Are all forms of ownership prone to tunneling? A meta ‐ analysis

Research Question/Issue: This study seeks to extend agency theory in the context of tunneling by exploring how different owner types seek private benefits of control. Spe-cifically, we examine how information asymmetries and board representation create different pressures for tunneling for state ‐ owned, business group ‐ owned, and family ‐ owned firms. We tested our hypotheses with a meta ‐ analytic structural equation model. Research Findings/Insights: Our findings show that the relationship between ownership and tunneling differs across owner types in terms of both directionality and magnitude. Our study offers a substantial theoretical contribution to the principal – principal problem literature by theorizing and testing variations of the problem among owners. Our study also advances our understanding of the role of ownership in firms. Theoretical/Academic Implications: Our findings have important theory ‐ building implications for the principal – principal literature. Controlling shareholders have been linked to performance outcomes or to tunneling in general, yet little is known about their comparative propensity to engage in tunneling. Our results, however, highlight that the propensity to engage in tunneling varies substantially among controlling shareholders. Furthermore, in the robustness checks, we disconfirm some of the assumptions of the principal – principal literature. This study demonstrates the need to theorize about specific types of ownership and reassess the core arguments of principal – principal theory. Practitioner/Policy study offers insights for policymakers interested in mitigating the risk of minority shareholders being expropriated by the controlling shareholder. the key driver of tunneling and of the firm, we offer on what policymakers do to the asymmetry of information.


| INTRODUCTION
Tunneling, or self-dealing, is an agency problem that occurs when a majority shareholder diverts a firm's wealth at the expense of other shareholders. Tunneling may take in many forms, such as asset transfers or intragroup loans, that aim to hide or remove valuable resources from the firm. The label originated from an incident in the Czech Republic when a firm's assets were literally removed through an underground tunnel (Johnson, La Porta, Lopes-de-Silanes, & Shleifer, 2000). Spurred in part by the work of Johnson et al. (2000), there has been a high level of interest in tunneling. However, the theory development on tunneling has been hampered by broad inconsistencies between predictions and empirical results. Although concentrated ownership is generally expected to facilitate tunneling, studies are sharply split between positive and negative findings (e.g., Huyghebaert & Wang, 2012;Jiang, Rao, & Yue, 2015;Lo, Wong, & Firth, 2010;Martins, Schiehll, & Terra, 2017;Peng & Jiang, 2010;Yang & Schwarz, 2016).
We looked more closely at this split by applying a more fine-grained conceptualization of ownership and its ensuing effects.
These inconsistencies are due to the fact that studies have focused on the extent of ownership as a driver of tunneling. In contrast, we offer a theoretical framework that explains how tunneling varies systematically across owner categories. Specifically, we examine how a key aspect of the problem-the reason why owners seek the private benefits of control-differs across owner types and its subsequent relationship to tunneling activity. We discuss two reasons for owners to engage in tunneling: extracting the private benefit of control to serve their own interests (i.e., self-serving private benefit) and benefiting other stakeholders, such as constituents, family members, and friends.
In the former, the controlling shareholder receives monetary benefits that are used to increase his or her own wealth, whereas in the latter, the extracted benefits are transferred to other stakeholders.
To reconcile the inconsistencies in this large body of research, we conducted meta-analytic structural equation modeling (MASEM).
MASEM is a more powerful tool than traditional meta-analyses (MAs), as it allows us to compare the different ownership forms concurrently while also including control variables. This analysis was based on 271 articles, 982 effect size estimates, and a sample of 66,590 observations. The MASEM revealed that tunneling varied substantially across owner types, both in terms of directionality and magnitude.
State-owned enterprises (SOEs) have a negative association with tunneling, whereas both business group (BG) ownership and family ownership have a positive relationship with tunneling. This association was largest for family ownership. We then used traditional MA techniques to run several robustness checks, which revealed that the results were unaffected by the time period, research field, or geographic focus of the respective studies.
This study expands our understanding of the role of ownership in organizations and makes several theoretical contributions. In much of the literature, ownership is often related to monitoring and appointments of new board members and top executives or to tunneling in general. In our study, however, we theorize the relationships between owners and their decisions to engage in tunneling. Our first contribution is to explain how and why the relationship between ownership and tunneling varies systematically across ownership types: The owners of large firms have different propensities to engage in tunneling, resulting in different levels of expropriation of minority shareholders. Our findings indicate that the most important driver of tunneling is the degree to which the controlling shareholder has access to and can leverage private information. The findings further suggest that being a controlling shareholder is a necessary but not a sufficient condition to engage in tunneling. In the end, some types of owners are better than others at resisting engaging in tunneling.
Therefore, governance policies regarding the issue-such as minority shareholder protection or board composition-vary in effectiveness based on the nature of the controlling shareholder. Furthermore, this study offers a second contribution to theory by assessing and disconfirming some key assumptions in the principal-principal (PP) literature, namely, the prominence of tunneling in emerging countries and that tunneling decreases as formal institutions develop (Young, Peng, Ahlstrom, Bruton, & Jiang, 2008). Our findings show that tunneling is not dependent on the institutional context of a country but on ownership type. We discuss the implications of our findings in Section 5 of the paper. Tunneling, or self-dealing, is the transfer of resources that benefits controlling shareholders at the expense of minority shareholders (Jiang & Peng, 2011;Johnson et al., 2000). Tunneling can be legal or illegal, depending on the specific actions taken and the host country's legal system. For example, a minority investor in France unsuccessfully sued when a dominant owner steered the business toward a familyowned subsidiary; in this case, the court ruled that the decision was legitimate (Johnson et al., 2000). However, in a case in Taiwan, multiple executives were arrested, and their firm was subsequently delisted following the discovery of tunneling activities (Yang & Schwarz, 2016).
Because tunneling is not directly observable, scholars use a number of proxies to assess its presence. One common indicator is wedge, which is measured by the divergence between the voting rights and cash-flow rights of controlling shareholders. This proxy increases information asymmetry and makes anti-self-dealing regulations less effective (Byun, Choi, Hwang, & Kim, 2013;Liu & Magnan, 2011;Miller, Breton-Miller, & Lester, 2013;Peng & Jiang, 2010). There are also several indirect proxies for the private benefits of control (e.g., Luo, Wan, & Cai, 2012;Zellweger, Kellermanns, Chrisman, & Chua, 2012), as well as accounting measures, such as inter-company transactions, including loans, account receivables, and preferential pricing, which signal the presence of an underlying problem (e.g., Boateng & Huang, 2016;Haß, Johan, & Müller, 2016;Jiang, Lee, & Yue, 2010;Lo et al., 2010;Yang, 2017). Overall, these measures capture the ability of the controlling shareholder to extract private benefits of control.

