CEO Duality and Firm Performance: Evidence from an Exogenous Shock to the Competitive Environment

Tina Yanga,*, Shan Zhaob

aSchool of Business, Villanova University, Villanova, PA, USA bGrenoble Ecole de Management, France

Abstract Regulators and governance activists are pressuring firms to abolish CEO duality (Chief Executive Officer also being the Chairman of the Board). However, the literature provides mixed evidence on the relation between CEO duality and firm performance. Using the exogenous shock of the 1989 Canada-United States Free Trade Agreement, we find that duality firms outperform non-duality firms by 3-4% when their competitive environments change. Further, the performance difference is larger for firms with higher information costs and better corporate governance. Our results underscore the benefits of CEO duality in saving information costs and making speedy decisions. JEL classification: G34; G38; K22 Keywords: CEO duality, firm performance, corporate governance, endogeneity, competitive environments

* We would like to thank Yung-Yu Ma and David Yermack for their generosity in sharing their data with us. We also thank Russell Chomiak, Paul Hanouna, Wan Hong, Steve Miller, Nancy Margolis, Shawn Mobbs, David Oesch, Sukesh Patro, Jesus Salas, and the seminar participants at the 2011 Asian Finance Association International Conference, the 2011 Financial Management meetings, the 2013 Conference of the Swiss Society for Financial Market Research, the 2013 Conference Twenty Years after Cadbury, Ten Years after Sarbanes-Oxley: Challenges of Corporate Governance, the 2013 International Finance and Banking Society Nottingham Conference, Lehigh University, and Villanova University. We gratefully acknowledge research support from the Center for Global Leadership at Villanova University. Corresponding author: Tina Yang, Finance Department, Villanova University, 800 Lancaster Avenue, Villanova, PA 19085, USA. Phone: (610) 519-5460. Fax: (610) 519-6881. E-mail: tianxia.yang@villanova.edu.

CEO Duality and Firm Performance: Evidence from an Exogenous Shock to the Competitive Environment

Abstract Regulators and governance activists are pressuring firms to abolish CEO duality (Chief Executive Officer also being the Chairman of the Board). However, the literature provides mixed evidence on the relation between CEO duality and firm performance. Using the exogenous shock of the 1989 Canada-United States Free Trade Agreement, we find that duality firms outperform non-duality firms by 3-4% when their competitive environments change. Further, the performance difference is larger for firms with higher information costs and better corporate governance. Our results underscore the benefits of CEO duality in saving information costs and making speedy decisions. JEL classification: G34; G38; K22 Keywords: CEO duality, firm performance, corporate governance, endogeneity, competitive environments

1

1. Introduction

In this paper, we study the effect of CEO duality on firm performance using an

exogenous shock to the competitive environment. (CEO duality refers to a board leadership

structure in which the Chief Executive Officer (CEO) is the Chairman of the Board (COB).) The

exogenous shock is the 1989 Canada-United States Free Trade Agreement (FTA), which

eliminated all tariffs and other trade barriers between the two countries. Three drivers motivate

our research question: the recent regulatory push to abolish CEO duality, the growing trend of

U.S. firms separating the CEO and COB titles, and the ambiguity in the literature about whether

CEO duality is beneficial or detrimental to firm performance. Specifically, firms that received

assistance under the 2008 Troubled Asset Relief Program (TARP) were required to separate the

CEO and COB titles. The U.S. Congress introduced several proposals in 2009 that called for the

titles to be split. The Dodd-Frank Act of 2010 required the Securities and Exchange Commission

(SEC) to issue rules mandating that listed firms disclose the reasoning behind their board

leadership structure.1 Under pressure, U.S. firms appear to have modified their decisions

regarding CEO duality. As Fig. 1 shows, from the 1970s until the early 1990s, over 80% of large

U.S. firms had combined titles. The figure fell to 54% in 2010.

The above trends are unsettling given the lack of clear theoretical predictions and mixed

empirical evidence on the performance impact of CEO duality. The main argument against CEO

duality (or dual leadership) is based on agency theory, which predicts that CEOs, as agents of

shareholders, do not always act in the best interests of shareholders. As the board of directors is

the apex of the decision control system of corporations, entrusting CEOs with the office of COB

exemplifies the ultimate conflict of interest. The main argument in favor of dual leadership

1 Prior to the Dodd-Frank Act, the SEC had already adopted the final rules in this regard on December 16, 2009, which became effective on February 28, 2010. Proxy Disclosure Enhancements are available at http://www.sec.gov/rules/final/ 2009/33-9089.pdf.

2

emphasizes the unparalleled firm-specific information of CEOs and firms ability to quickly

respond to changing environments due to unified leadership (Brickley, Coles, and Jarrell, 1997;

Larcker and Tayan, 2011). A large body of literature has developed over the years in an attempt

to solve empirically the debate over the efficacy of CEO duality. However, the evidence

produced so far is mixed due to endogeneity and heterogeneity challenges (Hermalin and

Weisbach, 2003; Adams, Hermalin, and Weisbach, 2010).

We add to the literature by analyzing the effect of CEO duality on firm performance in a

new framework that mitigates the endogeneity and heterogeneity issues. Specifically, by using an

exogenous shock to the competitive environment and relating post-shock firm performance to

pre-shock board leadership structure, our research design mitigates the endogeneity problem

caused by reverse causality, namely whether a certain board leadership structure improves firm

performance or whether better performing firms tend to adopt certain board leadership structures.

Heterogeneity problems arise because firms face different optimization problems and are driven

to heterogeneous solutions. For example, large boards can be beneficial to complex firms but

harmful to non-complex firms (Coles, Daniel, and Naveen, 2008). Conditioning the duality-

performance relation on firms responding to one common shock mitigates the heterogeneity

problem, because the shock disrupts the status quo and forces firms with different characteristics

to find new solutions to the same problem.2 Further, our research design mitigates unobserved

heterogeneity that potentially drives firms choices of board leadership structure by exploiting

performance differences across duality and non-duality firms under different tariff rates.

2 Our research design may not eradicate the heterogeneity problem, because firms of different characteristics likely incur different costs and benefits in solving the same problem and subsequently still find different optimization solutions. However, as the exogenous shock disrupts firms exiting equilibrium solutions to the choice problem of board leadership structure, it is useful to study the duality-performance relation in transition states (i.e., before firms find their new equilibrium solutions), particularly given that it is unclear ex-ante how the exogenous shock will impact the duality-performance relation for firms of different characteristics. See Coles, Daniel, and Naveen (2008) and Adams, Hermalin, and Weisbach (2010) for excellent discussion of the heterogeneity problem.

3

Our research design offers two additional benefits. First, although we study a conditional

effect, with the condition being increased competition and growing market opportunities, our

results have broad implications because more firms should operate under this condition as

globalization and technology advance. Second, our research design facilitates a direct test of the

main benefit of CEO duality. Dual leadership incurs lower information acquisition, transmission,

and processing costs than separate leadership, because CEOs accumulate unparalleled firm-

specific information through running the daily operation of the firm (Jensen and Meckling, 1995;

Brickley, Coles, and Jarrell, 1997). An exogenous shock that increases competition and broadens

market opportunities magnifies the information benefits of CEO duality for two reasons. First,

competition and new market opportunities increase the value of information, especially the value

of specific information, because information generates market power and specific information is

more costly to acquire and transfer and therefore generates larger and more sustained

information rents (Jensen and Meckling, 1995). Second, competition and new market

opportunities demand fast and frequent decision-making as under these market conditions

information becomes obsolete at a faster rate and the consequences of lost opportunities due to

delayed decisions become more severe (Christie, Joye, and Watts, 2003). Dual leadership allows

firms to make speedier decisions and react more quickly to new information than separate

leadership because the former eliminates an extra chain of command, namely the non-CEO COB

(Larcker and Tayan, 2011). These arguments, which we call the information advantage argument

of CEO duality, suggest that post trade liberalization, the value contribution of dual leadership is

greater for firms with high information costs than for firms with low information costs.

