Description
We examine whether the relationship between political connections and firm value is
moderated by the length of time firms have been politically connected. We find that compared
to firms with political connections for a short period, firms with political connections
for a long period have a smaller magnitude of negative stock price reaction to the 2008
General Election loss of the supermajority by the ruling party in Malaysia. We also find that
the smaller magnitude of negative stock price reaction is, in part, attributable to improvements
in board of director characteristics.
Corporate political connections and the 2008 Malaysian election
Simon Y.K. Fung
a,?
, Ferdinand A. Gul
b,1
, Suresh Radhakrishnan
c,2
a
School of Accounting and Finance, The Hong Kong Polytechnic University, Hung Hom, Hong Kong
b
School of Business, Monash University Malaysia, Malaysia
c
School of Management, The University of Texas at Dallas, Richardson, TX 75083-0688, USA
a b s t r a c t
We examine whether the relationship between political connections and ?rm value is
moderated by the length of time ?rms have been politically connected. We ?nd that com-
pared to ?rms with political connections for a short period, ?rms with political connections
for a long period have a smaller magnitude of negative stock price reaction to the 2008
General Election loss of the supermajority by the ruling party in Malaysia. We also ?nd that
the smaller magnitude of negative stock price reaction is, in part, attributable to improve-
ments in board of director characteristics. Furthermore, we ?nd that while the perfor-
mance subsequent to the General Election of politically connected ?rms is worse than
that of non-politically connected ?rms, ?rms with political connections for a long period
exhibit better performance than those connected for short periods. Collectively, the evi-
dence shows that the length of political connections is an important factor that moderates
economic value.
Ó 2015 Published by Elsevier Ltd.
Introduction
Prior research shows that ?rms’ political connections
are positively associated with ?rm value (Bliss & Gul,
2012; Boubakri, Guedhami, Mishra, & Saffar, 2012;
Faccio, 2006; Fisman, 2001; Johnson & Mitton, 2003). Li
and Zhang (2007) show how managers’ functional
experience and their political network help to enhance
performance of new ventures (also see Lester, Hillman,
Zardkoohi, & Cannella, 2008). A natural question that arises
is whether political connections continue to be as
important for ?rms during later periods as they are in
initial periods. Accordingly, we examine the following
two research questions: (a) Is the relationship between
?rms’ political connection and ?rm value different for
?rms with political connections for longer and shorter
periods? (b) What is the mechanism that drives the differ-
ential relationship?
We draw on institutional theory in sociology to develop
the hypothesis that the relationship between political con-
nections and ?rm value is moderated by the length of time
over which ?rms are politically connected (see Rona-Tas,
1994). Speci?cally, we posit that compared to ?rms that
are politically connected for a short period of time, ?rms
that are politically connected for a long period of time
are more likely to have built their credibility of adherence
to government policies which allows them access to the
informal political elite network (DiMaggio & Powell,
1983; Donaldson, 2008; Sanders & Tuschke, 2007). This
in turn would enable them to obtain important economic
resources such as talented managers and become more
self-sustaining. Thus, we expect ?rms with political con-
nections for a long period to be less severely affected by
the loss of political power by the ruling party (in termshttp://dx.doi.org/10.1016/j.aos.2015.04.001
0361-3682/Ó 2015 Published by Elsevier Ltd.
?
Corresponding author. Tel.: +852 2766 4246.
E-mail addresses: [email protected] (S.Y.K. Fung), [email protected]
(F.A. Gul), [email protected] (S. Radhakrishnan).
1
Tel.: +60 3 5514 4997.
2
Tel.: +1 972 883 4438.
Accounting, Organizations and Society 43 (2015) 67–86
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
of their supermajority) than ?rms with political connec-
tions for a short period.
We examine this hypothesis in the 2008 Malaysian
General Election setting where the ruling party – the
National Front or Barisan Nasional (BN) – unexpectedly
lost their longstanding supermajority. To identify politi-
cally connected ?rms, we update the list provided by
Johnson and Mitton (2003) and Faccio, Masulis, and
McConnell (2006). Firms that are identi?ed as politically
connected in both 1997/1998 and 2007 are classi?ed as
?rms with political connections for a long period (referred
to as old-politically connected ?rms); and ?rms that are
politically connected only in 2007 are classi?ed as ?rms
with political connections for a short period (referred to
as new-politically connected ?rms). In particular, we iden-
tify 122 politically connected ?rms out of which 39 are
new-politically connected ?rms.
Consistent with our hypothesis, we ?nd that the magni-
tude of negative stock market reaction to the ruling party’s
election loss of supermajority is less for the old-politically
connected ?rms than the new-politically connected ?rms.
To examine the mechanism through which the old-
politically connected ?rms insulate themselves from the
adverse consequences of the loss in supermajority power
by the ruling party, we investigate the change in board of
director characteristics and senior executive characteris-
tics over ten years across old- and new-politically con-
nected ?rms. We ?nd that the old-politically connected
?rms become more professional than the new-politically
connected ?rms. Interestingly, we ?nd that changes in
board professionalism exhibits improvement for old-
politically connected ?rms over the ten years, but execu-
tive professionalism does not. Furthermore, we ?nd that
the negative stock price reaction to the election result is
attenuated for those old-politically connected ?rms who
show improvements in professionalism, and not for those
who do not show such improvements. While, this evidence
provides insight into the mechanism through which polit-
ical connections enhance value, the access to managerial
talent that old-politically connected ?rms develop can by
itself help them to continue to extract rents even when
the direct bene?ts of political connection are removed. As
such, this evidence could be interpreted as a form of
cronyism as well.
Lastly, we attempt to address whether the improve-
ments in board professionalism could be related to crony-
ism, by examining the subsequent accounting and stock
market performance. In particular, we consider return on
assets and annual stock return for the three years subse-
quent to 2008 for the old- and new-politically connected
?rms. We ?nd that compared to the non-politically con-
nected ?rms, the accounting and stock market return per-
formance are worse for both the old- and new-politically
connected ?rms; but less so for the old-politically con-
nected ?rms than the new-politically connected ?rms.
This provides additional evidence that while old-politically
connected ?rms are more self-sustaining than the new-
politically connected ?rms, both of them exhibit worse
performance than non-politically connected ?rms.
The remainder of the paper is organized as follows: First
we provide the background literature and the research
question; second, we describe the Malaysian context and
the General Election in 2008, and develop the hypotheses;
third, we present the empirical analysis; and ?nally, we
provide some concluding remarks.
Background and hypotheses
Background and research question
The institutional view is based on the premise that
organizations adopt structures in response to their institu-
tional environments and hence is an important theoretical
lens in international business (see Carpenter & Feroz, 2001;
DiMaggio & Powell, 1983; North, 1990; Peng, Sun,
Pinkham, & Chen, 2009; Peng, Wang, & Jiang, 2008; Wan
& Hoskisson, 2003). Speci?cally, ?rms’ operations and
management practices are bound by formal and informal
country-speci?c institutions (see Ingram & Silverman,
2002). The formal and informal institutions in?uence mar-
ket competition (Khanna & Palepu, 1997; Peng, 2003;
Porter, 1990), and ?rms’ resource allocation decisions
either directly (North, 1990; Wan & Hoskisson, 2003), or
indirectly through management and corporate governance
practices (Aguilera & Jackson, 2003; La Porta, Lopez-
de-Silanes, Shleifer, & Vishny, 2000).
3
Managers’ ties with government of?cials, i.e., political
connections represent a unique economic resource in tran-
sition economies (Li & Atuahene-Gima, 2001). Because the
government controls signi?cant portions of strategic
resources and has considerable power to approve projects
and allocate resources, managers tend to maintain a ‘dis-
proportionately greater contact’ with government of?cials
(Child, 1994). Empirically, Nee (1992) ?nds that managers’
networking with local of?cials is associated with better
performance of new ventures in China. Similarly, Peng
and Luo (2000) show that the strength of ties between
?rms’ managers and government of?cials is positively
associated with ?rms’ performance in China (also see, Li
& Atuahene-Gima, 2002; Li & Zhang, 2007; Lu, 2000).
Thus, political connections are an informal institution that
provides direct bene?ts in the form of government favors
and subsidies. The natural question that arises is whether
political connections enable ?rms to become more or less
self-sustaining in the long-run.
4
Accordingly, the objective
of this study is to examine whether ?rms that are politically
connected for a long period of time become more self-
sustaining, i.e., insulated from the loss of direct bene?ts
arising from such connections.
3
Other studies have examined how ?rms and other market participants
(such as ?nancial analysts) respond to changing institutional conditions
(for example see, Fogarty & Rogers, 2005; Greenwood & Hinings, 1996;
Hoffman, 1999; Newman, 2000; Peng, 2003).
4
Loosely speaking, the research question is similar to the question of
whether children with privilege such as belonging to higher socio-
economic strata turn out to be ‘spoilt’ or ‘good citizens’ of society; given
that children in higher socio-economic strata have an advantage not only
from obtaining direct bene?ts of extra favors resulting from their elite
status, but also from obtaining education and developing networks that
enables them to become more self-sustaining. Our question is whether
politically connected ?rms are similar to spoilt-children or good citizens of
society.
68 S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86
The Malaysian Context
We examine the research question in the Malaysian set-
ting for two reasons. First, Faccio et al. (2006) document
that among listed ?rms in various countries, Malaysia
has the largest percentage of politically connected ?rms.
In addition, prior research documents the value-enhancing
role of political connections (see Gul, 2006; Johnson &
Mitton, 2003). Furthermore, political connections have
been prevalent for a long period of time in Malaysia, which
enables us to examine differences between ?rms with
political connections over long and short periods of time.
Second, the exogenous shock of the 2008 General
Election loss of the supermajority by the ruling party, i.e.,
National Front or Barisan Nasional (BN) provides a good
event to examine whether ?rm values change differentially
across ?rms with long and short periods of political con-
nections. The 12th General Election was held on March 8,
2008 in Malaysia. For about 40 years, up until 2008, the
ruling party, Barisan National (BN) held 198 of the 222
(roughly 90%) parliamentary seats. This supermajority of
holding more than two-thirds of the seats, i.e., gave BN
‘‘absolute power’’ and enabled them to pass important
laws relating to special economic privileges for politically
connected ?rms with relative ease. In the 2008 Election,
BN won only 140 (roughly 63%) which is slightly shy of
the supermajority of two-thirds. Although BN was
expected to win fewer seats, it was not expected to lose
its supermajority. It was not until the day of the General
Election, 8 March 2008 that the public and the stock mar-
ket realized that BN’s supermajority was lost based on exit
polls and early polling returns and sentiment (see
Anonymous, 2008b, 2008c).
5
This loss of BN’s supermajority led to many important
changes in the political landscape. For example, the former
Prime Minister, Dr. Mahathir Mohammad, quit UMNO (the
dominant party in BN) after the Election (Anonymous,
2008a), thus increasing uncertainties for BN. Ahmad
(2008) states, ‘‘[w]hile the BN continues to hold a simple
majority, a tectonic shift has taken place in Malaysian pol-
itics, and it was in many ways engineered by Anwar
Ibrahim. The victory means a new political future for the
world’s most economically advanced Muslim country, ush-
ering in newchapter in Muslim democracy’’. This change in
political landscape and the relatively signi?cant presence
of the opposition in the parliament was expected to curtail
cronyism, i.e., BN’s ability to have a free hand in pursuing
?nancial and economic policies that favor politically
connected ?rms (Bendeich, 2008). In addition, the
increased scrutiny by the opposition was also expected to
impede the ruling party’s ability to continue awarding
contracts, licenses and subsidies to politically connected
?rms.
