Why corporate board insiders still matter: Evidence using aggregate earnings shocks

Corporate governance research lacks clarity on why, when, and what types of firms appoint non-CEO board insiders because of the primary academic focus on board independence. While executive monitoring is relevant, it is equally salient to understand why firms appoint non-CEO insiders to the boards. This concern is especially relevant when firms face financial difficulties during periods of macro hardship. Using the corporate socialization theory, we suggest that board insiders likely generate firm-specific private information by utilizing their long-tenured firm-focused corporate experience. It would result in a unique form of non-fungible expertise that firms would find valuable during macro hardships. Employing a difference-in-difference research design, consistent with our theory, we document that when firms experience a cross-country macro hardship such as an aggregate earnings shock , they appoint long-tenured and firm-focused non-CEO board insiders. We further show that financially distressed firms have a higher demand for such directors.


INTRODUCTION
Agentic board composition theory (Fama & Jensen, 1983;Jensen & Meckling, 1976) suggests that board insiders possess "private information." Therefore, within the monitoring-advising diarchy-a standard feature of the board's resource-based view as well (Pfeffer & Salancik, 1978;Haynes & Hillman, 2010)board insiders likely play advisors since they provide firms-specific private information (Adams & Ferreira, 2007). According to this theory, private information enables effective and efficient corporate decision-making i (Harris & Raviv, 2008;Mace, 1971). Nevertheless, this theory provides an insufficient explanation of why firms appoint non-CEO board insiders.
Before independent directors' executive monitoring became the dominant corporate governance agenda (Granado-Peir o & L opez-Gracia, 2017), most firms used to have multiple board insiders on their corporate boards. However, as focus shifted to executive monitoring, corporate board reforms across several countries led firms to reduce the board insiders' appointments (Fauver et al., 2017;Zorn et al., 2017). However, as Figure 1 documents, this trend has reversed in favor of board insiders in some countries but not in others. Therefore, it is vital to ask why firms appoint non-CEO corporate board insiders to their board?
The theoretical mechanism behind board insider selection (Baysinger & Hoskisson, 1990), especially when it relates to non-CEOs, has not received sufficient academic attention barring a few exceptions (Coles et al., 2008). Still, non-CEO board insiders constitute 25% of the corporate boards globally. When we include the CEOs, this figure is as high as 36%. Therefore, our objective in this study is threefold. First, we suggest why board insiders matter to the firms by developing a corporate socialization theory, which aids private information generation. Next, building on our theorization, we empirically test when firms appoint board insiders by exploiting the probability that a country experiences a macro shock that proxies for periods of macro hardship. Finally, we assess if there is a particular class of firms that are more likely to appoint board insiders revealing their director selection preference during a period of macro shock.
Endogeneity concerns generally make it challenging to disassociate the reason behind a specific board configuration (Hermalin & Weisbach, 1998;Wintoki et al., 2012). Here we suggest that a macro shock would work as a cross-country source of exogenous variation, especially for firms in the industrial sector (Carey et al., 2012). It forms the basis for when board insiders should matter and to whom. Our central proposition is as follows: exogenous shock that comes from macro events such as a financial crisis (Carey et al., 2012)-similar to the global COVID19 pandemic (Spinelli & Pellino, 2020)-has the unique clarifying ability to distill it for the firms, issues, and decisions that are necessary and important from what is in vogue or fashionable under normal conditions (Kemper & Martin, 2010;Ayudhya et al., 2019). Using our identification strategy, ii which we call "aggregate earnings shock" iii (AES), we exploit the probability of some countries experiencing a stronger macro shock and include them within our treatment sample. In a difference-in-difference (DID) model (Card & Krueger, 1994), we then compare the board configuration differences in the treated sample and the control sample-which did not experience such a macro shock (i.e., AES)-both before and after the shock period. With this, we also impair generalizability shortcomings highlighted in prior literature (Desender et al., 2013) and reduce external environment-driven board selections (Filatotchev & Nakajima, 2010).
In this study, we develop four main hypotheses. Our first hypothesis suggests that firms are likely to appoint more board insiders if their private information is deemed useful in the post macro shock period. Our second hypothesis suggests that financially distressed firms would have a higher demand for board insiders in the post-macro shock period since they suffer the doublejeopardy of experiencing firm-specific financial market difficulties during a macro hardship period. We define financially distressed firms as those firm-year observations that exhibit negative financial market returns for two consecutive years. Using a 23 countries sample of 42,867 firm-year observations from 6,385 unique industrial firms, we document results consistent with these hypotheses.
Drawing on corporate socialization theory, we discuss why board insiders matter to the firms in our third and fourth hypotheses. Our third hypothesis suggests that board insiders with long-tenured firm experience are more likely to be appointed to the board of financially distressed firms in the post-AES period. We consider a director as long-tenured if they have longer tenure in the firm than on the board. We expect that with longer firm tenure, board insiders are more likely to develop firmspecific non-fungible expertise. On the other hand, generic industry experience or technical knowledge such as financial experience is fungible expertise (Sauerwald et al., 2016) since firms can acquire them from outside or from other sources.
Our fourth hypothesis suggests that financially distressed firms require more firm-focused rather than selffocused (Masulis & Mobbs, 2011) board insiders in the post-macro shock period. There is no evidence of clear benefits from "busy" directors (Fich & Shivdasani, 2006;Field et al., 2013). Therefore, financially distressed firms facing macro shock are likely to evaluate how firmfocused or available potential board insiders would be to their present firm when selecting them for board appointments. We adopt Masulis and Mobbs' (2011) identification strategy and define self-focused non-CEO board insiders as those "certified" executives who have at least one outside appointment besides their home firm. Thus, we define firm-focused non-CEO board insiders as those "uncertified" executives who do not have any other external appointment. We find broad support for both these propositions using our sample. We ensure our results' robustness using the propensity score matching (PSM) technique (Guo & Fraser, 2015), alternative measures of the dependent variables, among other tests.
While agentic board composition theory explains the asymmetry of incentives between insiders and investors, its insufficient explanation of why board insiders get appointed on a board, especially non-CEOs, leaves a large research gap. We contribute to the board configuration research by developing the corporate socialization theory, which predicts why the boards appoint insiders. Using it, we suggest and find that in the post-macro F I G U R E 1 Board insiders across AES treatment/control samples Notes: This figure presents the equal-weighted simple average of the proportions of Board Insiders across AES Treatment (experienced an aggregate earnings shock) and control samples within the 2004-2015 timeline. The figure shows that the monotonic decline in the proportion of Board Insiders reversed only in the AES Treatment sample firms by 2011 after the full impact of the earnings shock, starting the year 2008, was realized through the firms' financial reporting shock period, firms appoint more long-tenured firmfocused board insiders to benefit from their non-fungible private information. Financially distressed firms have a higher demand for such board insiders during such difficult times.
Our findings have implications for the theory of the board insiders' private information generation processes (Adams & Ferreira, 2007) and public policies concerning corporate governance (Fauver et al., 2017). These results should be especially relevant to academics and other stakeholders in countries where the board configuration has systematically reversed, without much notice, in favor of board insiders post experiencing an AES.
Here onwards, we divide our study into five main sections. In the next section, we discuss the literature and develop our testable hypotheses. Next, we discuss our study's design. Following that, we report and discuss our results. In the penultimate section, we summarize all the endogeneity and robustness tests we have performed. We conclude this study by discussing our contributions, limitations, and future directions of the literature.

