Toward a cleaner project procurement: Evaluation of construction projects’ vulnerability to corruption in developing countries

Accepted Manuscript Toward a Cleaner Project Procurement: Evaluation of Construction Projects’ Vulnerability to Corruption in Developing Countries

Emmanuel Kingsford Owusu, Albert P.C. Chan, Ernest Ameyaw PII:

S0959-6526(19)30141-6

DOI:

10.1016/j.jclepro.2019.01.124

Reference:

JCLP 15511

To appear in:

Journal of Cleaner Production

Received Date:

15 August 2018

Accepted Date:

11 January 2019

Please cite this article as: Emmanuel Kingsford Owusu, Albert P.C. Chan, Ernest Ameyaw, Toward a Cleaner Project Procurement: Evaluation of Construction Projects’ Vulnerability to Corruption in Developing Countries, Journal of Cleaner Production (2019), doi: 10.1016/j.jclepro.2019.01.124

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Toward a Cleaner Project Procurement: Evaluation of Construction Projects’ Vulnerability to Corruption in Developing Countries Emmanuel Kingsford Owusu1; Albert P. C. Chan2; Ernest Ameyaw (Ph.D)3 1Doctoral

Candidate, Department of Building and Real Estate, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong

2 Chair

Professor, Department of Building and Real Estate, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong 3Lecturer,

School of Engineering, Environment and Computing,

Coventry University, Coventry CV3 1NZ, UK Abstract While the measurement and predictability of corruption are reported to be difficult due to its clandestine nature, this paper attempts to develop a measurement model for evaluating the proneness of the procurement process of construction projects to corruption. The model is developed using a soft computing approach (i.e. the Fuzzy Synthetic Evaluation technique) to assess the levels of vulnerability with respect to the stages of corruption and their corresponding activities within the procurement process of construction projects. A review of related literature produced a 21-activity list encapsulated within the four stages of the procurement process. The list was evaluated by procurement and construction practitioners through an expert survey. The analysis of the survey results revealed an overall project vulnerability index which suggests that construction projects executed in developing countries are relatively susceptible to corrupt practices. The main stages of the process that were identified to be vulnerable were the contract stage (CTS) and the post-contract phase (PCP). The developed model is, possibly, the first measuring tool that evaluates and predicts a project’s susceptibility to corruption and provides useful insights that can inform project parties, policymakers and anti-corruption activists in their efforts to implement necessary actions aimed at curbing the incidence of corrupt practices in construction projects in the developing countries.

Keywords: Construction Project Procurement; Corruption; Developing Countries; Fuzzy Synthetic Evaluation 1

ACCEPTED MANUSCRIPT 1. Introduction Corruption has been identified to be a continuously evolving phenomenon in the domain of construction management (Chan and Owusu, 2017; Shan et al., 2017). This evolution is reported to be the result of diverse determinants, including the causal factors of corruption, the contextual irregularities or vulnerabilities to corruption associated with projects, the barriers that hamper the effectiveness of anticorruption measures and the likelihood for a projects’ parties to engage in corruption (Le et al., 2014; Shan et al., 2015; Owusu et al., 2017). Categorically, the causal factors comprise regulatory-specific causes, statutory-specific factors, psychosocial-specific factors, project-specific and organizationalspecific factors as well as with other determinants which stem from cultural and social settings of organizations and institutional structures, economic policies and political environments (Trapnell, 2015; Zhang et al., 2016; Owusu et al., 2017). The variables within these constructs are often reported and examined in construction management related studies to be the underlying indicators for measuring the extent of corruption in a given context. While past studies (e.g., Sohail and Cavil, 2008; Shan et al., 2017; Owusu and Chan, 2018; Deng et al., 2003) have examined the various constructs of corruption in the context of construction procurement and management, few efforts (eg. Le et al. 2014; Tabish and Jha 2011) have been made to measure the projects’ vulnerabilities to corruption. However, no efforts have been made towards the development of an index tool for measuring or predicting the vulnerability of construction projects to corrupt practices in the public sector and the extent of such corruption. Consequently, most sectors, including public construction sectors from different countries, depend on general indexes, such as the Transparency International’s Corruption Perception Index, World Bank’s Worldwide Governance Indicators (WGI) and the Global Competitiveness Index (GCI) on Government Integrity by World Economic Forum, to determine the degree of corruption in different sectors of the specific measured countries (Ameyaw et al., 2017; Krishnan, 2009). Since these indexes do not reflect the specificity of the performance of the public sectors and only give overall estimations on examined countries, it is difficult to estimate the actual or approximated contributions from the public construction sector which is reported to be the second most corrupt sector in the world (Kottasova, 2014). Therefore, different reported set of irregularities emanating from contractual and procurement works including the 2

ACCEPTED MANUSCRIPT execution and maintenance of construction and other infrastructure related projects are mostly left unexamined or partially examined by designated public departments such as the audit departments that also report on the overall financial or performance irregularities on different public sectors, including the construction and procurement sectors. Moreover, since these audits are normally conducted by general auditors and not necessarily specialized construction auditors, the identified irregularities may not be detailed enough to capture the specific conditions on the ground, such as specific forms of corrupt practices with their associated causal instigators and the specific stages of the construction process where the reported irregularities are identified. The early detection of the risk factors and the causative factors that propel incidence of corrupt practices can help abate the manifestation, proliferation and the unpleasant effects associated with corruption (Owusu et al., 2017). However, due to the difficulty in measuring corruption, it may seem impossible to establish measures to detect the likelihood of a project’s vulnerability to corruption (Le et al., 2014). Given these shortcomings, this study attempts to develop a fuzzy evaluation model for measuring and predicting the proliferation and likelihood of corrupt practices at different stages of the procurement process of construction projects. Additionally, the present study maintains that the estimated indexes of the individual stages of the procurement process can be integrated into an overall vulnerability index to determine the susceptibility of a construction project to corrupt practices. To develop the model, this study was informed by the following objectives: 1) to identify the various activities that are performed at the specific stages of the procurement process; 2) to examine the susceptibility levels of the individual activities within their respective stages, and 3) to develop the measurement model for assessing the vulnerability level of construction projects using the fuzzy synthetic evaluation (FSE) technique. Theoretically, this study contributes to the body of knowledge regarding the systematic approaches of measuring the various indicators of corruption in construction and other infrastructure procurement as it is, arguably, the first study to employ soft computing techniques (i.e., the FSE approach) to estimate the susceptibility patterns of the various stages of the procurement process as well as develop a simple yet standardized approach to facilitate similar estimations in future works. Practically, even though a rigorous technique is employed, the model is developed in such a manner 3

ACCEPTED MANUSCRIPT that is easily understandable and can be adopted by practitioners (e.g. policymakers and auditors) for detecting and/or measuring the vulnerability levels of the various procurement activities and their respective stages of the procurement process to corruption. The model can also form the basis for researchers to develop more comprehensive tools that extend beyond the boundaries of the procurement process for predicting, measuring and offering effective prevention measures for corrupt practices right from the definition of project’s requirements through to contract close-out. The rest of the study is structured as follows: section 2 presents the literature review which focuses on the existing work on corruption measurement and the procurement process. Sections 3 and 4 present the study’s methodology and discussion respectively. Section 5 presents the study’s limitations and recommendations for future studies, and the final section concludes the study.

2. Measuring Corruption Analogous to the definitions of existing indexes, the measurement of corruption in this context is defined as the techniques of estimating, documenting and reporting the performances of countries and institutions regarding the proliferation or control of corruption (Heinrich and Hodess, 2011). In other words, the approximate computations of the process by which delegated authority is subverted and the resources provided under that specific authority are lost, expended or exploited in an unauthorised or illegitimate manner can be classified as the measurement of corruption. However, given the clandestine nature of corrupt practices, it is often difficult to detect, estimate and document corrupt acts with accuracy, confidence and diminutive amount of resources (Knack, 2006). On the other hand, the measurement of anti-corruption measures or models, unlike corruption, is considered to be relatively easier and straightforward. This is due to the supposition that unlike corruption forms which are relatively difficult to detect and measure, the measurement of anticorruption strategies considers constructs that are stipulated to check and expunge corrupt practices (UNDP, 2015). They, include variables such as legal stipulations, policy frameworks, institutional arrangements, practices, and mechanisms associated with the mentioned constructs (Trapnell, 2015). Therefore, whereas the measurement of corruption takes into consideration or measures the susceptibility levels, estimated costs and the associated effects of corruption, anti-corruption models 4

ACCEPTED MANUSCRIPT measure the efficiency and performance of a state or an institution against corruption (UNDP, 2015). These two measurements are, therefore, needed to draw explicit views of individual states and institutions regarding the prevalence and extirpation of corrupt practices (Shan et al., 2015; Trapnell, 2015; Philp, 2016). In order to develop both corruption and anti-corruption indexes, Trapnell (2004) suggest the consideration of four input variables in order to build a near-perfect and rigorous model. Some of these variables are somewhat or to a greater degree adopted by top global institutions such as the World Bank, Transparency International (TI) and World Economic Forum to estimate the levels of corruption in diverse contexts. The first construct to be considered concerns the type of data which include perceptions and experiences (administrative and external experiences). The second construct is related to the methods of soliciting for the primary data. They include general and experts’ surveys, monitoring and evaluation systems, crowdsourcing, field or compliance testing, and indicator-driven case studies. The third construct focuses on the required methodologies to be adopted such as social audit, citizens report cards and community scorecards as well as risk and integrity assessments. However, most of these composite measures, especially with regard to the indexes developed on the basis of perceptionbased data are often criticized as erroneous or imprecise with the justification that perceptions are often antipodal facts and, thus, may not represent the truth or at least may possess an appreciable level of distortion (provide reference/s). Moreover, , given the sensitive and serious nature of corruption, the ‘scars’ it leaves behind after its incidence can tarnish the image of the context or the domain in which it occurred, even when pragmatic efforts have been taken to expunge its spread. This case, however, tends to produce biased results anytime a general survey is conducted (Trapnell, 2015; Kaufmann et al., 2011). However, expert-driven data based on experience are regarded to possess a relatively greater degree of reliability due to the class of respondents involved (Trapnell, 2015). Since perception data are largely limited because respondents are typically unaware of the actual condition on the ground, the use of the expert’s survey experiences, external assessments and administrative stipulations such as public reports and law records which are more objective are preferred in order to ensure increased accuracy and reliability of

5

ACCEPTED MANUSCRIPT models (UNDP, 2015). This study also relied on expert-driven data to instigate a relatively higher degree of reliability. This is explicated further at the research design section.

