THE QUALITY OF CORPORATE GOVERNANCE AND DIRECTORS’ ELECTIONS

How to cite this paper: Berthelot, S., Coulmont, M., & Gagné, V. (2022). The quality of corporate governance and directors’ elections. Corporate Ownership & Control,


INTRODUCTION
The many financial scandals of the 2000s led to the introduction of more stringent regulations on listed companies' governance practices in most Western countries, as illustrated by the Sarbanes-Oxley Act (SOX) (2002) Government, 2002) in Canada (as cited in Héroux & Roussy, 2020). Some countries, like the United States, adopted a mandatory rule-based approach to corporate governance, while others, like Canada, opted for a principles-based "comply or explain" approach. As part of the latter approach, regulatory authorities developed an overarching set of principles, accompanied by suggested "best practices" (Salterio, Conrod, & Schmidt, 2013). Firms may voluntarily adopt the recommended best practices (i.e., comply) or "explain" how they will comply with the governance principle underlying each unadopted practice (Luo & Salterio, 2014). The disciplinary power of this approach lies in the required public disclosure of corporate governance practices, which enables investors and other stakeholders to evaluate the effectiveness of the firm's governance system and make informed decisions (Luo & Salterio, 2014).
Along with these more stringent regulations, a number of studies were conducted to assess the validity of the governance practices put forward in each type of approach (Zhang, 2007 & Salterio, 2014). The results of these studies are mixed. The discrepancies noted may be explained by the different practices studied, the endogeneity of the explanatory variables, the specific context of the country under study (Kumar & Zattoni, 2013) or the nature of the benefits expected from the governance practices examined. Some researchers have studied their impact on firm performance, while others have investigated whether they serve to mitigate investors' complaints (Aguilera & Cuervo-Cazurra, 2009).
Although directors' elections have been the focus of few studies, they are particularly important in the principles-based "comply or explain" approach. They are the mechanism that allows shareholders to express their dissatisfaction with the practices of a firm's board of directors, which is usually elected at annual general meetings. Since the board is the highest legal authority in a firm (Petrovic, 2008), its directors are apt to have a significant impact on the firm's choice of strategic directions; the appointment, oversight and monitoring of management; communications with investors; and the implementation of governance practices aimed at fulfilling these different roles. Should these roles be inadequately fulfilled, shareholders can express their dissatisfaction through directors' elections with a view to introducing the necessary changes. The object of this study is therefore to determine whether shareholders take the quality of a firm's governance practices into account when electing directors. Our sample is comprised of Canadian firms listed on the Toronto Stock Exchange. Like Berthelot et al. (2010), Bozec and Dia (2012), and Lin and Lin (2020), we use the corporate governance practices index published by The Globe and Mail to operationalise the quality of the governance practices.
The results of our analyses indicate that shareholders take the quality of governance practices into account when voting at directors' elections. However, these findings appear to be sensitive to the different aspects of governance practices examined. We also found the voting percentages to be very high, with an average of 97% (median = 98.2%) of the votes being cast in favour of the candidates nominated. These percentages raise a number of questions about the factors shareholders consider when casting their votes.
This paper contributes to the literature in two ways. First, it makes it possible to document a major component of corporate governance that has received little attention in prior studies. It is true that shareholders can vote with their feet by selling their shares, but directors' elections offer them a way to express their opinion while maintaining their investment. In countries that have adopted a governance principles-based "comply or explain" approach, it is certainly worthwhile examining the actual disciplinary impact of directors' elections. Our findings tend to show that they have a genuine effect. Second, our results also show that Canadian firms comply with the governance practices put forward by the business community and set out in The Globe and Mail index. The average score of firms in the sample is 72.27% (median = 73%). This relatively high percentage of compliance confirms observations by Salterio et al. (2013), who noted that an average 82% of the Canadian firms in their sample complied with the best practices recommended by the Canadian Securities Administrators, 4% of which complied through explanations (Salterio et al., 2013). Since the principles-based "comply or explain" approach does not require companies to implement defined practices, it minimises costs for all firms. Overall, our findings support the validity of this approach.