LITERATURE AND HYPOTHESES
Background on the corporate socialization theory Unlike independent directors, a firm gainfully employs board insiders on a full-time basis. Therefore, board insiders are not only corporate directors-responsible for the firm's fiduciary duties-but they are also the firm's employees.
Prior literature suggests that employees pass through a complex organizational socialization process (Fang et al., 2011;Van Maanen & Schein, 1979). It enables the employees to become aware of the firm's operations and culture (Camerer & Vepsalainen, 1988). Firms spend corporate resources to acclimatize their employees with the firm's corporate culture, especially at the point of entry. Corporate socialization occurs at several levels, including mentoring by supervisors, interaction with co-workers, and formal corporate training programs (Kammeyer-Mueller & Wanberg, 2003). Proactive corporate employees acquire information about the firm and their roles and responsibilities through many nonstructured interpersonal and non-interpersonal sources (Ostroff & Kozlowski, 1992).
Some proactive corporate employees receive positive supervisor feedback (Seibert et al., 1999). They also likely receive greater corporate responsibilities (Grant & Ashford, 2008). As department leaders, corporate employees become immersed in the firm's corporate bureaucracy (Dugger, 1980). Through these channels of experiences, the corporate employees cultivate their insider network, create long-lived associations among peers and juniors, and create collegial comraderies (Mao et al., 2012). While climbing the corporate ladder, the corporate leaders widen their experience by managing corporate bureaucracies, enabling them to become firmspecific executive experts. It also places them in the directorial talent-pool once they acquire sufficient bureaucratic authority and seniority. It is a unique form of Nahapiet & Ghoshal (1998) type of social capital that individuals cannot replicate outside the firm (Sauerwald et al., 2016). This firm-specific expertise generated through long-tenured corporate experience is non-fungible. In other words, individuals cannot acquire such expertise over the firm, its bureaucracy, its operations, and strategy through any other means other than working there for extended periods. It is potentially the source of the firm-specific private information that board insiders possess and makes them suitable candidates for board appointments (Adams & Ferreira, 2007;Harris & Raviv, 2008).
Board insiders also likely bring domain experience and expertise. Industry experience and financial expertise are probably the most common and sought-after domain experience (Guner et al., 2008). Nevertheless, such expertise is fungible. iv In other words, the board could acquire such expertise not only from board insiders but also from independent directors. Moreover, the board could receive fungible expertise from other channels, such as through informal discussions with junior executives (Boivie et al., 2016) or consultants. To acquire fungible expertise, the firms need not appoint board insiders, especially non-CEO board insiders (Bebchuk & Fried, 2005), and dilute the board's monitoring strength (Raheja, 2005) and risk insider co-option (Coles et al., 2014). This line of argument would also suggest that senior corporate new-hires have a distinct disadvantage as they are likely to have fungible experience and expertise. However, they would still need time to cultivate their insider networks to generate for themselves a firm-specific non-fungible social capital.

When and to whom board insiders should matter
Before we address why boards appoint non-CEO board insiders, we set the stage by addressing when and which types of firms likely select them. We posit that a period of macro shock is likely to increase the demand for board insiders within the general population of firms v for three main reasons. First, the utility of the non-fungible private information held by board insiders (Adams & Ferreira, 2007) becomes crucial during post-macro shock periods. Second, during such a period, a balance of power tilted towards board insiders would enable them to take "collective responsibility" in decision-making (Feinberg, 1968) without the need for decision-delegation. That is because groups take collective responsibility, or "collective liability," when "members have mutual interests"; they are driven towards achieving "common objects," and they have "goods and harms" which are "necessarily collective and indivisible" (Feinberg,p. 677). Besides, when independent directors control the board's decision-making, they are likely to spend less effort acquiring firm-specific private information. It creates an additional risk of free-ridership (Harris & Raviv, 2008). Therefore, we state our Hypothesis 1 as follows: Hypothesis 1 Macro shock is positively associated with a higher representation of board insiders.
Here, financially distressed firms are a unique class of heterogeneous test-case firms whose board configuration decisions in the post-macro shock period require special attention. Going-concern firms facing macro difficulties generally have a host of organizational defenses, starting with a competent top executive team, effective board, robust liquidity position, and relatively stable financial performance. If anything, such firms are likely to act as safe havens for investors looking to park their investments during macro hardship periods (Johnson et al., 2000). Financially distressed firms during post-macro shocks face not only a challenging economic environment but also a lack of investors' confidence. At this stage, boards heavy on monitoring are likely to face a severe need for credible firm-specific private information and practical decision-making skills (Harris & Raviv, 2008). Therefore, in Hypothesis 2, we suggest that financially distressed firms in the post-macro shock conditions would have a higher demand for board insiders: Hypothesis 2 In the post-macro shock period, financially distressed firms appoint a higher proportion of board insiders.