2.1 The Procurement Process Different procurement systems may constitute different processes. However, the common activities involved in the traditional procurement process commences with the determination of what to purchase and end with the confirmation that the procured item or the final product received (i.e., goods, works or services) complies with the stipulated specifications (ISO, 2008; Ruparathna and Hewage, 2013). The primary consideration of every procurement system and process constitutes the method of price formation, the conditions of contract and the delivery method to be adopted (Eriksson and Westerberg, 2011). Construction managers and clients are, therefore, forced to decide on the type of procurement system to adopt right after determining the project’s aims and objectives (Sutt, 2011). According to Lædre et al. (2006), the failure to select a suitable procurement system may lead to adverse consequences of the entire process such as time and cost overruns and poor quality. In the course of strategizing a procurement process, Rowlinson and McDermott (2005) indicate that focus should not only be directed to the functional aspects of the process such as performance, conditions of contracts, law, client and the contract strategy but also external factors such as motivation, satisfaction, leadership, learning, political environment, sustainability, and culture. The basic activities that define each stage of the process are presented in Figure 1. Lastly, despite the studies conducted on the various forms of activities involved in the procurement process, few effort has been expended to assess vulnerability of the individual activities to corrupt practices, even though the leading setback to any procurement process is the exposure of the stages to corruption as well as the proliferation of corrupt practices throughout the entire process (Krishnan, 2010; Tabish and Jha, 2011; Doig, 2003). The specific activities with their associated stages captured under the procurement process are presented in Figure 1.

6

ACCEPTED MANUSCRIPT

Figure 1: The Procurement Process (adapted from Lester, 2007; Ruparathna and Hewage, 2014)

2.2 Corruption in Infrastructure Procurement (IP) IP forms the backbone of every economy globally, and it is critical to the survival and livelihood of humanity. Ranging from all kinds of structures (hospitals, roads, dams, etc.) to access to potable drinking water are all forms of infrastructure, and when budgets allocated to procure these needs of humanity are misappropriated, the net result is a socio-economic setback. Researching corruption in IP is therefore crucial. The World Bank defines corruption as the abuse of public (entrusted) power for private benefit (Tanzi, 1998). Corruption is regarded as the dark side of the relationship between personal interest and organizational goals (Tabish and Jha, 2011). In IP, corruption happens when the client’s representative assigned to carry out the IP activities collude with a third party, usually, suppliers or contractors to circumvent or break stipulated laws (second party) to satisfy personal interest (Boyd and Padilla, 2009). Conventional wisdom tells that “corruption does not happen in a vacuum but involves a medium to thrive” (Owusu and Chan 2017; Boyd and Padilla 2009). The procurement process represents one of the vulnerable vacuums for corrupt practices to take place. Boyd and Padilla (2009) also postulate that corruption which is often regarded as the converse of integrity is made up of three constituting factors, namely the demand side, that requests for an 7

ACCEPTED MANUSCRIPT illegitimate gain, the supply end that adheres to the request and grants it; and the condoning side that freely allows the demand and supply sides to have their way, but does nothing to prevent it. Several studies have concluded that public procurement is most prone to corruption (Ameyaw et al., 2017; Søreide, 2002; Transparency International, 2005; Tabish and Jha, 2011; Zhang et al., 2016) and is considered to be a vulnerable area to corrupt practices because of the large amount of public resources involved, the asymmetric information portraying the decision-making process and the intrinsic incompleteness of contracts (Cavalieri, 2016). Conventionally speaking, it has been identified that none of the IP stages (Figure 2) is immune to corrupt practices (Transparency International, 2005). The complexity of the procurement process in the pre-contract, contract and post-contract phases makes it vulnerable to corrupt practices (Heggstad et al., 2010). For instance, during the pre-qualification and tendering phases, it is possible for the client representatives to either bend or amend procurement rules to favour preferred tenderers in exchange for bribes (Ameyaw et al., 2017; Osei-Tutu et al., 2010). Unavoidably, the process of awarding a contract can also be highly influenced by the power relations of both economic and political organizations whose aim is to maintain the existing state of affairs (Sargiacomo et al., 2015, World Bank, 2013; Neupane et al., 2012). Other forms and examples of corrupt practices are reported later in this study. The impact of corruption on IP is grave. Corruption researchers have identified some negative impacts of this menace that impair the entire procurement process, leading to a decline in the lifespan of projects, abandonment of projects, and the collapse of completed infrastructures, etc. (Osei-Tutu et al., 2010; Boyd and Padilla, 2009). For instance, Wang et al. (1999) identified corruption as one of the primary risk factors in the Chinese BOT projects. In developing countries, several infrastructure projects have been abandoned due to corrupt practices in the public procurement sector (Mawenya, 2008; Bowen et al., 2012). Ameh and Odusami (2009) reported on a strong correlation between corruption prevalence and poor growth, development, and performance in developing countries. Therefore, governments who aim to improve their stock of infrastructure through the PPP strategy must fight corruption in public procurement before they can succeed in their quest to establish successful PPPs (Otairu et al., 2014). The last two decades have seen substantial contributions from scholars who have conducted and published many studies addressing the issue of corruption in IP. 8

ACCEPTED MANUSCRIPT

3. Research Design As mentioned in the previous section, expert data were solicited to determine the indicators of the model. First, primary data regarding the vulnerability of the procurement process were solicited from experts directly involved in procurement and construction management of construction works. A questionnaire survey was conducted to solicit views from the experts concerning the constructs and the indicators that are needed to arrive at the model. Prior to the development of the questionnaire, a thorough, systematic review of the literature (e.g. Le et al. 2014; Lester, 2007; Ruparathna and Hewage, 2014; Chan and Owusu 2017) was conducted to arrive at the variables or the activities within the procurement process which are used as the indicators to develop the constructs of the model. Since the aim of the study is to measure the indicators regarding the susceptibility of the procurement process, the review conducted led to the retrieval of 21 traditional activities captured under four procurement stages (Fig. 1). The 21 activities were considered as traditional, because they can be identified in most, if not all, of the procurement systems, comprising the design and construct method, management procurement, on-call contracting, guaranteed maximum price, full cost reimbursable, and total package options among others (Ruparathna and Hewage, 2013). The survey respondents were asked to rate the levels of vulnerability of each of the 12 activities based on a five-point grading scale (1 = very low vulnerability and 5 = very high vulnerability). Further, a section was provided in the questionnaire for the respondents to provide any known critical activity or stage of the procurement process that was not captured in the questionnaire. Questionnaires were adopted because they offer a valid and reliable source of information and is less costly (Hoxley, 2008). Moreover, a questionnaire survey, to a large extent, warrants anonymity and the protection of respondents’ data, especially on a sensitive topic of this nature (Chan et al., 2017; Ameyaw et al., 2017).

3.1 Survey Participants The respondents for this survey comprise expert practitioners involved in the procurement and delivery of construction projects in Ghana. Therefore, one of the requirements in getting the apropos experts was limited to the possession of knowledge on the processes involved in the procurement and management 9

ACCEPTED MANUSCRIPT of construction works, and familiarity with the dynamics and stipulations or unethical practices binding ethical or legal stipulations respectively. Purposive sampling approach was therefore adopted for this kind of study. The snowballing technique was also initiated at some point, because the respondents were requested to help disseminate the questionnaires to their expert colleagues. Even though the topic of corruption may be considered as a general concern and therefore the questionnaire could be answered by any person from the general public, it often becomes difficult to explicate, especially when viewed under a specific context. For instance, the subject of corruption observed in this study is captured under the domain of construction procurement and management. Therefore, any potential respondent should be able to understand the complexities involved in the procurement process from the pre-contract stage to the post-contract stage and the construction process which includes project design and execution, contract closeout, and dispute resolution. This kind of knowledge is not common to the general masses even though the subject of corruption remains a social issue and some may understand the concept of corruption in the general setting but not in a defined context of expertise. After the pilot study, expert survey, retrieval of the questionnaires and the assessment of the initial data to address any discrepancy that may negatively affect the data; 62 responses were regarded to be valid and suitable for further analysis. The survey was conducted practicing construction professionals including civil engineers, quantity surveyors, architects, contractors and academics. Whereas the academics involved were identified through their publications on the subject matter, the industrial experts are senior managers of both private and public sector domains and are involved in the procurement, execution and the management of construction and other infrastructure works. The overall biodata of the respondents is presented in tables 1 and 2.

10

ACCEPTED MANUSCRIPT Table 1: Respondents’ data Construct Sub-construct

Frequency

Sector

Public Private Both Total

20 30 12 62

Relative Frequency 32.26 48.39 19.35 100

Cumulative Frequency 32.26 80.65 100.00

Professional background

Civil Engineer Quantity Surveyor Contractor Architect Academics Total

17 31 4 7 3 62

27.42 50.00 6.45 11.29 4.84 100

27.42 77.42 83.87 95.16 100.00

Years of experience

Up to 10 years 11-20 years 21-40 years Total

45 12 5 62

72.58 19.35 8.06 100

72.58 91.94 100.00

Involvement in procurement stages

Single stage Multiple stages All stages Total

7 28 27 62

11.29 45.16 43.55 100

11.29 56.45 100.00

Position in organization

Head of Department Director of Works Senior Manager Supervisor Junior staff Total

5 8 17 28 4 62

8.06 12.90 27.42 45.16 6.45 100

8.06 20.97 48.39 93.55 100.00

3.2 Data Analysis and Results With the aim of examining over 21 procurement activities to develop the required indicators to finally arrive at the development of the measurement model, it is expedient to adopt the required techniques and tools to adequately execute the tasks at hand. Five different tools were adopted to conduct the pretest and the main analysis of the data (i.e., the reliability test tool was employed to assess the data reliability). After the reliability test was successfully conducted, the descriptive and the frequency tools in SPSS version 23 were also adopted to estimate the individual criticalities of the variables. The estimation of the respective mean values of each of the activities facilitated the computation of the criticalities of the constructs and the development of the fuzzy model. Finally, the FSE technique was employed to estimate the vulnerability indicator or index for each variable (procurement activity), the overall vulnerability index and the model to assess the vulnerability indexes for future projects. The 11

ACCEPTED MANUSCRIPT explications of the adoption and the rationale for the adoption of all the tools employed are conducted in the subsequent sections.

3.2.1 Pre-test The above-mentioned pre-test (i.e., the data reliability) was conducted, because it determines whether further analysis can be conducted. For instance, in the case where the reliability test produces an unfavourable result, it signals a clear indication that the data is statistically unreliable. In essence, the Cronbach alpha test examines the weighted correlation and the internal consistency among the variables of a given set of data to evaluate the reliability. Given a range of 0 to 1, where 0 represents no reliability and 1 indicates full reliability; the closer the alpha value is to 1, the higher the reliability and vice versa. However, according to Nunally (1978), the acceptable criterion for any statistical scale reliability should not be less than 0.7. The pre-test estimation of this study’s reliability was 0.9888 which indicates very strong reliability of the responses provided by the experts.

