INTRODUCTION
Acts of misconduct in business may undermine firms’ reputations (Kang, Reference Kang2008), dampen firms’ performance (Murphy, Shrieves, & Tibbs, Reference Murphy, Shrieves and Tibbs2009), jeopardize CEOs’ careers (Arthaud-Day, Certo, Dalton, & Dalton, Reference Arthaud-Day, Certo, Dalton and Dalton2006; Gomulya & Boeker, Reference Gomulya and Boeker2014; Semadeni, Cannella, Fraser, & Lee, Reference Semadeni, Cannella, Fraser and Lee2008), and spoil governmental relationships (Jonsson, Greve, & Fujiwara-Greve, Reference Jonsson, Greve and Fujiwara-Greve2009). After suffering the consequences, some firms may not commit acts of misconduct in the future. Unfortunately, many firms with records of misconduct (referred to as misdemeanants) commit similar misconduct in the future. Given these consequences, why do misdemeanants repeatedly commit similar acts of misconduct?
Although prior researchers in management and criminology have identified various drivers of recidivism at both the firm level (Baucus & Baucus, Reference Baucus and Baucus1997; Simpson & Koper, Reference Simpson and Koper1992; Zhou & Lu, Reference Zhou and Lu2016) and the individual level (Gendreau, Little, & Goggin, Reference Gendreau, Little and Goggin1996; Tangney, Stuewig, & Martinez, Reference Tangney, Stuewig and Martinez2014), they have largely assumed that misdemeanants are atomic entities and have overlooked the social interaction between misdemeanants and others. Misdemeanants interact with others through various social networks, which, in turn, will affect the misdemeanants’ perception of the costs and benefits of committing misconduct in the future (Stuart & Wang, Reference Stuart and Wang2016; Yiu, Xu, & Wan, Reference Yiu, Xu and Wan2014). Noting that social networks can help researchers to understand misconduct from a relational perspective, Simpson (Reference Simpson2013) has called on researchers to use social network perspective to study recidivism. However, limited attention, so far, has been paid to identifying the key drivers of the misdemeanant-to-recidivist process from a social network perspective.
To start to address this void, we employ social network theory to investigate what prevents misdemeanants from learning from prior acts of financial fraud, resulting in their becoming recidivists. Unlike prior research on recidivism that overlooks misdemeanants’ social interactions with others (Baucus & Baucus, Reference Baucus and Baucus1997; Simpson, Reference Simpson2013; Simpson & Koper, Reference Simpson and Koper1992; Zhou & Lu, Reference Zhou and Lu2016), we emphasize that misdemeanants are embedded in various social networks that affect the way they analyze the costs and benefits of recommitting financial fraud in the future (Maula, Keil, & Zahra, Reference Maula, Keil and Zahra2013). When the perceived costs outweigh the benefits, misdemeanants may avoid another financial fraud. In contrast, if the misdemeanants perceive that the benefits outweigh the costs, they are more likely to commit another fraud. Therefore, to open the black box of the misdemeanant-to-recidivist process, we propose a theoretical framework synergizing social network theory and cost–benefit analysis.
We emphasize that misdemeanants are embedded into two types of social networks, which affect the misdemeanant-to-recidivist process. Politically connected misdemeanants fear losing political capital if they commit future financial fraud, so they may perceive a higher cost of recommitting financial fraud in the future than misdemeanants without political ties. Interlock tie, as a key type of managerial tie, is the connection two firms have when they have in common one or more directors on their boards (Markóczy, Sun, Peng, & Ren, Reference Markóczy, Li Sun, Peng and Ren2013). Interlock ties are relevant to firm misconduct because such networks are composed of high-ranking executives who are responsible for various firm decisions, including financial fraud (Jiang, Reference Jiang2017; Sullivan, Haunschild, & Page, Reference Sullivan, Haunschild and Page2007). After committing financial fraud, misdemeanants may face various challenges, such as state-owned banks may retreat loans. Central misdemeanants can address the challenge via leveraging social capital to gain alternative financial support from interlock partners. So, a central misdemeanant may perceive lower cost of committing financial fraud, and more likely recommit financial fraud in future than peripheral one. Finally, densely connected subgroups are prevalent within social networks, and this kind of meso-level network structure, which is referred as a network community, is important for firms’ learning and strategic decision-making (Sytch & Tatarynowicz, Reference Sytch and Tatarynowicz2014). The punishment of interlock network community partners’ misconduct may drive misdemeanants to focus on the costs associated with a current financial fraud, thereby reducing the likelihood of their committing another.
Investigating the misdemeanant-to-recidivist process by integrating social network theory and cost-benefit analysis, we contribute to the current literature in two ways. First, prior research has examined several drivers of misconduct recidivism at both the firm and individual levels, such as sanction severity (Simpson & Koper, Reference Simpson and Koper1992), individual features (Tangney et al., Reference Tangney, Stuewig and Martinez2014), and firm misconduct experiences (Zhou & Lu, Reference Zhou and Lu2016). However, these studies largely assume that misdemeanants are atomic entities and overlook the role of social networks in recidivism. To help fill this lacuna, we integrate social network theory and cost-benefit analysis to open the black box of the misdemeanant-to-recidivist process. By doing so, we complement current knowledge about the drivers of corporate misconduct recidivism.
Second, we also contribute to literature in social network perspective, especially for those on the link between different types of ties, social network community, and firm behaviors (Jack, Reference Jack2005; Gulati, Sytch, & Tatarynowicz, Reference Gulati, Sytch and Tatarynowicz2012; Luo, Huang, & Wang, Reference Luo, Huang and Wang2012). These stream studies have mainly focused on firm behaviors, like recognizing opportunities, learning, and acquiring knowledge. By studying how social networks affect the possibility of financial misdemeanants recommitting another financial fraud, we have enriched the above stream research in terms of new firm behaviors.
THEORETICAL BACKGROUND AND HYPOTHESES
Financial Fraud Resurface: Cost-Benefit Analysis
Organizational misconduct, which is defined as ‘behavior in or by an organization that a social-control agent judges to transgress a line separating right from wrong’ (Greve, Palmer, & Pozner, Reference Greve, Palmer and Pozner2010: 56), has attracted researchers’ attention for its theoretical and practical value (Gomulya & Boeker, Reference Gomulya and Boeker2014; Jonsson et al., Reference Jonsson, Greve and Fujiwara-Greve2009; Murphy et al., Reference Murphy, Shrieves and Tibbs2009). As a prominent type of organizational misconduct, corporate financial fraud refers to ‘intentional misrepresentation of amounts or disclosures in the financial statements’ (Apostolou, Hassell, & Webber, Reference Apostolou, Hassell and Webber2000: 181). Financial fraud occurs when firms deceive investors or other stakeholders through false disclosure, profit inflation, failure to disclose material information, fictitious transactions, covering up systematic problems, and so on (Chen, Firth, Gao, & Rui, Reference Chen, Firth, Gao and Rui2006; Shi, Connelly, & Sanders, Reference Shi, Connelly and Sanders2016).
