Board Connections and Crisis Performance Family ST

Journal of Corporate Finance 64 (2020) 101630
Contents lists available at ScienceDirect
Journal of Corporate Finance

journal homepage: www.elsevier.com/locate/jcorpfin
Board connections and crisis performance: Family, state, and

T
political networks
Richard W. Carneya, , Travers Barclay Childa, Xiang Lib
⁎
a
China Europe International Business School, China
b
Boston College, USA
ARTICLE INFO ABSTRACT
JEL classification: We introduce a novel concept of network interactions in which board connections provide access
G3 to external spheres of political influence, state ownership, and family control. We posit this form
G14 of indirect access via board association enables connected firms to benefit from information privy
L14 to external networks while avoiding their resource-based costs of membership. Board network
data are assembled for 1290 East Asian firms and linked to hand-collected data on political
Keywords:
connections and corporate ownership around the 2008–09 crisis. Companies with board con-
Board networks
nections to state-owned firms and family business groups had greater crisis-period accounting
Firm performance
Political connections performance and stock returns. In countries with weak institutional development, board con-
Corporate ownership nections to politically connected firms were also beneficial.
Board networks have been demonstrated to boost firm performance during periods of economic stability and distress (Larcker
et al., 2013; Dass, Kini, Nanda, Onal, and Wang 2014). Leading explanations for this effect emphasize the role of board networks as
conduits of information flow. From this perspective, firms with many board connections are focal points of information within the
economy (Freeman, 1978). But what kind of information is transmitted through board networks to enhance firm performance? And
from which types of firms is this information likely to emanate?
Scholars have separately examined firm networks characterized by family control, state ownership, and political ties. A rich
literature has examined the benefits and costs of inclusion in these networks.1 But indirect association with them remains unexplored.
We postulate that interlocking directorates may constitute an important form of indirect association with family business groups,
SOEs, and politically connected firms. Board networks may facilitate access to the privileged information otherwise reserved for
group members. But importantly, the costs of family control, state ownership, and political ties are typically resource-based (e.g.
tunneling, pursuit of policy objectives, and campaign finance, respectively). So while board networks may transmit information-
based benefits of association with such groups, this form of indirect access simultaneously shields a firm from the concomitant
resource-based costs. This perspective may enhance our understanding of the established relation between board networks and
performance.
Our unique data allow us to identify firms with board interlocks to companies characterized by family control, state ownership,
and political ties. Throughout this paper, we refer to these novel types of network interactions as family networks, state networks, and
political networks, respectively. Crucially, firms with these types of board networks need not themselves be characterized by family/
Corresponding author.
⁎
E-mail address: carney.richard@ceibs.edu (R.W. Carney).

1
For family business groups, see Khanna and Palepu, 2002; Khanna and Yafeh, 2007. For state ownership, see Megginson and Netter, 2001;
Aivazian et al., 2005; Chen et al., 2009. For political connections, see Faccio, 2006; Goldman et al., 2009; Bertrand et al., 2018.
https://doi.org/10.1016/j.jcorpfin.2020.101630
Received 19 April 2018; Received in revised form 14 April 2020
Available online 22 April 2020
0929-1199/ © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/BY/4.0/).
R.W. Carney, et al. Journal of Corporate Finance 64 (2020) 101630
state ownership, nor by political ties. In fact, our identification relies on our ability to net out the direct benefits and costs of
membership in such groups, thereby exposing the indirect benefits of board association.
Our empirical analysis demonstrates firms with strong board networks have significantly greater accounting and stock perfor-
mance during the global financial crisis. By decomposing board networks into the above subtypes of network interactions, we
demonstrate the performance benefits of board networks are concentrated primarily in those providing access to family business
groups and SOEs. This finding adds nuance to a body of evidence on the effects of board interlocks implicitly regarded as homo-
geneous (see, for example, Larcker et al., 2013). Given well-cited research on the value of political connections (e.g. Fisman, 2001;
Faccio, 2006), it is surprising our political networks do not generally boost crisis performance. However, in economies with weak
institutions (as measured by GDP per capita, financial development, or investor protections), we do find political networks to improve
outcomes.
Our research design resembles Lins et al. (2017) which examines the impact of social capital on firm performance during a crisis.
Our identification strategy can therefore be expressed in nearly equivalent terms: We exploit the global financial crisis as a shock to
the market equilibrium, while board networks remain fixed in the short-run. The crisis gives rise to environmental conditions in
which the information accessed by board networks (and by family networks, state networks, and political networks) becomes po-
tentially more valuable. We thus observe whether these various types of network interactions boost accounting and financial per-
formance in this setting. Like Lins et al. (2017) we do not exploit a shock to the networks, and are consequently unable to identify
their direct impact on performance outcomes during regular periods of economic activity.
Because the crisis was unexpected, selection concerns are mitigated. Board networks are unlikely to be ex-ante selected on ex-post
crisis performance. Because networks are likely to be selected on contemporaneous performance, we invoke a dynamic model netting
out that source of selection bias and isolating crisis-contingent effects. Still, it is possible networks are selected on correlates of
performance-potential during the crisis. To alleviate such concerns we include a host of additional controls, and conduct a number of
robustness checks.
To understand the mechanism underlying our main effects, we explore three potential channels. First, we check whether networks
enable firms to maintain or boost sales, and find no evidence in support of this channel. Second, we test whether networks enable
firms to secure trade credit, and find family and political networks do lead to higher crisis-period trade credit. Third, we examine a
debt-related channel. We allow the performance effects of networks to vary between firms with below/above-median levels of
maturing debt (relative to assets). Our evidence suggests firms with high levels of maturing debt benefit from all types of board
networks during the crisis. Additional tests suggest, however, those benefits do not include direct reductions in the interest cost of
debt.
To carry out this project we build a large multi-country dataset mapping each firm's board networks, ultimate ownership
structure, and political ties. To identify the importance of each type of network interaction, we must focus on a region in which they
are known to be highly salient. We therefore choose East Asia as the setting for our study (Hamilton, 1996). Our sample consists of
1290 of the largest publicly traded firms across nine East Asian economies: Hong Kong, Indonesia, Japan, South Korea, Malaysia, the
Philippines, Singapore, Taiwan, and Thailand. To measure director networks, we use hand-collected data on approximately 29,000
directors. To determine whether firms are politically connected, we cross-reference our directors list with all politicians holding
national-level executive or legislative positions in the region. Finally, we use manually collected ultimate ownership data to identify
family and state-owned firms.
Our paper makes two broad contributions to the literature. First, we contribute significantly to existing research on board net-
works. Earlier research has highlighted the role of director/executive networks in transmitting information (e.g. Cohen et al., 2008;
Bizjak et al., 2009; Larcker et al., 2013; Faleye et al., 2014; Jiang et al., 2017). We extend this literature by theorizing which types of
board networks tap into valuable information repositories during a financial crisis. In related work, Dass et al. (2014) suggest
directors from related industries may confer important information in times of distress. We further establish the importance of board
network heterogeneity by highlighting the relevance of director interlocks with family, state, and politically connected firms. To
substantiate our theoretical conjectures, we present robust empirical evidence for heterogeneous effects of board connections, de-
pending on the external networks to which they grant access.
Our second contribution applies to the corporate finance literature more broadly. This paper constitutes one step towards con-
necting heretofore distinct branches of the literature examining different types of firm networks. To our knowledge, we assemble the
first dataset permitting joint exploration of board connections, political ties, and ownership structures. These data enable us to adopt
a holistic approach to the study of firm networks, with a special emphasis on their joint interaction. This innovation implies our
findings speak simultaneously to the separate literatures on family business groups, state ownership, and political ties. Existing bodies
of research have previously documented the benefits and costs related to each of the abovementioned phenomena (see Megginson
and Netter, 2001; Khanna and Yafeh, 2007; Goldman et al., 2009). Our paper contributes to these lines of inquiry by introducing
network interactions via board associations, and documenting the related benefits.
The remainder of this paper is structured as follows. Section 1 develops theory on the performance implications of family net-
works, state networks, and political networks. Section 2 describes data collection and variable construction. In Section 3 we examine
the impact of networks on accounting performance and stock returns during the crisis. Section 4 extends our main analysis by
exploring underlying mechanisms. Section 5 strengthens the causal interpretation of our findings by examining: director qualifica-
tions as a source of confound; crisis intensity as a moderating factor; and crisis-induced changes to networks. Section 6 concludes, and
an online appendix offers further technical robustness checks.
2
1. Theory of network interactions
1.1. The importance of networks to Asian business
The importance of business networks as an institutional medium by which East Asian economies are organized is well docu-
mented (Redding, 1996; Dacin and Delios, 2005; Carney, 2005; Greif and Tabellini, 2010). Professional, family, state, and political
networks are important to the organization of business throughout the region. However, countries are often characterized as pri-
vileging one type of network over others due to historical circumstances. For example, Chinese networks are commonly regarded as
being dominated by kinship ties, with business relations based on networking, guanxi (i.e. social connections). Chinese networks are
important not only to societies dominated by ethnically Chinese citizens (Taiwan, Singapore, and Hong Kong), but also to Southeast
Asian economies where there is a large Chinese diaspora (Redding, 1993).
Southeast Asian networks are usually considered to be kinship-based and/or organized around patron-client political relation-
ships. In Thailand, for example, family and kinship are the most important form of social bond in business networks (Suehiro and
Wailerdsak, 2014), and they form the core around which the country's dominant business groups (glum thurakit) have been organized
(Suehiro, 1989). In the Philippines, family forms the backbone of society with large and dense family networks common to family-
owned conglomerates (Wolters, 1999; Kondo, 2014).
In Indonesia and Malaysia, patron-client relations are regarded as having had a significant impact on the organization of business
networks. In a weak institutional environment like Indonesia, close personal relationships with powerful political patrons could
reduce the uncertainties involved in carrying on a business, and also grant access to public sector contracts (Carney et al., 2008;
Turner, 2007; Rademakers, 1998). In Malaysia, outside of ethnic Chinese businesses, inter-company networks are commonly state-led
and state-mediated (Gomez and Jomo, 1999). Public investment comprises a large fraction of total investment, so the reliance of
private actors on public investment proliferates patron-client business networks.
Inter-corporate ties that cement together a vast community of firms are widely viewed as dominating the structure of Japanese
business networks. These groups, known as keiretsu, have dominated the structure of the post-war Japanese economy although their
cohesiveness has varied across groups and time (Orru, 1996; Lincoln and Shimotani, 2010). An important characteristic of these
groups is the sharing of information, leading some scholars to label them “information clubs” (Imai, 1988). South Korean business
networks are regarded as being heavily influenced by elite business families that prospered through privileges granted by a strong
state. Japan's pre-World War II zaibatsu business groups have often been compared to the Korean chaebol of today (Lee and Shim,
2012). Both chaebol and zaibatsu are terms represented by the same Chinese characters and denote large family owned and centrally
controlled business networks. However, the chaebol are distinguished by their relations with the Korean state which has granted
these large groups special privileges (Hamilton, 1996).
1.2. Board networks and information
Generally speaking, funds flow from firms with a capital surplus to those requiring finance for productive investments. An
important barrier to the efficient allocation of capital is information frictions. When a financial crisis strikes, such frictions are
exacerbated (Mishkin, 1996, 1997; Mishkin and Hahm, 2000). A major business failure increases uncertainty and impairs the ability
of lenders to gauge the creditworthiness of borrowers.2 The initiation or expansion of business opportunities can also be affected. In a
crisis, it becomes harder to distinguish those with productive investment opportunities and healthy balance sheets from those without
(Morellec and Schurhoff, 2011; Lambert et al., 2012). Taken together, this lack of credit and reluctance to initiate new ventures or
modify existing contracts can magnify the negative performance implications of a crisis.
Existing research suggests director/executive networks might attenuate the above problems and thereby reduce the severity of a
crisis for individual firms by facilitating information flows (Bizjak et al., 2009; Larcker et al., 2013; Faleye et al., 2014; Jiang et al.,
2017). For example, Dass et al. (2014) demonstrate that board networks with directors from related industries can potentially bring
valuable information and access to professional contacts. Valuable information may include industry conditions and trends, thereby
facilitating better management of a firm's factors of production and protecting it against demand or supply shocks. Indeed, Dass et al.
(2014) find the effect of such directors on firm value is stronger when the information gap is larger. Given that information gaps
constitute a core problem of financial crises, we expect director networks to confer similar benefits in such a circumstance.
Through their capacity to reduce information asymmetries, board networks may grant firms better access to credit (Davis, 1991;
Stuart and Yim, 2010). Stearns and Mizruchi (1993), for example, demonstrate the types of financial institutions represented on a
firm's board affect its sources of financing. Highly connected boards may also have better information about business partners capable
of modifying terms of contracts, such as allowing for orders and payments to be temporarily altered (Uzzi, 1999). Knowing which
firms face a greater decline in the demand for their products/services, and which firms are less likely to make payments on time
allows the well-connected firm to make calibrated adjustments to its own financial and business operations (Dass et al., 2014).
In light of the above, it is natural to expect that the value of director interlocks is determined by the types of firms with which the
network enables exchange. Existing studies identify performance benefits due to director types satisfying specific environmental
needs, such as deregulation, young firms, or firms pursuing diversification strategies (Hillman et al., 2000; Kroll et al., 2007; Jones
2
Screening is more difficult under these conditions because prior methods for assessing borrower risk have proven problematic, and because of the
heightened risk of contagion.
3
et al., 2008). We therefore consider various measures of board centrality, each differing in terms of which external network is reached
via the director interlocks. In particular, we study director ties to politically connected firms, state-owned companies, and family
business groups. We refer to these subcategories of board networks as political networks, state networks, and family networks,
respectively.
1.3. Crisis performance and network interactions
A key benefit of board networks in comparison to networks linked via ownership regards the relative absence of an obligation to
transfer resources to other firms. For example, a family- or state-owned firm that controls another firm may possess the capacity to
divert resources from it (Hillman et al., 2009). During a crisis, these ownership costs are likely to rise as a parent firm with ownership
and control rights will exercise its power to divert resources from a subject firm in order to shore up its own financial position.3
However, firms linked exclusively via board networks enjoy autonomy to act on information received without concomitant ob-
ligations. In this context, firms can exploit their information channels to lessen uncertainty related to their economic environment
(Hillman et al., 2000). As uncertainty in the economic environment rises, the value of information also rises, thereby amplifying the
benefits provided via board networks.
1.3.1. Family networks

