17 / 372
臺大管理論叢
第
26
卷第
2
期
17
Table 6 reports the univariate analysis results. Panel A uses
IMS_D
to identify industry
expert. It shows both
BETR
and
CETR
are statistically lower amongst the industry expert
auditors than those of non-expert auditors from both t-test and Wilcoxon rank-sum test, and
BTD
of industry experts’ clients is statistically higher than that of non-experts’ clients, which
means the level of corporate tax avoidance is higher when auditor is an industry expert.
Panel B uses
IPS_D
to identify industry expert and demonstrates the same results. Overall,
Table 6 implies a strong relationship between the auditor industry expertise and clients’ tax
avoidance. However, this study needs further analysis to confirm this association, because
univariate analysis only uses the dummy variables to proxy auditor industry expertise and
does not consider other determinants of tax avoidance aside from auditor industry expertise.
Table 6 Univariate Analysis
Panel A: Use industry market share (
IMS_D
) to identify industry expert
Mean
Median
Expert
(N = 1001)
Non-expert
(N = 6691)
Dif
T-stat
Expert
(N = 1001)
Non-expert
(N = 6691)
Dif
Z-stat
BETR
0.1860 0.1942 -0.008**
-2.43 0.1799 0.1817 -0.002*
-1.89
CETR
0.2605 0.2673 -0.007*
-1.82 0.2802 0.2836 -0.003*
-1.92
BTD
17.8828 17.0053 0.877*** 14.21 17.7728 17.0558 0.717*** 12.17
Panel B: Use industry market share (
IPS_D
) to identify industry expert
Mean
Median
Expert
(N = 754)
Non-expert
(N = 6938)
Dif
T-stat
Expert
(N = 754)
Non-expert
(N = 6938)
Dif
Z-stat
BETR
0.1925 0.1995 -0.007*
-1.82 0.1798 0.1861 -0.006*** -2.78
CETR
0.2598 0.2755 -0.016*** -3.72 0.2822 0.2987 -0.017*** -3.36
BTD
17.3957 17.0883 0.307***
4.37 17.3384 17.1080 0.230***
3.61
Note: This table reports the univariate analysis results. The sample includes 7,692 firm-year
observations for the period 2008-2012. Panel A uses
IMS_D
to identify industry expert and
examines the difference of clients’ tax avoidance between industry expert sample and non-
expert sample through t-test and Wilcoxon rank-sum test. Panel B uses
IPS_D
to identify
industry expert and also examines the difference of clients’ tax avoidance between two
independent samples. T-statistics are in the parentheses. ***, **, * stand for a statistical
significant level of 1%, 5% and 10% respectively.