The remainder of the paper is organised as follows. The next section reviews the relevant literature. The third section describes the sample and the empirical model, while the fourth presents the main results. The fifth section provides a discussion of the main findings. Finally, the last section discusses the study's main conclusions, its limitations, and potential avenues for future research.

LITERATURE REVIEW
According to Cheffins (2013), the history of corporate governance dates from at least the founding of the East India Company, the Hudson's Bay Company, the Levant Company, and other major chartered companies launched in the 16th and 17th centuries. The establishment of these early companies clearly mirrors the agency, risk-sharing, and agent monitoring problems that can arise between shareholders and managers. These problems derive from the separation of ownership and control; shareholders own the firm, while managers control it (Kim, Nofsinger, & Mohr, 2010). The former seek a return on their investment even though it may imply a higher risk, whereas the latter seek to enhance their own well-being through minimising risk and increasing salaries, perks, and other benefits. These problems have been documented by many researchers (Bendickson, Muldoon, Liguori, & Davis, 2016) who examined such various issues as mechanisms to reduce excessive incentive compensation, executive stock options, the percentage of independent board members, and splitting the roles of CEO and board chair. Berle and Means (1932) were among the first to show that owners' ability to monitor managers declined as owners become increasingly diverse and spread out and that this problem that could only be redressed through government action (Bendickson et al., 2016).
Over the years and in the wake of numerous financial scandals, most Western countries introduced legislation that Aguilera, Goyer, and Kabbach de Castro (2013) attribute in large part to the strategy of international diversification of institutional investors from liberal market economies (Goyer, 2006;Clark & Wojcik, 2007). However, this legislation differs from one country to the next. Aguilera et al. (2013) attribute these differences to the structure of existing institutional arrangements. According to Shleifer and Vishny (1997), "much of the difference in corporate governance systems around the world stems from the differences in the nature of legal obligations that managers have to the financiers, as well as in the differences in how courts interpret and enforce these obligations" (p. 750). Various Western countries have adopted one of the two corporate governance regulation approaches. The mandatory rule-based approach, which is in the form of statutory rules intended to prohibit certain kinds of behaviour, is characterised by the use of a "one size fits all" approach designed to address common governance problems. The principles-based "comply or explain" approach recommends standards of best practice, leaving compliance determination to individual firms. In the latter approach, firms are not legally bound to adopt the best practices but can implement the governance's practices that best suit their particular contingencies and explain their choice (Aguilera et al., 2013). The disciplinary power of this approach lies in firms being legally required to disclose their governance practices (Luo & Salterio, 2014) and investors being able to appreciate these practices or the explanation. Depending upon a particular firm's actions, investors can sell their shares, do nothing, or keep their shares and try to effect change by communicating with management (Hirschman, 1970). Casting their votes at annual general meetings is one way for them to achieve this communication.
Encouraged by the European Union Directive 2006/46/EC, a number of countries adopted the principles-based "comply or explain" approach. In contrast, the United States government opted for the mandatory rule-based approach ( (Salterio et al., 2013). Although these requirements mandate corporate governance disclosures, recommended governance practices follow the "comply or explain" approach. According to these practices, the board should be composed of a majority of independent directors; nominating and compensation committees should be composed entirely of independent directors; independent directors should have the opportunity to hold regularly scheduled meetings without nonindependent directors and managers; the board should adopt a written mandate; the firm should draft a clear position description for the chair, the chair of each board committee and the CEO; provision should be provided for all new directors to have access to a comprehensive orientation; continuing education opportunities should be available to all directors; and the board should adopt and monitor a written code of business conduct and ethics. Lastly, National Policy 58-201 recommends that the effectiveness and contribution of the board, its committees and each individual director should be regularly assessed.