Hypotheses development
Differential market reaction to bad news across old- and new-
politically connected ?rms
We posit that ?rms with political connections for a
short period of time are different from those with such
connections for a relatively long period. As shown by prior
research, politically connected ?rms directly bene?t from
government favors and subsidies (see discussion in the pre-
vious section). In return, these ?rms are likely to pursue
policies and agendas that help the political actors – the
government of?cials and politicians – strengthen their
social and political agenda. For example, in Malaysia based
on the New Economic Policy (NEP), promulgated in 1971,
the government implemented policies to improve the eco-
nomic status of the Malays including setting up trust agen-
cies such as the National Trading Corporation (Perbadanan
National Bhd or PERNAS) to actively participate in the cor-
porate sector. Policies were also put in place to encourage
companies to appoint more Bumiputras (ethnic Malays) to
senior management positions. As a result, several ?rms
with political connections appointed Malays, who were
ex-government of?cials or ex-politicians, to senior execu-
tive and board positions (Case, 2005; Gomez, 2002;
Gomez & Jomo, 1997).
6
Building credibility of adherence to such government
policies takes time and ?rms that build such credibility
are likely to gain access to the informal political elite
club/network, which in turn provides managers of such
?rms with prestige and ‘high standing’ in society
(DiMaggio & Powell, 1983; Donaldson, 2008; Luo, 2003;
Sanders & Tuschke, 2007; Scott, 2008). The political elite
status is likely to enable ?rms to obtain important eco-
nomic resources. For example, ?rms that gain the political
elite status are likely to attract and retain talented man-
agers; talented managers would have an incentive to join
the politically connected ?rm to gain access to the political
elite network. Thus, ?rms that are politically connected for
a longer period are likely to not only obtain direct eco-
nomic bene?ts from government favors and subsidies,
but also obtain indirect bene?ts of attracting and retaining
high quality human capital and other important economic
resources that will enable them to be self-sustaining, i.e.,
become less dependent on the political connection (see
Rona-Tas, 1994). In comparison to this, ?rms that are polit-
ically connected for a short period are likely to obtain only
the direct economic bene?ts.
7
We refer to the ?rms with
political connections over long and short periods of time
as old and new-politically connected ?rms, respectively.
5
Before the Election, the general view was that ‘[w]idespread voter
apathy and disillusionment with politics will ensure that the United-
Malays-National-Organisation (UMNO)-led ruling Barisan Nasional (BN)
coalition breezes through the elections with a comfortable two-thirds
majority in the 222-member Parliament’ (Anonymous, 2008d), and ‘[n]o
new faces mean that the old policy and governance will be continued’
(Anonymous, 2008e). As such, BN’s subsequent loss in the Election is
considered as a ‘political tsunami. . . [j]ust like the natural tsunami that
never gives any sign before it happens. . . No one expected such results,
included BN and the opposition as there was no conspicuous sign before the
elections’ (Anonymous, 2008c). A similar view has been cited by The
Economist (Anonymous, 2008f), political analysts (e.g., Klein, 2011) and
even the US Congressional Research Service Report (Martin, 2008).
6
Case (1995) writing more than 15 years earlier points out that it
‘remains an open question whether this freely developed political savvy
will ripen into true business viability’ (p. 95).
7
These arguments parallel the arguments embedded in the surviving
elite thesis espoused by Rona-Tas (1994) in the sociology literature.
S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86 69
When the political apparatus that is the basis for the
political connection is subject to adverse events such as
the loss of supermajority in the General Election, ?rms with
political connections are not likely to obtain direct bene?ts.
However, the old-politically connected ?rms are likely to be
impacted less by the adverse event than the new-politically
connected ?rms, because the old-politically connected
?rms are likely to have used their membership in the infor-
mal political elite club/network to obtain key economic
advantages and resources such as talented human capital
and become self-sustaining. This empirical expectation is
stated in the following hypothesis:
Hypothesis 1. The impact of bad news with respect to
political connection is stronger for the new-politically
connected ?rms than the old-politically connected ?rms.
The alternative to Hypothesis 1 arises from the argu-
ment that the old-politically connected ?rms are more
entrenched because of their reliance on cronyism for a
longer period of time. However, we do not expect support
for this alternative hypothesis because as an emerging
economy with increased globalization and scrutiny by
international agencies such as International Transparency
and foreign institutional investors, all ?rms would have
an incentive to be more ef?cient. Furthermore, the alterna-
tive hypothesis posits that during the initial period of
political connection, the politicians and government of?-
cials help to enhance the ?rms’ value, and during the later
periods they obtain rents and favors from these ?rms. This
long-term quid pro quo is not likely to be sustained
because the managers and political actors may not have
a long horizon.
Mechanism of differential market reaction to bad news across
old- and new-politically connected ?rms
Hypothesis 1 is based on the notion that one possible
mechanism through which the old-politically connected
?rms insulate themselves from the bad news of election
loss, is to have gained access to better talented managers/
board of directors.
8
If this is the case, it is not only new-po-
litically connected ?rms that suffer the loss in ?rm value
from the bad news associated with the ruling party’s elec-
tion loss, but also those old-politically connected ?rms that
exhibit less improvement in managerial professionalism
over their time of political connection. This argument of
examining one possible mechanism is summarized in the
following hypothesis:
Hypothesis 2. The impact of bad news with respect to
political connection is stronger for old-politically con-
nected ?rms with less improvement in board/ executive
professionalism than old-politically connected ?rms with
more improvement in board/ executive professionalism.
The life cycle viewposits that young ?rms are likely to be
more susceptible to the loss of favors frompolitical connec-
tions than old ?rms (Peng &Luo, 2000). While under the life
cycle view, all old-politically connected ?rms would exhibit
insulationfromthe direct bene?ts and costs of suchconnec-
tions; Hypothesis 2 predicts that only those old-politically
connected ?rms that have improved their board/executive
professionalism will be insulated. Thus, Hypothesis 2 can
be viewed as a test for the alternative explanation of
Hypothesis 1 under the life cycle hypothesis.
Empirical analyses
Classifying politically connected ?rms
To identify politically connected ?rms (PCONN), we
start with the list of PCONN ?rms provided by Johnson
and Mitton (2003) and Faccio (2006) (hereafter referred
to as the JM_F List). In addition to the JM_F List, several
new ?rms with political connections have emerged under
the patronage of the ‘new political establishment’ such as
Kamaludin Abdullah, the son of Abdullah Badawi
(Barrock, 2007). We use the news reports in a leading
Malaysian business newspaper, The Edge to identify ?rms
that obtained political patronage in recent years. In addi-
tion, the government set up ‘government linked compa-
nies’ (GLCs) under Khazanah, an investment arm of
UMNO, which we include as politically connected ?rms.
9
For each of these ?rms, we examine the 2007 annual report,
which immediately precedes the event of General Election,
and classify the ?rm as politically connected if they satisfy
any of the following criteria: (a) had government cabinet
members and/or MPs as members of the board, or (b) had
signi?cant ownership by government or UMNO-linked orga-
nizations/ individuals, or (c) had politically connected indi-
viduals as their managers.
To test our hypotheses, we need to classify ?rms that
have been politically connected for long and short periods.
For this purpose, we use the 1996/1997 annual reports of
the ?rms identi?ed as politically connected in 2007.
10
We
require the politically connected ?rms to exist for ten years,
i.e., have a 1996/1997 annual report as well, so as to ensure
that our classi?cation of new- and old-politically connected
?rms do not proxy for ?rm age. This criterion helps to miti-
gate the potential concern that our results are driven by
the life cycle hypothesis. The variable PCONN = 1 for ?rms
that are politically connected in 2007. A PCONN ?rm is
tagged as having been politically connected for a long period,
i.e., Old-PCONN = 1 (New-PCONN = 0) if the ?rmsatis?es any
of the three criteria for being classi?ed as politically con-
nected in 1997/1998 as well. For a PCONN ?rm that satis?es
any of the three criteria only in 2007 but not in 1997/1998, it
is considered as having been politically connected for a short
period, i.e., New-PCONN = 1 (Old-PCONN = 0).
8
Even though the argument for attracting more talented managers is
applicable also for the operating line managers, we posit the hypothesis in
terms of the board and top management with the notion that good
management practices at the top-level will be re?ective of more talented
managers at lower levels as well.
9
Examples include Malaysia Airport Holding Berhad, Malaysian Airline
System Berhad, Tenaga Nasional Berhad etc. See the website for Khazanah
Nasional Berhad:http://www.khazanah.com.my/portfolio.htm.
10
We choose 1997/1998 because it was the year after the Asian ?nancial
crisis so as to mitigate survival issues; and furthermore, the availability of
annual reports from earlier years is very spotty.
70 S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86
In total, we identify 122 politically connected ?rms – 66
are from the JM_F list; 56 are identi?ed from newspaper
reports. Out of the 122 PCONN ?rms, there are 39 New-
PCONN ?rms and 83 Old-PCONN ?rms.
11,12
Out of the 56
additional politically connected ?rms, 17 are Old-PCONN
?rms and 39 are New-PCONN ?rms. Appendix A provides
the list of PCONN ?rms classi?ed as Old- and New-PCONN
and the following details: (a) whether the ?rm belongs to
the JM_F list, (b) the criteria that the ?rm satis?es to be clas-
si?ed as politically connected, and (c) the age of the ?rm –
the age of the ?rm is obtained from the ?rm’s website or
the ?rm’s history in the library of Bursa Malaysia.
A premise embedded in Hypothesis 1 is that both the
old- and new-politically connected ?rms have equally
strong political ties; if the strength of political connections
is less for the old-politically connected ?rms than the new-
politically connected ?rms; or political connections are
non-existent for the old-politically connected ?rms so
much so that they are similar to non-politically connected
?rms, then also we will ?nd support for the hypothesis.
To address this, we examine news reports in the mid-
2000s to ensure that the Old-PCONN ?rms continued to
obtain favors at the time of our analyses (see Appendix B
for a few examples).
Even though the ruling party (BN) had lost supermajor-
ity in the General Election, it still maintained a majority.
Thus, we could ?nd support for Hypothesis 1 if the favors
are more likely to occur for the Old-PCONN ?rms than
New-PCONN ?rms i.e., the entrenched cronyism effect. To
mitigate this potential concern, we examine news reports
after the Election to verify whether both Old- and New-
PCONN ?rms face adverse consequences after the
Election (see Appendix C for a few examples).
13
Managerial (board/executive) professionalism
To test Hypothesis 2, we measure the professionalismof
the board of directors and the top-executives, i.e., CEO and
Chairman. For board professionalism, we consider the fol-
lowing governance-related variables: Indpc is the propor-
tion of independent directors as disclosed in the annual
reports, Bsize is the total number of board members,
DiverseBOD is the proportion of board members who are
non-Malays, and IndChair is one if the chairman is an inde-
pendent director. For executive professionalism we include
the following variables: CeoNMalay (ChairNMalay) is one if
the CEO (chairman) is a non-Malay, CeoMBA (ChairMBA) is
one if the CEO (chairman) has an MBA degree or equiva-
lent, CeoOverseas (ChairOverseas) is one if the CEO (chair-
man) is educated overseas, CeoOthDir (ChairOthDir) is one
if the CEO (chairman) is a director of another company,
CeoIndExp (ChairIndExp) is one if the CEO (chairman) has
been in managerial positions in the same ?rm or other
?rms in the same industry.
We obtain this data from the annual reports of the
PCONN ?rms for 1996/1997 and 2007. While most of these
factors are governance and executive expertise variables
used in prior studies, the diversity factor of including
non-Malays as an indicator of managerial professionalism
is novel to the Malaysian context. This is because while
non-Malays are likely to be a board member or top-execu-
tive because of their merit, it is possible that Malays are in
the position to ful?ll the social and political agenda.
However, it should be noted that if the labor market is
mature then the Malays who are chosen to these top posi-
tions are also likely to exhibit expertise similar to the non-
Malays as well as help the PCONN ?rm build credibility of
adherence to the political agenda.
We compute the change in each of the 14 characteris-
tics of Old- and New-PCONN ?rms from 1997 to 2007
and give a score of one for an increase in the value of each
of the variable from 1997 to 2007. We then add up the
scores of the 14 factors to construct a professionalism
score (Score).
Events leading up to the General Election
The results of the General Election were announced on
March 10, 2008. As discussed earlier, the loss of the super-
majority by the ruling party was a surprise to the public
and the stock market. Even though the media reports indi-
cate that the loss of supermajority by BN was a surprise, it
could be possible that some events leading up to the
General Election could have provided an advance warning
about this outcome to the stock market and the stock mar-
ket reacted to that news for the Old-PCONN ?rms at that
time. As such, we consider such events in our research
design so as to emphasize that the loss of supermajority
was indeed an exogenous shock.
We identify three fact-based events that changed the
tracking polls considerably. These events are summarized
in Appendix D. On November 10, 2007 there was a mass
demonstration against the government held in Kuala
Lumpur (Event1). This demonstration was spear-headed
by Anwar Ibrahim, the former Deputy Prime Minister,
who was jailed for four years for various corruptioncharges.
By some estimates there were 40,000 people involved in the
demonstration and 15 demonstrators were arrested by
police. The second demonstration (Event2) of roughly
15,000 mostly ethnic Indian protestors was held on
November 25, 2007 and organized by the Hindu Rights
Action Front (HINDRAF). HINDRAF protested the marginal-
ization of ethnic Indians in Malaysia. The third event on
February 13, 2008 (Event3) was the announcement of the
General Election and the dissolution of the Parliament
(Anonymous, 2008d). These three events, i.e., Event1-
Event3 were unfavorable to the ruling party BN and are
11
We identify 17 GLCs out of which 7 are in the JM_F list and also ful?ll
the criteria of Old-PCONN.
12
Even though our procedure for identifying PCONN ?rms is extensive, it
is possible that some politically connected ?rms have not been identi?ed as
PCONN. In effect, while our procedure ensures that all PCONN ?rms are
indeed politically connected, we may have tagged some politically
connected ?rms as non-PCONN. This measurement error would bias
against ?nding the predicted negative stock market reaction for both the
New- and Old-PCONN ?rms.
13
Note that even if we gather data on all of such reported adverse events,
it will not provide the complete picture of adverse consequences. This is
because not all adverse events are newsworthy and hence many adverse
events may not be visible or transparent. As such, we search and obtain a
couple of examples to highlight the fact that the loss of supermajority by
the ruling party was indeed bad news for the PCONN ?rms.
S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86 71
likely to be indicative of the potential for loss of supermajor-
ity. While ex post, i.e., after Event4, Events1-Event3 can be
classi?ed as ‘‘bad news’’ events for BN, ex ante whether it
was as such is not clear. Thus, we do not have predictions
on the market’s reaction to Events1–3.
Sample and data
The sample contains 834 non-politically connected
?rms (PCONN = 0) listed on the Bursa Malaysia (Kuala
Lumpur Stock Exchange) with available data and 122 polit-
ically connected ?rms (PCONN = 1). Daily stock closing
prices and the market index, Kuala Lumpur Stock
Exchange Composite Index (KLCI) are obtained from
Datastream.
14
Stock returns for each company are calcu-
lated using the percentage change in daily closing prices,
adjusted for dividends, stock splits and rights issues (Fama,
Fisher, Jensen, & Roll, 1969). Market returns are calculated
based on the KLCI returns. We measure cumulative abnor-
mal returns (CARs) around the announcement date of the
General Election results (March 10, 2008) by subtracting
the KLCI’s value-weighted daily return.
15
Research design
To test Hypothesis 1, we estimate the following model
over a six-month period (121 days) from October 1, 2007
to March 31, 2008 (see Binder, 1985; Millon-Cornett &
Tehranian, 1989, 1990; Schipper & Thompson, 1983,
1985; Smith, Bradley, & Jarrell, 1986):
ðR
i
ÀR
f
Þ ¼ b
0
þb
1
Event1 þb
2
Event2 þb
3
Event3
þb
4
Event4 þb
5
PCONN þb
6
New-PCONN
þb
7
PCONNÂEvent1 þb
8
PCONNÂEvent2
þb
9
PCONNÂEvent3 þb
10
PCONNÂEvent4
þb
11
New-PCONNÂEvent1
þb
12
New-PCONNÂEvent2
þb
13
New-PCONNÂEvent3
þb
14
New-PCONNÂEvent4
þb
15
ðR
m
ÀR
f
Þ þb
16
logðSizeÞ þb
17
logðMBÞ
þb
18
logðLeverageÞ þb
19
logðFirmAgeÞ
þIndustry-fixed effects þerror ð1Þ
where R
i
is the daily return of ?rm i; R
f
is the daily return
on the Malaysian treasury bill; Event1–4 are the dummy
variables that take on a value of 1 for the three-day
window surrounding the four events leading up to the
General Election (see Appendix D)
16
; PCONN is one if
the ?rm is a politically connected ?rm in 2007; and
New-PCONN is one if the PCONN ?rm is not connected in
1997–1998 but is connected in 2007 (see Appendix B); R
m
is the daily return on KLCI; log(Size) is the log of the market
capitalization of the ?rm in the prior year, log(MB) is the log
of the market-to-book ratio in the prior year, and
log(Leverage) is the log of the long-term-debt to assets ratio
in the prior year. Since New-PCONN ?rms could be system-
atically younger ?rms than Old-PCONN ?rms, we include
log(FirmAge) which is the log of the ?rm age obtained from
the ?rms history to control for the potential age effects on
our results.
17
The standard errors of the coef?cient estimates can be
biased downwards because by design the ?rms are
repeated daily in the sample. In addition, macro events
such as the election results are also likely to be cross-sec-
tionally correlated. We use the robust standard errors clus-
tered by ?rm and date to mitigate the potential bias in
standard errors (see Gow, Ormazabal, & Taylor, 2010;
Petersen, 2009).
18
Based on earlier studies (e.g. Fisman, 2001), we expect
the coef?cient estimate on PCONN Â Event4, i.e., b
10
to be
negative, and based on Hypothesis 1 we expect the coef?-
cient estimate on New-PCONN Â Event4, i.e., b
14
to be
negative.
To test Hypothesis 2, we augment Eq. (1) by including
the change in managerial professionalism score.
Empirical results
Market reaction to the General Election results for Old- and
New-PCONN ?rms
Table 1, Panel A provides the mean cumulative abnor-
mal returns (CARs) for the PCONN ?rms (N = 122) and
non-PCONN ?rms (N = 834), as well as for the Old-
PCONN (N = 83) and New-PCONN (N = 39) ?rms, over dif-
ferent windows surrounding the General Election result
announcement date, i.e., March 10, 2008. While the mean
CAR for the event windows (À1, +1), (À1, +3) and (À1,
+5) are not signi?cantly different from zero for the non-
PCONN ?rms (e.g., t-statistic = 0.63 for the window À1,
+1), the mean CAR of PCONN ?rms for the event windows
(À1, +1), (À1, +3) and (À1, +5) are À3.12%, À4.32% and
À7.17% with corresponding t-statistics of À4.37, À5.65
and À7.36, respectively. This shows that the difference in
CAR across PCONN and non-PCONN, i.e., PCONN minus
non-PCONN is negative and statistically signi?cant, consis-
tent with ?ndings in prior studies (e.g., Fisman, 2001) that
political connections are positively associated with ?rm
value.
The columns on the right-hand side examine the differ-
ential market reaction for Old- and New-PCONN ?rms
based on Hypothesis 1. We ?nd that the difference in
Old-PCONN’s mean CAR and New-PCONN’s mean CAR for
the event windows (À1, +1), (À1, +3) and (À1, +5) are
14
For a few ?rms (less than one percent of our sample) with missing stock
price data in Datastream, we collect the data from Yahoo! Finance.
15
We also compute expected returns based on (1) the market model
estimated from March 1, 2007 to February 29, 2008; as well as (2) the ?rm-
speci?c average returns from March 1, 2007 to February 29, 2008 and we
obtain similar results. This is consistent with Brown and Warner’s (1985)
simulation ?ndings that results in short term event analyses are not
sensitive to the use of different expected return estimation methods.
16
We use a combined regression for the four events because the larger
panel data enhances statistical power. In essence, we give Events1–3 a
‘‘good’’ chance to be statistically signi?cant.
17
In unreported analysis we control for prior pro?tability (ROA), number
of segments and proportion of foreign sales and obtain similar results.
18
We cluster by month instead of day and cluster by ?rm alone and
obtain similar results.
72 S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86
À3.17%, À2.80% and À4.31%, with corresponding t-statis-
tics of À2.10, À1.72 and À2.09, respectively. This shows
initial support for Hypothesis 1. Furthermore, for the event
windows (À1, +1), (À1, +3) and (À1, +5) the mean CAR for
Old-PCONN (New-PCONN) are À2.11% (À5.28%), À3.43%
(À6.22%) and À5.80% (À10.11%) respectively, showing that
Table 1
Descriptive statistics and univariate tests of market reaction to election results.
Non-PCONN PCONN Diff Old-PCONN New-PCONN Diff
Panel A: Univariate tests of market reaction to election results
EvtCAR (À1, +1) 0.52% À3.12% À3.65% À2.11% À5.28% À3.17%
t-Stat 0.63 À4.37 3.33 À2.36 À4.79 À2.10
p-Value (0.53) (0.00) (0.00) (0.02) (0.00) (0.04)
EvtCAR (À1, +3) 0.85% À4.32% À5.17% À3.43% À6.22% À2.80%
t-Stat 0.99 À5.65 4.51 À3.78 À4.50 À1.72
p-Value (0.32) (0.00) (0.00) (0.00) (0.00) (0.09)
EvtCAR (À1, +5) 0.79% À7.17% À7.96% À5.80% À10.11% À4.31%
t-Stat 0.88 À7.36 6.00 À5.17 À5.53 À2.09
p-Value (0.38) (0.00) (0.00) (0.00) (0.00) (0.04)
Variable Non-PCONN (N = 834) PCONN (N = 122) Tests of
Mean Median Mean Median Means Medians
Panel B: Univariate tests of ?rm characteristics between PCONN and non-PCONN ?rms
Size 534.70 98.15 2808.07 849.97 3.10 (0.00) 10.3 (0.00)
MB 1.17 0.88 2.08 1.51 2.72 (0.01) 5.41 (0.00)
Leverage 0.10 0.10 0.12 0.11 0.64 (0.52) 0.64 (0.52)
FirmAge 22.98 18.00 33.82 32.00 5.83 (0.00) 6.66 (0.00)
ROA (%) 5.08 5.23 6.41 5.29 0.63 (0.53) 0.26 (0.79)
No. of segments 1.37 1.00 1.42 1.00 0.27 (0.79) 1.06 (0.29)
% Foreign sales 10.23 0.00 10.19 0.00 0.02 (0.98) 1.60 (0.11)
Cash?ow per share 0.12 0.03 0.30 0.11 3.56 (0.00) 3.57 (0.00)
Sales volatility 0.24 0.19 0.24 0.18 0.29 (0.77) 0.65 (0.52)
Sales growth 0.94 0.19 0.43 0.17 1.01 (0.31) 0.79 (0.43)
Variable Old-PCONN (N = 83) New-PCONN (N = 39) Tests of
Mean Median Mean Median Means Medians
Panel C: Univariate tests of ?rm characteristics between Old-PCONN and New-PCONN ?rms
Size 2775.55 737.74 2877.28 992.55 0.05 (0.96) 0.76 (0.45)
MB 2.43 1.62 1.35 1.37 2.28 (0.03) 1.35 (0.18)
Leverage 0.12 0.14 0.12 0.04 0.02 (0.98) 0.63 (0.53)
FirmAge 35.89 35.00 29.41 24.00 1.74 (0.08) 2.58 (0.01)
ROA (%) 9.06 5.56 0.78 4.39 2.40 (0.02) 1.56 (0.12)
No. of segments 1.45 1.00 1.36 1.00 0.30 (0.77) 0.32 (0.75)
% Foreign sales 10.75 0.00 9.00 0.00 0.42 (0.67) 0.59 (0.55)
Cash?ow per share 0.29 0.11 0.33 0.13 0.29 (0.77) 0.34 (0.73)
Sales volatility 0.25 0.19 0.23 0.17 0.66 (0.51) 0.29 (0.77)
Sales growth 0.50 0.16 0.28 0.18 0.88 (0.38) 0.34 (0.73)
Notes
1. In Panel A, the sample contains 956 ?rms. The columns ‘Non-PCONN’ and ‘PCONN’ represent ?rms with no political connections (N = 834) and ?rms
with political connections (N = 122) respectively. In addition, the column ‘New-PCONN’ includes ?rms with political connections for a short-period
(N = 39), and column ‘Old-PCONN’ includes ?rms with political connections for a long-period (N = 83).
