Governance structures, unions, and theirrelation to CEO pay∗

Muhammad Umar Boodoo†

January 4, 2013

Abstract

Corporate scandals and the advocacy of austerity measures have led toincreased board oversight and increased stakeholder outcry over corporateexecutive compensation packages. This paper, using a sample of firms listedon the S&P/TSX composite index, finds while Boards of Canadian com-panies have become more ethical and effective, the extra scrutiny has beenassociated with an increase in the magnitude of the non-equity componentsof CEO compensation. CEOs have been able to extract higher salaries andbonuses even as their companies steer through the financial and economiccrisis. Using a fixed-effects-specification, this paper further finds that com-panies that tend to be more heavily unionized pay their CEOs less in termsof the equity components of the compensation package. Overall, the pa-per suggests that CEOs have been recession-proof in that despite the recentstock market difficulties, they have been able to hold more or less similarlevels of compensation. Governance structures and unions have not exertedenough pressure to curb the trend.

∗I would like to thank Matt Fullbrook, Antonio Spizzirri and David Comrie at the ClarksonCentre for Business Ethics and Board Effectiveness (CCBE) for sharing their CEO compensationand corporate governance data. I would also like to thank Saudia Ali and Una Tsang from theOntario Ministry of Labour for their support. For their helpful comments, I thank all seminar par-ticipants at the Centre for Industrial Relations and Human Resources at the University of Toronto.†Muhammad Umar Boodoo is at the Centre for Industrial Relations and Human Re-

sources at the University of Toronto. Please address all correspondence to muham-mad.boodoo@mail.utoronto.ca. This paper is still in its initial stages of development and allhelpful comments will be duly appreciated.

1

1 Introduction

The Principal-Agent Model starts from the premise that principals and agents have

varying interests, which implies that the principal has to incent the agent by align-

ing the interests of both parties in order to reach a higher optimal value of the rela-

tionship for the principal. Modern corporations are mostly owned by small share-

holders and the task of hiring and incentivizing managers and other executives is

delegated to an elected board of Directors, acting on behalf of the shareholders.

Such delegation and its impacts on firm performance and executive compensation

packages has been studied by scholars from various fields of research and practice

and while some (e.g., Rosen (1990); Hubbard (2005); Gabaix and Landier (2008)

point to the fact that the labor market is efficient and rewards executives for their

differential talent, others (e.g., Jensen (1993); Hall and Murphy (2003); Bebchuk

and Fried (2006) argue that such delegation mechanism to the board leads to ex-

ecutive compensation actually differing from their true labor market values.

If labor markets are inefficient in terms of allocating and rewarding managers,

in particular CEOs, this then begs the following questions: How crucial is board

effectiveness in terms of setting up compensation packages? How do other stake-

holders such as the media, the public and employees react to and influence board

decisions? Do labor unions, which stand for pay fairness and more equality at

work, exert significant political pressure on the compensation committee of the

board of Directors? This paper attempts to answer these questions by analyzing

governance mechanisms, union pressure and their relationship with the compen-

2

sation of CEOs of firms listed on the S&P/TSX composite index from 2005 to

2010. The post 2002 period and the ensuing enactment of the Sarbanes-Oxley

Act saw major changes being imposed on firms’ corporate governance structure.

While Canadian firms were not immediately affected, the new rules had an impact

on their governance systems since the various provincial securities commissions

followed suit with changes in governance requirements. Many of the big Canadian

corporations are cross-listed in the United States and they had to alter their gover-

nance structures as well. The ripple effect was felt across the board even though

the changes affected firms differentially. However, even though the governance of

TSX-listed firms improved over the period 2005-2010, the paper finds that it did

little to curb the rising trend of CEO compensation. In fact, this paper finds that

CEOs had to be compensated at higher levels because of the extra pressure and

scrutiny that they had to face - a finding akin to Hermalin (2005).

An additional finding of this paper is that CEO compensation seems to have

withstood the recession which started in 2007. Canadian firms were mildly hit by

the recession which struck their main US market but, in line with the report by

Mackenzie (2011), this paper also adds weight to the popularized belief that CEO

pay has essentially been recession-proof. Such reports attract the attention of other

company stakeholders and rile up calls to bring changes to corporate practices in

the interest of fairness. The media and employees, in particular unionized employ-

ees, tend to focus their attention to excess annual pay (see Core et al. (2008)and try

to pressure corporations in altering their executive pay strategy given that lower-

level employees are being asked to foot the bill of the economic downturn. This

3

paper finds that unions tend to be predominant in larger but not necessarily higher-

performing firms. Even though unions negatively affect the equity component of

CEO pay, they seem to play no part in explaining variations in the levels of CEO

compensation.

This paper belongs to the stream of studies that have looked into the non-

economic determinants of CEO compensation (e.g. Yermack (1996); Core et al.

(1999); Bertrand and Mullainaithan (2001); DiNardo et al. (2000); Singh and

Agarwal (2002). Jensen and Murphy (1990), in their seminal paper on the de-

terminants of CEO compensation, posited that certain actors can exert political

constraints that can indirectly influence CEO pay. By and large, most studies have

focused on board structures and very few papers have looked into the impact of

indirect political constraints exercised by the media as well as employees, who

are major company stakeholders. One can surmise that the primary reason for

this paucity of studies is that measuring employee interest in CEO compensation

is empirically difficult and big corporations in the United States have very few

strong unions which can exert pressure on compensation committees. This paper,

turning to Canada (a country with a relatively high union density vis-a-vis the

US), considers the governance structures of corporations listed on the S&P/TSX

composite index and also includes this employee dimension by comparing union-

ized and non-unionized firms and by considering the extent of unionization within

a company.

The rest of the paper is organized as follows. Section II looks at the extant lit-

erature examining the components and determinants of CEO compensation. Sec-

4

tion III describes data, the model and hypotheses to evaluate the link between

governance, unionization and CEO compensation, while Section IV discusses the

results of the model. Section V gives an overview of the limitations and further

development of the study and Section VI concludes.