| What precipitates tunneling?
Agency problems arise when two transaction partners have competing interests and there are no mechanisms in place to constrain opportunistic behavior. A "principal-agent problem" occurs when the self-interested party has minimal equity at stake, whereas a "PP problem" occurs when both parties hold equity positions. The latter scenario involves controlling shareholders acting opportunistically against noncontrolling shareholders (Young et al., 2008).
Two mechanisms make engaging in tunneling possible. The relevance of these two mechanisms varies based on the reason for the tunneling. In the case of the self-serving private benefits of control, information asymmetry is a necessary condition for actors to engage in opportunistic behavior (Eisenhardt, 1989). In the case of self-serving tunneling, the controlling shareholder of a corporation aims to remunerate itself beyond what it is entitled to. Bergh, Ketchen, Orlandi, Heugens, and Boyd (2019) identified a set of factors that can create information asymmetries. First, key information may be unobservable or may have uncertain qualities. For example, some scholars have argued that a firm's true cash flow cannot be accurately estimated by outsiders, and the lack of transparency creates an opportunity for tunneling (Martins et al., 2017). Second, there can be structural barriers to collecting information: Collecting information becomes more complex as a firm grows or becomes more diversified. Larger firms will have more subsidiaries, facilitating the transfer of resources to subunits. This advantage can provide the controlling shareholder with greater access to the firm's cash flow and more options to opaquely tunnel the firm's resources away.
In the case of the non-self-serving private benefits of control, board representation is a critical mechanism for the controlling shareholder. Companies engage in non-self-serving tunneling when they need to benefit their stakeholders, such as family and friends (Yang & Schwarz, 2016) or the government (Huyghebaert & Wang, 2012).
Board representation reduces board oversight and increases information asymmetries because the board is captured by the controlling shareholder's affiliated directors. Board representation can be used to pressure the board into engaging in tunneling activities in order to favor a desired stakeholder through justifying the action of tunneling with the indirect benefits the firm might receive. Indeed, the percentage of directors affiliated with controlling shareholders tends to be high (Dahya, Dimitrov, & McConnell, 2008): For example, affiliated directors represent 30%-55% of the board members in SOEs (Cheung, Rau, & Stouraitis, 2010;Huyghebaert & Wang, 2012;Lee & Wang, 2017), 21%-53% in family firms (Arosa, Iturralde, & Maseda, 2010;Jones, Makri, & Gomez-Mejia, L. R., 2008;Yeh & Woidtke, 2005), and 20%-25% in BGs (Hearn, Strange, & Piesse, 2017;Pombo & Gutiérrez, 2011). In comparison, the percentage of directors affiliated with shareholders is lower when there is no dominant shareholder -at~10%-13% (Jones et al., 2008;Mobbs, 2013). A key difference between self-serving and non-selfserving tunneling is that the latter occurs more sporadically than the former does, making it less severe on the company. Shareholders who engage in tunneling to benefit themselves do so more regularly than those who do it to benefit other stakeholders (Atanasov, Black, & Ciccotello, 2014), leading to differences in the relevance of the relationship between tunneling activity and ownership type, as discussed in greater detail below.
Owners can engage in one, the other, or both types of tunneling simultaneously. In the latter case, the relationship between tunneling and ownership is the strongest because it is affected by both reasons for tunneling. In the next section, we develop a theoretical framework that explains why tunneling activity will differ systematically across ownership types.