In conjunction with these regulations, some financial market players, such as proxy advisors (e.g., Institutional Shareholder Services Inc. and Governance Metrics International), commercial governance rating agencies, and even the media, proposed numerous other "good" governance practices (Sauerwald, (Bozec & Dia, 2012). This index includes all the elements of Canadian National Policy 58-201 (Corporate Governance Guidelines), except for the adoption and monitoring of a written code of business conduct and ethics by the board, as well as several other elements respecting board composition, executive compensation, shareholder rights, disclosures on director relationships, board member biographies and age, voting results for directors and say-on-pay, director attendance at board and committee meetings, the value of directors' equity holdings, and provision for a retirement age policy. Table 1 presents the content of The Globe and Mail's corporate governance index. An interesting feature of this index is that, unlike the other governance rating systems developed by proxy advisors (e.g., Institutional Shareholder Services Inc. and Governance Metrics International) and commercial governance rating agencies, it is available to all investors at a relatively low cost, which means that specialists are not the only ones who can afford to purchase costly data (Berthelot et al., 2010). Splitting of the roles of board chair and CEO. 5 Presence of women board members and disclosure of a plan for female representation on the board. 5 Presence of a system for evaluating directors' performance, a CEO succession plan and a continuing education plan. 8 Provision for independent directors to meet without management. 3

Shareholding and compensation (29 marks out of 100)
Requirement to hold shares or units, values and minimum number.
13 Quality of the disclosure of the CEO's compensation agreement, disclosure of targets to be reached, justification of CEO's performance, existence of an anti-hedging policy. 7 Utilisation and justification of the composition of a peer comparison group for the CEO; goals to be attained other than a higher share price. 5 Presentation of CEO and executives' compensation for the last five years. 4

Shareholder rights (27 marks out of 100)
Implementation of say-on-pay.
3 Report of the annual voting results for each item in the proxy circular.
1 Provision for a "claw-back" policy, a holding period for shares after a CEO's departure, a policy on compensation when there is a change of control, opportunity for the shareholders to meet with the board. 4 Stock option dilution effect, option grant rate, publication of the dilution effect on common shares. 8 Directors' compensation based on firm performance. 1 All shareholders have equal voting rights.
10 Disclosure (10 marks out of 100) Disclosure of the directors' biographies, relationships, attendance rate at board meetings; results of directors' elections for say-on-pay. 5 Disclosure of the number/value of shares held and the establishment of a minimum number/value to be held. 3 Disclosure of directors' age and retirement policy. 2 According to Hirschman (1970), shareholders have three options open to them when they are dissatisfied with corporate behaviour (Berthelot & Coulmont, 2020). They can exit (sell their shares), maintain the status quo (hold on to their shares) or communicate with management (Hillman, Shropshire, Certo, Dalton, & Dalton, 2011; Berthelot & Coulmont, 2020). Several studies have investigated ties between corporate governance practices, operationalised on the basis of various indices or through individual practices, and different accounting (e.g., ROA, ROE, EPS) and financial measures (e.g., market-to-book, Tobin's Q, market value added). The findings of these studies are mixed, making it difficult to determine to what extent shareholders take firms' governance practices into account in the prices attributed to their shares, mainly because of the diversity of the governance practices and measures studied. Another means of assessing the extent to which shareholders take corporate governance practices into account is to examine their voting at directors' elections, which is one of the ways they can communicate their satisfaction or dissatisfaction with firms or their representatives. Since one of the directors' key roles is to implement good governance practices to reach the firm's organisational goals and, in turn, to ensure its sustainability, it is worth investigating whether the exercise of this important shareholder right (Shleifer & Vishny, 1997) is really used to approve or disapprove of the directors' performance in terms of governance practices We thus propose the following hypothesis: H1: The "in favour" votes at directors' elections are positively related to the quality of corporate governance practices.