3.2.2 Determining Mean values Following the examination of scale reliabilities, the individual levels of the proneness of the IP activities to corruption were determined using the mean scores (MS) approach. As this tool has also been widely employed in both CM and corruption related research studies, this study employed the MS technique to determine the vulnerability indexes of the individual activities within the stages of the procurement process. The mean values range from 1 to 5 (extrapolated from the Likert-grade points). Therefore, the closer the MS of any activity is to 5, the higher the level of its susceptibility. The aim is to determine which activities are more vulnerable to corrupt practices so as to divert specific attention towards the expurgation of corrupt practices at the critical stages of the procurement process.

3.3 Explication and Application of the FSE Technique The FSE technique is regarded as a branch of the fuzzy set theory developed by Zadeh (1965) which is often employed to quantify multi-attributes and multivariate (Osei-Kyei and Chan 2017; Hu et al. 2016). According to Ameyaw and Chan (2017) and Khatri et al. (2011), the FSE is a fuzzy multicriteria 12

ACCEPTED MANUSCRIPT decision-making approach whereby individual factors of a construct are synthesized into a single score. Moreover, the opinions on the vulnerability levels of the procurement process by the experts are considered to be uncertain and typically subjective (Shan et al., 2015). Accordingly, FSE uses linguistic expressions (terms or variables) to represent and capture experiential knowledge of the survey respondents (Boussabaine, 2014), which helps to resolve uncertainty and subjectivity associated with the responses of the survey respondents (Ameyaw and Chan, 2015). The use of linguistic variables enables the respondents to qualitatively assess the vulnerability levels of the 21 procurement activities (Boussabaine 2014). The FSE technique employs the application of membership degrees in a given set instead of a strict true or false membership (Tah and Carr, 2000). Simply put, rather than using absolute terms or values such as 0 and 1 to represent an elements association to a fuzzy set, the FSE tool expresses the element’s belongingness to a fuzzy set in terms of varying degree of relation. The degree or extent of membership can, therefore, consider any value within a closed range of 0 and 1 and the obtained value characterizes the degree or measure to which the element belongs to a fuzzy set (Ameyaw et al., 2015; Tar and Carr, 2000; Kasirolvalad et al., 2006). While it is impossible to meaningfully represent these linguistic terms with a single value, the FSE technique allows for these variables to be expressed in a mathematical logic (Zadeh, 1975; Lam et al., 2007; Ameyaw and Chan, 2015). Thus, enabling the FSE tool to be used for quantifying and modelling the linguistic terms for vulnerability levels of the procurement phase. Representing and modelling experiential knowledge using linguistic variables within FSE facilitates decision-making (Lam et al., 2007; Boussabaine, 2004; Ameyaw and Chan, 2015). The evaluation of the irregularities associated with the procurement process of construction infrastructure works is considered to be a multifaceted issue, clouded with uncertainty and imprecision (Ameyaw and Chan, 2015; Tah and Carr, 2000). The difficulty associated with the assessment of the susceptibility levels stem from the experts' subjective judgement coupled with the vague non-mathematical assertions (Tah and Carr, 2000; Ameyaw and Chan, 2015) regarding the susceptibility levels of the individual activities and stages of the procurement process. Subjectivity often emanates from partial and unquantifiable information

13

ACCEPTED MANUSCRIPT (Sadiq and Rodriguez, 2004). The FSE technique has the capacity to rectify the vagueness present in the experts' subjective assertions and rationalizes the fuzziness associated with the the susceptibility assessments. Thus, the FSE converts the subjective expressions of the experts into a much more objective, quantifiable and crisp output (Ameyaw and Chan, 2015; Lo et al., 1999). Moreover, as discussed by Yeung et al. (2010) and Liu et al. (2013), the FSE technique is noted for evaluating multicriteria decision process. Issues regarding decision-making are considered multi-criteria granted that there are several criteria, inaccurate data, uncertainties, several decisions makers and inadequate information surrounding the process of making the decision (Ameyaw and Chan, 2015). Thus, conjecturing the susceptibility level of a project's procurement process to corruption remains a multicriteria decision process where the activities encapsulated in the respective stages or constructs are evaluated by multiple experts per the aforementioned linguistic variables. In evaluating a project's procurement susceptibility to corruption, the individual activities that are interrelated are categorized under their respective stages or constructs (Lo1999, Ameyaw and Chan, 2015). Thus, a stage is made up of one or more activities as presented in Fig. 1. However, the overall susceptibility level is determined by defuzzifying the membership grades of the individual stages that make up the procurement process. This is explicated later at the FSE application section. This technique was employed by Shan et al. (2015) to examine corruption in the Chinese public projects and has been employed to by other researchers (e.g., Ameyaw and Chan 2015; Osei-Kyei et al. 2017) to examine different subject matters within the domain of construction management research as well as other corruption-related studies. The assessment of the overall vulnerability level of a project to corruption encapsulates the examination of the individual principal constructs of the procurement process (i.e., PSC, CTS, CAS, PCP) levels of vulnerability on one level and the activities within each construct or procurement stage on another level. In this instance, each principal component or construct is examined with respect to its vulnerability level which leads to the quantification of the overall vulnerability level of the procurement process. The multi-level fuzzy evaluation technique is often adopted to evaluate this multi-construct and multilevel challenges inherent in assessing project’s vulnerability to the incidence of corrupt practices. Moreover, as explained earlier, given that the determination of the susceptibility or 14

ACCEPTED MANUSCRIPT vulnerability index of an infrastructure project is by nature fuzzy and often drawn on the subjective judgement of the experts, the FSE technique is deemed suitable (Boussabaine 2014; Ameyaw et al. 2015). This study also encapsulates three levels of vulnerabilities which are: vulnerabilities associated with procurement activities (level 1); vulnerabilities associated with procurement stages of constructs (level 2) and the overall vulnerability associated with the entire procurement process (level 3). Inferring from these three levels, it is, therefore, apropos to employ the FSE technique to develop a vulnerability index and assessment model. Lastly, The FSE technique was most preferable as it provides an objective index as compared to the ordinary weighting method. The steps for developing the vulnerability index are outlined below.

3.3.1. Developing an assessment Index System Given the four constructs of the procurement process, the evaluation system for computing the index can be developed by establishing the construct stages as the first or primary level index system as V= (V1, V2, V3… Vm) (Ameyaw and Chan 2015; Shao 2004). In this context, they are labelled as (VPCS, VCTS, VCAS and VPCP). The variables or activities within their respective procurement constructs are also defined as secondary or second level index system as: VPCS = {𝑣𝑃𝐶𝑆1, 𝑣𝑃𝐶𝑆2, 𝑣𝑃𝐶𝑆3, 𝑣𝑃𝐶𝑆4, 𝑣𝑃𝐶𝑆5, 𝑣𝑃𝐶𝑆6} VCTS = {𝑣𝐶𝑇𝑆1, 𝑣𝐶𝑇𝑆2, 𝑣𝐶𝑇𝑆3, 𝑣𝐶𝑇𝑆4, 𝑣𝐶𝑇𝑆5} VCAS = {𝑣𝐶𝐴𝑆1, 𝑣𝐶𝐴𝑆2, 𝑣𝐶𝐴𝑆3, 𝑣𝐶𝐴𝑆4}, and VPCP = {𝑣𝑃𝐶𝑃1, 𝑣𝑃𝐶𝑃2, 𝑣𝑃𝐶𝑃3, 𝑣𝑃𝐶𝑃4, 𝑣𝑃𝐶𝑃5, 𝑣𝑃𝐶𝑃6} The index systems are considered as the input variables for the fuzzy synthetic analysis (Ameyaw et al. 2017). The respondents were therefore required to rank the individual activities using the 5-point Likert grading system as V= (1,2,3,4,5) where 1= very low (VL), to 5= extremely high (EH). Adopting the five-point grading scale in this study is consistent with past studies (Shan et al. 2015; Zhao et al. 2015; Osei-Kyei and Chan 2017). The Likert scale representing the linguistic terms which facilitates the judgement of the experts involved in the study and the determination of the membership functions are presented in Table 2. 15

ACCEPTED MANUSCRIPT

Table 2: Input Variables’ Linguistic Terms Scale Term Degree of capability 1 Very low 0 – 0.25 2 Low 0 – 0.50 3 Neutral 0.25 – 0.75 4 High 0.50 – 1.00 5 Extremely high 0.75 – 1.00 Adapted from Ameyaw et al. (2017)

Constant 0.125 0.250 0.500 0.750 0.875

3.3.2 Estimations of Input Variables’ Weightings. The individual weightings of the variables or activities within the four stages of the procurement process were determined using the normalization technique. The formula for determining the normalized values or the specific weighting is given as (Lo, 1999): 𝑤𝑖 =

𝑀𝑖

5

, 0 < 𝑤𝑖 < 1, 𝑎𝑛𝑑 ∑𝑖 = 1𝑤𝑖 = 1

(1)

∑ 5 𝑀𝑖 𝑖=1

Where 𝑤𝑖 represents the weighting of the activities within the procurement stages or constructs i; 𝑀𝑖, represents the mean index or value of the specific activity or associated stage i generated from the survey analysis. The function set of the weights is thus given by: Wi = {𝑤1, 𝑤2, 𝑤3,… 𝑣𝑚}

(2)

Therefore, using the ‘check for proof of delivery’ (𝑃𝐶𝑃4), which is the fourth activity in the postcontract phase and equation (1), the weighting is calculated as follows: 3.76

3.76

𝑊𝑃𝐶𝑃4 = 3.48 + 3.51 + 3.48 + 3.76 + 3.45 + 3.39 = 21.07 = 0.178 Using the same formula, the remaining weightings of the individual activities within their respective constructs are evaluated. The summation of the weightings of a group of activities within the same construct must be equal to the value of one. A typical example using the PCP construct is presented below:

16

ACCEPTED MANUSCRIPT 6

∑𝑊

𝑃𝐶𝑃

= 0.165 + 0.167 + 0.165 + 0.178 + 0.164 + 0.161 = 1.00

𝑘=1

17

Table 3: MF for Stages and respective activities of the Procurement Process Procurement Procurement Activities Stages 1 Pre-Contract Define requirements, vpcs1 Procurement process planning and strategy development, 2 stage vPCS2 3 Pre-tender survey, vPCS3 4 Obtaining necessary approvals, vpcs4 5 Soliciting tenders, vpcs5 6 Receipt of tenders, vpcs6 Pre-tender meeting (Establishing Evaluation Criteria, 7 Contract Evaluation Plan, Evaluation Criteria), vcts1 Stage 8 Tender evaluation (review to approve or reject bids), vcts2 9 Select contractor, vcts3 10 Award contract/Purchase order, vcts4 11 Preparation and Signing of Contract, vcts5 Contract 12 Issuing contract amendments, vcas1 administrati Monitor Progress, vcas2 13 on stage 14 Follow up delivery, vcas3 15 Administer Progress payments, vcas4 16 File final action Contractor agreement to final claim, pcp1 Post contract 17 Issue final contract amendment, vpcp2 phase 18 Complete of financial audits, vpcp3 19 Check for proof of delivery, vpcp4 20 Return of performance bonds and close-out, vpcp5 Ensure completeness and accuracy of file documentation, 21 vpcp6 Total mean and weighting values