We study 1,926 U.S. firms from 1979 to 1998 and find that post trade liberalization

duality firms outperform non-duality firms. Among the firms that experience elimination of

4

import tariffs, dual leadership increases Tobins Q by 2.95%. Among firms that experience

elimination of export tariffs, dual leadership increases Tobins Q by 3.83%. The results are

robust to various robustness checks, including controlling for firm fixed effects and firm-level or

industry-level clustering, temporary changes in board leadership structure due to CEO

succession, and additional operating and governance variables that potentially affect firm

performance. To test the information advantage argument of CEO duality, we partition firms

based on the level of information costs before trade liberalization. Consistent with the

information advantage argument, the performance contribution of dual leadership is stronger for

firms with high information costs than for firms with low information costs.

An alternative explanation for our findings is that the exogenous shock to the competitive

environment reduces agency costs and that duality firms disproportionately benefit from such

reduction. To test this possibility, we partition the sample based on the strength of corporate

governance before trade liberalization. As firms with weak governance have higher agency costs

(Shleifer and Vishny, 1997; Hartzell and Starks, 2003), we should then find a stronger duality

effect in the partition of firms with weaker governance if our results are driven by a reduction in

agency costs. We find the opposite. The positive effect of dual leadership is stronger for firms

with higher total institutional ownership, higher holdings by block institutional investors, and

higher holdings by the top five institutional investors.

As the literature is replete with evidence that competition is the enemy of the sloth

(Nickell, 1996; Bertrand and Mullainathan, 2003; Giroud and Mueller, 2010), we also test

whether dual leadership has positive effects on firm performance after controlling for efficiency

gains due to increased competition. We find that post trade liberalization firms that receive tariff

concessions and improve efficiency or reduce input costs have higher Tobins Q. Importantly,

5

the performance contribution from dual leadership exceeds the performance contribution from

efficiency gains.

We also examine the time trend of dual leadership from 1988 to 1998, as evolutionary

theory predicts that firms adopt the winning governance arrangements over time. We observe a

steady increase in the percentage of firms that combined the CEO and COB titles from 1988 to

1993. However, that percentage has been declining since 1993, which is one year after the launch

of the global movement by regulators toward the separation of CEO and COB titles (Dahya,

Garcia, and Bommel, 2009, p. 180).3

Our paper makes several contributions. First, the literature lacks clear evidence on the

impact of CEO duality on firm performance due to the endogeneity and heterogeneity issues

(Hermalin and Weisbach, 2003; Adams, Hermalin, and Weisbach, 2010). We use an exogenous

shock to a firms competitive environment, thereby mitigating these econometric challenges, and

find evidence that CEO duality is beneficial to firm performance when competition intensifies

and market opportunities expand. Second, although arguments in favor of dual leadership are

well developed, the literature lacks empirical evidence directly linking the benefits of dual

leadership to firm performance. We provide direct evidence for the information benefits of dual

leadership. Third, our results, together with Bloom and Van Reenen (2007), shed light on some

seemingly puzzling phenomena in practice. Firms have been under enormous pressure to abolish

CEO duality for over two decades. In many countries, separate leadership has become the norm.

For example, in the late 1980s, a majority of U.K. firms combined the CEO and COB titles.

Now, less than 5% of U.K. firms still do. In contrast, U.S. firms and some U.S. investors have

3 Between 1992-2004, reports sponsored by national governments, major stock exchanges, or both in at least 16 countries outside of the U.S. recommended splitting the CEO and COB titles. The global movement toward the separation of the CEO and COB positions can be traced back to the 1992 Cadbury Report, a report issued by a committee appointed by the Conservative Government of the United Kingdom (Dahya, Garcia, and Bommel, 2009).

6

been reluctant to embrace the practice.4 We find that duality firms outperform non-duality firms

when competitive environments change. Bloom and Van Reenen (2007) find that the U.S. has

one of the most competitive markets in the world. Our results, viewed together with those of

Bloom and Van Reenen (2007), suggest that firms adopt the winning governance structure in

response to their competitive environments. Therefore, our results have important policy

implications given the strong push for U.S. firms to abolish CEO duality.

The rest of the paper is organized as follows: Section 2 presents the arguments in favor of

and in opposition to dual leadership and summarizes the literature on the duality-performance

relation; Section 3 discusses the research design; Section 4 describes the data collection and the

sample; Section 5 and 6 present the empirical results; finally, Section 7 concludes.

2. Institutional background

2.1. Arguments regarding the costs and benefits of dual leadership

The arguments against dual leadership (or in favor of separate leadership) are largely

based on the agency theory. CEOs of modern corporations have decision rights but not control

rights over shareholder capital. As a result, CEOs have conflicting interests and do not always

act to maximize shareholder value. The board of directors is the apex of the decision control

system of modern corporations, which mitigates agency problems due to the separation of

ownership and control (Fama and Jensen, 1983). Having CEOs lead this decision control

hierarchy likely compromises the effectiveness of the control system and exemplifies the

ultimate conflict of interest. Supporting this conflict-of-interest argument, empirical studies find

4 Morgan, Poulsen, Wolf, and Yang (2011) find that mutual funds support 90% of the shareholder proposals that aim to declassify the board, but support only 34% of the proposals that call to separate the CEO and COB positions. A more recent example is the controversy over whether to strip Jamie Dimon (the CEO of J.P. Morgan Chase) of his COB title at the 2013 shareholder meeting. While CalPERs and proxy advisory firms like Institutional Shareholders Services and Glass, Lewis & Co are in favor of the shareholder proposal to separate the dual roles, the legendary investor, Warren Buffet, and proxy advisory firms like Egan-Jones Proxy Services are against it. Also see Section 7 for more discussion on this issue.

7

that when CEO and COB titles are combined, CEO compensation is higher and the sensitivity of

CEO turnover to firm performance is lower (Core, Holthausen, and Larcker, 1999; Goyal and

Park, 2002). Proponents of separate leadership also argue that this arrangement allows the CEO

to focus on running the business, while the COB focuses on running the board. An independent

and experienced COB can also be a valuable resource and a sounding board for the CEO

(Dalton, Daily, Johnson, and Ellstrand, 1998).

The arguments in support of dual leadership emphasize the unparalleled firm-specific

knowledge of CEOs and the benefits of strong stewardship. As CEOs may often have the best

specific knowledge of the strategic challenges and opportunities facing the firm (Jensen and

Meckling, 1995, p. 13), a CEO who is also in charge of the board should be able to coordinate

board actions and implement strategies more swiftly, giving the firm a competitive edge

particularly in tough business conditions. As acquiring and transmitting firm-specific

information can be costly, dual leadership potentially enjoys large cost savings by eliminating

information transferring and processing costs associated with non-CEO chairmen. Consolidated

power also provides clarity regarding the leadership and direction of the firm, which promotes

effective dealing with external parties (Dalton, Daily, Johnson, and Ellstrand, 1998).

Additionally, the COB title is an integral part of the CEO incentive contract. If a firm does not

award the additional title of COB, its CEO may be less motivated to work and even consider

leaving the firm.5 Firms that practice separate leadership may also have a more difficult time

recruiting new CEOs than firms that combine the titles (Larcker and Tayan, 2011).

Separating the dual roles can interfere with succession planning, i.e., the retiring CEO

remains on the board as the COB and relinquishes the COB title to the new CEO only after the

5 Jamie Dimon, the CEO of J.P. Morgan Chase, said that he may leave the bank if shareholders vote to separate the CEO and COB positions at the 2013 annual meeting (Reuters, May 11, 2003).

8

new CEO successfully passes the probationary period (Brickley, Coles, and Jarrell, 1997;

Brickley, Coles, and Linck, 1999). Dual leadership saves certain costs that separate leadership

creates. Extra compensation to COBs can be sizable. Walt Disney paid $550,732 to its non-

executive chairman in fiscal year 2009. Installing non-CEO COBs creates its own agency

problems in the form of who monitors the monitor (Brickley, Coles, and Jarrell, 1997).6

In summary, it is not theoretically obvious whether dual or separate leadership is more

beneficial to firm performance. Therefore, the efficacy of CEO duality is an empirical question.