Sample
The study sample is composed of all Canadian companies covered by The Globe and Mail corporate governance ratings for 2015 and 2016 for which the proxy circulars and annual general meeting vote results were available on the www.sedar.com 1 database. The financial data were derived from the S&P Capital IQ database. The corporate governance index is composed of all the firms listed on the S&P TSX composite index, which includes some 240 firms with the largest market capitalisation on the Toronto Stock Exchange. As previously indicated, these index ranking scores are calculated using a 100-point scale comprising four components (see Table 1

Empirical model
To our knowledge, with the exception of studies by Cai, Garner, and Walkling (2009) and Berthelot and Coulmont (2018), little research has focused on the determinants of shareholder votes in directors' elections. Accordingly, the regression model used to examine whether shareholders take into account the quality of the governance practices instituted by directors when casting their votes in directors' elections is based on the model developed by Berthelot and Coulmont (2018) and Cai et al. (2009), using data that could be collected in the Canadian context. This ordinary least squares (OLS) regression model includes the average of in favour votes cast by shareholders at annual general meetings for all directors on the board of the firm as the dependent variable, the corporate governance index score as the independent variable to test our hypothesis, and several underlying corporate governance firm characteristics, including CEO compensation and some financial performance variables as control variables. The CEO compensation growth and financial performance variables are control variables associated with key director roles -such as monitoring managers as fiduciaries of stakeholders, particularly by establishing a compensation scheme that aligns managers' and shareholders' interests (controlled by CEO total compensation growth (ΔTCOMPit), including cash and non-cash compensation components or short-term and longterm components) and advising and assisting managers in the firm's strategic planning and the implementation of such plans (Johnson, Daily, & Ellstrand, 1996) in order to ensure the company's financial performance and sustainability. This latter role is controlled by two financial performance measures; an accounting measure, return on assets (ROAit), and a financial market measure, stock return (RETURNit). The model also includes other control variables that could have an impact on the shareholders' votes. One of these is revenue growth (GROWit), which measures whether the firm is in a growth stage, in which case it could be more focused on long-term profitability and less on the return on assets (ROAit). In contrast to Cai et al.
(2009), we did not include the percentage of independent directors and the fact that the CEO is board chair because these variables are included in the corporate governance index score. However, we included the number of board members (SIZEBOARDit) and the logarithm of the firm's total assets (LnASSETit). The size of the board should reflect the diversity and range of the directors' ability to make decisions and the logarithm of the firm's total assets, which is a measure that controls for firm size. As larger companies receive more media coverage, it is possible that more information about these companies is available and that this affects the shareholders' votes when electing directors. We also added the variable BLOCKit that controls whether the firm is a concentrated shareholding structure. Lastly, we also included a dummy variable that controls the fiscal year. This OLS regression model is expressed as follows (equation (1) (2) where, VOTEit is the average of the votes cast by shareholders in the election of directors at the annual general meeting for year t; CGITit is the total score of the firm on The Globe and Mail's corporate governance index for year t; ROAit is the return on assets at the end of year t; RETURNit is the stock return for year t; ΔTCOMPit is the growth of CEO total compensation for year t; GROWit is the revenue growth (total assets at the end of year t/total assets at the end of the previous year) of year t; SIZEBOARDit is the number of members on the board of firm i at the end of year t; BLOCKit is a dummy variable equal to the value of 1 if at least one investor holds more than 10% of shares and 0 otherwise at the annual general meeting covering the year t; LnASSETit is the logarithm of the total assets of firm i at the end of year t; YEARSit is a dummy variable equal to the value of 1 if the fiscal year is 2016 and 0 otherwise, and is the error term. If H1 is confirmed, 1 , the regression coefficient associated with the corporate governance index score will be positive and significant. In this case, the average of the shareholder votes cast at the directors' election would be higher when the firm's corporate governance practices obtain a better score in The Globe and Mail's corporate governance index.