| Ownership
One challenge in synthesizing prior works is the diversity of owner types discussed in different studies. There are several types of possible owners, each of which has a distinct set of goals and priorities (Boyd & Solarino, 2016) and differing implications for both the underlying rationale and the extent of tunneling associated with each type.
We present a set of testable propositions, from the weakest to the strongest association, to capture these differences.

| State-owned enterprises
Despite the declining involvement of government in business since the 1980s, SOEs are still common, even in mature economies. The majority of research on SOEs has explored the effects of ownership on their performance, primarily in the context of emerging economies (Boyd & Solarino, 2016). In part, this reflects the argument that state ownership may buffer weak institutions associated with emerging economies (Inoue, Lazzarini, & Musacchio, 2013). The benefit of strong government ties, however, can be offset by the competing goals of state or national interests. Indeed, a state can leverage its direct ownership in SOEs as an alternative means of benefiting particular constituents (Okhmatovskiy, 2010). The state will pressure SOEs to contribute to employment, growth, equity, regional development, social care, and other areas (Ding, Zhang, & Zhang, 2007;Shen & Lin, 2009). Indeed, state ownership is associated with valuedestroying-related party transactions that aim to tunnel resources out of the SEOs (Cheung, Jing, Lu, Rau, & Stouraitis, 2009;Jiang et al., 2015;Lee & Xiao, 2004) to benefit their constituents; thereby, the SOEs engage in non-self-serving tunneling.
First, because their financing largely depends on other stateowned sources of financing, SOEs are less constrained by market forces than other forms of ownership, have less need to listen to other shareholders' needs compared with any other ownership form, and are less constrained by supporting the state's noneconomic goals.
Second, compared with other types of firm, the careers and rewards of SOE managers largely depend on whether they succeed in fulfilling government goals (Milhaupt & Lin, 2013;Wang, Hong, Kafouros, & Wright, 2012). Many managers are also themselves politicians or bureaucrats affiliated with the controlling political party (Fan, Wong, & Zhang, 2007). Consequently, managers extract benefits from the firm to benefit particular political constituencies (Shleifer, 1998;Shleifer & Vishny, 1997). Indeed, politicians have an incentive to transfer value from listed SOEs to entities owned by or affiliated with the state in order to redeploy that value to specific stakeholders, thus benefiting the state consensus. Specifically, politicians' goals are dictated by their public interests (Shleifer & Vishny, 1994, 1997, and to meet these goals, they will subsidize other underperforming SOEs or make transfers to the central government to finance social policies, ensure that their constituents are satisfied, and thereby improve their chances of reelection or reappointment by engaging in tunneling activities in order to favor the government (Huyghebaert & Wang, 2012). Finally, SOEs tunnel resources sporadically when the government needs funding for specific projects. For example, during the 2008 financial crisis, ENI, the Italian state-controlled energy company, transferred €200 million to the ENI Foundation: This money was used to finance a government-sponsored social care program.
Because most of the ENI board members were state-affiliated directors, it was easier for the SOE to approve tunneling to benefit the government, thus representing an instance of non-self-serving tunneling. This led to our first hypothesis: Hypothesis 1. State ownership is positively associated with tunneling activities.