18

Code

Mean

Weighting

2.63 3.05

NValue 0.00 0.27

PCS1 PCS2 PCS3 PCS4 PCS5 PCS6 CTS1

3.15 3.52 4.02 3.35 3.40

0.33 0.56 0.88 0.46 0.49

0.160 0.178 0.204 0.170 0.182

CTS2 CTS3 CTS4 CTS5 CAS1 CAS2 CAS3 CAS4 PCP1 PCP2 PCP3 PCP4 PCP5 PCP6

4.00 4.21 3.74 3.35 3.13 3.39 2.97 3.68 3.48 3.51 3.48 3.76 3.45 3.39

0.87 1.00 0.70 0.46 0.32 0.48 0.22 0.66 0.54 0.56 0.54 0.72 0.52 0.48

0.214 0.225 0.200 0.179 0.238 0.257 0.226 0.279 0.165 0.167 0.165 0.178 0.164 0.161

Total mean of stages

Weighting of stages

19.72

0.271

18.70

0.257

13.17

0.181

21.07

0.290

72.66

1.000

0.133 0.155

ACCEPTED MANUSCRIPT 3.3.3 Determining the Membership Functions of the Input Variables for the Activities Membership functions (MFs) in fuzzy set theory represent the extent or degree (within the range of 0 and 1) of an element’s membership in a fuzzy set (Xu et al. 2010). Thus, regarding the MFs of the input variables, the FSE technique employs the application of membership degrees in a given set instead of a strict true or false membership (Tah and Carr, 2000). Rather than using absolute terms or values such as 0 and 1 to represent an elements association to a fuzzy set, the FSE tool expresses the element’s belongingness or membership to a fuzzy set in terms of varying degree of relation. The degree or extent of membership can, therefore, consider any value within a closed range of 0 and 1; the obtained value characterizes the degree or measure to which the element belongs to a fuzzy set (Ameyaw et al., 2015; Tar and Carr, 2000; Kasirolvalad et al., 2006). They are derived from the respondents’ assessments of the 21 procurement activities using the Likert scale (as discussed). According to Ameyaw and Chan (2016), it is appropriate to designate the various levels where the membership functions are derived. Therefore, as established earlier, the linguistic terms for examining the input variables (i.e., the procurement activities) against the vulnerabilities constructs were determined using the 5-point grading system as 𝑙 = (1,2,3,4,5), where 𝑙1 = very low, 𝑙2 = low, 𝑙3 = moderate, 𝑙4 = high, 𝑙5 = very high. The membership function of a given vulnerability construct was derived using the formula below:

𝑀𝐹𝑣𝑖𝑛 =

𝑥1𝑣

𝑖𝑛

𝑙1

+

𝑥2𝑣

𝑖𝑛

𝑙2

+

𝑥3𝑣

𝑖𝑛

𝑙3

+

𝑥4𝑣

𝑖𝑛

𝑙4

+

𝑥5𝑣

𝑖𝑛

𝑙5

𝑥1𝑣

𝑖𝑛

= 𝑣𝑒𝑟𝑦 𝑙𝑜𝑤 +

𝑥2𝑣

𝑖𝑛

𝑙𝑜𝑤

𝑥3𝑣

𝑖𝑛

(3)

+.... + 𝑣𝑒𝑟𝑦 ℎ𝑖𝑔ℎ

Where 𝑀𝐹𝑣𝑖𝑛 represents the membership function (MF) of a specific activity of the procurement process 𝑣𝑖𝑛; 𝑥𝑦𝑣𝑖𝑛(y=1,2,3,4,5) indicates the percentage of a given score y for an activity of a given construct as assigned by the experts (i.e., 𝑣𝑖𝑛); and 𝑥𝑦𝑣𝑖𝑛/𝑙1 denotes the relation between 𝑥𝑦𝑣𝑖𝑛and its respective grade alternative. Therefore, referring to equation 3, the membership function of a specific construct can be written as:

𝑀𝐹𝑣𝑖𝑛 = (𝑥1𝑣𝑖𝑛, 𝑥2𝑣𝑖𝑛, 𝑥3𝑣𝑖𝑛, 𝑥4𝑣𝑖𝑛, 𝑥5𝑣𝑖𝑛).

(4)

19

ACCEPTED MANUSCRIPT Therefore, using the contractor selection stage (CTS3) as an example owing to the ratings by the experts (i.e., 0.00%, 1.60%, 12.90%, 48.40%, 37.10%), the membership function is evaluated as follows:

0.00

0.02

0.13

0.48

0.37

𝑀𝐹𝑐𝑡𝑠3 = 𝑁𝑜𝑡 𝑉𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑙𝑒 + 𝐿𝑒𝑠𝑠 𝑉𝑢𝑙𝑛𝑒𝑟𝑎𝑙𝑒 + 𝑁𝑒𝑢𝑡𝑟𝑎𝑙 + 𝑉𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑙𝑒 + 𝐸𝑥𝑡𝑟𝑒𝑚𝑒𝑙𝑦 𝑉𝑢𝑙𝑛𝑒𝑟𝑎𝑏𝑙𝑒

The MF is, therefore, presented as (0.00, 0.02, 0.13, 0.48, 0.37). The individual members within the function range between 0 and 1, and the summation of the members must be equal to 1:

∑5

𝑥 𝑘 = 1 𝑘𝑣𝑖𝑛

(5)

=1

This was repeated for all the procurement activities within their respective constructs in order to determine their respective membership functions. The MF of all the individual activities of the procurement process were classified under level three as presented in Table 5.

3.3.4 Determination of the Membership Functions for the Constructs/Stages (Level 2) The next step after the determination of the membership functions at level three was an evaluation of the membership function at level two. That is, membership function for the main procurement stages or constructs. The derivation of the membership function is derived by the formula below:

(6)

𝐷 = 𝑊𝑖●𝑅𝑖

Where 𝑊𝑖 represents the individual weightings of all the activities within their respective constructs or stages and 𝑅𝑖 represents the fuzzy evaluation matrix. Following a similar approach used previously in eqn. (1), the weightings were estimated using the formula below:

𝑀𝐹𝑣𝑖𝑛 =

𝑀𝑖

5

, 0 < 𝑤𝑖 < 1, 𝑎𝑛𝑑 ∑𝑖 = 1𝑤𝑖 = 1

(7)

∑ 5 𝑀𝑖 𝑖=1

20

ACCEPTED MANUSCRIPT

Using CTS as an example, the weightings for the constructs were estimated as follows:

18.70

18.70

𝑊𝐶𝑇𝑆 = 19.72 + 18.70 + 13.17 + 21.07 = 72.66 = 0.257

The remaining three constructs (i.e., the PCS, CAS and PCP) were computed in a similar as described above . In calculating for the individual weightings of the constructs, the mean scores of each stage or construct was normalized to ascertain their respective weightings where the summations of all the weightings equate to 1. For instance, the calculation for the four stages or constructs were evaluated as follows: 19.72

19.72

13.17

13.17

21.07

21.07

𝑊𝑃𝐶𝑆 = 19.72 + 18.70 + 13.17 + 21.07 = 72.66 = 0.271 𝑊𝐶𝐴𝑆 = 19.72 + 18.70 + 13.17 + 21.07 = 72.66 = 0.181 𝑊𝑃𝐶𝑃 = 19.72 + 18.70 + 13.17 + 21.07 = 72.66 = 0.290

∑4

𝑤𝑖 = 0.271 + 0.257 + 0.181 + 0.290 = 1

𝑖=1

Therefore, the weightings can be presented as follows: 𝑤𝑖 = (𝑤1,𝑤2,𝑤3,…𝑤𝑖) = (0.271, 0.257, 0.181, 0 .290)

3.3.5 Establishment of the Multi-level and Multi-criteria FSE Model Given that the evaluation of the susceptibility levels of the procurement process is a multi-criteria (that is, from activity to activity and stage to stage) and a multi-level (that is, from the activity level to the construct or stage level), the evaluation involves three primary stages. The first stage deals with the establishment of the MFs and the estimated weighted functions (𝑤𝑖) of the individual activities (for example: PCS1, PCS2, CTS4, CTS3, CAS1). This is known as the lower level and is based on the assessments from the experts’ survey. The second stage also deals with the establishment of the membership and weighted functions of the individual constructs or stages (PCS, CTS, CAS and PCP). 21

ACCEPTED MANUSCRIPT The evaluation of the respective impacts of the stages are therefore calculated at this level. The third and final stage estimates the overall vulnerability index of the procurement process which is presented by a single index or value. Therefore, with the establishment of the MFs and the estimated 𝑤𝑖 of the individual activities (3rd level) which is obtained from the experts’ responses, the estimations of the constructs’ indexes (2nd level) and the overall vulnerability index (1st level) are discussed. With reference to eqn. (6), to determine the vulnerability levels of the respective stages of the procurement process, 𝐷𝑖, which represents a fuzzy matrix is first established for each of the procurement stages after the derivation of the membership functions of the activities within the stages. Therefore, following eqn. 4 [𝑀𝐹𝑎𝑖𝑛 = (𝑥1𝑎𝑖𝑛 + 𝑥2𝑎𝑖𝑛 + 𝑥3𝑎𝑖𝑛 + 𝑥4𝑎𝑖𝑛 + 𝑥5𝑎𝑖𝑛)] the membership functions for all the activities within their respective stages can be demonstrated in a fuzzy matrix presented below:

| ||

𝑀𝐹𝑣𝑖1 𝑥1𝑣𝑖1 𝑀𝐹𝑣𝑖2 𝑥1𝑣𝑖2 𝑅𝑖= 𝑀𝐹𝑣𝑖3 = 𝑥1𝑣𝑖3 ⋯ ⋯ 𝑀𝐹𝑣𝑖𝑛 𝑥1𝑣𝑖𝑛

𝑥2𝑣𝑖1 𝑥2𝑣𝑖2 𝑥2𝑣𝑖3 ⋯ 𝑥2𝑣𝑖𝑛

𝑥3𝑣𝑖1 𝑥3𝑣𝑖2 𝑥3𝑣𝑖3 ⋯ 𝑥3𝑣𝑖𝑛

𝑥4𝑣𝑖1 𝑥4𝑣𝑖2 𝑥4𝑣𝑖3 ⋯ 𝑥4𝑣𝑖𝑛

|

𝑥5𝑣𝑖1 𝑥5𝑣𝑖2 𝑥5𝑣𝑖3 ⋯ 𝑥4𝑣𝑖𝑛

(8)