Fungible versus non-fungible expertise
In the post-macro shock period, what expertise board insiders would bring to the financially distressed firm? We have previously suggested that board insiders have two channels to generate private information: fungible and non-fungible expertise. In Hypothesis 3, we posit that financially distressed firms in the post-macro shock period would appoint board insiders with non-fungible expertise to benefit from their ability to generate firmspecific private information. Moreover, firms would find less resistance from outside investors as the board insiders in the financially distressed firms are likely to find themselves in a unique position where their incentives and that of the stakeholders are the same, that Is, the survival of the firm during periods of macro shock: Hypothesis 3 In the post-macro shock period, financially distressed firms appoint a higher proportion of board insiders with nonfungible expertise.

Self-focused versus firm-focused board insiders
Agentic board configuration theory suggests that the decision to appoint board insiders faces severe incentive alignment problems (Bebchuk & Fried, 2005;Fama & Jensen, 1983). Here Acharya et al. (2011) have argued that the firm's survival and performance beyond the present CEO's tenure is of concern to the junior non-CEO executives. These non-CEO executives, during periods of financial and economic stress, could not only enforce corporate control via "internal governance" (Acharya et al., 2011) but offer viable corporate succession planning (Ocasio, 1999). However, when deciding to appoint board insiders, the firm and the investors have two types of non-CEO options. Appoint board insiders who have pre-established market credibility through their outside appointments or appoint board insiders whose primary focus is the firm, but they are untested outside. The picture that emerges from prior literature suggests that "busy" directors may not be effective board monitors (Fich & Shivdasani, 2006), but they could bring invaluable external network resources (Field et al., 2013;Pfeffer & Salancik, 1978). Masulis and Mobbs (2011) have shown that some self-focused board insiders with outside appointments such as "Certified Inside Directors" results in positive organizational outcomes.
Nevertheless, these self-focused board insiders, with full-time jobs, have a higher burden of responsibilities and time pressure than independent directors who sit on multiple boards. Therefore, firms facing financial distress during macro shock would want board insiders who are not only firm-focused but have an interest in its long-term well-being out of self-preservation and career-concern (Acharya et al., 2011). The firm-focused board insiders are also likely to have their ears-to-the-ground, and as a result, they are likely to possess superior quality nonfungible private information. Therefore, in Hypothesis 4, we posit that financially distressed firms in the postmacro shock period are likely to appoint firm-focused board insiders: Hypothesis 4 In the post-macro shock period, financially distressed firms appoint a higher proportion of firm-focused board insiders.

Sample
We collect the data for this study from five sources. Our global corporate board data is from the BoardEx database. BoardEx has emerged as a premier source of quality cross-country corporate governance data across several disciplines due to the availability of high-quality director-level profiles. The financial accounting and market data are from the Worldscope database. We supplement our database by hand-collecting gender quota timings for each country in the sample. We collect the publicly available country-level data such as gross domestic product (GDP)/capita and foreign direct investment (FDI)/Ccapita from the World Bank database. To control institutional differences across countries, we utilize the time-varying country-level data made available by Guillén and Capron (2016). The Appendix contains a detailed list of all variables along with their sources.
The construction of our sample starts with the BoardEx database. To reduce noise and increase the reliability of our regression estimates, we apply some exclusionary criteria. The BoardEx universe consisted of director-level data from publicly listed firms from over a hundred countries. Most of these countries are small economies with only a few firm-year observations. We retained only those BoardEx countries in our sample with over eight years of continuous data with more than 100 firm-year observations. We eliminate firm-year observations with negative Book Value since they face significantly higher default risk. We also exclude firmyear observations with missing control variables. The macro shock that emerged from the financial crisis has its roots in the financial services sector despite its governance structures (Adams, 2012). Therefore, the exogenous shock may not be the sole reason to drive the board reorganization within this sector. Furthermore, governments own and operate utility sector firms across many countries. In a cross-country setting, there is no robust method to control such variations. Thus, we retained only the industrial sector firms vi with at least $100 million in Total Assets (US$, nominal vii ) for all our hypothesis testing.
The timeline of our data initiates in the year 2002. The corporate governance data available on BoardEx before this year is of inferior quality. We allow some countries, such as Australia, Austria, China, and Singapore, to enter the database as late as 2005. We exclude countries whose data initiates in later years. We document statistically significant differences between non-Winsorized and Winsorized continuous variables using a two-sample mean difference t-test and Wilcoxon signed-rank test. Therefore, we use one-percentile (1 and 99%) two-sided Winsorized continuous variables for all our main tests.

Dependent variables
Firm-level dependent variables At the firm-level, we have three main dependent variables. Our primary board variable is Board Insiders. We disaggregate board insiders into Self-focused Insiders and Firm-focused Insiders. Self-focused Insiders are the non-CEO "certified" insiders who have secured at least one outside directorship (Masulis & Mobbs, 2011). Firmfocused Insiders are non-CEO board insiders who do not have any other outside affiliation. Board Insiders is the proportion of executive directors per Board Size. We use a similar method to calculate other board variables, that Is, Self-focused Insiders and Firm-focused Insiders.