2.2. Mixed evidence on the relation between dual leadership and firm performance

Pi and Timme (1993) study 112 U.S. banks from 1987 to 1990 and find a higher return

on assets for those banks with separate titles. Brickley, Coles, and Jarrell (1997) study 661 U.S.

firms in the 1989 Forbes compensation survey and find that firms with separate leadership do

not perform better. Indeed, they conclude that duality firms are associated with better accounting

performance. To compare with Pi and Timme (1993), Brickley, Coles, and Jarrell (1997)

separately study banks and thrifts and find no significant differences in performance across firms

with different board leadership models. Palmon and Wald (2002) study announcements of board

leadership changes from 1896 to 1996. They find that small firms experience negative abnormal

returns when changing from dual to separate leadership, while large firms experience positive

abnormal returns. The results of Palmon and Wald (2002) are in contrast to Faleye (2007b), who

finds that dual leadership increases Tobins Q for complex firms, but decreases it for non-

complex firms. Dalton, Daily, Johnson, and Ellstrand (1998) conduct a meta-analysis of 31

studies, concluding that CEO duality does not affect performance and firm size does not

moderate the duality-performance relation.

6 In Appendix A, we give some examples of arguments that firms make to support their decisions of having a dual leadership structure, including that it promotes clarity regarding the leadership of the firm, facilitates succession planning, and enhances more effective business planning and execution.

9

Dey, Engel, and Liu (2011) study performance consequences of combining or splitting

the titles of CEO and COB from 2001 to 2009. They find that firms combining (splitting) the

titles have higher (lower) announcement returns and better (worse) post-announcement

performance. Larcker, Ormazabal, and Taylor (2011) investigate the market reaction to

legislative and regulatory actions from 2007 to 2009, including two proposed regulations that

would ban CEO duality. They find that the abnormal returns of these events are not related to the

presence of CEO duality, which is inconsistent with the view that CEO duality destroys value. In

1992, the Cadbury Committees Code of Best Practice called on U.K. firms to separate the titles

of CEO and COB. Using this external shock, Dahya, Garcia, and Bommel (2009) test whether

firm performance improved after the separation. They fail to find any performance improvement.

3. Research design

3.1. The Canada-United States Free Trade Agreement of 1989

On January 2, 1988, U.S. President Ronald Reagan and Canadian Prime Minister Brian

Mulroney signed the Canada-United States Free Trade Agreement (FTA), which eliminated

tariffs and other trade barriers between the two countries. To take effect, the FTA had to be

approved by the U.S. Congress and the Canadian Parliament. While it passed the U.S. Congress

smoothly, the FTA encountered strong opposition in Canada. Mulroneys Progressive

Conservative Party controlled the House, but the Senate, which had a Liberal Party majority,

refused to ratify the FTA until Canadians voted on the issue in a national election. Mulroney was

forced to dissolve the Parliament and called a general election. Although more Canadians were

against the FTA than in favor of it, Mulroneys party won the election as they benefitted from

being the only party in favor of the agreement, while the opposition parties split the anti-free-

trade vote. The FTA took effect on January 1, 1989. Since the passage of the FTA was

10

improbable and unexpected, it qualifies as an exogenous shock (Brander, 1991; Thompson,

1993; Guadalupe and Wulf, 2010). The FTA is also a clean policy experiment, untainted by

confounding events such as macroeconomic shocks or financial crises (Trefler, 2004).

The FTA has been shown to significantly impact the competitive environment of U.S.

firms. For example, Clausing (2001) finds that the FTA significantly increases U.S. imports from

Canada, and that the increase is larger for goods with greater tariff reductions. In addition, the

FTA is associated with substantial employment loss, labor productivity gains, and changes in

price-cost margin (Trefler, 2004; Guadalupe and Wulf, 2010). As Appendix B shows, tariff

reductions on Canadian imports to the U.S. and U.S. exports to Canada can be as high as 36%

and 48%, respectively.

3.2. Empirical method

Like other corporate finance research, an analysis of the impact of dual leadership on

firm performance faces the challenges of endogeneity and heterogeneity (Adams, Hermalin, and

Weisbach, 2010; Wintoki, Linck, and Netter, 2012). Ideally to tackle these challenges we would

like to have an exogenous shock to board leadership structure. However, to the best of our

knowledge, such exogenous shocks to U.S. firms do not exist at this time. Therefore, like other

researchers (see, e.g., Frsard, 2010; Gormley, Matsa, and Milborun, 2013), we use an

exogenous shock that disrupts the relation between the endogenous choice variable and the

outcome variable.7 The idea is that because the exogenous shock of the FTA is largely

unexpected, it is difficult to argue that firms optimally choose board leadership structure

beforehand to deal with the problems posed by the new competitive environment. Therefore,

relating pre-existing board leadership structure to post-shock firm performance mitigates the

7 Frsard (2010) uses tariff cut shocks to examine the effect of cash holdings on firm performance. Gormley, Matsa, and Milborun (2013) explore a shock to workplace exposures to carcinogens and examine the effect of CEO compensation on corporate risk-taking. Also see Zingales (1998).

11

endogeneity problem. As the exogenous shock only affects some firms, we also alleviate the

unobserved firm heterogeneity problem through benchmarking performance changes of firms

affected by the exogenous shock against performance changes of firms unaffected or less

affected by the exogenous shock. More specifically, we estimate the following baseline model to

assess the impact of dual leadership on firm performance:

Tobins Qit = 1duali*post89*tariffi + 2post89*tariffi + Xit + dt + di + it, (1)

where i indexes firms, duali is a dummy that equals one if the firm has a stable board

leadership structure of the CEO being the COB, post89 is a dummy that equals one from 1989

onwards, tariffi is the average tariff rate for firm i during the period of 1986-1988, X is a vector

of firm characteristics such as firm size, dt are year dummies that control for time trends, di are

firm fixed effects that absorb time-invariant industry and firm heterogeneities, and it is the error

term adjusted for heteroskedasticity and firm-level clustering.8 It is worth noting that without the

triple interaction term, this framework is a standard differences-in-differences specification with

continuous treatment that other studies (see, e.g., Guadalupe and Wulf, 2010) have used to

exploit the quasi-natural experiment of the 1989 trade liberalization.

To mitigate the endogeneity problems, we need to use board leadership in the event year

or just before (Low, 2009; Guadalupe and Wulf, 2010). Further, Brickley, Coles, and Jarrell

(1997) show that many instances of changes in board leadership are transient due to CEO

succession. To mitigate these problems, we require sample firms to exist in 1988 and follow a

stable board leadership model, i.e., no change in board leadership structure for more than 80% of

firm years for a minimum of four years from 1988 to 1998. For example, a firm with four years

of board data is classified as following a stable board leadership model only if it had the same

board leadership status during 1988-1991, a firm with nine years of board data is classified as 8 Our results hold if we use industry-level clustering instead of firm-level clustering.

12

following a stable board leadership model if it changed board leadership structure only once

during 1988-1996, and a firm with more than ten years of board data is classified as following a

stable board leadership model if it changed board leadership structure only twice during 1988-

1998.9

We use three measures of tariff concessions (tariffi) to capture the changes in competitive

environments: 1) the FTA-mandated U.S. tariff rate reduction (tariff_import) on Canadian

imports, 2) the FTA-mandated Canadian tariff rate reduction on U.S. exports (tariff_export), and

3) a score (tariff_rank) that we construct based on the rankings of the import and export tariff

rates. Specifically, we separately rank firms with positive import tariff rates and positive export

tariff rates into two groups. Firms with above-average tariff rates receive a score of two, while

firms with below-average tariff rates receive a score of one. As a firm can have an above-average

import tariff rate and an above-average export tariff rate, the maximum value of tariff_rank is

four. A firm with a zero import tariff rate and a zero export tariff rate receives a score of zero.

Therefore, tariff_rank captures the aggregate effect that the 1989 FTA has on a firm in terms of

both increased competition and expanded market opportunities. While all tariffs were scheduled

to go to zero after 1989, some tariff reductions took effect immediately and others were to be

phased out over ten years. This phase-out schedule is a potential source of endogeneity. To

address this problem, we follow Guadalupe and Wulf (2010) and treat all industries equally

regardless of their phase-out schedules by exploiting the differential tariff rates during 1986-

1988. In other words, the tariff rates (tariffi) are the average tariff rates of 1986-1988.

We choose Tobins Q as our primary measure of firm performance because it is the

conventional proxy for firm performance in corporate finance literature, including studies on the

9 We also use board leadership structure in 1988 as a robustness check. Our results hold using this alternative definition of duali. We discuss this robustness check in more detail in Section 5.1.3.

13

impact of the board of directors on firm performance (see, e.g., Hermalin and Weisbach, 1991;

Yermack, 1996; Faleye, 2007a). Further, Tobins Q measures the present value of all future cash

flows over the replacement cost of tangible assets. Therefore, it captures all aspects of firm

operation including managerial entrenchment (Lang and Stulz, 1994). We control for other firm

characteristics that might affect Tobins Q, including firm size (the natural logarithm of total

book assets), current-year ROA, one-year and two-year lagged ROA, growth opportunities (sales

growth over the past three years), capital structure (long-term debt over total book assets), and

risk (annualized daily stock return volatility). We define these and other variables and their

computation in Table 1.

4. Sample

As noted earlier, the FTA was implemented in 1989, but had a phase-out schedule of ten

years. Lileeva and Trefler (2010) report that by 1996 the tariff was down to less than one-fifth of

its 1988 level, and by 1998 all tariffs were eliminated. Therefore, we choose 1979-1998 as our

sample period to have an equal number of years before and after 1989. Import tariff data come

from Feenstra (1996), while export tariff data come from Trefler (2004). Tariff rates are

aggregated from the commodity level to the level of the four-digit Standard-Industrial-

Classification (SIC) codes. To get firm-level tariff rates, we first obtain segment sales and the

four-digit SIC codes associated with each segment from the Compustat Segments Database, then

weigh the tariff rates at the level of the four-digit SIC codes by firms segment sales and sum the

weighted rates.10

10 Using segment sales and four-digit SIC codes from the Compustat Segments Database to compute weighted tariff rates yields a more precise measure of tariff concessions than using tariff rates based on the four-digit SIC codes from the Compustat North America Database. This is because the Compustat North America Database assigns the four-digit SIC codes to a firm based on the greatest value of product shipments for a product group. As an example, if a firm produces two products with 40% of its shipments in SIC 2046 and 60% in SIC 6519, the assigned SIC code would be 6519. The import tariff rate for SIC 2046 is 0.045 and for SIC 6519 is zero. Thus, using four-digit SIC

14

Table 2 provides key summary statistics for tariff rates in 1988. We report the data only

for 1988 because it is the year immediately preceding the implementation of the 1989 FTA and

because the firm-level tariff rates (tariffi) are constant throughout the sample period under our

research design. Two patterns are worth noting. First, if a firm receives the FTA-mandated

import tariff reduction, the firm also likely receives the FTA-mandated export tariff reduction.

Indeed, while the FTA affects 66% of the sample, 80% of the 66% sample firms receive both

import and export tariff reductions. Appendix B confirms this overlap. For example, the Rubber

and Plastics Footwear industry (SIC=3021) has the highest import tariff rate and one of the

highest export tariff rates. Second, when firms receive both import and export tariff reductions,

their import and export tariff rates are higher than the rates of firms that receive only one type of

tariff concessions. Because of the high correlation between import and export tariff rates, we do

not include tariff_import and tariff_export simultaneously in the regressions, but use tariff_rank

instead.

To be selected for our sample, the firm cannot be a utility or a financial institution, must

follow a stable board leadership model (as defined in 3.2), and must have positive values of total

assets and net sales, daily stock returns for at least a quarter of the fiscal year, and Compustat

data before 1989. We obtain financial data from Compustat North America, segment sales from

Compustat Segments, and stock returns from CRSP. Board and ownership data come from the

SEC Compact Disclosure Database. To the best of our knowledge, the Disclosure database

provides the earliest coverage for board structure of publicly traded firms, which starts in 1988.

Although Disclosures data coverage does not go back to 1979, which is the start of our sample

period, it is not a concern to our multivariate tests because our research design requires using

codes from the Compustat North America Database will yield a tariff rate of zero, while using segment sales and four-digit SIC codes from the Compustat Segments Database will yield a tariff rate of 0.018 (0.045*0.4=0.018).

15

board structure in the event year or just before (the event year for our study is 1989). The final

sample consists of 1,926 unique firms (25,246 firm years) from 1979 to 1998, or 1,180 unique

firms (16,156 firm years) that have stable dual leadership and 746 unique firms (9,090 firm

years) that have stable separate leadership. Table 3 reports the summary statistics of key firm

characteristics. To mitigate the problem of extreme outliers, we winsorize all operating variables

at the 1% level at both tails except for Tobins Q, total assets, and the number of business

segments, which will be in natural logarithm form in the regressions. The percent of the sample

firms with stable dual leadership is 64%, which is lower than the percentage (about 80%) that

other studies report for similar time periods (Fig. 1). This occurs because the SEC Compact

Disclosure database covers a much larger and more comprehensive set of firms, and small firms

are more likely to have separate leadership (Linck, Netter, and Yang, 2008).

Fig. 2 contrasts time trends of median values of Tobins Q for firms with stable duality

status against firms with stable non-duality status. As we defined earlier, firms with stable

duality status (duality firms) have the CEO as the COB for more than 80% of firm years for a

minimum of four years from 1988 to 1998, while firms with stable non-duality status (non-

duality firms) are those who have a non-CEO COB for more than 80% of firm years for a

minimum of four years from 1988 to 1998. We report median values of Tobins Q to mitigate the

problem of outliers. We graph the time trends separately for firms impacted by the 1989 FTA

(Panel A) and firms not impacted by the 1989 FTA (Panel B).

A cursory inspection of Panel A suggests that for firms impacted by the 1989 FTA, those

with dual leadership have similar valuation as those with separate leadership prior to 1989. Post

the trade liberalization, Tobins Q increases for firms of both types of leadership models,

although the increase appears to be larger for duality firms. Fig. 2 Panel B shows that for the

16

sample not impacted by the 1989 FTA, duality and non-duality firms exhibit similar trends in

Tobins Q. In this sample, we do not observe an obvious shift in trends nor a clear separation in

Tobins Q between duality and non-duality firms around 1989 as we do for the sample that is

impacted by the 1989 FTA. It is interesting to note that for the sample not impacted by the 1989

FTA, duality firms have higher Tobins Q than non-duality firms for most of the sample years.

This pattern is in line with the existing evidence that better performing CEOs are rewarded the

additional title of the COB (Brickley, Coles, and Jarrell, 1997). Overall, Fig. 2 offers suggestive

evidence that duality firms outperform non-duality firms when the business environment

becomes more competitive and dynamic.

5. Main results

5.1. Impact of dual leadership on firm performance

Columns (1), (2), and (3) of Table 4 report regression results from the baseline model,

Eq. (1), using import tariff rates, export tariff rates, and tariff ranks, respectively. The coefficient

of dual*post89*tariff_import is significantly positive in Column (1). The average import tariff

rate for the sample used in the regression is 1.69%. This result suggests that duality firms, which

receive the FTA-mandated tariff cuts on Canadian imports, experience an increase of 2.95% in

Tobins Q, compared to non-duality firms that also receive the FTA-mandated tariff cuts. Using

export tariff rates, we find similar results. The coefficient of dual*post89*tariff_export is

significantly positive in Column (2). Given that the average export tariff rate for the sample used

in the regression is 2.85%, the result suggests that dual leadership increases Tobins Q by 3.83%.