A family business group is a collection of legally independent firms under the ownership of a single family. A distinctive feature of
such groups is the existence of internal markets allowing for the allocation of capital among member firms. Studies of Korean chaebol
for example have documented negative outcomes associated with tunneling (Bae et al., 2002; Joh, 2003; and Baek et al., 2006),
which may be exacerbated during a crisis (Bertrand et al., 2002).4
Other scholars have reported that family groups may actually be helpful by exploiting their financing and reputation advantages
to fund young high-growth firms with limited cash flows at the bottom of pyramidal ownership structures.5 Khanna and Yafeh (2007)
also argue that links between group firms may provide for mutual insurance which can help to dampen shocks; and Jia et al. (2013)
provide evidence for this phenomenon in the Chinese context.
A firm with director ties to a family group is well positioned to secure some benefits of association without incurring costs. Such
firms may selectively act upon valuable information while obviating the need to transfer resources. Without being subject to own-
ership by a controlling family shareholder, a firm will not face costs associated with tunneling and other forms of predatory behavior.
The extension of benefits may also be reduced, such as inter-firm resource transfers (e.g., to young, high-growth firms with limited
cash flows) (Gopalan et al., 2007; Morck et al., 2005; Almeida et al., 2011). But during a crisis, the likelihood for tunneling is greater
than the prospect of receiving financial assistance.
At the same time, firms with family networks may be especially well positioned to receive information regarding the capacity of
group firms to calibrate their financial and business operations (Khanna and Yafeh, 2007). The board connection may serve as a
conduit of information on the feasibility of modifying existing contracts with group firms; being extended trade credit in relation to
orders and payments; or temporarily reducing the size of interest payments on debt (Uzzi, 1999; Stuart and Yim, 2010).
1.3.2. State networks

Firms that are state-owned confront similar types of resource diversion costs as family-owned firms described above. Additionally,
state-owned firms also face costs due to the pursuit of government policy objectives unaligned with the firm's optimal financial
strategy (Dinc, 2005; Tian and Estrin, 2007). During a crisis, the costs of state ownership may become substantial as the state
prioritizes policy objectives, such as elevated employment levels or extending loans to troubled firms (Megginson and Netter, 2001).
These costs may be offset by access to state resources that dampen the need to implement cost-cutting measures (Chen et al., 2009).
Firms with state networks may receive valuable information enabling access to resources typically reserved for state-owned firms.
At the same time, these firms are not subject to an obligation to implement government policy directives, nor do they face resource
diversion. State networks can provide similar benefits to family networks when they access numerous firms within a state-owned
group. State networks may transmit information on the willingness of various firms to make advance payments; to delay payments for
receivables; to move forward purchases (thereby forestalling a decline in sales); or to grant an extension of trade credit (Uzzi, 1999;
Stuart and Yim, 2010). If a financial institution is a member of the state-owned group, then state networks may also convey in-
formation regarding capacity to delay debt repayment, or to reduce the cost of debt (Stearns and Mizruchi, 1993). Further in-
formation may regard government capacity to extend preferential treatment or to offer various forms of financing assistance during
the crisis period.
1.3.3. Political networks

Firms with politically connected directors face somewhat similar tradeoffs as firms that are state-owned. These firms may ex-
change bribes or campaign finance for favorable treatment by the state (Shleifer and Vishny, 1994; Lester et al., 2008). Benefits may
3
Granted some firms (such as those critical to a parent firm's operations) may receive financial assistance.
4
Tunneling in the context of business groups refers to the transfer of assets/profits across firms within the group. Related work on tunneling
focuses on transfers across shareholders within (rather than across) firms (Lemmon and Lins, 2003; Johnson et al., 2000; Baek et al., 2004).
5
See Morck et al. (2005) and Khanna and Yafeh (2007) for surveys of this literature.
4
include greater access to government financing (Khwaja and Mian, 2005; Duchin and Sosyura, 2012; Li et al., 2008) or bailout funds
(Faccio et al., 2006); increased procurement of government contracts (Goldman et al., 2013); protection from market competition
(Bunkanwanicha and Wiwattanakantang, 2009); and favorable regulatory treatment (Berkman et al., 2010). Firms have been shown
to benefit from their political connections in both emerging and high-income economies (Fisman, 2001; Johnson and Mitton, 2003;
Ferguson and Voth, 2008; Goldman et al., 2009; Cingano and Pinotti, 2013; Acemoglu et al., 2016).
Despite the advantages above, political connections may also impose pressures to pursue policy objectives at odds with firm
performance, like elevated employment levels (Naughton, 2009). An additional drawback of political connections is the incentive for
managers to expropriate for personal gain. Faccio et al. (2006) find politically connected firms more likely to be bailed out, and they
also exhibit worse performance. Chekir and Diwan (2013) find in Egypt that politically connected firms display worse accounting
performance. Moreover, politically connected firms face costs arising from their affiliation with patronage networks allied to in-
cumbent political leaders (i.e., patron-client relationships) (Johnson and Mitton, 2003). During a crisis, costs are likely to rise in order
to fulfill political leaders' objectives of preserving the stability of their political rule.
Firms with political networks (but no direct political connection) are one step removed from political influence, and do not face
pressures to implement costly measures for political expediency. Instead, they can benefit from information privy to political insiders,
thereby gaining privileged access to credit and other government-provided benefits. Despite these benefits, firms with political
networks will lack access to information concerning multiple lines of business (as with family or state networks). This limits the
availability of information regarding adjustments to accounts payable or the extension of trade credit, for example.
2. Data
We assemble a unique dataset capturing various types of network interactions among East Asian firms. In particular, we document
board connections to firms characterized by family business group affiliation, state ownership, and political ties. Due to the sub-
stantial resources devoted to hand-collecting this data, our sample faces constraints to cross-sectional breadth and temporal depth.
Still, we present detailed board network data for 1290 firms, which are among the 200 largest companies by 2008 market capita-
lization in each of nine East Asian economies.6
Our regional coverage includes Hong Kong (accounting for 133 sample firms), Indonesia (169), Japan (126), South Korea (133),
Malaysia (281), the Philippines (98), Singapore (116), Taiwan (107), and Thailand (127). Within East Asia, we choose these nine
economies because they have a sufficiently large number of listed firms with data available. But data availability on board members
still precludes us from including all target firms in our sample. Our results are therefore subject to a sample selection bias towards
firms with greater disclosure practices. It is unclear how this selection may bias later findings, but we have no theoretical justification
to suspect strong directional bias. Given the nature of our selection (missing data), we unfortunately cannot conduct a sensitivity
analysis to determine whether effect sizes are conditional on likelihood of inclusion in our sample.
We use cross-sectional network data for our sample in 2008. These are then supplemented with other cross-sectional and panel
data throughout the analysis. In the remainder of this section we first discuss mapping board networks, then describe our data on
family business groups, state ownership, and political connections. Then we formally introduce our measures of network interactions,
and conclude the section by describing data on additional firm characteristics. Descriptive statistics for all firm data are presented in
Table 1.
2.1. Board connections
To measure board networks we count the number of board interlocks - instances in which a firm's board member or executive is
shared with another firm in the economy. This method of measuring board networks follows Bizjak et al. (2009) and Larcker et al.
(2013). To construct this measure, we require board members and executives data for each sample firm. We focus on the largest 200
publicly-traded firms within each country, annual report availability permitting. Annual reports are taken from Worldscope, OSIRIS,
and company websites. In gathering data from these sources, we amass a pool of approximately 29,000 names affiliated with our
sample firms. We algorithmically match directors/executives across firms to depict each firm's board network. Subsequently, we
conduct manual verification to correct for inconsistencies across annual reports arising from transliteration.
Gathering time-varying board network data is not feasible in our case since we require hand-collected data linking board con-
nections, ownership structures, and political connections. Existing studies focusing exclusively on board networks have benefitted
from the BoardEx database which provides extensive details on individual directors of listed firms. For our region and time period of
interest, however, directors data are not widely available from that source. Crisis-period (2008–09) data for the countries in our study
is limited to only 250 firms in BoardEx. Due to minimal coverage, Indonesia, South Korea, the Philippines, and Thailand would need
to be dropped from a study based on BoardEx data. The resulting sample would therefore lack the size and institutional variation
necessary for identifying our network interactions of interest. Consequently, to conduct our analysis we rely on manually collected
data on board networks, ownership structures, and political connections.
6
For Malaysia, our sample is drawn from the largest 300 listed firms.
5
Table 1
Descriptive statistics.
N mean SD min max
ROA 9426 1.91 4.11 −14.35 16.36

avg return 29,966 −0.09 0.74 −11.28 16.98
board network 1290 4.4 5.6 0 38
family network 1290 1.38 3.15 0 26
state network 1290 0.87 2.01 0 18
political network 1290 0.28 1.01 0 14
family 1097 0.24 0.43 0 1
state 1290 0.17 0.37 0 1
political 1290 0.04 0.21 0 1
boardsize 1290 11.3 4.9 0 34
listing 1273 1992 8.8 1973 2007
concentration 1145 34.2 26 0 100
institutional 1145 1.2 7.5 0 100
blockholder 1150 0.59 0.49 0 1
tangible 8166 0.33 0.24 0.00 0.90
cash 7369 0.07 0.08 0.00 0.47
liabilities 9692 5.43 18.24 0.00 145.01
volatility 9893 0.03 0.03 0.00 0.16
leverage 8154 0.52 0.27 0.03 1.46
ROE 8173 2.69 6.77 −32.81 28.97
sales 7987 11.42 2.28 0.00 18.05
size 8272 13.55 2.08 8.98 18.77
mature debt 7980 0.10 0.29 0.00 1.19
account payable 1246 0.09 0.08 0.00 0.50
skill 3541 65.58 24.56 0 100
cv 3549 10.38 3.73 2 25
experience 3513 7.48 4.48 0.25 28.84
affiliations 3540 0.92 0.83 0 5.36
Data are for 1290 firms spread across nine East Asian countries. Network and corporate governance data (panels 2–4) are assembled by the authors,
and are cross-sectional for 2008. Other financial/accounting data span from 2007 Q2 until 2009 Q1. ROA: return on assets (quarterly, percentage
points). average return: daily average return (monthly average, percentage points). board network: number of interlocking directorates. Family
network: director ties to family-owned firms. State network: director ties to state-owned firms. Political network: director ties to politically connected
firms. State: indicates if firm has government blockholder. Political: indicates if firm has politically connected director. Family: indicates if firm
belongs to business group. Boardsize: number of board members. Listing: year of IPO. concentration: control rights of largest blockholder (%).
institutional: control rights of largest financial institution blockholder. Blockholder: indicates if any entity controls more than 10% of shares. Tangible:
tangible assets - ratio of investment in plants, property, and equipment to total assets. Cash: relative to total assets. Liabilities: total liabilities - short
and long term obligations. Volatility: realized return volatility (quarterly, standard deviation of daily realized return). leverage: ratio of total li-
abilities to total assets. ROE: return on equity (quarterly, percentage points). sales: total sales (log). size: total assets (log). mature debt: debt maturing
within one year (scaled by assets). account payable: calculated over assets. Skill: % of board members who have either an industry-specific or
financial background. cv: total number of board members with publicly disclosed background. Experience: average number of years each board
member has been on the board. Affiliations: average number of other corporate affiliations for the board member.
2.2. Family business groups, state ownership, and political connections
We consider a firm to form part of a family group if it falls under the direct or indirect ownership of that group. Each group
includes at minimum two of our large listed sample firms, and also includes other non-sample firms. This definition is consistent with
Khanna and Palepu (2002) and Masulis et al. (2011). If two sample firms share a common family shareholder, they are deemed
affiliated to a family group.7 To build our family group indicator we draw on previously collected ownership data from Carney and
Child (2013). Following common practice (see Claessens et al., 2000; Morck et al., 2005), we use a 10% threshold of outstanding
share ownership to depict control. The shares need not be held directly, and are often held via other publicly traded firms in which
the group enjoys majority control.8 When a pyramid arises, we calculate ultimate ownership as the smallest stake in the chain of
control (following Classens et al. 2000).
We identify state-owned firms as those in which the state holds a controlling share (Megginson and Netter, 2001). Again we use a
10% threshold to define control, and ownership by the state may be direct or indirect (via a number of other entities, including
sovereign wealth funds). Precise ultimate ownership calculations are carried out as described above. An exhaustive explanation of the
construction of ownership figures is available in Carney and Child (2013). The sources for ownership data include the ThomsonONE
Worldscope, Bureau Van Dijk OSIRIS, and LexisNexis databases; company websites; stock exchange filings; and media reports.
We consider a firm to be politically connected if a director or executive has political ties (i.e. he/she simultaneously occupies a
7
Our proportion of group-affiliated listed firms per country closely resembles the analogous figures reported in Masulis et al. (2011).
8
Either the family group or the intermediate listed firm may be situated abroad, although this is rarely the case empirically.
6
position as minister or MP). To identify such connections, we combine hand-collected data on politicians with aforementioned data
on board members and executives. The data on politicians were assembled from various government websites, hardcopy publications,
and email correspondence with public officials (Carney and Child, 2013). A complete list of sources used for this data is available in
Table A1. For each country, every member of the legislative and executive branches of government has been recorded. We sup-
plement an algorithm with manual verification to cross-reference the list of politicians with board members to detect firms with
political ties.9 Our method of identifying political connections follows Faccio (2006), except our coverage includes all directors (in
addition to top executives), but does not identify the political status of directors' friends and family. By restricting the measure to
encompass only directors/executives who concurrently hold political office, we can be confident about their access to political in-
formation of potential relevance to a firm.10
2.3. Network interactions
Our analysis is centered on the network of interlocking directorates. The nodes of this network are firms, and the links represent
cross-directorships between corresponding firms. We calculate degree centrality - the sum of board interlocks with other firms (each
interlock weighted according to the number of board members involved). With respect to the flow of information, an actor with a
relatively high measure of degree centrality is a focal point of communication and information (Freeman, 1978). An actor with a low
measure of degree centrality would be at the periphery of the network's information flow, isolated from direct communication with
others in the network. Hence, degree centrality is important as a baseline index of potential communication activity and information
exposure.
We decompose degree centrality into subcomponents reflecting director ties to various types of firm groups. These subcomponents
constitute network interactions. Specifically, we examine the overlap between board networks and networks of: political influence
(p); state ownership (s); and family business group affiliation (f). We also trace board connections to firms unrelated to all the above
networks (o). In what follows, we formally define our general measure of board networks and each network interaction. To begin, our
degree centrality measure is calculated as:
D = Gu
where G is the matrix of interlocking directorates, and u is the unit vector. The symmetric matrix G is of dimension N × N, where N is
the number of sample firms within the subject country, and also the length of u. Each element Gij indicates the number of common
board members between firm i and firm j.
Our decomposition of degree centrality is very straightforward. Let pi, si, fi, and oi be firm-level indicators of political influence;
state ownership; family affiliation; and the absence thereof, respectively. For each of these dimensions, we encode a column vector of
length N whose elements are the corresponding binary indicators. Let us denote these column vectors as p, s, f, and o. The resulting
decomposition then simply takes the following form.11
Dd = Gd d {p, s, f, n}
The element Did therefore reflects how many board connections firm i shares with other firms possessing network connections in
domain d. Alternatively put, Dd provides an index of communication potential and information exposure with respect to firms
characterized by d.
Descriptive statistics for each board network type are offered in Table 2, broken down by country. For each network interaction,
there is considerable variation both across and within countries. At the same time, these networks often occur in tandem at the firm
level. Table 3 reports Pearson's correlation coefficients between the various types of board networks, which are all positive and
significant at the 1% level.
2.4. Firm characteristics
Financial data for our analysis are obtained from Datastream. Our key performance measures are: (i) return on assets (ROA),
measured quarterly and calculated as the ratio of net income to total assets (expressed in percentage points); and (ii) daily stock
returns, averaged monthly and expressed in percentage points. Our control variables include: public listing date (year)12; tangible
assets (ratio of investment in plants, property, and equipment to total assets, quarterly); firm size (log of total assets, quarterly); total
liabilities (quarterly); cash (relative to total assets, quarterly); ROE (quarterly); sales (log of total sales, quarterly); debt maturing
within 1 year (relative to total assets, FY2007); cost of goods sold (relative to total assets, quarterly); leverage (ratio of total liabilities
to total assets, quarterly); stock return volatility (standard deviation of daily stock returns, quarterly); and industry classification (13
sectors, based on two-digit SIC codes in accordance with Campbell, 1996). In addition to the above, we use previously hand-collected
data from Carney and Child (2013) to report on board size (number of board members); blockholders (indicating whether a single
entity holds more than 10% of outstanding shares); ownership concentration (control rights of the largest owner); and institutional
9
In our sample, it is never the case that foreign politicians (from within East Asia) fill domestic board positions.
10
We identify political ties for 4% of sample firms, while Faccio (2006) identifies political ties for 5% of firms across the same set of countries.
11
In a strict sense, the set {Dp, Ds, Df, Do} is not a pure decomposition of D because pi, si, and fi are not mutually exclusive indicators.
12
Year of establishment is unavailable for some countries in our sample, so we opt instead for the listing date as a proxy for firm age.
7
Table 2
Networks across East Asia.
Country N Board network Family network State network Political network
mean SD max mean SD max mean SD max mean SD max
Hong Kong 133 5.12 6.1 33 2.62 4.51 26 1.00 1.41 6 0.67 1.37 6
Indonesia 169 1.64 3.31 23 0.95 2.64 17 0.14 0.38 2 0.22 1.09 9
Japan 126 1.84 2.33 15 0.07 0.42 3 0.09 0.31 2 0.00 0.00 0
South Korea 133 2.5 2.8 21 1.09 1.37 6 0.15 0.40 2 0.02 0.15 1
Malaysia 281 7.35 6.61 37 1.07 1.94 8 2.15 3.09 18 0.36 0.74 5
Philippines 98 8.52 8.91 38 5.33 6.16 21 0.71 1.59 10 0.20 0.81 6
Singapore 116 3.52 3.24 15 0.59 1.66 12 1.28 2.40 11 0.57 1.90 14
Taiwan 107 1.6 2.22 12 0.21 1.11 7 0.14 0.46 3 0.00 0.00 0
Thailand 127 5.11 5.04 23 1.58 3.15 19 0.73 1.99 11 0.29 1.16 8
Data are for 1290 firms across nine East Asian economies. All network data are assembled by the authors, and are cross-sectional for 2008. Table
reports country-level statistics on board networks, family networks, state networks, and political networks. Minimum values are everywhere 0. board
network counts the amount of board/executive interlocks. Political network counts the amount of board/executive interlocks with politically-con-
nected firms. Family network counts the amount of board/executive interlocks with family-controlled firms. State network counts the amount of
board/executive interlocks with state-owned firms.
Table 3
Network correlations.
board network political network family network
Political network 0.3300*

