight is to see that DB has both a higher mean (.133 versus 1.04) and higher dispersion rate (a standard deviation of respectively .162 versus .117) than the traditional DBB delivery method. So for the DB projects in the sample, the mean difference between pre-calculated result and final profit seems to be higher than for DBB projects. This could point to a higher level of financial risk for DB projects.

With a one-way ANOVA the single impact of delivery method on the difference between pre-calculated result and final profit is tested. On average, DB delivery systems (M=0.13. SE = 0.016) seem to have a higher difference between the corrected difference between pre-calculated and final profit than DBB delivery systems (M=0.10. SE=0.007). This observed difference is marginally significant: F(1, 351) = 3.544, p = 0.061.

(24)

Frequency Mean Standard Deviation

DBB 251 .104 .117

DB 102 .133 .162

Total 353 .113 .132

Table 7: mean and standard deviation for both delivery methods on pre-calculated versus final profit

Working company

When a more detailed look is given on the means and standard deviations for each working company (Table 8: means and standard deviations for each working company and delivery method on ), it becomes clear that this is the case for all working companies, except Infra-MW. There are only 11 projects of this working company part of this research, of which only one project had a DB delivery method. Furthermore, the working company C & H shows a very large difference. For this working company only 2 DB projects are included, both of them having a large difference between pre-calculated and final profit.

Also the single impact of the working company on pre-calculated versus final profit is tested separately. Although the impact of delivery system was marginally significant on our outcome variable, the main effect of working company on our outcome variable is statistically significant with F(5, 347) = 5.215, p= .000. Frequ encies Mean Std. Deviation DBB DB DBB DB Total DBB DB Total Infra-MW 10 1 ,093 ,080 ,092 ,107 ,101 Infra-NW 69 7 ,092 ,109 ,094 ,083 ,104 ,085 Infra-ZW 119 63 ,092 ,103 ,096 ,079 ,117 ,094 Environm ent 5 5 ,160 ,206 ,183 ,135 ,146 ,135 Railroads 30 24 ,177 ,199 ,187 ,238 ,249 ,241 C & H 18 2 ,100 ,250 ,115 ,108 ,057 ,113 Total 251 102 ,104 ,133 ,113 ,117 ,162 ,132

Table 8: means and standard deviations for each working company and delivery method on pre-calculated versus final profit

(25)

When delivery method and working unit are combined in a two-way ANOVA to test the interaction effect, it becomes clear that the interaction effect between the two is not statistically significant with F(5, 341) = .441, p = .818.

Client

Another important variable that could have a major impact on financial risk and return is the client. The table below shows remarkable differences in the mean and standard deviation of the outcome variable for each client (Table 9: means and standard deviations for each client and delivery method on pre-calculated versus final profit). An interesting picture seems to emerge in the interaction between client and delivery method: for the municipalities and provinces, it seems as if the delivery method has almost no influence on the outcome variable, whereas for Prorail and companies, the mean is bigger for DB projects than for DBB projects. For projects that have Rijkswaterstaat (water district board) as client, it is exactly the opposite: DBB projects have a higher mean than DB projects. Yet, it must be noted that for the latter client, only 4 projects are included with the DB as delivery method.

Also the single effect of client on the output variable pre-calculated versus final profit is tested separately. Just like the working company, the impact of client is statistically significant with F(3, 349) = 2.769, p = 0.042). A second series of post-hoc analyses on client revealed that the difference between municipalities and Prorail was also significant (p = .028); Prorail has a higher mean on our first outcome variable (Table 9: means and standard deviations for each client and delivery method on pre-calculated versus final profit).

Frequencies Mean Std. Deviation DBB DB DBB DB Total DBB DB Total Municipals / Provinces 160 20 ,105 ,099 ,104 ,124 ,102 ,122 Com panies 56 64 ,092 ,125 ,110 ,098 ,135 ,120 Rijksw aterstaat 16 4 ,118 ,053 ,105 ,105 ,030 ,098 Prorail 19 14 ,125 ,243 ,175 ,122 ,288 ,213 Total 251 102 ,104 ,133 ,113 ,117 ,162 ,132 Table 9: means and standard deviations for each client and delivery method on pre-calculated versus final profit

(26)

A two-way ANOVA with client and delivery method shows that there is also a marginally significant interaction effect between client and delivery method on the difference between the expected and realised profit, with F(3, 345) = 2.258, p = .081.