When using tariff ranks, we again find corroborating results. As defined in Section 3.2, tariff

rank is a score based on the rankings of a firms import and export tariff rates. An above-average

tariff rate receives a score of two, a below-average tariff rate receives a score of one, and a zero

17

tariff rate receives a score of zero. Therefore, with the coefficient of dual*post89*tariff_rank

being 0.028, the result suggests that an increase of one rank is associated with an increase of

2.79% in Tobins Q. The economic size of the valuation impact of dual leadership is similar to

that of board size. Yermack (1996) finds that expanding an eight-member board by one director

is associated with a decrease of 4% in Tobins Q.11

Estimation results of our control variables are qualitatively similar to the existing

literature. For example, Tobins Q is negatively and significantly related to firm size and stock

return volatility, similar to the findings in Anderson and Reeb (2003). Tobins Q is positively

and significantly related to sales growth and ROA, similar to the findings in Yermack (1996) and

Anderson and Reeb (2003).

5.1.1. Impact of dual leadership, controlling for additional operating and governance variables

We re-estimate Eq. (1), controlling for other operating and governance variables that

potentially impact Tobins Q. Trade liberalization likely has a smaller effect on diversified firms

(Frsard and Valta, 2012). A plethora of academic and anecdotal evidence shows (see, e.g.,

Frsard, 2010) that large cash reserves give firms a competitive edge when business

environments suddenly change. Therefore, as a robustness check, we add to our baseline model

the natural logarithm of the number of business segments and zCash. Following Frsard (2010),

we compute zCash as last years cash-to-assets ratio minus the industry-year mean over the

industry-year standard deviation. As Columns (1), (3), and (5) of Table 5 show, our results hold

when we control for these aspects of firm operation using import tariff rates, export tariff rates,

and tariff ranks, respectively. In Columns (2), (4), and (6) of Table 5, we also control for the

potential effects of board size, board composition, D&O ownership, and institutional ownership.

11 Board leadership, board size, and board composition are arguable three most heavily analyzed board features. So far, the literature has not found any significant relation between board composition and firm performance (see, e.g., Hermalin and Weisbach, 1991; Hermalin and Weisbach, 2003; Adams, Hermalin, and Weisbach, 2010).

18

As the SEC Compact Disclosure database starts the data coverage in 1988, we set the 1979-1987

values of board size, board composition, D&O ownership, and institutional ownership to be the

same as the 1988 values. Inclusion of these governance variables does not change our results.

5.1.2. Impact of dual leadership on ROA and ROE

As discussed in Section 3.2, we believe that Tobins Q is the best measure of firm

performance for the purpose of our study. However, for completeness, we also test the impact of

dual leadership on operating performance using the accounting measures of return on assets

(ROA) and return on equity (ROE). ROA and ROE are earnings before interests and taxes

(EBIT) over book value of total assets and book value of common equity, respectively. A few

caveats are in order. First, while competition unambiguously promotes efficiency (Nickell,

1996), its impact on profitability is less clear (Nickell, 1996; Giroud and Mueller, 2010; Frsard

and Valta, 2012). Second, accounting measures are prone to managerial manipulation. For

example, Balakrishnan and Cohen (2011) show that in response to competition firms may

increase or decrease financial misreporting. As Table 6 shows, dual leadership has no significant

impact on ROA and ROE for firms that receive tariff concessions on Canadian imports.

However, dual leadership has a significantly positive impact on ROE for firms that receive tariff

reduction on U.S. exports. We also find a significantly positive effect of dual leadership on ROE

using tariff_rank.

5.1.3. Additional robustness checks

For robustness, we use an alternative definition of duali. Specifically, we replace the

measure of stable board leadership structure, which is defined in Section 3.2, with the board

leadership of 1988, which is a dummy that takes the value of one if a firm combined the CEO

and COB positions in 1988. We re-run the baseline model using this dummy and obtain similar

19

results. For example, firms that receive import tariff concessions experience an increase of

2.13% in Tobins Q, if they had a dual leadership structure in 1988. Firms that receive export

tariff concessions experience an increase of 3.62% in Tobins Q, if they have a dual leadership

structure in 1988. When we use tariff ranks, we find that an increase of one rank is associated

with an increase of 2.10% in Tobins Q. In each case, the significant level is at the 1% level or

better.

The FTA affects only the tradable sector (i.e., SIC codes up to 4000). Therefore, as

another robustness check, we re-run the baseline regression, Eq. (1), for two scenarios: 1) firms

whose primary SIC codes are less than 4000; and 2) manufacturing firms (i.e., firms whose

primary SIC codes are between 2000-3999). This restriction reduces the number of observations

to 15,883 and 17,806 firm years for the first and the second scenarios, respectively. Our results

remain qualitatively the same in each scenario. As we discussed in Section 4, we choose to study

firms of all SIC codes, because we want to assess the impact of the FTA on a firm regardless of

whether the FTA affects the primary industry to which the firm is assigned. Further, duality and

non-duality firms not affected by the 1989 FTA serve as useful control groups for duality and

non-duality firms affected by the 1989 FTA.

We are not aware of any shocks that occurred in 1989 and impacted board leadership

structure. Studies such as Rechner and Dalton (1991), Brickley, Coles, and Jarrell (1997), and

Faleye (2007b) collectively show that the percent of firms with dual leadership is stable for the

period from 1978 to 1995. Nevertheless, as another robustness check, we add dual*post89 to the

baseline model to control for the possibility that other shocks contemporaneous with the FTA

could systematically affect duality and non-duality firms. The coefficient of

dual*post89*tariff_export is still significantly positive, but dual*post89 interacted with

20

tariff_import and tariff_rank are no longer significant. Multicollinearity likely causes the latter

two predictors to lose their significance. The variable of dual lacks variation across time. Given

that we already control for time and firm fixed effects, adding dual*post89 to the baseline

regression introduces substantial noise in estimating variables interacted with dual. Consistent

with this idea, the Wald test shows that dual*post89*tariff_import is jointly significant with

post89*tariff_import and dual*post89 at the 1% level and dual*post89*tariff_rank is jointly

significant with post89*tariff_rank and dual*post89 at the 1% level. Further, adding dual*post89

does not improve the value of adjusted R-squared, suggesting that dual*post89 is an unnecessary

predictor variable.12

To summarize, we find strong evidence that dual leadership is superior to separate

leadership when firms business environments become more competitive and dynamic. We find

it instructive that the effect of dual leadership is stronger among firms impacted by export tariff

concessions than among firms impacted by import tariff concessions. As Appendix B and Table

2 show, tariff reductions are steeper on U.S. exports than on Canadian imports. Thus, this pattern

is also consistent with the idea that dual leadership is more conducive to better firm performance,

as steeper tariff reduction likely induces larger shock to firms competitive environments.

5.2. Does dual leadership matter more for firms with high information costs?

According to the literature (see e.g., Brickley, Coles, and Jarrell, 1997), the main benefit

of dual leadership is its information advantage over separate leadership. As a CEO has

unparalleled firm-specific information, combining the CEO and COB titles incurs lower

information acquisition, transmission, and processing costs than separating the titles.

Competition magnifies the information benefits of dual leadership, because slow information

transmitting and processing delays decision-making and when competition is intense, 12 Results of these and the following robustness checks are available upon request.

21

information becomes obsolete at a faster rate and the consequences of lost opportunities due to

delayed decision-making become more severe (Christie, Joye, and Watts, 2003). Consistent with

this idea, the economic literature shows that firms decentralize when they need to respond

quickly to product market competition and when local managers have an information advantage

over their headquarters (Acemoglu, Aghion, Lelarge, Van Reenen, and Zilibotti, 2007; Bloom,

Sadun, and Van Reenen, 2010; Guadalupe and Wulf, 2010). To test the information advantage

argument, we study next the impact of dual leadership on firm performance, partitioning the

sample based on the levels of information costs in 1988. If competition magnifies the

information benefits of dual leadership, we should find a stronger duality effect for firms with

high levels of information costs than for firms with low levels of information costs post trade

liberalization.