| Business group ownership
A BG is a set of firms that are ostensibly independent. These firms are coordinated by a central actor to achieve mutual objectives. Coordination happens through multiple ties, including ownership, economic means, and/or social relations (Khanna & Rivkin, 2001;Yiu, Lu, Bruton, & Hoskisson, 2007). These independent companies are often listed on the stock market themselves and have client-supplier relationships among them. The specific structure of BGs varies widely across regions, and BGs can be found in both emerging and mature economies, but all BGs have a centralized actor and share a common group objective (Yiu et al., 2007). For these reasons, we discuss all of them as a single ownership category. 1 In some emerging economies, most BGs are family controlled.
The key difference between a family firm and a BG lies in the structure of the organization and in the business in which the family firms operate. In family firms, only the holding tends to be listed, and subsidiaries are directly managed firms, whereas in BGs, multiple firms are listed and are coordinated and not controlled by a central actor. At the same time, family firms tend to invest more than BGs do in unrelated businesses to hedge the risk for the family wealth and have fewer within-subsidiary client-supplier relationships than BGs do.
Controlling shareholders of BGs are less prone to engaging in tunneling for non-self-serving purposes than SOEs are, but they are more prone to engaging in self-serving tunneling. First, BG owners will contribute toward supporting state goals until these goals are beneficial for the BG itself, but the size and geographical scope of BGs make them less dependent on state or local community support. Second, as mentioned above, BGs have fewer affiliated directors, thereby reducing the chances of the board being influenced by different stakeholders' interests.
At the same time, we expect that the potential for controlling shareholders to abuse information asymmetries is higher in BGs than in SOEs, as the organizational structure of BGs gives them an advantage when it comes to engaging in tunneling. BGs have a higher degree of unrelated diversification, which is intentionally designed to spread business risk across different industries: The controlling shareholder of a BG seeks to exploit market opportunities and will create a number of related business units to achieve such a purpose (Chang & Hong, 2002). The size and diversification of BGs generate a more complex organizational structure that allows for intragroup business transactions (e.g., goods, services, and capital). The latter are exploited to divert private benefits to controlling shareholders by diverting resources to companies in which they own more cash-flow rights (Ang, Cole, & Lin, 2000;Bae, Kang, & Kim, 2002;Chang, 2003).
BGs are thus ideally placed to engage in self-serving tunneling because (a) a controlling shareholder can arrange inter-company deals more easily than other types of owners and (b) due to within-group client-supplier relationships, other shareholders face many more hurdles in attempting to trace such inter-company transfers and assess whether they are genuine or are being used to tunnel resources. This is especially true as wedges (the difference between cash-flow and control rights) increase (Bae et al., 2002;Khanna & Yafeh, 2007).
Therefore, because BGs are more prone to engaging in self-serving tunneling, whereas SOEs are more prone to engaging in non-selfserving tunneling, we postulate that Hypothesis 2. Business groups will have a stronger association with tunneling than state-owned enterprises.

| Family ownership
Family firms are the most common form of ownership in all regions of the world (Astrachan & Shanker, 2003;Claessens, Djankov, & Lang, 2000;La Porta, Lopez-de-Silanes, & Shleifer, 1999). Such firms are not limited to small or medium enterprises, as firms controlled by their founder or a descendant of the founder dominate stock exchanges worldwide (La Porta et al., 1999).  (Craig & Dibrell, 2006) and are therefore willing to divert part of the firm's resources to benefit their stakeholders, such as local communities or governments, to reinforce their status. This is because the identity of the family owner is so closely tied to that of the organization that external stakeholders perceive the firm as an extension of the family itself, connecting its name and reputation to the product it sells (Bingham, Dyer, Smith, & Adams, 2011). For example, studies have found that controlling families in polluting industries are more likely to adopt environmentally friendly practices than nonfamily firms (Berrone, Cruz, Gómez-Mejía, & Larraza-Kintana, 2010). Other cases revealed that controlling families will, at times, tunnel resources away from the firm to fulfill the needs of some stakeholders. For instance, the owner of Wrightbus diverted over £15 million in company resources to a local church in 6 years, driving the company into administration (BBC, 2019; Simpson, 2019). Because the board was captured by family members and their affiliates, non-self-serving tunneling that benefited the firm's image was more likely to happen.
At the same time, family firms desire the self-serving private benefits of control and often appoint family members as company officers to provide them with an informational advantage over other shareholders. These information advantages can be used to expropriate minority shareholders (Anderson & Reeb, 2003;Lemmon & Lins, 2003).
Such an information advantage, alongside multiple voting-right share structures, creates incentives and the opportunities to exploit other shareholders, providing the family with benefits beyond what it would expect given the family's equity share in the company. Finally, the family derives its income from the firm itself but not from its stock price.
Therefore, family members are not averse to tunneling, which negatively affects a firm's market valuation (Claessens et al., 2000;La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 2002;Lins, 2003). Furthermore, the family control over the firm makes the latter not subject to the market for corporate control, thereby removing a constraining factor on the family members' behavior. Therefore, family firms have the incentive and the opportunity to tunnel resources to fulfill the needs of external stakeholders (who are engaging in non-self-serving tunneling), as well as the opportunity and the access to the necessary information that gives them an advantage in engaging in self-serving tunneling. This leads to the following hypothesis: Hypothesis 3. Family ownership will be more strongly associated with tunneling than state-owned enterprises and business groups.