2. PCONN is an indicator variable that equals one for politically connected ?rms, and zero otherwise. A ?rm is politically connected if any of the following
three criterion is satis?ed in 2007: either (a) had government cabinet members and/or MPs as members of the board, or (b) had signi?cant ownership
by government or UMNO-linked organizations/individuals, or (c) had politically connected individuals as their managers. They are identi?ed based on
(1) the ?rm list provided in Johnson and Mitton (2003) and/ or Faccio (2006), and (2) the search for recent news reports and articles. A PCONN ?rm is
tagged as having been politically connected for a long period, i.e., Old-PCONN = 1 (New-PCONN = 0) if the ?rm satis?es the three criteria for being clas-
si?ed as politically connected in 1997/1998 as well, and is considered as having been politically connected for a short period, i.e., New-PCONN = 1 (Old-
PCONN = 0) if the PCONN ?rm satis?es the three criteria only in 2007 but not 1997/1998. See Appendix A for identi?cation strategy for both Old-PCONN
and New-PCONN ?rms.
3. EvtCAR is the cumulative abnormal returns for the different windows centered around the General Election results announcement date of March 10,
2008 (Day 0).
4. Panel B (Panel C) tests the mean and median differences between Non-PCONN and PCONN (between Old-PCONN and New-PCONN) on a range of ?rm
characteristics: Size is the market capitalization, MB is the market to book ratio, Leverage is the total debts to total assets, FirmAge is the age of the ?rm
since its incorporation, ROA is the returns on assets, No. of Segments is the number of geographical segments, % Foreign Sales is the percentage of for-
eign sales to total sales, Cash?ow per share is the total cash ?ow divided by the number of outstanding shares, Sales volatility is the standard deviation
of sales over the past three years, and Sales growth is the percentage change in sales from year t À 1 to year t. t-values and p-values are based on two-
tailed tests and are reported in italics (p-values are in parentheses).
S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86 73
the stock market reaction to both Old- and New-PCONN
?rms are signi?cantly negative, which provides some fur-
ther con?dence that Old-PCONN ?rms are politically
connected.
Table 1, Panels B and C provide the means and medians
of ?rm characteristics. Panel C shows that Old-PCONN and
New-PCONN ?rms are similar on most dimensions of ?rm
characteristics and risk pro?les, except that Old-PCONN
?rms have higher market to book ratios and pro?tability
on average, and are older than New-PCONN ?rms. For this
reason we control ?rm age in our estimations to ensure
that our results are not driven by the difference in ?rm
age between the two groups.
19
Table 2 presents the results of estimating Eq. (1). In the
left set of columns we estimate Eq. (1) without the New-
PCONN variables to provide a benchmark. In estimating
Eq. (1) presented in the right set of columns, we ?nd that
the coef?cient estimate on the interaction between
PCONN Â Event4 is À0.007 (t-stat = À11.350) and the
interaction between New-PCONN Â Event4 is À0.023 (t-
stat = À32.7). The negative coef?cient of PCONN Â Event4
is consistent with prior studies in political connection
(see Fisman, 2001). The negative coef?cient on New-
PCONN Â Event4 supports Hypothesis 1 and shows that
New-PCONN ?rms are more adversely affected by the rul-
ing party’s loss of supermajority in the General Election. In
terms of economic signi?cance, the New-PCONN ?rms are
roughly 3.29 [=0.023/0.007] times more adversely affected
than Old-PCONN ?rms.
The market reaction is more negative for PCONN ?rms
than Non-PCONN ?rms for Event3 and Event4 (see
coef?cients PCONN Â Event3 and PCONN Â Event4); and,
interestingly, although these preceding events may be
indicative of the loss in the popularity of the ruling party,
we ?nd that the market reacts more positively for PCONN
?rms than non-PCONN ?rms for Event1 and Event 2 (see
coef?cients PCONN Â Event1 and PCONN Â Event2). This
could be because while the event of dissolution of the
Parliament (Event3) and the result of the Election
(Event4) are both related directly to the Election, the mass
demonstrations against the ruling party captured by Event1
and Event2 are indicative of the power of the ruling party.
Further tests controlling for ?rm age of Old- and New-PCONN
?rms
The life cycle hypothesis predicts that young ?rms are
likely to be more susceptible to the loss of favors from
political connections than the old ?rms (Peng & Luo,
2000), and it is possible that our measure of old- and the
new-politically connected ?rms are simply a proxy for ?rm
age and young ?rms captured by our measure of new-po-
litically connected ?rms are less resistant to adverse
shocks. Even though we control for FirmAge in Eq. (1), this
may not be suf?cient as it only controls for the average
effect of ?rm age. In Table 3, we conduct a number of tests
to further ensure that our results are not driven by the dif-
ference in age of the Old- and New-PCONN ?rms. This is
important to rule out the alternative explanation of the life
cycle hypothesis.
First, we consider PCONN ?rms below median age as
young PCONN ?rms; accordingly, Young-PCONN is an indi-
cator variable that is one if the ?rm-age is below the med-
ian ?rm age, and 0 otherwise. We include Young-PCONN as
an additional variable in Eq. (1) with the corresponding
interactions with the events. If New-PCONN is correlated
with young PCONN ?rms, then the results documented in
Table 2 is likely to be driven by Young-PCONN. The result
of this estimation is provided in the leftmost column of
Table 3. The result shows that while Young-PCONN Â
Event4 is signi?cantly negative, New-PCONN Â Event4 is
also negative and signi?cant. In particular, the coef?cient
estimate (À0.021) is similar to the results reported in
Table 2 (À0.023), which indicates that the results in
Table 2 are not driven by young ?rms.
Second, if our results are driven by younger ?rms being
more prone to economic shocks, then we should not ?nd
results when we exclude young PCONN ?rms in our sam-
ple. To examine this possibility, we exclude all PCONN
?rms (both Old- and New-PCONN ?rms) with below med-
ian ?rm age (less than 32 years). With this smaller sample
we re-estimate Eq. (1); the results are reported in the mid-
dle set of columns in Table 3 and show that the results are
similar to those discussed in Table 2.
Third, we further control for the age difference between
Old- and New-PCONN ?rms by matching each New-
PCONN ?rm to an Old-PCONN ?rm based on ?rm age and
total assets. We match each New-PCONN ?rm to an
Old-PCONN ?rm because there are fewer New- than
Old-CONN ?rm. The results are reported in the rightmost
set of columns of Table 3 and are also similar to those
discussed in Table 2.
20
Overall, these results suggest that our ?ndings reported
in Table 2 are not likely to be driven by difference in ?rm
age across Old- and New-PCONN ?rms, i.e., New-PCONN
?rms are younger and thus more sensitive to economic
shocks.
Change in board/executive professionalism across Old- and
New-PCONN ?rms
Table 4, Panel A shows the increase/ improvements in
the 14 variables of board/ executive professionalism across
Old- and New-PCONN ?rms from 1997 to 2007. The Score
variable in the last row is the total of the increase in the 14
components. The mean Score of Old-PCONN is 1.831
compared to 0.821 of New-PCONN; this suggests that the
Old-PCONN ?rms make improvements on roughly two of
the dimensions while the New-PCONN ?rms make
19
Note that since we employ a short window event study methodology,
such characteristics should not affect the results. Nevertheless, in a
sensitivity analysis, we repeat our analyses by controlling for these
characteristics and obtain similar results.
20
In unreported robustness tests we ?nd similar results when we re-
estimate Eq. (1) by doing each of the following: (1) exclude non-PCONN
?rms with below median ?rm age, (2) exclude only New-PCONN ?rms with
below median ?rm age, (3) use the ?rm age cutoff of 15 years to de?ne
young ?rms instead of the median age. We do not consider the matched
sample for our primary research design, so as to preserve a reasonable
sample size for the test of Hypothesis 2.
74 S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86
improvements on one dimension.
21
Roughly 30%, 40% and
50% of the Old-PCONN ?rms show improvements in board
independence, board size and diversity, respectively; com-
pared to only roughly 10–20% of New-PCONN ?rms.
Roughly 10% Old-PCONN ?rms change from Malay CEO to
non-Malay CEO, while only 3% of New-PCONN ?rms exhibit
this change. Also, we ?nd that 7% of the Old-PCONN ?rms
have improvements in the industry expertise of the
Chairman but such improvement is not observed for New-
PCONN ?rms. All the other changes are not different
between the two groups. Overall, these results suggest that
compared to the New-PCONN ?rms, board professionalism
of the Old-PCONN ?rms improved over time. Overall, while
our professionalism measure captures both board (gover-
nance) characteristics and senior executive characteristics,
empirically we observe that the differences in professional-
ism between Old- and New-PCONN ?rms are driven by the
board (governance-related) professionalism. This is likely
because the board-related characteristics are proportions
and thus a continuous variable; while the CEO/Chair related
variables are indicator variables. Thus, the change in the
indicator variables may not be powerful enough to capture
improvement when compared to the board-related vari-
ables. We continue our analysis with the caveat of an under-
lying assumption in our analysis is that improvements in
board characteristics percolate to the managerial cadre of
the ?rm as well. Overall, our results on changes in board/
executive professionalism are attributable to changes in
board professionalism.
Table 4, Panel B compares the proportion of Old-PCONN
(New-PCONN) ?rms that exhibit an increase in the Score
measure above the mean in the sample over the ten-year
time period, i.e., 1997 to 2007. We use the mean of the
Score among the Old-PCONN to classify the Old- and
New-PCONN ?rms as High and Low Score. We ?nd that
slightly more than half (45 out of 83 ?rms) of the Old-
PCONN ?rms are classi?ed as High Score. However, less
than one-third (10 out of 39 ?rms) of the New-PCONN
?rms are classi?ed as High Score, and the difference in
Table 2
Multivariate tests of market reaction to election results.
Dependent variable = R
i
– R
f
Eq. (1) without New-PCONN Eq. (1)
Coef t-stat p-value Coef t-stat p-value
Intercept 0.002 1.110 0.269 0.002 1.110 0.268
Event1 0.001 1.030 0.304 0.001 1.030 0.304
Event2 À0.010 À7.980
We examine whether the relationship between political connections and firm value is
moderated by the length of time firms have been politically connected. We find that compared
to firms with political connections for a short period, firms with political connections
for a long period have a smaller magnitude of negative stock price reaction to the 2008
General Election loss of the supermajority by the ruling party in Malaysia. We also find that
the smaller magnitude of negative stock price reaction is, in part, attributable to improvements
in board of director characteristics.