2 Review of the literature

2.1 Economic determinants and rewards for human capital

Economists who start with the premise that the labor market is efficient and re-

wards talent based on productivity and human capital have examined and devel-

oped tractable general equilibrium models that posit, inter alia, that small infinites-

imal differences in human capital and talents of executives (mostly CEOs) within

the already small labor market for CEOs explain why CEO pay is so high and why

CEO pay varies tremendously between corporations (e.g. Rosen (1990); Hubbard

(2005); Gabaix and Landier (2008). This line of the literature looks at the impact

of economic determinants such as firm size, profitability and industry effects and

also brings in detailed human capital determinants of CEO compensation. The

CEOs of the big corporations in the US and Canada are mostly males, relatively

of the same age and hold similar academic and professional credentials. However,

some do tend to be slightly younger than others and do tend to have accomplished

different goals in their careers leading up to their appointment as CEO.1 Such

1Gordon Nixon was one of the youngest CEOs when he was appointed to the job at the RoyalBank of Canada (RBC). However, prior to his ascension to the position he currently holds, hehad been the head of RBC Securities and had headed major international projects. This compares

5

small but consequential differences is what lead to the big variation in CEO com-

pensation across firms, according to these studies. This is close to the predictions

of tournament theory where small differences determine who wins the tourna-

ment and who the losers are. The ensuing payoff for the winner usually tends to

be much higher than his/her incremental productivity relative to the losers.

While theoretically sound, most empirical studies have refrained from using

this approach due to differences in terms of the measures used to quantify hu-

man capital. Infinitesimal differences are difficult to capture and such differences

might not be symmetrically rewarded by firms within or across industries. Gabaix

and Landier (2008) calibrate a model where they exploit how small differences in

CEO talent lead to big differences in company profitability. The authors subtly

argue that the marginal productivity of certain CEOs make the latter deserving of

the high pay that they end up receiving.

2.2 Governance and CEO Compensation

Analysis of the compensation committee, the board of directors and general gover-

nance structures have given rise to a number of papers, mostly in Finance journals,

establishing the link between several governance attributes and CEO compensa-

tion. Governance attributes include, but are not limited to, independence of board

members, relation of such members with company affiliate, committee indepen-

dence, CEO/Chair split and Board/Director Evaluation. The relationship between

to Jeffrey Inmelt who succeeded Jack Welch at General Electric at a higher age but with lower(observed)credentials.

6

such attributes and CEO compensation is mixed in the literature. Jensen (1993)

argues that compensation decisions should be delegated to outside directors be-

cause they are better able to make rational and unbiased judgments about these

situations. However, the same paper also notes that outside and so-called inde-

pendent directors tend to be nominated by the CEO and as such may have a duty

of obligation towards the latter. In such cases, the unbiasedness may be called into

question. Indeed, Main et al. (1995) suggests that CEOs “manage” their compen-

sation committees to extract higher pay for a given level of performance. In the

same light, Hermalin (2005) argues that the surge in CEO pay is due to tighter cor-

porate governance: to compensate for the extra scrutiny from the board members,

CEOs get compensated with higher pay. Newman and Mozes (1999) do not find

evidence that CEO compensation is greater in companies that have “insiders” on

the compensation committee. The authors, however, do find evidence that the re-

lationship between CEO pay and performance is more favorable toward the CEO,

at the expense of shareholders, when there are “insiders” on the compensation

committee.

Most papers, however, report a negative correlation between the strength and

indepdendence of governance structures and CEO compensation. Core et al.

(1999), for example, use governance and compensation data for publicly-traded

US companies and find that CEOs earn greater compensation when governance

structures are weak, which exacerbates the agency problem. CEOs tend to be com-

pensated at a higher level if they also chair the board of directors, if the board size

is bigger and if the board is composed of a number of outside directors who have

7

been appointed by the CEO. Given such governance problems and corporate scan-

dals, the United States Securities and Exchange Commission embarked upon a

series of requirement alterations much in line with the provisions of the Sarbanes-

Oxley Act. Using these regulatory changes as a quasi-experiment, Chhaochharia

and Grinstein (2009) use a difference-in-difference approach and report that firms

that were most affected by Board provisions emanating from the Sarbanes-Oxley

Act drastically reduced their levels of CEO Compensation, i.e. those companies,

which had to improve their governance structures to comply with the new rules,

significantly reduced the level of pay of their CEOs. Even though some scholars

argue that CEOs may be compensated at a higher rate because of stricter gover-

nance mechanisms, most papers in this area of the literature have reported that

stronger governance is associated with lower “excessive” pay.

2.3 Employees/Unions acting as implicit regulators of CEO pay

Employees represent a powerful stakeholder group who may heavily influence the

HR department as well as the compensation committee on the board of directors.

Usually, if employees are represented by some form of a union, they tend to act as

a politically powerful counterbalance to managerial power. In their seminal paper,

Jensen and Murphy (1990) set out how unions may impact workplace-level poli-

cies and, by extension, apply considerable pressure on the corporate headquarters’

executives and directors. Unions position themselves as fairness activists and if

they perceive that executives are being overpaid over the heads of workers/union

members, they can demonstrate their show of strength mainly through threats to

8

disrupt the activities at the workplace. Such union pressure can be formally ex-

ercised during collective bargaining negotiations through public awareness cam-

paigns via blogs, social networking websites, political maneuvering and public

policy research.

At the negotiating table, unions may also be able to anchor higher (make fewer

concessions and bargain more strongly) if they have clear evidence that the earn-

ings of executives and the CEO, in particular, have been going up during the pe-

riod covering the previous collective bargaining agreement (Jensen and Murphy

(1990)).2 DiNardo et al. (2000) uses a sample of publicly-traded US companies

between 1971 and 1982 and initially report that union strength, measured in terms

of union density at the firm level, is negatively correlated with CEO compensa-

tion, i.e. the stronger the union, the lower is the level of CEO compensation.