We then broke down the corporate governance index into its four subcomponents to examine whether shareholders take some issues into account more than others when electing directors. The regression model used was thus as follows in equation (2), where, CGIBCit is the corporate governance index (CGI) sub-index score on board composition for firm i in year t; CGISCit is the CGI sub-index score on shareholding and compensation for firm i in year t; CGISRit is the CGI sub-index score on shareholder rights for firm i in year t, and CGIDit is the CGI sub-index score on board governance disclosure for firm i in year t. Table 2 breaks down the study observations according to the activity sector. As can be seen, close to 60% of the observations refer to firms in the energy, materials, and financial sectors. These sectors are also over-represented in the Toronto Stock Exchange S&P TSX Composite Index. The consumer discretionary and industrial sectors are next, with some 10% of observations each, followed by the other sectors with lower percentages.  Table 3 presents the descriptive statistics for the corporate governance index measures. Overall, the firms in the sample post relatively high scores. The mean scores obtained with the total corporate governance index (CGITit) is 72.44% and the median is 73%. The firms in the sample presented lower scores in the corporate governance index sub-score associated with shareholding and compensation (CGISCit) (mean = 19.97/29 or 68.86%) and higher scores in the corporate governance index sub-score associated with disclosure (CGIDit) (mean = 8.15/10.5 or 80.95%).  The return on assets (ROAit) of the sample firms is an average of 8% (median = 1.6%), while the average return on shares (RETURNit) is 12.7% (median = 8.3%). Average sales growth (GROWit) is 5% (median = 3.2%).

index (CGI) sub-index score on board composition for firm i in year t; CGISCit is the CGI sub-index score on shareholding and compensation for firm i in year t; CGISRit is the CGI sub-index score on shareholder rights for firm i in year t; CGIDit is the CGI sub-index score on board governance disclosure for firm i in year t.
These firms are relatively large with average total assets (ASSETit) of CAD$45.904 billion (median = 4.779 billion).

* in millions of Canadian dollars. VOTEit is the average of the votes cast by shareholders in the election of directors at the annual general meeting for year t; ROAit is the return on assets at the end of year t; RETURNit is the stock return for year t; ΔTCOMPit is the growth of CEO total compensation for year t; GROWit is the revenue growth (total assets at the end of year t/total assets at the end of previous year) for year t; SIZEBOARDit is the number of members on the board of firm i at the end of year t; ASSETit is the logarithm of the total assets of firm i at the end of year t.
The correlation coefficients between the variables included in the regression analyses are set out in Table 5. Apart from the variables associated with the corporate governance index (CGITit, CGIBCit, CGISCit, CGITSRit, and CGIDit) showing correlation coefficients from 0.379 to 0.853 and the firm's total assets (ASSETit) showing a correlation coefficient of 0.637 with the board size (SIZEBOARDit), the other correlation coefficients are relatively weak.

.01. VOTEit is the average of the votes cast by shareholders in the election of directors at the annual general meeting for year t; ROAit is the return on assets at the end of year t; RETURNit is the stock return for year t; ΔTCOMPit is the growth of CEO total compensation in year t; GROWit is the revenue growth (total assets at the end of year t/total assets at the end of previous year) in year t; SIZEBOARDit is the number of members on the board of firm i at the end of year t; BLOCKit is a dummy variable equal to the value of 1 if at least one investor holds more than 10% of shares and 0 otherwise at the annual general meeting covering year t; LnASSETit is the logarithm of the total assets of firm i at the end of year t; YEARSit is a dummy variable equal to the value of 1 if the fiscal year is 2016; CGITit is the total score of the firm on The Globe and Mail's corporate governance index in the year t and 0 otherwise; CGIBCit is the corporate governance index sub-index score on board composition for firm i in year t; CGISCit is the corporate governance index sub-index score on shareholding and compensation for firm i in year t; CGISRit is the corporate governance index sub-index score on shareholder rights for firm i in year t, and CGIDit is the corporate
governance index sub-index score on board governance disclosure for firm i in year t; and ε i is the error term. Table 6 presents the results of the regression analyses intended to corroborate H1. None of the models show a variance inflation factor (VIF) higher than 2, which seems to indicate that there is little multicollinearity between the variables included in these models (Hair, Black, Babin, & Anderson, 2009).