Where the elements are presented by 𝑥1𝑣𝑖1. Using the contract stage (CTS) as an example, the elements in the fuzzy matrix can be presented through Eqn. 8 as in:

| ||

𝑀𝐹𝑣21 0.15 0.13 𝑀𝐹𝑣22 0.02 0.03 𝑅𝑐𝑡𝑠= 𝑀𝐹𝑣23 = 0.00 0.02 𝑀𝐹𝑣23 0.10 0.06 𝑀𝐹𝑣25 0.08 0.11

0.16 0.18 0.13 0.08 0.32

0.31 0.48 0.48 0.52 0.34

0.26 0.29 0.37 0.24 0.15

|

It must be emphasized that the FSE technique comprises three levels of the membership functions (i.e., from the third level to the first level). The computations at this section are used to arrive at level 2. The derivation of the fuzzy matrix 𝐷𝑖 is arrived with the application of the function set of the individual weightings 𝑤𝑖 = (𝑤1,𝑤2,𝑤3,…𝑤𝑖) of the activities within their respective stages or construct i as presented below:

22

ACCEPTED MANUSCRIPT

|

𝑥1𝑣𝑖1 𝑥1𝑣𝑖2 𝐷𝑖=𝑊𝑖●𝑅𝑖=(𝑑𝑖𝑛,𝑑𝑖𝑛,𝑑𝑖𝑛,…𝑑𝑖𝑛)=(𝑤1,𝑤2,𝑤3,…𝑤𝑖)● 𝑥1𝑣𝑖3 ⋯ 𝑥1𝑣𝑖𝑛

𝑥2𝑣𝑖1 𝑥2𝑣𝑖2 𝑥2𝑣𝑖3 ⋯ 𝑥2𝑣𝑖𝑛

𝑥3𝑣𝑖1 𝑥3𝑣𝑖2 𝑥3𝑣𝑖3 ⋯ 𝑥3𝑣𝑖𝑛

𝑥4𝑣𝑖1 𝑥4𝑣𝑖2 𝑥4𝑣𝑖3 ⋯ 𝑥4𝑣𝑖𝑛

|

𝑥5𝑣𝑖1 𝑥5𝑣𝑖2 𝑥5𝑣𝑖3 ⋯ 𝑥4𝑣𝑖𝑛

=(𝑑𝑖1,𝑑𝑖2,𝑑𝑖3,…𝑑𝑖𝑛)

(9)

Where 𝑑𝑖𝑛represents the grade alternative membership degree, vin with regard to a specific stage of the procurement process; and ● connotes the fuzzy composite operation (Ameyaw and Chan 2016; Lo et al. 1999). Taking into consideration the individual weightings of the activities within 𝑐𝑡𝑠 (i.e., construct CTS), the matrix 𝑅𝑐𝑡𝑠 is normalized using equation 9 above.

| |

𝑀𝐹𝑎21 𝑀𝐹𝑎22 𝐷𝑐𝑡𝑠 = 𝑊𝑐𝑡𝑠●𝑅𝑐𝑡𝑠=(𝑤𝑐𝑡𝑠1,𝑤𝑐𝑡𝑠2,𝑤𝑐𝑡𝑠3,𝑤𝑐𝑡𝑠4,𝑤𝑐𝑡𝑠5) × 𝑀𝐹𝑎23 𝑀𝐹𝑎23 𝑀𝐹𝑎25

|

0.15 0.02 𝑇ℎ𝑒𝑟𝑒𝑓𝑜𝑟𝑒, 𝐷𝑎𝑐𝑡𝑠 = (0.182,0.214,0.225,0.200,0.179) × 0.00 0.10 0.08

0.13 0.03 0.02 0.06 0.11

0.16 0.18 0.13 0.08 0.32

0.31 0.48 0.48 0.52 0.34

0.26 0.29 0.37 0.24 0.15

|

= (0.06, 0.07, 0.17, 0.43, 0.27)

The same approach was used to derive the membership function for all the other respective constructs or stages at level two. Table 4, therefore, presents the derivations of the membership functions at both level two and three.

23

Table 4: MF for Stages of the Procurement Process Procurement Code Activities’ MF for Level 3 Process Weightings Pre-Contract PCS1 0.133 0.29, 0.16, 0.23, 0.27, 0.05 stage PCS2 0.155 0.13, 0.24, 0.19, 0.32, 0.11 PCS3 0.160 0.15, 0.18, 0.23, 0.29, 0.16 PCS4 0.178 0.11, 0.13, 0.11, 0.42, 0.23 PCS5 0.204 0.03, 0.05, 0.11, 0.48, 0.32 PCS6 0.170 0.13, 0.10, 0.24, 0.35, 0.18 0.182 0.15, 0.13, 0.16, 0.31, 0.26 Contract Stage CTS1 CTS2 0.214 0.02, 0.03, 0.18, 0.48, 0.29 CTS3 0.225 0.00, 0.02, 0.13, 0.48, 0.37 CTS4 0.200 0.10, 0.06, 0.08, 0.52, 0.24 CTS5 0.179 0.08, 0.11, 0.32, 0.34, 0.15 CAS1 0.238 0.13, 0.19, 0.16, 0.42, 0.08 Contract administration CAS2 0.257 0.10, 0.15, 0.19, 0.40, 0.16 CAS3 0.226 0.15, 0.21, 0.24, 0.34, 0.06 stage CAS4 0.279 0.06, 0.10, 0.16, 0.45, 0.23 PCP1 0.165 0.08, 0.13, 0.19, 0.42, 0.18 Post contract PCP2 0.167 0.05, 0.18, 0.18, 0.39, 0.19 phase PCP3 0.165 0.08, 0.08, 0.24, 0.47, 0.13 PCP4 0.178 0.05, 0.08, 0.18, 0.45, 0.24 PCP5 0.164 0.10, 0.08, 0.24, 0.44, 0.15 PCP6 0.161 0.08, 0.18, 0.18, 0.40, 0.16 Note: MF connotes membership function

MF for Level 2 0.13, 0.14, 0.18, 0.37, 0.19

Constructs’ Weightings 0.13

0.06, 0.07, 0.17, 0.43, 0.27

0.06

0.11, 0.16, 0.19, 0.41, 0.14

0.11

0.07, 0.12, 0.20, 0.43, 0.18

0.07

24

MF for Level 1 0.09, 0.12, 0.19, 0.41, 0.20

ACCEPTED MANUSCRIPT After determining each of the constructs’ fuzzy matrix, the vulnerability level of each of the procurement stage can be estimated using the formula below (eqn. 10).

5

𝑉𝐿𝑖 = ∑𝑘 = 1𝐷 × 𝐿𝑡 = (𝐷'1,𝐷'2,𝐷'3,𝐷'4,𝐷'5) × (1,2,3,4,5)

(10)

1 ≤ VLi ≤ 5

Since an analogous formula is employed to estimate the overall vulnerability index for the entire procurement process, the formula is better explained at the overall estimation section. However, using the formula above, the individual index for each stage (i.e. the vulnerability level for the individual procurement stages) is computed as follows:

𝑉𝐿𝑝𝑐𝑠 = [(0.13 × 1) + ( 0.14 × 2) + ( 0.18 × 3) + ( 0.37 × 4) + ( 0.19 × 5)] = 3.34, for the precontract phase;

𝑉𝐿𝑐𝑡𝑠 = [(0.06 × 1) + ( 0.07 × 2) + ( 0.17 × 3) + ( 0.43 × 4) + ( 0.27 × 5)] = 3.77 for the contract phase;

𝑉𝐿𝑐𝑎𝑠 = [(0.11 × 1) + ( 0.16 × 2) + ( 0.19 × 3) + ( 0.41 × 4) + ( 0.14 × 5)] = 3.30, for the contract administration stage; and,

𝑉𝐿𝑝𝑐𝑝 = [(0.07 × 1) + ( 0.12 × 2) + ( 0.20 × 3) + ( 0.43 × 4) + ( 0.18 × 5)] = 3.51 for the postcontract phase.

3.3.6 Estimating the overall vulnerability Index The weighted mean method was adopted for this study since it reserves the vulnerability effects of all the procurement activities as well as constructs, an upper threshold of 1 is needed in the event of the normalization of both the activities and the constructs and also the disparities between their respective 25

ACCEPTED MANUSCRIPT weightings are nominal (Hsiao, 1998; Lo et al. 2016; Ameyaw and Chan 2016). This selection criterium is as well adopted widely in the domain of fuzzy multi-criteria decision making (Ameyaw et al. 2016; Yeung et al. 2010). According to Hsaio (1998), the weighted mean {M(●,Ꚛ)} approach is defined as:

{

}

𝑚 𝑑𝑖𝑛 = 𝑚𝑖𝑛 1, ∑𝑖 = 1𝑤𝑖𝑛𝑥𝑘𝑣𝑖𝑛 , 𝑛 = 1,2,.....,𝑓

(11)

The notation of 1 as the upper threshold for the normalized weightings is done using the symbol “Ꚛ”and the normalization of the weighted functions causes the operator ‘Ꚛ’ to regress when real numbers are added such that:

𝑚

𝑑𝑖𝑛 = ∑𝑖 = 1𝑤𝑖𝑛𝑥𝑘𝑣𝑖𝑛, 𝑛 = 1,2,.....,𝑓

That is, the operation relapses to {M (●,Ꚛ)} and this model is termed as the weighted mean score (Hsiao 1998). The computed matrix, therefore, forms the foundational fuzzy matrix to calculate the overall vulnerability index for the procurement of infrastructure projects. Therefore, using the formula below:

|||

𝐷1 𝑑11 𝐷2 𝑑21 𝑅= 𝐷 = 𝑑 3 31 𝐷4 𝑑41

𝑑12 𝑑22 𝑑32 𝑑42

𝑑13 𝑑23 𝑑33 𝑑43

𝑑14 𝑑24 𝑑34 𝑑44

|

𝑑15 𝑑25 𝑑35 𝑑45

(12)

Where 𝑅 represents the fuzzy matrix for estimating the overall vulnerability index, 𝐷𝑖 (𝑖 = 𝑣1,𝑣2,𝑣3,𝑣4) characterizes the obtained evaluated matrix. A typical illustration is presented below using the final generated fuzzy evaluation matrix of the procurement constructs at level 2 which will then be used to determine the overall vulnerability index of the procurement process. This is demonstrated below:

26

ACCEPTED MANUSCRIPT

| ||

𝐷𝑝𝑠𝑐 0.13 0.14 𝐷𝑐𝑡𝑠 0.06 0.07 = 0.11 0.16 𝑅= 𝐷 𝑐𝑎𝑠 0.07 0.12 𝐷𝑝𝑐𝑝

0.18 0.17 0.19 0.20

0.37 0.43 0.41 0.43

0.19 0.27 0.14 0.18

|

Using equation 12 above, the fuzzy matrix, 𝑅, is afterwards normalized through the determined weighting function set of the individual constructs of the procurement process (𝑤1,𝑤2,𝑤3,𝑤4)to estimate the fuzzy evaluation matrix of the final stage, as explicated in the previous section:

|

𝑑11 𝑑21 𝐷𝑖 = 𝑊𝑖●𝑅𝑖= (𝑤1,𝑤2,𝑤3,𝑤4)𝑥 𝑑 31 𝑑41

𝑑12 𝑑22 𝑑32 𝑑42

𝑑13 𝑑23 𝑑33 𝑑43

𝑑14 𝑑24 𝑑34 𝑑44

|

𝑑15 𝑑25 ' ' ' ' ' 𝑑35 = (𝐷 1,𝐷 2,𝐷 3,𝐷 4,𝐷 5)(13) 𝑑45

Using the generated weighting function set estimated through Eqn. 10, where 𝐷𝑖 represents the fuzzy evaluation matrix or the membership functions for the procurement stages (𝑖.𝑒., 𝐷'1,𝐷'2,𝐷'3,𝐷'4,𝐷'5), the fuzzy matrix can be quantified using the selected grade alternative (L=1,2,3,4,5) using the formula below:

∑5

𝐷 × 𝐿𝑡 = (𝐷'1,𝐷'2,𝐷'3,𝐷'4,𝐷'5) × (1,2,3,4,5)

1 ≤ CVI ≤ 5

𝑘=1

| ||

0.271 0.13 0.14 0.257 0.06 0.07 𝑅 = 0.181 = 0.11 0.16 0.290 0.07 0.12

0.18 0.17 0.19 0.20

0.37 0.43 0.41 0.43

(14)

|

0.19 0.27 0.14 = (0.09, 0.12, 0.19, 0.41, 0.20) 0.18

𝑃𝑉𝐼𝑜𝑣𝑒𝑟𝑎𝑙𝑙 = [(0.09 × 1) + ( 0.12 × 2) + ( 0.19 × 3) + ( 0.41 × 4) + ( 0.20 × 5)] = 3.54

3.3.7 Development of the Procurement Vulnerability Model In developing the procurement vulnerability model, the vulnerability index for each of the constructs are merged to develop a linear equation model. The technique is adopted due to its ability and flexibility to allow for different ordinal grading scales (i.e., either 9, 7 or 5-point scale to be used in assessing the

27

ACCEPTED MANUSCRIPT vulnerability index of procurement processes in developing countries which share similar characteristics with Ghana. Moreover, another advantageous justification for adopting the linear model approach is that it is clear, concise, logical, and easily understandable. It, therefore, allows practitioners, policy-makers, anti-corruption institutions and researchers to adopt or use it without difficulty. However, prior to the development of the linear model for the evaluation of the overall procurement vulnerability index for infrastructure projects in Ghana or other developing regions, it was needful to normalize the individual procurement vulnerability index of the procurement stages in order to equate the summation of all the constructs to one (Osei-Kyei and Chan 2017). This can enable the evaluation or the estimation of any of the overall procurement vulnerability index of either a proposed or existing project irrespective of the grading scale of linguistic terms adopted. Table 5 presents the normalized values of the procurement constructs. Table 5: Stages of the Procurement Process Code

IP Stages

Weighting

Coefficients

Coefficient Symbols

PSC

Pre-Contract stage

3.34

0.240

𝐶𝑝𝑠𝑐

CTS

Contract Stage

3.77

0.271

𝐶𝑐𝑡𝑠

CAS

Contract administration stage

3.30

0.237

𝐶𝑐𝑎𝑠

PCP

Post contract phase

3.51

0.252

𝐶𝑝𝑐𝑝

Total

13.92

1.000

Therefore, using the normalized values, the linear equation model of the estimation of the overall PVI for IP’s in Ghana or other similar developing countries is presented below:

𝑃𝑉𝐼𝑜𝑣𝑒𝑟𝑎𝑙𝑙 = 𝐶𝑝𝑠𝑐[𝑃𝐶𝑆] + 𝐶𝑐𝑡𝑠[𝐶𝑇𝑆] + 𝐶𝑐𝑎𝑠[𝐶𝐴𝑆] + 𝐶𝑝𝑐𝑝[𝑃𝐶𝑃]

(15)

= 0.240[PCS]+ 0.271[CTS]+0.237[CAS]+0.252[PCP]

In equation 15, the coefficients assigned to each procurement construct correspond with their respective normalized values and according to the equation. CAS turned out to be the construct with the highest coefficient value followed by PCP due to their respective high weightings obtained from the FSE technique. 28

ACCEPTED MANUSCRIPT

4. Discussion The indexes obtained from the FSE technique for each of the procurement constructs or stages explicitly revealed two out of the four constructs to be vulnerable to corrupt practices. They are contract stage and the post-contract stage. That notwithstanding, it must be emphasized that some of the vulnerability indexes of the individual activities within the specified constructs vary. Hence, whereas activities such as the preparation and signing of contract at the contract stage may be regarded as less vulnerable even though its respective construct (i.e., the contract stage) is determined to be vulnerable, activities such as the solicitation of tenders is regarded to be one of the highly vulnerable activities to the incidence of corrupt practices even though its respective construct (i.e., pre-construct stage) is identified to be relatively less vulnerable. A further discussion on each of the constructs or stages are subsequently presented.

4.1 The Pre-Contract Stage The pre-contract stage encompasses the definition of the project to be constructed through to the receipt of tenders (Ruparathna and Hewage, 2013). However, other practices regarding specific project may conduct the pre-tender meeting to establish the evaluation criteria that would be employed to assess tenders received (Lester, 2007). Therefore, CTS1 can be captured under PCS depending on the project plan. The susceptibility index recorded at this stage is identified to be relatively low mainly due to the limited number of parties involved at this stage as compared to the other constructs. However, despite the low-level index recorded by the construct, two activities out of the six under this construct were identified to be vulnerable to corrupt practices. They are the solicitation of tenders which also happened to be the highest susceptible activity under the PCS construct and second highest in terms of the vulnerability indices of all the 21 activities of the procurement process. With a mean index of 4.06, the Organization for Economic Co-operation and Development (OECD) (2016) revealed that this activity is often exposed to corruption risks such as intentional declination of notifying the public for bid invitation or disclosing vital information on the bid requirements to a favourite bidder to give him a higher chance of outbidding other potential suitable bidders. Therefore, at the pre-contract stage, most 29

ACCEPTED MANUSCRIPT of these risks are initiated by the client’s representatives or parties within the sourcing board. In the public domain, it is often instigated by government official who serves as representatives (Le et al., 2014). Government officials are frequently reported to be major contributors to the proliferation of corruption in the public sector and at this particular stage of the procurement process (Owusu et al., 2017). The penultimate susceptible activity recorded at this stage is the obtaining of the required approvals (PCS4) either needed to commence a project’s execution or during the course of execution. Often plagued with diverse forms of corrupt practices such as solicitation, influence peddling and facilitation payments, some scholars have attributed the manifestation of these forms to causative factors ranging from regulatory-specific (e.g., multifarious licenses and permits as well as defective regulatory system) to statutory-specific causes such as the ineffectiveness of coordination in government departments and licensing authorities as well as the poor documentation of records (Chan and Owusu, 2017; Ling and Tran, 2012; Le et al., 2014). A respondent reported on the proliferation of collusion (regarded as a fraudulent act often characterized by the undisclosed arrangement among a group of projects’ representatives who meet to conspire to commit a deceitful act with the primary aim of obtaining illegitimate benefits such as financial gains (Chan and Owusu, 2017) at this stage. However, this identified shortfall was not attributed to a specific activity but as a general issue within the pre-contract stage.

4.2 The Contract Stage and the Contract Administration Stage The contract stage was found to be the most susceptible stage of the procurement process to corruption as indicated by the respondents. This stage covers activities ranging from the pre-tender meeting to the evaluation of retrieved bids as well as the selection and the award of the contract to the suitable contractor. The activities within this construct obtained relatively higher vulnerability indexes, making the CTS construct the most vulnerable procurement process in developing countries. Some identified risk factors such as conflict of interest, bid suppression and rotation, cover bidding, cartels, ghosting and bid rigging are manifested at this stage and during the execution of the individual activities (World Bank, 2013). As pointed in the previous section, the causative factors underpinning the incidence of these variants of corrupt acts can be attributed to project-specific causes such as the lack of proactive 30

ACCEPTED MANUSCRIPT mechanisms to stop or limit the proliferation of corrupt acts; over competition in tendering process, overclose relationship and weak procurement structures among many others (Owusu et al., 2017; Zhang et al., 2017; Le et al., 2014). Due to the high level of susceptibility, both the fuzzy weighting and the model coefficient for the CTS construct were calculated to be 3.77 and 0.271 (more than a quarter percentage out of four constructs). These values were identified to be the highest estimated values at both levels. In effect, any development of anti-corruption measures aimed at expurgating corrupt practices in the procurement process should consider targeting and expending greater efforts at the contract and post-contract phases as compared to the other two. This will either save time or create more time as well as allow the maximum output of the expended efforts and resources. As explained earlier, the contract administration stage consists of four distinct but interrelated activities. This stage can also be termed as the post-tender or the contract management stage (Park and Kim 2018). In escending order of highly susceptible activities the topmost is the administration of progress payments with a mean index of 3.39; monitoring of project’s progress, 3.39, issuing of contract amendments and the follow up of project delivery with mean indexes of 3.13 and 2.97 respectively. The overall fuzzy index generated for the construct was 3.30 which indicates moderate or neutral vulnerability. Given the overall index, it must be emphasized that this particular construct is as well vulnerable to the incidence of corrupt practices, although moderately. Therefore, in advancing strategic measures to curb the incidence of corrupt practices should not only consider the vulnerable stages (i.e., the contract and the post-contract stages) but also on reducing both the index and the real-life incidence of the varying forms of corrupt practices at this stage of the procurement process with their respective causative instigators.