The cost and benefit is an essential framework to investigate the motivation of firms committing financial fraud (Mishina, Dykes, Block, & Pollock, Reference Mishina, Dykes, Block and Pollock2010; Zhang, Bartol, Smith, Pfarrer, & Khanin, Reference Zhang, Bartol, Smith, Pfarrer and Khanin2008). On the one hand, there may be benefits of committing financial fraud because it may boost a firm's financial performance in the short run, which, in turn, may increase decision makers’ personal wealth (Shi et al., Reference Shi, Connelly and Sanders2016; Zhang et al., Reference Zhang, Bartol, Smith, Pfarrer and Khanin2008). The benefits of financial fraud in the short run motivate decision makers to engage in such actions. On the other hand, financial fraud also inflicts various costs on firms, including undermining the firms’ legitimacy and spoiling relationship with government and state-owned banks (Jonsson et al., Reference Jonsson, Greve and Fujiwara-Greve2009; Kang, Reference Kang2008). Firms are more likely to commit financial fraud when decision makers perceive that the benefits outweigh the costs (Greve et al., Reference Greve, Palmer and Pozner2010; Mishina et al., Reference Mishina, Dykes, Block and Pollock2010). In contrast, if firms perceive that the costs outweigh the benefits, they are less likely to commit financial fraud. Hence, cost-benefit analysis is a relevant theoretical lens to view the misdemeanant-to-recidivist process.
Based upon cost-benefit analysis frame, scholars have examined the effects of firm performance, executive compensation, and external governance on organizational misconduct and financial fraud (Mishina et al., Reference Mishina, Dykes, Block and Pollock2010; Zhang et al., Reference Zhang, Bartol, Smith, Pfarrer and Khanin2008). However, this research produces relatively inconsistent findings on why firms commit financial fraud. For example, Zhang et al. (Reference Zhang, Bartol, Smith, Pfarrer and Khanin2008) found that firms are more likely to manipulate earnings when decision makers are unable to achieve goals through legitimate means, while Mishina et al. (Reference Mishina, Dykes, Block and Pollock2010) showed that high-performing firms are also more likely to engage in illegal activities due to decision makers’ loss aversion. Similarly, while Yiu, Wan, and Xu (Reference Yiu, Wan and Xu2018) found that external governance as one type of informal institution can deter corporate financial fraud, Shi, Connelly, and Hoskisson (Reference Shi, Connelly and Hoskisson2017) found that external governance may lead to financial fraud due to high expectations from stakeholders. One reason accounting for this inconsistency may be that prior researchers seemingly adopt a static perspective on firm misconduct and overlook the interdependence between these acts. Firms may commit financial misconduct more than once, and these acts are interrelated, rather than independent.
To dynamically explore financial fraud, we examine financial fraud resurface, which refers to the process where a financial misdemeanant commits another fraud in future. Financial fraud resurface hazard refers to the likelihood that a firm with one act of financial fraud will commit another one in the future, and emphasizes the connection between, and dynamics of, firms’ financial fraud. Such concepts will help us to open the black box between firms’ acts of misconduct, thereby deepening our understanding of the antecedents of corporate financial fraud. Additionally, these concepts will enhance our knowledge of the misdemeanant-to-recidivist process, which complements the current research about what drives firms to commit misconduct (Krishnan & Kozhikode, Reference Krishnan and Kozhikode2015; Stuart & Wang, Reference Stuart and Wang2016; Tang, Qian, Chen, & Shen, Reference Tang, Qian, Chen and Shen2015), and about recidivism (Baucus & Baucus, Reference Baucus and Baucus1997; Gendreau et al., Reference Gendreau, Little and Goggin1996; Simpson, Reference Simpson2013; Simpson & Koper, Reference Simpson and Koper1992; Tangney et al., Reference Tangney, Stuewig and Martinez2014; Zhou & Lu, Reference Zhou and Lu2016).
The likelihood of misdemeanants’ continually committing financial fraud is dependent on how they interpret and perceive the costs and benefits of committing another fraud in the future. When misdemeanants perceive the costs of committing another financial fraud to be high, they will more likely focus on the threats and therefore not recommit financial fraud (Yiu et al., Reference Yiu, Xu and Wan2014). In contrast, when misdemeanants perceive the benefits of committing another fraud to be high, they are more likely to interpret financial fraud as a good opportunity to gain illegal benefits and thus are more likely to commit another fraud in the future. What will affect misdemeanants’ perceptions of the costs and benefits of committing another financial fraud in the future?
Financial Misdemeanants’ Social Networks: Political Ties and Interlock Ties
Social networks may affect how they analyze the costs and benefits of committing another financial fraud, because social networks can shape misdemeanants’ perceptions and attention. Interrelationships with external organizations provide access to different viewpoints and may direct attention to different aspects of the organization's environment, allowing decision makers to focus their time and efforts on certain problems and solutions (Lawrence & Lorsch, Reference Lawrence and Lorsch1967; Ocasio, Reference Ocasio1997). Social networks offer communication channels through which firms interact (Uzzi, Reference Uzzi1997) and help firms to acquire information from external environments (Maula et al., Reference Maula, Keil and Zahra2013). De Carolis and Saparito (Reference De Carolis and Saparito2006) argue that social networks enhance entrepreneurial opportunities by triggering entrepreneurs’ perception over risk. Thus, interactions with external actors will shape entrepreneurs’ perceptions and attention to certain issues (Maula et al., Reference Maula, Keil and Zahra2013; Ocasio, Reference Ocasio1997). In this study, we suggest that misdemeanants’ perceptions on the costs and benefits of recommitting financial fraud may be influenced by two types of social networks: political ties and interlock ties.
Political ties refer to the guanxi and connection that misdemeanants have with government officials. Governments in emerging economies play important roles in economic activities by setting regulatory policies and controlling valuable resources (Peng & Luo, Reference Peng and Luo2000). Political ties provide firms with crucial access to policy information and scarce resources and function as a substitute for formal institutional support, thereby reducing policy uncertainty (Xin & Pearce, Reference Xin and Pearce1996). Therefore, because officials may avoid relationships with misdemeanants to protect their reputations and images, politically connected misdemeanants may be more likely to perceive the potential costs of losing political capital by recommitting financial fraud in the future.
