Zakari Mustapha, Clinton Aigbavboa and Wellington Didi ThwalaDepartment of Construction Management & Quantity Surveying,

University of Johannesburg, South AfricaEmail: zakari.mustapha1967@gmail.com

Abstract:Health and Safety (H&S) compliance has been problem to majority of Small andMedium-

A face-to-facemethod of questionnaire administration was adopted among the SMEs contractors inGhana. Findings from Structural Equation Modelling (SEM) analysis confirmed that

policy features on the H&S compliance. Hence, strong in predicting H&S compliance.

through the use of SEM. The practical contribution of the article is the application ofthe safety policy model by contracts and its enforcement by the Ministry of WaterResources, Works and Housing.

Keywords:Construction, Compliance, Health and Safety, Policy

1 Introduction

The well-being of workers among Small and Medium-Sized Enterprises (SMEs) contractors isaffected by the high rate of accidents at their workplaces. SMEs contractors are faced withdifficulties in sourcing financial resources, expertise and staff which have significant effect onsafety regulations compliance (Department of Occupational Safety and Health Annual Report(DOSH, 2008). The said difficulties have limited their purchasing power for equipment andtraining of employees (Ofori & Toor, 2012; International Labour Office (ILO, 2005). SMEscontractors H&S compliance in relation to safety policy were found to be lacking. Moreover,SMEs contractors form the bulk of the contractors in Ghana and provide operational flexibilityto the larger firms as sub-contractors (Ofori & Toor, 2012; ILO, 2005). It is presumed that the

seof this study is to carry out a confirmatory factor analysis of safety policy features for use inH&S compliance study among SMEs contractors in Ghana. The article begins with anoverview of a literature review on the topic in question. The methodology adopted for the studyis presented, followed by the findings of the research. Finally, conclusions are drawn and

safety policy features and provides significant insight into SMEs contractors H&S complianceimprovement.

1.1

Report from the International Labour Organisation (ILO, 2005) has shown that majority oflarge companies concentrate on a few specialized core areas due to demand for flexibilityarising from globalization. Outsourcing and subcontracting have been attributed to the largenumber of SMEs, micro-enterprises and self-employed workers (ILO, 2005). Most of the SMEscontractors operate in the informal economy beyond any coverage by the formal OccupationalSafety and Health (OSH) or inspection services because they are not adequately covered bysafety and health legislation (ILO, 2005). Occupational hazards and risks are recognized to bemore widespread in SMEs contractors than in large enterprises due to their unwillingness toseek relevant advice on H&S inspection (ILO, 2005). Construction H&S as posited by Furter(2011) has become one of the top ten risks. Research has also shown that legislation or targetedregulations can influence H&S performance of either a project, industry or a stakeholder(Construction Industry Development Board (cidb, 2008). It is not sufficient to provide safeequipment, systems and procedures if the culture is not conducive to a healthy and safe workingenvironment (Institution of Occupational of Safety and Health (IOSH, 2004). Since, culturecreates a homogeneous set of assumptions and decision premises in which compliance occurswithout surveillance (Grote, 2007).

It is also argued that a positive culture leads to both improved H&S as well as organisationalperformance (Dingsdag, Biggs, Sheahan & Cipolla, 2006). Behaviour is a product of culturejust as much as accidents are a product of the prevailing culture (Wiegmann, Zhang, vonThaden, Sharma & Mitchell, 2002). Wiegmann et al. (2002) further argued that sustainedimprovement in H&S is not possible without cultural change. OSH culture can be described interms of the informal, cultural aspects of an organisation. The latter can have an impact on howOSH is perceived and dealt with, and on whether people are aware of OSH-related issues andact in a safe and healthy way (European Agency for Safety and Health at Work (EU-OSHA,2011). 'OSH culture' - can be seen in terms of the relationship between organisational culture

positive or negative. This is possible through the following two steps: i. Setting the values andnorms, and underlying beliefs and convictions, through which workers deal with or disregardrisks; ii. Influencing the conventions for (safe or unsafe, healthy or unhealthy) behaviour,interaction, and communication (EU-OSHA, 2011). OSH culture can be assessed as part of aprocess of organisational improvement. It is also perceived and dealt with among workers inan organisation and whether workers are aware of OSH-related issues and act in a safe andhealthy way. The knowledge and information gained from such a cultural approach, can in turnbe very useful in the process of changing OSH-related policies, processes, and practices stepby step, adapting them to the existing local context and culture, and eventually leading to betterOSH performance (EU-OSHA, 2011).