Following the literature, we use three variables to measure the level of information costs:

research and development (R&D) expenditure over sales, advertising expenditure over sales, and

intangible assets over total book assets. Studies establish that R&D involves specialized inputs

that are unique to the investing firm and is a powerful proxy for information that is privy to

insiders and costly to transfer and process (Levy, 1985; Titman and Wessels, 1988; Aboody and

Lev, 2000). In terms of advertisement, firms that spend heavily on advertisement likely have

more unique products and non-standardized production inputs (Levy, 1985; Titman and Wessels,

1988). As an advertising campaign is frequently tied to a specific product or company,

advertising itself also creates an intangible asset that is non-transferable (Grullon, Kanatas, and

Kumar, 2006). These considerations suggest that advertising expenditures should be positively

associated with information transmission and processing costs. Lastly, intangible assets likely

capture the level of soft information, which is costly to transfer across parties and over

22

geographic distance (Alam, Chen, Ciccotello, and Ryan, 2012). Further, intangible assets such as

company reputation, customer relationships, and intellectual properties do not have obvious

physical value. Their value critically depends on a firms ability to adapt to the changing

environment and capitalize on new opportunities.13

Table 7 reports the regression results when we separately run the baseline regression, Eq.

1, for firms with high and low information costs. For brevity, we tabulate regression results only

for tariff_rank, as it captures the aggregate shock of both import and export tariff reductions to

firms competitive environments and results using tariff_import and tariff_export are frequently

similar. Columns (1) and (2) report the results using R&D spending as the proxy for the level of

information costs. Following the literature (see, e.g., Himmelberg, Hubbard, and Palia, 1999;

Linck, Netter, and Yang, 2008), we replace missing R&D spending with zero. Contrary to the

information advantage argument of dual leadership, the coefficient estimate of

dual*post89*tariff_rank is significantly positive for firms with low levels of R&D spending and

is insignificant for firms with high levels of R&D spending. We find the same pattern when

using tariff_import. However, we find evidence consistent with the information advantage

argument of dual leadership when using tariff_export. Specifically, the coefficient estimate of

dual*post89*tariff_export is 3.115 with a p-value of 0.006 for the sample with high R&D

spending and is 1.058 with a p-value of 0.005 for the sample of low R&D spending. Using R&D

spending has one limitation, because firms are not required to disclose R&D expenditures if they

13 A prime example is Eastman Kodak. In 1976, Kodak accounted for 90% of film and 85% of camera sales in America. An innovation giant of its day, Kodak had invented the first digital camera just a year earlier. Despite the fact that it pioneered the digitization revolution and predicted in 1979 that digital would replace film by 2010, Kodak was too slow to change. As Rick Braddock (a Kodak director from 1987 to 2012) puts it, the mindset of the company was ready for the challenge: it was Batten down the hatches. We sold the healthcare business and we started the process of developing a digital response. But the way the market shifted was dramatically faster than we had anticipated or than Id ever seen (Financial Times, 2 April 2012). By contrast, Fujifilm, which had been a long-time competitor to Kodak and enjoyed a strong, long-time monopoly on camera film in Japan like Kodak did in the U.S., was able to adapt to the changing world. For the fiscal year of 2011, Fujifilm reported a net income of 769 million, while Kodak reported a net loss of 764 million. Kodak declared bankruptcy on January 19, 2012.

23

are immaterial. Although using the 1988 values has the econometric benefit of conditioning the

duality-performance relation on pre-determined levels of R&D spending, estimation noise arises

if certain firms disclose R&D spending in some years and do not in others. Therefore, as a

robustness check, we partition the sample based on whether a firms R&D is greater than zero in

a given year. The median value of R&D spending is zero for the full sample period and for the

1988 year. Using the median value of all firm years to partition the sample has the additional

advantage of having similar numbers of observations in each sub-sample. We find evidence

consistent with the information advantage argument using all three tariff rates.

Columns (3) and (4) of Table 7 report the results using advertising spending as the proxy

for the level of information costs. Consistent with the information advantage argument of dual

leadership, the coefficient estimate of dual*post89*tariff_rank is significantly positive for firms

with high levels of advertising spending and is insignificant for firms with low levels of

advertising spending. We find similar results using tariff_import and tariff_export. Similar to

R&D spending, for further robustness, we replace missing advertising spending with zero,

partition the sample based on the median value of a firms advertising spending in a given year,

and re-run the baseline model. Again, we find similar results when using all three tariff rates.

Columns (5) and (6) of Table 7 report the results from using intangible assets as the

proxy for the level of information costs. Again, the results support the information advantage

argument of dual leadership. Our results hold using import and export tariff rates. In untabulated

results, we use an alternative measure for intangible assets, which is one minus the ratio of

tangible assets over total book assets. Following Berger, Ofek, and Swary (1996), we compute

tangible assets by taking the sum of 0.715*receivables, 0.547*inventory, and 0.535*Property,

Plant and Equipment. We find similar results using this alternative measure of intangible assets.

24

One potential concern is that the levels of R&D spending, advertising spending, and

intangible assets are systematically related to tariff rates. In such a case, we may capture the

effect of dual leadership associated with changes in the competitive environment instead of

different levels of information costs. To mitigate this concern, we include in the regressions only

firms with above-average tariff rates and firms that are not impacted by the 1989 FTA. Our

results remain qualitatively the same.

5.3. Can reduction in agency costs explain the results?

In the preceding section, we find evidence that the information advantages of the CEO

contribute to the positive effect of dual leadership on firm performance. The main argument

against dual leadership is based on the agency theory, which predicts that CEOs, as agents of the

shareholders, do not always act in the best interests of the shareholders. As product market

competition is arguably the best governance device in minimizing the agency costs (Shleifer and

Vishny, 1997), an alternative explanation may account for our results. Specifically, the

exogenous shock of the 1989 FTA increases competition, forcing firms to reduce agency costs,

and duality firms disproportionately benefit more than non-duality firms from competition-

induced agency cost reduction.

To test this alternative explanation, we partition the sample based on the strength of

corporate governance in 1988, the year immediately preceding the 1989 trade liberalization, and

re-run the baseline regression. If reduction in agency costs drives our results, we should find a

stronger duality effect in the sample of firms with weak governance than in the sample of firms

with strong governance.14 Following the literature (see, e.g., Shleifer and Vishny, 1997; Hartzell

14 We assume that firms with strong corporate governance have lower agency costs and firms with weak corporate governance have higher agency costs. This assumption is consistent with a large body of empirical evidence (see, e.g., Hartzell and Starks, 2003; Agrawal and Chadha, 2005; Dittmar and Mahrt-Smith, 2007; Cornett, Marcus, Tehranian, 2008).

25

and Starks, 2003; Frsard and Valta, 2012), we use institutional ownership, stock holdings by

block institutional investors, and stock holdings by top five institutional investors to proxy for

the strength of corporate governance. As the SEC Compact Disclosure Database does not have

detailed institutional ownership data, we obtain the data needed for our regression analysis in this

section from the Thomson-Reuters Institutional Holdings (13F) Database. As Table 8 shows, we

find that the positive effect of dual leadership is larger and more significant for firms with strong

corporate governance than for firms with weak corporate governance. Therefore, the results do

not support the agency cost reduction argument. Our findings are consistent with Morck,

Stangeland, and Yeung (2000) and Amore and Zaldokas (2012), who find that firms with weak

corporate governance perform poorly when competition intensifies.

5.4. A horse race between dual leadership and efficiency gains

The literature is replete with empirical evidence that competition is the enemy of sloth

(Nickell, 1996, p. 724; Bertrand and Mullainathan, 2003; Giroud and Mueller, 2010). Therefore,

it is instructive to examine whether the duality effect remains when we consider efficiency gains

due to increased competition. For this test, we run a horse race by adding to the baseline model

various proxies for sloth including sales per employee, overhead costs, input costs, and employee

wage. Sales per employee is a turnover ratio that directly measures employee productivity and is

a common proxy for firm efficiency (see, e.g., Vining and Boardman, 1992; Clark, 1984).

Overhead costs is the ratio of selling, general and administrative expenses to sales. Input costs is

the ratio of the costs of goods sold to sales. Employee wage is the staff expense to employees.

Giroud and Mueller (2010) find that overhead costs, input costs, and employee wage all increase

once managers are insulated from competition and the takeover market. Similar to our treatment

of other operating variables, we winsorize sales per employee, overhead costs, and the ratio of

26

the costs of goods sold at the 1% level at both tails and use the natural logarithm of wage to

mitigate the outlier problem. As the results using tariff_import, tariff_export, and tariff_rank are

similar, we tabulate regression results only for tariff_rank.