However, after controlling for industry and firm fixed effects, their results become

insignificant and even suggest a positive relationship between union strength and

CEO compensation. The latter is more akin to the hypothesis in Hermalin (2005)

where one would think that CEOs have to be compensated for the extra pressure

of having to deal with stronger unions. Singh and Agarwal (2002), in a contextu-

ally different paper, looks at 86 Canadian mining and manufacturing firms listed

on the TSX in 1996. After controlling for firm size and performance, the authors

find that union presence is positively correlated with CEO base salary, but uncor-

related with the other components of CEO pay. These findings give impetus to

2The AFL-CIO website, for example, has a database dedicated to CEO compensation. Work-ers and unions can verify the compensation of their CEOs. See http://www.aflcio.org/Corporate-Watch/CEO-Pay-and-the-99

9

the “ratchet” effect whereby unionized firms, which are generally associated with

higher wages than non-unionized ones, drive up the labor costs starting from the

workers and going all the way through to the executives. Gomez and Tzioumis

(2006) extend the previous literature on unionization and CEO compensation by

using a panel of publicly-listed US companies between 1992 and 2001. The au-

thors match the compensation data with the data on workplace unionization from

the Bureau of National Affairs (BNA) and conclude that union presence is signif-

icantly negatively correlated with stock option values and hence total CEO com-

pensation. Although there is a positive relationship between CEO base salary and

union presence, this effect is small and is, in fact, wiped out, after accounting for

stock options.

Board members and employees are not the only stakeholders of a publicly-

traded corporation. The media is a powerful stakeholder that can attract negative

public attention and foster negative reputation toward a company that is engaging

in “excessive” executive compensation while employees and shareholders are re-

spectively rebuked with significantly low wages and low dividends/returns. Core

et al. (2008) look at the attention the press devotes to excessive CEO compensa-

tion and conclude that even though the press engages in sensationalism by picking

only on CEOs with sizable stock options, it does not force directors to lower CEO

compensation or indeed increase CEO turnover. The effect of the press on CEO

compensation appear to be, therefore, mitigated.

This paper attempts to combine governance and unionization in order to cap-

ture the effects of both mechanisms on CEO compensation. While not establishing

10

a causal relationship, this paper looks at firms in the post financial scandal period

and their CEO compensation practices during a recessionary period to track down

how differences in governance practices and unionization were associated with

differences in CEO compensation within and across firms.

3 Data, Variables and Methodology

3.1 Data

While boards and compensation committees can reward their CEOs in various

forms, most compensation packages consist of a base salary, an annual bonus,

stock options, restricted stock units, long-term incentive plans, other cash pay-

ments (severance etc.) and defined pension benefits or contributions. Data on

governance and compensation were generously provided to me by the Clarkson

Centre for Business Ethics and Board Effectiveness (CCBE). They collected data

on CEO compensation for the firms listed on the S&P/TSX Composite in 2010

and have been tracking the compensation of those CEOs as far back as 2005. The

advantage of tracking compensation this way is that my model and estimates will

not be biased by firms entering and leaving the sample after a short period of

time. Firms in my sample have existed for a rather lengthy period of time and

are credible on the stock exchanges. The CCBE compensation data consists of

salary, bonuses, options, restricted stock units, long-term incentive plans, and to-

tal compensation. Each component is evaluated at the beginning of the year when

compensation contracts are set out and published in a mandatory corporate proxy

11

circular. Typically, for a given company/CEO, the salary is fixed, the bonus is

tied to accounting results and the CCBE follows most academic researchers and

values stock options and restricted stock units using the Black-Scholes model.

The CCBE has also been collecting and using their governance data to rate

companies across Canada. Their estimates are very meticulously constructed and

are in fact subject to further scrutiny from Corporate Directors and the national

newspaper The Globe and Mail, which publishes the overall governance scores

for each company in its yearly publication called Board Games. The governance

score is a medley of governance attributes properly weighted to construct an index

that best proxies the prevalent governance structure at each company during those

6 years of data. Essentially, each company in a given year is allotted a score of

100 and for each governance “miss”, marks are deducted off this top score of 100.

The union data was much more difficult to assemble. Companies that are listed

on the S&P/TSX Composite are large companies that have multiple small work-

places where unions are certified. Matching workplaces to big corporations is

very tedious and requires manual checks because name strings can be misleading.

Further, collective bargaining is a provincial jurisdiction and the unionization data

was put together from the Ontario Ministry of Labour. Ontario is Canadas biggest

province and this is perhaps why Canada is best suited for studies that match

provincial-level workplace data to higher level corporate data. Most companies

that are listed on the S&P/TSX Composite have their headquarters in Ontario and

as such they would have a significant number of subsidiaries in Ontario. While

restricting unionization to Ontario only is not the most appropriate methodology,

12

it is a defensible position in that corporations would be most heavily scrutinized

by their nearest monitors.

The financial data(economic determinants of CEO pay) were out together by

carefully matching company ticker and names with the data from Compustat. The

S&P/TSX Composite had 204 firms listed in 2010 and I initially follow all of

them in their compensation packages back to 2005. However, with certain income

funds and companies not being evaluated by the CCBE in terms of governance and

with missing values emanating from Compustat, my final sample consists of 732

observations in the fixed-effects specification. All nominal data are adjusted for

inflation using 2009 as the base year.

3.2 Variables

The compensation variables that will be used in this study have been covered in the

previous section and consist of salary, bonus, options, restricted stock units, long-

term incentive plans, and total compensation. The economic determinants of CEO

compensation have almost always been size and some performance measures. I

follow previous the literature and look to proxy firm size with different measures

including sales, total assets and market capitalization. Such data is obtained from

Compustat at the firm-year unit of analysis. Performance measures include return

on assets (such returns are before interest and taxes) and return on equity. These

measures are also calculated from Compustat. The only human capital variable I

am looking at in this study is tenure, which I measure in months held on the job

at the firm-year level of analysis.

13

The governance attributes that will be used in this paper are: CEO/Chair

split, audit and compensation committee independence, and Director Evaluation.3

CEO/Chair split is an attempt to measure whether the Board can operate indepen-

dently of the main actors/management of the company. A large deduction (-10)

is made from the overall score of 100 if the CEO and Chair of the company are

the same person. The CCBE deducts 7 points if there is a split between the CEO

and the Chair but they are still related (usually because of some Board interlocks).

A smaller deduction (-5) is made if the company has appointed an independent

lead director to chair Board meetings. The best scenario in terms of governance

structure is where a company has a fully independent lead director and a chair and

a CEO who are not the same individual. In such a case, no point is deducted from

the overall score of 100.