The key issue for employers, business managers and OSH professionals to strive for excellencein the field of OSH, is to ensure that occupational accidents and work related ill health areprevented as much as possible, and that safe and healthy behaviour among all employees ispromoted (EU-OSHA, 2011). Policy formulation, implementation and monitoring are theresponsibility of government and are vital indicators that determine compliance of H&S among

H&S policy statement details out how it willensure a healthy and safe work environment. Individual policies need to be developed forspecific hazards and issues. Policies should be supported by procedures that provide the step-by-step instructions on how policies will be achieved (Construction Design and ManagementRegulations 2007 (CDM 2007). It has also been indicated in Section 2 of Health and Safety atWork (HSW) Act 1974 that if an organization employs more than five people, it must have awritten H&S policy (CDM 2007). However, the latent construct for the study is Contrators

Safety Policy (CSP) and its indicator variables are safe storage of equipment, safe and healthywork environment, safe storage of formwork and false work and do not service equipmentwhich is in operation. The first step towards the management systems approach to OSH and isreflected in the Occupational Safety and Health Convention of 1981 (No. 155). Although, theAct deals with OHS and working environment in a comprehensive manner, but it is largely apolicy rather than a prescriptive instrument. The Occupational Safety and Health Conventionof 1981 (No. 155) also provide priority to the formulation, implementation and periodic reviewof a national policy to prevent accidents and injury to health arising from or that is linked withoccurrence of accident in the course of work. It also seeks to minimize, as far as possible thecauses of hazards inherent in the working environment (ILO, 2005). Moreover, the scope andcoverage of OSH provisions has evolved from a focus on industrial safety to one on workplacesafety and health, from protection to prevention and assessment of risks. Modern standardsreflect not only on collective responsibilities to workplace safety and health, but also therespective roles, rights, responsibilities and areas for cooperation of and between employers,workers and their representatives (ILO, 2005).

It is the responsibility of the H&S personnel to provide general H&S advice, and also advicerelating to construction H&S issues (Lingard & Rowlinson, 2005; Carpenter, 2006a) foremployees. Occupational Health and Safety (OHS) is core to the successful long-termsustainability of any business and fortunately in South Africa, Health and Safety (H&S) is alegislatively compliant criterion, enforced by the OHS Act 85 of 1993 and the Department ofLabour (Action Training Academy, 2014). Health policy is best formulated through rigorousand objective assessment of data. Modern health policy poses complex legal, ethical and socialquestions. Hence, the goal of health policy is to protect and promote the health of individualsand the community. Government officials can accomplish this objective in ways that respecthuman right (Gostin, n.d.).

Table 1:

Source: Researcher

2 Research Methodology

This section presents the methodology used during the questionnaire administration. From theeight- hundred questionnaires administered, 558 questionnaires were returned at the end of thesurvey. The sample size of 588 obtained for the study is considered as large (Kline, 2005). Asample size less than 100 cases will be difficult to analyse when Structural Equation Mdeling(SEM) is used as an analytical tool (Harris & Schaubroeck, 1990; Kline, 2005). The minimumsample required for SEM analysis should be 200 respondents (Bollen, 1989). This requirementwas used by Bentler and Chou (1987). The variable ratio of an ideal SEM model has beensuggested by Tong (2007) to be at least 5:1. This implies, a SEM model with 10 observedvariables should have more than 50 respondents. There are 66 hypothesised observed variablesand the ratio to sample size for the current study is 8.45:1. Therefore, variable ratio to sample

Latent construct CSP construct Indicator variables Label

(CSP)Four (4) dependentvariables.Five (5) independentvariables.Eight (8) free parameters.Five fixed non-zeroparameters.

Safe storage of equipment CSP 1Safe and healthy workenvironment

CPS 2

Do not service equipmentwhich is in operation

CPS 3

Safe storage of formwork andfalse work

CPS 4

size meets the requirement recommendation in literature by Tong (2007). The sample data of558 was finally taken through random sampling before carrying out an Exploratory FactorAnalysis (EFA) and Confirmatory Factor Analysis (CFA) respectively. 269 samples wererealised for the EFA analysis and 289 samples for CFA analysis. The study adopted aquantitative method of data collection. A face-to-face method questionnaire administration wascarried out among Small and Medium-Sized Enterprises (SMEs) contractors in the major cities(Accra-Tema, Kumasi and Takoradi) in Ghana. Structural Equation Modeling (SEM) softwareversion 6.2 was employed in the data analysis. The factor structure of the constructs wasassessed by the SEM software. The conceptual variables were then tested as a prior using SEMof the questionnaire survey results (Hu & Bentler, 1999).

3.1 Model testing

The maximum likelihood method was used for the construct parameters. Consideration was

significance deviation from normality, the Satorra-Bentlet scale statistics (robust) would beused as these have been found to perform adequate under such conditions (Bentler,1990). Theconstruct validity for the variables was conducted to demonstrate the extent to which theconstructs hypothetically relate to one another in order to establish the score reliability. The

on the examination of the residual covariance matrix from CFA output results as opposed tothe correlation matrix. The MI ensures that the attributes would relate to the same set ofobservations in the same way. While the covariance matrix establishes the variables that

ormed,all indicator variables with an unacceptable high residual covariance matrix greater than 2.58were dropped. This implies that the identified indicator variables do not sufficiently measure

importance in other cultural contextsand previous studies.

Bryne (2006) and Joreskog and Sorbom (1998) opined that residual covariance matrix greaterthan 2.58 are considered large. Therefore, in order for a variable to be described as well-fittingincovariance matrix should be systematically and centred on zero (Bryne, 2006; Joreskog &Sorbom, 1998). This procedure was adopted as a means to ensure that the indicator variableswere measuring the latent constructs. The assumption of measurement invariance is mostlytested in CFA (Meredith, 1993) in order to allow for comparison of indicator variables underthe same condition. In the current article, CFA was used for the assessment of measurementinvariance across latent variables. This procedure was adopted by Aigbavboa and Thwala(2013) and Musonda (2012).

4 Findings and Discussion

This section provided demographic information on the individual respondents and the firms.

including their individual information and the company information. A total of 269 sampleswere realised for the EFA after the random sampling. The responses represent 82.2% (N = 221)males and 17.8% (N = 48) females, as shown in Table 2.

Table 2: GenderGender Frequency Percent (%)Male 221 82.2

Female 48 17.8Source: Researcher

Table 3 shows that majority of the respondents (32.7%; N=88) were between the ages of 26-30years, followed by the age group of 31-35years (20.8%) and the aged range 20-26years and36-40 years constituted (14.1%) each respectively of the sample.

Table 3: Age GroupAge Group Frequency Percent (%)Less than 20years 7 2.620-25years 38 14.126-30years 88 32.731-35years 56 20.836-40years 38 14.141-45years 23 8.646years and above 18 6.7Missing 1 0.4

Source: Researcher

The highest education level of the majority of the sample respondents was National Diplomaor Certificate (34.9%; N = 94) and the least of the sample respondents was Post-Graduatedegree (10.0%; N=27) as shown in Table 4.

Table 4: Highest QualificationQualification Frequency Percent (%)Senior School Certificate 42 15.6National Diploma or Certificate 94 34.9

90 33.5Post-Graduate 27 10.0Missing 16 5.9

Source: Researcher

Table 5 shows that a large number of the respondents (38.3%) have worked in the firm between2 -5years, followed by year group of 6-10years (28.6%) and the year range 31years and aboveconstituted the least (1.5%) of the sample.

Table 5: Tenure in the FirmTenure Frequency Percent (%)2-5years 103 38.36-10years 77 28.611-15years 32 11.916-20years 30 11.221-25years 11 4.126-30years 10 3.731years and above 4 1.5Missing 2 0.7

Source: Researcher

Majority of the respondents (27.1%; N=73) indicated their firms have been in existence in therange of 15-20 years, followed by the year range of 5-9 years; N=64) and the firm with the yearrange of 21-30 years constituted the least (11.9 %; N=32) of the sample as shown in Table 6.

Table 6: Existence of Firm

Existence Frequency Percent (%)5-9years 64 23.810-14years 62 23.0

Source: Researcher

Table 7 shows that majority of the respondents were employed by the private firms (62.1%;N=167), followed by public (24.5%, N=66) and the sole proprietorship has the least number ofrespondents (13.0; N=35) of the sample.

Table 7: Firm Ownership

Ownership Frequency Percent (%)Private 167 62.1Public Liability 66 24.5Sole Proprietorship 35 13.0Missing 1 0.4

Source: Researcher

Majority of the ongoing projects were under public liability (50.9%; N=137), followed byprivate firms (39.8%; N=107) and the sole proprietorship constituted the least (5.6%; N=15) ofthe sample as shown in Table 8.

Table 8: Ongoing Projects

Ongoing Frequency Percent (%)Private 107 39.8Public Liability 137 50.9Sole Proprietorship 15 5.6Missing 10 3.7

Source: Researcher

Table 9 Majority of the respondents (66.9%; N=180) were carrying out Building ConstructionWorks, followed by Civil Engineering Works (30.1%; N= 81) and Other Works has the leastof the respondents (1.5%; N=4) of the sample.

Table 9: Type of Projects

Type Frequency Percent (%)Building Construction 180 66.9Civil Engineering 81 30.1Other 4 1.5Missing 4 1.5

Source: Researcher

Majority of the respondents (43.9%; N= 118) indicated the National level as the geographicalspread of their firm, followed by Regional (26.8%; N= 72) and the International constituted theleast number of respondents (7.8%; N= 21) of the sample as shown in Table 10.

Table 10: Geographical Spread

Geographical Spread Frequency Percent (%)Metropolitan 57 21.2Regional 72 26.8National 118 43.9International 21 7.8Missing 1 0.4

Source: Researcher

15-20years 73 27.121-30years 32 11.931years and above 35 1.3.0Missing 3 1.1

Table 11 show that a large number of the respondents (31.2%; N=84) worked with D2/K2 classof contractors, followed by D3/K3 class (30.9 %; N=83) and the D4/K4 class constituted theleast (9.3%; N=25) of the sample.

Table 11: Classification of FirmClassification Frequency Percent (%)D1/K1 67 24.9D2/K2 84 31.2D3/K3 83 30.9D4/K4 25 9.3Missing 10 3.7

Source: Researcher

D1/K1 - D4/K4 (Building & Civil Engineering Contractors) classification, from the lowest tothe highest based on their financial standing, equipment holding, qualification and number ofpermanent employees.

4.1

Table 1 shows four (4) indicator variables (CSP 1, CSP 2, CSP 3 and CSP 4) with acceptableresidual covariance matrix. The acceptability of residual covariance matrix was as a result ofits symmetrical and centred on zero (Byrne, 2006:94; Joreskog & Sorbom, 1988) deemed well-fit for a model. The four-indicator model provides good measures of residual matrix andevidence of convergent validity. The assessment of the CSP model goodness-of-fit was basedon the four indicator variables. Some scholars (Bollen, 1998; Bryne, 2006; MacCallum,Browne & Sugawara, 1996) have suggested a minimum of four indicator variables. Therefore,the data obtained fall within the minimum requirement. Robust maximum likelihood method

deviated significantly from normality (Yuan, Lambert and Fouladi = 262.0696). Theexamination of the Bentler-Weeks structure representation for the approved construct revealedthat CSP construct has four (4) dependent variables, five (5) independent variables and eight(8) free parameters. The number of fixed non-zero parameters was five (5).

Table 12 shows that the sample data on CSP measurement model yield an S 2 of 3249.5with 1861 degrees of freedom. The associated p-value was determined to be 0.0000. The chi-square value advocated that the difference between the sample data and the postulated

chi-square value was determined to be 1.75. The normed chi-square is the procedure of dividingthe chi-square by the degree of freedom. The normed values up to 3.0 or 5.0 are recommended(Kline, 2005). The ratio of S 2 to the degree of freedom was lower than the lower limitvalue of 3.0 suggesting a good fit of the data to the construct (Byrne, 2006).

Fit Index Cut-off value Estimate Comment

S 2 3249.5

df 1861 Good fit

CFI 0.794 Acceptable

RMSEA Less than 0.05 withconfidence interval (CI)0.00-

0.051Good fit

95%

NFI 0.629 Acceptable

NNFI Greater than 0.80. 0.777 Acceptable

RMSEA 95% CI 0.048: 0.054 Acceptablerange

Source: Researcher

Table 12 shows the goodness-of-fit indexes. The comparative fit index (CFI) of 0.794 wasfound to be slightly lower than the cut-off value for good fit model. A model is said to be goodfit if the CFI is greater or equal to the cut-off value of 0.95 (Hu & Bentler, 1999; Joreskog &Sorbom, 1998). This indicates a drop (difference of 0.156) in the CFI value, hence the modelcan be described to have an acceptable fit (Schreiber et al., 2006; Hu & Bentler, 1999) thoughnot well fitting. However, the robust mean square error of approximation (RMSEA) with 95per cent confidence interval was found to be 0.051 (lower bound value = 0.054 and the upperbound value =0.048) which is within the acceptable range for a good fit model (MacCallum etal., 1996). Moreover, both the normed fit index (NFI) and non-normed fit index (NNFI) werefound to be within the acceptable range of 0.629 and 0.777 respectively (Schreiber et al., 2006;Hu & Bentler, 1999). Evaluation of RMSEA (95% CI), CFIs, NFIs and NNFIs indicated anacceptable fit of the measurement model, but not very good for a government support featuresfactor.

4.2health and safety compliance

Determination of the internal consistency for the CSP measurement model was made possiblethrough the examination of the Rireliability. Kline (2005) posited that the desired multivariate reliability coefficient should fall

examined to determine the score reliability. The Rio coefficient of internal consistency was

found to be above the minimum value 0.70 at 0.937. High levels of internal consistency andinternal reliability were as shown in Table 13.

The examination of the magnitude of the parameter coefficients led to the determination of theconstruct validity. Hence, high parameter coefficients greater than 0.50 indicate a close relationbetween the factor and the indicator variable. Hair, Anderson, Tatham and Black (1998) positedthat a parameter coefficient of 0.50 is interpreted as 25 per cent of the total variance in theindicator variable being explained by the variable (factor). In this case, a parameter coefficienthas to be between 0.50 and 0.70 or greater to explain about 50 per cent of the variance in anindicator variable. Hence, the inspection of the standardized parameter coefficient shown inTable 13 shows that they were significantly high (values from 0.747 to 0.604). The minimumestimate of 0.604 suggested that the measured factor accounts for 9.540 of the Z-statistics inpredicting the overall health and safety (H&S) compliance. The Z-statistics for each indicatorvariables by the endogenous variables revealed that the scores were significant at 5 percentlevel.

Table 13: Reliability and construct validity of CSP model

Indicator

Variable

Stad. Z- Stat. R2 FactorLoading

Sig.@5%level?

CSP 1CSP 2CSP 3CSP 4

0.471

0.787

0.629

0.573

6.50910.71010.0219.506

0.602

0.779

0.381

0.605

0.5632

0.5187

0.5898

0.4125

YesYesYesYes

Source: Researcher

0.937; Rio coefficient = 0.964

(Robust statistical significance at 5% level)

** SEM analysis norm (Kline, 2005) One variable loading per latent factor is set equal to1.0 in order to set the metric for that factor

*Parameter estimates are based on standardized solutions

Moreover, the assessment of the inter-factor correlation (R2

policy feature indicator measures revealed that only one indicator value was close to the desired

compliance. The inter-factor correlation test of statistics (Z-stats) which functions as a Z-statistics test shows that the estimate is significantly different from zero. However, the R2 didnot significantly measure the R2 variable. The statistical assessment of the score results showedthat the influence of this factor on the R2 variable was weak (indirect). This is not withstanding,the fact that the combined results revealed that it has a good indirect association in theprediction of the overall H&S compliance.

