13 also in line with prior research eg guiso et al

13
Also in line with prior research (e.g., Guiso et al. 2006 , 2008 ) are the correlations between economic
development ( GDPPERCAP ) on the one hand and religiosity ( RELIGIOSITY ) (r ¼± 0.65) and societal trust ( SOCTRUST ) (r ¼
0.61) on the other. Moreover, and also consistent with prior research (e.g., Hooghiemstra et al. 2015 ), we observe high
correlation coefficients between GDP per capita ( GDPPERCAP ) and collectivism (IGC) (r ¼± 0.75). Finally, we find negative
correlations between SOCTRUST on the one hand and PD (r ¼± 0.61) and IGC (r ¼± 0.65) on the other. To reduce concerns
about multicollinearity, we obtain variance inflation factors. These VIFs are all less than 5.0, and the average VIFs in the
analyses are less than 2.7, indicating that multicollinearity should not be an issue. To minimize the impact of extreme values,
we winsorize each of the continuous variables used in the regressions at the top and bottom 1 percent.
Regression Results
Table 3 presents the coefficients and Huber-White robust standard errors (in brackets) from ordinary least squares
regressions. In Column (1) we present the results of a ‘‘ controls only ’’ analysis. In Columns (2) to (5), we add IGC, PD,
RELIGIOSITY, and SOCTRUST, respectively. Finally, Column (6) presents the results based on the full model.
In Column (2) of Table 3, we see that the coefficient of IGC is significantly negative ( b ¼± 5.93; p-value , 0.01). In the
full model (Column (6)), we find a comparable result ( b ¼± 8.81; p-value , 0.01). These findings, first of all, lend support to
H1 and suggest that the level of (in-group) collectivism and the extent to which auditors follow fraud risk assessment
procedures are negatively associated. More specifically, this finding lends support for our notion that auditors may be less likely
to engage in rigorous fraud risk assessment procedures so as to avoid the risk of having to put probing, confrontational
questions to management—i.e., that maintaining harmony in relationships is prevalent in collectivistic cultures.
Furthermore, Column (3) of Table 3 shows that the association between PD and COMPLIANCE is not significant ( b ¼
4.07; n.s.). Column (6) of Table 3 shows a marginally significant and positive association between COMPLIANCE and PD ( b ¼
7.22; p-value , 0.10). These findings contrast with the univariate comparison reported in Panel C of Table 1. The results
suggest that, after controlling for other firm-level and country-level variables, power distance is no longer associated with
COMPLIANCE and that smaller differences in PD are not associated with differences in COMPLIANCE . Taken together, we
therefore reject H2, which predicted an association between auditor compliance with fraud risk assessment procedures and
power distance.
14
The lack of a significant result seems to suggest that this association may be a more complex relation. It may
be that alternative and potential opposing loyalties to power exist. For example, the literature on client-auditor relationships
suggests that the client has the option to switch auditor and, hence, may be seen as the more powerful party (e.g., McKinnon
1984 ; Patel et al. 2002 ; Lin and Fraser 2008 ). Consequently, in high power distance countries, auditors may be