Delivery year

The following table shows the means of the absolute difference between pre-calculated profit and final profit for every year, corrected for revenue (Table 10: means and standard deviations for each year and delivery method on pre-calculated versus final profit). Although there is some variation in the outcome variable over time, the means of DB projects are higher than for the DBB projects; this is the case for every single year. The same is true for the dispersion, except for the year 2013, where the standard deviation of DBB projects seems to be a bit higher.

In order to make sure that different years (2009 to 2013) with different market conditions are not influencing this research, also the influence of the different years is tested. There was no significant impact of year on the difference between pre-calculated result and final profit with F(4, 348) = 1.572, p = .181.

Frequencies Mean Std. Deviation DBB DB DBB DB Total DBB DB Total 2009 9 2 ,123 ,170 ,132 ,116 ,127 ,113 2010 64 37 ,120 ,138 ,127 ,104 ,213 ,152 2011 92 32 ,088 ,097 ,090 ,075 ,080 ,076 2012 25 10 ,072 ,202 ,109 ,079 ,214 ,142 2013 60 21 ,123 ,143 ,128 ,179 ,122 ,166 Total 250 102 ,104 ,133 ,113 ,117 ,162 ,132 Table 10: means and standard deviations for each year and delivery method on pre-calculated versus final profit

There is no statistically significant interaction effect between the year and delivery method on the outcome variable when a two-way ANOVA is used, with F(4, 343) = 1.254, p = .288.

(27)

Project size

When the project size (four revenue categories) is used as predictive variable, the table below (Table 11: means and standard deviations for each project size category for both delivery methods on pre-calculated versus final profit) shows that the difference between pre-calculated and final profit is relatively bigger for smaller projects than for larger projects. The impact of revenue is not significant with F(4, 348) = 1.039, p = .387.

Frequencies Mean Std.

Deviation

DBB DB Total DBB DB Total DBB DB Total 0 - 0,5 m ln. 136 58 194 .110 .153 .123 .095 .195 .134 0,5 - 1 m ln. 58 20 78 .109 .119 .111 .176 .131 .165 1 - 5 m ln. 48 17 65 .090 .108 .095 .091 .070 .086 5 - 10 m ln. 9 2 11 .062 .060 .062 .070 .028 .064 > 10 m ln. 5 5 .084 .084 .053 .053 Total 251 102 353 .104 .133 .113 .117 .162 .132 Table 11: means and standard deviations for each project size category for both delivery methods on pre-calculated versus final profit

A two-way ANOVA shows that there is no interaction effect between project size and delivery method with F(3, 344) = .311, p = .818.

5.2.2 Regression analysis

Also a regression analysis is done to have a specific look at the influence of each working company and client separately. Also delivery method is a separate variable. Project revenue is added as the only continuous variable. Because the other variables are categorical, dummy variables are created. For working company, Infra-ZW is the control group and for client Municipalities and provinces is the control group. I chose these as control group because they have the highest amount of data. The regression model assumes that the independent variables are not correlated with each other. The Pearson coefficient was calculated for the independent variables for the ordinary least squares regression with pre-calculated versus final profit as dependent variable. The results are shown in the table below (Table 12: Pearson correlations for all variables with pre-calculated versus final profit)). When a Pearson correlation is above 0.8, multicollinearity is very likely to exist. Two variables show a high correlation: Prorail as client and

(28)

Railroads as working company (r=.756). This is easy explainable, since Prorail is the main client of the Railroads working company. Prorail will be excluded for the regression analysis.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (1) Diifference profit before/after 1 (2) Revenue -.085 1 (3) Delivery Method .100 .141 1 (4) Infra-MW -.028 -.014 -.078 1 (5) Infra-NW -.076 .048 -.227 -.094 1 (6) Environm ent .091 -.062 .080 -.031 -.089 1 (7) Railroads .238 .316 .146 -.076 -.223 -.073 1 (8) C & H .004 -.083 -.102 -.044 -.128 -.042 -.104 1 (9) Com panies -.017 -.060 .387 -.094 -.172 .202 .027 .031 1 (10) Rijksw aterstaat -.014 -.044 -.048 -.044 -.069 -.042 -.104 .311 -.176 1 (11) Prorail .151 .318 .096 -.058 -.168 -.055 .756 -.079 -.230 -.079 1 Table 12: Pearson correlations for all variables with pre-calculated versus final profit

The table below (Table 13: Results of the ordinary least squares with pre-calculated versus final profit) shows the results with the effects of every dummy variable on the outcome variable “pre-calculated profit versus final profit”. Here, the working companies are compared to Infra -ZW and the working companies to Municipalities.

It is interesting to see that the delivery method has a significant influence with p = .020. The Beta of .127 indicates that the risk (pre-calculated result versus final profit) is higher for DB projects compared to DBB projects. Especially the working companies Environment (with p = .027 and Beta .117) and Railroads (with p = .000 and Beta = .303) have a significant influence: both show a higher risk level, compared to Infra-ZW. Also the impact of revenue (the only continuous predictive variable) is significant (with p = .000 and Beta -.195). The larger projects in the sample show a smaller difference between pre-calculated versus final profit.

(29)

Variable Standardized Beta

Std. Error t-value Sig (Constant) .012 8.564 .000 Revenue -.195 .000 -3.560 .000 Delivery m ethods .127 .017 2.185 .030 Infra-MW -.001 .040 -.013 .989 Infra-NW .027 .018 .479 .632 Environm ent .117 .042 2.225 .027 Railroads .303 .021 5.344 .000 C & H .048 .031 .874 .383 Com panies -.109 .016 -1.865 .063 Rijksw aterstaat -.012 .031 -.226 .821 Table 13: Results of the ordinary least squares with pre-calculated versus final profit

Adjusted R-squared for this model is .089. The value of the F-statistic is 4.818.

5.3 Analysis of forecasted profit variation

The second analysis looks not only at the pre-calculation before the start of the project and the final profit at the end of the project, but also at the forecasts during the execution phase. Every quarter the working companies report the expected profit for each project. This analysis wants to compare the financial uncertainty (the highest and lowest forecast given) of the projects for each delivery method. So the outcome variable that will be tested here can actually be seen as the ‘bandwidth’ of the profit forecast. For each project, this is the difference between the lowest (most pessimistic) and highest (most optimistic) expected profit, corrected for revenue. In the analysis, the impact of client, working companies and year are included in the model as predictor variables (both separately and in interaction with each other).

The table below shows several descriptive numbers of this outcome variable compared between the two delivery methods (Table 14: Descriptives for both delivery methods on profit bandwidth). Comparable to the numbers in the first analysis, the DB projects show a higher mean (.16) for ‘bandwidth’ than the DBB projects (.131) and likewise, the dispersion is higher for DB projects than for DBB projects, with standard deviations of respectively .164 and .135. A one-way ANOVA is used to test the single effect of delivery method on the outcome variable

(30)

containing the bandwidth of the profit forecast for each project. Even though the means show there is difference in the forecasts for both delivery systems, this difference is only marginally significant with F(1, 323) = 2.733, p= .099.

Frequency Mean Std. Dev. Low er Quartile Median Upper Quartile DBB 235 0.131 0.135 0.050 0.090 0.150 DB 90 0.160 0.164 0.060 0.120 0.203 Total 325 0.139 0.144 0.060 0.100 0.160 Table 14: Descriptives for both delivery methods on profit bandwidth

Working company

However, when a closer look is taken at these differences and the impact of working companies is taken into account (Table 15: means and standard deviations for each working company and delivery method on the profit bandwidth), it becomes clear that this is not the same for every working company: for Infra-MW and Railroads the mean “bandwidth” under DBB is larger than for the DB delivery method, which is in contradiction to the general numbers reported above. Like for the first research question, the single impact of working company on the profit bandwidth is tested using a one-way ANOVA. The impact of working company has a significant impact on the bandwidth of the forecasted results with F(5, 319) = 7.914. p = .000. Post-hoc analyses using the Bonferroni correction, showed that for working company, the difference between Railroads and Infra-NW was significant (p = .000). Furthermore the difference between Railroads and Infra-ZW was significant (p = .000). In both cases, the working company Railroads has a higher mean on the outcome variable.