Two firms are interlocked if they have one or more common directors on their boards, which allows the firms to access information, knowledge, and resources (Gulati & Westphal, Reference Gulati and Westphal1999; Markóczy et al., Reference Markóczy, Li Sun, Peng and Ren2013; Shipilov, Greve, & Rowley, Reference Shipilov, Greve and Rowley2010). Interlock ties, like the misdemeanant's connections and guanxi with other firms, represent key social capital for the misdemeanant. One essential and relevant feature in interlock networks is centrality, which refers to the extent to which the focal actor can directly connect with others in the network (Bonacich, Reference Bonacich1987; Grewal, Lilien, & Mallapragada, Reference Grewal, Lilien and Mallapragada2006; Pappas & Wooldridge, Reference Pappas and Wooldridge2007). Interlock network centrality reflects the firms’ status advantage in the network, which provides firms access to social capital (Shi, Sun, Pinkham, & Peng, Reference Shi, Sun, Pinkham and Peng2014). Social capital from interlock networks can also help firms to gain support when it is in trouble (Nahapiet & Ghoshal, Reference Nahapiet and Ghoshal1998). After committing financial fraud, misdemeanants face various costs that keep them from acquiring financial support from authority and government. Central misdemeanants can employ social capital to dilute the above costs, while peripheral misdemeanants cannot. So, a central misdemeanant may perceive lower cost of committing financial fraud, and be more likely to recommit financial fraud in the future than a peripheral one.
The other relevant feature in interlock networks is the behavior of the community partners. Network communities are dense, non-overlapping structural groups within a network. Nodes in a community are connected more to each other than to nodes outside the community (Knoke, Reference Knoke2009). Network community partners will pay attention to each other's behaviors in the same community (Clement, Shipilov, & Galunic, Reference Clement, Shipilov and Galunic2018). Clement et al. (Reference Clement, Shipilov and Galunic2018) argues that, due to the close connection of the community members, they have more interaction with, and are better able to observe the behavior of, their community partners compared with firms outside the community. Recent research has highlighted that the network community to which a firm belongs can affect the firm's knowledge acquisition and learning (Baum, Shipilov, & Rowley, Reference Baum, Shipilov and Rowley2003; Gulati et al., Reference Gulati, Sytch and Tatarynowicz2012; Sytch & Tatarynowicz, Reference Sytch and Tatarynowicz2014). Sytch and Tatarynowicz (Reference Sytch and Tatarynowicz2014) found that firms can improve invention productivity by exchanging knowledge and information with other firms within the same network community because of the dense connections among community members. In our research setting, our network includes all listed firms in China, and the network community is defined as a group of Chinese listed firms closely connected through interlock ties. Nodes in the network community include bad apples (misdemeanants) and good apples (those without financial fraud records). The punishment of financial fraud committed by community partners may drive misdemeanants to pay attention to financial fraud punishment and therefore perceive the costs of recommitting financial fraud in the future.
Misdemeanant-to-Recidivist Process: The Effect of Political Ties
Connections to government officials may affect the extent to which firms commit financial fraud (Peng & Luo, Reference Peng and Luo2000; Stuart & Wang, Reference Stuart and Wang2016). Stuart and Wang (Reference Stuart and Wang2016) propose that, because political ties can reduce the chance of being scrutinized by regulators, thereby allowing firms that commit fraud to emerge unscathed, politically connected firms are more likely to commit fraud. However, we argue that political ties may reduce the likelihood of misdemeanants recommitting financial fraud and therefore decrease the financial fraud resurface hazard.
First, misconduct has negative spillover effects, so governmental officials may try to avoid firms with records of fraud to protect their reputations, legitimacy, and status (Benjamin & Podolny, Reference Benjamin and Podolny1999; Jonsson et al., Reference Jonsson, Greve and Fujiwara-Greve2009; Kang, Reference Kang2008). Government officials are more likely to avoid misdemeanants, or will cut relationships with them, because reputations and images are fundamental to their political careers. In addition, governments in emerging economies, such as China, play an important role in economic activities by setting regulatory policies and controlling resources (Peng & Luo Reference Peng and Luo2000). Considering the significant costs of losing political capital in an emerging country, politically connected misdemeanants may be more likely than those without political ties to perceive the costs and risk of committing another financial fraud in the future.
Second, after committing financial fraud, the misdemeanant's future decisions will be closely monitored by the public, the media, and regulators (Kang, Reference Kang2008). Officials who have connections to the misdemeanants may also be thrust into the limelight. From the perspective of the government officials, it may be unwise to continue to protect the misdemeanants during this sensitive time. Therefore, the best choice for these officials may be to avoid the misdemeanants instead of protecting them. Officials may also encourage the misdemeanants to behave well in the future to protect their own images and legitimacy.
Finally, the Chinese Stock Exchanges started in 1997, and therefore, their history is relatively short. Laws regarding illegal financial acts are ambiguous and equivocal, which may impair the ability of misdemeanants to investigate the laws if they wish to. However, politically connected misdemeanants may have better access to information about the laws, allowing them to investigate more effectively and reducing the likelihood of their committing another fraud in the future. Thus, we hypothesize the following:
Hypothesis 1:
Political ties will be negatively related to the misdemeanant's financial fraud resurface hazard.
Misdemeanant-to-recidivist process: The effect of interlock networks
A misdemeanant's centrality in the interlock network is positively related to the financial fraud resurface hazard. First, interlock network centrality reflects the firms’ status, which provides firms access to social capital (Shi et al., Reference Shi, Sun, Pinkham and Peng2014). Firm status is largely a function of social capital that its directors can bring to bear on its strategic challenges. Social capital from interlock networks can also facilitate firms to gain help when it confronts challenges, and is directly linked to firms’ ability to secure resources from external actors (Nahapiet & Ghoshal, Reference Nahapiet and Ghoshal1998). Therefore, high status can strengthen a firm's intention to engage in behaviors that an external audience might view negatively. Phillips and Zuckerman (Reference Phillips and Zuckerman2001) have found that high status may boost the possibility of actors engaging in behaviors that deviate from accepted business norms. Cowen and Marcel (Reference Cowen and Marcel2011) have found that firms with high status will be more likely to keep the directors who have involved misconduct than those in moderate status.