Director-level dependent variables
We measure non-fungible expertise using the length of firm-specific experience. Long-run firm-specific experience is non-fungible expertise. A director can acquire this expertise only by staying employed by the firm. We measure firm-specific expertise BI Firm Experience by coding a dummy variable as "1" if the Board Insider has longer tenure in the firm than on the board. It is a time-varying measure as the coding switches to "1" only in those years in which the firm tenure is higher than board tenure, "0" otherwise. This way, we mechanically exclude externally hired executive directors who simultaneously join the board and the firm. Alternatively, we use the second measure, BI Net Firm Experience (ln), which is the log difference between the time served within the firm and on the board to ensure the primary reported result's robustness.
Alternatively, we code two fungible expertise that Board Insiders are also likely to possess. They are industry experience depth, and financial expertise. We consider these as fungible expertise, as individuals can acquire them by being employed by a rival firm or across different sectors. Moreover, the firm could receive such expertise by hiring consultants and other similar channels without adding another insider to the board. We measure Board Insiders' industry experience depth (BI Ind. Exp. Depth) by counting the number of Fama-French 48 industries a director has worked in the past, per 48. A higher figure would indicate that the director has a broader industry experience depth. This figure is timevarying since it increases if the director accepts more board appointments outside their own Fama-French 48 industry. However, the measure's value would not increase if the director only accepts new board positions within their industry. We measure financial expertise (BI Fin. Expertise) of a Board Insider if they have the following job titles: CFO, Finance Director, Financial Manager, Accounting Specialist, Investment Director, Controller, etcetera. This measure is also a time-varying measure as we code a BI Fin. Expertise as "1" only starting the year they have acquired the necessary corporate designation.

Main explanatory variables
Country-level macro shock timings We identify countries that experienced a macro shock using AES timing using the Zivot & Andrews' (1992) unit-root test for a single structural break. viii Our choice of an AES is motivated by the emerging accounting literature on aggregate earnings' information value (Konchitchki & Patatoukas, 2014;Shivakumar & Urcan, 2017). Specifically, Konchitchki and Patatoukas (2014) and Shivakumar and Urcan (2017) have suggested that aggregate earnings have significant explanatory power regarding the national GDP and future inflation.
This study proposes that a significant break or a shock in the aggregate earnings series across countries could trigger board re-configurations in treated countries only. Therefore, we perform the countryby-country structural break analysis on the industrial firms' equal-weighted aggregate earnings (aggregated using operating earnings). This way, we identify the countries experiencing a significant change in the mean and trend. Both in the short-term (temporary break) and if the shift is permanent.
This identification strategy has several merits. Here, the use of accounting earnings is deliberate. Earnings have to meet a considerably higher verifiability threshold than the market data, such as stock prices, which vary widely due to short-term market sentiments. Thus, firms may not take strategic decisions such as changes to their board configuration based solely on short-term market movements. Besides, CEOs have significant discretion over such a performance measure. They have an incentive to keep the firm's earnings series as stable as possible, especially during the macro hardship period (Healy & Wahlen, 1999;Leuz et al., 2003). It reduces the risk of over-identification of countries into the treatment sample, the main concern for our analysis. We suggest that a temporary or a systematic shock to the aggregate earnings series could prompt strategic board re-configuration action by insiders and investors alike. ix We avoid using generic GDP growth rates to model macro shock because of the arm's length at which such data is generated and a broader set of consumption indicators involved in producing such a series. Our use of the Zivot-Andrews unit-root test helps detects a change in the mean and trend in our aggregate earnings series because it assumes a single break-point in the data and econometrically minimizes the t-statistic to detect its location. By making the selection of the breaking point a part of the econometric process, we make the tacit ex ante assumption of the break-year unnecessary. x We consider the firm-year observations from countries whose aggregate earnings series experienced a significant shift as the AES Treatment sample and code them "1" irrespective of the year. Therefore, firm-year observations that were coded "0" are from the Control sample countries. For the Post variable, we code "1" starting the year 2009 for all countries, "0" otherwise. That is because most countries experienced an AES beginning in the year 2008. Thereby, we accommodate for a year's lag. This timeline is broadly consistent with financial economics research (Carey et al., 2012;Kalemli-Özcan et al., 2016). Therefore, we code AES Treatment x Post as "1" for treatment countries, which experienced a shock in the aggregate earnings series starting the year in which Zivot-Andrews unit-root test detected a significant shift in mean and trend, "0" otherwise. Since some countries are likely to experience macro shock starting slightly different years, we expect the treatment of an AES to be akin to a staggered adoption (Athey & Imbens, 2021).

Firm-level explanatory variable
We code a firm-year as financially distressed if the firm's returns are below zero (i.e., negative) for two consecutive years. The dummy Financial Loss is coded "1" in the second year of the two successive years when the firm has a negative return. Therefore, Financial Loss is a timevarying variable that switches to "1" only in the second consecutive year of enduring a financial market loss. To test Hypothesis 2, in addition to the Financial Loss variable, we also use the Board Insiders as an additional explanatory variable.