Frequencies Mean Std. Deviation

DBB DB DBB DB Total DBB DB Total Infra-MW 10 1 ,133 ,100 ,130 ,152 ,144 Infra-NW 69 7 ,104 ,159 ,109 ,081 ,083 ,082 Infra-ZW 119 63 ,111 ,127 ,116 ,082 ,120 ,096 Environm ent 5 5 ,194 ,226 ,210 ,207 ,130 ,164 Railroads 30 24 ,261 ,218 ,242 ,269 ,248 ,258 C & H 18 2 ,129 ,290 ,139 ,108 ,112 Total 251 102 ,131 ,160 ,139 ,135 ,164 ,144 Table 15: means and standard deviations for each working company and delivery method on the profit bandwidth

(31)

A two-way ANOVA with both delivery method and working company as predictive variables shows that there is no significant interaction effect with F(5, 313) = .778, p = .566.

Client

The table below shows the means for the outcome variable for both DBB as DB contracts for all four client categories (Table 16: means and standard deviation for each client and delivery method on profit bandwidth). At first sight, the numbers seem to follow a pattern that is quite comparable to the results of our first analysis, described in paragraph 5.2. For all client types, with the exception of Rijkswaterstaat (water district board), the forecasted profit bandwidth is larger for DB delivery methods are used than for the traditional DBB delivery method.

Also the single impact of client on profit bandwidth is statistically significant with F(3, 321) = 5.275, p = .001. A second series of post-hoc analyses on client revealed that the difference between Prorail and all three other clients is statistically significant: municipalities and provinces (p=.001), companies (p = .017) and Rijkswaterstaat (p = .026). Again, Prorail scores the highest on the outcome variable.

Frequencies Mean Std. Deviation DBB DB DBB DB Total DBB DB Total Municipals / Provinces 160 20 ,120 ,153 ,123 ,127 ,096 ,124 Com panies 56 64 ,138 ,146 ,142 ,139 ,136 ,137 Rijksw aterstaat 16 4 ,124 ,055 ,109 ,111 ,033 ,102 Prorail 19 14 ,208 ,254 ,228 ,184 ,285 ,230 Total 251 102 ,131 ,160 ,139 ,135 ,164 ,144 Table 16: means and standard deviation for each client and delivery method on the profit bandwidth

A two-way ANOVA with client and delivery method as predictive variables shows that there is no significant interaction effect with F(3, 317) = .602, p = .614.

Delivery year

Also the variation over the years 2009 unto 2013 is quite comparable to the outcome of the analysis of pre-calculation and final profit. Especially 2011 en 2012 show lower means than the other years. For 2012 the difference between DBB and DB is maximal. Also the impact of

(32)

different years are tested as predictive variable and have a marginally significant impact with F(4, 320) = 2.149, p = .075.

Frequencies Mean Std. Deviation DBB DB DBB DB Total DBB DB Total 2009 9 2 ,0733 ,0750 ,0736 ,0436 ,0636 ,0439 2010 64 37 ,0597 ,0807 ,0662 ,0514 ,1200 ,0793 2011 92 32 ,0469 ,0473 ,0470 ,0411 ,0327 ,0390 2012 25 10 ,0471 ,0820 ,0574 ,0378 ,0754 ,0530 2013 60 21 ,0710 ,0716 ,0712 ,0861 ,0530 ,0779 Total 250 102 ,0564 ,0677 ,0595 ,0563 ,0793 ,0636 Table 17: means and standard deviations for every year and delivery method on profit bandwidth

A two-way ANOVA shows that there is no significant interaction effect between delivery year and delivery method on the outcome variable with F(.4, 315) = .885, p = .473.

Project size

The effect of the four revenue categories on the profit bandwidth is quite comparable with the outcomes of the first analysis (pre-calculated versus final profit). The table below (Table 16: means and standard deviation for each client and delivery method on the profit bandwidth) shows that the risks in the execution phase seem to be bigger for small projects. Of course these risks are corrected for project size. The impact of revenue (clusterend in four categories) on the profit bandwidth is not significant with F(4, 320) = 1.644, p = .163.