Corporate political connections and the 2008 Malaysian election
Simon Y.K. Fung
a,?
, Ferdinand A. Gul
b,1
, Suresh Radhakrishnan
c,2
a
School of Accounting and Finance, The Hong Kong Polytechnic University, Hung Hom, Hong Kong
b
School of Business, Monash University Malaysia, Malaysia
c
School of Management, The University of Texas at Dallas, Richardson, TX 75083-0688, USA
a b s t r a c t
We examine whether the relationship between political connections and ?rm value is
moderated by the length of time ?rms have been politically connected. We ?nd that com-
pared to ?rms with political connections for a short period, ?rms with political connections
for a long period have a smaller magnitude of negative stock price reaction to the 2008
General Election loss of the supermajority by the ruling party in Malaysia. We also ?nd that
the smaller magnitude of negative stock price reaction is, in part, attributable to improve-
ments in board of director characteristics. Furthermore, we ?nd that while the perfor-
mance subsequent to the General Election of politically connected ?rms is worse than
that of non-politically connected ?rms, ?rms with political connections for a long period
exhibit better performance than those connected for short periods. Collectively, the evi-
dence shows that the length of political connections is an important factor that moderates
economic value.
Ó 2015 Published by Elsevier Ltd.
Introduction
Prior research shows that ?rms’ political connections
are positively associated with ?rm value (Bliss & Gul,
2012; Boubakri, Guedhami, Mishra, & Saffar, 2012;
Faccio, 2006; Fisman, 2001; Johnson & Mitton, 2003). Li
and Zhang (2007) show how managers’ functional
experience and their political network help to enhance
performance of new ventures (also see Lester, Hillman,
Zardkoohi, & Cannella, 2008). A natural question that arises
is whether political connections continue to be as
important for ?rms during later periods as they are in
initial periods. Accordingly, we examine the following
two research questions: (a) Is the relationship between
?rms’ political connection and ?rm value different for
?rms with political connections for longer and shorter
periods? (b) What is the mechanism that drives the differ-
ential relationship?
We draw on institutional theory in sociology to develop
the hypothesis that the relationship between political con-
nections and ?rm value is moderated by the length of time
over which ?rms are politically connected (see Rona-Tas,
1994). Speci?cally, we posit that compared to ?rms that
are politically connected for a short period of time, ?rms
that are politically connected for a long period of time
are more likely to have built their credibility of adherence
to government policies which allows them access to the
informal political elite network (DiMaggio & Powell,
1983; Donaldson, 2008; Sanders & Tuschke, 2007). This
in turn would enable them to obtain important economic
resources such as talented managers and become more
self-sustaining. Thus, we expect ?rms with political con-
nections for a long period to be less severely affected by
the loss of political power by the ruling party (in termshttp://dx.doi.org/10.1016/j.aos.2015.04.001
0361-3682/Ó 2015 Published by Elsevier Ltd.
?
Corresponding author. Tel.: +852 2766 4246.
E-mail addresses: [email protected] (S.Y.K. Fung), [email protected]
(F.A. Gul), [email protected] (S. Radhakrishnan).
1
Tel.: +60 3 5514 4997.
2
Tel.: +1 972 883 4438.
Accounting, Organizations and Society 43 (2015) 67–86
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
of their supermajority) than ?rms with political connec-
tions for a short period.
We examine this hypothesis in the 2008 Malaysian
General Election setting where the ruling party – the
National Front or Barisan Nasional (BN) – unexpectedly
lost their longstanding supermajority. To identify politi-
cally connected ?rms, we update the list provided by
Johnson and Mitton (2003) and Faccio, Masulis, and
McConnell (2006). Firms that are identi?ed as politically
connected in both 1997/1998 and 2007 are classi?ed as
?rms with political connections for a long period (referred
to as old-politically connected ?rms); and ?rms that are
politically connected only in 2007 are classi?ed as ?rms
with political connections for a short period (referred to
as new-politically connected ?rms). In particular, we iden-
tify 122 politically connected ?rms out of which 39 are
new-politically connected ?rms.
Consistent with our hypothesis, we ?nd that the magni-
tude of negative stock market reaction to the ruling party’s
election loss of supermajority is less for the old-politically
connected ?rms than the new-politically connected ?rms.
To examine the mechanism through which the old-
politically connected ?rms insulate themselves from the
adverse consequences of the loss in supermajority power
by the ruling party, we investigate the change in board of
director characteristics and senior executive characteris-
tics over ten years across old- and new-politically con-
nected ?rms. We ?nd that the old-politically connected
?rms become more professional than the new-politically
connected ?rms. Interestingly, we ?nd that changes in
board professionalism exhibits improvement for old-
politically connected ?rms over the ten years, but execu-
tive professionalism does not. Furthermore, we ?nd that
the negative stock price reaction to the election result is
attenuated for those old-politically connected ?rms who
show improvements in professionalism, and not for those
who do not show such improvements. While, this evidence
provides insight into the mechanism through which polit-
ical connections enhance value, the access to managerial
talent that old-politically connected ?rms develop can by
itself help them to continue to extract rents even when
the direct bene?ts of political connection are removed. As
such, this evidence could be interpreted as a form of
cronyism as well.
Lastly, we attempt to address whether the improve-
ments in board professionalism could be related to crony-
ism, by examining the subsequent accounting and stock
market performance. In particular, we consider return on
assets and annual stock return for the three years subse-
quent to 2008 for the old- and new-politically connected
?rms. We ?nd that compared to the non-politically con-
nected ?rms, the accounting and stock market return per-
formance are worse for both the old- and new-politically
connected ?rms; but less so for the old-politically con-
nected ?rms than the new-politically connected ?rms.
This provides additional evidence that while old-politically
connected ?rms are more self-sustaining than the new-
politically connected ?rms, both of them exhibit worse
performance than non-politically connected ?rms.
The remainder of the paper is organized as follows: First
we provide the background literature and the research
question; second, we describe the Malaysian context and
the General Election in 2008, and develop the hypotheses;
third, we present the empirical analysis; and ?nally, we
provide some concluding remarks.
Background and hypotheses
Background and research question
The institutional view is based on the premise that
organizations adopt structures in response to their institu-
tional environments and hence is an important theoretical
lens in international business (see Carpenter & Feroz, 2001;
DiMaggio & Powell, 1983; North, 1990; Peng, Sun,
Pinkham, & Chen, 2009; Peng, Wang, & Jiang, 2008; Wan
& Hoskisson, 2003). Speci?cally, ?rms’ operations and
management practices are bound by formal and informal
country-speci?c institutions (see Ingram & Silverman,
2002). The formal and informal institutions in?uence mar-
ket competition (Khanna & Palepu, 1997; Peng, 2003;
Porter, 1990), and ?rms’ resource allocation decisions
either directly (North, 1990; Wan & Hoskisson, 2003), or
indirectly through management and corporate governance
practices (Aguilera & Jackson, 2003; La Porta, Lopez-
de-Silanes, Shleifer, & Vishny, 2000).
3
Managers’ ties with government of?cials, i.e., political
connections represent a unique economic resource in tran-
sition economies (Li & Atuahene-Gima, 2001). Because the
government controls signi?cant portions of strategic
resources and has considerable power to approve projects
and allocate resources, managers tend to maintain a ‘dis-
proportionately greater contact’ with government of?cials
(Child, 1994). Empirically, Nee (1992) ?nds that managers’
networking with local of?cials is associated with better
performance of new ventures in China. Similarly, Peng
and Luo (2000) show that the strength of ties between
?rms’ managers and government of?cials is positively
associated with ?rms’ performance in China (also see, Li
& Atuahene-Gima, 2002; Li & Zhang, 2007; Lu, 2000).
Thus, political connections are an informal institution that
provides direct bene?ts in the form of government favors
and subsidies. The natural question that arises is whether
political connections enable ?rms to become more or less
self-sustaining in the long-run.
4
Accordingly, the objective
of this study is to examine whether ?rms that are politically
connected for a long period of time become more self-
sustaining, i.e., insulated from the loss of direct bene?ts
arising from such connections.
3
Other studies have examined how ?rms and other market participants
(such as ?nancial analysts) respond to changing institutional conditions
(for example see, Fogarty & Rogers, 2005; Greenwood & Hinings, 1996;
Hoffman, 1999; Newman, 2000; Peng, 2003).
4
Loosely speaking, the research question is similar to the question of
whether children with privilege such as belonging to higher socio-
economic strata turn out to be ‘spoilt’ or ‘good citizens’ of society; given
that children in higher socio-economic strata have an advantage not only
from obtaining direct bene?ts of extra favors resulting from their elite
status, but also from obtaining education and developing networks that
enables them to become more self-sustaining. Our question is whether
politically connected ?rms are similar to spoilt-children or good citizens of
society.
68 S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86
The Malaysian Context
We examine the research question in the Malaysian set-
ting for two reasons. First, Faccio et al. (2006) document
that among listed ?rms in various countries, Malaysia
has the largest percentage of politically connected ?rms.
In addition, prior research documents the value-enhancing
role of political connections (see Gul, 2006; Johnson &
Mitton, 2003). Furthermore, political connections have
been prevalent for a long period of time in Malaysia, which
enables us to examine differences between ?rms with
political connections over long and short periods of time.
Second, the exogenous shock of the 2008 General
Election loss of the supermajority by the ruling party, i.e.,
National Front or Barisan Nasional (BN) provides a good
event to examine whether ?rm values change differentially
across ?rms with long and short periods of political con-
nections. The 12th General Election was held on March 8,
2008 in Malaysia. For about 40 years, up until 2008, the
ruling party, Barisan National (BN) held 198 of the 222
(roughly 90%) parliamentary seats. This supermajority of
holding more than two-thirds of the seats, i.e., gave BN
‘‘absolute power’’ and enabled them to pass important
laws relating to special economic privileges for politically
connected ?rms with relative ease. In the 2008 Election,
BN won only 140 (roughly 63%) which is slightly shy of
the supermajority of two-thirds. Although BN was
expected to win fewer seats, it was not expected to lose
its supermajority. It was not until the day of the General
Election, 8 March 2008 that the public and the stock mar-
ket realized that BN’s supermajority was lost based on exit
polls and early polling returns and sentiment (see
Anonymous, 2008b, 2008c).
5
This loss of BN’s supermajority led to many important
changes in the political landscape. For example, the former
Prime Minister, Dr. Mahathir Mohammad, quit UMNO (the
dominant party in BN) after the Election (Anonymous,
2008a), thus increasing uncertainties for BN. Ahmad
(2008) states, ‘‘[w]hile the BN continues to hold a simple
majority, a tectonic shift has taken place in Malaysian pol-
itics, and it was in many ways engineered by Anwar
Ibrahim. The victory means a new political future for the
world’s most economically advanced Muslim country, ush-
ering in newchapter in Muslim democracy’’. This change in
political landscape and the relatively signi?cant presence
of the opposition in the parliament was expected to curtail
cronyism, i.e., BN’s ability to have a free hand in pursuing
?nancial and economic policies that favor politically
connected ?rms (Bendeich, 2008). In addition, the
increased scrutiny by the opposition was also expected to
impede the ruling party’s ability to continue awarding
contracts, licenses and subsidies to politically connected
?rms.
Hypotheses development
Differential market reaction to bad news across old- and new-
politically connected ?rms
We posit that ?rms with political connections for a
short period of time are different from those with such
connections for a relatively long period. As shown by prior
research, politically connected ?rms directly bene?t from
government favors and subsidies (see discussion in the pre-
vious section). In return, these ?rms are likely to pursue
policies and agendas that help the political actors – the
government of?cials and politicians – strengthen their
social and political agenda. For example, in Malaysia based
on the New Economic Policy (NEP), promulgated in 1971,
the government implemented policies to improve the eco-
nomic status of the Malays including setting up trust agen-
cies such as the National Trading Corporation (Perbadanan
National Bhd or PERNAS) to actively participate in the cor-
porate sector. Policies were also put in place to encourage
companies to appoint more Bumiputras (ethnic Malays) to
senior management positions. As a result, several ?rms
with political connections appointed Malays, who were
ex-government of?cials or ex-politicians, to senior execu-
tive and board positions (Case, 2005; Gomez, 2002;
Gomez & Jomo, 1997).
6
Building credibility of adherence to such government
policies takes time and ?rms that build such credibility
are likely to gain access to the informal political elite
club/network, which in turn provides managers of such
?rms with prestige and ‘high standing’ in society
(DiMaggio & Powell, 1983; Donaldson, 2008; Luo, 2003;
Sanders & Tuschke, 2007; Scott, 2008). The political elite
status is likely to enable ?rms to obtain important eco-
nomic resources. For example, ?rms that gain the political
elite status are likely to attract and retain talented man-
agers; talented managers would have an incentive to join
the politically connected ?rm to gain access to the political
elite network. Thus, ?rms that are politically connected for
a longer period are likely to not only obtain direct eco-
nomic bene?ts from government favors and subsidies,
but also obtain indirect bene?ts of attracting and retaining
high quality human capital and other important economic
resources that will enable them to be self-sustaining, i.e.,
become less dependent on the political connection (see
Rona-Tas, 1994). In comparison to this, ?rms that are polit-
ically connected for a short period are likely to obtain only
the direct economic bene?ts.
7
We refer to the ?rms with
political connections over long and short periods of time
as old and new-politically connected ?rms, respectively.
5
Before the Election, the general view was that ‘[w]idespread voter
apathy and disillusionment with politics will ensure that the United-
Malays-National-Organisation (UMNO)-led ruling Barisan Nasional (BN)
coalition breezes through the elections with a comfortable two-thirds
majority in the 222-member Parliament’ (Anonymous, 2008d), and ‘[n]o
new faces mean that the old policy and governance will be continued’
(Anonymous, 2008e). As such, BN’s subsequent loss in the Election is
considered as a ‘political tsunami. . . [j]ust like the natural tsunami that
never gives any sign before it happens. . . No one expected such results,
included BN and the opposition as there was no conspicuous sign before the
elections’ (Anonymous, 2008c). A similar view has been cited by The
Economist (Anonymous, 2008f), political analysts (e.g., Klein, 2011) and
even the US Congressional Research Service Report (Martin, 2008).
6
Case (1995) writing more than 15 years earlier points out that it
‘remains an open question whether this freely developed political savvy
will ripen into true business viability’ (p. 95).
7
These arguments parallel the arguments embedded in the surviving
elite thesis espoused by Rona-Tas (1994) in the sociology literature.
S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86 69
When the political apparatus that is the basis for the
political connection is subject to adverse events such as
the loss of supermajority in the General Election, ?rms with
political connections are not likely to obtain direct bene?ts.
However, the old-politically connected ?rms are likely to be
impacted less by the adverse event than the new-politically
connected ?rms, because the old-politically connected
?rms are likely to have used their membership in the infor-
mal political elite club/network to obtain key economic
advantages and resources such as talented human capital
and become self-sustaining. This empirical expectation is
stated in the following hypothesis:
Hypothesis 1. The impact of bad news with respect to
political connection is stronger for the new-politically
connected ?rms than the old-politically connected ?rms.
The alternative to Hypothesis 1 arises from the argu-
ment that the old-politically connected ?rms are more
entrenched because of their reliance on cronyism for a
longer period of time. However, we do not expect support
for this alternative hypothesis because as an emerging
economy with increased globalization and scrutiny by
international agencies such as International Transparency
and foreign institutional investors, all ?rms would have
an incentive to be more ef?cient. Furthermore, the alterna-
tive hypothesis posits that during the initial period of
political connection, the politicians and government of?-
cials help to enhance the ?rms’ value, and during the later
periods they obtain rents and favors from these ?rms. This
long-term quid pro quo is not likely to be sustained
because the managers and political actors may not have
a long horizon.
Mechanism of differential market reaction to bad news across
old- and new-politically connected ?rms
Hypothesis 1 is based on the notion that one possible
mechanism through which the old-politically connected
?rms insulate themselves from the bad news of election
loss, is to have gained access to better talented managers/
board of directors.
8
If this is the case, it is not only new-po-
litically connected ?rms that suffer the loss in ?rm value
from the bad news associated with the ruling party’s elec-
tion loss, but also those old-politically connected ?rms that
exhibit less improvement in managerial professionalism
over their time of political connection. This argument of
examining one possible mechanism is summarized in the
following hypothesis:
Hypothesis 2. The impact of bad news with respect to
political connection is stronger for old-politically con-
nected ?rms with less improvement in board/ executive
professionalism than old-politically connected ?rms with
more improvement in board/ executive professionalism.
The life cycle viewposits that young ?rms are likely to be
more susceptible to the loss of favors frompolitical connec-
tions than old ?rms (Peng &Luo, 2000). While under the life
cycle view, all old-politically connected ?rms would exhibit
insulationfromthe direct bene?ts and costs of suchconnec-
tions; Hypothesis 2 predicts that only those old-politically
connected ?rms that have improved their board/executive
professionalism will be insulated. Thus, Hypothesis 2 can
be viewed as a test for the alternative explanation of
Hypothesis 1 under the life cycle hypothesis.
Empirical analyses
Classifying politically connected ?rms
To identify politically connected ?rms (PCONN), we
start with the list of PCONN ?rms provided by Johnson
and Mitton (2003) and Faccio (2006) (hereafter referred
to as the JM_F List). In addition to the JM_F List, several
new ?rms with political connections have emerged under
the patronage of the ‘new political establishment’ such as
Kamaludin Abdullah, the son of Abdullah Badawi
(Barrock, 2007). We use the news reports in a leading
Malaysian business newspaper, The Edge to identify ?rms
that obtained political patronage in recent years. In addi-
tion, the government set up ‘government linked compa-
nies’ (GLCs) under Khazanah, an investment arm of
UMNO, which we include as politically connected ?rms.
9
For each of these ?rms, we examine the 2007 annual report,
which immediately precedes the event of General Election,
and classify the ?rm as politically connected if they satisfy
any of the following criteria: (a) had government cabinet
members and/or MPs as members of the board, or (b) had
signi?cant ownership by government or UMNO-linked orga-
nizations/ individuals, or (c) had politically connected indi-
viduals as their managers.
To test our hypotheses, we need to classify ?rms that
have been politically connected for long and short periods.
For this purpose, we use the 1996/1997 annual reports of
the ?rms identi?ed as politically connected in 2007.
10
We
require the politically connected ?rms to exist for ten years,
i.e., have a 1996/1997 annual report as well, so as to ensure
that our classi?cation of new- and old-politically connected
?rms do not proxy for ?rm age. This criterion helps to miti-
gate the potential concern that our results are driven by
the life cycle hypothesis. The variable PCONN = 1 for ?rms
that are politically connected in 2007. A PCONN ?rm is
tagged as having been politically connected for a long period,
i.e., Old-PCONN = 1 (New-PCONN = 0) if the ?rmsatis?es any
of the three criteria for being classi?ed as politically con-
nected in 1997/1998 as well. For a PCONN ?rm that satis?es
any of the three criteria only in 2007 but not in 1997/1998, it
is considered as having been politically connected for a short
period, i.e., New-PCONN = 1 (Old-PCONN = 0).
8
Even though the argument for attracting more talented managers is
applicable also for the operating line managers, we posit the hypothesis in
terms of the board and top management with the notion that good
management practices at the top-level will be re?ective of more talented
managers at lower levels as well.
9
Examples include Malaysia Airport Holding Berhad, Malaysian Airline
System Berhad, Tenaga Nasional Berhad etc. See the website for Khazanah
Nasional Berhad:http://www.khazanah.com.my/portfolio.htm.
10
We choose 1997/1998 because it was the year after the Asian ?nancial
crisis so as to mitigate survival issues; and furthermore, the availability of
annual reports from earlier years is very spotty.
70 S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86
In total, we identify 122 politically connected ?rms – 66
are from the JM_F list; 56 are identi?ed from newspaper
reports. Out of the 122 PCONN ?rms, there are 39 New-
PCONN ?rms and 83 Old-PCONN ?rms.
11,12
Out of the 56
additional politically connected ?rms, 17 are Old-PCONN
?rms and 39 are New-PCONN ?rms. Appendix A provides
the list of PCONN ?rms classi?ed as Old- and New-PCONN
and the following details: (a) whether the ?rm belongs to
the JM_F list, (b) the criteria that the ?rm satis?es to be clas-
si?ed as politically connected, and (c) the age of the ?rm –
the age of the ?rm is obtained from the ?rm’s website or
the ?rm’s history in the library of Bursa Malaysia.
A premise embedded in Hypothesis 1 is that both the
old- and new-politically connected ?rms have equally
strong political ties; if the strength of political connections
is less for the old-politically connected ?rms than the new-
politically connected ?rms; or political connections are
non-existent for the old-politically connected ?rms so
much so that they are similar to non-politically connected
?rms, then also we will ?nd support for the hypothesis.
To address this, we examine news reports in the mid-
2000s to ensure that the Old-PCONN ?rms continued to
obtain favors at the time of our analyses (see Appendix B
for a few examples).
Even though the ruling party (BN) had lost supermajor-
ity in the General Election, it still maintained a majority.
Thus, we could ?nd support for Hypothesis 1 if the favors
are more likely to occur for the Old-PCONN ?rms than
New-PCONN ?rms i.e., the entrenched cronyism effect. To
mitigate this potential concern, we examine news reports
after the Election to verify whether both Old- and New-
PCONN ?rms face adverse consequences after the
Election (see Appendix C for a few examples).
13
Managerial (board/executive) professionalism
To test Hypothesis 2, we measure the professionalismof
the board of directors and the top-executives, i.e., CEO and
Chairman. For board professionalism, we consider the fol-
lowing governance-related variables: Indpc is the propor-
tion of independent directors as disclosed in the annual
reports, Bsize is the total number of board members,
DiverseBOD is the proportion of board members who are
non-Malays, and IndChair is one if the chairman is an inde-
pendent director. For executive professionalism we include
the following variables: CeoNMalay (ChairNMalay) is one if
the CEO (chairman) is a non-Malay, CeoMBA (ChairMBA) is
one if the CEO (chairman) has an MBA degree or equiva-
lent, CeoOverseas (ChairOverseas) is one if the CEO (chair-
man) is educated overseas, CeoOthDir (ChairOthDir) is one
if the CEO (chairman) is a director of another company,
CeoIndExp (ChairIndExp) is one if the CEO (chairman) has
been in managerial positions in the same ?rm or other
?rms in the same industry.
We obtain this data from the annual reports of the
PCONN ?rms for 1996/1997 and 2007. While most of these
factors are governance and executive expertise variables
used in prior studies, the diversity factor of including
non-Malays as an indicator of managerial professionalism
is novel to the Malaysian context. This is because while
non-Malays are likely to be a board member or top-execu-
tive because of their merit, it is possible that Malays are in
the position to ful?ll the social and political agenda.
However, it should be noted that if the labor market is
mature then the Malays who are chosen to these top posi-
tions are also likely to exhibit expertise similar to the non-
Malays as well as help the PCONN ?rm build credibility of
adherence to the political agenda.
We compute the change in each of the 14 characteris-
tics of Old- and New-PCONN ?rms from 1997 to 2007
and give a score of one for an increase in the value of each
of the variable from 1997 to 2007. We then add up the
scores of the 14 factors to construct a professionalism
score (Score).
Events leading up to the General Election
The results of the General Election were announced on
March 10, 2008. As discussed earlier, the loss of the super-
majority by the ruling party was a surprise to the public
and the stock market. Even though the media reports indi-
cate that the loss of supermajority by BN was a surprise, it
could be possible that some events leading up to the
General Election could have provided an advance warning
about this outcome to the stock market and the stock mar-
ket reacted to that news for the Old-PCONN ?rms at that
time. As such, we consider such events in our research
design so as to emphasize that the loss of supermajority
was indeed an exogenous shock.
We identify three fact-based events that changed the
tracking polls considerably. These events are summarized
in Appendix D. On November 10, 2007 there was a mass
demonstration against the government held in Kuala
Lumpur (Event1). This demonstration was spear-headed
by Anwar Ibrahim, the former Deputy Prime Minister,
who was jailed for four years for various corruptioncharges.
By some estimates there were 40,000 people involved in the
demonstration and 15 demonstrators were arrested by
police. The second demonstration (Event2) of roughly
15,000 mostly ethnic Indian protestors was held on
November 25, 2007 and organized by the Hindu Rights
Action Front (HINDRAF). HINDRAF protested the marginal-
ization of ethnic Indians in Malaysia. The third event on
February 13, 2008 (Event3) was the announcement of the
General Election and the dissolution of the Parliament
(Anonymous, 2008d). These three events, i.e., Event1-
Event3 were unfavorable to the ruling party BN and are
11
We identify 17 GLCs out of which 7 are in the JM_F list and also ful?ll
the criteria of Old-PCONN.
12
Even though our procedure for identifying PCONN ?rms is extensive, it
is possible that some politically connected ?rms have not been identi?ed as
PCONN. In effect, while our procedure ensures that all PCONN ?rms are
indeed politically connected, we may have tagged some politically
connected ?rms as non-PCONN. This measurement error would bias
against ?nding the predicted negative stock market reaction for both the
New- and Old-PCONN ?rms.
13
Note that even if we gather data on all of such reported adverse events,
it will not provide the complete picture of adverse consequences. This is
because not all adverse events are newsworthy and hence many adverse
events may not be visible or transparent. As such, we search and obtain a
couple of examples to highlight the fact that the loss of supermajority by
the ruling party was indeed bad news for the PCONN ?rms.
S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86 71
likely to be indicative of the potential for loss of supermajor-
ity. While ex post, i.e., after Event4, Events1-Event3 can be
classi?ed as ‘‘bad news’’ events for BN, ex ante whether it
was as such is not clear. Thus, we do not have predictions
on the market’s reaction to Events1–3.
Sample and data
The sample contains 834 non-politically connected
?rms (PCONN = 0) listed on the Bursa Malaysia (Kuala
Lumpur Stock Exchange) with available data and 122 polit-
ically connected ?rms (PCONN = 1). Daily stock closing
prices and the market index, Kuala Lumpur Stock
Exchange Composite Index (KLCI) are obtained from
Datastream.
14
Stock returns for each company are calcu-
lated using the percentage change in daily closing prices,
adjusted for dividends, stock splits and rights issues (Fama,
Fisher, Jensen, & Roll, 1969). Market returns are calculated
based on the KLCI returns. We measure cumulative abnor-
mal returns (CARs) around the announcement date of the
General Election results (March 10, 2008) by subtracting
the KLCI’s value-weighted daily return.
15
Research design
To test Hypothesis 1, we estimate the following model
over a six-month period (121 days) from October 1, 2007
to March 31, 2008 (see Binder, 1985; Millon-Cornett &
Tehranian, 1989, 1990; Schipper & Thompson, 1983,
1985; Smith, Bradley, & Jarrell, 1986):
ðR
i
ÀR
f
Þ ¼ b
0
þb
1
Event1 þb
2
Event2 þb
3
Event3
þb
4
Event4 þb
5
PCONN þb
6
New-PCONN
þb
7
PCONNÂEvent1 þb
8
PCONNÂEvent2
þb
9
PCONNÂEvent3 þb
10
PCONNÂEvent4
þb
11
New-PCONNÂEvent1
þb
12
New-PCONNÂEvent2
þb
13
New-PCONNÂEvent3
þb
14
New-PCONNÂEvent4
þb
15
ðR
m
ÀR
f
Þ þb
16
logðSizeÞ þb
17
logðMBÞ
þb
18
logðLeverageÞ þb
19
logðFirmAgeÞ
þIndustry-fixed effects þerror ð1Þ
where R
i
is the daily return of ?rm i; R
f
is the daily return
on the Malaysian treasury bill; Event1–4 are the dummy
variables that take on a value of 1 for the three-day
window surrounding the four events leading up to the
General Election (see Appendix D)
16
; PCONN is one if
the ?rm is a politically connected ?rm in 2007; and
New-PCONN is one if the PCONN ?rm is not connected in
1997–1998 but is connected in 2007 (see Appendix B); R
m
is the daily return on KLCI; log(Size) is the log of the market
capitalization of the ?rm in the prior year, log(MB) is the log
of the market-to-book ratio in the prior year, and
log(Leverage) is the log of the long-term-debt to assets ratio
in the prior year. Since New-PCONN ?rms could be system-
atically younger ?rms than Old-PCONN ?rms, we include
log(FirmAge) which is the log of the ?rm age obtained from
the ?rms history to control for the potential age effects on
our results.
17
The standard errors of the coef?cient estimates can be
biased downwards because by design the ?rms are
repeated daily in the sample. In addition, macro events
such as the election results are also likely to be cross-sec-
tionally correlated. We use the robust standard errors clus-
tered by ?rm and date to mitigate the potential bias in
standard errors (see Gow, Ormazabal, & Taylor, 2010;
Petersen, 2009).
18
Based on earlier studies (e.g. Fisman, 2001), we expect
the coef?cient estimate on PCONN Â Event4, i.e., b
10
to be
negative, and based on Hypothesis 1 we expect the coef?-
cient estimate on New-PCONN Â Event4, i.e., b
14
to be
negative.
To test Hypothesis 2, we augment Eq. (1) by including
the change in managerial professionalism score.
Empirical results
Market reaction to the General Election results for Old- and
New-PCONN ?rms
Table 1, Panel A provides the mean cumulative abnor-
mal returns (CARs) for the PCONN ?rms (N = 122) and
non-PCONN ?rms (N = 834), as well as for the Old-
PCONN (N = 83) and New-PCONN (N = 39) ?rms, over dif-
ferent windows surrounding the General Election result
announcement date, i.e., March 10, 2008. While the mean
CAR for the event windows (À1, +1), (À1, +3) and (À1,
+5) are not signi?cantly different from zero for the non-
PCONN ?rms (e.g., t-statistic = 0.63 for the window À1,
+1), the mean CAR of PCONN ?rms for the event windows
(À1, +1), (À1, +3) and (À1, +5) are À3.12%, À4.32% and
À7.17% with corresponding t-statistics of À4.37, À5.65
and À7.36, respectively. This shows that the difference in
CAR across PCONN and non-PCONN, i.e., PCONN minus
non-PCONN is negative and statistically signi?cant, consis-
tent with ?ndings in prior studies (e.g., Fisman, 2001) that
political connections are positively associated with ?rm
value.
The columns on the right-hand side examine the differ-
ential market reaction for Old- and New-PCONN ?rms
based on Hypothesis 1. We ?nd that the difference in
Old-PCONN’s mean CAR and New-PCONN’s mean CAR for
the event windows (À1, +1), (À1, +3) and (À1, +5) are
14
For a few ?rms (less than one percent of our sample) with missing stock
price data in Datastream, we collect the data from Yahoo! Finance.
15
We also compute expected returns based on (1) the market model
estimated from March 1, 2007 to February 29, 2008; as well as (2) the ?rm-
speci?c average returns from March 1, 2007 to February 29, 2008 and we
obtain similar results. This is consistent with Brown and Warner’s (1985)
simulation ?ndings that results in short term event analyses are not
sensitive to the use of different expected return estimation methods.
16
We use a combined regression for the four events because the larger
panel data enhances statistical power. In essence, we give Events1–3 a
‘‘good’’ chance to be statistically signi?cant.
17
In unreported analysis we control for prior pro?tability (ROA), number
of segments and proportion of foreign sales and obtain similar results.
18
We cluster by month instead of day and cluster by ?rm alone and
obtain similar results.
72 S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86
À3.17%, À2.80% and À4.31%, with corresponding t-statis-
tics of À2.10, À1.72 and À2.09, respectively. This shows
initial support for Hypothesis 1. Furthermore, for the event
windows (À1, +1), (À1, +3) and (À1, +5) the mean CAR for
Old-PCONN (New-PCONN) are À2.11% (À5.28%), À3.43%
(À6.22%) and À5.80% (À10.11%) respectively, showing that
Table 1
Descriptive statistics and univariate tests of market reaction to election results.
Non-PCONN PCONN Diff Old-PCONN New-PCONN Diff
Panel A: Univariate tests of market reaction to election results
EvtCAR (À1, +1) 0.52% À3.12% À3.65% À2.11% À5.28% À3.17%
t-Stat 0.63 À4.37 3.33 À2.36 À4.79 À2.10
p-Value (0.53) (0.00) (0.00) (0.02) (0.00) (0.04)
EvtCAR (À1, +3) 0.85% À4.32% À5.17% À3.43% À6.22% À2.80%
t-Stat 0.99 À5.65 4.51 À3.78 À4.50 À1.72
p-Value (0.32) (0.00) (0.00) (0.00) (0.00) (0.09)
EvtCAR (À1, +5) 0.79% À7.17% À7.96% À5.80% À10.11% À4.31%
t-Stat 0.88 À7.36 6.00 À5.17 À5.53 À2.09
p-Value (0.38) (0.00) (0.00) (0.00) (0.00) (0.04)
Variable Non-PCONN (N = 834) PCONN (N = 122) Tests of
Mean Median Mean Median Means Medians
Panel B: Univariate tests of ?rm characteristics between PCONN and non-PCONN ?rms
Size 534.70 98.15 2808.07 849.97 3.10 (0.00) 10.3 (0.00)
MB 1.17 0.88 2.08 1.51 2.72 (0.01) 5.41 (0.00)
Leverage 0.10 0.10 0.12 0.11 0.64 (0.52) 0.64 (0.52)
FirmAge 22.98 18.00 33.82 32.00 5.83 (0.00) 6.66 (0.00)
ROA (%) 5.08 5.23 6.41 5.29 0.63 (0.53) 0.26 (0.79)
No. of segments 1.37 1.00 1.42 1.00 0.27 (0.79) 1.06 (0.29)
% Foreign sales 10.23 0.00 10.19 0.00 0.02 (0.98) 1.60 (0.11)
Cash?ow per share 0.12 0.03 0.30 0.11 3.56 (0.00) 3.57 (0.00)
Sales volatility 0.24 0.19 0.24 0.18 0.29 (0.77) 0.65 (0.52)
Sales growth 0.94 0.19 0.43 0.17 1.01 (0.31) 0.79 (0.43)
Variable Old-PCONN (N = 83) New-PCONN (N = 39) Tests of
Mean Median Mean Median Means Medians
Panel C: Univariate tests of ?rm characteristics between Old-PCONN and New-PCONN ?rms
Size 2775.55 737.74 2877.28 992.55 0.05 (0.96) 0.76 (0.45)
MB 2.43 1.62 1.35 1.37 2.28 (0.03) 1.35 (0.18)
Leverage 0.12 0.14 0.12 0.04 0.02 (0.98) 0.63 (0.53)
FirmAge 35.89 35.00 29.41 24.00 1.74 (0.08) 2.58 (0.01)
ROA (%) 9.06 5.56 0.78 4.39 2.40 (0.02) 1.56 (0.12)
No. of segments 1.45 1.00 1.36 1.00 0.30 (0.77) 0.32 (0.75)
% Foreign sales 10.75 0.00 9.00 0.00 0.42 (0.67) 0.59 (0.55)
Cash?ow per share 0.29 0.11 0.33 0.13 0.29 (0.77) 0.34 (0.73)
Sales volatility 0.25 0.19 0.23 0.17 0.66 (0.51) 0.29 (0.77)
Sales growth 0.50 0.16 0.28 0.18 0.88 (0.38) 0.34 (0.73)
Notes
1. In Panel A, the sample contains 956 ?rms. The columns ‘Non-PCONN’ and ‘PCONN’ represent ?rms with no political connections (N = 834) and ?rms
with political connections (N = 122) respectively. In addition, the column ‘New-PCONN’ includes ?rms with political connections for a short-period
(N = 39), and column ‘Old-PCONN’ includes ?rms with political connections for a long-period (N = 83).