Control
In this study, we employ four sets of control variables: (i) board; (ii) CEO; (iii) firm; and (iv) country controls. Following standard corporate governance research (Masulis & Mobbs, 2011;Zorn et al., 2017), we use Board Size (natural logarithm), Board Gender Diversity, Board Tenure, and Board's Outside Affiliations as board controls. The latter two variables are normalized using Board Size (all our results are qualitatively similar if we use natural logarithm instead). We identify if the CEOs hold dual roles (CEO Duality) using the "individual role" data available in the BoardEx database. Other CEO controls are Woman CEO and CEO Turnover. If a firm-year has two CEOs, then we retain the outgoing CEO because of legacy considerations. Following prior literature (Masulis & Mobbs, 2011;Zorn et al., 2017), we employ firm-level controls such as Tobin's Q, Total Assets (ln), profitability (Operating ROA), Total Liabilities, Interest Rate, CapX, R&D, Cash Holdings, Ownership, Cross-Listed, Business Segments (ln) and Geographic Segments (ln). We include the IFRS Accounting Standard dummy and Big4 Auditors dummy to control the firm-level mandatory and voluntary adoption of high-quality accounting standards across the world and financial reporting quality, which requires insiders' expertise.
Given that we employ cross-country data to test all our hypotheses, it is vital to control observable variations across institutional settings and countries. Several countries in our sample allow or mandate a two-tier board system with a separate supervisory board (Denis & McConnell, 2003;Ferreira & Kirchmaier, 2013). To control this country-level variation, we code a dummy variable Dual Board as "1" if the country allows or mandates a two-tier board structure, "0" otherwise. All results are qualitatively similar if, instead of Dual Board, we use the proportion of supervisory directors. Guided by previous research (Fauver et al., 2017), we use GDP and FDI as our main time-varying country controls. The seminal study by La Porta et al. (1998) has shown that countries have institutional differences owing to their legal origins, resulting in variations in their minority shareholder protection. However, due to their identification strategy's static nature, we employ Guillén & Capron's (2016) data on time-varying cross-country minority-shareholder protection scores (SRI) to control for institutional differences. Given a country's unique institutional arrangement, the board's labor market could lead to significant unobserved variations, creating supply-side constraints (Knyazeva et al., 2013). Therefore, to control for such variations, we create a variable called Indp. Board Market. We measure it as the difference between the number of rookie independent directors (defined by no prior board experience within listed firms) and the number of independent directors who drop-off the board labor market (defined by those directors who never reappear in any of the subsequent year's corporate board, irrespective of the firm). We normalize the resulting figure by the total number of board seats in a country-year. Several countries have passed laws to mandate women's representation on corporate boards. It could likely alter the board's structure across the affected countries (Ahern & Dittmar, 2012). To control for it, we code a dummy variable Gender Quota as "1" starting the year a country passed a law to mandate women's representation on corporate boards, "0" otherwise. In the Appendix, we describe the method employed to create all the variables used in this study.
The difference-in-difference empirical model To test Hypothesis 1, we construct our staggered, dynamic cross-sectional DID model (Card & Krueger, 1994). Using DID, we exploit the comparison between the treatment sample (which experienced a significant AES) and the control sample (which did not). Furthermore, we also exploit the AES's staggered timing (identified using a break in the country-level aggregate earnings series) in the treatment sample against the same period in the control sample. Given that no single firm can drive the country to experience an AES, the selection of its country-level break dates is exogenous to the post-period board configuration decisions (Zivot & Andrews, 1992). Therefore, the categorization of countries in the treatment or the control samples for the DID model is also randomized. xi We present the basic structure of the DID model as follows: Here, we identify the dependent variable as the firm i, across t years. In this model, the coefficient β 3 , associated with the Treatment Â Post represents the post-AES period's incremental effect. Here x and z are a set of contemporaneous and lagged covariates. Since board configuration is a contractual process, any change to its structure requires time-lags. Therefore, we lag the controls, such as firm size, performance, etc., by one year (the x controls). However, while time-varying, some controls do not change within the firm over the years (the z controls). Chief among them is the business and geographic segments. We do not lag these variables when including them in the controls. Here, m and n identified the maximum number of lagged and non-lagged controls.
We use the Year fixed-effects (Year FE) to control for time-varying changes. We control time-invariant unobserved factors such as corporate culture (Camerer & Vepsalainen, 1988) using group effects (Unit FE) such as within-firm fixed-effects in the firm-level analysis and industry fixed-effect in the director-level analysis. Thus, ν is the white noise with an assumed normal distribution. We estimate the model using an ordinary jeast square (OLS) estimator. xii We test Hypotheses 2-4 by modifying the basic DID, as discussed later.

MAIN RESULTS
In In Table 2, we report the sample descriptive statistics. We find that an average board is composed of nine directors (Panel A). Firm-focused Insiders hold over 17% of the directorships, whereas the Self-focused Insiders occupy about 7.6% of board seats. xix In Panel B of

À0.031
Notes: In this table, we present the sample statistics. Obs. is the number of firm-year observations per country. Firms identify the number of unique firms per country. Next, we report the year in which each country enters our dataset. AES BEGINS identifies the countries and the year in which countries experienced AES using the Zivot & Andrews (1992) test. DUAL BOARD identifies countries that allow or mandate dual board structure. Following that, we report the GDP per capita (in current $U.S. thousands) and FDI per capita (in current $U.S. millions), respectively. Next, we report the average Guillén & Capron (2016). In the next column, we report the board independence market's average expansion rate (INDP BOARD MARKET). GENDER QUOTA identifies the country that has passed a board gender reform law.    In the full model (Model 4), the results indicate that the coefficient loading on the AES Treatment Â Post is statistically significant and positive at conventional confidence levels (β = 0.012, p < 0.00). This coefficient loading indicates that the proportion of Board Insiders increased by 3.34% in the treated sample compared to the control sample in the post-AES period. xxi In the full model, the coefficient loading on the modified interaction AES Treatment Â Post Â Financial Loss is also positive and statistically significant (β = 0.013, p < 0.00). It indicates an additional 3.62% increase in the levels of Board Insiders in the post-AES period, which is in addition to the 3.34 increase we documented earlier as an average effect. It means that Financial Loss firms increased their Board Insiders levels by about 7% in the post-AES period. At the bottom of the table, we report the F-test for the significance of interactions' incremental explanation. The p-values of the F-test are consistently significant for all models (p < 0.00). The median In this table, we present the difference-in-difference determinants of board configuration model estimates using the within-firm fixed-effect OLS estimator. We report the two-tailed p-values underneath the coefficients within parenthesis. We report the method used to construct all variables in Appendix A. Variables are country (C), firm (I), and time-indexed (T), as shown in the table. We report the model's adjusted R-square and the p-value of the model's overall F-Stat. In the F-test for an incremental explanation, the baseline refers to a model without the main interaction of interest. We cluster the robust standard errors at the firm-level. We report the statistical significances as follows: (Two-tailed) * p < 0.1, ** p < 0.05, *** p < 0.01.