Frequencies Mean Std. Deviation DBB DB DBB DB Total DBB DB Total 0 - 0,5 m ln. 125 47 .142 .194 .156 .130 .203 .154 0,5 - 1 m ln. 54 19 .131 .132 .131 .179 .126 .166 1 - 5 m ln. 47 17 .109 .127 .114 .090 .063 .083 5 - 10 m ln. 9 2 .093 .070 .089 .067 .042 .062 > 10 m ln. 5 .092 .092 .041 .041 Total 235 90 .131 .160 .139 .135 .164 .144 Table 18: means and standard deviations for each project size category for both delivery methods on project profit bandwidth

(33)

There is no significant interaction effect between project size and delivery method with F(3, 316) = .583, p = 627.

5.3.2 Regression analysis

Also a regression analysis is done with each working company, client and delivery method as a separate variable. The Pearson coefficient was calculated for the independent variables for the ordinary least squares regression with the bandwidth of the profit forecast as dependent variable. The results are shown in the table below (Table 19: Pearson correlations for all dummy variables with the profit bandwidthFout! Verwijzingsbron niet gevonden.). Again, Prorail and Railroads show a high correlation (r=.766). Also here, Prorail will be excluded from the analysis.

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (1) Profit bandwidth 1 (2) Revenue -.107 1 (3) Delivery Method .092 .164 1 (4) Infra-MW -.011 -.020 -.078 1 (5) Infra-NW -.113 .035 -.225 -.103 1 (6) Environm ent .088 -.068 .089 -.033 -.098 1 (7) Railroads .309 .322 .168 -.081 -.236 -.077 1 (8) C & H .001 -.076 -.103 -.041 -.120 -.039 -.095 1 (9) Com panies .016 -.045 .391 -.091 -.153 .223 .033 .008 1 (10) Rijksw aterstaat -.050 -.041 -.030 -.045 -.069 -.043 -.104 .331 -.165 1 (11) Prorail .206 .320 .119 -.062 -.181 -.059 .766 -.073 -.225 -.080 1 Table 19: Pearson correlations for all dummy variables with the profit bandwidth

The table below (Table 20: Results of the ordinary least squares with the profit bandwidth) shows the results with the effects of every dummy variable on the outcome variable “lowest versus highest forecasted profit”. The influence of delivery methods is only marginally significant with p = .118 and a standardized Beta of .093. The Beta shows that the financial risk in the execution phase is a bit higher for DB projects. Comparable with the outcome of the first analysis, again the working companies Environment (p = .037, Beta = .112) and Railroads (p = .000 and Beta = .388) show a significant influence. This means that these working companies show more risk (fluctuation) in the financial forecasts compared to Infra-ZW. It is interesting to see the big influence of revenue (the only continuous predictive variable) with p = .000 and a standardized Beta of -.241. This Beta is interesting because it shows that large projects have a

(34)

smaller risk in the profit forecasts. This is interesting because large projects have a longer execution phase which would make a bigger risk to be expected.

Variable Standardized

Beta

Std. Error _t-value Sig

(Constant) .013 9.714 .000 Revenue -.241 .000 -4.322 .000 Delivery methods .093 .019 1.569 .118 Infra-MW .021 .042 .401 .689 Infra-NW .014 .019 .238 .812 Environment .112 .044 2.097 .037 Railroads .388 .023 6.712 .000 C & H .049 .038 .877 .381 Companies -.072 .018 -1.216 .225 Rijkswaterstaat -.038 .035 -.683 .495

Table 20: Results of the ordinary least squares with the profit bandwidth

Adjusted R-squared for this model is 0.139. The value of the F-statistic is 6.830.

5.4 Analysis of financial return

For the third analysis we pay closer attention to the financial return of the projects as percentage of the revenue for each project. Main focus of research will be the influence of the delivery method on the financial return of infrastructural projects. Besides the delivery methods, the variables client, working company, delivery year and project size will be included again. The outcome variable for this research is the final profit of each project as percentage of the total revenue.

The table below (Table 21: descriptives for both delivery methods on financial return) shows again the descriptive information on the difference between DBB and DB contracts on the mean, standard deviation and quartiles of the financial return. Both the mean and the quartiles show that DB project have a higher return percentage than the traditional DBB contracts. It must be noted that the DBB projects contain one outlier with a financial loss that is eight times the total revenue: 818%. This outlier is excluded for further analysis because it led to a skewed distribution. Because the loss was discovered in an early phase, the loss was also part of the pre-calculated result, which makes it no outlier for the first two analyses.