The audit and compensation committee independence work in the same way.

If a Board member sitting on either of these committees is related (through inter-

locks or through kinship) to a senior executive of the company, then the CCBE

deducts 10 points off the overall score for this variable. These two committees

are important because full-independence of a company’s audit and compensation

committees is necessary to ensure that executive compensation and company ac-

counting are handled without conflict of interest between management and share-

holders.3I avoid using the Independence of Directors as a measure because of the fact that many of

those independent directors may be handpicked by the CEO. However, results when independenceis used as a governance attribute yield no different economic results to the ones reported in thispaper. I also avoid using the CCBE overall score because it is not very intuitive to interpret giventhe number of different attributes that make up the overall score.

14

Evaluations are a key component of Board and Director performance. Evalu-

ation processes generate feedback mechanisms that look into Board and Director

performances and also create a benchmark against which Directors’ performances

(attendance at meetings, recommendations etc.) can be ascertained. In order to

receive a perfect score in this category, a company in a given year must implement

regular and formal evaluation processes for both the Board as a whole, and each

of its individual Directors. The CCBE deducts 10 points if there is no full-Board

evaluation, 5 points if there is no Director evaluation, and 0 otherwise.

In terms of unionization at the firm-year level, I utilize the Ontario Ministry

of Labour data and if any workplace that is a subsidiary of a firm has at least one

collective agreement with a union, then that firm is classified as being unionized in

the year that the first collective agreement was effective. Years preceding the first

year of the earliest collective agreement are coded as non-unionized firm-year

observations, whereas years following the first collective agreement are coded

as unionized firm-year dummy variables. This assumption does not impose a

major constraint on the robustness of the results given that it is very difficult to

decertify unions after they are formed to represent members at a workplace. More

importantly though, I construct an index of union density at the company-year

level. This index is calculated by dividing the number of unionized employees by

the total number of employees that are under contract from a particular company

in a particular year.

Table 1 shows the means of the aforementioned variables by year. Panel A

shows how the mean and median compensation components varied by year. In-

15

deed, the observation by Mackenzie (2011) seems to hold in that CEO compensa-

tion seems to have been recession-proof. The recession hit during the year 2007

and CEO compensation showed no significant decrease in real 2009 dollars. In

fact, the average CEO compensation even rose during the recessionary period. The

trends are almost similar whether we measure central tendency using the mean or

the median. All components of CEO pay (salary, bonus, options, restricted stock

units, long-term incentive plans) seem to have withheld the recession.4

Average (mean) total compensation rose by approximately 20% from 2005

to 2010 and even though there was a fall post 2008, the trend picked up again in

2010. This rise was accounted for by increases in all the components of CEO com-

pensation during the same time period from 2005 to 2010 (salary - 13%; bonus -

24%; options - 28%; restricted stock units - 77%; long-term incentive plans fell by

78% although they never represented a large share of total CEO compensation.)

The median measures reveal lower absolute values of CEO compensation but the

trends are almost similar. Average (median) total compensation rose by as much

as 52% from 2005 to 2010 and the most alarming was that the average value of

options rose by about 500%. This, of course, falls in line with the once-very-

popular trend of rewarding CEOs with options rather than fixed salaries in order

to further align their interests with those of the shareholders. In a nutshell, CEO

compensation not only withstood the recession, but it also seems to have outpaced

the recession.4There is an additional component of CEO compensation termed “other”. It is relatively in-

significant (contains mostly zeros) and has been left out of this analysis.

16

This observation and the fact that returns to assets and equity (Panel B) fell

post 2007 further leads support to the hypothesis put forth by Bebchuk and Fried

(2006) in that pay does not necessarily vary with performance.5 On average, CEOs

seem to be highly paid despite the across-the-board poor performance. Of course,

this is a descriptive-level analysis that does not take into account industry effects

and firm fixed effects and other confounding variables but they do give an idea

about where the importance of good corporate governance and monitoring comes

in.

Indeed, Panel C shows that new regulations in the US and Canada have led

to companies adopting better governance practices: fewer corporations had the

CEO also chair the board meetings as the years passed by, and companies seemed

to favor either appointing lead Directors or actually selecting independent Direc-

tors from the market. In terms of committee independence, even though the audit

committees seem to have become relatively less fully independent, the compensa-

tion committee, on the other hand, did become more independent with time. The

pressure by regulators (direct or implicit) seems to have forced boards to ring in

changes to the compensation committee memberships. Union density is very low

amongst these TSX-listed companies. One of the reasons is that these companies

are usually very large and employ many professionals who are not keen to union-

izing. Another reason is that the data looks only at Ontario-level unionization and,

5Note that I am measuring the economic and non-economic determinants of pay from 2004to 2009 given that compensation contracts are usually set out at the beginning of every year. Soit seems plausible that the compensation contract set out for 2005 would be based on the firmperformance in 2004 and in earlier years.

17

therefore, underestimates the extent of unionization at the firm level. Nonetheless,

given that most of the companies in the sample have their headquarters and main

operations in Ontario, this does not represent a serious limitation.

Another interesting descriptive statistic to look at would be the share of total

CEO pay in terms of salary, bonus, stock options, restricted stock units, and long-

term incentive plans, i.e. how did the composition change with time? Gomez

and Tzioumis (2006) report that the share of options in total CEO compensation

rose during the late 1990s and early 2000s and this is what led to the massive rise

in total CEO compensation. However, looking at Canadian companies post the

regulatory changes, this paper finds that the cash non-equity components of CEO

pay (salary + bonus) still account for a high percentage of total CEO compen-

sation. Figure 1 shows that on average, salary and bonus together accounted for

approximately 40 % of total CEO compensation. Options and Restricted Stock

Units, however, account for over 50% of total CEO compensation. This implies

that there exist huge equity-based payouts to the CEO in order to align his/her

interests more closely to those of the shareholders.