4.3 The Post Contract Stage The post contract stage was identified to be the second highest susceptible construct. With their respective overall index and associated model coefficient values of 3.51 and 0.252, signifying more than 25% of the overall susceptibility index of the procurement process, the activities within the PCP construct that render the entire construct vulnerable to the incidence of corrupt practices include the checking of proof for delivery or an assessment to ascertain whether the project has been executed as 31

ACCEPTED MANUSCRIPT specified in the contract. Other activities include the issuance of final contractual amendment and the completion of final audits among other activities (see Tables 3 and 4). Even though there are wide reported cases and studies on the causal factors and the associated effects of corrupt practices in the procurement process, only few efforts have been made to research the specificity of the procurement activities and their relationships with various constructs under the subject of corruption (Owusu et al., 2017). Consequently, it is extremely challenging to make determinations and estimations on the degree of corrupt practices (Shan et al., 2015; Ameyaw et al., 2017). This is because, whereas the practices of corruption are constantly evolving, resulting in new forms and their associated causal factors, there is an unparalleled development of calculative measures instigated and enforced to curtail their incidence and effects (Chan and Owusu, 2019; Bowen et al., 2012). In other words, the current anti-corruption measures or frameworks may not be comprehensive enough to exterminate the specific forms or corruption at the specific stages of the procurement process. Therefore, even though there are several policies and stipulations on the entire procurement process that guide different jurisdictions, studies on measures developed to specifically target the expurgation of the proliferation of corrupt practices as well as the susceptibility of the procurement process remain limited (Owusu et al., 2018; Shan et al., 2017). According to the findings of this study, not only do the stages or constructs of the procurement process vary in terms of their degree of susceptibility and the proliferation of corruption but also in terms of the corrupt activities prevalent within the stages (Stansbury and Stansbury, 2008; Chan and Owusu, 2017). Therefore, this study maintains that the development of anti-corruption measures and frameworks towards the annihilation of variants of corrupt practices in the procurement process should be specific and directed towards the individual procurement activities rather than being generic. A final point to be made here concerns the work of Ferewer et al. (2017) which tested some indicators within the procurement process, specifically at the decision to contract stage, the definition of contract requirement stage, the contract stage, the contract award stage, and the contract implementation and periodic supervision and monitoring. Regarding the decision to contract stage, Ferewer et al. (2017) pointed out that while this stage is intended to facilitate the decision to procure the needed goods, works and services, there is a high tendency to deviate from the policy rationale or 32

ACCEPTED MANUSCRIPT the actual need to procure. The deviation is therefore aimed at creating possible outlet to illegally channel resources or benefits either to an organization or the parties behind the act (OECD, 2007; Ferewer et al., 2017; Owusu et al., 2017). One key indicator that was identified at the requirement definition stage was the tendency for project consultants or public officials to design the tender to suit a favorite bidder since the procurement process design is formulated by these parties (Ferewer et al., 2017; Stansbury and Stansbury, 2008; Sohail and Cavil, 2007; Tabish and Jha, 2008). As a result, the entire tendering process is secretly distorted. According to Soreide (2002), it is not uncommon for public officials to decide which person or firm gets invited to tender in a competitive bidding procurement system. This, results in diverse forms of discriminatory corrupt practices including favouritism, cronyism, nepotism, and patronage among other forms of corruption (Chan and Owusu, 2017; Brown and Loosemore, 2016; Bowen et al., 2012). These forms of corrupt practices often manifest in the procurement process as a result of some critical causal factors including the complex nature of projects (Owusu et al., 2017; Krishnan, 2009), fierce competition (Le et al., 2016; Zhang et al., 2017), complex contractual structure (18;34), overclose relationship (Ling et al., 2014; Chan et al., 2003), and the inadequate proactive measures to mitigate corruption especially in developing countries like Ghana (Ameyaw and Chan, 2017). Regarding the remaining phases (i.e., from the contract administration stage to the post-contract stage), diverse examples triggered by their unique causal instigators have been identified by different studies in different contexts. For instance, at the pre-qualification and tender phase, the noted examples which corroborate the views of experts include price fixing, bid rigging, unreasonably shorter bidding time, inadequate tender advertisement, ghosting and influence peddling (Chan and Owusu, 2017; Ameyaw et al., 2017; Le et al., 2014; Locatelli et al., 2018). Moreover, Owusu and Chan (2019) purported that in developing holistic measures to mitigate the incidence and proliferation of corrupt practices in procurement, execution and management of infrastructure works, it is expedient to examine four vital indicators: (1) the causal instigators of corrupt practices (Shan et al., 2017b; Owusu et al., 2017), the contextual risk indicators or procurement irregularities (Tabish and Jah, 2011; Le et al., 2014), the criticality of the forms within a context (Ameyaw et al., 2017; Shan et al., 2017a; Chan and Owusu, 2017) and the factors that hamper the effectiveness of anti-corruption measures. However, 33

ACCEPTED MANUSCRIPT while corruption cases remain incessant in the media, with fresh cases evolving in different departments of both public and private sectors, empirical studies attributed to exploring the mentioned constructs to help limit corruption in the context of developing countries such as Ghana remain very limited. And as reported by Owusu et al. (2018), wherever the subject of and practice of corruption is not thoroughly examined, the tendency for that context to vulnerable to the incidence and practices of corruption is very high.

5. Limitations and Recommendations for future studies Admittedly, the construction process is a highly complex one; however, only the procurement section was examined in this study. Therefore, whereas germane and practical conclusions may be drawn to make significant contributions towards the justification of the adoption of the model developed in this paper in different contexts, it must be noted that the results obtained, and the model developed cannot be generalized to represent the entire construction process. Further studies should consider examining the other stages of the construction process other than the procurement which has been reported in this study. Moreover, another primary limitation of this study is attributed to the overly generalization or the applicability of the results to different contexts. Simply put, since the study primarily highlights the case of the Ghanaian context, the adoption of the linear model for estimating the vulnerability indexes of other countries should be conducted with caution as explicated above. Also, per the theory behind the measurement tools for estimating corruption index, Trapnell (2014) revealed two foundational constructs that can be considered to establish more rigorous and explicit results. They are the estimation of corrupt practices and the estimation of anti-corruption measures. This study only focused on the measurement of corruption (i.e., the susceptibility of the procurement stages to the incidence of corrupt practices) and not on anti-corruption measures. Therefore, it is recommended that future studies explore the measurement indexes for anti-corruption measures since such a study will contribute to a to the holistic measurement or estimation and determination of a firm’s or a state’s condition regarding corruption.

34

ACCEPTED MANUSCRIPT 6. Conclusion This study examined the susceptibility of the procurement process of infrastructure related projects in Ghana using the FSE technique. The study aimed to ascertain the extent to which the activities within the procurement process get exposed to the incidence of corrupt practices so that a more specific and comprehensive efforts can be taken to address the vulnerability level of each of the activity within the process. It also aimed at developing an assessment tool for determining the vulnerability indexes of projects. The results obtained were not altogether unsurprising as Ghana is regarded by several measurement indexes such as the corruption perception index (CPI) by the Transparency International (TI) as one of the neutrally corrupt countries regarding the proliferation of corrupt practices in the public sector. However, since the TIs CPI for developing countries, in general, and Ghana, in particular, constitutes the assessment on general perception of corruption proliferation and not specifically on project procurement, this study contributes to the body of knowledge on corruption, and particularly in the field of infrastructure procurement by examining the vulnerabilities of projects’ procurement process to the incidence of corrupt practices in the developing countries. The study further established the indexes for the individual procurement stages, thereby informing researchers and practitioners about how prone the stages are regarding corruption and their respective contributions towards the overall estimation of the vulnerability index. The weightings of the stages were normalized to facilitate the development of the linear model which can be adopted to estimate the vulnerability index of existing or proposed projects. This can enhance the facilitation process of predicting how susceptible projects are to corruption and the necessary measures to take to limit or expunge corruption in Ghana and other developing countries. Generally, the most common techniques for measuring corruption adopts the solicitation of the general perception using the average or mean ratings to indicate the levels of criticalities. This study, however, employs a soft computing technique to examine the vulnerability of corruption in project procurement and develops a linear measurement model for estimating and predicting the vulnerability index of proposed or existing projects – the first of its kind to be conducted and reported in the field of construction management. The results also indicate that the activities undertaken at the contract stage should be given attention in order to reduce the number of the risk indicators that could potentially 35

ACCEPTED MANUSCRIPT expose a proposed or an ongoing project to the incidence of corruption. This study contributes to the body of knowledge on the ways of measuring the various indicators of corruption in infrastructure procurement and is, arguably, the first to employ soft computing techniques (i.e., the FSE approach) to estimate the susceptibility patterns of the various stages of the procurement process as well as develop an easy to adopt-and-use, yet standardized approach to facilitate similar estimations in future works. Practically, even though a rigorous technique is employed, the model is developed in a manner that is easily understandable and can be adopted by practitioners such as policymakers and auditors for detecting and/or measuring the vulnerability indexes of various procurement activities and their respective stages of the procurement process to the incidence of corruption. The model can as well form the basis for researchers to develop more comprehensive tools that extend beyond the boundaries of the procurement process for predicting, measuring and offering effective measures for corrupt practices right from the definition of project’s requirements through to project execution to contract close-out. Lastly, this study contributes to a more deepened understanding of the various means of measuring corruption in the domain of project procurement and management.

Acknowledgements This paper forms part of a research project entitled “Dynamic Evaluation of Corruption in Public Infrastructure Procurement: A Comparative Study of Emerging and Established Economies,” from which other deliverables have been produced with different objectives but sharing common background, methodology and same research participants. We express our sincere gratitude to the Research Grants Council (RGC) of Hong Kong and to the Hong Kong Polytechnic University for funding this study. We are also grateful to the industrial and academic experts involved in this study for their invaluable input, support and motivation (for both pilot study and the main survey).

References Ameyaw, E. E., & Chan, A. P. (2016). A fuzzy approach for the allocation of risks in public–private partnership water-infrastructure projects in developing countries. Journal of Infrastructure Systems, 22(3), 04016016.