In our study, a central misdemeanant with high status can access social capital through interlock networks, which can help the misdemeanant engage in problems after committing financial fraud. For instance, financial fraud keeps misdemeanants from acquiring critical financial support from authority and state-own banks. Ren, Au, and Birtch (Reference Ren, Au and Birtch2009) find that in China, social network centrality will boost loan guarantee ties between firms, including receiving loan guarantees from other firms and giving loan guarantees to other firms. Central misdemeanants can employ interlock networks to acquire financial support, thereby diluting the above-mentioned cost. Also, many partners may choose to endorse the central misdemeanant, and warrant their financial support (Nahapiet & Ghoshal, Reference Nahapiet and Ghoshal1998), because they still rely on the central misdemeanant's resources and knowledge. In addition, social network centrality can serve a signal to outsiders about a senders’ true quality (Podolny, Reference Podolny2001). Ozmel, Reuer, and Gulati (Reference Ozmel, Reuer and Gulati2013) found that new ventures’ social network centrality can serve as signal to other organizations, like research institutions, about its high capability, thereby more likely forming strategic alliance with them. Even if interlocked partners, after committing financial fraud, may be unable to help central misdemeanants, high interlock centrality as a positive signal can help misdemeanants to acquire needed support from other sources. So, a central misdemeanant may perceive lower cost of committing financial fraud, and more likely to recommit financial fraud in future than a peripheral one.
Second, prior success and high status may engender hubris, which leads managers to believe in their own infallibility and become more risk seeking (Hayward & Hambrick, Reference Hayward and Hambrick1997). Mishina et al. (Reference Mishina, Dykes, Block and Pollock2010) imply that hubristic managers tend to believe they can outsmart regulatory authorities and avoid detection of their illegal behaviors, thus increasing the likelihood that they will engage in corporate illegality. In our study, high status in interlock networks may breed hubris in decision makers who are misdemeanants. These individuals tend to ignore the costs and risks of financial fraud and pay more attention to the potential benefits and opportunities of recommitting financial fraud in the future. Therefore, central misdemeanants in interlock networks may be more likely than peripheral misdemeanants to focus on the benefits of recommitting financial fraud and therefore will be more likely to recommit financial fraud.
Chinese unique political and business environments may account for the distinct mechanisms through which misdemeanant's political ties and interlock ties affect financial fraud resurface. Chinese political environments, such as media report, central government, and public sensation, are harsh for corrupted officials. Staying away from the stigmatized is vital for Chinese officials to obtain political capital and promotions (White, Reference White2005). In contrast, because the Chinese financial market is still developing (Peng & Luo, Reference Peng and Luo2000), the definition of financial fraud may be vague and ambiguous to interlocked partners. Also, interlocked ties can benefit interlocked partners with key business resources (Nahapiet & Ghoshal, Reference Nahapiet and Ghoshal1998). Hence, interlocked partners may still connect with central financial misdemeanants. In short, in Chinese, setting political ties are more sensitive to financial frauds than interlock ties.
Thus, we hypothesize the following:
Hypothesis 2:
Interlock network centrality will be positively related to the misdemeanant's financial fraud resurface hazard.
Misdemeanant-to-Recidivist Process: The Effect of Interlock Network Communities
The punishment of financial frauds committed by interlock network community partners will guide misdemeanants’ attention to the costs of financial frauds, thereby reducing the likelihood of their committing another fraud in the future. First, the network community is a densely connected group, and partners in the community pay great attention to each other (Knoke, Reference Knoke2009). Clement et al. (Reference Clement, Shipilov and Galunic2018) argue that, due to the close connection of the community members, they are better able to observe the behavior of their community partners than they are others who are not in the community. The financial fraud punishment of interlock network community partners may lead misdemeanants to perceive the potential threat and costs of recommitting financial frauds. When network community partners are punished, misdemeanants may expect the same punishment if they commit another financial fraud in the future (Bandura, Reference Bandura, London and Rosenham1968, Reference Bandura1977). Given the high expected cost of financial fraud, misdemeanants will be deterred from committing another fraud. Yiu et al. (Reference Yiu, Xu and Wan2014) observed that a firm is deterred from committing misconduct if its peers in the same industry are punished.
Moreover, when interlock network community partners are punished for financial fraud, the community members may accumulate experience and knowledge about such misconduct and communicate this knowledge to their partners (Sytch & Tatarynowicz, Reference Sytch and Tatarynowicz2014). By acquiring relevant experience and knowledge from punished community partners, misdemeanants may effectively investigate, learn from, and explore current financial fraud, thus becoming more likely to avoid committing another fraud in the future. Therefore, we hypothesize the following:
Hypothesis 3:
The punishment of the financial frauds of interlock network community partners will be negatively related to the misdemeanant's financial fraud resurface hazard.
METHODS
Sample and Data
To test our theories, we collected all listed firms in the Chinese Shanghai and Shenzhen Stock Exchanges that committed financial fraud from 2005 to 2014. Several reasons make our sample appropriate. First, regulators from the Shanghai and Shenzhen Stock Exchanges will have investigated and presented the details of listed firms that have committed financial fraud, which will provide us with an appropriate empirical context to examine whether those financial misdemeanants will commit another fraud in the future. Second, the Chinese Stock Exchanges started in 1997 with 720 listed firms. Therefore, the Chinese Stock Exchanges are in nascent stages, and their institutional environments are rather dynamic and flexible. Additionally, China has a well-documented institutional void, and social networks can work as an alternative for formal institutions (Peng & Luo, Reference Peng and Luo2000; Xin & Pearce, Reference Xin and Pearce1996). Such an institutional context will facilitate the testing of how social networks affect financial fraud resurface.
We collected data from two sources. First, we gathered data regarding Chinese listed firms’ financial fraud and firm characteristics data from the China Stock Market & Accounting Research Database (CSMAR). The financial fraud related data included the date of the financial fraud, the fine, and the punishment of community members. Firm characteristics included corporate governance, firm performance, and firm size. Second, since external institutions are relevant to firms’ misconduct, we collected external institution indexes from the National Economic Research Institution (NERI) (Fan, Wang, & Zhu, Reference Fan, Wang and Zhu2011). Several studies have suggested that multiple data sources can assist researchers to explore their questions from a more comprehensive aspect and avoid perception-perception bias (Scandura & Williams, Reference Scandura and Williams2000).
Our dataset includes 1197 financial misdemeanants that committed 2382 financial fraud from January 1st 2005 to December 31st 2014. Specifically, 643(53.71%) of them committed one financial fraud, 263 (21.97%) of them committed two financial frauds, and 291(24.32%) of them committed more than two financial frauds.