T A B L E 3 Firm-level analysis-determinants of board configuration
T A B L E 4 Director-level analysis-director selection channels AES Treatment x post (2009)   Notes: This table presents the director-level difference-in-difference determinants of board configuration model estimates using the OLS estimator. We report the twotailed p-values underneath the coefficients within parenthesis. We report the method used to construct all variables in Appendix A. Variables are country (C), firm (I), director (J), and time-indexed (T), as shown in the table. We report the model's adjusted R-square and the p-value of the model's overall F-Stat. In the F-test for an incremental explanation, the baseline refers to a model without the main interaction of interest. We cluster the robust standard errors at the firm-level. We report the statistical significances as follows: (Two-tailed) * p < 0.1, ** p < 0.05, *** p < 0.01.
Vvariance inflation factor (VIF) across all models is below 4, indicating a low probability of multicollinearity. Nearly all firms in our sample have their CEOs on the board. Therefore, it is likely that non-CEOs drive almost all of the variations in Board Insiders in the post-AES period. These results provide support to our Hypotheses 1 xxii and 2, respectively. These results are also an endorsement of our country-sorting method based on the AES. Figure 1, which shows the Board Insiders' raw proportion across the AES Treatment and Control samples, gives visual support to our regression results. In Panel B of Table 3, we disaggregate the Board Insiders into Self-focused Insiders and Firm-focused Insiders. We calculate both variables without including the CEOs on the board. Our results suggest that Firmfocused Insiders mainly drive the demand for board insiders among the Financial Loss firms (β = 0.009, p < 0.01) rather than Self-focused Insiders xxiii (β = 0.004, p = 0.06). The coefficient mean-difference test (Paternoster et al., 1998) across the regression estimates in the reported main models suggests that the relative increase in the Firm-focused Insiders compared to Selffocused Insiders is significant at a 10% confidence level. In the robustness test, when we use PSM samples (reported in the Supporting Information document) or we adopt the bootstrapping technique (with 50 resamplings; untabulated), we find that the AES Treatment Â Post Â Financial Loss loads non-significantly when the dependent variable is Self-focused Insiders. When the dependent is Firm-focused Insiders, the coefficient remains large and significant at better confidence levels (β = 0.011; p-value<0.00). In this case, the mean difference test is significant at the 1% level. Therefore, we find support for Hypothesis 4. Table 4 addresses the first mechanism of why Financial Loss firms selected some Board Insiders in the post-AES period. We test this using the AES Treatment Â Â Financial Loss's loading on the director-level Board Insiders' non-fungible expertise (BI Firm Experience). For comparison, we report results using fungible expertise, too, such as industry-experience depth (BI Ind. Exp. Depth) and financial expertise (BI Fin. Expertise). We construct this test at the director-level data to control for observable demographics, work experience, and expertise covariates of the directors, which got omitted in the firmlevel analysis. As discussed in earlier sections, we also include all other board-, CEO, firm-and country-level controls in these tests. Since we estimate the regression models at the director-level, we control time-invariant unobserved factors using industry fixed-effects. AES Treatment Â Post Â Financial Loss loads positively and statistically significantly only for the non-fungible expertise BI Firm Experience (β = 0.005, p < 0.05). Our results are qualitatively similar if we use logit or probit estimators. For brevity, we do not report these results. These results are qualitatively similar if we use the continuous variable to measure BI Net Firm Experience (ln). We report this result in the Supporting Information document. There we also report regression estimates, which show that Firm-focused Insiders drive the non-fungible expertise results. When taken together, these results provide support for Hypothesis 3.

ENDOGENEITY AND ROBUSTNESS ANALYSIS
To further ensure our results' robustness, we adopt the nearest neighbor PSM technique (Guo & Fraser, 2015) for the firm-level analysis. We ensure the robustness of our director-level analysis using other dependent variables. Besides, we perform a battery of other robustness tests. Some of them are as follows: we perform sensitivity tests using alternative sub-samples, model specifications, AES timings, system GMM estimator (untabulated) (Wintoki et al., 2012), and placebo tests. In the interest of brevity, we discuss and report these tests in the Supporting Information document. All our inferences find support from these additional analyses.

