Confirmatory Factors Analysis

4.2 Reliability and Validity Analysis

This paper employs the Cronbach’s α coefficient to analyze the reliability between the variable scales. Moreover, we adopt the suggestion of Cuieford (1965) and Nunnally (1978). If Cronbach’s α coefficient is greater than 0.7, there is high internal consistency; on the contrary, if the Cronbach’s α coefficient is lower than 0.35, the internal consistency is low. A coefficient between 0.35 and 0.7 shows fair reliability. The coefficient for moral judgment is 0.61, moral maturity is 0.61, and others are above 0.70. This paper also analyzes the composite reliability and variance extracted by Confirmatory Factor Analysis (CFA). As shown in Table 4, the composite reliability of variables in this paper is above the 0.6 acceptable level, and variance extracted is also above the 0.5 level (Fornell and Larcker, 1981). Hence, the analysis result implies the internal consistency is within the acceptable range.

For the validity, good convergent validity should have: (1) factor loading above 0.5 and should reach the significance level; (2) composite reliability should exceed 0.6; (3) variance extracted should exceed 0.5 (Fornell and Larcker, 1981).

After eliminating items with factor loading lower than 0.5, the result shows one load-factor. The first-order factor analysis for the model displays that the goodness of fit for this model are GFI=0.90, AGFI=0.89, NFI=0.89, CFI=0.92 which are all above 0.89. This implies that the convergent validity for all variables is within the acceptable range (Bagozzi and Yi, 1988).

4.3 Overall Model Analysis

This paper employs Amos 16.0 to conduct the structure equation modeling (SEM) in order to understand the causal relationship of the structure model and goodness of fit for the research model. The SEM has two basic models, which are the measured model and the structural model. Measured model is also known as Confirmatory Factor Analysis (CFA). Its purpose is to assess the reliability and effect between measured variables and latent variables and predict the significance level of parameters. Structural model is also known as the causal model which is a multivariate statistics technique based on regression. Its purpose

74

is to inspect the causal relationship between patent variables. Under these two models, researchers are able to obtain the most suitable causal model by repeated tests when they fail to obtain the needed observed information for their models.

4.3.1 Model Fit Analysis

According to the suggestion of Bagozzi and Yi (1988), the fitness analysis of the model includes preliminary fit criteria, overall model fit, and fit of internal structure of model.

Preliminary fit criteria should meet the following: (1) the measurement error (ε) should not be negative; (2) factor loading (λ) should be within 0.5 and 0.95; (3) factor loading should reach the significance level (Bagozzi and Yi, 1988).

Table 4 shows that the factor loading for each variable falls between 0.50 and 0.95, and there is no negative in measurement error. Overall, it is within the acceptable range.

The main purpose is to assess the model fit between the overall model and observed data. As shown in Table 5, the actual values of this paper are close to or within the standard. So this paper has reasonable overall model fit.

Table 5 Overall Model Fit

Goodness of Fit Index Criterion Actual Value

CMIN/DF < 3 2.01

GFI > 0.9 0.90

AGFI > 0.9 : Good

> 0.8: Acceptable 0.89 NFI > 0.9 : Good

> 0.8: Acceptable 0.89

CFI > 0.9 0.92

RMSEA < 0.05 0.05

RMR < 0.05 0.04

According to Bagozzi and Yi (1988), the fit of internal structure of the model should have: (1) factor loading greater than 0.5 and reaches significance level; (2) composite reliability greater than 0.6; (3) variance extracted greater than 0.5. Table 4 shows that factor loading for each variable falls between 0.50 and 0.95, the composite reliability is greater than 0.6, and variance extracted is greater than 0.5. Hence, the internal structure is within the acceptable range.

4.3.2 Path Analysis

Table 6 presents the analysis result for the paths of each variable. Figure 2 displays the analysis result for SEM.

(1) Correlation between moral perception, moral judgment, and moral intention In accordance with the path analysis result of Table 6, moral perception has a significant positive relation with moral judgment (path coefficient = 0.55), and moral judgment also has a significant positive relation with moral intention (path coefficient = 0.80). The empirical result proves that internal auditors have higher moral judgment on audit issues if they recognize the importance and magnitude of consequences of the issues. Similarly, when they have higher moral judgment on ethical issues, internal auditors are able to come up with a behavior intention that is appropriate for this judgment. Hence, hypotheses H1a and H1b are supported. The empirical result proves the propositions of Rest (1986), and O’Leary and Stewart (2007).

(2) Correlation between moral intensity and moral decision-making model

Table 6 presents moral intensity only having a positive effect on moral judgment (path coefficient = 0.43). Empirical results find that when internal auditors have higher moral intensity on audit issues, the degree of damage of these issues, probability of happening, and immediacy are closely related. If internal auditors confront an audit issue that relates to the interests of the entire society, their legal and moral responsibilities (judgment) are heavier. This means higher moral intensity would strengthen the standard of the internal auditor’s moral judgment. So hypothesis H2b is supported. However, moral intensity has no direct influence on moral perception (path coefficient = 0.08) and moral intention (path coefficient = 0.22). To infer its reason, this paper believes this is

76

caused by the heavier legal and moral responsibilities of auditors in all countries (Sweeney, Arnold, and Pierce, 2010). Hence, moral intensity directly affects the moral judgment factors in the moral decision-making process.

(3) Correlation between moral maturity and moral decision-making model

Similar to the previous hypothesis testing of moral intensity on the moral decision-making model, moral maturity only has positive influence on moral judgment (path coefficient = 0.17), so hypothesis H3b is supported. Empirical results find that internal auditors with higher moral maturity also have prosocial behavior. They tend to make behavioral decisions which conform to the social norms. So they will think twice about the impact of “reducing audit quality”

which affects their moral judgment positively. Nevertheless, moral maturity has no significant and direct influence on moral perception (path coefficient = 0.13) and moral intention (path coefficient = 0.03). The possible reason may be that internal auditors try to cover up their perception of the ethical issues to avoid hurting others when there are any controversial issues. Moral maturity has no significant influence on moral intention because people with higher moral maturity tend to be considerate of the interests of others in order to prevent hurting another person. So, moral intention is derived from the moral judgment s t a g e . H y p o t h e s i s H3 i s n o t s u p p o r t e d .

(4) Correlation between obedience pressure and moral decision-making model For the correlation between obedience pressure and moral recognition, the result shows that obedience pressure has a significant negative relation with moral recognition (path coefficient = -0.58), so hypothesis H4a is supported. The result also explains that internal auditors tend to eliminate audit reports that are unfavorable to the corporations when they are under invisible, indirect potential pressure, which includes the chance of promotion, job relocation, and abolishment controlled by supervisors (Coram et al., 2008). For this reason, internal auditors may reduce the audit quality. Therefore, internal auditors must have sufficient moral perception when facing “audit issues.” Besides, obedience pressure also has a significant negative relation with moral judgment (path coefficient = -0.31),

hypothesis H4b is supported. The result also explains that negative obedience pressure would force internal auditors to make inappropriate auditing judgments.

When the internal auditors could not implement the normal audit program, it will result in avoiding discovering any deficiency (normal judgment). The empirical result confirms the finding of DeZoort and Lord (1997).

Figure 2