| Sample and coding
To create the MASEM correlation matrix, we conducted a structured content analysis by sampling a set of prominent journals in management, international business, accounting, economics, and finance over 39 years . We searched Web of Science, Scopus, and Google Scholar for the keywords "principal-principal," "private benefits of control," "wedge," "control divergence," "pyramidal structure," "tunnelling," "tunneling," "self-dealing," "business group ownership," "BG ownership," "state own*," "SOE," and "family ownership" in the title, keywords, or abstract. This resulted in 2,136 results for Web of Science, 548 for Scopus, and we manually searched the first 40 pages of Google Scholar. As a second step, we removed sources that appeared in more than one database. Third, we manually examined each of the abstracts and retained the articles that were relevant to the study, totaling 693 articles. We erred in the direction of including a source that may not be relevant, rather than excluding a source that may be relevant. We then identified empirical articles that reported the relevant statistics for computing meta-analytic effect sizes (correlations or t-tests). We did not search for unpublished papers, as the "file drawer problem" has been found to have minimal influence on the outcome of MAs (Dalton, Aguinis, Dalton, Bosco, & Pierce, 2012).
The final sample comprised 271 articles, 2 producing a total of 982 effect sizes and a harmonic mean of 66,590 observations. Although MAs can be affected by sample dependence across individual studies, this potential concern is mitigated by three aspects of our sample: our use of a 39-year time horizon, sampling across multiple disciplines, and the broad range of geographic regions in our article pool. All articles were coded by a single expert rater. A second rater coded a subset of articles to assess reliability. The overall reliability was 1.

| Ownership
We categorized ownership forms into separate owner types, following Boyd and Solarino's (2016) study: SOEs are most frequently defined as a categorical variable, but some studies have measured this variable as the percentage of equity held by the local or national government. Family-owned firms are operationalized in multiple ways, including the degree of equity held, various ownership thresholds (e.g., 5%, 15%, 25%, or 50% of equity), or by dichotomous measures.
BGs have also been operationalized in multiple ways, including the percentage of group ownership, group affiliation (dummy variable), and the presence of corporate blockholders. We coded the ownership categories based on the original paper's definition and as mutually exclusive categories.

| Control variables
MASEM allows researchers to include control variables, which help to rule out alternative explanations (Bergh et al., 2016;Combs, Crook, & Rauch, 2019). We included three control variables: firm size, firm leverage, and institutional investors. Firm size is often associated with organizational complexity, which in turn might facilitate tunneling opportunities. It has been measured as the (log) number of employees or total assets. Firm leverage can constrain the ability of the controlling shareholder to engage in tunneling. Institutional investors are pervasive in the global landscape and are considered as being able to exert a direct influence through their "voice" and an indirect influence through their "exit" managerial behaviors (Appel, Gormley, & Keim, 2016;Edmans, 2014). We collected data on the presence of pressure-resistant investors (PRIs) (Brickley, Lease, & Smith, 1988).
PRIs do not have ongoing business relations with their holdings, and these include mutual funds, hedge funds, and pension funds. By definition, their arm's-length relationship can raise concerns and challenge management (David, Hitt, & Gimeno, 2001). These owners influence firm performance and strategies by engaging in both formal (Bharath, Jayaraman, & Nagar, 2013) and informal negotiations with the board and executives (Appel et al., 2016). PRIs, for example, can pressure the board to replace an underperforming CEO (Brav, Jiang, Partnoy, & Thomas, 2008;Del Guercio, Seery, & Woidtke, 2008) or limit executive compensation (Brav et al., 2008;Ertimur, Ferri, & Muslu, 2010). They can also extend their influence by obtaining board seats, thus further increasing their voice (Klein & Zur, 2009). PRIs are measured by the overall equity held by all investors or by the largest PRIs.  (Borenstein, Hedges, Higgins, & Rothstein, 2005). We used Mplus 8 (Muthén & Muthén, 2017) for all structural models reported in the manuscript. The harmonic mean is the average of the number of observations for each pairwise correlation in the dataset and is the recommended value to be used in structural equation models. For this study, this value was 66,590. Table 1 presents the meta-analytic correlation matrix used for our analysis with no adjustments (e.g., the reliability is set to 1.0).

