Independent Directors and Corporate Financial Performance – A Hong Kong Perspective (A DBA Dissertation Completed in August 2005)
Chapter 4 – Analysis
4.1 Summary of
Sampling Results
Statistical testing of the data
returned results rejecting the null hypotheses H01 (including
both H01a and H01b), H02
and H03. Whereas the outcomes in respect of H02
might be regarded more equivocal than those in respect of the other two
hypotheses, the evidence overall has been sufficient to justify acceptance of
the thrust of the three alternate hypotheses developed in Chapter 2 that a
relationship exists between board composition and corporate financial
performance, and that the strength of relationship varies among certain groups
of companies, though the strength of relationship found for H3 was in
opposite direction to HR3. The three findings are,
respectively, the proportion or number of independent directors in the board of
directors is positively associated with company financial performance; the
relationship between board composition and company financial performance is
stronger in growth-oriented companies than non-growth-oriented companies; and
the relationship between board composition and company financial performance is
stronger in companies majority-owned by mainland Chinese interests than
companies majority-owned by non-mainland Chinese interests.
Whilst the 2003 HKEx list of
companies included 861 on the Main Board and 187 on the GEM, the following
companies were excluded from the sample for analysis:
·
Financial institutions – these are separately
regulated by the regulatory bodies, a process which might distort the
relationship between independent directors and company financial performance.
Any possible relationship might have been moderated or mediated by the
additional regulations. Moreover, the components of these companies’ financial
statements are different from those of other companies, making, for example,
ROA, a less useful comparative measure. This could arise by virtue of the
specific line of business of financial institutions which principally involves
accepting deposits or premiums to making loans or investments. Consequently, a
direct comparison of the accounting returns of financial institutions with
those of other industries might be misleading.
·
Companies with missing data including –
o
Those with losses in the previous year, hence
for which a meaningful P/E could not be calculated.
o
Those suspended for trading on 24 September
2004, the reference date for market data, or failed to be listed on 24
September 2004, for which MV did not exist.
·
Companies with out-dated financial data – those
for which latest published annual reports (when data collection for this study
was completed in March 2005) were for years ended earlier than 31 July 2004.
It might be argued, for example,
that those companies with losses in the previous year, or suspended from
trading on our reference date for market data, were such as a result of bad CG.
However, for the sake of validity and reliability, this study required its
sample to comprise strictly only companies with comparable data on the same
chosen date. Any deviations from that criterion would likely compromise the
validity and reliability of its research results. It is, therefore, decided
that the aforementioned three groups of companies be excluded from further
analysis. After the exclusion, the sample size was 628 (see Table 4.1):
Table
4.1 – Sample Composition
Company Category
|
Number
|
%
|
|
Main
|
Large
Companies (excl. CEs & CAs)
|
36
|
5.7%
|
Medium
Companies (excl. CEs & CAs)
|
115
|
18.3%
|
|
Small
Companies (excl. CEs & CAs)
|
295
|
47.0%
|
|
Main
& GEM
|
CEs
|
78
|
12.4%
|
CAs
|
53
|
8.4%
|
|
GEM
|
Excl.
CEs & CAs
|
51
|
8.1%
|
Total
|
628
|
100.0%
|
|
4.2 Hypothesis 1
HR1 states that
the proportion or number of independent directors in the board of directors is
associated with company financial performance. Conversely, H01
assumes that the proportion or number of independent directors in the board of
directors is not associated with company financial performance.
4.2.1 Preliminary
Scatterplots
Simple linear
regression works well only if the two variables tested have a linear relationship.
A scatterplot is capable of revealing the relationship between two variables
through the pattern formed by the full set of dots. Therefore, before
running regression analyses, as a preliminary test for linear relationship, the
whole sample was used to construct the following scatterplots, with % of NEDs,
as the x-axis, plotting against ROE, ROA, P/E and MV/BV respectively (see Fig.
4.1):
Figure
4.1 – Scatterplots of the Whole Sample
Figure 4.1A – ROE versus % of NEDs Figure 4.1B – ROA versus % of
NEDs
Figure 4.1C – P/E versus % of NEDs Figure 4.1D – MV/BV versus %
of NEDs
The four scatterplots in Fig.
4.1 showed that a broadly linear relationship might exist between the
independent and dependent variables. It was also apparent that outliers, and in
some cases extreme outliers, existed. Outliers are notorious for causing the size of a correlation coefficient
to understate or exaggerate the strength of the relationship between two
variables (Huck, 2004), and thus cannot be ignored. Accordingly, the
issue of outliers will be specifically dealt with in subsequent sections.
4.2.2 Preliminary
Descriptive Statistics
It is helpful to gain some basic
understanding of the data by exploring their distributional shape, central
tendency and dispersion. Some selected descriptive statistics are shown in Table
4.2:
Table
4.2 – Descriptive Statistics of the Whole Sample
Table 4.2 reveals that the
mean % of NEDs of all the 628 companies was 0.4362 and the mean % of
INEDs was 0.3211. The mean No. of NEDs was 3.66 while the mean No.
of INEDs was 2.55. The medians of these four proxies for independent
director were reasonably close to the means, indicating that the sampling
distributions were reasonably symmetric. This was also evident by looking at the
skewness figures which ranged from 0.375 to 1.732. The respective standard
deviations of these four proxies with symmetric shape are therefore useful
indicators.
In contrast, the four proxies for
company financial performance skewed far to the right. The means of ROE,
ROA, P/E and MV/BV were 0.3108, 0.1575, 53.31 and 1.85,
while the respective medians were much lower at 0.0978, 0.514, 12.30 and 1.06.
The positive skewness ranging from 12.549 to 24.617 were indeed very high,
confirming the existence of extreme outliers. Since all the four proxies were
ratios, they could be extremely high were the denominators exceptionally small.
Such situations could arise when earnings, equity value or total asset value in
the previous financial year were abnormally low, which do happen occasionally
in the business world. It is therefore not surprising to find outliers in these
variables. For skewed distribution, medians and interquartile ranges are more
informative than means and standard deviations.
Despite the existence of
outliers, in order to get a complete picture, it is still useful to perform
simple linear regression for the whole sample to get a feel for the
relationships between the variables.
4.2.3 Preliminary
Linear Regression Analyses
All the Main Board and GEM
companies were used in running simple linear regressions. As shown in Table
4.3, no significant relationships were identified since all p-values
(2-tailed significance level) were greater than 0.05:
Table 4.3 –
Preliminary Linear Regression Results
Independent
variable
|
Dependent
variable
|
r
|
R2
|
Sig.
(2-tailed)
|
%
of NEDs
|
ROE
|
-0.025
|
0.001
|
0.534
|
ROA
|
-0.029
|
0.001
|
0.475
|
|
P/E
|
0.013
|
0.000
|
0.746
|
|
MV/BV
of Equity
|
0.067
|
0.004
|
0.094
|
|
%
of INEDs
|
ROE
|
0.008
|
0.000
|
0.851
|
ROA
|
0.012
|
0.000
|
0.762
|
|
P/E
|
-0.013
|
0.000
|
0.744
|
|
MV/BV
of Equity
|
0.009
|
0.000
|
0.818
|
|
No.
of NEDs
|
ROE
|
-0.035
|
0.001
|
0.381
|
ROA
|
-0.043
|
0.002
|
0.288
|
|
P/E
|
0.005
|
0.000
|
0.908
|
|
MV/BV
of Equity
|
0.050
|
0.003
|
0.207
|
|
No.
of INEDs
|
ROE
|
-0.029
|
0.001
|
0.469
|
ROA
|
-0.032
|
0.001
|
0.428
|
|
P/E
|
-0.026
|
0.001
|
0.515
|
|
MV/BV
of Equity
|
0.007
|
0.000
|
0.858
|
To isolate the effects of size,
growth-orientation and shareholder background, the whole sample was divided
into different groups for running separate regression analyses. Different
groups could then be compared in respect of the strength of relationship
between the tested variables. The groups included:
·
Main Board – All Companies (excluding CEs and
CAs)
·
Main Board – Large Companies (excluding CEs and
CAs)
·
Main Board – Medium Companies (excluding CEs and
CAs)
·
Main Board – Small Companies (excluding CEs and
CAs)
·
GEM – All Companies (excluding CEs and CAs)
·
Main Board – High MV/BV Companies (excluding CEs
and CAs)
·
Main Board – Low MV/BV Companies (excluding CEs
and CAs)
·
Main Board and GEM – CEs
·
Main Board and GEM – CAs
High MV/BV companies were defined
for the purposes of this study to delineate those companies with a MV/BV
exceeding 3, which represented a high ratio. By implication, low MV/BV
companies were those with MV/BV ratios of 3 or less. As such, 52 high MV/BV
companies and 394 low MV/BV companies were identified in the 446 non-CE/CA Main
Board companies.
No statistically significant
relationships were identified in the majority of regression runs. The only significant
associations found were in respect of Main Board – Small Companies (excluding
CEs and CAs) between % of INEDs and P/E (r = 0.117, p-value
= 0.045); Main Board and GEM – CEs between % of NEDs and MV/BV (r
= 0.311, p-value = 0.006); Main Board and GEM – CAs between % of NEDs
and MV/BV (r = 0.292, p-value = 0.034), as well as between
% of INEDs and MV/BV (r = 0.352, p-value = 0.010).
However, these associations must
be viewed with caution. As discussed earlier, outliers existed when the whole
sample of 628 companies was viewed as a group. Indeed, outliers existed in each
of the aforementioned sub-groupings of companies. The presence of outliers
could seriously distort the regression line between the independent and
dependent variables in each of the company category. The significant
relationships identified might not in fact be significant, and the
insignificant relationships in the majority of cases might not in fact be the
case. As described earlier in Section 4.2.2, the four proxies for company financial
performance were ratios, and hence could be extremely high were the
denominators exceptionally small. Such situations could arise when earnings,
equity value or total asset value in the previous financial year were
abnormally low. It is therefore not surprising to find outliers in these
variables. However, the presence of these outliers could seriously distort
regression results.
4.2.4 Sample
Without Outliers
Consequently, all outliers were
eliminated for another round of regression analyses. Outliers in this study
were defined as data above the upper fence (Q3 + 1.5 x Interquartile Range), or
below the lower fence (Q1 – 1.5 x Interquartile Range) in a boxplot. After such
elimination, the sample size was reduced from the original 628 companies to as follows
(see Table 4.4):
Table 4.4 –
Sample Composition Without Outliers
Company Category
|
Number
|
%
|
|
Main
|
Large
Companies (excl. CEs & CAs)
|
33
|
6.3%
|
Medium
Companies (excl. CEs & CAs)
|
97
|
18.6%
|
|
Small
Companies (excl. CEs & CAs)
|
238
|
45.6%
|
|
Main
& GEM
|
CEs
|
68
|
13.0%
|
CAs
|
45
|
8.6%
|
|
GEM
|
Excl.
CEs & CAs
|
41
|
7.9%
|
Total
|
522
|
100.0%
|
|
4.2.5 Scatterplots
The reduced sample was used to
construct scatterplots similar to those presented earlier in respect of the
original sample, again with % of NEDs, as the x-axis, plotting against ROE,
ROA, P/E and MV/BV respectively (see Fig. 4.2):
Figure
4.2 – Scatterplots of the Whole Sample Without Outliers
Figure 4.2A – ROE versus % of NEDs Figure 4.2B – ROA versus % of NEDs
Figure 4.2C – P/E versus % of NEDs Figure 4.2D – MV/BV versus % of
NEDs
Fig. 4.2 reveals no
curvilinear or other complex patterns in the scatterplots. As expected, the
data were much more clustered together than before the elimination of outliers.
Straight line relationships might then be expected to exist.
4.2.6 Descriptive
Statistics
Descriptive statistics in a
pattern similar to those reported in Section 4.2.2 are shown in Table 4.5:
Table
4.5 – Descriptive Statistics of the Whole Sample Without Outliers
The means and medians of all the
variables became reasonably close. Their closeness was supported by the degrees
of skewness which ranged from 0.275 to 1.703. That showed that the sampling
distributions of all the variables were reasonably symmetric.
4.2.7 Linear
Regression Analyses
Compared with the sample before
eliminating outliers, many more significant associations, with p-values
less than 0.05, were found after simple linear regression runs (see Table
4.6):
Table 4.6 –
Linear Regression Results
Company
Category
|
Indep.
Var.
|
Dep. Var.
|
r
|
p-value
|
Main
and GEM – All Companies
|
%
of NEDs
|
P/E
|
0.162
|
0.000
|
MV/BV
|
0.152
|
0.000
|
||
No.
of NEDs
|
ROE
|
-0.097
|
0.026
|
|
ROA
|
-0.096
|
0.028
|
||
P/E
|
0.268
|
0.000
|
||
MV/BV
|
0.172
|
0.000
|
||
No.
of INEDs
|
P/E
|
0.141
|
0.001
|
|
Main
Board – All Companies
|
%
of NEDs
|
ROE
|
-0.116
|
0.025
|
(excl.
CEs & Cas)
|
No.
of NEDs
|
ROE
|
-0.157
|
0.002
|
ROA
|
-0.142
|
0.006
|
||
P/E
|
0.236
|
0.000
|
||
No.
of INEDs
|
P/E
|
0.117
|
0.025
|
|
Main
Board – Medium Companies
|
%
of INEDs
|
ROE
|
0.245
|
0.016
|
(excl.
CEs & CAs)
|
ROA
|
0.214
|
0.036
|
|
Main
Board – Small Companies
|
%
of NEDs
|
ROE
|
-0.164
|
0.011
|
(excl.
CEs & CAs)
|
MV/BV
|
-0.181
|
0.005
|
|
No.
of NEDs
|
ROE
|
-0.218
|
0.001
|
|
ROA
|
-0.175
|
0.007
|
||
MV/BV
|
-0.200
|
0.002
|
||
No.
of INEDs
|
ROE
|
-0.147
|
0.023
|
|
ROA
|
-0.166
|
0.010
|
||
MV/BV
|
-0.165
|
0.011
|
||
GEM
– All Companies
|
%
of NEDs
|
MV/BV
|
0.332
|
0.034
|
(excl.
CEs & CAs)
|
No.
of NEDs
|
MV/BV
|
0.506
|
0.001
|
Main
Board – High MV/BV Companies
|
%
of NEDs
|
P/E
|
0.325
|
0.031
|
(excl.
CEs & CAs)
|
No.
of NEDs
|
P/E
|
0.355
|
0.018
|
Main
Board – Low MV/BV Companies
|
%
of NEDs
|
ROE
|
-0.143
|
0.008
|
(excl.
CEs & CAs)
|
ROA
|
-0.108
|
0.046
|
|
No.
of NEDs
|
ROE
|
-0.178
|
0.001
|
|
ROA
|
-0.144
|
0.008
|
||
P/E
|
0.217
|
0.000
|
||
No.
of INEDs
|
P/E
|
0.137
|
0.011
|
|
Main
& GEM – CEs
|
%
of NEDs
|
MV/BV
|
0.306
|
0.011
|
Main
& GEM – CAs
|
%
of NEDs
|
ROE
|
0.315
|
0.035
|
ROA
|
0.302
|
0.043
|
||
MV/BV
|
0.318
|
0.033
|
||
%
of INEDs
|
ROE
|
0.442
|
0.002
|
|
ROA
|
0.530
|
0.000
|
||
No.
of NEDs
|
MV/BV
|
0.311
|
0.038
|
(Note: R2 not
separately shown with p-values lower than 0.05)
Additional scatterplots were
drawn in respect of each of the relationship identified above. Other than
linear relationship, no evidence suggesting curvilinear or other complex
relationships existed.
To further support the existence
of linear relationship, it is necessary to check the residuals to ensure they
fulfil the assumptions of normality, zero mean, homogeneity of variance and
independence (Carver & Nash, 2005). Thus, residuals plots were drawn to
examine the appropriateness of the models. In all the identified relationships,
the residuals lay along a 45° upward sloping diagonal line in the respective
normal probability plots, revealing that the residuals were normally
distributed. However, the scatterplots of the standardised residuals compared
with the standardised estimated values showed that in most of the identified
relationships, the residuals “fanned out” either from left to right or from
right to left. These patterns indicated that the homogeneity of variance
assumption might have been violated, thus making the models concerned
inappropriate. Excluding these “suspicious” cases, the remainder showed
residuals randomly scattered in an even, horizontal band around a residual
value of zero[1]. These satisfied the four
aforementioned assumptions and could therefore be considered reliable. They
included (see Table 4.7):
Table 4.7 –
Linear Regression Results Satisfying Regression Assumptions
Company
Category
|
Indep.