Family network 0.7417* 0.3021*
State network 0.2665* 0.2620* 0.0872*
Reported numbers are Pearson correlations (* means significant at the 1% level). Data are for the 1290 firms for which network data are
available across all networks. Board network counts the amount of board/executive interlocks. Political network counts the amount of board/
executive interlocks with politically-connected firms. Family network counts the amount of board/executive interlocks with family-controlled
firms. State network counts the amount of board/executive interlocks with state-owned firms.
ownership (percentage of shares held by financial institutions). All monetary values are expressed in terms of USD. Accounting
measures and volatility are winsorized at the 1% and 99% levels.
3. Performance implications of board networks
3.1. Networks and accounting performance
Throughout this section we operationalize firm performance as quarterly ROA. Our composite board network measure is degree
centrality, ultimately decomposed into family networks, state networks, political networks, and other networks. Following Lins et al.
(2017), the crisis period we examine includes Q4 2008 and Q1 2009. This period directly follows the collapse of Lehman Brothers on
September 15th 2008. Since the crisis arrived as a shock, our identification suffers less from selection bias. For example, it is not
intuitive that firms ex-ante select their board networks on the basis of crisis-performance. However, board networks may well be
selected on traits correlated with crisis performance (e.g. regular performance). Throughout this section we hope to allay such
concerns; additional threats to identification are addressed in section five and an online appendix.
3.1.1. Cross-sectional model

To understand the correlation between firm performance and board networks during the global financial crisis, we first report
simple bivariate regressions for Q4 2008 (the hardest-hit quarter).13 For ease of interpretation, we include country-industry fixed
effects. We thus examine whether networked firms enjoyed higher performance during the global financial crisis than non-networked
country-industry counterparts. The statistical model we estimate in Table 4 is:
yijk = Nijk + jk + ijk
In this specification, ROA (y) of firm i in industry j of country k is determined only by the network measure (N) and the country-
industry effect (γ). The network measure N alternates across columns 1–5 between degree centrality and its various subcomponents
(N ∈ {D, Df, Ds, Dp, Do}).
The first column of Table 4 shows firm performance during the 2008 crisis strongly positively correlates with board network
13
Results are similar for Q1 2009.
8
Table 4
Firm performance and board networks.
(1) (2) (3) (4) (5) (6) (7) (8)
Board network 0.0800***

(0.0274)
Family network 0.0927* 0.0890* 0.108* 0.163**
(0.0473) (0.0504) (0.0606) (0.0820)
State network 0.744** 0.668** 1.616** 2.797***
(0.290) (0.295) (0.644) (0.746)
Political network −0.0160 −0.103 −0.0488 −0.119
(0.136) (0.148) (0.167) (0.236)
Other network 0.0888** 0.0708* 0.0701 0.0330
(0.0388) (0.0399) (0.0506) (0.0559)
Network controls √ √
Firm controls √
Observations 1178 1178 1178 1178 1178 1178 1033 753
R-squared 0.192 0.189 0.191 0.186 0.190 0.196 0.202 0.336
Data are for 1178 firms spread across nine East Asian countries. Financial indicators come from Worldscope and Datastream. Network data are
assembled by the authors. Dependent variable is quarterly return on assets (ROA), expressed in percentage points. Explanatory variables are various
measures of networks (defined in Sections 2.1–2.3). In column 7, we additionally control for network controls: state ownership, family ownership,
and political connections. In column 8, we additionally control for other firm characteristics (listed in Section 3.1.1). Data are cross-sectional for the
fourth quarter of 2008 - the crisis period following the collapse of Lehman Brothers. All specifications include country-industry fixed effects. The
standard errors reported in parentheses are clustered by country-industry (***p < 0.01, **p < 0.05, *p < 0.1).
centrality. Columns 2–5 decompose board networks into the various subcomponents (i.e. network interactions). Performance is
positively correlated with board connections to family, state, and ordinary firms. Because these network dimensions are interrelated
(see Table 3), we include them together in column 6 (formally, N becomes the vector [Df, Ds, Dp, Do]). Again we find a similar
relationship. But the estimates in column 6 are likely subject to omitted variable bias. Obvious sources of confound include family
control, state ownership, and political ties. A state-owned company is likely to have interlocking directorates with other SOEs
(correlation 0.61), and the same logic extends to family business groups (correlation 0.56) and politically connected firms (corre-
lation 0.37). Accordingly, in column 7 we include a vector of controls, Xijk, capturing state ownership, family control, and political
ties (hereafter referred to as network controls). The estimated model takes the form:
yijk = Nijk + Xijk + jk + ijk
When including network controls, family and state networks remain positively correlated with crisis performance, but ordinary
board networks are no longer significant. Other firm characteristics may also drive crisis-performance and be correlated with board
networks. Following related work by Anderson and Reeb (2003), Gorton and Rosen (1995), Dass et al. (2014), Erkens et al. (2012),
Adams et al. (2004), Masulis and Mobbs (2011), Jensen (1986), and De Jong (2002), we control for: firm age; firm size; board size;
institutional ownership; blockholders; ownership concentration; return volatility; leverage; tangible assets; liabilities; and cash.14
Adding the aforementioned controls to the vector Xijk, we estimate a more flexible model in column 8. Family networks and state
networks are highly correlated with crisis performance, and coefficient magnitudes are larger under this controlled specification. But
because our sample size has also declined in column 8 (due to missing control data) we cannot determine whether changes in effect
sizes/significance are due to improved precision or sample selection. Overall, however, these results suggest observable firm char-
acteristics are unlikely to explain the performance differential between firms with strong networks and those without.
3.1.2. Dynamic model

In the preceding section we show crisis performance is strongly correlated with board networks, and board connections to family
and state-owned firms in particular. We also provide evidence that board networks are not proxying for omitted firm characteristics.
An important remaining question is whether these correlations are restricted to the crisis period. To the extent that board networks
and performance are correlated over time, our argument against selection deteriorates. We claim crisis-performance alone is unlikely
to determine a firm's network established prior to the crisis. But regular firm performance is harder to rule out as a potential selection
criterion. If regular performance is correlated with crisis-performance, and we focus exclusively on the crisis period, we would
misinterpret our results as causal when they could be rooted in selection.
To address this concern we invoke a dynamic model with a 2-year sample starting from Q2 2007. Following Lins et al. (2017), Q4
2008 and Q1 2009 form our crisis period. All preceding quarters comprise the benchmark. We are interested in crisis-contingent
effects - the interaction term between networks and the crisis. A general network effect introduces a performance wedge between
networked and non-networked firms during non-crisis periods. That wedge may be attributable to selection effects. The crisis-con-
tingent network effect captures the magnitude of increase in that wedge during the crisis. That increase is net of selection on time-
invariant characteristics. We control for country-industry, and allow for country-industry shocks during the crisis period. This rules
14
Refer to Section 2.4 for precise definitions of our control variables.
9
out the explanation that networked firms happen to be concentrated in country-industries relatively unscathed by the crisis.
Moreover, we explicitly control for firm performance over the preceding eight quarters to capture any time-varying performance-
related selection into treatment. We include time period dummies to net out common time shocks. To account for interdependencies,
standard errors are clustered at the country-industry level.15 The panel model we estimate is:
yijkt = Nijk Ct + Nijk + jk Ct + jk + yijk, t + t + ijkt (1)
where C is a crisis dummy indicating whether period t is either Q4 2008 or Q1 2009. Hence, we allow both the network effects and
the country-industry effects to vary during the crisis. It is the effects captured in β which are of primary interest.
The first column of Table 5 suggests the performance of firms with strong board networks was buoyed following the collapse of
Lehman Brothers. Even in comparison to country-industry counterparts with similar performance trajectories, firms with strong
board networks fared better. During regular periods of economic activity, however, there is no performance differential between
firms with strong and weak board networks (conditional on a strict set of controls). To understand which types of board connections
yield benefits during the crisis, we next unpack our composite board network measure into its constituent parts. In columns 2–5 the
board network measure alternates between various types of network interactions. Similar to our cross-sectional results, we find the
beneficial impacts to be concentrated in family and state networks. In column 6 we include all networks simultaneously. Again we
find state and family networks to be the strongest determinants of crisis performance. Because of the aforementioned correspondence
between family control and family networks; between state ownership and state networks; and between political ties and political
networks; in column 7 we extend our model to account for the network controls.
yijkt = Nijk Ct + Nijk + Xijk Ct + Xijk + jk Ct + jk + yijk, t + t + ijkt (2)
The network controls in Xijk are also interacted with the crisis dummy, to allow for crisis-specific shocks. The effects we detect are
thus marginal effects beyond any direct performance implications of such characteristics (in or out of crises).16 Our results are robust
to this flexible panel model. Our final column 8 adds to Xijk the firm controls described in the preceding section. Again, our results
remain robust to this flexible specification.17 During the crisis period, a one-standard-deviation increase to family network size
implied a 0.57 percentage point boost to quarterly ROA (relative to a crisis mean and standard deviation of 0.50% and 4.7%,
respectively). By comparison, a one-standard-deviation increase to the size of state networks yielded a 3.2 percentage point increase
in quarterly ROA.
3.2. Networks and stock returns
In the preceding section we demonstrate accounting performance was buoyed by family and state networks during the global
financial crisis. To determine whether financial markets value board access to family and state-owned firms, we next examine the
impact of network interactions on stock performance (i.e. raw returns, following Lins et al., 2017). To this end we estimate the
following model:
yijkt = Nijk Ct + Nijk + jk Ct + jk + t + ijkt (3)
where the outcome y is average daily stock returns, measured each month (t) and expressed in percentage points. We again allow for
level-shifts in country-industry effects (γ) during the crisis and, as before, β is our coefficient of interest. Standard errors are clustered
at the country-industry level.
In column 1 of Table 6 we first estimate the general impact of board networks on stock performance. We find the effect to be
positive and significant during the crisis. Next, in columns 2–5 we unpack that effect into the various network subtypes. As with
accounting performance, we find the positive impact on financial performance to be driven by family and state networks. Including
all network interactions at once in column 6 yields a similar result. In column 7 we condition the estimates on crisis-contingent
network controls and find our results robust. Finally, in column 8 we include a vector of firm controls deemed important in earlier
research: firm size (Dass et al., 2014); return volatility (Goyal and Santa-Clara, 2003); leverage and return on equity (Jensen, 1986;
De Jong, 2002; Acemoglu et al., 2016). In this final specification we again find family and state networks to be robust determinants of
stock performance during the crisis.
On average our sample firms incurred a one-third percentage point decline in daily stock returns during the crisis (relative to pre-
crisis means). Our estimates from column 8 therefore suggest a one-standard-deviation increase to the size of state networks
15
Because firm clusters are nested within country-industry clusters, it is not surprising our results are robust to firm-level clusters. With only nine
countries, it is inadvisable to cluster over so few (unbalanced) clusters (Cameron and Miller, 2015).
16
As a point of theoretical interest, the net impacts of family group affiliation, state ownership, and political connections during the crisis are
typically indistinguishable from zero in our setting. This may be due to costs of association outweighing benefits during the crisis, or adverse
selection during periods of regular economic activity.
17
Notably, we find no evidence that board networks influence performance outside of the crisis. At first glance this may appear inconsistent with
the findings of Larcker et al. (2013), but the latter results are identified using within-firm variation in boardroom centrality. Our network effects
remain poorly identified during periods of regular economic activity, as we rely on the crisis shock to introduce environmental conditions in which
the beneficial features of board networks become salient. Accordingly, our documented effects of board networks on firm performance are truly
conditional on this period of high uncertainty and financial dislocation.
10
Table 5
Impact of board networks on crisis performance.
(1) (2) (3) (4) (5) (6) (7) (8)
Board network × crisis 0.0313**

(0.0127)
Family network × crisis 0.0614*** 0.0589*** 0.0517** 0.182*
(0.0209) (0.0209) (0.0255) (0.0941)
State network × crisis 0.379** 0.363** 0.558*** 1.591**
(0.155) (0.150) (0.191) (0.628)
Political network × crisis 0.0585 −0.0433 −0.0202 −0.0675
(0.0688) (0.0651) (0.0664) (0.115)
Other network × crisis 0.00551 −0.000200 −0.00360 0.0225
(0.0196) (0.0206) (0.0285) (0.0258)
Board network 0.00258
(0.00600)
Family network 0.00693 0.00373 0.0146 0.0816**
(0.00950) (0.0108) (0.0150) (0.0395)
State network −0.000602 −0.0106 −0.130 0.0786
(0.158) (0.158) (0.231) (0.371)
Political network 0.0283 0.0297 0.0449 −0.0493
(0.0245) (0.0301) (0.0295) (0.0620)
Other network −0.00153 −0.000565 −0.00248 −0.0286
(0.00602) (0.00669) (0.0101) (0.0233)
Firm controls √
Observations 6990 6990 6990 6990 6990 6990 5914 4452
R-squared 0.554 0.554 0.553 0.554 0.553 0.555 0.573 0.235
assembled by the authors. Dependent variable is quarterly return on assets (ROA) expressed in percentage points. Explanatory variables of interest
are various measures of networks (defined in Sections 2.1–2.3), interacted with the crisis period to capture the crisis-contingent network effect.
Network data are cross-sectional for 2008. In column 7, we additionally control for network controls: state ownership, family ownership, and
political connections and their interaction terms with the crisis period. In column 8, we additionally control for other firm characteristics (listed in
Section 3.1) and their interaction terms with the financial crisis period. Observations run from second quarter 2007 until first quarter 2009, leaving
a panel of eight quarters. All specifications include country-industry fixed effects, as well as crisis-period country-industry shocks. Each specification
contains eight quarterly lags of the dependent variable. The standard errors reported in parentheses are clustered by country-industry (***p < 0.01,
**p < 0.05, *p < 0.1).
cushioned that decline by more than one-half.18 A standard-deviation increase to family networks offset the negative crisis shock by
approximately 7%. Earlier work establishes that director networks generate equity value (Larcker et al., 2013), also during times of
distress (Dass et al., 2014). Our results add to this knowledge by demonstrating higher crisis-period stock returns for firms with state
and family networks. Meanwhile, some scholars have associated the onset of state ownership and family control with greater
shareholder value (Bunkanwanicha et al., 2013; Karolyi and Liao, 2017). Our results suggest part of that value may be captured even
through indirect board association with state- and family-owned groups.
3.3. Political networks and institutions
Earlier work has found political connections to be valuable in emerging economies (e.g., Fisman, 2001; Faccio, 2006). Conducting
business in that context can be challenging because of the prevalence of institutional voids (Khanna and Palepu, 1997). Such voids
inhibit the spread of information; intermediaries responsible for verifying, disseminating, and analyzing financial information (e.g.,
accounting firms, credit rating agencies, and banks) are either absent or ineffective. Moreover, in environments with insecure
property rights, disclosing information about financial performance can make a firm vulnerable to predatory behavior by other actors
(Durnev et al., 2009), with negative implications likely magnified in times of uncertainty. To alleviate the above risks in emerging
markets, political connections can protect firms from exploitation and grant privileged access to information and resources. Thus,
political networks may also prove more valuable in weak institutional environments. To assess whether institutional development
moderates the benefits of political networks, we allow their impact to vary across three institutional factors: GDP per capita, investor
protections, and financial development.19 For each measure of institutional development we define a binary variable, Ik, indicating
whether the value for country k is greater than or equal to the median level of our sample in 2008. In Table 7 we estimate the
following model.
18
Precisely, we multiply the standard deviation of state networks (2.01) and the effect size (0.0936%) to obtain +0.188%, which offsets more
than half the average crisis decline.
19
GDP per capita data are from the IMF World Economic Outlook Database. The investor protection measure is from La Porta et al. (2006), and
reflects a combination of liability, disclosure, and antidirector rights. Financial development is based on Rajan and Zingales (1998), and is measured
by the ratio of stock market capitalization and domestic credit to GDP.
11
Table 6
Impact of board networks on stock returns.
(1) (2) (3) (4) (5) (6) (7) (8)
Board network × crisis 0.00414**

(0.00179)
Family network × crisis 0.00734** 0.00634** 0.00563* 0.00724**
(0.00306) (0.00301) (0.00313) (0.00330)
State network × crisis 0.0716*** 0.0683*** 0.0644** 0.0936***
(0.0184) (0.0182) (0.0247) (0.0289)
Political network × crisis 0.0139 0.000805 0.00344 −0.00880
(0.0110) (0.00914) (0.0109) (0.0109)
Other network × crisis 0.000903 0.000190 −0.00257 −0.00608*
(0.00254) (0.00258) (0.00348) (0.00356)
Board network −0.00112
(0.00104)
Family network −0.00158 −0.000725 0.000986 0.00161
(0.00131) (0.00147) (0.00221) (0.00242)
State network −0.0165 −0.0147 −0.0164 −0.0263*
(0.0103) (0.00974) (0.0125) (0.0141)
Political network −0.00706 −0.00499 −0.00522 0.000560
(0.00453) (0.00493) (0.00462) (0.00553)
Other network −0.000271 −0.000372 −0.00153 0.00131
(0.00131) (0.00142) (0.00194) (0.00203)
Crisis −0.137*** −0.129*** −0.134*** −0.123*** −0.122*** −0.143*** −0.123*** 0.0269
(0.0234) (0.0206) (0.0199) (0.0209) (0.0233) (0.0238) (0.0254) (0.115)
Firm controls √
Observations 22,845 22,845 22,845 22,845 22,845 22,845 19,634 17,880

R-squared 0.013 0.013 0.012 0.013 0.012 0.013 0.013 0.045
Data are for 1178 firms spread across nine East Asian countries. Stock price data come from Datastream. Network data are assembled by the authors.
Dependent variable is raw return expressed in proportions. Explanatory variables of interest are various measures of networks (defined in Sections
2.1–2.3), interacted with the crisis period to capture the crisis-contingent network effect. In column 7, we additionally control for network controls:
state ownership, family ownership, and political connections, interacted with the crisis period. In column 8, we additionally control for other firm
characteristics including ROE, leverage, stock price volatility, firm size and their interaction terms with the financial crisis period. Network data are
cross-sectional for 2008. Observations run from February 2007 to March 2009 (inclusive), leaving a panel of twenty-six months. All specifications
include country-industry fixed effects, as well as a crisis-period dummy. The standard errors reported in parentheses are clustered by country-
industry (***p < 0.01, **p < 0.05, *p < 0.1).
Table 7
Political networks and institutional strength.
(1) (2) (3)
GDP per capita Investor Protection Financial Development
Institutions × political network × crisis −0.177** −0.221** −0.234***

(0.0761) (0.0897) (0.0799)
Political network × crisis 0.137** 0.191** 0.191***
(0.0536) (0.0787) (0.0691)
Institutions × political network 0.0512 0.0707 0.0762*
(0.0535) (0.0434) (0.0451)
Political network 0.00668 −0.0150 −0.0166
(0.0217) (0.0197) (0.0192)
Observations 6990 6990 6990

R-squared 0.549 0.549 0.549
are links with political firms (defined in Sections 2.1–2.3), interacted with the crisis period to capture the crisis-contingent network effect. We set
independent variables (GDP per capita, Investor Protection, Financial Development) to one if it is at or higher than the median value of these nine
countries and zero otherwise. Network data are cross-sectional for 2008. Observations run from second quarter 2007 until first quarter 2009, leaving
a panel of eight quarters. All specifications include country-industry fixed effects, as well as crisis-period country-industry shocks. Each specification
contains eight quarterly lags of the dependent variable. The standard errors reported in parentheses are clustered by country-industry (***p < 0.01,
**p < 0.05, *p < 0.1).
12
yijkt = 0 Nijk Ct + 1 Nijk Ct Ik + 0 Nijk + 1 Nijk Ik + jk Ct + jk + yijk, t + t + ijkt
Each column of Table 7 invokes a separate measure of institutional development, yet all yield similar results. Political networks
significantly boost crisis performance in countries with weak institutions, but that effect dissipates in strong institutional environ-
ments.20 Under weak institutions, a standard-deviation increase to the size of political networks boosted quarterly ROA by
0.14–0.19% during the crisis. Earlier research has documented the value of direct political connections in emerging economies
(Fisman, 2001; Faccio, 2006), including during times of crisis (Johnson and Mitton, 2003). Our results add to that literature by
documenting the value of indirect association to politicians via director interlocks.21
4. Mechanisms
In this section we test three potential channels through which our different types of network interactions may generate value. The
first mechanism concerns the ability of firms to adjust trade credit in relation to accounts payable (Garcia-Teruel and Martinez-
Solano, 2010). Trade credit received represents a source of short-term financing which may be used to finance a significant portion of
the firm's current assets. Several studies have explained that trade credit provides a higher degree of financial flexibility than bank
loans and that it can overcome a variety of financial constraints, such as when credit from financial institutions is not available
(Danielson and Scott, 2004; Huyghebaert et al., 2007; Garcia-Teruel and Martinez-Solano, 2010). Firms belonging to a large own-
ership group (i.e. family- or state-owned firms) will have information regarding a comparatively large volume of sales across many
different businesses, expanding the opportunities for the extension of trade credit. Table 8 examines the relation between each type of
board network and trade credit during the crisis period.22 Columns 2 and 4 demonstrate that family and political networks exhibit a
significant positive relationship to trade credit. While those effects are not robust to the most flexible model in column 6, we contend
this is one possible channel through which benefits are conferred through these networks.
We next explore a debt financing channel (see Lins et al., 2017; Cohn and Wardlaw, 2016; Almeida et al., 2012). Firms with
political connections or state ownership may have preferential access to government resources, such as bank credit, and may convey
information about government-linked financial institutions that will modify financing arrangements. Financial institutions owned by
family groups may also extend financing assistance, but they would be more constrained than a government-linked entity with
greater resources and stronger political motivation to stem the impact of a crisis. Stearns and Mizruchi (1993), for example, de-
monstrate the types of financial institutions represented on a board affect the financing obtained by firms. We explore whether
similar effects extend to board networks by examining whether firms with strong networks enjoy a lower cost of debt, and whether
firms with high debt maturity draw greater benefits from the network.
In Table 9 we examine whether the benefits of board networks are greater for firms with high levels of debt maturity. We build a
firm-level indicator of debt maturity equal to 1 if FY2007 debt maturing within 1 year (i.e. during the crisis) is greater than the
sample median. We then test whether this measure moderates the impact of board networks on performance during the crisis.23
Table 9 shows that high-debt firms with family, state, and political networks exhibit better ROA performance during the crisis. Next,
in Table A2 we test whether board networks may reduce the cost of interest on debt during the crisis. We find little evidence in
support of this mechanism, suggesting the benefits derived from board networks by high-debt firms do not include directly lower
interest rates.
As a final potential channel, director networks may inform firms which businesses are capable of preserving their orders (or even
increasing purchases), which would be reflected in a smaller decline in sales growth. Dass et al. (2014) find that firms with directors
from related industries better anticipate and navigate sales shocks because they can more effectively manage their inventories. In the
20
Notably, we find no significant correlation between political networks and firm performance during non-crisis quarters. It is therefore worth
repeating our identification strategy does not permit valid inference based on the estimates ω0 or ω1 - both net effects subject to selection-based
endogeneity.
21
Taken together, Tables 5 and 7 imply political networks are effective at buffering performance in weak institutional environments, while state
networks buoy firm performance in general. This corresponds to a common reliance on patronage networks under weak institutions. For example,
private business owners linked to Malaysia's political incumbent Mahathir incurred far lower costs during the 1997 crisis than those allied to the
exiled political challenger, Anwar Ibrahim (Johnson and Mitton, 2003). By contrast, state-owned networks are likely to confer benefits regardless of
institutional strength (e.g., Singapore and Malaysia SOEs both received state support during the 97 crisis). While the Malaysia episode is a dramatic
illustration of the benefits of direct political ties under weak institutions, our results are consistent with this narrative.
22
Throughout Table 8 we control for firm characteristics that have been found to affect trade credit and are potentially correlated with networks.
The effects of these potential confounds are also allowed to vary during the crisis. Control variables include cash holdings, investments, purchases,
sales (log), and debt maturing within one year (all scaled by total assets). The existence of matching asset liquidity to the maturity of liabilities in a
firm is an important factor that positively impacts firm performance in general (Morris, 1976; Myers, 1977). We control for investment in current
assets because firms with more investment may require more supplier financing. Following Deloof and Jegers (1999) we control for cash holding.
Motivated by Garcia-Teruel and Martinez-Solano (2010), to control for the quantity of credit offered to customers we include cost of goods sold as a
proxy for purchases. The broader literature establishes that growth opportunities can reshape financing demands (Deloof and Jegers, 1999;
Niskanen and Niskanen, 2006). We therefore control for sales, expecting firms with higher sales to have higher demands for trade credit. Li et al.
(2008) document that well-connected firms have better access to external financing, so we control for debt maturity as firms find it difficult to roll
over maturing debt, and networks may be more important for these firms and thus increase the usage of trade credit.
23
Denoting our measure of debt maturity as Mijk, the estimated model is: yijkt = β0NijkCt + β1NijkCtMijk +
ω0Nijk + ω1NijkMijk + αjkCt + γjk + ψ0MijkCt + ψ1Mijk + ρyijk, t− + δt + εijkt
13
Table 8
Networks and trade credit.
(1) (2) (3) (4) (5) (6)
Board network × crisis 0.00182

(0.00190)
Family network × crisis 0.00468** 0.00132
(0.00217) (0.00984)
State network × crisis −0.00759 −0.00603
(0.00841) (0.00917)
Political network × crisis 0.00779** 0.00486
(0.00359) (0.0128)
Other network × crisis 0.00176 0.00184
(0.00214) (0.00208)
(0.00195)
Family network −0.00951 −0.00399
(0.00723) (0.0182)
State network 0.0132 0.0114
(0.0162) (0.0168)
Political network −0.0149 −0.0107
(0.0160) (0.0267)
Other network −0.00231 −0.00241
(0.00225) (0.00229)
Observations 1374 1374 1374 1374 1374 1374

R-squared 0.494 0.493 0.491 0.494 0.492 0.499
assembled by the authors. Trade credit is calculated as accounts payable over assets. Control variables are cash holdings over assets, investment in
current assets divided by assets, the purchases over assets, logarithm of sales and debt maturing within one year divided by total assets. Explanatory
variables of interest are various measures of networks (defined in Sections 2.1–2.3), interacted with the crisis period to capture the crisis-contingent
network effect. Network data are cross-sectional for 2008. Observations run from second quarter 2007 until first quarter 2009, leaving a panel of
eight quarters. All specifications include country-industry fixed effects, as well as crisis-period country-industry shocks. Each specification contains
eight quarterly lags of the dependent variable. The standard errors reported in parentheses are clustered by country-industry (***p < 0.01,
**p < 0.05, *p < 0.1).
context of a crisis, governments may deploy fiscal stimulus measures, and both political and state networks could convey useful
information about policy developments. Accordingly, Table A3 tests whether firms with network interactions undergo smaller de-
clines in sales growth during the crisis. We find no evidence in support of this channel.
5. Robustness
Thus far we have addressed endogeneity by introducing a dynamic model, employing a strict set of controls, and setting our
analysis in the unexpected crisis period. Still, additional challenges to identification persist. An online appendix rules out time-
invariant firm-level confounders, and demonstrates insensitivity to country and industry outliers. The present section is reserved for
robustness checks of greater theoretical interest: (1) director qualifications as a source of confound; (2) crisis intensity as a mod-
erating factor; and (3) crisis-induced changes to networks.
5.1. Director qualifications
Highly-qualified directors likely hold multiple directorships, including those at family and state-owned firms. Meanwhile, qua-
lified directors should be adept at buoying firm performance and navigating financial crises, due to superior monitoring and advising
capabilities. Therefore, director qualifications may underlie our main results as an important confounding factor.24 To acknowledge
this type of concern, we collect quarterly data on director qualifications and conduct a number of supplemental tests.
We assemble data on director qualifications based on three firm-level measures from Datastream. The first variable, skill, captures
the percentage of board members with an industry-specific or financial background. A second measure, CV, indicates the total
number of board members with publicly disclosed professional backgrounds. A third proxy, experience, measures the average tenure
of current board members. Unfortunately, the above measures of director qualification are available for only a subsample of 206
24
For example, consider a qualified director holding board positions at family-owned and widely-held firms. As discussed in Section 1.3.1, the
family-owned firm may face tunneling costs during the crisis, muting any monitoring or advising benefits associated with our director. The widely-
held firm, by contrast, would only benefit from enhanced monitoring and advising, and would be characterized by family networks due to the
connections of the qualified director. In this scenario our analysis would erroneously attribute strong performance of the widely-held firm to family
networks, but it would in fact be rooted in director qualifications.
14
Table 9
Debt maturity and network effects.
(1) (2) (3) (4) (5)
Debt maturity × board network × crisis 0.000157*

(8.54e-05)
Debt maturity × family network × crisis 0.000152**
(7.05e-05)
Debt maturity × state network × crisis 0.00108*
(0.000601)
Debt maturity × political network × crisis 0.000788***
(6.16e-05)
Debt maturity × other network × crisis −0.000294
(0.000304)
(0.0212)
Family network × crisis −0.00957
(0.0326)
State network × crisis 0.251
(0.209)
Political network × crisis 0.0574
(0.0678)
Other network × crisis 0.0347**
(0.0151)
(0.0121)
Family network 0.0231
(0.0174)
State network 0.00578
(0.225)
Political network 0.0512
(0.0524)
Other network −0.0102*
(0.00567)
Debt maturity × crisis −0.000828*** −0.000802*** −0.000799*** −0.000824*** −0.000729***
(0.000136) (0.000166) (0.000163) (0.000139) (0.000232)
Observations 5717 5717 5717 5717 5717

R-squared 0.547 0.547 0.547 0.547 0.546
are various measures of networks (defined in Sections 2.1–2.3), interacted with the crisis period to capture the crisis-contingent network effect. For
each firm, we set debt maturity to one if debt maturing within one year of fiscal year-end 2007 divided by total assets is at or above the 50th
percentile for the sample and zero otherwise. Network data are cross-sectional for 2008. Observations run from second quarter 2007 until first
quarter 2009, leaving a panel of eight quarters. All specifications include country-industry fixed effects, as well as crisis-period country-industry
shocks. Each specification contains eight quarterly lags of the dependent variable. The standard errors reported in parentheses are clustered by
country-industry (***p < 0.01, **p < 0.05, *p < 0.1).
firms. Relative to our full sample, this subsample is reasonably balanced along financial and industry characteristics, but Table A4
shows Japanese firms comprise 57% of the subsample. Before checking director qualification as a potential confound, we therefore
must re-establish benchmark results for the restricted sample.
Columns 1 of Tables A5-A6 reproduces tests from columns 8 of Tables 5-6, but for the restricted sample of 206 firms. Roughly
consistent with results from our full sample, we find state networks to boost ROA, and find both state and family networks to improve
stock returns. Political networks actually appear to negatively affect crisis performance in this subsample, however, according to both
ROA and stock returns.25 Relative to the column 1 benchmark, we next explore the confounding role of director qualifications. In
columns 2–4 of Tables A5-A6 we add to the benchmark our measures of director qualifications and their interaction with the crisis
period. In columns 5 we include all qualification measures simultaneously. Our results suggest director qualifications do not sys-
tematically influence ROA or stock returns during the crisis period. Moreover, effect sizes and standard errors of network interactions
remain reasonably stable when including director qualifications as a potential confound. Thus, it appears director qualifications do
not significantly confound the role of state and family networks, at least within this restricted sample.
For additional reassurance, we conduct further tests on a broader sample of East Asian companies. For 817 out-of-sample firms in
25
Among these 206 firms, most political connections are identified in Hong Kong and Singapore (comprising 1/4 of the subsample). It is possible
that managers in the region overestimated government assistance (based on their 1997 experience), contributing to suboptimal decisions and
performance declines. In countries with weak institutions, firms may draw upon political networks for assistance under such conditions. But in
countries with strong institutions (e.g., HK and Singapore), such assistance may not be forthcoming.
15
the region, Datastream offers firm-quarter averages of director qualifications and directors' external corporate affiliations. We regard
this affiliations measure as a proxy for board networks, and demonstrate its importance for crisis performance. In columns 1 of Tables
A7-A8 affiliations are found to significantly improve performance during the crisis period (consistent with the effect of board net-
works in columns 1 of Tables 5-6). In columns 2–5 we allow for the confounding role of director qualifications, yet the estimated
impact of affiliations remains positive and significant across all specifications. Our coefficient of interest remains stable in Table A7,
but effect sizes are dampened in Table A8 (where qualifications are also found to boost stock performance).26 Director qualifications
could therefore account for some of the network effects we earlier identify in Table 6. But taken together, the evidence of this section
suggests director qualifications are not sufficiently important to constitute an alternative interpretation of our main results.
5.2. Crisis intensity
During a crisis, family and state owned firms may be compelled to divert resources towards parent companies, while politically
connected firms may incur costs helping politicians preserve their rule. We posit that firms with board associations to family, state, or
politically connected companies do not have an obligation to transfer resources to external actors. At the same time, such firms will
glean information from board associations to help calibrate their crisis response. The value of such information rises as uncertainty in
the economic environment escalates - in tandem with the severity of a crisis. So according to our theory, the beneficial effects of
network interactions should be stronger for firms located in countries hit harder by the crisis.
To verify whether data support our theory in the above respect, we build five measures of crisis intensity following Lane and
Milesi-Ferretti (2011). With data from the IMF World Economic Outlook, we calculate for each economy the crisis-period (2008–09)
growth rates of GDP, private consumption, investment, imports, and exports. We then construct binary indicators of crisis intensity
equal to one if the corresponding growth rate is less than or equal to the sample median, and zero otherwise. In Table A9 we allow for
heterogeneous effects of network interactions according to crisis severity. Columns are differentiated by the growth rate upon which
crisis intensity is defined. In Sections 3.1 and 3.2, state and family networks were found to be most beneficial to crisis performance in
our setting. It is therefore reassuring we find both of these network interactions more impactful in settings characterized by severe
crises.
5.3. Crisis impact on networks
Our network data is accurate for the end of 2008, but we treat networks as static throughout the sample period (Q2 2007 - Q1
2009). Measurement error therefore constitutes a potential threat to identification. While board composition and ownership struc-
tures are persistent, they do in fact change. In the context of our study, such changes could be triggered by the crisis itself, which
would imply early observations are mismeasured. If inaccuracies were randomly distributed across firms, this would cause at-
tenuation bias in the regular network effects, and consequent overestimation of (marginally additive) network effects during the
crisis. If measurement error were nonrandom, our estimates would be subject to directional bias. In this subsection we therefore
endeavor to test whether the crisis meaningfully impacted firm networks.
The absence of time-varying network data is at the core of this issue, so we cannot directly observe whether the crisis altered
networks within our sample. For the out-of-sample firms from Section 5.1, however, we can examine the crisis impact on affiliations -
our time-varying board network proxy. This measure reflects the average number of external affiliations at the board level, so it is
subject to change in the event of significant director turnover. In Table A10 we estimate:
Nijkt = Ct + jk + ijkt (4)
where C is the crisis indicator, N captures affiliations, t indexes the quarter (between Q2 2007 and Q1 2009), and the fixed effects γ
vary across columns. In Panel A the network outcome is expressed in level terms; in Panel B it is expressed in first differences
(between t and t − 1); and in Panel C the outcome constitutes a binary indicator for whether any change (positive or negative)
occurred. Across all model specifications we find no evidence to suggest the crisis significantly influenced external affiliations of
directors. It is therefore unlikely the crisis meaningfully impacted board networks during the crisis (at least for this particular
sample).
Beyond board networks, our primary interest regards state networks, family networks, and political networks. To determine
whether the crisis altered these networks, we must also examine its impact on ownership structures and political appointments.
Practically no elections or transfers of political power occurred during the crisis, so political networks remained as constant as board
networks.27 Ownership structures, on the other hand, are generally subject to greater fluidity. To test whether the crisis affected
family and state ownership, we leverage data from Carney et al. (2020) tracing ownership for 238 sample firms between 2000 and
26
Interestingly, director skill appears to negatively influence accounting performance (Table A7) but positively affect stock performance (Table
A8). Because a large proportion of companies are family or state-owned, experienced directors are likely to be loyal to the dominant owner rather
than outside shareholders. Insofar as a crisis magnifies trade-offs between insiders and outsiders, such loyalty can contribute to the suboptimal
allocation of resources, leading in turn to a decline in ROA. At the same time, positive stock performance may be due to mock compliance with the
“letter of the law” regarding director qualifications, but not the “spirit of the law” in terms of the abovementioned incentives (Walter, 2008).
27
Thailand is an exception, with a national election on 15 December 2008. Importantly, Table A13 of the Online Appendix shows that excluding
Thailand from our main analysis does not meaningfully affect results.
16
2009. We estimate a variant of eq. 4 with annual ownership outcomes (defining 2008/09 as the crisis). Similar to Table A10, we
estimate the crisis impact on (i) an indicator for family or state blockholders; (ii) the first-difference of that indicator; and (iii) an
indicator for whether any change to the presence of family or state blockholders occurred. Just as in Table A10, for each outcome we
alternate country, industry, and country-industry fixed effects. In untabulated results we find the crisis to be an insignificant de-
terminant of family and state ownership, across all the above specifications.28 Taken together, the results of this section suggest
family, state, and political networks in East Asia were not meaningfully affected by the crisis.
6. Conclusion
This paper introduces a novel concept of board network interactions. By gathering unique data across 1290 East Asian firms, we
identify companies with board interlocks to networks characterized by family business groups, state ownership, and political con-
nections. We find stronger crisis performance among companies with board interlocks to family and state-owned firms. Based on
several indicators of institutional strength, we find board interlocks to politically connected firms most beneficial when institutions
are weak.
Our results are based on data from East Asia, but the prevalence of business networks in other regional contexts makes our
findings of general interest. For example, interlocking directorates are commonplace in Latin America (Cardenas, 2014), Europe (Van
Veen and Kratzer, 2011) and the United States (Larcker et al., 2013). Faccio (2006) finds widespread evidence of political con-
nections among shareholders and top corporate officers in a study of 47 countries. And Khanna and Yafeh (2007) document the
prevalence of family business groups around the world. In effect, wherever these business networks coexist, our findings are of
relevance.
From our study we draw some managerial and policy implications. Our findings suggest firm networks can result in higher trade
credit, and the benefits of networks are greater under high levels of debt maturity. This implies managers would gain from identi-
fying, cultivating, and maintaining director ties to organizations capable of modifying finance or trade credit arrangements.
Policymakers may target firms lacking networks as they are more vulnerable to underperform during a crisis. This is particularly
concerning if those firms offer products or services of national strategic importance (e.g., drug manufacturing, financial services). If
the government seeks a level playing field, it should ensure equitable access to information on support from regulatory agencies or
state-owned organizations. Maximal transparency about which policies or programs are likely to benefit or harm companies/in-
dustries in the near future would reduce the disproportionate benefits conferred to firms with access to privileged information via
director networks.
Acknowledgements
The authors would like to acknowledge very helpful feedback from Professor Cumming and three anonymous referees during the
review process. For constructive comments and advice at various stages of the project, we thank Remco Oostendorp, Chris Elbers,
Peter Lanjouw, Yishay Yafeh, Eric Bartelsman, Matthew Jackson, Adam Szeidl, Takeo Hoshi, Sweder van Wijnbergen, Mario Schabus,
Nick Longford, Zacharias Sautner, Simas Kucinskas, Phong Ngo, Melissa Chunmei Lin, and two anonymous referees. We also thank
seminar participants at the Tinbergen Institute, the China Europe International Business School (CEIBS), HEC Montreal, the
Universitat Pompeu Fabra (UPF), and the CEPR-BREAD-PODER Conference at the London School of Economics. This paper was partly
funded under the grant "Policy Design and Evaluation Research in Developing Countries" Initial Training Network (PODER), under
the Marie Curie Actions of the EU's Seventh Framework Programme (Contract Number: 608109).
Appendix A. Appendix
Table A1
Sources for politicians data.
Country Sources
Panel A: Parliament Members

Hong Kong Legislative Council www.elections.gov.hk/elections/legco2004
Hong Kong Election Committee Me- webb-site.com/dbpub
mbers
Indonesia www.mpr.go.id/profil/anggota
Japan www.weblio.jp/wkpja/content
South Korea Source file emailed to us by the National Assembly
Malaysia National Assembly psephos.adam-carr.net
Malaysia Senate www.parlimen.gov.my
Philippines House www.congress.gov.ph/download/archives/mem\_14th.pdf
Philippines Senate en.wikipedia.org/wiki/14th\_Congress\_of\_the\_Philippines
Singapore www.parliament.gov.sg/history/11th-parliament
(continued on next page)
28
Interested readers may contact the authors for tabulated results.
17
Table A1 (continued)
Country Sources
Taiwan www.ly.gov.tw/en/03\_leg/legList.action
Thailand www.senate.go.th/inforcenter/documents/infosection23\_1.pdf
Panel B: Ministers
All countries except Hong Kong United States Central Intelligence Agency. Various years. Chiefs of State and Cabinet Members of Foreign Governments.
Washington, DC.
Hong Kong Executive (Secretaries) webb-site.com/dbpub
Table A2
Networks and cost of debt.
(1) (2) (3) (4) (5) (6)
Board network × crisis −0.00931*

(0.00513)
Family network × crisis −0.00622* 0.0162
(0.00347) (0.0214)
State network × crisis 0.00270 0.0417
(0.0371) (0.0653)
Political network × crisis −0.225 −0.245
(0.215) (0.235)
Other network × crisis −0.00498 −0.00693
(0.00447) (0.00475)
(0.00454)
(0.00432) (0.0209)
State network −0.0159 −0.0473
(0.0325) (0.0652)
Political network 0.193 0.214
(0.213) (0.231)
Other network −0.00401* −0.00216
(0.00239) (0.00285)
Observations 4461 4461 4461 4461 4461 4461

R-squared 0.039 0.039 0.039 0.055 0.039 0.057
assembled by the authors. Dependent variable is interest rate measured by interest expense divided by total debt. Control variables are leverage,
firm size, firm age and debt maturing within one year divided by total assets (following Adams et al., 2004; Masulis and Mobbs, 2011; Duchin, Ozbas
and Sensoy, 2010; and Harford et al., 2014). Explanatory variables of interest are various measures of networks (defined in Sections 2.1–2.3),
interacted with the crisis period to capture the crisis-contingent network effect. Network data are cross-sectional for 2008. Observations run from
second quarter 2007 until first quarter 2009, leaving a panel of eight quarters. All specifications include country-industry fixed effects, as well as
crisis-period country-industry shocks. Each specification contains eight quarterly lags of the dependent variable. The standard errors reported in
parentheses are clustered by country-industry (***p < 0.01, **p < 0.05, *p < 0.1).
Table A3
Networks and sales growth.
(1) (2) (3) (4) (5) (6)

(0.0756)
Family network × crisis 0.551 0.768
(0.599) (0.684)
State network × crisis 0.0898 0.544
(0.415) (0.702)
Political network × crisis −1.703 −2.565
(1.627) (1.832)
Other network × crisis −0.00699 −0.0363
(0.0184) (0.0377)
(0.0753)
18
(1) (2) (3) (4) (5) (6)

(0.633) (0.711)
State network −0.727 −1.441
(0.566) (1.034)
Political network 1.469 2.339
(1.604) (1.723)
Other network 0.00715 0.0118
(0.0299) (0.0301)
Observations 5527 4523 4523 4523 4523 4523

R-squared 0.072 0.112 0.112 0.112 0.112 0.112
assembled by the authors. Dependent variable is sales growth. Explanatory variables of interest are various measures of networks (defined in
Sections 2.1–2.3), interacted with the crisis period to capture the crisis-contingent network effect. Network data are cross-sectional for 2008.
Observations run from second quarter 2007 until first quarter 2009, leaving a panel of eight quarters. All specifications include country-industry
fixed effects, as well as crisis-period country-industry shocks. Each specification contains eight quarterly lags of the dependent variable. The
standard errors reported in parentheses are clustered by country-industry (***p < 0.01, **p < 0.05, *p < 0.1).
Table A4
Country distributions for full sample and datastream subsample.
Country Full Sample Datastream Subsample
Frequency Percent Frequency Percent
Hong Kong 133 10.31 26 12.62

Indonesia 169 13.1 5 2.43
Japan 126 9.77 117 56.8
Korea 133 10.31 19 9.22
Malaysia 281 21.78 8 3.88
Philippines 98 7.6 1 0.49
Singapore 116 8.99 24 11.65
Taiwan 107 8.29 3 1.46
Thailand 127 9.84 3 1.46
Total 1290 100 206 100
This table reports distributions for our main sample of firms (N = 1290) and the Datastream subsample (N = 206). Datastream subsample consists
of all sample firms for which director qualifications data are available.
Table A5
Director qualifications, networks, and ROA.
(1) (2) (3) (4) (5)
Family network × crisis 0.0274 0.0295 0.0250 0.0302 0.0346

(0.0407) (0.0405) (0.0431) (0.0406) (0.0435)
State network × crisis 0.821* 0.778* 0.694* 0.825* 0.674*
(0.454) (0.449) (0.385) (0.458) (0.396)
Political network × crisis −0.190* −0.188* −0.120 −0.188* −0.103
(0.107) (0.106) (0.0758) (0.107) (0.0746)
Other network × crisis −0.0368 −0.0423 −0.0149 −0.0349 −0.0142
(0.0313) (0.0345) (0.0331) (0.0327) (0.0365)
Skill × crisis 3.324 10.41
(6.609) (10.27)
Skill 5.282* 5.940*
(2.810) (3.495)
Experience × crisis 24.48 12.75
(28.75) (27.10)
Experience −3.679 −11.98
(20.98) (25.24)
cv × crisis −43.50* −84.71
(24.23) (55.73)
cv −5.567 −0.757
(8.129) (19.23)
19
(1) (2) (3) (4) (5)
Network controls √ √ √ √ √
Firm controls √ √ √ √ √
Observations 1352 1153 990 1347 907
R-squared 0.758 0.760 0.748 0.759 0.752
Data are for 206 firms across nine East Asian economies. Financial indicators and director qualifications data are from Worldscope and Datastream.
Network data are assembled by the authors. Dependent variable is quarterly return on assets (ROA) expressed in percentage points. Explanatory
variables of interest are various measures of networks (defined in Sections 2.1–2.3) and director qualifications (including skill, CV, and experience),
interacted with the crisis period. Skill: percentage of board members who have either an industry-specific or a strong financial background. cv: total
number of board members with publicly disclosed professional backgrounds. Experience: average number of years each board member has been on
the board. Observations run from second quarter 2007 until first quarter 2009, leaving a panel of eight quarters. We suppress network and firm
controls to save space. All specifications include country-industry fixed effects, as well as crisis-period country-industry shocks. Each specification
contains eight quarterly lags of the dependent variable. The standard errors in parentheses are clustered by country-industry (***p < 0.01,
**p < 0.05, *p < 0.1).
Table A6
Director qualifications, networks, and stock returns.
(1) (2) (3) (4) (5)
Family network × crisis 0.0182*** 0.0172*** 0.0198*** 0.0187*** 0.0201***

(0.00587) (0.00586) (0.00651) (0.00587) (0.00606)
State network × crisis 0.0817** 0.0778** 0.0732* 0.0846** 0.0739*
(0.0382) (0.0382) (0.0388) (0.0385) (0.0400)
Political network × crisis −0.0286** −0.0298** −0.0309** −0.0277** −0.0324**
(0.0118) (0.0120) (0.0138) (0.0117) (0.0132)
Other network × crisis −0.00634 −0.00917 −0.00734 −0.00585 −0.0111
(0.00566) (0.00639) (0.00765) (0.00561) (0.00832)
Skill × crisis 0.308 0.106
(0.652) (1.015)
Skill 0.0340 0.232
(0.209) (0.244)
Experience × crisis 5.079 5.762
(3.844) (3.633)
Experience −1.975 −2.721
(1.784) (1.880)
cv × crisis −2.413 −3.851
(1.695) (3.449)
cv −1.033 0.486
(0.792) (2.579)
Observations 4081 3411 2972 4057 2641

R-squared 0.075 0.071 0.074 0.076 0.073
Data are for 206 firms across nine East Asian economies. Stock price and director qualifications data are from Datastream. Network data are
assembled by the authors. Dependent variable is average daily stock returns. Explanatory variables of interest are various measures of networks
(defined in Sections 2.1–2.3) and director qualifications (including skill, CV, and experience), interacted with the crisis period. Skill: percentage of
board members who have either an industry-specific or a strong financial background. cv: total number of board members with publicly disclosed
professional backgrounds. Experience: average number of years each board member has been on the board. Control variables are firm characteristics
including ROE, leverage, stock price volatility and firm size. Network data are cross-sectional for 2008. Observations run from February 2007 to
March 2009 (inclusive), leaving a panel of twenty-six months. We suppress network and firm controls to save space. All specifications include
country-industry fixed effects, as well as a crisis-period dummy. The standard errors in parentheses are clustered by country-industry (***p < 0.01,
**p < 0.05, *p < 0.1).
20
Table A7
Director qualifications, corporate affiliations, and ROA.
(1) (2) (3) (4) (5)
Affiliations × crisis 0.531** 0.503* 0.510* 0.532** 0.522*

(0.259) (0.278) (0.258) (0.258) (0.281)
Affiliations 0.00259 0.00200 0.00316 0.00255 0.00269
(0.00214) (0.00203) (0.00220) (0.00214) (0.00204)
Skill × crisis −17.43* −10.92
(9.263) (9.823)
Skill 0.691 3.170
(9.494) (9.845)
Experience × crisis −194.8*** −182.4***
(43.75) (43.58)
Experience −50.39 −68.11*
(42.48) (40.74)
cv × crisis −0.839 −13.92
(50.58) (52.00)
cv 42.23 45.66
(41.82) (43.36)
Observations 3509 3330 3470 3505 3290

R-squared 0.421 0.420 0.426 0.421 0.426
Data are for 734 firms across nine East Asian economies. Data come from Worldscope and Datastream. Dependent variable is quarterly return on
assets (ROA) expressed in percentage points. Explanatory variables of interest are corporate affiliations and director qualifications (including skill,
CV, and experience), interacted with the crisis period. Skill: percentage of board members who have either an industry-specific or a strong financial
background. cv: total number of board members with publicly disclosed professional backgrounds. Experience: average number of years each board
member has been on the board. Observations run from second quarter 2007 until first quarter 2009, leaving a panel of eight quarters. All speci-
fications include country-industry fixed effects, as well as crisis-period country-industry shocks. Each specification contains eight quarterly lags of
the dependent variable. The standard errors in parentheses are clustered by country-industry (***p < 0.01, **p < 0.05, *p < 0.1).
Table A8
Director qualifications, corporate affiliations, and stock returns.
(1) (2) (3) (4) (5)
Affiliations × crisis 0.0341*** 0.0165*** 0.0153*** 0.0124*** 0.00980***

(0.00610) (0.00441) (0.00433) (0.00456) (0.00366)
Affiliations −0.00310** −0.00130 −0.000687 −0.000581 −0.000290
(0.00123) (0.00111) (0.00129) (0.00105) (0.00117)
Skill × crisis 0.433*** 0.129
(0.0534) (0.108)
Skill 0.154*** 0.195***
(0.0437) (0.0465)
Experience × crisis 3.665*** 1.432**
(0.507) (0.703)
Experience −0.691*** −0.442*
(0.244) (0.253)
cv × crisis 3.231*** 1.726**
(0.385) (0.721)
cv −0.102 −0.137
(0.268) (0.284)
Observations 17,070 17,018 16,914 17,044 16,836

R-squared 0.018 0.022 0.021 0.022 0.023
Data are for 720 firms across nine East Asian economies. Data come from Datastream. Dependent variable is average daily stock returns. Explanatory
variables of interest are corporate affiliations and director qualifications (including skill, CV, and experience), interacted with the crisis period. Skill:
percentage of board members who have either an industry-specific or a strong financial background. cv: total number of board members with
publicly disclosed professional backgrounds. Experience: average number of years each board member has been on the board. Control variables are
firm characteristics including ROE, leverage, stock price volatility and firm size. Network data are cross-sectional for 2008. Observations run from
February 2007 to March 2009 (inclusive), leaving a panel of twenty six months. All specifications include country-industry fixed effects, as well as a
crisis-period dummy. The standard errors reported in parentheses are clustered by country-industry (***p < 0.01, **p < 0.05, *p < 0.1).
21
Table A9
Crisis intensity moderating the impact of networks on firm performance.
(1) (2) (3) (4) (5)
GDP consumption investment import export
Family network × crisis × crisis intensity 0.0461** 0.0846** 0.0406 0.0786* 0.00171
(0.0179) (0.0424) (0.0422) (0.0428) (0.0408)
State network × crisis × crisis intensity 0.0323** 0.0667*** 0.0404** 0.0446*** 0.0476***
(0.0151) (0.0177) (0.0174) (0.0129) (0.0121)
Political network × crisis × crisis intensity 0.0675 0.206* 0.0694 0.0500 0.115
(0.145) (0.113) (0.188) (0.152) (0.127)
Other network × crisis × crisis intensity −0.0449 −0.0698* −0.0340 −0.0802 −0.00743
(0.0327) (0.0362) (0.0520) (0.0489) (0.0420)
Family network × crisis 0.0438 0.0444 0.0462 0.124* 0.0348
(0.0310) (0.0298) (0.0424) (0.0677) (0.0403)
State network × crisis 0.281 0.271* 0.341 0.178 0.0510
(0.172) (0.158) (0.209) (0.208) (0.203)
Political network × crisis −0.0439 −0.0460 −0.0461 −0.0433 −0.0565
(0.105) (0.0537) (0.168) (0.0963) (0.0459)
Other network × crisis −0.0369 −0.137 −0.214 0.106 0.268
(0.236) (0.249) (0.259) (0.267) (0.275)
Observations 6990 6990 6990 6990 6990

R-squared 0.550 0.550 0.550 0.550 0.550
Data are for 1178 firms across nine East Asian economies. Financial indicators come from Worldscope and Datastream. Network data are assembled
by the authors. Dependent variable is quarterly return on assets (ROA) expressed in percentage points. We set crisis intensity equal to one if the
corresponding growth rate (in GDP, private consumption, investment, imports, or exports) is less than or equal to the median growth rate for
2008–09, and zero otherwise. Growth rates are from the IMF World Economic Outlook Database. Observations run from Q2 2007 until Q1 2009,
leaving a panel of eight quarters. All specifications include country-industry fixed effects, as well as crisis-period country-industry shocks. Each
specification contains eight quarterly lags of the dependent variable. The standard errors reported in parentheses are clustered by country-industry
(***p < 0.01, **p < 0.05, *p < 0.1).
Table A10
Crisis impact on corporate affiliations.
(1) (2) (3)
Panel A: Levels
Crisis 0.0210 0.00520 0.0145
(0.0136) (0.0148) (0.0105)
Fixed effect industry country industry × country
R-squared 0.265 0.281 0.642
Panel B: Change
Crisis 0.0258 0.0274 0.0273
(0.0194) (0.0202) (0.0198)
R-squared 0.036 0.004 0.071
Panel C: Change Indicator

Crisis 0.178 0.177 0.182
(0.170) (0.168) (0.173)
R-squared 0.047 0.030 0.062
Data are for 817 firms across nine East Asian economies. Corporate affiliations data come from Datastream. Dependent variable is a
board network proxy - the average number of external corporate affiliations among board members. Crisis is a dummy indicator for
Q4 2008 and Q1 2009. The estimated model is Nijkt = β crisist + fixed effect + εijkt. Panel A (B) outcome is expressed in terms of
levels (first-differences). Panel C outcome is an indicator for whether a change to corporate affiliations took place between t-1 and t.
Observations run from second quarter 2007 until first quarter 2009, leaving a panel of eight quarters. Columns are differentiated by
the level of fixed effects aggregation. The standard errors reported in parentheses are clustered by country-industry (***p < 0.01,
**p < 0.05, *p < 0.1).
22
Appendix B. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.jcorpfin.2020.101630.
References
Acemoglu, Daron, Johnson, Simon, Kermani, Amir, Kwak, James, Mitton, Todd, 2016. The value of connections in turbulent times: evidence from the United States. J.
Financ. Econ. 121 (2), 368–391.
Adams, Renee, Almeida, Heitor, Ferreira, Daniel, 2004. Powerful CEOs and their impact on corporate performance. Rev. Financ. Stud. 18 (4), 1403–1432.
Aivazian, Varouj A., Ge, Ying, Qiu, Jiaping, 2005. Can corporatization improve the performance of state-owned enterprises even without privatization? J. Corp. Finan.
11 (5), 791–808.
Almeida, Heitor, Park, Sang Yong, Subrahmanyam, Marti, Wolfenzon, Daniel, 2011. The structure and formation of business groups: evidence from Korean Chaebols.
J. Financ. Econ. 99 (2), 447–475.
Almeida, H., Campello, M., Laranjeira, B., Weisbenner, S., 2012. Corporate debt maturity and the real effects of the 2007 credit crisis. Crit. Financ. Rev. 1 (1), 3–58.
Anderson, Ronald C., Reeb, David M., 2003. Founding family ownership and firm performance: evidence from the S&P 500. J. Financ. 58 (3), 1301–1328.
Bae, Kee-Hong, Kang, Jun-Koo, Kim, Jin-Mo, 2002. Tunneling or value added? Evidence from mergers by Korean business groups. J. Financ. 57 (6), 2695–2740.
Baek, Jae-Seung, Kang, Jun-Koo, Park, Kyung Suh, 2004. Corporate governance and firm value: evidence from the Korean financial crisis. J. Financ. Econ. 71 (2),
265–313.
Baek, Jae-Seung, Kang, Jun-Koo, Lee, Inmoo, 2006. Business groups and Tunneling: evidence from private securities offerings by Korean Chaebols. J. Financ. 61 (5),
2415–2449.
Berkman, Henk, Cole, Rebel A., Fu, Lawrence J., 2010. Political connections and minority-shareholder protection: evidence from securities-market regulation in China.
J. Financ. Quant. Anal. 45 (6), 1391–1417.
Bertrand, Marianne, Mehta, Paras, Mullainathan, Sendhil, 2002. Ferreting out Tunneling: An application to Indian business groups. Q. J. Econ. 117 (1), 121–148.
Bertrand, M., Kramarz, F., Schoar, A., Thesmar, D., 2018. The cost of political connections. Rev. Financ. 22 (3), 849–876.
Bizjak, John, Lemmon, Michael, Whitby, Ryan, 2009. Option backdating and board interlocks. Rev. Financ. Stud. 22 (11), 4821–4847.
Bunkanwanicha, Pramuan, Wiwattanakantang, Yupana, 2009. Big business owners in politics. Rev. Financ. Stud. 22 (6), 2133–2168.
Bunkanwanicha, Pramuan, Fan, Joseph P.H., Wiwattanakantang, Yupana, 2013. The value of marriage to family firms. J. Financ. Quant. Anal. 48 (2), 611–636.
Cameron, Colin, Miller, Douglas L., 2015. A Practitioner’s guide to cluster-robust inference. J. Hum. Resour. 50 (2), 317–373.
Campbell, John, 1996. Understanding Risk and Return. J. Polit. Econ. 104 (2), 298–345.
Cardenas, Julian, 2014. Are Latin America’s corporate elites transnationally interconnected? A network analysis of interlocking directorates. Global Netw. 15 (4),
424–445.
Carney, Michael, 2005. Globalization and the renewal of Asian business networks. Asia Pac. J. Manag. 22 (4), 337–354.
Carney, Richard W., Child, Travers B., 2013. Changes to the ownership and control of east Asian corporations between 1996 and 2008: the primacy of politics. J.
Financ. Econ. 107 (2), 494–513.
Carney, Michael, Dieleman, Marleen, Sachs, Wladimir, 2008. The value of social capital to family Enterprises in Indonesia. In: Phan, Phillip, Venkataraman, Sankaran,
Velamuri, S. Ramakrishna (Eds.), Entrepreneurship in Emerging Regions Around the World: Theory, Evidence and Implications. Edward Elgar Publishers,
Cheltenham.
Carney, Richard W., Child, Travers B., Liu, Waiman, Ngo, Phong, 2020. The Dynamism of Partially State-Owned Enterprises in East Asia. (Working Paper).
Chekir, Hamouda, Diwan, Ishac, 2013. Crony Capitalism in Egypt. (CID Working Paper).
Chen, G., Firth, M., Xu, L., 2009. Does the type of ownership control matter? Evidence from China’s listed companies. J. Bank. Financ. 33 (1), 171–181.
Cingano, Federico, Pinotti, Paolo, 2013. Politicians at work: the private returns and social costs of political connections. J. Eur. Econ. Assoc. 11 (2), 433–465.
Claessens, Stijn, Djankov, Simeon, Lang, Larry H.P., 2000. The separation of ownership and control in east Asian corporations. J. Financ. Econ. 58 (1–2), 81–112.
Cohen, Lauren, Frazzini, Andrea, Malloy, Christopher, 2008. The small world of investing: board connections and mutual fund returns. J. Polit. Econ. 116 (5),
951–979.
Cohn, Jonathan B., Wardlaw, Malcolm I., 2016. Financing constraints and workplace safety. J. Financ. 71 (5), 2017–2058.
Dacin, Tina, Delios, Andrew, 2005. Editors’ introduction: the special issue on networks in the Asia Pacific. Asia Pac. J. Manag. 22 (4), 315–320.
Danielson, Morris G., Scott, Jonathan A., 2004. Bank loan availability and trade credit demand. Financ. Rev. 39 (4), 579–600.
Dass, Nishant, Kini, Omesh, Nanda, Vikram, Onal, Bunyamin, Wang, Jun. 2014. Board expertise: do directors from related industries help bridge the information gap?
Rev. Financ. Stud. 27 (5), 1533–1592.
Davis, Gerald, 1991. Agents without principles? The spread of the poison pill through the Intercorporate network. Adm. Sci. Q. 36 (4), 583–613.
Deloof, Marc, Jegers, Marc, 1999. Trade credit, corporate groups, and the financing of Belgian firms. J. Bus. Financ. Acc. 26 (7–8), 945–966.
Dinc, I. Serdar, 2005. Politicians and banks: political influences on government-owned banks in emerging markets. J. Financ. Econ. 77 (2), 453–479.
Duchin, Ran, Sosyura, Denis, 2012. The politics of government investment. J. Financ. Econ. 106, 24–48.
Duchin, R., Ozbas, O., Sensoy, B.A., 2010. Costly external finance, corporate investment, and the subprime mortgage credit crisis. Journal of financial economics 97
(3), 418–435.
Durnev, A., Errunza, V., Molchanov, A., 2009. Property rights protection, corporate transparency, and growth. J. Int. Bus. Stud. 40 (9), 1533–1562.
Erkens, David, Hung, Mingyi, Matos, Pedro, 2012. Corporate governance in the 2007-2008 financial crisis: evidence from financial institutions worldwide. J. Corp.
Finan. 18 (2), 389–411.
Faccio, Mara, 2006. Politically connected firms. Am. Econ. Rev. 96 (1), 369–386.
Faccio, Mara, Masulis, Ronald, McConnell, John, 2006. Political connections and corporate bailouts. J. Financ. 61 (6), 2597–2635.
Faleye, Olubunmi, Kovacs, Tunde, Venkateswaran, Anand, 2014. Do better-connected CEOs innovate more? J. Financ. Quant. Anal. 49 (5–6), 1201–1225.
Ferguson, Thomas, Voth, Hans-Joachim, 2008. Betting on Hitler-the value of political connections in Nazi Germany. Q. J. Econ. 123 (1), 101–137.
Fisman, Raymond, 2001. Estimating the value of political connections. Am. Econ. Rev. 91 (4), 1095–1102.
Freeman, Linton C., 1978. Centrality in social networks conceptual clarification. Soc. Networks 1 (3), 215–239.
Garcia-Teruel, Pedro J., Martinez-Solano, Pedro, 2010. A dynamic perspective on the determinants of accounts payable. Rev. Quant. Finan. Acc. 34 (4), 439–457.
Goldman, Eitan, Rocholl, Jorg, So, Jongil, 2009. Do politically connected boards affect firm value? Rev. Financ. Stud. 22 (6), 2331–2360.
Goldman, Eitan, Rocholl, Jorg, So, Jongil, 2013. Political connections and the allocation of procurement contracts. Rev. Financ. 13 (5), 1617–1648.
Gomez, Edmund Terence, Jomo, K.S., 1999. Malaysia’s Political Economy: Politics, Patronage and Profits. Cambridge University Press, New York, NY.
Gopalan, Radhakrishnan, Nanda, Vikram, Seru, Amit, 2007. Affiliated firms and financial support: evidence from Indian business groups. J. Financ. Econ. 86 (3),
759–795.
Gorton, Gary, Rosen, Richard, 1995. Corporate control, portfolio choices, and the decline of banking. J. Financ. 50 (5), 1377–1420.
Goyal, Amit, Santa-Clara, Pedro, 2003. Idiosyncratic risk matters!. J. Financ. 58 (3), 975–1007.
Greif, Avner, Tabellini, Guido, 2010. Cultural and institutional bifurcation: China and Europe compared. Am. Econ. Rev. 100 (2), 135–140.
Hamilton, Gary (Ed.), 1996. Asian Business Networks. Walter de Gruyter, Berlin.
Harford, Jarrad, Klasa, Sandy, Maxwell, William F., 2014. Refinancing risk and cash holdings. J. Financ. 69 (3), 975–1012.
Hillman, Amy J., Cannella, Albert A., Paetzold, Ramona L., 2000. The resource dependence role of corporate directors: strategic adaptation of board composition in
23
response to environmental change. J. Manag. Stud. 37 (2), 235–256.

Hillman, Amy J., Withers, Michael C., Collins, Brian J., 2009. Resource dependence theory: a review. J. Manag. 35 (6), 1404–1427.
Huyghebaert, N., Van de Gucht, L., Van Hulle, C., 2007. The choice between Bank debt and trace credit in business start-ups. Small Bus. Econ. 29 (4), 435–452.
Imai, Kenichi, 1988. The corporate network in Japan. Jpn. Econ. Stud. 16, 3–37.
Jensen, Michael, 1986. Agency costs of free cash flow, corporate governance, and takeovers. Am. Econ. Rev. 76 (2), 323–329.
Jia, Nan, Shi, Jing, Wang, Yongxiang, 2013. Coinsurance within business groups: evidence from related party transactions in an emerging market. Manag. Sci. 59 (10),
2295–2313.
Jiang, Chao, Kubick, Thomas R., Miletkov, Mihail K., Wintoki, M. Babajide, 2017. Offshore expertise for onshore companies: director connections to island tax havens
and corporate tax policy. Manag. Sci. 64 (7), 3241–3268.
Joh, Sung Wook, 2003. Corporate governance and firm profitability: evidence from Korea before the economic crisis. J. Financ. Econ. 68 (2), 287–322.
Johnson, Simon, Mitton, Todd, 2003. Cronyism and capital controls: evidence from Malaysia. J. Financ. Econ. 67 (2), 351–382.
Johnson, Simon, Boone, Peter, Breach, Alisdair, Friedman, Eric, 2000. Corporate governance in the Asian financial crisis. J. Financ. Econ. 58 (1), 141–186.
Jones, Carla D., Makri, Marianna, Gomez-Mejia, Luis R., 2008. Affiliate directors and perceived risk bearing in publicly traded, family-controlled firms: the case of
diversification. Entrepreneur. Theory Pract. 32 (6), 1007–1026.
De Jong, Abe, 2002. The disciplining role of leverage in Dutch firms. Eur. Financ. Rev. 6 (1), 31–62.
Karolyi, Andrew, Liao, Rose, 2017. State Capitalism’s global reach: evidence from foreign acquisitions by state-owned companies. J. Corp. Finan. 42, 367–391.
Khanna, Tarun, Palepu, Krishna, 1997. Why focused strategies may be wrong for emerging markets. Harv. Bus. Rev. 75 (4), 41–43.
Khanna, Tarun, Palepu, Krishna, 2002. Is group affiliation profitable in emerging markets? An analysis of diversified business groups. J. Financ. 55 (2), 867–891.
Khanna, Tarun, Yafeh, Yishay, 2007. Business groups in emerging markets: paragons or parasites? J. Econ. Lit. 45 (2), 331–372.
Khwaja, Asim Ijaz, Mian, Atif, 2005. Do lenders favor politically connected firms? Rent provision in an emerging financial market. Q. J. Econ. 120 (4), 1371–1411.
Kondo, Mari, 2014. Philippines. In: Witt, Michael, Redding, Gordon (Eds.), Oxford Handbook of Asian Business Systems. Oxford University Press, New York, NY.
Kroll, Mark, Walters, Bruce A., Le, Son A., 2007. The impact of board composition and top management team ownership structure on post-IPO performance in young
entrepreneurial firms. Acad. Manag. J. 50 (5), 1198–1216.
Lambert, Richard, Leuz, Christian, Verrechia, Robert, 2012. Information precision, information asymmetry, and the cost of capital. Rev. Financ. 16 (1), 1–29.
Lane, P.R., Milesi-Ferretti, G.M., 2011. The cross-country incidence of the global crisis. IMF Econ. Rev. 59, 77–110.
Larcker, David, So, Eric, Wang, Charles, 2013. Boardroom centrality and firm performance. J. Account. Econ. 55 (2–3), 225–250.
Lee, Moosung, Shim, Jae-Seung, 2012. The Korean Economic System: Government, Big Business and Financial Institutions. Routledge, London.
Lemmon, Michael L., Lins, Karl V., 2003. Ownership structure, corporate governance, and firm value: evidence from the east Asian financial crisis. J. Financ. 58 (4),
1445–1468.
Lester, Richard H., Hillman, Amy, Zardkoohi, Asghar, Jr, Cannella, Albert, A., 2008. Former government officials as outside directors: the role of human and social
capital. Acad. Manag. J. 51 (5), 999–1013.
Li, Hongbin, Meng, Lingsheng Wang, Qian, Zhou, Li-An., 2008. Political connections, financing and firm performance: evidence from Chinese private firms. J. Dev.
Econ. 87 (2), 283–299.
Lincoln, J., Shimotani, M., 2010. Business Networks in Postwar Japan: Whither the Keiretsu? In: The Oxford Handbook of Business Groups. Oxford University Press,
New York.
Lins, Karl V., Servaes, Henri, Tamayo, Ane, 2017. Social capital, trust, and firm performance: the value of corporate social responsibility during the financial crisis. J.
Financ. 72 (4), 1785–1824.
Masulis, Ronald, Mobbs, Shawn, 2011. Are all inside directors the same? CEO entrenchment or board enhancement. J. Financ. 66 (3), 823–872.
Masulis, Ronald, Pham, Peter Kien, Zein, Jason, 2011. Family business groups around the world: financing advantages, control motivations and organizational choices.
Rev. Financ. Stud. 24 (11), 3556–3600.
Megginson, William, Netter, Jeffry, 2001. From state to market: a survey of empirical studies on privatization. J. Econ. Lit. 39 (2), 321–389.
Mishkin, Frederic, 1996. Understanding Financial Crises: A Developing Country Perspective. NBER Working Papers 5600. National Bureau of Economic Research.
Mishkin, Frederic, 1997. The Causes and Propagation of Financial Instability: Lessons for Policy Makers. Proceedings - Economic Policy Symposium - Jackson Hole,
Federal Reserve Bank of Kansas City, pp. 55–96.
Mishkin, Federic, Hahm, Joon-Ho, 2000. Causes of the Korean Financial Crisis: Lessons for Policy. NBER Working Papers 7483. National Bureau of Economic
Research.
Morck, Randall, Wolfenzon, Daniel, Yeung, Bernard, 2005. Corporate governance, economic entrenchment, and growth. J. Econ. Lit. 43 (3), 655–720.
Morellec, Erwan, Schurhoff, Norman, 2011. Corporate investment and financing under asymmetric information. J. Financ. Econ. 99 (2), 262–288.
Morris, James R., 1976. On corporate debt maturity strategies. J. Financ. 31 (1), 29–37.
Myers, Stewart C., 1977. Determinants of corporate borrowing. J. Financ. Econ. 5 (2), 147–175.
Naughton, Barry, 2009. Loans, firms, and steel: is the state advancing at the expense of the private sector? China Leadership Monit. 30, 1–10.
Niskanen, Jerki, Niskanen, Myrvi, 2006. The determinants of corporate trade credit policies in a Bank-dominated financial environment: the case of Finnish small
firms. Eur. Financ. Manag. 12 (1), 81–102.
Orru, Marco, 1996. Practical and theoretical aspects of Japanese business networks. In: Hamilton, Gary (Ed.), Asian Business Networks. de Gruyter, Berlin, pp.
201–227.
Porta, La, Rafael, Lopez-De-Silanes, Shleifer, Andrei, 2006. What works in securities laws? J. Financ. 61 (1), 1–32.
Rademakers, Martijn, 1998. Market Organization in Indonesia: Javanese and Chinese family business in the Jamu industry. Organ. Stud. 19 (6), 1005–1027.
Rajan, Raghuram G., Zingales, Luigi, 1998. Financial dependence and growth. Am. Econ. Rev. 88 (3), 559–586.
Redding, Gordon, 1993. The Spirit of Chinese Capitalism. de Gruyter, New York, NY.
Redding, Gordon, 1996. Weak organizations and strong linkages: managerial ideology and Chinese family business networks. In: Hamilton, Gary (Ed.), Asian Business
Networks. Oxford University Press, New York, NY.
Shleifer, Andrei, Vishny, Robert W., 1994. Politicians and firms. Q. J. Econ. 109 (4), 995–1025.
Stearns, Linda Brewster, Mizruchi, Mark S., 1993. Board composition and corporate financing: the impact of financial institution representation on borrowing. Acad.
Manag. J. 36 (3), 603–618.
Stuart, Toby, Yim, Soojin, 2010. Board interlocks and the propensity to be targeted in private equity transactions. J. Financ. Econ. 97 (1), 174–189.
Suehiro, Akira, 1989. Capital Accumulation in Thailand, 1855–1985. Silkworms Books, Chiang Mai.
Suehiro, Akira, Wailerdsak, Natenapha, 2014. Thailand. In: Redding, Gordon, Witt, Michael (Eds.), Oxford Handbook of Asian Business Systems. Oxford University
Press, New York, NY.
Tian, Lihui, Estrin, Saul, 2007. Debt financing, soft budget constraints and government ownership: evidence from China. Econ. Transit. 15 (3), 461–481.
Turner, Sarah, 2007. Small-scale Enterprise livelihoods and social Capital in Eastern Indonesia: ethnic Embeddedness and exclusion. Prof. Geogr. 59 (4), 407–420.
Uzzi, Brian, 1999. Social Embeddedness in the creation of financial capital. Am. Sociol. Rev. 64 (4), 481–505.
Van Veen, Kees, Kratzer, 2011. National and international interlocking directorates within Europe: corporate networks within and among fifteen European countries.
Econ. Soc. 40 (1), 1–25.
Walter, Andrew, 2008. Governing Finance: East Asia’s Adoption of International Standards. Cornell University Press, Ithaca, NY.
Wolters, Oliver William, 1999. History, Culture, and Region in Southeast Asian Perspectives. Cornell University Press, Ithaca, NY.
24

Board Connections and Crisis Performance Family ST

Uploaded by

Copyright:

Available Formats

Board Connections and Crisis Performance Family ST

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Board Connections and Crisis Performance Family ST

Uploaded by

Copyright:

Available Formats

Journal of Corporate Finance 64 (2020) 101630

Contents lists available at ScienceDirect

Journal of Corporate Finance

Board connections and crisis performance: Family, state, and

ARTICLE INFO ABSTRACT

E-mail address: carney.richard@ceibs.edu (R.W. Carney).

1. Theory of network interactions

1.1. The importance of networks to Asian business

1.2. Board networks and information

1.3. Crisis performance and network interactions

1.3.1. Family networks

1.3.2. State networks

1.3.3. Political networks

2.1. Board connections

ROA 9426 1.91 4.11 −14.35 16.36

2.2. Family business groups, state ownership, and political connections

2.3. Network interactions

2.4. Firm characteristics

mean SD max mean SD max mean SD max mean SD max

Political network 0.3300*

3. Performance implications of board networks

3.1. Networks and accounting performance

3.1.1. Cross-sectional model

Board network 0.0800***

3.1.2. Dynamic model

3.2. Networks and stock returns

Board network × crisis 0.0313**

3.3. Political networks and institutions

Board network × crisis 0.00414**

Observations 22,845 22,845 22,845 22,845 22,845 22,845 19,634 17,880

GDP per capita Investor Protection Financial Development

Institutions × political network × crisis −0.177** −0.221** −0.234***

Observations 6990 6990 6990

yijkt = 0 Nijk Ct + 1 Nijk Ct Ik + 0 Nijk + 1 Nijk Ik + jk Ct + jk + yijk, t + t + ijkt

Board network × crisis 0.00182

Observations 1374 1374 1374 1374 1374 1374

5.1. Director qualifications

Debt maturity × board network × crisis 0.000157*

Observations 5717 5717 5717 5717 5717

5.2. Crisis intensity

5.3. Crisis impact on networks

Panel A: Parliament Members

(1) (2) (3) (4) (5) (6)

Board network × crisis −0.00931*

Observations 4461 4461 4461 4461 4461 4461

(1) (2) (3) (4) (5) (6)

Board network × crisis 0.0501

(1) (2) (3) (4) (5) (6)

Family network −0.631 −0.842

Observations 5527 4523 4523 4523 4523 4523

Country Full Sample Datastream Subsample

Frequency Percent Frequency Percent

Hong Kong 133 10.31 26 12.62

(1) (2) (3) (4) (5)

Family network × crisis 0.0274 0.0295 0.0250 0.0302 0.0346

(1) (2) (3) (4) (5)

(1) (2) (3) (4) (5)

Family network × crisis 0.0182*** 0.0172*** 0.0198*** 0.0187*** 0.0201***

Observations 4081 3411 2972 4057 2641

Institutions × political network × crisis −0.177 −0.221 −0.234***

Family network × crisis 0.0182* 0.0172* 0.0198* 0.0187* 0.0201***

Affiliations × crisis 0.0341* 0.0165* 0.0153* 0.0124* 0.00980***