4.3 Discussion of results

internal reliability and the construct validity criteria. The Rio value was above the minimumvalue of 0.70 (Table 13). The construct validity criteria were justified by the magnitude and

policy feature indicator revealed that four indicator variables passed the test and were used forthe assessmen -of-fit.Moreover, the indicator variables (safe storage of equipment, safe and healthy workenvironment, safe storage of formwork and false work and do not service equipment which isin operatiFurther assessment of the Z-statistics accounted for each measure by the indicator variablesrevealed that the scores were significant, since two Z-statistics values were close to 10.00 andtwo were above 10.00. These results suggest that the direct influence of these variables on theH&S compliance was strong (direct).

The government is responsibility for the H&S policy formulation. Moreover, theimplementation and monitoring of H&S policy among contractors by government officials will

These findingsconcur with the findings of Lingard and Rowlinson, 2005; Carpenter, 2006. Conducting aconfirmatofeatures is very important because of its application in H&S study among contractors in Ghana.

The analysis of confirmatory factor analysis made it possible to characterize and identifyspecifically the factors of which have statistically significantinfluence on the contractors in Ghana. Hence, contractors will find it important to implementand monitor the safety policy formulated by the government in relation to their establishedsafety policy to ensure H&S compliance.

5 Conclusion and Recommendations

The postulated prior was analysed using SEM software with EQS version 6.2. The SEMprocess was undertaken as both EFA and CFA of the prior variables. The CFA analysisrevealed that four indicator variables were successful in the factorial validity test conducted.

measurement model goodness-of-fit are found to be a contributing factor to Health and Safety(H&S) compliance. Further findings indicated that the Z-statistics for the four indicatorvariables were within the acceptable range. The robust fit indexes had an acceptable fit, whileRMSEA value and the RMSEA with 95 per cent confidence interval produced an acceptablerange. Moreover, the parameter estimates were statistically significant and dealt with

adequate fit to the sample data. The CFA result shows that only few variables were classified

ferent researchmethods on the determinants of support from the government are in agreement with the aboveview.

This research supports the theory confirmation that measurement of indicator variables shouldbe the first stage of theory testing. It is therefore recommended that a checklist of items defining

It becomesparamount to integrate H&S policy into the management systems of SMEs contractors at alllevels of construction. Moreover, the effective formulation and implementation of H&S policy,regular education and training within the H&S policy should be encouraged both by thegovernment and the parties involved. The formulation of the government policy and its

their employees. Moreover, SEM software with EQS version 6.2 should be used to analyse thevariables that may be considered in the future for the development of H&S compliance projects.The Ministry of Water Resources, Works and Housing should enforce the use of safety policyby the contractors.

6 References

Aigbavboa, C.O. and Thwala, W.D. (2013) Confirmatory factorial validity of neighborhoodfeatures among South African low-income housing occupants, Journal of Economics andBehavioral Studies, 5(12), pp. 825-837.

Action Training Academy (2014). Occupational Health and Safety Compliance-Training-Equipment, available at: http://www.actiontrainig.co.za [accessed 08 April 2015].

Bentler, P.M. (1990) Comparative fit indexes in structural models, Psychological Bulletin,107(2), pp. 238-246.

Bentler, P.M. & Chou, C.P. (1987). Practical Issues in Structural Modeling. SociologicalMethods and research, 16(1), pp. 78-117.

Bollen, K.A. (1998) Structural Equations with Latent Variables, John Wiley & Sons, Inc., NewYork.

Byrne, B.M. (2006) Structural Equation Modelling with EQS- Basic Concepts, Applicationsand Programming, Lawrence Erlbaum Associates, Mahwah.

Carpenter, J. (2006) Developing guidelines for the selection of designers and contractors underConstruction (Design and Management) Regulations 1994, Norwich, Health and SafetyExecutive.

Construction (Design and Management) Regulations 2007 (CDM 2007) Managing health andsafety in construction. Approved Code of Practice.

Construction Industry Development Board (cidb, 2008) Development throughPartnership.Construction Health and Safety in South Africa, available at:http://www.asocsa.org [accessed 08 April 2015].

Department of Occupational Safety and Health Annual Report (DOSH) (2008), Department ofOccupational Safety and Health: Labour and Human Resources Statistics 2008, availableat: http://www.mohr.gov.my/ (accessed 20 February 2015).