| RESULTS
Hypotheses 1-3 proposed that the effect on tunneling would vary according to the owner type. The first step of our analysis was, therefore, to treat ownership as a latent construct, with SOEs, BGs, and family firms as multiple indicators of a common dimension. Institutional investors and firm size were included as separate dimensions.
This model did not converge despite multiple iterations, indicating that this configuration was not supported by the data. We proceeded by testing the MASEM, treating each ownership form as distinct from the others. Table 2a presents the standardized MASEM results with the adjustments for measurement reliability set to 0.80. The correlations with the reliability set to 1.0 -a more conservative test-are presented in Table 2b.
In Hypothesis 1, we proposed that state ownership is positively related to tunneling. SOEs showed a negative and significant effect (γ = −0.043, p = .001). Consequently, Hypothesis 1 is not supported.
In Hypothesis 2, we postulated that the association between BG ownership and tunneling is stronger than that between state ownership and tunneling. The BG's effect is positive and significant (γ = 0.073, p < .001). Furthermore, the confidence interval does not overlap with that of the SOEs (−0.045/−0.039 and 0.070/0.076 for SOEs and BGs, respectively), supporting Hypothesis 2.
In Hypothesis 3, we assert that the association between family ownership and tunneling is stronger than that for SOEs and BGs. Family firms showed the largest effect, with an effect size of 0.237 (p < .001). Additionally, we found that there was no overlap between the confidence interval of the relationship between the ownership types and tunneling (confidence interval: 0.234/0.239). Thereby, the MASEM results indicate that Hypothesis 2 and Hypothesis 3 are

| Robustness checks
We ran supplementary analyses to assess the robustness of our findings using 82 effect size estimates and 398,000 observations from 50 studies. First, we ran a traditional MA using the ownership form as a moderator. Additionally, we tested for possible moderators, including the disciplinary focus, the time period of the articles, and the level of development of the country. The results of these analyses are reported in Table 4. First, we examined whether effect sizes differed systematically based on the research field. Table 4a presents the results of the comparisons for finance, international business, and management publications. We did not find sufficient articles from economics or accounting to include in this comparison. Although effect sizes appeared larger for management-based articles, the Q value of 1.19 was not significant (p = .55), demonstrating that these differences were not statistically meaningful. Second, we split the sample in half to compare more recent versus older studies. As indicated in Table 4b, this moderation test produced a Q value of 0.46, which was again nonsignificant. As an alternative test for temporal effects, we also ran a meta-regression of the year of publication as a predictor of effect size magnitude. As shown in Table 4c, the slope coefficient was also nonsignificant (p = .64). Then, we tested whether the effect would systematically differ between emerging and advanced nations, as previous studies have suggested that tunneling is more likely to occur in less developed institutional environments (Young et al., 2008). We excluded studies mixing advanced and emerging nations from our analysis. As shown in Table 4d, the moderation effect by the level of institutional development was not significant (Q = 0.40). Finally, we assessed whether regulations could worsen or lessen tunneling (Table 4e). We employed the Guillén and Capron (2016) Shareholder Protection Index, which measures the state's capacity to implement shareholder protection rules. 3 The results indicate that the state's capacity to implement shareholder protection rules does not moderate the ability of the controlling shareholders to engage in tunneling.
Overall, the results from the robustness checks increase our confidence in the findings of our main analysis.

