Page 81 - 33-2

NTU Management Review Vol. 33 No. 2 Aug. 2023

3. Methodology

We use a sample of Taiwan listed firms from 2013 to 2018 to analyze the
We use a sample of Taiwan listed firms from 2013 to 2018 to analyze the hypotheses
hypotheses in this study. We employ the ordinary least squares regression based
in this study. We employ the ordinary least squares regression based on the following
on the following equation:
equation:

���
= + ��� + + ��� × +
���
�
�
�
���
�
���
�
+ + + + +
� ��� � ��� � ��� � ��� � ���
+ ��� + ��� + + ���
��
��
��
���
��
+ + + + . (1)
���
���
��
��
���
We use textual narratives written in Mandarin Chinese as a model
We use textual narratives written in Mandarin Chinese as a model of readability of
to estimate the readability of content from the sample firms’ annual reports. Using the
readability to estimate the readability of content from the sample firms’ annual
model established by Sung, Chen, Lee, Cha, Tseng, Lin, Chang, and Chang (2013),
reports. Using the model established by Song et al., (2013), the formula for
the formula for computing Readability takes into account the total number of difficult
computing takes into account the total number of difficult words
words (multiplied by 0.01), the percentage of simple sentences (multiplied by -0.86), the
(multiplied by 0.01), the percentage of simple sentences (multiplied by -0.86), the
logarithmic mean of content word frequency (multiplied by -1.45), and the total number
logarithmic mean of content word frequency (multiplied by -1.45), and the total
of personal pronouns (multiplied by 0.02), with a constant value of 4.53 added to the sum.
number of personal pronouns (multiplied by 0.02), with a constant value of 4.53
The use of more complex or obscure words, as well as a higher proportion of complex
added to the sum. The use of more complex or obscure words, as well as a higher
sentence structures, can result in an increase in Readability. That is, a textual narrative
proportion of complex sentence structures, can result in an increase in
with a higher Readability is less readable and requires a higher level of education from
. That is, a textual narrative with a higher is less
the reader to be understood. Next, to measure the degree of tax avoidance, we adopt the
effective tax rate and the long-term cash effective tax rate as proxies. Then, to make the
readable and requires a higher level of education from the reader to be understood.
results more easily interpretable, we multiply each of these proxies by -1. The higher the
Next, to measure the degree of tax avoidance, we adopt the effective tax rate and
value of the converted results is, the higher the degree of tax avoidance is. If the degree of
the long-term cash effective tax rate as proxies. Then, to make the results more
tax avoidance is positively associated with the readability score (β > 0), H1 is supported.
easily interpretable, we multiply each of these proxies by −1. The higher the value
1
H2 focuses on the coefficient β in the interaction between tax avoidance and auditor
3
of the converted results is, the higher the degree of tax avoidance is. If the degree
industry specialization. If auditor industry specialization serves as an effective corporate
of tax avoidance is positively associated with the readability score ( >0), H1 is
governance mechanism, they attenuate the negative relationship between the degree of tax
�
supported. H2 focuses on the coefficient in the interaction between tax
avoidance and financial readability (β < 0). �
3
avoidance and auditor industry specialization. If auditor industry specialization
serves as an effective corporate governance mechanism, they attenuate the
negative relationship between the degree of tax avoidance and financial readability
( <0).
�
73
4. Findings
The empirical results indicate that the higher the degree of tax avoidance, the
lower the readability of firms’ financial statements. However, if a firm retains an
industry specialist auditor, the damage caused by tax avoidance to financial
38