Bivariate analysis as statistical analysis assesses the empirical relationship between two variables. It is helpful in preliminary tests of hypothesis of associations. In this regard, bivariate correlation was conducted between the latent variables considering their composite index (mean) calculated using SPSS version 20. SPSS was also used to examine the relationships between the latent variables. The associations of the principal constructs were also examined taking into account the composite indexes of the latent variables forming each principal construct. The latter was used for the preliminary tests of the hypotheses before formally testing them using the PLS-PM. Bivariate analysis of the latent variables and principal (higher order) constructs are displayed in Tables 6.25, 6.26 and 6.27, respectively.
Table 6.25 of Pearson‟s correlation for the 14 latent variables, used in the measurement model of PLS, demonstrates that 86 out of the 91 (95%) correlations are significant, of which 80 are positively correlated and 6 negatively. The preliminary diagnostic process gives support to the hypothetical model. With regard to the bivariate analysis of the principal constructs that form the inner model, Table 6.26 shows that all the six correlations are significantly and positively correlated. This result agrees with the five hypotheses established in the research. The preliminary support of the bivariate correlation results is briefly presented below taking into account the hypothetical model developed in Chapter Three. The model assumes that the board structure as a principal (higher level) construct with three latent variables, namely, board composition, board independence, and board committees will directly influence the board process, board service and control roles; with the board process, board service role and control roles standing as principal constructs, each with its reflective latent variables. The model also shows that there is a relationship between the board process and board service and control roles. The tables below show the correlation coefficients of the latent variables of the conceptual model and the correlations between the principal latent constructs.
**,*. Correlation is significant at the 0.01 and 0.05 level (2-tailed), respectively.
SComp SBInd SComm PrCom PrCon PrCog PrBrA SerAd SerNwR SerNwI SerSp BCont OCont SCont SComp (LV1) Pearson Correlation 1 Sig. (2-tailed) N 106 SBInd Pearson Correlation .374** 1 Sig. (2-tailed) .000 N 104 104 SComm Pearson Correlation .340** .298** 1 Sig. (2-tailed) .000 .002 N 106 104 106 PrCom Pearson Correlation .428** .515** .435** 1 Sig. (2-tailed) .000 .000 .000 N 106 104 106 106 PrCon Pearson Correlation -.284** -.296** .061 -.263** 1 Sig. (2-tailed) .003 .002 .534 .007 N 105 103 105 105 105 PrCog Pearson Correlation .457** .476** .502** .704** -.269** 1 Sig. (2-tailed) .000 .000 .000 .000 .006 N 105 103 105 105 104 105 PrBrA Pearson Correlation .519** .456** .465** .632** -.384** .670** 1 Sig. (2-tailed) .000 .000 .000 .000 .000 .000 N 106 104 106 106 105 105 106 SerAd Pearson Correlation .538** .371** .398** .519** -.116 .562** .584** 1 Sig. (2-tailed) .000 .000 .000 .000 .239 .000 .000 N 105 103 105 105 104 104 105 105 SerNwR Pearson Correlation .460** .206* .339** .398** -.007 .334** .443** .556** 1 Sig. (2-tailed) .000 .040 .001 .000 .945 .001 .000 .000 N 101 100 101 101 100 100 101 100 101 SerNwI Pearson Correlation .264** .194* .503** .461** -.004 .401** .446** .433** .502** 1 Sig. (2-tailed) .007 .049 .000 .000 .964 .000 .000 .000 .000 N 104 103 104 104 103 103 104 103 100 104 SerSp Pearson Correlation .559** .550** .491** .611** -.286** .690** .763** .734** .602** .549** 1 Sig. (2-tailed) .000 .000 .000 .000 .004 .000 .000 .000 .000 .000 N 103 102 103 103 102 102 103 102 99 102 103 BCont Pearson Correlation .547** .507** .532** .635** -.192 .624** .663** .672** .535** .559** .777** 1 Sig. (2-tailed) .000 .000 .000 .000 .051 .000 .000 .000 .000 .000 .000 N 105 104 105 105 104 104 105 104 101 104 103 105 OCont Pearson Correlation .489** .557** .337** .654** -.310** .556** .634** .586** .524** .314** .701** .711** 1 Sig. (2-tailed) .000 .000 .000 .000 .001 .000 .000 .000 .000 .001 .000 .000 N 105 104 105 105 104 104 105 104 101 104 103 105 105 SCont Pearson Correlation .529** .539** .366** .591** -.361** .605** .688** .588** .442** .322** .775** .712** .694** 1 Sig. (2-tailed) .000 .000 .000 .000 .000 .000 .000 .000 .000 .001 .000 .000 .000 N 106 104 106 106 105 105 106 105 101 104 103 105 105 106
Table 6. 26: Pearson Correlations among the principal constructs
**. Correlation is significant at the 0.01 level (2-tailed).
Table 6. 27: Pearson correlation coefficients of structural and process latent variables
**,*. Correlation is significant at the 0.01 and 0.05 level (2-tailed), respectively
As shown in Table 6.27 above, the board structure‟s latent variables (Scomp, SBInd and Scomm) are significantly correlated with board process‟s latent variables (PrCom, PrCon, PrCog, and PrBrA) except for SComm and PrCon. The structural latent variables are positively correlated with commitment, cognitive conflict and boardroom activity latent variables; and negatively correlated with the process conflict latent variable. These values give preliminary evidence of the relationship between board structure and board process latent variables. The results show that board structure affects board process. That is, a properly structured board in terms of composition, independence, and active committee might also be active in the board process expressed in terms of commitment, critical debate, and good sprit in the boardroom.