Regression analysis results
The calculations for Model 1 exclude the corporate governance index (CGITit) and its subcomponents. As can be seen, this model, which includes only the control variables, explains 10.2% of the variance in the average shareholders' votes in directors' elections at annual general meetings. The coefficient associated with CEO total compensation growth (ΔTCOMPit) is negative and significant with an error threshold of 0.05, while the coefficient associated with board size (SIZEBOARDit) is positive and significant with an error threshold of 0.05. These results are consistent with those of Berthelot and Coulmont (2018). Shareholders' votes are less in favour of director nominees when CEO compensation growth is more significant. These findings thus tend to show that shareholders use directors' elections to express their dissatisfaction with substantial increases in CEO compensation.
The positive and significant coefficient associated with board size could reflect shareholders' possible appreciation of the broader variety of competencies that large boards represent or the greater range of individual director's resources. These results are constant in the six models. Apart from the coefficient of the variable associated with a shareholder structure where at least one shareholder holds more than 10% of the share capital (BLOCKit), which is marginally positive and significant with an error threshold of 0.10, the coefficients are not significant.
Model 2 is comprised of the same variables as Model 1, with the addition of the total corporate governance index (CGITit). The explanatory threshold of the variance (adjusted R 2 ) of the average of "in favour" shareholder votes rose by 0.057, climbing to 0.159. The coefficient of the variable associated with the total corporate governance index (CGITit) is positive as predicted and significant with an error threshold of 0.001, which corroborates H1. Shareholders thus appear to take the quality of the governance practices implemented by directors into account when casting their votes. On the whole, Model 2 tends to show that shareholders do not consider directors to be responsible for firm performance since the coefficients of the variables respecting return on assets (ROAit) and stock return (RETURNit) are not significant. However, they do take into account the control that directors are expected to exercise over the growth of CEO compensation and the quality of governance practices they must implement.
Models 3 to 6 replicate the analyses in Model 2 but include each of the individual sub-indices (CGIBCit, CGISCit, CHISRit, and CGIDit) rather than the total corporate governance index (CGITit). As Model 3 shows, the coefficient of the sub-index associated with board composition is positive and significant, with an error threshold of 0.001. The explanatory threshold of the variance (adjusted R 2 ) in the average "in favour" votes is 0.192. Shareholders, therefore, appear to grant a certain measure of importance to directors' independence, split CEO/Chair roles, close-knit blocks of directors, the presence of women on the board, the implementation of a policy on gender diversity for the board and senior managers, the implementation of a system to evaluate board performance, the opportunity for independent directors to meet without management, CEO succession planning, and directors' education and training.
Model 4 includes the sub-index associated with shareholding and compensation (CGISCit). The coefficient of this variable is positive and significant with an error threshold of 0.001. Shareholders thus seem to take into account shareholding and executive compensation issues when electing directors. This sub-index refers more specifically to the requirements for directors and the CEO to hold shares or share units, the value of their equity holdings, the implementation of an antihedging policy, and the transparency and policies associated with CEO compensation. The explanatory threshold of the variance (adjusted R 2 ) of the average shareholder votes "in favour" of the directors nominated is 0.151.