Dingsdag, D.P., Biggs, H.C., Sheahan, V.L. and Cipolla, C.J. (2006) A construction safetycompetency framework: Improving OHS performance by creating and maintaining a safetyculture, Brisbane: Cooperative Research Centre for Construction Innovation. DisasterManagement Institute (DMI) nd. Accident Causation Theories.

European Agency for Safety and Health at Work (2011) Occupational Health and SafetyCulture Assessment- A Review of Main Approaches and Selected Tools, WorkingEnvironment Information, Working Paper, ed. Terence N. Taylor, T.N. (EU-OSHA),Luxembourg: Publications Office of the European, available at: http://europa.eu/[accessed02 May 2016].

Occupational Health and Safety, Environment and Quality Management, available at:http://www. Sheqafriac.com [accessed 26 February 2016].

Gostin, L. (n.d.)

Biomedicine, Washington, DC: The National Academies Press, 1995, available at:http://www.nap.edu/r(accessed [accessed 02 May 2016].

Grote, G. (2007) Understanding and assessing safety culture through the lens of organizationalmanagement of uncertainty, Safety Science, 45: pp. 637-652.

Hair, J.F., Anderson, R.E., Tatham, R.L. and Black, W.C. (1998) Multivariate Data Analysis,5th edn, Prentice Hall, Englewood Cliffs, New Jersey.

Harris, M.M. and Schaubroeck, J. (1990) Confirmatory modeling in OB/HRM: Technicalissues and applications. Journal of Management Information Systems, 16, pp. 337-360.

Hu, L. and Bentler, P.M. (1999) Cut-off criteria for fit indices in covariance structure analysis:conventional criteria versus new alternatives, Structural Equation Modelling, Vol. 6(1),pp. 1-55.

International Labour Office (2005) Promotional framework for occupational safety and healthInternational Labour Conference, 93rd Session 2005, Report IV (1), International LabourOffice Geneva.

Institution of Occupational of Safety and Health (IOSH, 2004) Promoting a positive culture -a guide to health and safety culture, 4.2. Leicestershire.

Joreskog, K.G. and Sorbom, D. (1998) PRELIS: A Program for Multivariate Data Screeningand Data Summarization. A Preprocessor for LISREL, 2nd edn, Scientific Software,Mooresville.

Kline, R.B. (2005) Principles and Practice of Structural Equation Modeling, 2nd edn, GuilfordPress, New York.

Lingard, H. and Rowlinson, S. (2005) Occupational Health and Safety in Construction ProjectManagement. Spon Press 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN.

MacCallum, R.C., Browne, M.W. and Sugawara, H.M. (1996) Power analysis anddetermination of sample size for covariance structure modelling. Psychological Methods,1, pp. 130-149.

Meredith, W. (1993) Measurement invariance, factor analysis and factorial invariance.Psychometrika, 58, pp. 525-543.

Musonda, I. (2012). Construction Health and Safety (H&S) Performance Improvement. Aclient-Centred Model. Unpublished DPhil in Engineering Management. University ofJohannesburg. South Africa, available at: http://www.ujdigispace.uj.ac.za [accessed 02June 2014].

Ofori1, G. and Toor, S.R. (2012) Leadership development for construction SMEs, inProceedings of the POC 2012 Conference, Working Paper Proceedings, EngineeringProject Organizations Conference, Rheden, The Netherlands, 10-12 July 2012.

Schreiber, J.B., Nora, A., Stage, F.K., Barlow, E.A. and King, J. (2006) Reporting structuralequation modeling and confirmatory factor analysis results: A review, The Journal ofEducational Research, 99(6), pp.323-338.

Tong, D.Y. (2007) An Empirical Study of E-Recruitment Technology Adoption in Malaysia:Assessment of Modified Technology Acceptance Model, Multimedia University,Malaysia.

Wiegmann, D. A., H. Zhang, T. L. von Thaden, G. Sharma, and A. A. Mitchell (2002) ASynthesis of Safety Culture and Safety Climate Research, University of Illinois AviationResearch Lab Technical Report ARL-02-03/FAA-02-2.