Contributions
This study offers new insights on why, when, and to whom board insiders should matter in a cross-country difference-in-difference empirical setting. Our first contribution suggests why board insiders matter to the firms. It is a salient question since most prior studies related to board insiders focus on the agency risk associated with them using the firms' financial performance as its revealing mechanism (Coles et al., 2008;Masulis & Mobbs, 2011;Zorn et al., 2017). It leaves a broad research gap in our understanding of board insiders on why they receive board appointments, especially non-CEOs. Prior research posits that board insiders might have private information, but it is unclear if it is a helpful selection mechanism (Baysinger & Hoskisson, 1990;Adams & Ferreira, 2007;Boivie et al., 2016). We suggest two non-mutually exclusive selection mechanisms by building on the corporate socialization theory. First, board insiders develop a unique social capital by gaining non-fungible expertise over the firm through their longrun tenure as employees and executives. This intra-firm local network-based social capital (Nahapiet & Ghoshal, 1998;Sauerwald et al., 2016) is difficult for outsiders, such as independent directors or consultants, to acquire. Second, board insiders' firm-focus rather than self-focus is quite relevant for the director selection decisions, especially during special conditions, such as in the post-AES period. It also relates to the non-fungible private information generation process since firmfocused board insiders are likely to have their ears-to-the-ground rather than have excessive demands on their executive time through outside appointments. Consistent with our theory, we document that firms appoint longtenured and firm-focused board insiders in the post-AES period.
In our second contribution, we focus on to whom board insiders should matter the most. We show that financially distressed firms, which face the doublejeopardy of weaker investors' perception during a period of macro hardship, appoint more long-tenured and firmfocused board insiders. The financially distressed firms' board appointments during the post-AES period are especially revealing since insiders find a rare moment of agency convergence with its stakeholders. It is based on self-preservation and the firm's survival. During such a period, we document that these firms have a stronger demand for long-tenured firm-focused board insiders, which is in addition to the average firms' needs. It suggests that despite considerable monitoring challenges (Adams & Ferreira, 2007), such directors could provide assistance and future direction through their non-fungible private information sources.
Our third contribution is on when board insiders should matter to the firms. Owing to the endogeneity challenges imposed on the empirical corporate governance research (Hermalin & Weisbach, 1998;Wintoki et al., 2012), it is generally difficult to ascertain with a high degree of certainty why some firms have a specific board configuration (Bhagat & Black, 2001;Boivie et al., 2016). An emerging strand of empirical literature has responded to this challenge by using unique identification strategies such as "sudden deaths" and "CEOonly" boards (Nguyen & Nielsen, 2010;Zorn et al., 2017). Others have exploited regulatory changes (Ahern & Dittmar, 2012;Choi et al., 2007;Duchin et al., 2010;Fauver et al., 2017). We contribute to this literature by adding a new cross-country context of the aggregate earnings shocks. Using this setting, we document that even non-R&D intensive firms (Coles et al., 2008) appoint more board insiders in the postmacro shock period. We causally infer these results due to the low endogeneity risk.

Implications, limitations, and future directions
Several implications and limitations follow. Studying the US context, Zorn et al. (2017) have documented that a significant proportion of firms have resorted to the "CEO-only" board. Globally, policymakers have invested considerable regulative and legislative time and effort to increase the board's monitoring outcomes through higher board independence (Fauver et al., 2017). Nevertheless, in this study, we document that firms have reversed the declining trend of the board insiders in the AES treatment countries during the post-period. Most interestingly, we notice that globally, most regulators have adopted board gender diversity as the primary regulatory instrument (see Table 1) instead of focusing on the board's monitoring needs. While board gender diversity is a vital dimension of the board, regulators should increase their vigilance to ensure the board insiders' increase does not adversely affect the firms' stakeholders by being co-opted by the CEOs (Coles et al., 2014). On their part, investors must assess how the newly minted board insiders are likely to use their non-fungible expertise and long-tenured experience to benefit the firms' performance and policies by taking collective responsibility (Feinberg, 1968). Or are they harbingers of agency risks?
Our study has some other limitations too. These results illuminate the non-CEO board insider selection strategy within the industrial sector firms, especially in the post-AES period. Our study leaves open questions about what personality types, gender, social identities, behavioral traits, or corporate culture (Camerer & Vepsalainen, 1988;Ostroff & Kozlowski, 1992) enable or hinder certain kinds of corporate employees or department leaders from rising to the board positions. Future studies could explore these questions using the AES context. Finally, studies could also explore how AES or, more generally, the financial crisis affects the financial and utility sector firms' governance and performance in a cross-country setting.