2. PCONN is an indicator variable that equals one for politically connected ?rms, and zero otherwise. A ?rm is politically connected if any of the following
three criterion is satis?ed in 2007: either (a) had government cabinet members and/or MPs as members of the board, or (b) had signi?cant ownership
by government or UMNO-linked organizations/individuals, or (c) had politically connected individuals as their managers. They are identi?ed based on
(1) the ?rm list provided in Johnson and Mitton (2003) and/ or Faccio (2006), and (2) the search for recent news reports and articles. A PCONN ?rm is
tagged as having been politically connected for a long period, i.e., Old-PCONN = 1 (New-PCONN = 0) if the ?rm satis?es the three criteria for being clas-
si?ed as politically connected in 1997/1998 as well, and is considered as having been politically connected for a short period, i.e., New-PCONN = 1 (Old-
PCONN = 0) if the PCONN ?rm satis?es the three criteria only in 2007 but not 1997/1998. See Appendix A for identi?cation strategy for both Old-PCONN
and New-PCONN ?rms.
3. EvtCAR is the cumulative abnormal returns for the different windows centered around the General Election results announcement date of March 10,
2008 (Day 0).
4. Panel B (Panel C) tests the mean and median differences between Non-PCONN and PCONN (between Old-PCONN and New-PCONN) on a range of ?rm
characteristics: Size is the market capitalization, MB is the market to book ratio, Leverage is the total debts to total assets, FirmAge is the age of the ?rm
since its incorporation, ROA is the returns on assets, No. of Segments is the number of geographical segments, % Foreign Sales is the percentage of for-
eign sales to total sales, Cash?ow per share is the total cash ?ow divided by the number of outstanding shares, Sales volatility is the standard deviation
of sales over the past three years, and Sales growth is the percentage change in sales from year t À 1 to year t. t-values and p-values are based on two-
tailed tests and are reported in italics (p-values are in parentheses).
S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86 73
the stock market reaction to both Old- and New-PCONN
?rms are signi?cantly negative, which provides some fur-
ther con?dence that Old-PCONN ?rms are politically
connected.
Table 1, Panels B and C provide the means and medians
of ?rm characteristics. Panel C shows that Old-PCONN and
New-PCONN ?rms are similar on most dimensions of ?rm
characteristics and risk pro?les, except that Old-PCONN
?rms have higher market to book ratios and pro?tability
on average, and are older than New-PCONN ?rms. For this
reason we control ?rm age in our estimations to ensure
that our results are not driven by the difference in ?rm
age between the two groups.
19
Table 2 presents the results of estimating Eq. (1). In the
left set of columns we estimate Eq. (1) without the New-
PCONN variables to provide a benchmark. In estimating
Eq. (1) presented in the right set of columns, we ?nd that
the coef?cient estimate on the interaction between
PCONN Â Event4 is À0.007 (t-stat = À11.350) and the
interaction between New-PCONN Â Event4 is À0.023 (t-
stat = À32.7). The negative coef?cient of PCONN Â Event4
is consistent with prior studies in political connection
(see Fisman, 2001). The negative coef?cient on New-
PCONN Â Event4 supports Hypothesis 1 and shows that
New-PCONN ?rms are more adversely affected by the rul-
ing party’s loss of supermajority in the General Election. In
terms of economic signi?cance, the New-PCONN ?rms are
roughly 3.29 [=0.023/0.007] times more adversely affected
than Old-PCONN ?rms.
The market reaction is more negative for PCONN ?rms
than Non-PCONN ?rms for Event3 and Event4 (see
coef?cients PCONN Â Event3 and PCONN Â Event4); and,
interestingly, although these preceding events may be
indicative of the loss in the popularity of the ruling party,
we ?nd that the market reacts more positively for PCONN
?rms than non-PCONN ?rms for Event1 and Event 2 (see
coef?cients PCONN Â Event1 and PCONN Â Event2). This
could be because while the event of dissolution of the
Parliament (Event3) and the result of the Election
(Event4) are both related directly to the Election, the mass
demonstrations against the ruling party captured by Event1
and Event2 are indicative of the power of the ruling party.
Further tests controlling for ?rm age of Old- and New-PCONN
?rms
The life cycle hypothesis predicts that young ?rms are
likely to be more susceptible to the loss of favors from
political connections than the old ?rms (Peng & Luo,
2000), and it is possible that our measure of old- and the
new-politically connected ?rms are simply a proxy for ?rm
age and young ?rms captured by our measure of new-po-
litically connected ?rms are less resistant to adverse
shocks. Even though we control for FirmAge in Eq. (1), this
may not be suf?cient as it only controls for the average
effect of ?rm age. In Table 3, we conduct a number of tests
to further ensure that our results are not driven by the dif-
ference in age of the Old- and New-PCONN ?rms. This is
important to rule out the alternative explanation of the life
cycle hypothesis.
First, we consider PCONN ?rms below median age as
young PCONN ?rms; accordingly, Young-PCONN is an indi-
cator variable that is one if the ?rm-age is below the med-
ian ?rm age, and 0 otherwise. We include Young-PCONN as
an additional variable in Eq. (1) with the corresponding
interactions with the events. If New-PCONN is correlated
with young PCONN ?rms, then the results documented in
Table 2 is likely to be driven by Young-PCONN. The result
of this estimation is provided in the leftmost column of
Table 3. The result shows that while Young-PCONN Â
Event4 is signi?cantly negative, New-PCONN Â Event4 is
also negative and signi?cant. In particular, the coef?cient
estimate (À0.021) is similar to the results reported in
Table 2 (À0.023), which indicates that the results in
Table 2 are not driven by young ?rms.
Second, if our results are driven by younger ?rms being
more prone to economic shocks, then we should not ?nd
results when we exclude young PCONN ?rms in our sam-
ple. To examine this possibility, we exclude all PCONN
?rms (both Old- and New-PCONN ?rms) with below med-
ian ?rm age (less than 32 years). With this smaller sample
we re-estimate Eq. (1); the results are reported in the mid-
dle set of columns in Table 3 and show that the results are
similar to those discussed in Table 2.
Third, we further control for the age difference between
Old- and New-PCONN ?rms by matching each New-
PCONN ?rm to an Old-PCONN ?rm based on ?rm age and
total assets. We match each New-PCONN ?rm to an
Old-PCONN ?rm because there are fewer New- than
Old-CONN ?rm. The results are reported in the rightmost
set of columns of Table 3 and are also similar to those
discussed in Table 2.
20
Overall, these results suggest that our ?ndings reported
in Table 2 are not likely to be driven by difference in ?rm
age across Old- and New-PCONN ?rms, i.e., New-PCONN
?rms are younger and thus more sensitive to economic
shocks.
Change in board/executive professionalism across Old- and
New-PCONN ?rms
Table 4, Panel A shows the increase/ improvements in
the 14 variables of board/ executive professionalism across
Old- and New-PCONN ?rms from 1997 to 2007. The Score
variable in the last row is the total of the increase in the 14
components. The mean Score of Old-PCONN is 1.831
compared to 0.821 of New-PCONN; this suggests that the
Old-PCONN ?rms make improvements on roughly two of
the dimensions while the New-PCONN ?rms make
19
Note that since we employ a short window event study methodology,
such characteristics should not affect the results. Nevertheless, in a
sensitivity analysis, we repeat our analyses by controlling for these
characteristics and obtain similar results.
20
In unreported robustness tests we ?nd similar results when we re-
estimate Eq. (1) by doing each of the following: (1) exclude non-PCONN
?rms with below median ?rm age, (2) exclude only New-PCONN ?rms with
below median ?rm age, (3) use the ?rm age cutoff of 15 years to de?ne
young ?rms instead of the median age. We do not consider the matched
sample for our primary research design, so as to preserve a reasonable
sample size for the test of Hypothesis 2.
74 S.Y.K. Fung et al. / Accounting, Organizations and Society 43 (2015) 67–86
improvements on one dimension.
21
Roughly 30%, 40% and
50% of the Old-PCONN ?rms show improvements in board
independence, board size and diversity, respectively; com-
pared to only roughly 10–20% of New-PCONN ?rms.
Roughly 10% Old-PCONN ?rms change from Malay CEO to
non-Malay CEO, while only 3% of New-PCONN ?rms exhibit
this change. Also, we ?nd that 7% of the Old-PCONN ?rms
have improvements in the industry expertise of the
Chairman but such improvement is not observed for New-
PCONN ?rms. All the other changes are not different
between the two groups. Overall, these results suggest that
compared to the New-PCONN ?rms, board professionalism
of the Old-PCONN ?rms improved over time. Overall, while
our professionalism measure captures both board (gover-
nance) characteristics and senior executive characteristics,
empirically we observe that the differences in professional-
ism between Old- and New-PCONN ?rms are driven by the
board (governance-related) professionalism. This is likely
because the board-related characteristics are proportions
and thus a continuous variable; while the CEO/Chair related
variables are indicator variables. Thus, the change in the
indicator variables may not be powerful enough to capture
improvement when compared to the board-related vari-
ables. We continue our analysis with the caveat of an under-
lying assumption in our analysis is that improvements in
board characteristics percolate to the managerial cadre of
the ?rm as well. Overall, our results on changes in board/
executive professionalism are attributable to changes in
board professionalism.
Table 4, Panel B compares the proportion of Old-PCONN
(New-PCONN) ?rms that exhibit an increase in the Score
measure above the mean in the sample over the ten-year
time period, i.e., 1997 to 2007. We use the mean of the
Score among the Old-PCONN to classify the Old- and
New-PCONN ?rms as High and Low Score. We ?nd that
slightly more than half (45 out of 83 ?rms) of the Old-
PCONN ?rms are classi?ed as High Score. However, less
than one-third (10 out of 39 ?rms) of the New-PCONN
?rms are classi?ed as High Score, and the difference in
Table 2
Multivariate tests of market reaction to election results.
Dependent variable = R
i
– R
f
Eq. (1) without New-PCONN Eq. (1)
Coef t-stat p-value Coef t-stat p-value
Intercept 0.002 1.110 0.269 0.002 1.110 0.268
Event1 0.001 1.030 0.304 0.001 1.030 0.304
Event2 À0.010 À7.980