Model 5 shows that the coefficient of the subindex related to shareholder rights (CGISRit) is not significant and the adjusted R 2 of this model is 0.10, as it is in Model 1. Shareholders, therefore, seem to grant less importance to the following types of disclosures: the implementation of a say-on-pay vote on executive compensation; annual voting results for each item in the report on voting results in the proxy circular; a provision to "claw back" bonus payments to the CEO if wrongdoing is discovered; a provision for a holding period for shares after a CEO leaves the firm to ensure there is a performance "tail" to the CEO's work; requirement for a double trigger before paying compensation; permitting equity units to vest for top executives upon a change of control; excessively diluted stock options; an excessive annual stock option grant rate; a vesting period before options can be exercised; the firm's year-end dilution level of stock options as a percentage of shares outstanding; the prior year's rate for option grants as a percentage of shares outstanding; a provision for awarding stock options to directors; and equal voting rights for all shareholders. It should be noted that the founders of a number of Canadian firms in the sample hold categories of shares conferring more voting rights than the shares issued and outstanding on the Toronto Stock Exchange. This may at least partly explain these non-significant results in relation to shareholder rights.
The coefficient associated with the sub-index on disclosure (CGIDit) presented in Model 6 is positive and significant at 0.001, indicating that shareholders take into account the disclosure practices put forward by the board when electing directors. The information covered by the sub-index specifically relates to relationships between directors; the age and biography of each director; the voting results in the prior year's board elections; whether the directors attend all meetings and the company removes directors with poor attendance; the total accumulated value (a dollar amount, not just the number of units held) of the directors' equity holdings, including shares and vested deferred share units; whether each director's share ownership meets (or fails to meet) the required share ownership guideline; and whether or not the firm has a retirement age policy for directors, as well as the details of this policy. Model 6 explains 15% (adjusted R 2 ) the variance of the mean of the shareholder vote "in favour" of the set of directors nominated at the annual general meeting.

DISCUSSION
Overall, our analyses, including the total corporate governance index and three of its sub-indices, support hypothesis (H1). Although the percentages of votes for the nominees are very high, our findings show that the shareholders' votes are not only procedural in nature, but that they also reflect, at least in part, shareholders' expectations. It may thus be concluded that acquiring or selling shares is not the only way shareholders communicate their expectations. They actively participate in electing directors through votes expressing their approval (or disapproval) of how the directors fulfil their duties and responsibilities.
Of all the different elements included in the corporate governance index, those associated with board composition (e.g., the percentage of independent board members and its committees, the splitting of the roles of board chair and CEO, the presence of women board members, the presence of a system for evaluating directors' performance, a CEO succession plan, and a continuing education plan) appear to have a stronger impact on shareholder votes. Note that these are essentially the same elements set out in the Canadian Securities Administrators' Policy 58-201. Our results thus seem to corroborate the importance that these practices represent for shareholders of Canadian companies. They also support the rationale for regulation by the Canadian Securities Administrators.
When electing directors, shareholders also take into account firms' shareholding and compensation practices (e.g., the requirement to hold shares or units, values and minimum number, the quality of the disclosure of the CEO's compensation agreement, the utilisation and justification of the composition of a peer comparison group for the CEO, and the presentation of the CEO and executives' compensation for the last five years) in a way that is complementary to the growth of CEO compensation since, according to the results observed, it negatively impacts shareholders' votes. Our observations of shareholder votes substantiate the importance granted to these practices by financial market actors such as institutional investors, proxy advisors, shareholder activists, and so forth. In fact, this stakeholder consideration supports the assumption that these practices are being institutionalised in firms without the need for specific reinforcement (for some of the elements studied) of the principle-based "comply or explain" regulation or even mandatory rule-based regulation.
Practices associated with the disclosure of the information about directors (e.g., disclosure of their biographies, relationships, the attendance rate at board meetings; the number/value of shares held and the establishment of a minimum number/value to be held; age and retirement policy) are also reflected in shareholders' votes at directors' elections. Shareholders thus seem to appreciate firms' transparency about director nominees.