| DISCUSSION
PP problems and tunneling have been widely discussed in the literature, and conventional corporate governance assumes that concentrated ownership will be a source of tunneling (Young et al., 2008).
However, to date, there has been no differentiation according to the type of owner nor for the motives for engaging in tunneling. Although the theoretical prediction for expropriation is quite clear, the empirical findings are mixed, and there is less agreement about which specific kind of ownership could lead to the highest level of expropriation; these circumstances call for a more fine-grained theorization of the PP problems.
We discuss the existence of two reasons for controlling shareholders to engage in tunneling and map these reasons against the different types of controlling shareholders. First, we clarified the reasons owners engage in tunneling: Controlling shareholders engage in tunneling to benefit themselves or their stakeholders. Benefiting themselves implies continuous rather than sporadic tunneling, thereby making it more severe. Second, we discussed how not all controlling shareholders weigh up their reasons for engaging in tunneling equally.
Benefiting their stakeholders is important in the context of state ownership and family ownership, whereas benefiting themselves is important in BG ownership and family ownership. On the basis of this rationale, we developed a set of hypotheses arguing that tunneling is more strongly associated with certain types of concentrated ownership. Our results demonstrated that the three types of concentrated ownership did not lead to the same level of expropriation of minority shareholders. The pattern of relationships differs across owner types: State ownership does not seem to be related to the expropriation of minority shareholders. Family ownership generally strengthens conflict, whereas BG ownership yielded weaker findings, even though both of these ownership types were, on average, positively related to tunneling. These disparate findings suggest that rather than assuming the existence of a general predisposition to (or not to) expropriate minority shareholders, scholars should theorize the relationship between tunneling and each type of owner in their analyses.
Our findings indicate that the most important driver of tunneling is the degree to which the controlling shareholder has access to and been seen as mostly passive and has generally been linked to performance outcomes with weak effect sizes (Boyd & Solarino, 2016). Our results, however, highlight that tunneling has a significant effect on firms and suggest that controlling shareholders actively lead firms to engage in tunneling activities.
The MASEM analysis further evidenced that institutional investors play an important role in mitigating tunneling. Prior studies have shared the assumption that institutional investors would only suffer losses from tunneling (e.g., Dyck & Zingales, 2004;Johnson et al., 2000;Mitton, 2002;Young et al., 2008 Note: We also have publications from economics and accounting, but they are too few to run a meaningful analysis. Studies mixing developed and emerging countries were excluded. the boundaries of a theory or, as in our case, the importance and consistency of effect size across multiple independent variabledependent variable relationships. In particular, we tested whether the effect differed systematically between emerging and advanced nations, as previous studies have suggested that tunneling is more likely to occur in less developed institutional environments (Young et al., 2008). We did not find a meaningful difference between the two environments. Similarly, the state's capacity to implement shareholder protection rules does not seem to mitigate tunneling either.
Another robustness check revealed that the magnitude of effect sizes does not change over time, again suggesting that the effect size of tunneling did not change over the 39-year period we examined, regardless of the improvement in the institutional environment in terms of minority shareholder protection, the rule of law, and so forth.
These robustness checks suggest that assumptions about the role of the institutional environment need to be reassessed.

| Policy implications
Our results indicate that it is possible to draw some counterintuitive policy implications regarding board composition. As the key driver of tunneling appears to be access to private information and knowledge of the firm, a possible solution to mitigating tunneling would be to employ more long-tenured directors. This is at odds with current trends. In recent years, countries have tended to limit the tenure of outside directors (Bonini, Deng, Ferrari, & John, 2017 (Miner, 2003). Further theorizing about the domain of owner conflicts is needed and must not assume homogeneity across owners.
Some work has been done in this area (e.g., Appel et al., 2016) (Platteau, 2000). Future research should assess how the informal institutions of a country are related to tunneling. For example, are family owners more or less prone to engaging in tunneling in countries that place a greater emphasis on family values? Or are the different ownership forms more or less prone to engaging in tunneling in countries that place more value on relational business transactions versus arm's-length transactions? It is therefore important that the assessment of the assumptions behind tunneling is explored, not only from a formal institution perspective but also from an informal institution's point of view.
Additionally, future studies should assess whether the interaction effect between formal and informal institutions moderates the relationship between ownership and tunneling. Helmke and Levitsky (2004), building on the ideas of North (1990North ( , 1991, argue that informal institu- Some preliminary work has been done in this area, but the work is limited to single-country studies (e.g., Boateng & Huang, 2016

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of this article.
How to cite this article: Solarino AM, Boyd BK. Are all forms of ownership prone to tunneling? A meta-analysis. Corp