3.3 Methodology

This paper attempts to explain variations in levels and rates of change in CEO pay

emanating differences in governance structures and unionization. The following

fixed-effects specification is run over an unbalanced panel of 202 firms in the years

18

2005 to 20106:

lnCit = α+βlnF irmSizet−1+γlnAssetReturnt−1+3∑

j=1

δCEO − ChairSplitij,t−1

+ ζAuditCommIndependencet−1 + ηCompCommIndependencet−1

+ φDirectorEvalt−1 + µUnionDensityt−1 + ε (1)

where Cit can represent either one of base salary, annual bonus, stock options,

restricted stock units, long-term incentive plans or total compensation of the CEO

of firm i at time t. As standard in the literature, all continuous variables have

been log-transformed. This serves at least two purposes: 1) it smoothes the com-

pensation measures given that, from Table 1, it was apparent that the dispersion

between observations was rather high, and 2) it makes the interpretation easier

in terms of elasticities. Note that the explanatory variables have been lagged by

1 period given that the compensation variables is set at the beginning of time t.

It is more likely that compensation contracts set at time t would look to the per-

formance at time t − 1 as an indicator of how much leverage the compensation

committee has versus the CEO. The logarithmic transform of asset return is ac-

tually calculated by taking the log of 1 plus asset return.7 This is because some

returns (net incomes) are actually negative and would, therefore, lead to the log

6Note that because I am running a panel fixed-effects estimation, I can dismiss industry fixedeffects in the regression and other fixed firm characteristics. I am also leaving out tenure becausethe tenure data is limited to a few companies. I have run regression on the sample of data thatincludes tenure and the results are qualitatively similar.

7I avoid using stock return as a measure because it is highly correlated with asset return andmost papers in this literature (see Core et al. (2008)) prefer asset return as a measure of profitability.

19

transformation to be undefined or missing.

The size variable is a rather open-ended one in the literature. Different proxies

have been used to measure firm size but I will pick market capitalization over

sales and total assets since my sample (and the S&P/TSX Composite in general)

is made up of firms that fall into different 2-digit NAICS categories but a sizeable

majority of such firms lie in mining, manufacturing and finance and insurance.

Sales for mining firms are not the best proxy for size. An accounting measure

such as total assets and/or market capitalization is more robust and captures size

more effectively given the sample of firms in question.8

Based on the previous literature, I expect that the sign of the

β coefficient for firm size will be positive and around 0.35, based on the ex-

tensive review of the literature in Murphy (1999). I do not expect the

γ coefficient (asset return) to be significantly correlated with the salary com-

ponent, but I do expect it to be positively related to the bonus, stock options and

restricted stock units components and, hence, also to total compensation. In terms

of the governance parameters, I expect the signs of each δ is progressively nega-

tive, implying that if the more split the CEO and the Chair, the less will be total

compensation and its components. Similarly, if the audit and compensation com-

mittees are independent, then I expect CEO compensation to be lower. The union

density measure is intriguing. Gomez and Tzioumis (2006)find that union pres-

ence leads to lower total compensation, in particular to lower bonuses and stock

8I have calculated regressions where size is proxied through sales and total assets and theresults are qualitatively similar.

20

options for CEOs even though union presence is weakly associated with higher

salaries for CEOs. Singh and Agarwal (2002), on the other hand, find that union

density is positively correlated with CEO salary but uncorrelated with the other

components of CEO pay. If unions are active and are politicizing fairness slogans,

one could expect that they would negatively affect the equity components of CEO

pay (which are the major share of total CEO compensation as depicted in Figure

1).

4 Results and Discussion

As indicated in the first row of Table 2, return on assets, a measure of profitability,

yields no significant result when the dependent variable is total compensation,

total equity compensation(options+restricted stock+long-term incentive plans) or

total non-equity compensation(salary+ bonus). CEO pay seems to be unrelated

to firm profitability. It is important to note here that the period 2005 to 2010

was a period where many of the firms on the S&P/TSX Composite Index suffered

negative net incomes. Despite that, CEO pay stayed the same or even increased as

shown in Table 1. As such, this could explain why the results are insignificant for

return on assets. Irrespective of the level of profitability, CEO pay was unaffected.

The elasticity of total CEO compensation with respect to firm size is 0.286,

slightly lower than the typical estimate of 0.35 documented in the review of Mur-

phy (1999). This is driven mainly by the large effect of size on equity compensa-

tion (elasticity of 0.488 - though not significant) and on non-equity compensation

21

(elasticity of 0.256).

The governance results indicate that CEOs who also chair the Board are com-

pensated at a lower rate than CEOs who have to answer to an independent Chair.

When the Chair is independently chosen, for example, the CEO makes 45% more

than a CEO who also chairs his/her Board. Most of this ‘gain’ is driven by non-

equity compensation, i.e. CEOs, who have to answer to a Lead Director or to a

completely independent Chair, are paid higher salaries and bonuses than CEOs

who chair their Boards. This results falls in line with the predictions and results

from Hermalin (2005) who argued that CEOs are extracting higher cash com-

ponents of compensation because of the high scrutiny they now face. Another

way of phrasing the same phenomenon is to posit that CEOs prefer to negotiate

higher salaries and bonuses (which are a safer bet than stock options and restricted

stock) when they are under pressure to answer to independent Chairs. In the same

vein, when there are independent Director evaluations, CEOs tend to make more

than their counterparts who work in companies whose Boards have not yet imple-

mented fully independent Director evaluations. Independent compensation com-

mittees do not have that big of an effect in terms of CEO compensation. One

would surmise that the more independent the audit and compensation commit-

tees, the less will be the level of CEO compensation However, this is not the case

and the only significant result is that independent compensation committees lead

to CEOs being paid approximately 17% less than their counterparts with regards

to their non-equity components of pay. This is perhaps because Boards are wary

of compensation contracts and whether independent or not, compensation com-

22

mittees set compensation structures that closely follows the market trend.

Unionization leads to lower total compensation but the result is not signifi-

cant. What is especially significant is the fact that higher union density leads to

lower equity-based compensation of CEOs. Companies that are heavily unionized

or companies that become more heavily unionized pay their CEOs less in terms

of equity components than their counterparts. Table 4 indicates that the major

reason for this negative relationship between union density and equity compensa-

tion arises because of low restricted stock units being granted to CEOs of more

heavily unionized companies. Restricted stock, being stock that is transferable

under certain conditions, are lower for heavily unionized companies. CEOs per-

haps cannot meet all the conditions that are laid out to gain such stock or they

simply substitute them for ‘safer’ alternatives such as bonuses and options when

they are negotiating their contracts. More research is needed to decipher the rea-

sons behind the significantly high negative relationship between union density and

restricted stock units granted.