Measurement
Financial fraud dormancy
Financial fraud dormancy refers to the time that misdemeanants will spend between two consecutive financial frauds. We used the number of days between two consecutive misdemeanants’ financial frauds to measure financial fraud dormancy. We began following misdemeanants from January 1st 2005 (which is the earliest day of the time range) and recorded the date of each financial fraud event for each misdemeanant. We measured the number of days between two consecutive financial frauds or the number of days until December 31st 2014 (the end of our study period) if the event is a misdemeanant's last financial fraud during our study period. This approach is called right-censored in the terminology of event history studies.
Misdemeanants’ political ties
To measure misdemeanants’ political ties, we took into account China's institutional features and defined a politically connected firm as one with a board chairperson or CEO who is currently holding or previously held a position in either the central government, local government, National People's Congress, local people's congress, Chinese People's Political Consultative Conference, or local people's political consultative conference (Sun, Mellahi, Wright, & Xu, Reference Sun, Mellahi, Wright and Xu2015; Zheng, Singh, & Mitchell, Reference Zheng, Singh and Mitchell2015). To be specific, we encoded misdemeanants’ political ties into an ordinal variable to capture the different levels of political connection status. A leading/assisting role of townships or equivalents is coded as 1, a leading/assisting role of counties or equivalents is coded as 2, a leading/assisting role of prefectures or equivalents is coded as 3, a leading/assisting role of provinces or equivalents is coded as 4, and a leading/assisting role of state or equivalents is coded as 5. If the CEO and board chair have no work experience in governmental agencies, this variable is coded as 0. As a robust test, we also dichotomized this variable into 0 and 1, and the results still hold.
Interlock network construction
To calculate the misdemeanants’ interlock network characteristics, we first constructed a global interlock network among all Chinese listed firms for each year. Using the affiliated relationship between all board directors and all listed firms, we built a two-mode network. Next, we converted this two-mode network into a one-mode network among all listed firms. The network construction is programmed in R.
Misdemeanants’ interlock network centrality
Several studies have suggested that eigenvector centrality is an effective indicator of the power and influence that a node has within its network (Grewal et al., Reference Grewal, Lilien and Mallapragada2006; Pappas & Wooldridge, Reference Pappas and Wooldridge2007). Eigenvector centrality of a node is proportional to the sum of the eigenvector centrality of all other nodes connected to it in the network (Bonacich, Reference Bonacich1987).
Let A = (a ij)n×n represent the adjacency matrix of the interlock network, where a ij = 1 if node i is connected with node j and a ij = 0 otherwise. According to Bonacich (Reference Bonacich1987), the eigenvector centrality x i of node i can be defined by its neighbors’ eigenvector centrality as the following formula:
$$x_i = \displaystyle{1 \over \lambda} \mathop \sum \limits_{t\in M\lpar i \rpar } x_t$$where M(i) is the set of all the nodes directly connected with node i, and λ is a constant. Using the matrix notation, the eigenvector centrality x of all the nodes in the interlock network satisfies the following equation:
$$ {\bi Ax} \equals \lambda {\bi x}$$This equation means that the eigenvector centrality is the eigenvector of the adjacency matrix A associated with a particular eigenvalue λ.
Punishment of misdemeanants’ interlock network community partners
To measure the punishment of misdemeanants’ interlock network community partners, we first used the fast greedy algorithm to divide the global interlock network into communities (Clauset, Newman, & Moore, Reference Clauset, Newman and Moore2004). Figure 1 shows the community structure of the interlock networks in 2014 for illustration. To keep the figure simple, we excluded communities whose size is less than 65. The points on the circle represent different firms, and firms within the same community are grouped together on the circle (in Figure 1, a same color of points represent an interlock network community). The lines in the circle represent the interlock ties between firms.
Figure 1. Interlock network communities of Chinese listed firms in 2014.
Second, based on the interlock network community, we measured the punishment of misdemeanants’ interlock network community partners as the number of nodes in the community that have been punished due to financial fraud one year before the misdemeanants’ current financial fraud.
Controls
We also controlled several variables that may affect the financial fraud resurface hazard, including misdemeanants’ punishment for the financial fraud (Barnett, Reference Barnett2014; Sampath, Rahman, & Gardberg, Reference Sampath, Rahman and Gardberg2013), misdemeanants’ organizational features (Harris & Bromiley, Reference Harris and Bromiley2007; Schnake & Williams, Reference Schnake and Williams2008), and misdemeanants’ formal institutions (Groysberg, Lin, & Serafeim, Reference Groysberg, Lin and Serafeim2016). First, punishment can impose financial cost on misdemeanants and affect their expectations of financial cost derived from future punishment. So, we controlled the punishment for misdemeanants’ current financial fraud with the natural logarithm of the fine amount. Second, we controlled misdemeanants’ features that may influence the recurrence of financial fraud. Firms with better economic performance may attract more attention from authorities and investors (Marquis & Bird, Reference Marquis and Bird2018), thereby increasing the possibility of being caught if they commit another financial fraud. We measured misdemeanants’ economic performance with ROA and the development capability with the growth rate of net income. Additionally, we controlled for misdemeanants’ age (which is the number of years since misdemeanants’ establishment) and size (which is measured by misdemeanants’ employee number). Firm age and firm size may affect firms’ speed to respond to environment change due to inertia (Hannan & Freeman, Reference Hannan and Freeman1984; Kelly & Amburgey, Reference Kelly and Amburgey1991), so new and small misdemeanants may take short time to change (if they are willing to) their ill routines that lead to financial fraud, and thereby avoiding future misconducts. To correct for the highly skewed nature of the number of employees, we measured the misdemeanant's size using the natural logarithm of the number of employees. TMT size is the number of managers in the top management team. Independent directors can serve as key watchdog of firm behaviors, especially those harming investors (Chen, Firth, Gao, & Rui, Reference Chen, Firth, Gao and Rui2006). So, we controlled the number of independent directors of misdemeanants. The ownership structure is another important governance factor, especially supervision share ratio and state share ratio. Supervision share ratio and state share ratio may motivate supervisors of board and Chinese authorities to scrutinize firm unethical behaviors more closely (Firth, Rui, & Wu, Reference Firth, Rui and Wu2011). Therefore, we included the supervision share ratio and the state share ratio to capture the influence of the ownership structure on financial fraud resurface hazard. CEO duality may trigger agency issues because CEOs’ power is lack of supervision, which may lead to unethical decisions (Firth et al., Reference Firth, Rui and Wu2011). So, we controlled for CEO duality of misdemeanants, which is coded as 1 if the CEO is also the chairperson of the board; otherwise, it is coded as 0. Misdemeanant's alters may also commit financial frauds, thereby affecting the alters’ capability to provide financial supports to focal misdemeanant, which may affect focal misdemeanant's willingness of recommitting another financial misconduct in future. So, we included the ratio of misdemeanant's alters that also committed financial fraud, which equals the number of alters that committed financial fraud divided by the total number of alters. Finally, firms’ behaviors are embedded in and shaped by institutions (Yiu et al., Reference Yiu, Wan and Xu2018), so we controlled the misdemeanant's formal institutions with its province's marketization index. This index has been widely used in many studies to capture the qualities of the regional institutional environment (Shi, Sun, & Peng, Reference Shi, Sun and Peng2012; Sun, Hu, & Hillman, Reference Sun, Hu and Hillman2016).
Analytical Methodology
One way to model whether misdemeanants may commit another financial fraud is the OLS or logistic regression. Two primary drawbacks are embedded in such traditional regression methods. First, ordinary regressions cannot easily incorporate changes in the value of explanatory variables over time. Creating financial fraud dummy variables for every day for every misdemeanant (up to 10 years) would be highly cumbersome and may cause multicollinearity. Second, no satisfactory way exists to cope with right-censored cases using the OLS (Staw & Hoang, Reference Staw and Hoang1995). Since the observational window following the current financial fraud is finite, no response may be observed during the observational window. Conducting a logistic regression on a categorical dependent variable that distinguishes those misdemeanants that perpetrate another financial fraud from those that do not will retain information in two categories. However, logistic regression cannot incorporate the memory effect of the occurrence of the current financial fraud. The memory effect (measured by length of the dormancy period between the two consecutive financial frauds) is particularly important in our study, since we expect that two consecutive acts of misconduct are interconnected and interdependent.
EHA can resolve the two drawbacks by not modeling the response time, but rather by modeling the hazard rate (the likelihood that a response will be observed at time t, given that no response occurred prior to time t) (Yu & Cannella, Reference Yu and Cannella2007). First, EHA can effectively answer our research question regarding the effects of misdemeanants’ social networks on the financial fraud resurface hazard. Since EHA models a conditional likelihood, it is not biased by right censoring (Staw & Hoang, Reference Staw and Hoang1995). Second, given a hazard function, we can directly calculate the financial fraud dormancy, which is the expected number of days between two consecutive financial frauds. A significant increase in the hazard rate can be directly interpreted as a significant decrease in financial fraud dormancy. This finding will help us to unveil the connection between the financial fraud resurface hazard and financial fraud dormancy in one model.
Given the two advantages of the EHA, we used the continuous-time EHA to model the financial fraud resurface hazard. Our analyses were derived from the Cox proportional hazards regression models (Cox, Reference Cox1972; Katila & Shane, Reference Katila and Shane2005; Ozmel & Guler, Reference Ozmel and Guler2015) in the following form:
$$\ln \lsqb {h_i\lpar t \rpar } \rsqb = \ln \lsqb {h_0\lpar t \rpar } \rsqb + \mathop \sum \nolimits^ \beta _k \times \lsqb {X_{ik}\lpar t \rpar } \rsqb ,\;$$where h 0(t) is the baseline hazard function, and X ik(t) is the value of the kth covariate (independent variable) for firm i at the time of the current financial fraud t. A key benefit of the Cox model is that no assumptions are made as to the distribution of the baseline hazard function. We also ran the PH test, and the results show that the Cox model is appropriate. All regressions were processed by STATA 14.0 with stcox. To engage the tied failures, we employed Efron approximation calculation (Efron, Reference Efron1977; Hertz-Picciotto & Rockhill, Reference Hertz-Picciotto and Rockhill1997). We also used other types of approximation calculation to test the robustness of our findings.
Addressing Endogeneity
Sample selection bias
Sample selection bias is a systematic error due to a non-random sample of a population, which causes some members of the population to be less likely to be included than others, resulting in a biased sample (Bock, Opsahl, George, & Gann, Reference Bock, Opsahl, George and Gann2012; Heckman, Reference Heckman1979). Before testing any hypotheses, we first considered sample selection bias by testing which types of firms are likely to commit financial fraud. To test for sample selection bias, we created a dummy variable for all listed firms in the Chinese Shanghai and Shenzhen Stock Exchanges that committed financial fraud from 2005 to 2014. The dummy variable takes 1 if a firm has committed a financial fraud and 0 otherwise. We estimated a binary logit model to check which types of firms are likely to commit financial fraud and are therefore included in our sample as misdemeanants. The results of the sample selection bias test are presented in Table 1, showing that the variables relating to firm characteristics are insignificant. Hence, sample section bias is not a serious concern in our study.
Table 1. Sample selection bias test
Notes: 1. Year dummies were included.
2. DV: Firm financial fraud (binary logit model).
Reverse causality bias
Reverse causality bias refers to the case of when arguing that the variance of X gives rise to the variance of Y, we can also see plausible arguments that the direction is from Y to X. We address this concern by timely lagging the consequence, misdemeanants committing another financial fraud, behind the hypothesized causes, misdemeanants’ social networks (Xiao & Tsui, Reference Xiao and Tsui2007).
Addressing Modeling Dependence
We engaged in three types of modeling dependence (space, time, and firm), since observations are not independent in our sample (Marquis & Qian, Reference Marquis and Qian2013). Spatial correlation occurs when firms are located in the same province, and they may have opportunities to compete against or cooperate with each other. Thus, the firms could be dependent. Second, since we used panel data from 2005 to 2014, firms in the same year may have familiar external environments, such as governmental policies, industry dynamics, and even global competition. Hence, firms in the same year may be dependent. Finally, firm correlation emerges when some firms have committed financial fraud more than twice, meaning that an observation may appear more than once in our sample. These observations are dependent. To control for these three types of autocorrelation, we ran EHA models with clustering robust standard errors on firm, province, network, and year levels, and then compared their effects. The results show that each approach produces similar results. Thus, we have appropriately approached modeling dependence.
RESULTS
Table 2 provides the means, standard deviations, and pairwise correlations for all variables used in our analysis. The correlation of misdemeanants’ political ties, misdemeanants’ interlock network centrality, and the punishment of misdemeanants’ interlock network community partners are related at relatively low levels (r < 0.22). The highest correlation is between TMT size and firm size (r = 0.36). These results suggest that few collinear variables exist (r > 0.80) (Kennedy, Reference Kennedy2003). We also tested the VIFs for all models. The mean VIFs are 1.51, which is below the proposed threshold value of 10 (Marquis & Qian, Reference Marquis and Qian2013). In short, multicollinearity is not a serious issue.