BStruct BProcess BServrole BControle
BStruct Pearson Correlation 1 Sig. (2-tailed) N 104 BProcess Pearson Correlation .679** 1 Sig. (2-tailed) .000 N 102 104 BServrole Pearson Correlation .679** .740** 1 Sig. (2-tailed) .000 .000 N 96 95 97 BControle Pearson Correlation .737** .746** .819** 1 Sig. (2-tailed) .000 .000 .000 N 104 103 97 105
Principal Construct Board process
Board structure
LV PrCom PrCon PrCog PrBrA
SComp .428** -.284* .457** 519** SBInd .515** -.296** .476** .456** SComm .435** .061 .502** .465**
Table 6. 28: Pearson’s correlation coefficients of structural, service & control roles latent variables
Principal Construct Board service role Board control role
Board structure
LV SerAd SerNwR SerNwI SerSp BCont OCont Scont SComp .538** .460** .264** 559** .547** .489** .529** SBInd .371** .206* .503** .550** .507** .557** .539** SComm .398** .339** .503** .491** .532** .337** .366** **. Correlation is significant at the 0.01 level (2-tailed)
The table above shows the extent of relationship between board structural latent variables and board service and control role latent variables. The correlation coefficients provide preliminary support of the significant and positive association between each of the structural latent variables (Scomp, SBInd and Scomm) and each of the service (SerAd, SerNwR, SerNwI and SerSp) and control role (BCont, OCont and SCont) latent variables. The significant positive correlation of the structural latent variables and the service and control latent variables may not be surprising as properly structured boards are expected to accomplish all service and control tasks effectively.
Table 6. 29: Pearson’s correlation coefficients of process, service & control roles latent variable
Principal Construct Board service role Board control role
Board Process
LV SerAd SerNwR SerNwI SerSp BCont OCont Scont PrCom .519** .398** .461** .611** .635** .654** .591** PrCon -.116 -.007 -.004 -.286** .624** -.310** -.361** PrCog .562** .334** .401** .690** .624** .556** .605**
PrBrA .584** .443** .446** .763** .663** .634** .688** **. Correlation is significant at the 0.01 (2-tailed)
Table 6.29 shows the relationship between the board process and board service and control roles. Except for the process conflict (PrCon) variable and three service role variables (SerAd, SerNwR & SeRNwI), the remaining variables show significant correlations. PrCon is significantly but negatively correlated with SerSp, OCont, and Scont but positively related with BCont. These associations seem logical because when the intensity of process conflict gets higher, it negatively affects (is detrimental to) the strategic participation, output and strategic control roles while positively influencing the behavioral control role. The remainders of the board process latent variables (PrCom, Prcog and PrBrA) are positively and significantly related to the service and control role latent variables, at the 0.01 significance level. This means that high level of commitment; cognitive conflict and good boardroom spirit probably allow boards to execute their service and control roles more
effectively, in terms of advice provision, resource generation, image building, strategic participation, behavioral , output, and strategic control.
Turning to the principal (higher level) constructs, the same results have been obtained as above but in their aggregate forms. The summary table of the bivariate correlation between the constructs is presented below.
Table 6. 30: Pearson’s correlation coefficients of board structure, board process, board service & control roles main constructs
Principal construct Board Process (Bprocess) Board service role (BServrole) Board control role (BControle) Board Structure (BStruct) .679** .679** .737** Board Process (Bprocess) 1 .740** .746** Board service role (BServrole) .740** 1 .819**
**. Correlation is significant at the 0.01 level (2-tailed).
The correlation coefficients between the latent constructs are high, significant and positive, as expected, providing preliminary evidence for the hypothesized relationships. Examining the relationship from the table above, it is noted that board structure is positively and significantly related to board process, board service role and board control role. This result is consistent with the results given in Tables 6.27 and 6.28 above showing that a properly structured board can positively influence the board process, board service role and board control role. In turn, the board process with high coefficients is positively and significantly related to the board service role and control roles. That is, a board that is in the right board process is expected to accomplish both the board service and control roles rightfully. The correlation coefficient between board service and control roles is the highest value recorded in this study (.819**) demonstrating that a board that is active in performing the service tasks can also be active in performing all the control tasks.
Once the quality of the measurement model is assured and preliminary test of relationships performed, the next step of analysis would be estimation of the specified structural model and formal tests of the hypotheses using Smart PLS-PM. This is discussed in Chapter Eight.
6.6 Summary
This chapter presented the evaluation of the reliability and validity of the measurement model as a prerequisite to the assessment of the structural model, (testing hypotheses). That is, if the measurement model cannot pass tests of
reliability and validity, it is not possible to assess the structural model or perform hypothesis testing. The measurement purification was done in two stages. Firstly, the factorial dimensionality of the indicator variables with respect to their latent variables was carried out through exploratory factor analysis using principal component analysis (PCA). Secondly, construct reliability and validity tests were performed using PLS outer (measurement) model evaluation on those items retained after EFA. The two processes led to the removal of a total of 20 items (14 items through EFA and 6 items by CFA). The PCA regrouped the indictor variables into several constructs. The processes ensured the reliability and validity of the measurement model, which serves as a basis for the measurement of the structural model presented in Chapter Eight. The profile of sample-1 respondents and descriptive statistics of those manifest and latent variables that passed the validation tests are also presented in this chapter. Before the structural model evaluation that tests the hypotheses established in Chapter Three, bivariate correlation analysis was carried out as a preliminary test of the appropriateness of the structural model relationships hypothesized.
Before presenting the structural model evaluation, the next chapter will examine the perceptions of both the governing bodies (Sample-1) and group of stakeholders (Sample-2), separately and in aggregate terms, regarding the current corporate governance systems and practices in Ethiopia. The analysis is believed to complement the structural model results.
Chapter 7 Perception Survey of Corporate Governance Practices: