Professional Documents
Culture Documents
Earnings Conservatism and Audit Committee Financial
Earnings Conservatism and Audit Committee Financial
Abstract
Using an Australian sample of 494 rm-year observations, this study nds that
accounting nancial expertise is the primary type of expertise that inuences
earnings conservatism, rather than nonaccounting nancial expertise. The
association between accounting nancial expertise and conservatism holds only
when the accounting nancial expert(s) on audit committees is (are) independent. Overall, results suggest that audit committee accounting nancial
expertise is important in recognising the asymmetrical timeliness of losses.
Findings provide a better understanding of the dynamics between audit
committee nancial expertise and earnings conservatism and demonstrate the
importance of accounting nancial expertise in improving nancial reporting
quality.
Key words: Conservatism; Corporate governance; Audit committee nancial
expertise
JEL classification: M41, M42, M48, G38
doi: 10.1111/ac.12042
1. Introduction
The primary objective of this study is to examine the association between
audit committee nancial expertise and earnings conservatism within an
Australian capital market setting. Debate about audit committees and the
The authors gratefully acknowledge the valuable comments from Steven Cahan (Editorin-Chief) and two anonymous reviewers. The authors also acknowledge the many
helpful suggestions of Robert Durand, Ross Taplin, Kamran Ahmed, Sue Wright,
Asheq Rahman, Ann Tarca, Ray Da Silva Rosa, Paul Gerrans, Paul Mather, Manzurul
Alam and both the discussant and participants at the AFAANZ 2012 conference,
seminar participants at the Curtin Business School 2012, the UWA Business School
2012 and the La Trobe Accounting Research Seminar 2012.
Received 16 March 2013; accepted 22 July 2013 by Steven Cahan (Editor in Chief).
2013 AFAANZ
280 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
2013 AFAANZ
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 281
Only one prior study of note (i.e. Krishnan and Visvanathan, 2008) examines the audit
committee nancial expertise and earnings conservatism association for a sample of the
US rms.
3
In 1999, the SEC, in its proposed rule, suggested a narrower denition of nancial
expertise (i.e. an audit committee member with professional qualications and/or
experience as a public accountant, auditor, principal or chief nancial ocer, controller,
principal or chief accounting ocer). However, subsequently in 2003, in its nal rule, the
SEC dened nancial experts broadly incorporating nonaccounting nancial experts
(i.e. an audit committee member with experience as a CEO or president of a for-prot
rm) (Securities and Exchange Commission, 2003). In this study, whenever the term
SEC (2003) denition is used, it refers to the broader denition of nancial experts in
SECs nal rule in 2003.
4
ASX CGC 2003 Recommendation 4.2 requires at least one member of the audit
committee to be a nancial expert and denes a nancial expert as a qualied
accountant or other nance professional with experience of nancial and accounting
matters. This denition is more stringent than the one prescribed by regulators in other
international settings (e.g. Blue Ribbon Committee, 1999; Securities and Exchange
Commission, 2003). However, the ASX CGC 2003 denition of nancial expertise is
similar to the one initially proposed by the SEC in 1999.
2013 AFAANZ
298 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 283
284 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
Davidson et al. (2004), for example, identify a signicant positive stock price
reaction to the appointment of an accounting nancial expert but not to a
nonaccounting nancial expert. Similarly, Dhaliwal et al. (2010) and Krishnan and Visvanathan (2008) examine the importance of both accounting and
nonaccounting nancial expertise and conclude that only accounting nancial
experts are associated with higher accruals quality and accounting conservatism respectively. Using an Australian sample, Baxter and Cotter (2009)
document a positive relationship between the existence of an accounting
nancial expert on the audit committee and the rms accruals quality.
Although the majority of past nancial expertise studies show that accounting
nancial expertise is the primary type of expertise that improves nancial
reporting quality, nonaccounting nancial expertise is also found to be a
valuable governance mechanism in improving audit committee eectiveness
(McDaniel et al., 2002; Goh, 2009). McDaniel et al. (2002) document that, in
comparison with accounting nancial experts, nonaccounting nancial experts
are more likely to raise concerns about high-salience nancial statement items
that are nonrecurring in nature and which receive press attention. In addition,
Goh (2009) nds that audit committee members with nonaccounting nancial
expertise are more eective in rectifying material internal control weaknesses.
As suggested by past research, both accounting expertise and nonaccounting
nancial expertise enhance the governance expertise of audit committee
members.
In relation to accounting nancial expertise, consistent with past literature,
we expect that with a more rened and in-depth knowledge of accounting
principles, practices and processes, an accounting nancial expert will be better
able to identify and recommend the most appropriate conservative accounting
policies. On the other hand, although nonaccounting nancial experts do not
have the necessary educational and professional qualications to be classied
as a nancial expert, they nevertheless have valuable knowledge both in
nancial (i.e. monitoring the preparation of nancial statements and certifying
them annually) and nonnancial matters (i.e. overseeing management and the
internal control structure). With knowledge both in accounting matters and
corporate management, nonaccounting nancial experts will be able to
identify, recommend or assess the reasonableness of managements use of
conservative accounting policies. Consistent with the usefulness of both
accounting and nonaccounting nancial experts in promoting conservative
accounting practices, we test the following hypotheses:
H2a: Audit committees accounting financial expertise is positively associated
with earnings conservatism.
H2b: Audit committees nonaccounting financial expertise is positively associated
with earnings conservatism.
2013 AFAANZ
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 285
A stratied random approach is used to control for potential size biases, and the 5-year
period is selected as it occurs after the introduction of key corporate governance reforms
in Australia (i.e. ASX CGC 2003).
Consistent with prior research (Ball et al., 2000; Givoly et al., 2007), nancial (133),
insurance (10), utility (30), initial public oering (106) and trust (92) rms are excluded.
Foreign incorporated and domiciled rms (64) are also excluded as nancial statements
of these entities are not necessarily prepared in accordance with the typical disclosure
requirements of ASX listed rms. To avoid undue inuences of unexpected share price
changes, 222 rms not continuously listed on the ASX throughout the entire observation
period (i.e. rms delisted and subsequently relisted) are eliminated from the initial
sample. Finally, 381 rms are excluded due to missing data.
2013 AFAANZ
286 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
Table 1
Sample selection and industry breakdown
Panel A: Sample selection
Number of rms listed on ASX as at 1 January 2004
Exclusions
Financial institutions
Insurance rms
Utility rms
IPO rms
Trusts
Foreign incorporated rms
Firms that are not continuously listed
Missing data
Total number excluded
Sample pool for random selection
Number randomly selected by quartiles per year
Excluded due to missing data
Final usable sample
Panel B: Sample firm break down by industry
ASX Industry
Consumer discretionary
Consumer staples
Energy
Health care
Industrials
Information technology
Materials
Telecommunication services
Total
2,128
(133)
(10)
(30)
(106)
(92)
(64)
(222)
(381)
100*5
No. Firm-Year
Observations
80
19
35
70
128
34
115
13
494
(1,038)
1,090
500
(6)
494
% of Sample
16.194
3.846
7.085
14.170
25.911
6.883
23.279
2.632
100
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 287
288 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
where Xjt is the operating prot after tax deated by market value of equity at
beginning of the scal year, DRjt is an indicator variable equal to 1 if RRAjt is
negative, and RRAjt is the annual share returns for the rm from 3 months
after the previous scal year to 3 months after the current scal year, adjusted
for the All Ordinaries Index over the same period.
To test H1, Equation (2) is developed. If audit committee members
with overall nancial expertise inuence conservatism, the coecient on
7
To calculate the earnings conservatism variables, share price data are obtained from
DataStream whilst accounting-related information is obtained from the FinAnalysis/
DatAnalysis Aspect Huntley Financial Database. Data for the three types of audit
committee nancial expertise are hand collected from the 2004-2008 annual reports of
each selected rm. Annual reports were obtained from the Annual Reports Collection
(Connect 4 Pty Ltd) and the FinAnalysis/DatAnalysis Aspect Huntley Financial
Database.
Although the Basu (1997) model is the most commonly used method to measure the
timeliness of earnings conservatism, it is not without limitations (Ball and Shivakumar,
2005; Dietrich et al., 2007). Dietrich et al. (2007), for example, argue that the dierence
in slopes predicted and reported by the Basu (1997) timeliness model reects sample
variance ratio and sample truncation biases. In addition, Ball and Shivakumar (2005)
note that the reverse regression approach assumes asymmetrical and ecient reaction to
economic news. Therefore, taking the limitations of the Basu (1997) model into account
and in an eort to capture an additional dimension to earnings conservatism, an
alternative measure of earnings conservatism proposed by Ball and Shivakumar (2005)
is also used.
2013 AFAANZ
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 289
To test H2a and H2b, Exptjt in Equation (2) is replaced with Expt_Accjt,
Expt_CEOjt, and nally, to determine the incremental eects of accounting and
nonaccounting nancial expertise on conservatism, Exptjt is replaced with both
Expt_Accjt and Expt_CEOjt concurrently. See Appendix 1 for the denitions of
all variables.
3.4.2. Accrual-based loss recognition
Whilst the Basu (1997) timeliness of earnings to news model focuses on the
timely recognition of negative news in share prices, Ball and Shivakumar (2005)
introduced a model based on the extent to which accruals are timely in
reecting cash ows. In the Ball and Shivakumar (2005) model, operating cash
ows are used to determine bad news and good news. Specically, Ball and
Shivakumar (2005) argue that conservatism exists when negative cash ows are
recognised earlier than positive cash ows. Equation (3) details the Ball and
Shivakumar (2005) model where the coecient on CFOjt*DCFOjt is predicted
to be positive and signicant if conservatism exists:
ACCjt b0 b1 DCFOjt b2 CFOjt b3 CFOjt DCFOjt ejt
where ACCjt is accruals (i.e. dierence between operating prot and cash ow
from operations) scaled by the book value of total assets at the beginning of the
scal year, DCFOjt is an indicator variable equal to 1 if CFOjt is negative, CFOjt
is cash ow from operating activities scaled by the book value of total assets at
the beginning of the scal year.
9
290 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
To test H2a and H2b, Exptjt in Equation (4) is replaced with Expt_Accjt,
Expt_CEOjt, and nally, to determine the incremental eects of accounting and
nonaccounting nancial expertise on conservatism, Exptjt is replaced with both
Expt_Accjt and Expt_CEOjt concurrently. See Appendix 1 for the denition of
all variables.
4. Empirical results
4.1. Descriptive statistics and correlations
Table 2 reports descriptive statistics for the nal useable sample. Average
deated operating prot after tax (Xjt) for the nal useable sample is negative,
whilst the median value is positive. In comparison, average market-adjusted stock
returns (RRAjt) for the nal useable sample are positive with the median being
negative. Importantly, the negative (positive) skewness of earnings (returns) is
consistent with the asymmetric timeliness of earnings and returns (Basu, 1997).
The mean and median for total accruals as a proportion of total assets (ACCjt) are
negative (i.e. 0.49 and 0.18 respectively). The results for ACCjt suggest the
existence of conservatism in the nancial reporting by rms in the sample.
The average (median) number and proportion of directors (#Expt_Accjt and
%Expt_Accjt) on the audit committee deemed as having accounting nancial
expertise are 0.97 (1.00) and 31.39 per cent (33.33 per cent), respectively. For
the nal useable sample, the average (median) number of audit committee
members (#Expt_CEOjt) with prior experience as a CEO or managing director
is 0.45 (0.00). This represents an average proportion of audit committee
members having prior experience as a CEO or managing director (%
Expt_CEOjt) of 14.10 per cent. A total of 379 (or 76.72 per cent) rm-year
observations have at least one director on the audit committee either with
accounting or nonaccounting nancial expertise (see Exptjt). In total, 274 (or
55.47 per cent) rm-year observations in the nal useable sample have at least
one director on the audit committee with accounting nancial expertise (see
Expt_CEOjt). There are 179 (or 36.24 per cent) rm-year observations in the
nal useable sample that have at least one director on the audit committee with
2013 AFAANZ
2013 AFAANZ
Yes
280
214
379
263
274
Pooled
sample
494
494
494
494
494
Variable
DRjt
DCFOjt
Exptjt
Ind_Exptjt
Expt_Accjt
56.680
43.320
76.721
53.239
55.466
0.021
0.095
0.178
0.134
1.000
33.333
0.000
0.000
3.000
2.000
66.667
38,682
2.140
1.850
0.081
0.074
0.492
0.328
0.970
31.390
0.450
14.095
3.163
1.730
54.626
562,016
3.510
2.239
494
494
494
494
494
494
494
494
494
494
494
494
494
494
Xjt
RRAjt
ACCjt
CFOjt
#Expt_Accjt
%Expt_Accjt
#Expt_CEOjt
%Expt_CEOjt
ACSIZEjt
#ACINDjt
%ACINDjt
BVTAjt ($000)
MTBjt
LEVjt
Median
Mean
Pooled
sample
Variables
Table 2
Descriptive statistics
0.356
1.134
0.974
0.702
0.890
28.847
0.695
21.344
0.946
1.297
37.778
1,561,076
6.970
2.814
Standard
deviation
214
280
115
231
220
No
2.707
1.225
8.991
0.188
0.000
0.000
0.000
0.000
1.000
0.000
0.000
61
19.130
10.730
Minimum
43.320
56.680
23.279
46.761
44.534
1.578
19.490
2.232
9.760
3.000
100.000
3.000
100.000
6.000
6.000
100.000
14,929,000
110.170
36.490
Maximum
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 291
2013 AFAANZ
179
133
92
87
57
243
494
494
494
494
494
494
Expt_CEOjt
ACCONLYjt
CEOONLYjt
BothExptjt
DUALITYjt
Industryjt
36.235
26.923
18.623
17.611
11.538
49.190
70.526
75.258
77.000
78.000
84.000
2004
2005
2006
2007
2008
56.842
59.739
66.000
68.000
70.000
Expt_Accjt
34.737
36.082
37.000
35.000
38.000
Expt_CEOjt
20.000
26.804
26.000
29.000
33.000
ACCONLYjt
16.666
20.408
19.000
16.000
21.000
CEOONLYjt
315
361
402
407
437
251
No
16.842
14.432
18.000
20.000
19.000
BothExptjt
63.765
73.076
81.377
82.388
88.462
50.810
All continuous variables (except log transformed variables) are winsorised at one per cent and 99 per cent. Refer to Appendix 1 for the denition
of variables.
Exptjt
Year
Yes
Pooled
sample
Variable
Table 2 (continued)
292 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 293
2013 AFAANZ
0.590**
0.095***
0.256**
0.031
0.096
0.335**
0.077
0.397**
0.141**
0.068
0.033
Xjt [1]
ACCjt [2]
Exptjt [3]
Expt_Accjt [4]
Expt_CEOjt [5]
ACSIZEjt [6]
ACINDjt [7]
Dualityjt [8]
LnBVTAjt [9]
MTBjt [10]
LEVjt [11]
Industryjt [12]
0.148**
0.244**
0.085
0.039
0.306**
0.080
0.406
0.122**
0.049
0.051
0.641**
[2]
0.494**
0.415**
0.121**
0.261**
0.095***
0.279**
0.015
0.090***
0.062
0.135**
0.156**
[3]
0.062
0.159**
0.462**
0.158**
0.509**
0.086
0.005
0.034
0.317**
0.341**
0.494**
[4]
0.168**
0.113***
0.132**
0.233**
0.002
0.058
0.068
0.071
0.140**
0.415***
0.062
[5]
0.225**
0.531**
0.027
0.030
0.026
0.336**
0.372**
0.261**
0.462**
0.113***
0.045
0.034
0.078
0.133**
0.156**
0.180**
0.058
0.119**
0.154**
0.018
0.030
0.035
[7]
[6]
0.355**
0.076
0.035
0.124**
0.125**
0.219**
0.095***
0.158**
0.132**
0.136**
0.225**
[8]
0.110***
0.008
0.001
0.552**
0.676**
0.284**
0.516**
0.226**
0.147**
0.536**
0.330**
[9]
0.042
0.097***
0.035
0.140**
0.049
0.012
0.110***
0.039
0.131**
0.136**
0.004
[10]
0.052
0.202**
0.266**
0.154**
0.132**
0.143**
0.073
0.118**
0.027
0.313
0.035
[11]
0.202
0.086
0.062
0.034
0.068
0.069
0.026
0.124
0.007
0.054
0.004
[12]
Tests are two-tailed. ** and *** denote signicance at the 5 per cent and 1 per cent condence levels, respectively. Refer to Appendix 1 for the
denition of variables.
[1]
Variables
Table 3
Pearsons and Spearmans correlation analyses
294 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
2013 AFAANZ
0.056
0.124
0.111
0.256
Intercept
DRjt
RRAjt
RRAjt*DRjt
Expt_ACEjt
DRjt*Exptjt
RRAjt*Exptjt
RRAjt*DRjt*Exptjt
Expt_Accjt
DRjt*Expt_Accjt
RRAjt*Expt_Accjt
RRAjt*DRjt*Expt_Accjt
Expt_CEOjt
DRjt*Expt_CEOjt
RRAjt*Expt_CEOjt
RRAjt*DRjt*Expt_CEOjt
Control Variables
Yeark
Adjusted-R2
N
F-Statistic
(2.661)***
(0.846)
(0.384)
(1.966)**
(1.231)
(2.211)**
(4.110)***
(3.602)***
0.501
0.238
0.096
0.660
0.047
0.072
0.060
0.362
0.485
0.239
0.103
0.579
0.027
0.102
0.005
0.394
Included
Included
0.637
494
25.735***
Included
Included
0.634
494
25.387***
(Expt_Accjt)
(Exptjt)
(2.681)***
(0.877)
(0.421)
(1.746)*
(0.546)
(1.368)
(0.067)
(2.366)**
Column III
Column II
(2.343)**
(1.078)
(0.113)
(1.928)*
0.011 (0.236)
0.083 (1.223)
0.047 (0.780)
0.147 (1.231)
Included
Included
0.637
494
25.699***
0.448
0.304
0.030
0.361
(CEOjt)
Column IV
(2.371)**
(1.040)
(0.028)
(1.789)*
0.050 (1.176)
0.081 (2.466)**
0.062 (4.199)***
0.368 (3.549)***
0.010 (0.225)
0.086 (1.242)
0.050 (0.827)
0.461 (1.215)
Included
Included
0.635
494
22.959***
0.472
0.306
0.007
0.464
(Both Expt_Accjt
and CEOjt)
Column V
*, ** and *** denote signicance at the 10 per cent, 5 per cent and 1 per cent condence levels, respectively. Coecient values are shown outside
the parentheses with t-statistics (robust and clustered) values shown inside the parentheses. Results presented in Columns IV are based on
Equations (1) and (2). Refer to Appendix 1 for the denition of variables.
Included
Included
0.115
494
22.427***
(2.214)**
(2.933)***
(7.128)
(3.678)***
(Base model)
Variables
Column I
Table 4
Regression analysis using the timeliness of earnings to news measure (Basu, 1997)
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 295
296 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
earnings. All reported results are based on HuberWhite robust standard errors
clustered at both year and rm level. Results in Table 4 Column I based on the
original Basu (1997) asymmetric timeliness of earnings model (without
inclusion of independent and control variables) indicate the existence of
conservatism in general as the b3 coecient on the two-way interaction term
RRAjt*DRjt is signicant and positive.
With respect to the variable of interest, audit committee overall nancial
expertise, Table 4 Column II results indicate that the coecient (i.e. k4) on
RRAjt*DRjt*Exptjt is signicant at the 5 per cent level. Table 4 Column III
results indicate that RRAjt*DRjt*Expt_Accjt is signicant when examining the
relationship between only accounting nancial expertise (Expt_Accjt) and
conservatism. However, when accounting nancial expertise (Expt_Accjt) is
replaced with nonaccounting nancial expertise (Expt_CEOjt), the coecient
on RRAjt*DRjt*Expt_CEOjt is not signicant (see Table 4 Column IV). In
contrast, the coecient on RRAjt*DRjt*Expt_Accjt (see Table 4 Column V) is
positive and statistically signicant at the 1 per cent level when both Expt_Accjt
and Expt_CEOjt are included in the model. Overall, Table 4 results suggest that
accounting nancial expertise is the primary type of expertise that promotes the
asymmetrical timeliness of earnings (i.e. swifter recognition of economic losses
relative to gains).
With respect to control variables, audit committee size and independence are
negatively associated with the asymmetrical timeliness of conservatism across
Table 4 Columns IIV. Duality, rm size and leverage are not signicant in any
of the regression results. In contrast, a rms growth options are found to be
positively and signicantly associated with conservatism across all regressions
reported in Table 4. Finally, Table 4 Columns IIV results show that the
adjusted R2 ranges from 11.5 per cent for the base model to over 63 per cent for
the augmented models.
4.2.2. Accrual-based loss recognition
Table 5 presents regression results based on the accrual-based asymmetric
timeliness of accruals measure developed by Ball and Shivakumar (2005) and
modied to measure the incremental conservatism associated with audit
committee nancial expertise. Results shown in Table 5 Column I utilise the
base model derived by Ball and Shivakumar (2005) as dened in Equation (3).
The directionality (i.e. positive) of the coecient on CFOjt*DCFOjt (i.e. b3) is
as predicted. The coecient is statistically signicant indicating that when
accruals are lower, cash ows are higher, suggesting the presence of earnings
conservatism. The b3 coecient is positive and statistically signicant across
the remaining regressions shown in Table 5 Columns II V after the inclusion
of independent and control variables.
With regard to the audit committee nancial expertise measure, the d4
coecient on the three-way interaction term CFOjt*DCFOjt*Exptjt is
2013 AFAANZ
2013 AFAANZ
(1.506)
(0.468)
(1.613)
(3.080)***
0.183
0.078
0.252
0.987
Included
Included
0.958
494
324.796***
(0.735)
(1.170)
(11.190)***
(2.515)**
0.091
0.132
2.255
1.470
(0.828)
(1.341)
(11.868)***
(2.739)***
0.020 (0.624)
0.022 (0.545)
0.135 (3.213)***
0.136 (0.994)
Included
Included
0.957
494
314.970***
0.120
0.147
2.297
1.587
(CEOjt)
Column IV
(0.814)
(1.234)
(11.557)***
(2.566)**
0.017 (1.369)
0.051 (0.359)
0.242 (1.405)
0.937 (2.868)***
0.019 (0.482)
0.019 (0.469)
0.132 (2.893)***
0.122 (0.884)
Included
Included
0.958
494
289.782***
0.100
0.138
2.303
1.506
(Both Expt_Accjt
and CEOjt)
Column V
*, ** and *** denote signicance at the 10 per cent, 5 per cent and 1 per cent condence levels, respectively. Coecient values are shown outside
the parentheses with t-statistics (robust and clustered) values shown inside the parentheses. Results presented in Columns IV are based on
Equations (34). Refer to Appendix 1 for the denition of variables.
Included
Included
0.727
494
436.969***
(0.416)
(1.325)
(11.412)***
(2.633)***
(0.228)
(1.146)
(0.264)
(1.844)*
Included
Included
0.957
494
312.648***
0.052
0.147
2.275
1.554
0.008
0.104
0.035
0.758
(0.262)
(1.731)*
(32.364)***
(12.784)***
0.009
0.084
1.097
1.588
Intercept
DCFOjt
CFOjt
CFOjt*DCFOjt
Expt_ACEjt
DCFOjt*Exptjt
CFOjt*Exptjt
CFOjt*DCFOjt*Exptjt
Expt_Accjt
DCFOjt* Expt_Accjt
CFOjt*Expt_Accjt
CFOjt*DCFOjt*Expt_Accjt
Expt_CEOjt
DCFOjt*Expt_CEOjt
CFOjt*Expt_CEOjt
CFOjt*DCFOjt*Expt_CEOjt
Control Variables
Yeark
Adjusted-R2
N
F-Statistic
(Expt_Accjt)
(Exptjt)
(Base model)
Column III
Column II
Variables
Column I
Table 5
Regression analysis using the accrual-based loss recognition measure (Ball and Shivakumar, 2005)
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 297
298 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 299
k1 k
10
According to ASX CGC 2003 Recommendation 2, an independent director is a nonexecutive director who is not a member of management and who is free of any business
or other relationship that could materially interfere with or could reasonably be
perceived to materially interfere with the independent exercise of their judgement (ASX
Corporate Governance Council, 2003).
2013 AFAANZ
300 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
Ind_Exptjt and Non_Ind_Exptjt in Equations (5) and (6) are replaced with
Ind_ACCjt/Non_Ind_ACCjt and Ind_CEOjt/Non_Ind_CEOjt when measuring
the inuence of independent/nonindependent accounting nancial expertise and independent/nonindependent nonaccounting nancial expertise on
conservatism. Finally, to determine incremental eects, Ind_Exptjt and
Non_Ind_Exptjt in Equations (5) and (6) are replaced with both Ind_ACCjt/
Non_Ind_ACCjt and Ind_CEOjt/Non_Ind_CEOjt concurrently. See Appendix 1
for the denition of all variables.
Tables 6 (timeliness of earnings to news model) and 7 (accruals-based loss
recognition model) present results considering the impact of independent and
nonindependent audit committee nancial experts on conservatism. The results
from both Table 6 Column I and Table 7 Column I reveal that independent
overall nancial expertise is positively associated with conservatism. Results
from Columns II and IV of both Tables 6 and 7 suggest that the association
between conservatism and accounting expertise is most pronounced when the
audit committee accounting expert/s are also independent. However, nonaccounting nancial experts are consistently not associated with conservatism
regardless of their level of independence. Results suggest that accounting
nancial experts who are independent are most likely to promote conservative
accounting practices.
4.3.2. Alternative measures of earnings conservatism
Several robustness and sensitivity tests are undertaken to check the resilience
of the main results reported in Tables 4 and 5. Past studies (Givoly and Hayn,
2000; Khan and Watts, 2009) suggest that a limitation of using the Basu (1997)
and Ball and Shivakumar (2005) models is an over reliance on the calculation
of average measures of conservatism for both time and across rms reducing
the generalisability of ndings.
To further test the strength of the association between audit committee
eectiveness and conservatism, the C-SCORE approach developed by Khan
and Watts (2009) is used, where a higher C-SCORE indicates greater
conservatism. Consistent with the main ndings presented in Table 4, results
(untabulated) using the Khan and Watts (2009) C-SCORE approach
continue to show a signicant association between conservatism and the
presence of audit committee members with overall nancial expertise
(t = 2.011) and accounting nancial expertise (t = 3.362). In addition, to
further corroborate the main ndings, a second accrual-based approach
(CON-ACC) developed by Givoly and Hayn (2000) is used. This approach
is measured as income before extraordinary items less cash ows from
operations plus depreciation divided by average total assets, averaged over a
ve-year period before year t. Using the Givoly and Hayn (2000) approach,
a lower CON-ACC indicates the existence of conservatism. When the CONACC measure is used, results (untabulated) show that presence of audit
2013 AFAANZ
2013 AFAANZ
(0.681)
(1.515)
(0.258)
(3.053)***
(0.316)
(0.035)
(0.144)
(0.404)
0.063
0.026
0.028
0.618
0.018
0.003
0.023
0.106
(2.171)**
(0.521)
(0.005)
(0.205)
0.025 (0.405)
0.023 (0.242)
0.003 (0.033)
0.003 (0.018)
0.034 (0.562)
0.001 (0.006)
0.079 (1.041)
0.097 (0.690)
Included
Included
0.642
494
21.524***
0.440
0.170
0.001
0.135
Column III
(2.243)**
(0.429)
(0.511)
(0.084)
0.098 (0.899)
0.156 (1.023)
0.086 (0.578)
0.766 (2.999)***
0.049 (0.579)
0.039 (0.306)
0.132 (0.995)
0.040 (0.132)
0.007 (0.083)
0.041 (0.357)
0.075 (0.697)
0.079 (0.963)
0.040 (0.650)
0.001 (0.016)
0.099 (1.274)
0.111 (0.703)
Included
Included
0.641
494
23.668***
0.048
0.148
0.151
0.059
Column IV
*, ** and *** denote signicance at the 10 per cent, 5 per cent and 1 per cent condence levels, respectively. Coecient values are shown outside
the parentheses with t-statistics (robust and clustered) values shown inside the parentheses. Results presented in Columns IV are based on
Equation (5). Refer to Appendix 1 for the denition of variables.
Included
Included
0.620
494
28.503***
Included
Included
0.632
494
26.821***
(2.385)**
(0.267)
(0.217)
(0.276)
0.475
0.080
0.062
0.170
0.222
0.032
0.042
0.155
0.056
0.011
0.042
0.656
0.066
0.042
0.028
0.732
Intercept
DRjt
RRAjt
RRAjt*DRjt
Ind_Exptjt
DRjt*Ind_Exptjt
RRAjt*Ind_Exptjt
RRAjt*DRjt*Ind_Exptjt
Non_Ind_Exptjt
DRjt*Non_Ind_Exptjt
RRAjt*Non_Ind_Exptjt
RRAjt*DRjt*Non_Ind_Exptjt
Ind_ACCjt
DRjt*Ind_ACCjt
RRAjt*Ind_ACCjt
RRAjt*DRjt*Ind_ACCjt
Non_Ind_ACCjt
DRjt*Non_Ind_ACCjt
RRAjt*Non_Ind_ACCjt
RRAjt*DRjt*Non_Ind_ACCjt
Ind_CEOjt
DRjt*Ind_CEOjt
RRAjt*Ind_CEOjt
RRAjt*DRjt*Ind_CEOjt
Non_Ind_CEOjt
DRjt*Non_Ind_CEOjt
RRAjt*Non_Ind_CEOjt
RRAjt*DRjt*Non_Ind_CEOjt
Control Variables
Yeark
Adjusted-R2
N
F-Statistic
(1.706)*
(0.215)
(0.186)
(0.392)
(1.982)**
(0.347)
(2.471)**
(2.829)***
(2.334)**
(1.434)
(1.794)*
(1.104)
Column II
Column I
Variables
Table 6
Regression analysis using the timeliness of earnings to news (Basu, 1997) and the independence of nancial experts
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 301
0.142
0.029
1.047
1.679
0.148
0.295
0.066
1.389
0.146
0.238
0.119
0.376
Intercept
DCFOjt
CFOjt
CFOjt*DCFOjt
Ind_Exptjt
DCFOjt*Ind_Exptjt
CFOjt*Ind_Exptjt
CFOjt*DCFOjt*Ind_Exptjt
Non_Ind_Exptjt
DCFOjt*Non_Ind_Exptjt
CFOjt*Non_Ind_Exptjt
CFOjt*DCFOjt*Non_Ind_Exptjt
Ind_ACCjt
DCFOjt*Ind_ACCjt
CFOjt*Ind_ACCjt
CFOjt*DCFOjt*Ind_ACCjt
Non_Ind_ACCjt
DCFOjt*Non_Ind_ACCjt
CFOjt*Non_Ind_ACCjt
CFOjt*DCFOjt*Non_Ind_ACCjt
Ind_CEOjt
DCFOjt*Ind_CEOjt
CFOjt*Ind_CEOjt
CFOjt*DCFOjt*Ind_CEOjt
Non_Ind_CEOjt
DCFOjt*Non_Ind_CEOjt
CFOjt*Non_Ind_CEOjt
CFOjt*DCFOjt*Non_Ind_CEOjt
Control Variables
Yeark
Adjusted-R2
N
F-Statistic
2013 AFAANZ
(1.891)*
(0.342)
(0.054)
(5.743)***
(1.264)
(0.091)
(0.905)
(1.402)
0.159
0.036
0.005
1.088
0.115
0.010
0.114
0.657
Included
Included
0.967
494
378.126***
(0.745)
(0.539)
(7.511)
(6.562)***
0.151
0.069
0.912
1.629
Column II
(0.023)
(0.428)
(23.745)***
(7.393)***
0.071 (0.803)
0.034 (0.255)
0.155 (1.388)
0.447 (0.782)
0.006 (0.052)
0.003 (0.016)
0.024 (0.238)
0.066 (0.121)
Included
Included
0.902
494
376.032***
0.009
0.038
1.043
1.811
Column III
(0.697)
(0.598)
(5.739)***
(6.056)***
0.122 (1.262)
0.020 (0.170)
0.061 (0.495)
1.060 (5.600)***
0.059 (0.511)
0.026 (0.177)
0.003 (0.015)
0.567 (1.394)
0.115 (1.052)
0.040 (0.305)
0.155 (1.031)
0.104 (0.294)
0.015 (0.130)
0.089 (0.665)
0.052 (0.521)
0.173 (0.687)
Included
Included
0.966
494
376.954***
0.145
0.090
1.023
1.626
Column IV
*, ** and *** denote signicance at the 10 per cent, 5 per cent and 1 per cent condence levels, respectively. Coecient values are shown outside
the parentheses with t-statistics (robust and clustered) values shown inside the parentheses. Results presented in Columns IV are based on
Equation (6). Refer to Appendix 1 for the denition of variables.
Included
Included
0.944
494
372.918***
(0.247)
(0.246)
(24.584)
(6.709)***
(1.343)
(0.769)
(0.739)
(3.780)***
(1.349)
(0.640)
(1.245)
(0.929)
Column I
Variables
Table 7
Regression analysis using the accrual-based loss recognition measure (Ball and Shivakumar, 2005) and the independence of nancial experts
302 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 303
11
The main tests are also reperformed using alternative proxies for key variables
underpinning the Basu (1997). Researchers (Ahmad-Zaluki et al., 2007) argue that
listed rms that constitute the whole market represent a variety of characteristics that
cannot be easily captured by a single index. Following Ahmad-Zaluki et al. (2007),
market-adjusted returns are recalculated using three alternative indices in an eort to
better control for size and industry biases. First, market-adjusted returns are
calculated using the ASX 200 Index. Second, to better control for size, marketadjusted returns for rms included in the (i) 1st quartile are based on the All
Ordinaries Index, (ii) 2nd quartile are based on the ASX 200 Index, (iii) 3rd quartile
are based on the Mid-Cap 50 Index and (iv) 4th quartile are based on the Small
Ordinaries Index. Finally, in relation to industry dierences, market-adjusted returns
are calculated using an industry index relevant to the major industry sector of each
rm. In addition, regression results are also reperformed using raw returns. Tests
performed using alternative measures provide results highly similar to Table 4
ndings.
2013 AFAANZ
304 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
committee. These results are consistent with the main ndings tabulated in
Tables 4 and 5.12
4.3.4. Controlling for endogeneity bias
A concern expressed in the corporate governance literature during the past
decade involves the issue of endogeneity between governance factors and other
contracting mechanisms (Larcker and Richardson, 2004; Brown et al., 2011).13
Based on the corporate governance literature (Beekes et al., 2004; Lim, 2011),
ndings from this study may be subject to omitted variable biases resulting
from the possible omission of other governance mechanisms that can inuence
nancial reporting quality. Although the main ndings of this study are robust
to various control variables, additional tests were performed to identify any
endogenous determination of audit committee nancial expertise that impacts
its relationship with conservative accounting practices.
Consistent with Frankel et al. (2006) and Sun and Cahan (2012), a two-stage
least squares (2SLS) approach is adopted to control for potential endogeneity
problems. The 2SLS requires the identication and use of an instrument
(exogenous variable) in the rst stage which has an impact on the conservative
accounting practices of rms only through audit committee nancial expertise
without any direct inuence on levels of conservatism (Wooldridge, 2010).14
12
Tests were reperformed using continuous proxy measures (i.e. proportion of overall/
accounting/nonaccounting nancial expertise to the total number of audit committee
members) for nancial expertise of audit committee members. Also, a more restrictive
criterion is applied when dening a nancial expert. Specically, each rm is given a
score of 1 only if at least two audit committee members are nancial experts (including
overall, accounting and nonaccounting expertise). Results of additional tests are
qualitatively the same when alternative measures for the audit committee nancial
expertise are used. In addition, further tests are conducted using other control variables
(e.g. board size, board independence, board expertise, board tenure, audit committee
diligence). Again, tests using the additional measures yield highly consistent ndings to
those reported in Tables 4 and 5.
13
Endogeneity can arise either from simultaneous causality where causality runs in both
directions between the dependent and predictor variables or from unobserved
heterogeneity where important predictor variables are omitted from the regression
model (Wooldridge, 2010; Brown et al., 2011). It is unlikely that this study suers from
endogeneity problems arising from simultaneous causality as it is improbable that
conservatism accounting practices will inuence the eectiveness of audit committee
nancial expertise.
14
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 305
Consistent with Matolcsy et al. (2012), this study uses nomination committee
(i.e. existence, size and independence of nomination committee) characteristics
as instrumental variables. This is because it is more likely that a nomination
committee is associated with audit committee characteristics (including
nancial expertise) without having a direct eect on a rms accounting
practices as the nomination committee is primarily responsible for appointing
board and board subcommittee members (including the audit committee) with
the requisite skills and competencies. The 2SLS regressions are run for all three
models based on the three dierent measures of audit committee nancial
expertise (Exptjt, Expt_Accjt and Expt_CEOjt). In the rst stage, it is estimated
whether each of the three types of audit committee expertise is associated with
nomination committee characteristics. Then, the predicted values (denoted
PExptjt, PExpt_Accjt and PExpt_CEOjt) are obtained for each type of nancial
expertise measure. Subsequently, the predicted values of nancial expertise
from the rst stage are then regressed against conservatism in the second stage
of the 2SLS.
The 2SLS tests yield results signicantly comparable with main results
reported in Tables 4 and 5. Specically, untabulated statistics for Basu (1997)
timeliness of earnings to news measure are t = 1.986 for PExptjt, 2.232 for
PExpt_Accjt and 1.304 for PExpt_CEOjt, respectively. With respect to the Ball
and Shivakumar (2005) accrual-based loss recognition model, the untabulated
statistics are t = 1.843 for PExptjt, 2.610 for PExpt_Accjt and 1.117 for
PExpt_CEOjt, respectively. Thus, results of the second-stage regressions are
largely unchanged after allowing for endogeneity.