Table 2. Descriptive statistics
Note: N = 1700. Correlations above 0.04 are significant at 0.05 or below.
Table 3 presents the results of the Cox regressions. Model 1 provides a baseline Cox model with only control variables. Before testing our hypotheses, we found some interesting results. First, financial misdemeanant's profitability is negatively related with its financial fraud resurface hazard (β = –0.074, p = 0.000). To make sense the coefficient of cox models, it should be transformed into a percentage by using 100[exp(β) - 1]. In terms of continuous variables, this formula shows the percentage change in the hazard rate if independent variable varies one-unit (Staw & Hoang, Reference Staw and Hoang1995). Based upon this method, we find that financial fraud resurface hazard will decrease by 7.13% if financial misdemeanant's profitability increases by one unit, all else remaining equal. This may be because when misdemeanants are developing in good shape, they will perceive more cost if they soon commit another financial fraud. Second, marketization index of the province where financial misdemeanants locate is negatively related with financial fraud resurface hazard (β = –0.058, p = 0.034). Financial fraud resurface hazard will decrease by 5.63% if marketization index increases by one unit, all else remaining equal. This may be because those financial misdemeanants in province with high marketization will confront quick and severe punishment, and this will increase their perceived cost of committing another in future.
Table 3. Results of Cox regression of financial fraud resurface hazard
Notes: 1. +p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001.
2. Models report coefficients, robust standard errors, and p values.
3. Nine year-dummies are included, but not reported here.
However, some control variables, such as CEO duality, independent director number, and state share ratio, which are suggested to have effect on financial fraud in prior research (Chen et al., Reference Chen, Firth, Gao and Rui2006; Firth et al., Reference Firth, Rui and Wu2011), did not show significant results in this research. One possible reason is that prior studies mainly focused on the link between these variables and single financial fraud, rather than repeated financial frauds. Committing and recommitting financial frauds may be affected by CEO duality and other governance control variables in different ways. For example, in terms of single financial fraud, CEO duality represents as a governance structure to help firms to detect possible misconducts, while CEO duality may not significantly affect misdemeanants’ evaluation on the cost and benefit of recommitting financial fraud in future. Another interesting finding is that penalty cannot significantly reduce financial fraud resurface in our research setting. The possible reason is that Chinese authority's punishment on financial misconduct is mostly symbolic, or too soft from misdemeanants’ perspective. Finally, the possible reason for insignificant results of firm age and firm size may be that such two variables have complicated or even contradicted theoretical meanings. On the one hand, firm age and firm size can be viewed as index of organization inertia (Hannan & Freeman, Reference Hannan and Freeman1984; Kelly & Amburgey, Reference Kelly and Amburgey1991). On the other hand, such two variables can also be viewed organization resource (Shan, Reference Shan1990; Thornhill & Amit, Reference Thornhill and Amit2003). Organization inertia and resource may affect financial fraud resurface hazard in different directions. So, we suspect the complicated theoretical meanings account for their insignificant results.
Hypothesis 1 states that misdemeanants’ political ties are negatively related with its financial fraud resurface hazard. The result of Model 2 shows that misdemeanant's political ties are negatively related with its financial fraud resurface hazard (β = –0.092, p = 0.008). Financial fraud resurface hazard will decrease by 8.79% if misdemeanants’ political ties increase by one unit, all else remaining equal. Therefore, Hypothesis 1 was supported.
Hypothesis 2 states that misdemeanants’ interlock network centrality is positively related with its financial fraud resurface hazard. The result of Model 2 shows that misdemeanants’ interlock network centrality is positively related with its financial fraud resurface hazard (β = 0.041, p = 0.078). Financial fraud resurface hazard will increase by 4.19% if misdemeanants’ interlock network centrality increases by one unit, all else remaining equal. Therefore, Hypothesis 2 was supported.
Hypothesis 3 states that the punishment of misdemeanants’ interlock network community partners is negatively related with the financial fraud resurface hazard. The result of Model 2 shows that the punishment of misdemeanants’ interlock network community partners is negatively related with its financial fraud resurface hazard (β = –0.136, p = 0.005). Financial fraud resurface hazard will decrease by 12.71% if the punishment of misdemeanants’ interlock network community partners increases by one unit, all else remaining equal. Therefore, Hypothesis 3 was supported.
Robust Tests
The main effect retest
Tie failures mean that at least two firms commit the second financial fraud at the same time, which will, in turn, significantly increase the calculation time. To address the problem, we also used approximation calculation proposed by Breslow (Reference Breslow1974) and exact-marginal calculation. The results support Hypothesis 1 and Hypothesis 3, but only partially support Hypothesis 2. To be specific, cox model with Breslow calculation shows that misdemeanant's political ties are negatively related with its financial fraud resurface hazard (β = –0.093, p = 0.008), misdemeanants’ interlock network centrality is positively related with its financial fraud resurface hazard (β = 0.042, p = 0.078), and the punishment of misdemeanants’ interlock network community partners is negatively related with its financial fraud resurface hazard (β = –0.136, p = 0.005). Model with exact-marginal calculation shows that misdemeanant's political ties are negatively related with its financial fraud resurface hazard (β = –0.093, p = 0.072), misdemeanants’ interlock network centrality is positively related with its financial fraud resurface hazard (β = 0.049, p = 0.215), and the punishment of misdemeanants’ interlock network community partners is negatively related with its financial fraud resurface hazard (β = –0.136, p = 0.065). We also employed other proportional hazard functions, including the exponential regression, the Gompertz regression, and the Weibull regression. These results also support our main findings.
Cluster in different levels
We also ran the Cox models with different levels, including firm, province, network, and year levels. All of the results support Hypothesis 1 and Hypothesis 3 but only partially support Hypothesis 2.
Overall, our data support Hypothesis 1 and Hypothesis 3 but only partially support Hypothesis 2.