One could go about comparing the characteristics of heavily unionized com-

panies with those that are less heavily unionized. If heavily unionized companies

are less profitable, then this might be the reason behind the lower levels of CEO

compensation in those companies. However, simple regression shows that this

is not the case. Figure 2 shows that there is an insignificant positive relation-

ship between union density and return on assets (profitability). Heavily unionized

companies are not driving down profits, and hence, not driving down either eq-

uity or non-equity components of compensation. What is certainly intriguing is

23

that the larger companies, as measured by market capitalization,9 are more heav-

ily unionized than the smaller companies. Unions seem to be better willing and

able to organize in larger companies than in smaller ones. As such, it is not clear

why the negative relationship between restricted stock units and union density is

so high and significant. Further research might reveal the reasons behind the low

values of restricted stock options granted to CEOs of heavily unionized compa-

nies. Qualitative studies might be a first avenue to explore the reasons behind this

finding.

5 Limitations of study

The fixed-effects estimates in this paper do overcome some selection bias and the

effect of other unobservable characteristics that may have biased the results of

previous papers in this field of research. However, this paper suffers from limita-

tions and in this section, I will run over some of those limitations and how I plan

to overcome them in a revised version of this paper. The first major problem re-

volves around the fact that too many observations are lost when matching across 3

different sets of data. The compensation variables have very few missing data but

the governance ones do have some missing data, especially towards the beginning

of the study period. Not all firms that were listed on the S&P/TSX Composite in

2010 were listed back in 2004 and, hence, there are missing governance values

for such firms. Mean imputation and/or multiple imputation techniques can be

9This also hold when total assets and/or revenue are used to proxy size

24

used to fill such missing data so that most observations are kept in the analysis

to add precision and reduce sensitivity to bias. However, these methods may not

be suitable because they would assume that governance necessarily improved or

stayed at the mean for particular kinds of companies. The best way would be to

obtain compensation data on firms listed on the S&P/TSX Composite Index back

from 2004 to 2008. Instead of filling in the governance data with ad-hoc imputa-

tions, real compensation data can be collected for companies that were listed on

the S&P/TSX Composite Index back from 2004 to 2008 but which were de-listed

in 2010. This would involve heavy manual undertaking and would require care-

ful analysis of proxy circulars and company annual reports. I plan to follow the

CCBE’s method and add this data to the analysis.

With regards to unionization (potentially the major contribution of this paper),

the measure is simply not precise enough. Unions vary in strength and in the na-

ture and degree of activism at the workplace. Deeper analysis into the Ministry

of Labour data is needed to measure union entrance and union strength. Once the

data is cleaned, I could measure the impact of a new union being certified for the

first time on CEO compensation. A more complex index of union strength will

include, amongst other things, variables such as level of bargaining unit, interna-

tional or non-international nature of union, and number of bargaining members.

A further limitation is that this study does not relate a causal model of CEO

compensation. Unlike Chhaochharia and Grinstein (2009) who use a difference-

in-difference approach to capture the effect of regulatory changes on CEO com-

pensation, this paper’s findings are merely correlational in nature. It would per-

25

haps be a better idea to go back further in time to see how regulatory changes

brought about in Canada in the aftermath of Sarbanes-Oxley affected the CEO

compensation practices of firms operating within the Canadian jurisdiction. Again,

more data would be needed to cover the pre- and post-Sarbanes-Oxley Act.

6 Conclusion

This paper has attempted to look into the relationship between governance, union-

ization and the various components of CEO compensation in Canada. Control-

ling for size, performance, firm and year fixed-effects, this paper finds that CEO

compensation is only midly influenced by governance mechanisms. Higher CEO

compensation is associated with stronger governance reflecting the hypothesis that

CEOs tend to be compensated for the extra scrutiny that they now have to face. In

particular, if the CEO has to report to an independent Chair, his non-equity com-

pensation(salary plus bonus) are higher by a factor of 32% compared to a CEO

who also chairs his/her Board. Further the total compensation (equity plus non-

equity) of a CEO who has to report to an independent Chair is 45% higher than

a CEO who chairs his/her Board. Independent compensation committees palliate

the non-equity component of CEO compensation by a factor of 16.8%, i.e. if the

compensation committee is independent, a CEO earns 16.8% less in salaries and

bonuses. However, the effect on independent committees on total compensation

is insignificant, reflecting either that governance structures have improved enough

such that small variations in their attributes are not leading to major differences in

26

CEO pay or that governance structures do not matter with regards to CEO com-

pensation.

This paper has also looked at the association between unions and the various

components of CEO pay. Higher union density are associated with lower CEO

total compensation, in particular with lower equity-based compensation. This is

not due to the fact that heavily-unionized firms are less profitable or that heavily-

unionized firms are smaller. Indeed, heavily-unionized firms are usually large

and relatively profitable and the fact that their CEOs make less than the CEOs of

firms with low union density suggests that perhaps unions are having an effect on

compensation committees and on Boards of Directors. More research is needed

to understand how and why unions can affect CEO compensation either directly

or indirectly.