Overall, robustness and sensitivity tests provide further support to the main
ndings reported in Tables 4 and 5 having yielded similar results.15
5. Conclusion
In response to major corporate scandals (e.g. HIH, Enron, WorldCom,
One.Tel), the Australian Securities Exchange Corporate Governance Council
introduced a best practice guide known as the Principles of Good Corporate
Governance and Best Practice Recommendations placing greater responsibil15
The time period (i.e. 20042008) of this study transcends the International Financial
Reporting System (IFRS) adoption by Australian rms in 2005. Thus, to control for the
eect of IFRS adoption on the association between audit committee nancial expertise
and earnings conservatism, another robustness test was performed where it is
investigated whether rms in the pre-IFRS period were more conservative than in the
post-IFRS period. The data are partitioned between pre-IFRS (20042005) and postIFRS (20062008) periods and regressions are rerun. Regressions are also rerun using
IFRS as an indicator variable that equals 1 for scal years 20062008 and 0 for scal
years 20042005. Results remain signicantly similar to the main results reported
suggesting that IFRS adoption did not change the demand for conservative accounting
practices by Australian rms.
2013 AFAANZ
306 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 307
308 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
Givoly, D., C. K. Hayn, and A. Natarajan, 2007, Measuring reporting conservatism,
The Accounting Review 82, 65106.
Goh, B. W., 2009, Audit committees, board of directors, and remediation of material
weaknesses in internal control, Contemporary Accounting Research 26, 549579.
Goh, B. W., and D. Li, 2011, Internal controls and conditional conservatism, The
Accounting Review 86, 9751005.
Goodwin-Stewart, J., and P. Kent, 2006, Relation between external audit fees,
audit committee characteristics and internal audit, Accounting and Finance 46,
387404.
Hair, J. F., R. E. Anderson, R. L. Tatham, and W. C. Black, 1995, Multivariate Data
Analysis, 4th edn (Prentice Hall, Englewood Clis, NJ).
Hartmann, F. G. H., and F. Moers, 1999, Testing contingency hypotheses in budgetary
research: an evaluation of the use of moderated regression analysis, Accounting,
Organisations and Society 24, 291315.
Hentschel, L., and S. P. Kothari, 2001, Are corporations reducing or taking risks with
derivatives?, Journal of Finance and Quantitative Analysis 36, 93118.
Karamanou, I., and N. Vafeas, 2005, The association between boards and audit
committees with management earnings forecasts: an empirical analysis, Journal of
Accounting Research 43, 453486.
Khan, M., and R. L. Watts, 2009, Estimation and empirical properties of a rm-year
measure of accounting conservatism, Journal of Accounting and Economics 48, 132
150.
Klein, A., 2002, Economic determinants of audit committee independence, The
Accounting Review 77, 435452.
Krishnan, G. V., and G. Visvanathan, 2008, Does the SOX denition of an accounting
expert matter? The association between audit committee directors accounting
expertise and accounting conservatism, Contemporary Accounting Research 25, 827
858.
Lafond, R., and S. Roychowdhury, 2008, Managerial ownership and accounting
conservatism, Journal of Accounting Research 46, 101135.
Lafond, R., and R. L. Watts, 2008, The information role of conservatism, The
Accounting Review 83, 447478.
Lara, J. M. G., B. G. Osma, and F. Penalva, 2009, Accounting conservatism and
corporate governance, Review of Accounting Studies 14, 161201.
Larcker, D. F., and S. Richardson, 2004, Fees paid to audit rms, accrual choices and
corporate governance, Journal of Accounting Research 42, 625658.
Lim, R., 2011, Are corporate governance attributes associated with accounting
conservatism?, Accounting and Finance 51, 10071030.
Matolcsy, Z. P., J. V. Tyler, and P. A. Wells, 2012, Is continuous disclosure associated
with board independence?, Australian Journal of Management 37, 99123.
McDaniel, L., R. Marint, L. A. Maines, and M. E. Peecher, 2002, Evaluating nancial
report quality: the eects of nancial expertise vs. nancial literacy, The Accounting
Review 77, 139167.
Securities and Exchange Commission, 2003, Final Rule: Audit Committee Disclosure
(Government Printing Oce, Washington, DC).
Srinivasan, S., 2005, Consequences of nancial reporting failure for outside directors:
evidence from accounting restatements and audit committee members, Journal of
Accounting Research 43, 291334.
Sun, J., and S. F. Cahan, 2012, The economic determinants of compensation committee
quality, Managerial Finance 38, 188205.
Watts, R., 2003, Conservatism in accounting - Part I: explanations and implications,
Accounting Horizons 17, 207221.
2013 AFAANZ
N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310 309
Wooldridge, J. M., 2010, Econometric Analysis of Cross Section and Panel Data (MIT
Press, Boston, MA).
Xie, B., W. N. Davidson, and P. J. DaDalt, 2003, Earnings management and corporate
governance: the role of the board and audit committee, Journal of Corporate Finance
9, 295316.
Denition
Xjt
DRjt
RRAjt
ACCjt
DCFOjt
CFOjt
Exptjt
Expt_Accjt
Expt_CEOjt
Ind_Exptjt
Non_Ind_Exptjt
Ind_ACCjt
Non_Ind _ACCjt
2013 AFAANZ
310 N. Sultana, J-L. W. Mitchell Van der Zahn/Accounting and Finance 55 (2015) 279310
Appendix 1 (continued)
Variables
Denition
Ind_CEOjt
Non_Ind _CEOjt
#Expt_Accjt
%Expt_Accjt
#Expt_CEOjt
%Expt_CEOjt
ACCONLYjt
CEOONLYjt
BothExptjt
ACSIZEjt
ACINDjt
DUALITYjt
BVTAjt
LnBVTAjt
MTBjt
LEVjt
Industryjt
Yeark
2013 AFAANZ