As Table 9 shows, our results hold after controlling for different measures of sloth, with

the exception of employee wage. The variable of dual* post89*tariff_rank is insignificant in the

wage regression, likely because the number of observations is small. Many firms do not report

staff expense thereby reducing the number of observations in the wage regression by nearly 90%.

In the regressions of other proxies for sloth, dual*post89*tariff_rank is highly significant.

Further, the coefficient size is similar to that in the baseline model in Column (3) of Table 4. We

also find that firms that increase employee turnover and lower input costs after the trade

liberalization experience a larger increase in Tobins Q. These results are consistent with the

prevailing view that competition promotes efficiency. Importantly, the economic impact of dual

leadership is larger than those proxies for sloth. For example, Column (1) suggests that duality

firms with a tariff rank of two experience an increase in Tobins Q that is 2.61% higher than

duality firms with a tariff rank of one. For comparison, an improvement of 10 percentage points

in sales per employee is associated with an increase of 1.12% in Tobins Q. (Measures of sloth

are in decimals.)15

6. Additional results

6.1. Duality effect or survival bias

Studies have found that CEOs with more decision-making power are associated with

more variable firm performance (Adams, Almedia, and Ferreira, 2005). If duality firms 15 As a robustness check, we exclude dual*post89*tariff_rank from the regressions and re-estimate the effect of sloth*post89*tariff_rank. The idea is that if dual*post89*tariff_rank and sloth*post89*tariff_rank are highly correlated, this may cause a downward bias of the coefficient estimates of sloth*post89*tariff_rank. We do not find such is the case; we find that the coefficient estimates of sloth*post89*tariff_rank are stable. For example, when we re-run the regression in Column (1) excluding dual*post89*tariff_rank, the coefficient estimate of sloth*post89*tariff_rank is 0.113 with 1% significance.

27

experience larger variances in firm performance, then a sudden change in firms competitive

environments may disproportionately eliminate a larger number of poorly performing duality

firms than poorly performing non-duality firms. In such a case, our results may arise from a

survival bias instead of a detection of true performance enhancement due to dual leadership. To

assess this possibility, we study corporate failure rate. If we do not find a disproportionately

higher corporate failure rate for duality firms than non-duality firms among those firms that are

affected by trade liberalization, then the survival bias is not a concern for our results.

We construct the sample of failed firms using two different approaches. The first

approach uses Compustat and CRSP. Specifically, we first identify our sample firms that file for

Chapter 7 or Chapter 11bankcruptcy using Compustat data item STALT. We find 42 unique

firms. We then verify Compustat bankruptcy records against CRSP delisting data, namely

whether CRSP data item HDLRSN is coded 02 or 03. A code of 02 indicates that the firm

is in bankruptcy, while 03 indicates liquidation. Since a firm can continue to have stock price

data after filing for bankruptcy, a large gap sometimes exists between the delisting date in CRSP

and the date of the last financial statement in Compustat. For our sample, the mode of this gap is

three years, with a maximum of 14. To ensure that we robustly test the survival bias concern, we

want to be as aggressive as possible in identifying corporate failures for our sample period. Thus,

we treat a delisting firm as failed during our sample period if the gap between the delisting date

in CRSP and the date of the last financial statement in Compustat is less than or equal to four

years. Seventy-five percent of our sample firms are delisted within four years of the last financial

statement. We identify an additional 83 unique bankrupt firms using this process. Therefore, we

identify 125 corporate failures from 1988 to 1998 or 120 corporate failures post trade

liberalization (1989-1998). We use the year of the last financial statement as the year in which

28

the firm fails. We report corporate failure rates using the first approach for the sample period of

1988-1998 in Table 10 Panel A. We choose 1988 as our starting year, because to study the

impact of the 1989 FTA we require our sample firms to have financial data in 1988.

As the first approach is a crude way to identify corporate failures (e.g., it does not have

precise bankruptcy dates and may not include all bankruptcies or may incorrectly include non-

bankruptcy events), we construct the second sample of failed firms using a proprietary

database.16 The advantage of this database is that it contains all bankruptcies filed by U.S.

publicly traded firms and the bankruptcy filing dates. The disadvantage of the database is that it

covers only part of our sample period, 1991-1998. However, this partial coverage should not be a

concern, as we care about corporate failures in years after trade liberalization much more than in

years before. We report the time trend of corporate failure rates using the second approach in

Table 10 Panel B.

To provide another benchmark, we also report corporate failure rate for firms not

impacted by the 1989 FTA. If the relative failure rates of duality vs. non-duality firms that are

impacted by the 1989 FTA are similar to the relative failure rates of duality vs. non-duality firms

that are not impacted by the 1989 FTA, then it provides additional assurance that survival bias is

not a concern to our study. As Table 10 shows, the two approaches tell a consistent story. Of the

firms that are impacted by the 1989 FTA, duality firms do not have a higher failure rate than

non-duality firms after the trade liberalization. Additionally, the relative failure rates of duality

vs. non-duality firms that are impacted by the 1989 FTA are similar to the relative failure rates of

duality vs. non-duality firms that are not impacted by the 1989 FTA. We observe similar patterns

16 For more details about this database, please refer to Lemmon, Ma, and Tashjian (2009). We deeply appreciate and thank Yung-Yun Ma for his generosity in sharing with us his manually-collected data.

29

when we separately analyze firms impacted by import tariff concessions and firms impacted by

export tariff concessions. In summary, survival bias does not appear to drive our results.

6.2. Event study results

We conduct an event study to assess the market perception of the value contribution of

dual leadership as the FTA came into effect. As there are no clear event dates, we follow the

long-run event study methodology in Chhaochharia and Grinstein (2007) and Wintoki (2007)

and calculate excess portfolio returns for an extended event window using the following four-

factor model:

Rpt-Rft =p + 1(Rmt- Rft) + 2SMBt + 3HMLt + 4MOMt + it, (2)

where Rft is the risk-free rate. The first three factors, (Rmt-Rft), SMBt, and HMLt, are based

on Fama and French (1993) and measure the market excess return, the differences in returns

between portfolios of small and large stocks, and the differences in returns between portfolios of

high and low book-to-market stocks, respectively. The fourth factor, the momentum factor

(MOMt), is based on Carhart (1997). It measures the differences in returns between a portfolio of

stocks with high returns in the past year and a portfolio of stocks with low returns in the past

year. Rpt is the equally weighted portfolio of duality firms or non-duality firms that are affected

by the 1989 FTA. p is the daily excess portfolio returns relative to the four factors. To get an

annualized rate, we multiply p by 252 trading days. If the market perceives that the 1989 FTA

benefits duality firms more than non-duality firms, then a position long in the portfolio of duality

firms and short in the portfolio of non-duality firms should yield positive returns. We use daily

stock returns adjusted for delisting returns to perform our study. Based on our review of the

events, we choose four different event windows. Results are reported in Table 11.

30

As Table 11 shows, we do not find any significant results for any of our event windows.

One reason for the non-results could be that negotiations regarding a free-trade agreement

encapsulate an event that spans too long a period for our methodology to detect any statistically

significant effects. A free-trade regime between the U.S. and Canada had been on the working

agendas of both governments since the early 1900s. Canadian Prime Minister Mulroney formally

requested that the U.S. and Canada explore the possibility of a comprehensive free trade

agreement on September 26, 1985. It could also be that the U.S. stock market did not view the

implementation of the 1989 FTA as a significant event. Although the 1989 FTA was extremely

contentious in Canada to the extent that the Canadian general election of 1988 was largely fought

on this single issue, the FTA did not garner much attention in the U.S. It passed without any

fanfare in the House by a vote of 366 to 40 and in the Senate by a vote of 83 to nine. In fact,

polls show that up to 40% of Americans were unaware that the FTA had been signed compared

to 3% of Canadians.17

6.3. Time trends of board leadership

Evolutionary theory of organizations predicts that firms adopt governance arrangements

that give them the competitive edge (see, e.g., Kole and Lehn, 1997; Fama and Jensen, 1983). If

dual leadership is a superior leadership structure in a competitive environment, we should expect

more firms to adopt this structure over time. In this section, we examine the time trend of dual

leadership in search of further corroborating evidence for our main results in Table 4.