36

ACCEPTED MANUSCRIPT Ameyaw, E. E., Chan, A. P., Owusu-Manu, D. G., & Coleman, E. (2015). A fuzzy model for evaluating risk impacts on variability between contract sum and final account in governmentfunded construction projects. Journal of Facilities Management, 13(1), 45-69. Ameyaw, E. E., Pärn, E., Chan, A. P., Owusu-Manu, D. G., Edwards, D. J., & Darko, A. (2017). Corrupt practices in the construction industry: Survey of Ghanaian experience. Journal of Management in Engineering, 33(6), 05017006. Ameh, O. J., & Odusami, K. T. (2009). Professionals’ ambivalence toward ethics in the Nigerian construction industry. Journal of professional issues in engineering education and practice, 136(1), 9-16. Bowen, P. A., Edwards, P. J., & Cattell, K. (2012). Corruption in the South African construction industry: A thematic analysis of verbatim comments from survey participants. Construction Management and Economics, 30(10), 885-901. Boussabaine, A. (2013). Risk pricing strategies for public-private partnership projects. John Wiley & Sons. Boyd, J. M., & Padilla, J. D. (2009). FIDIC and integrity: A status report. Leadership and Management in Engineering, 9(3), 125-128. Cavalieri, M., Guccio, C., & Rizzo, I. (2016). On the role of environmental corruption in healthcare infrastructures: An empirical assessment for Italy using DEA with truncated regression approach. Health Policy. Chan, A. P. C., Darko, A., Olanipekun, A. O., & Ameyaw, E. E. (2017). Critical barriers to green building technologies adoption in developing countries: The case of Ghana. Journal of Cleaner Production, 172, 1067-1079. Chan, A. P., & Owusu, E. K. (2017). Corruption Forms in the Construction Industry: Literature Review. Journal of Construction Engineering and Management, 143(8), 04017057. Deng, X., Tian, Q., Ding, S., & Boase, B. (2003). Transparency in the procurement of public works. Public Money and Management, 23(3), 155-162. Eriksson, P. E., & Westerberg, M. (2011). Effects of cooperative procurement procedures on construction project performance: A conceptual framework. International Journal of Project Management, 29(2), 197-208. Heggstad, K., Frøystad, M., & Isaksen, J. (2010). The basics of integrity in procurement: A guidebook. Chr. Michelsen Institute, Bergen, Norway. Heinrich, F., & Hodess, R. (2011). Measuring corruption. Handbook of global research and practice in corruption, 18-33. Hoxley, M. (2008). “Questionnaire design and factor analysis.” Advanced Research Methods in the Built Environment, A. Knight and L. Ruddock, eds., Wiley-Blackwell, Chichester, UK. Hsiao, S. W. (1998). Fuzzy logic based decision model for product design. International Journal of Industrial Ergonomics, 21(2), 103-116. Hu, Y., Chan, A. P., Le, Y., Xu, Y., & Shan, M. (2016). Developing a program organization performance index for delivering construction megaprojects in China: Fuzzy synthetic evaluation analysis. Journal of Management in Engineering, 32(4), 05016007. International Organization for Standardization (ISO). (2008). “Construction procurement, part 1: Processes, methods and procedures.” Building, (http://www.bsigroup.com/upload/Standards&Publications/Building/Draft_BSISO108451.pdf) (July 17, 2017). Kaufmann, D., Kraay, A., & Mastruzzi, M. (2011). The worldwide governance indicators: methodology and analytical issues. Hague Journal on the Rule of Law, 3(2), 220-246.

37

ACCEPTED MANUSCRIPT Khatri, K. B., Vairavamoorthy, K., & Akinyemi, E. (2011). Framework for computing a performance index for urban infrastructure systems using a fuzzy set approach. Journal of Infrastructure Systems, 17(4), 163-175. Kim, J., & Hastak, M. (2018). Social network analysis: Characteristics of online social networks after a disaster. International Journal of Information Management, 38(1), 86-96. Knack, S. (2006). Measuring corruption in Eastern Europe and Central Asia: a critique of the crosscountry indicators. The World Bank. Kottasova I. (2014) World's most corrupt industries. CNN Money. Available at: https://money.cnn.com/2014/12/02/news/bribery-foreign-corruption/index.html (Assessed on 2nd May 2018). Krishnan, C. (2009). Combating corruption in the construction and engineering sector: The role of transparency international. Leadership and Management in Engineering, 9(3), 112-114. Krishnan, C. (2010). Tackling corruption in the construction. Available at https://www.transparency.org.uk/wpcontent/plugins/downloadattachments/includes/download.php?id=1032. Accessed on 15th October 2017 Lædre, O., Austeng, K., Haugen, T. I., & Klakegg, O. J. (2006). Procurement routes in public building and construction projects. Journal of construction engineering and management, 132(7), 689696. Le, Y., Shan, M., Chan, A. P., & Hu, Y. (2014a). Investigating the causal relationships between causes of and vulnerabilities to corruption in the Chinese public construction sector. Journal of construction engineering and management, 140(9), 05014007. Le, Y., Shan, M., Chan, A. P., & Hu, Y. (2014b). Overview of corruption research in construction. Journal of Management in Engineering, 30(4), 02514001. Ling, F. Y. Y., & Tran, P. Q. (2012). Effects of interpersonal relations on public sector construction contracts in Vietnam. Construction Management and Economics, 30(12), 1087-1101. Mawenya, A. S. (2008). Preventing corruption in Africa. occasional paper, SAIIA, Johannesburg. Neupane, A., Soar, J., & Vaidya, K. (2012). The potential of e-procurement technology for reducing corruption. International Journal of Information Technology and Management, 11(4), 273287. Neupane, A., Soar, J., & Vaidya, K. (2014). An empirical evaluation of the potential of public eprocurement to reduce corruption. Australasian Journal of Information Systems, 18(2). Nunnally, J. C. (1978). Psychometric Theory. McGraw-Hill Book Company, pp. 86-113, 190-255 Organization for Economic Co-operation and Development (OECD) (2016). Preventing Corruption in Public Procurement. Available at: http://oecd.org/gov/ethics/Corruption-Public-ProcurementBrochure.pdf. (Accessed on 15 May 2018) Osei-Kyei, R., & Chan, A. P. (2017). Developing a project success index for public–private partnership projects in developing countries. Journal of Infrastructure Systems, 23(4), 04017028. Osei-Tutu, E., Badu, E., & Owusu-Manu, D. (2010). Exploring corruption practices in public procurement of infrastructural projects in Ghana. International Journal of Managing Projects in Business, 3(2), 236-256. Otairu, A., Umar, A. A., Zawawi, N. A. W. A., Sodangi, M., & Hammad, D. B. (2014). Slow adoption of PPPs in developing countries: Survey of Nigerian Construction professionals. Procedia Engineering, 77, 188-195. Owusu, E. K., Chan, A. P., & Shan, M. (2017). Causal Factors of Corruption in Construction Project Management: An Overview. Science and engineering ethics, 1-31. 38

ACCEPTED MANUSCRIPT Park, S. H., & Kim, Y. S. (2018). An assessment of contract management capabilities for overseas construction projects. KSCE Journal of Civil Engineering, 22(7), 2147-2158. Philp, M. (2016). Corruption Definition and Measurement. In Measuring corruption (pp. 61-72). Routledge. Rowlinson, S., & McDermott, P. (Eds.). (2005). Procurement systems: A guide to best practice in construction. Routledge. Ruparathna, R., & Hewage, K. (2013). Review of contemporary construction procurement practices. Journal of management in engineering, 31(3), 04014038. Sargiacomo, M., Ianni, L., D’Andreamatteo, A., & Servalli, S. (2015). Accounting and the fight against corruption in Italian government procurement: A longitudinal critical analysis (1992– 2014). Critical Søreide, T. (2002). Corruption in public procurement. Causes, consequences and cures. Chr. Michelsen Intitute. Shan, M., Chan, A. P., Le, Y., & Hu, Y. (2015). Investigating the effectiveness of response strategies for vulnerabilities to corruption in the Chinese public construction sector. Science and engineering ethics, 21(3), 683-705. Shan, M., Chan, A. P., Le, Y., & Hu, Y. (2015a). Investigating the effectiveness of response strategies for vulnerabilities to corruption in the Chinese public construction sector. Science and Engineering Ethics, 21(3), 683–705. Shan, M., Chan, A. P., Le, Y., Xia, B., & Hu, Y. (2015b). Measuring corruption in public construction projects in China. Journal of Professional Issues in Engineering Education and Practice, 141(4), 05015001. Shan, M., Chan, A. P. C., Le, Y., Hu, Y. & Xia, B. (2017). Understanding collusive practices in Chinese construction projects. ASCE Journal of Professional Issues in Engineering Education and Practice, 143(3), 05016012. Shan, M., Le, Y., Yiu, K.T.W., Chan, A.P.C., & Hu, Y. (2017). Investigating the underlying factors of corruption in the public construction sector: Evidence from China. Science and Engineering Ethics, 23(6), 1643–1666. Shao, R. (2004). A multi-level fuzzy synthetic evaluation on investment programs in shipping. In Applications of Advanced Technologies in Transportation Engineering (2004) (pp. 497502). Sutt, J. (2011). Manual of construction project management: For owners and clients, Wiley, Chichester, U.K Tabish, S. Z. S., & Jha, K. N. (2011). Analyses and evaluation of irregularities in public procurement in India. Construction Management and Economics, 29(3), 261-274. Tabish, S. Z. S., & Jha, K. N. (2012). The impact of anti-corruption strategies on corruption free performance in public construction projects. Construction Management and Economics, 30(1), 21-35. Tah, J. H. M., & Carr, V. (2000). A proposal for construction project risk assessment using fuzzy logic. Construction Management & Economics, 18(4), 491-500. Tanzi, V. (1998). Corruption around the world: Causes, consequences, scope, and cures. Staff Papers, 45(4), 559-594. Transparency International, (2005). Global Corruption Report-2005, Pluto Press, London. Trapnell, S. E. (2015). User’s Guide to Measuring Corruption and Anti-corruption. United Nations Development Programme. Available at: http://www. undp. org/content/undp/en/home/librarypage/democratic-governance/anticorruption/user-s-guide--measuring-corruption-and-anticorruption. Html (Assessed on 5th March 2018) 39

ACCEPTED MANUSCRIPT United Nations Development Programme (UNDP) (2015). Global Integrity. Innovation for Transparency and Accountability. User’s Guide to Measuring Corruption and AntiCorruption. UNDP Global Anti-Corruption Initiative (GAIN), New York. Wang, S. Q., Tiong, R. L., Ting, S. K., & Ashley, D. (2000). Political risks: analysis of key contract clauses in China's BOT project. Journal of construction engineering and management, 125(3), 190-197. World Bank (2013). Fraud and Corruption Awareness Handbook: A handbook for civil servants Involved in public procurement. Washington: International Bank for Reconstruction and Development. Available at: http://documents.worldbank.org/curated/en/309511468156866119/pdf/877290PUB0Frau00B ox382147B00PUBLIC0.pdf (Accessed on 15th March 2018) Xu, Y., Chan, A. P., & Yeung, J. F. (2010). Developing a fuzzy risk allocation model for PPP projects in China. Journal of construction engineering and management, 136(8), 894-903. Yeung, J. F. Y., Xu, Y., Chan, A. P. C., Chan, D. W. M., Wang, S. Q. and Ke, Y. (2010). Developing a risk assessment model for PPP projects in China—A fuzzy synthetic evaluation approach. Automation in Construction, 19(7), 929-943. Zadeh, L. A. (1965). Information and control. Fuzzy sets, 8(3), 338-353. Zhang, B., Le, Y., Xia, B., & Skitmore, M. (2016). Causes of Business-to-Government Corruption in the Tendering Process in China. Journal of Management in Engineering, 05016022.

40

ACCEPTED MANUSCRIPT

Highlights 

Projects executed in developing countries are relatively susceptible to corruption.



The contract and the post-contract phase are identified to be most susceptible



The developed model can be employed by other developing countries



Specific extirpation measures are required rather than generic frameworks