Var.
|
Dep. Var.
|
r
|
p-value
|
Main
and GEM – All Companies
|
%
of NEDs
|
P/E
|
0.162
|
0.000
|
MV/BV
|
0.152
|
0.000
|
||
Main
Board – Medium Companies
|
%
of INEDs
|
ROE
|
0.245
|
0.016
|
(excl.
CEs & CAs)
|
ROA
|
0.214
|
0.036
|
|
GEM
– All Companies
|
%
of NEDs
|
MV/BV
|
0.332
|
0.034
|
(excl.
CEs & CAs)
|
No.
of NEDs
|
MV/BV
|
0.506
|
0.001
|
Main
Board – High MV/BV Companies
|
%
of NEDs
|
P/E
|
0.325
|
0.031
|
(excl.
CEs & CAs)
|
||||
Main
& GEM – CEs
|
%
of NEDs
|
MV/BV
|
0.306
|
0.011
|
Main
& GEM – CAs
|
%
of NEDs
|
ROE
|
0.315
|
0.035
|
ROA
|
0.302
|
0.043
|
||
MV/BV
|
0.318
|
0.033
|
||
%
of INEDs
|
ROE
|
0.442
|
0.002
|
|
ROA
|
0.530
|
0.000
|
The p-values of these
findings were less than 0.05, and indeed in some cases were less than 0.01,
hence the first null hypothesis was rejected at a significance level of 5%. In
other words, independent directors were found to be associated with company
financial performance. It could be seen that the % of NEDs, for listed
companies as a whole in Hong Kong, was positively associated with market-based
returns including P/E and MV/BV of equity. The % of INEDs
for medium-sized companies on the Main Board was positively related to
accounting returns including ROE and ROA. Market-based returns
were also positively associated with % or No. of NEDs for
companies listed on the GEM, high MV/BV Main Board companies and CEs. It is of
particular note that for CAs both accounting returns and market-based returns
were found to be positively associated with independent directors.
4.3 Hypothesis
2
Since the outliers identified
could seriously distort any linear relationship, only the sample excluding
outliers was used for further statistical testing of hypotheses HR2
and HR3. The respective correlation coefficients found with
GEM companies were compared against those found with Main Board companies,
using Fisher’s r-to-z transformation. Fisher’s r-to-z transformation
is chosen because of its ease of calculation and comprehension. Likewise, high
MV/BV Main Board companies, as another proxy for growth-oriented companies,
were compared with low MV/BV Main Board companies for sensitivity analysis. In
addition, to control for the possible effect of size, GEM companies were
separately compared with three sub-groups of Main Board companies, namely, the large,
medium, and small. For a two-tailed test, if the absolute value
of z is greater than 1.96, a 5% significance level is attained (Bogartz,
1994). The following significant differences were found (see Table 4.8):
Table 4.8 – Z
Test between Growth-Oriented Companies and Non-Growth-Oriented Companies
Indep. Var.
|
Dep. Var.
|
z value of
the difference
|
|
GEM
with Main – All Companies
|
%
of NEDs
|
MV/BV
|
2.2366
|
(excluding
CEs & CAs)
|
No.
of NEDs
|
MV/BV
|
3.0361
|
GEM
with Main – Medium Companies
|
%
of INEDs
|
ROE
|
-2.7862
|
(excluding
CEs & CAs)
|
ROA
|
-2.2669
|
|
No.
of NEDs
|
MV/BV
|
3.4633
|
|
GEM
with Main – Small Companies
|
%
of NEDs
|
MV/BV
|
3.0203
|
(excluding
CEs & CAs)
|
No.
of NEDs
|
ROE
|
1.9800
|
ROA
|
1.9990
|
||
MV/BV
|
4.3471
|
Overall, the positive
relationship between NEDs and MV/BV in GEM companies was
significantly stronger than that in Main Board companies as a whole. The same
was true between GEM and medium-sized Main Board companies as well as between
GEM and small Main-Board companies. The positive association between No. of
NEDs and accounting returns was significantly stronger in GEM companies
than small Main-Board companies. On the contrary, the positive association
between % of INEDs and accounting returns was significantly stronger in
medium-sized Main Board companies than GEM companies. However, no significant
difference was identified between GEM and large Main Board companies, or
between high MV/BV and low MV/BV Main Board companies.