27

Table 1A: Summary statistics for components of CEO Compensation

Mean of Components of CEO Pay (All in 2009 constant Dollars )

2005 2006 2007 2008 2009 2010 Total

Salary 610889.9 614521.1 598273.6 602470.1 655042.4 688842.4 626461.9

(465169.6) (594938.4) (406361.5) (416724.6) (420832.6) (506102.7) (475070.2)

Bonus 601680.1 663121.8 765874.9 871522.9 706330.3 746616.6 723101.4

(1062480.9) (986322.3) (1178257.2) (1483258.6) (1001686.2) (1303068.6) (1181293.3)

Options 945667.7 1140489.2 1141573.1 1249764.2 1077914.5 1207779.4 1123802.1

(3670591.1) (2619821.8) (2029369.3) (2620572.5) (1776501.9) (2106378.3) (2588399.3)

Restricted Stock Units 418553.2 421513.2 633565.9 672389.9 730397.3 739332.4 593024.4

(1762910.5) (1087622.9) (1431174.5) (1463900.8) (2165470.8) (1641754.4) (1616378.9)

Long-term Incentive Plans 33471.2 77590.0 29735.4 15084.1 19278.3 7258.3 31526.0

(185619.4) (639898.6) (224087.5) (130741.2) (170194.0) (70213.1) (310008.3)

Total Compensation 3014734.0 3293839.5 3576537.5 3644496.9 3461070.9 3635228.1 3424949.0

(6105236.7) (5420617.9) (3971648.1) (4508033.3) (4091235.3) (3934958.7) (4800453.3)

Median of Components of CEO Pay (All in 2009 constant Dollars )

2005 2006 2007 2008 2009 2010 Total

Salary 534579.4 524289.6 513004.5 507671.3 563557.5 575404.0 534579.4

Bonus 308853.3 340788.3 421437.8 377162.5 346227.5 423230.9 367862.0

Options 59263.5 283687.3 339557.7 298282.2 278200 399491.7 267097.3

Restricted Stock Units 0 0 0 0 0 0 0

Long-term Incentive Plans 0 0 0 0 0 0 0

Total Compensation 1480363.9 1681916.4 2322220.5 2074378.8 2069638.8 2249187 1953454.5

Top Panel: Means reported; standard errors in parentheses

28

Table 1B: Summary Statistics for Financial and Economic Determinants of CEO pay

2004 2005 2006 2007 2008 2009 Total

Total Assets($ millions) 1709.7 1795.5 2007.5 1933.3 2008.8 2008.7 1903.7

(3540.5) (3343.6) (3630.8) (2952.9) (3055.5) (3037.8) (3282.4)

Sales($ millions) 2525.7 2722.3 3022.8 3176.6 3635.7 3297.4 3053.3

(5157.9) (5553.0) (6125.9) (6453.6) (7134.9) (6686.1) (6205.2)

Market Cap($ millions) 3070.4 3762.9 4510.7 5046.3 3434.5 4797.1 4077.4

(6017.5) (7524.0) (8888.5) (10631.9) (7770.9) (9813.9) (8528.6)

Return on assets(%) 2.341 3.196 4.702 3.623 3.038 2.017 3.190

(13.16) (12.79) (11.33) (11.59) (11.91) (9.852) (11.82)

Return on equity(%) 4.806 11.44 12.19 9.069 7.494 5.678 8.555

(56.14) (24.04) (19.74) (24.50) (22.34) (16.95) (29.84)

29

Table 1C: Summary Statistics for Governance Structures and Unionization

2004 2005 2006 2007 2008 2009 Total

No CEO-Chair Split 0.0977 0.0762 0.0773 0.0681 0.0615 0.0542 0.0735

(0.298) (0.266) (0.268) (0.253) (0.241) (0.227) (0.261)

CEO-Chair Split/Chair is Related 0.163 0.152 0.0918 0.0785 0.0923 0.0783 0.112

(0.370) (0.360) (0.289) (0.270) (0.290) (0.269) (0.315)

CEO-Chair Split/Lead Director Appointed 0.209 0.251 0.290 0.304 0.308 0.301 0.275

(0.408) (0.435) (0.455) (0.461) (0.463) (0.460) (0.447)

CEO-Chair Split and fully independent 0.530 0.520 0.541 0.550 0.538 0.566 0.540

(0.500) (0.501) (0.500) (0.499) (0.500) (0.497) (0.499)

Related Director on Audit Committee 0.260 0.238 0.203 0.220 0.215 0.247 0.231

(0.440) (0.427) (0.403) (0.415) (0.412) (0.433) (0.421)

Full Audit Committee Independence 0.740 0.762 0.797 0.780 0.785 0.753 0.769

(0.440) (0.427) (0.403) (0.415) (0.412) (0.433) (0.421)

Related Director on Compensation Committee 0.251 0.220 0.203 0.152 0.138 0.102 0.182

(0.435) (0.415) (0.403) (0.360) (0.346) (0.304) (0.386)

Full Compensation Committee Independence 0.749 0.780 0.797 0.848 0.862 0.898 0.818

(0.435) (0.415) (0.403) (0.360) (0.346) (0.304) (0.386)

No Individual Director Evaluation 0.186 0.565 0.425 0.414 0.421 0.337 0.393

(0.390) (0.497) (0.496) (0.494) (0.495) (0.474) (0.489)

Full individual Director Evaluation 0.814 0.435 0.575 0.586 0.579 0.663 0.607

(0.390) (0.497) (0.496) (0.494) (0.495) (0.474) (0.489)

Union Density 0.0130 0.00895 0.00751 0.00778 0.0237 0.00979 0.0117

(0.0475) (0.0326) (0.0296) (0.0298) (0.254) (0.0425) (0.111)

30

0

.1

.2

.3

.4

(Mea

n) P

ropo

rtio

n of

Toa

l Pay

2005 2006 2007 2008 2009 2010

Fig1: Proportion of Compensation Components

Salary Bonus OptionsRSU LTIP

31

Table 2: Total compensation, equity and non-equity compensation and their correlates

Total Compensation(log) Equity Compensation(log) Non-Equity Compensation(log)

Return on Assets 0.0968 1.106 0.0838

(0.102) (0.900) (0.0727)

Market Capitalization 0.286∗∗∗ 0.488 0.256∗∗∗

(0.0719) (0.635) (0.0512)

No CEO-Chair Split - - -

CEO-Chair Split/Chair is Related 0.202 1.991 0.155

(0.266) (2.345) (0.189)

CEO-Chair Split/Lead Director Appointed 0.514∗∗ 3.150 0.316∗∗

(0.222) (1.958) (0.158)

CEO-Chair Split and fully independent 0.449∗ 3.539∗ 0.323∗

(0.237) (2.089) (0.169)

Audit Committee Independence (dummy) −0.107 1.181 −0.0431

(0.134) (1.184) (0.0956)

Compensation Committee Independence (dummy) −0.144 −0.916 −0.168∗

(0.131) (1.154) (0.0932)