ENDNOTES
i It does not rule out the possibility that independent directors could also provide quality advice. Nonetheless, to do so, they would have to depend on the same board insiders for firm-specific private information, giving them a greater context to facilitate quality advice (Adams & Ferreira, 2007).
ii Corporate governance literature faces a pronounced risk of endogeneity, where concerns about reverse causality make an inference of regression results from a causal perspective difficult if not impossible (Hermalin & Weisbach, 1998). A significant macro shock faced by some countries due to the spill-over effect of the 2008-2010 global financial crisis (or AES) allows us to test the board configuration choices in a difference-in-difference setting (Card & Krueger, 1994). It provides a cleaner interpretation of the results. iii This literature posits that aggregate earnings have information value for the macro-economy (see, e.g., Konchitchki &Patatoukas, 2014 andShivakumar &Urcan, 2017). We posit that an unexpected short-term or permanent AES would affect how corporate boards would configure themselves in the treatment countries compared to the firms in the control countries, which did not experience such a shock (statistically not significant). However, we do not form a strong hypothesis about the direction of the shock in the aggregate earnings series, as that is not the purpose of this study. iv Using similar theoretical logic, experience, and expertise from various kinds of board diversity, including but not limited to gender, is also fungible. It means that the board need not appoint women or other markers of a diverse board through the board insider channel to receive information processing advantages that come with diversity (Pirson & Turnbull, 2011). v Coles et al. (2008) show that R&D-intensive firms, which require more advice owing to the firm's complexity, appoint a higher proportion of insiders within their boards. vi We deleted all financial (6,000-6,999) and utility sector (4,900-4,950) firms from our sample after identifying them using the two-digit SIC codes matched with Fama & French's (1997) industry classifiers (FF48 Industry). vii The time-series length of our panel dataset is 14 years. We assume that relative purchasing power parity holds during this period (Taylor, 2002). We used the yearly-average Home/US$ currency rates to convert local currencies into US$ equivalents. The historic exchangerate data is also from the World Bank database. viii Zivot and Andrews' (1992) test draws on the unit-root literature whose basis was laid by Nelson and Plosser (1982) and Perron (1989). Zivot and Andrews' (1992) study also includes an empirical application of their test with various macroeconomic series. They have real and nominal Gross National Product (GNP), Employment, Industrial Production, etcetera, using original US data from the Nelson and Plosser (1982) paper. See Hansen (2001) for a survey of the literature. ix Given that board configuration decision tends to persist over the years (Hermalin & Weisbach, 1998), even a temporary AES could trigger board configuration realignments. We do not distinguish between a temporary or a permanent shock to the aggregate earnings series for simplicity.
x Studies, usually from the US, do not use any such treatment to identify the recent global financial crisis's timeline because of the broad consensus about its origins and timings (Adams, 2012;Carey et al., 2012). However, in cross-country data, sorting countries into treatment and control samples based on which ones experienced significant macro shock without such a treatment would require considerable valuejudgment. We avoid such arbitrary selection by employing the Zivot-Andrews test (1992). xi It helps address the concerns expressed about the "endogenous" selection of the treatment sample resulting in questions raised about the DID-type models' econometric validity (Bertrand et al., 2004). xii During the estimation process, we mean-centered all firm-year and country-year variables at their respective levels (Hofmann & Gavin, 1998). xiii As discussed in the Supporting Information (available from the authors), for robustness, we also create a separate set of aggregate earnings series by averaging the operating earnings of an extended sample, which includes financial and utility sector firms. Our structural break results remained mostly consistent. Adopting a conservative approach for our main tests, we use the industrial sector break years to match our sample closely. xiv We confined the annual aggregate earnings series to match the BoardEx's timeline. Elongating the timeline deeper into the 1990s would potentially include the Dot-Com Bubble, also originating in the US, and its cascading effects in other countries. The present Zivot-Andrews test is ill-equipped to track two structural breaks in the time-series and would, therefore, necessitate other tests. Also, corporate board data is unavailable from BoardEx to match a longer timeline. Our timeline matches the conventionally known sub-prime mortgage-induced global financial crisis originating in the US and is accepted by standard financial economics research (Adams, 2012;Carey et al., 2012). xv There is little doubt that US experienced a financial crisis starting September 2008 when Lehman Brothers filed for Chapter 11 under US bankruptcy protection laws. However, we adopt a conservative approach based on the cash-heavy aggregate earnings series. Our Zivot-Andrews test captures the break date for the US in the year 2010. This timing is reasonable given that most emerging and advanced countries practice accruals accounting (Leuz et al., 2003). It allows firms to book income in advance, which is likely to materialize to a reasonable degree of confidence. In this way, the AES's full impact is likely to affect the US firms with a lag. The Zivot-Andrews test captured it in 2010. Kalemli-Özcan et al. (2016) have suggested that Greece, Ireland, Spain, Portugal, and Italy have experienced a "sovereign debt" crisis (p. 1) during the same period. Our Zivot-Andrews test captures break dates for Ireland (2008), Portugal (2008) and Spain (2008), if the statistical significance used to reject the null is weaker, at 10% (in the reported results, Spain is part of the control sample). For Greece, we notice that the aggregate earnings series declined monotonically from a high of 16 percent in 2004 to a low of 1.8% in 2013. Therefore, their aggregate earnings suffered a systematic yet gradual decline eliminating the potential of a sudden shock. Italian aggregate earnings were volatile throughout the time-period with steep and frequent ups-and-downs, thereby making it difficult to pinpoint an exact timeline of a shock as reflected in the non-significant result emerging from the Zivot-Andrews test. In our tests, we treat both Greece and Italy as control sample countries. xvi Of the nine countries, only two-Belgium and Ireland-were captured by the Zivot-Andrews test to have a structural break in 2004/2005. In 2009 too, these two countries suffered a significant decline in the mean. It matches our timeline from the rest of the sample countries. Therefore, we manually shifted their break dates to 2009. Our results remained unaffected from these minor subjective interventions. xvii A critical research design success of our method is that we could weaken the institutional homogeneity in the AES Treatment and Control subsamples, as countries from both legal-origins, common-law, and code-law, were represented in each (La Porta et al., 1998). It increases institutional variations in both sub-samples and reduces the risk of institutional homogeneity.
xviii Firms in almost all countries were affected by the financial crisis 2008-2010. Using Zivot-Andrews AES timings, our objective is not to identify all affected countries. Our objective is to separate those countries that were significantly affected from those that were not as significantly affected. Here the statistical test proposed by Zivot-Andrews contributes to identifying the significantly-affected from the not-so-significantly-affected countries through a unit-root test. We, therefore, cannot rule out some misidentification of countries in either treatment or control samples. Nevertheless, this method is superior to identifying firms into treatment/control samples based on a priori information about the crisis-affected countries without any appropriate econometric treatment. xix Masulis & Mobbs (2011) find that Self-focused Insiders hold about 10% of US directorships. xx These figures are understated as the sample average is measured by counting independent directors as well.
xxi Following Fauver et al. (2017), we calculate the coefficients' economic magnitude as follows: 3.34% = 0.012/0.359, where 0.012 is the coefficient loading on AES Treatment x Post in Model 4 of Table 3. Whereas 0.359 is the mean proportion of Board Insiders, as reported in Table 2, Panel A. Here onwards, we use the same method to calculate the economic magnitudes. xxii To rule out alternative explanations such as whether R&D-intensive firms drive our results, we interact the AES Treatment x Post with the proportion of R&D to Total Assets (instead of Financial Loss) and reestimate Model 4 as reported in Table 3. In this model, the interaction Treatment x Post x R&D does not load statistically significantly. Nevertheless, our primary interaction -AES Treatment x Postremains positive and statistically significant at conventional confidence levels (p < 0.05). This robustness check indicates that unlike Coles et al. (2008), R&D-intensive firms do not drive our primary results. xxiii To ensure that our results are not driven by declining board sizes, as a robustness check, we use the natural log of the number of Self-focused Insiders and Firm-focused Insiders, respectively, per firm-year as alternative dependent variables. All our results are consistent. These results are untabulated.