There was, therefore, sufficient
evidence to reject the second null hypothesis, though the relationship could be
either stronger or weaker in growth-oriented companies than non-growth-oriented
companies. Whether the relationship is stronger or weaker is dependent on the
size of the non-growth-oriented companies and whether the performance measure
was accounting-based or market-based.
4.4 Hypothesis
3
In respect of companies
majority-owned by mainland Chineses interests and companies majority-owned by
non-mainland Chinese interests, significant differences identified, as
represented by z value over 1.96, are tabulated below (see Table 4.9):
Table 4.9 – Z
Test between Companies Majority-Owned by Mainland Chinese Interests and
Companies Majority-Owned by Non-Mainland Chinese Interests
Groups Concerned
|
Indep. Var.
|
Dep. Var.
|
z value of
the difference
|
CEs
with Main – All Companies
|
%
of NEDs
|
MV/BV
|
2.6173
|
CEs
with Main – Small Companies
|
%
of NEDs
|
MV/BV
|
2.8548
|
No.
of NEDs
|
MV/BV
|
2.2062
|
|
CAs
with Main – All Companies
|
%
of NEDs
|
ROE
|
3.2897
|
ROA
|
2.9277
|
||
MV/BV
|
2.7161
|
||
%
of INEDs
|
ROE
|
3.2755
|
|
ROA
|
4.2153
|
||
MV/BV
|
2.3100
|
||
No.
of NEDs
|
ROE
|
2.5062
|
|
ROA
|
12.4160
|
||
MV/BV
|
2.0932
|
||
No.
of INEDs
|
ROE
|
2.3664
|
|
ROA
|
2.7166
|
||
MV/BV
|
2.0719
|
||
CAs
with Main – Large Companies
|
%
of INEDs
|
ROE
|
3.0692
|
ROA
|
3.4095
|
||
No.
of NEDs
|
ROA
|
7.0747
|
|
No.
of INEDs
|
ROE
|
2.1748
|
|
CAs
with Main – Medium Companies
|
%
of NEDs
|
MV/BV
|
1.9863
|
%
of INEDs
|
ROA
|
2.3109
|
|
No.
of NEDs
|
ROA
|
9.8166
|
|
MV/BV
|
2.6660
|
||
No.
of INEDs
|
MV/BV
|
2.1811
|
|
CAs
with Main – Small Companies
|
%
of NEDs
|
ROE
|
2.8115
|
ROA
|
2.4551
|
||
MV/BV
|
2.9308
|
||
%
of INEDs
|
ROE
|
2.7379
|
|
ROA
|
3.4896
|
||
No.
of NEDs
|
ROE
|
2.2902
|
|
ROA
|
9.7477
|
||
MV/BV
|
2.9991
|
||
No.
of INEDs
|
ROE
|
2.1863
|
|
ROA
|
2.4864
|
||
MV/BV
|
2.6039
|
The positive relationship between
% of NEDs and MV/BV in CEs was significantly stronger than in
respect of the Main Board companies in the aggregate. A similar significant
difference was identified between CEs and small Main Board companies in
relation to NEDs and MV/BV. No other significant difference was
found between CEs and Main Board companies of other groupings or in respect of
other variables.
Very significant differences in
performance relationships were identified when CAs were compared with Main
Board companies. The association between independent directors and company
financial performance was significantly stronger in CAs than in Main Board
companies. The differences were evident in the relationship between virtually
all proxies for independent directors and ROE, ROA and MV/BV,
when Main Board companies were considered as a whole or divided into
sub-groups.
The strong evidence leads to the
rejection of H03. It is noteworthy that, contrary to HR3,
the positive association between independent directors and company financial
performance was significantly stronger in companies majority-owned by mainland
Chinese interests than by non-mainland Chinese interests.
The findings will be discussed in
detail in Chapter 5. Conclusions and the ensuing policy implications will also
be drawn and explained.
[1] The residual plots can be obtained from the author upon request.
They are not annexed to constrain the size of the dissertation.
No comments:
Post a Comment