Furthermore, when electing directors, shareholders do not seem to take into account practices respecting shareholders rights (e.g., implementation of say-on-pay, reporting of the annual voting results for each item in the proxy circular, provision for a "claw-back" policy, a holding period for shares after a CEO's departure, a policy on compensation when there is a change of control, opportunity for the shareholders to meet with the board, the stock option dilution effect, option grant rate, publication of the dilution effect on common shares, the equal voting rights of shareholders, and directors' compensation based on firm performance). It should be noted that these practices are relatively diverse and that a number of them have been advocated by shareholder activists rather than by the Canadian Securities Administrators.
Our results add to the knowledge about the principle-based "comply or explain" approach for regulating public firms' corporate governance practices. Shareholders appear to be playing the role expected of them. Our findings also show that for certain practices put forward by the business community, shareholders have been stakeholders in processes that led to their institutionalisation. In summary, although shareholders seem to exercise their voting rights with caution (the mean of the in favour votes is very high), they also vote in a manner that can ensure the discipline of directors. ΔTCOMPit is the growth of CEO total compensation in year t; GROWit is the revenue growth (total assets at the end of year t/total assets at the end of previous year) in year t; SIZEBOARDit is the number of members on the board of firm i at the end of year t; BLOCKit is a dummy variable equal to the value of 1 if at least one investor holds more than 10% of shares and 0 otherwise at the annual general meeting covering year t; LnASSETit is the logarithm of the total assets of firm i at the end of year t; YEARSit is a dummy variable equal to the value of 1 if the fiscal year is 2016; CGITit is the total score of the firm on The Globe and Mail's corporate governance index in the year t and 0 otherwise; CGIBCit is the corporate governance index sub-index score on board composition for firm i in year t; CGISCit is the corporate governance index sub-index score on shareholding and compensation for firm i in year t; CGISRit is the corporate governance index sub-index score on shareholder rights for firm i in year t, and CGIDit is the corporate governance index sub-index score on board governance disclosure for firm i in year t; and εi is the error term.

CONCLUSION
Our research is based on a different approach than those used in previous studies relying on share price (Berthelot et al., 2010;Bozec & Dia, 2015;Swain & Kar, 2021) to determine how shareholders take firms' governance practices into account. Instead, our study examines shareholders' votes "in favour" of director nominees at annual general meetings. Our results tend to show that shareholders' votes express their approval (or disapproval) of how directors fulfil their role, particularly as concerns monitoring CEO compensation growth and the implementation of governance practices recognised in the business community. Our findings thus confirm the reality of shareholder democracy even though the percentages of votes cast "in favour" of nominees are very high.
This study has some limitations, one of which is that the sample includes only large Canadian firms listed on the Toronto Stock Exchange.
The study's conclusions should therefore be interpreted with this specific context in mind. In addition, the governance practices analysed are limited to those included in The Globe and Mail's corporate governance index, which although important in the Canadian context may prove less so in other legal settings. It should also be noted that only one candidate is nominated for each director's position in Canada, which could provide an explanation for over 97% of the votes being cast "in favour" of the nominee. Finally, the study examines shareholders registered on companies' shareholder registers on the reference date for sending proxy circulars. Since firms do not disclose any information that could identify these shareholders, no distinction can be made between the votes cast by small and larger shareholders.
Our study points out various future avenues of research. For example, it could be interesting to examine the percentages of shareholders' participation in director elections. In Canada, the percentages of "in favour" votes are calculated only on the total votes cast rather than on the total possible votes tied to outstanding shares. Shareholders' non-participation in director elections could be another way for them to express their dissatisfaction with directors' actions. Such an analysis would also make it possible to better determine how Canadian shareholder democracy works. Another potential avenue of research could be to examine whether shareholders' "in favour" votes are influenced by tactics to manage accounting results (Franzoi,  , and the inclusion of inopportune photos in corporate publications (e.g., annual report) where the board has a role to play. This research would make it possible to more effectively assess the impact of these tactics on perceptions of shareholder performance.