DISCUSSION
Theoretical Implications
By answering why misdemeanants repeatedly perpetrate financial fraud, we have made two major contributions to the literature. First, we have contributed to management and criminology research regarding misconduct recidivism in general (Baucus & Baucus, Reference Baucus and Baucus1997; Simpson, Reference Simpson2013; Simpson & Koper, Reference Simpson and Koper1992; Gendreau et al.,1996; Tangney et al., Reference Tangney, Stuewig and Martinez2014; Zhou & Lu, Reference Zhou and Lu2016). One the one hand, prior management research on recidivism mainly assumes misdemeanants are atomic entities. Synergizing social network perspective and cost-benefit analysis, we have built a theoretical framework to open the black box of how a misdemeanant becomes a recidivist: the misdemeanant-to-recidivist process. Our findings show that politically connected misdemeanants are less likely to recommit financial fraud, while misdemeanants with interlock network centrality are more likely to commit financial fraud. In addition, misdemeanants are less likely to recommit financial fraud when partners in the interlock network community are punished for financial fraud. Our findings have moved our understanding of management recidivism from an atomic perspective to a relational perspective. On the other hand, most prior criminology studies on recidivism have focused on individual offenders (Gendreau et al., Reference Gendreau, Little and Goggin1996; Tangney et al., Reference Tangney, Stuewig and Martinez2014). Considering the different natures of individual and cooperate crimes, Simpson (Reference Simpson2013) has called for more research on ‘knowledge about explanatory characteristics and firm-level variation in offending over time’ (319). Our findings on the drivers of firm financial fraud recidivism not only respond to this eager call but also enrich relevant criminology studies on recidivism (Baucus & Baucus, Reference Baucus and Baucus1997; Simpson & Koper, Reference Simpson and Koper1992).
Second, we also contribute to literature in social network perspective. One growing research in social networks is to identify how different types of ties will affect the behaviors of individuals, groups, and firms, such as weak tie and strong tie (Jack, Reference Jack2005; Ma, Huang, & Shenkar, Reference Ma, Huang and Shenkar2011), and political tie and managerial tie (Luo et al., Reference Luo, Huang and Wang2012). For instance, Luo et al. (Reference Luo, Huang and Wang2012) employed meta-analysis to study the different links, managerial tie, political tie, and firm performance. Our findings suggest that political tie and interlock tie as one type of managerial tie will affect repeated financial frauds in different ways. This finding will enrich the research focusing on how different types of ties affect firm behaviors differently. In addition, another growing research stream in social network is network community and organization behaviors. Research has highlighted that the network community in which the firm is embedded can affect the firm's knowledge acquisition, learning, and strategic behaviors (Baum et al., Reference Baum, Shipilov and Rowley2003; Gulati et al., Reference Gulati, Sytch and Tatarynowicz2012; Sytch & Tatarynowicz, Reference Sytch and Tatarynowicz2014). Our findings suggest that social network community will also affect the possibility of financial misdemeanants to commit another financial fraud in future. This finding also enriches current research on the link between network community and firm behaviors.
Practical Implications
One interesting finding is that the authorities’ punishment of current misconduct may not reduce the likelihood of misdemeanants committing further misconduct in the future (β = −0.004, p = n.s.). This finding contradicts common sense regarding the effectiveness of the authority's punishment of misdemeanants. Punishment may fail not because deterrence is not effective, but because the misdemeanants’ networks may shape their perceptions of the costs and benefits of recommitting misconduct in the future. In addition to punishment, the authorities are also encouraged to understand what kinds of punishment will shape the misdemeanants’ perceptions of the misconduct's costs and benefits, thereby reducing the likelihood of committing misconduct in the future. Therefore, our findings may spur a novel idea for authorities to manage misdemeanants and prevent them from becoming recidivists.
In addition, managing misconduct (including pollution, corporate social irresponsibility, and financial fraud) is essential for firm development and social welfare, but merely knowing what will drive firms to commit such acts of misconduct is insufficient. We must know why misdemeanants do not learn from prior misconduct and what drives misdemeanants to repeatedly make the same mistakes. Our research on misconduct resurface may help managers understand why they are unable to learn from current acts of misconduct but instead commit them repeatedly. Repeated misconduct dampens firms’ reputations and undermines social welfare and justice. Our findings show that social networks shape firms’ perceptions and expectations about future misconduct and therefore the likelihood of committing similar mistakes in the future. Understanding the drivers of and reasons for the misdemeanant-to-recidivist process can help managers handle current crises more effectively. To be specific, in China, building political connection with governmental officials will facilitate firms to monitor their unethical behaviors. Therefore, our findings unveil another positive function of political ties in terms of unethical behaviors governance, and we suggest firms to pay more attention to political ties in China.
Last but not least, our study also helps authorities of China and other developing economies to detect and predicate various repeated misconduct. Based upon our findings, not all misdemeanants, which are even punished by authorities, will stop committing another misconduct in future. Those without political ties, in interlock network centrality, and community partners without similar punishment will be more likely or sooner committing another similar misconduct in future. Considering the limited attention and resource that can be deployed to supervise misdemeanants, our findings can help to narrow authorities’ targets to keep eye on. This may boost authorities’ efficiency of supervising misconducts.
Limitations and Future Research Directions
Our results need to be interpreted considering the study's limitations. First, we do not differentiate between types of financial fraud, such as delayed release of financial information, material financial information omission, violation of the laws related to buying and selling stocks, and accounting transgressions. Different results when firms have committed the same type of financial fraud could affect the findings. Thus, future research is encouraged to retest our theories with specific types of financial fraud.
Second, we only study one type of managerial network, the interlock network, in the misdemeanors-to-recidivist process, which may limit the theory boundary. In addition to interlock networks, we encourage future research to study other types of managerial networks, such as the buyer-suppliers network and manager's formal and informal networks. Testing whether our theory holds in other managerial network settings will complement our research.
Finally, studies on misconduct resurface are in the nascent stage, and more work must be undertaken. In addition to social networks, what else will jeopardize misdemeanants learning from current acts of misconduct, resulting in transition to recidivists? Will different types of misconduct affect firms’ ability to learn? Because governmental punishment of misconduct may fail to prevent misconduct resurface, what other policies could authorities employ? Research from the institutional perspective on this issue would be rather promising. Studying the joint effects of social networks and punishment on misdemeanants’ current misconduct will also contribute to this research. Also, authoritarian authorities, including China, Singapore, Malaysia, and a number of African and Latin American countries, have been said to have great control on firms (Levitsky & Way, Reference Levitsky and Way2010; Marquis & Bird, Reference Marquis and Bird2018). Punishment on misconduct is a key way through which authoritarian authorities control firms. However, it is still unclear why authoritarian authorities fail to control firms considering their great power on firms. Viewing misconduct resurface as authoritarian authority’ control failure is another promising research direction.
CONCLUSION
Given the consequences of misconduct, why do certain misdemeanants repeatedly perpetrate fraud? Integrating social network theory and cost-benefit analysis, we investigate the misdemeanant-to-recidivist process. We find that interlock network centrality prevents misdemeanants from learning from current financial fraud, resulting in a high risk of recommitting in the future. In contrast, political ties and the punishment of network community partners will reduce the risk of recommitting fraud in the future.