Individual Director Evaluation (dummy) 0.269∗∗∗ 1.375∗ 0.111∗

(0.0901) (0.796) (0.0642)

Union Density −1.673 −22.73∗ −1.415

(1.508) (13.31) (1.075)

Constant 12.21∗∗∗ 3.262 11.86∗∗∗

(0.597) (5.266) (0.425)

Observations 732 732 732

R2 0.334 0.029 0.098

Standard errors in parentheses∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01

32

Table 3: Non-Equity Compensation Components and their correlates

Salary Bonus

Return on Assets 0.0138 0.167

(0.0414) (0.643)

Market Capitalization 0.0528∗ 1.568∗∗∗

(0.0292) (0.453)

No CEO-Chair Split - -

CEO-Chair Split/Chair is Related 0.114 3.279∗

(0.108) (1.674)

CEO-Chair Split/Lead Director Appointed 0.149∗ 3.395∗∗

(0.0900) (1.398)

CEO-Chair Split and fully independent 0.250∗∗∗ 4.269∗∗∗

(0.0961) (1.491)

Audit Committee Independence (dummy) 0.0496 −0.297

(0.0544) (0.845)

Compensation Committee Independence (dummy) −0.0851 −1.452∗

(0.0531) (0.824)

Individual Director Evaluation (dummy) 0.140∗∗∗ 0.912

(0.0366) (0.568)

Union Density −0.254 3.900

(0.612) (9.505)

Constant 12.55∗∗∗ −3.301

(0.242) (3.759)

Observations 732 732

R2 0.041 0.085

Standard errors in parentheses∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01

33

Table 4: Equity compensation components and their correlates

Options Restricted Stock LTIP

Return on Assets 0.627 0.437 0.227

(1.005) (0.767) (0.378)

Market Capitalization 0.447 0.0332 −0.0965

(0.708) (0.541) (0.266)

No CEO-Chair Split - - -

CEO-Chair Split/Chair is Related 0.749 −0.0362 0.857

(2.617) (1.998) (0.983)

CEO-Chair Split/Lead Director Appointed 2.607 0.0905 −0.0469

(2.185) (1.669) (0.821)

CEO-Chair Split and fully independent 2.791 2.264 0.0132

(2.332) (1.780) (0.876)

Audit Committee Independence (dummy) 0.0184 0.660 0.0885

(1.321) (1.009) (0.496)

Compensation Committee Independence (dummy) −1.511 0.486 0.118

(1.288) (0.984) (0.484)

Individual Director Evaluation (dummy) 1.059 0.921 0.167

(0.888) (0.678) (0.334)

Union Density −4.072 −22.05∗ 3.509

(14.86) (11.35) (5.583)

Constant 3.660 2.105 0.821

(5.877) (4.487) (2.208)

Observations 732 732 732

R2 0.022 0.010 0.009

Standard errors in parentheses∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01

34

0 .1 .2 .3 .4Union Density

Market Capitalization Fitted values

Fig2: Union Density and its relation to Firm Size

0 .1 .2 .3 .4Union Density

Return on Assets Fitted values

Fig2: Union Density and its relation to Asset Return

35

References

Bebchuk, L. A. and Fried, J. M. (2006). Pay without performance: Overview of

the issues. Academy of Management Journal, 20(1):5–24.

Bertrand, M. and Mullainaithan, S. (2001). Are CEOs rewarded for luck? The

one without principals are. Quarterly Journal of Economics, 116:901–932.

Chhaochharia, V. and Grinstein, Y. (2009). CEO Compensation and Board Struc-

ture. Journal of Finance, 64(1):231–261.

Core, J. E., Guay, W., and Larcker, D. F. (2008). The power of the pen and

executive compensation. Journal of Financial Economics, 88:1–25.

Core, J. E., Holthausen, R. W., and Larcker, D. F. (1999). Corporate governance,

chief executive officer compensation, and firm performance. Journal of Finan-

cial Economics, 51:371–406.

DiNardo, J., Hallock, K. F., and Pischke, J. F. (2000). Unions and the labor market

for managers. CEPR Working Paper No. 2418.

Gabaix, X. and Landier, A. (2008). Why has CEO Pay increased so much? Quar-

terly Journal of Economics, 1:49–100.

Gomez, R. and Tzioumis, K. (2006). What do Unions do to CEO Compensation?

Centre for Economic Performance, LSE, CEP Discussion Papers No. 720.

Hall, B. J. and Murphy, K. J. (2003). The trouble with stock options. Journal of

Economic Perspectives, 17:49–70.

36

Hermalin, B. (2005). Trend in Corporate Governance. Journal of Finance,

60:2351–2384.

Hubbard, G. R. (2005). Pay without performance: A market equilibrium critique.

Journal of Corporate Law, pages 717–710.

Jensen, M. (1993). The modern industrial revolution, exit, and the failure of inter-

nal control systems. Journal of Finance, 48:831–880.

Jensen, M. C. and Murphy, K. J. (1990). Performance pay and top-management

incentives. Journal of Political Economy, 98(2):225–264.

Mackenzie, H. (2011). Recession-Proof: Canada’s 100 best-paid CEOs. Cana-

dian Centre for Policy Alternatives, Growing GAP Project.

Main, G., O’Reilly, C., and Crystal, G. (1995). The CEO, the Board of Direc-

tors, and Executive Compensation: Economic and Psychological Perspectives.

Industrial and Corporate Change, 11:605–628.

Murphy, K. J. (1999). Executive compensation. In Ashenfelter, O. C. and Card,

D., editors, Handbook of Labour Economics, volume 3B, pages 2486–2563,

New York. Elsevier.

Newman, H. A. and Mozes, H. A. (1999). Does the composition of the compensa-

tion committee influence ceo compensation practices? Financial Management,

28(3):41–53.

37

Rosen, S. (1990). Contracts and the market for executives. NBER Working Paper

No. W3542.

Singh, P. and Agarwal, N. C. (2002). Union Presence and Executive Compensa-

tion: An Exploratory Study. Journal of Labour Research, 23(4):631–646.

Yermack, D. (1996). Higher market valuation of companies with a small board of

directors. Journal of Financial Economics, 40:185–211.

38