As we are interested in firms response to the 1989 FTA, we require for this investigation

that the sample firms exist in 1988, but relax the requirement that the firms follow a stable board

leadership structure. Consequently, we have 13,627 firm years in the sample impacted by the

FTA and 12,110 firm years in the sample not impacted by the FTA. In Fig. 3, we plot the time 17 http://www.enotes.com/u-s-canada-free-trade-agreement-1989-reference/u-s-canada-free-trade-agreement-1989

31

trends of board leadership and board composition partitioned by whether a firm is impacted by

the FTA. We add the time trend of board composition to our study, because we want to place the

time trend of board leadership in the broader context of the development of board structure.

Regardless of whether they are impacted by the FTA, more firms appear to adopt dual

leadership from 1988 through the early 1990s, which is consistent with both the evolutionary

arguments that firms adopt the governance structure that maximizes their survival chances and

our earlier findings that dual leadership is a superior leadership structure in competitive and

growing business environments. However, the upward trend stalls after 1993. If anything, more

firms appear to move to separate leadership in the latter part of our sample period. Some key

events may have contributed to this change in trend. In 1992, the U.K. issued the Cadbury

Report, which called on firms to abolish the practice of combined CEO and COB positions and

arguably started the global movement toward abolishing CEO duality (Dahya, Garcia, and

Bommel, 2009). In the same year, the Conference Board, a global business membership

organization headquartered in the U.S., issued a report recommending that firms separate the two

titles. Therefore, the time trend of dual leadership suggests that firms decisions regarding board

leadership structure are not only a function of the competitive pressure from the product market

but also of a broad range of socioeconomic factors. Consistent with the notion that firms are

under increasing pressure to establish independent boards, Fig. 3 also shows that the ratio of

outside directors on the board steadily increases throughout our sample period.

7. Conclusion and Discussion

Despite the large body of literature on board leadership, the evidence on the relation

between board leadership structure and firm performance is mixed due to the endogeneity and

heterogeneity challenges. In this paper, we employ a new framework that mitigates these

32

challenges and find that dual leadership is beneficial to firm performance when competition

increases and the product market expands. Further, the positive effect of dual leadership is larger

when firms have high levels of information costs and better corporate governance.

Our results shed light on some seemingly puzzling phenomena in practice. Firms have

been under enormous pressure to abolish CEO duality for over two decades. While firms in other

countries seem to be more amicable to the idea, U.S. firms have been reluctant to change. For

example, in the late 1980s, a majority of U.K. firms combined the CEO and COB titles (Dahya,

Garcia, and Bommel, 2009). Now, less than 5% of U.K. firms still do.18 In contrast, the majority

of U.S. firms still have a dual leadership structure. Certain investors are also less enthusiastic in

supporting the effort to separate the CEO and COB titles than some other governance initiatives.

For instance, Morgan, Poulsen, Wolf, and Yang (2011) find that mutual funds support 90% of

the shareholder proposals that aim to declassify the board, but support only 34% of the proposals

that call to separate the CEO and COB positions. Our findings help explain the reluctance on the

part of U.S. firms and certain investor groups to embrace the independent COB. Our results also

complement Bloom and Van Reenen (2007), who find that the U.S. has the best management

practice of the four countries (U.S., U.K., France and Germany) surveyed. Bloom and Van

Reenen also find that poor management practice is more prevalent when product market

competition is weak and that the U.S. has the most competitive market of the four countries.

One limitation of our study is that we do not distinguish amongst non-CEO Chairmen,

e.g., whether the Chairman is a former or present employee of the firm or is an independent

director. This limitation is attributable partially to data availability and partially to the fact that

the practice of having an independent chair is a recent phenomenon. Firms made the noticeable

18 Financial Reporting Council, Developments in Corporate Governance 2011, available at http://www.frc.org.uk/getattachment/5f4fada9-2a88-43a4-bbec-be15b6519e79/Developments-in-Corporate-Governance-2011-The-impact-and-implementation-of-the-UK-Corporate-Governance-and-Stewardship-Codes.aspx

33

move to adopt the practice of independent chairman after the passage of the Sarbanes-Oxley Act

of 2002. By 2007, just 13% of S&P500 firms have a truly independent chairman (PR Newswire,

July 30, 2012).19 Despite the data limitation, we believe that our results are useful in

understanding the existing literature, which has historically defined board leadership structure

similarly to this paper. Our results are also useful in explaining corporate behaviors during our

sample period as well as at the present time. As we mentioned earlier, the majority of U.S. firms

still combine the CEO and COB positions.

Importantly, our results highlight the link between the identity of the COB and his

influence on firm performance. Prior literature has long argued that CEOs possess unparalleled

firm-specific information, which gives them a unique advantage over non-CEO chairmen in

leading the board of directors. We provide evidence explicitly linking information costs to the

positive impact of dual leadership on firm performance. Favaro, Karlsson, and Neilson (2010)

report that [a]t the outset of the decade [2000], roughly half of the North American and

European CEOs entering office were named chairman and CEO. In 2009s incoming class, that

number had fallen to 16.5% in North America and 7.1% in Europe. The current push towards

more independent chairmen will inevitably result in a more heterogeneous distribution of non-

CEO chairmen. Future work is urgently needed to understand the identities and different

incentives of newly minted COBs, as well as how different types of COBs may have different

impacts on firm performance and corporate polices.

19 10 years later: Sarbanes-Oxley Act Continues to Shape Board Governance, available at http://www.prnewswire.com/news-releases/10-years-later-sarbanes-oxley-act-continues-to-shape-board-governance-164296516.html

34

Appendix A: Examples of Arguments Made by Firms in Support of Dual Leadership Argument 1: COB selection is part of the succession planning process. The CEO is the best person to set board agenda. Honeywell, Inc., in its 2003 proxy statement, notes that [t]he Company has no fixed rule as to whether these offices should be vested in the same person or two different people, or whether the Chairman should be an employee of the Company or should be elected from among the non-employee directors. The Board believes that this issue is part of the succession planning process and that it is in the best interests of the Company to make such a determination when it elects a new CEO. Under Honeywells Corporate Governance Guidelines, the Chairman establishes the agenda for each Board meeting. The Board believes that the CEO is in the best position to develop this agenda from among the many short-term and long-term issues facing Honeywell. available at http://www.sec.gov/Archives/edgar/data/773840/000095011703000983/a34157.txt Argument 2: Dual leadership provides clarity regarding the leadership of the firm. In their statement to oppose a shareholder proposal calling for a separate COB and CEO filed at the 2010 annual shareholder meeting, the board of directors of Goldman Sachs reasons that [t]he most effective leadership model for our firm at this time is to have the roles of CEO and Chairman combined this structure helps to ensure clarity regarding leadership of the firm, allows the firm to speak with one voice and provides for efficient coordination of Board action, particularly in times of market turmoil or crisis. The combination of the Chairmans ability to call and set the agenda for Board meetings with the CEOs intimate knowledge of our business, including our risk management framework, provides the best structure for the efficient operation of our Board process and effective leadership of our Board overall. This structure avoids potential confusion as to leadership roles and duplication of efforts that can result from the roles being separated, especially in complex firms like ours where the information necessary to make critical decisions is often in flux. available at http://www.sec.gov/Archives/edgar/data/886982/000119312510078005/ddef14a.htm Argument 3: Dual leadership promotes more effective business planning and execution. Office Depot, in their 2009 proxy statement, states that [t]he Board has given careful consideration to separating the roles of Chairman and Chief Executive Officer and has determined that the Company and its shareholders are best served by having Mr. Odland, serve as both Chairman of the Board and Chief Executive Officer. Mr. Odlands combined role as Chairman and Chief Executive Officer promotes unified leadership and direction for the Board and executive management and it allows for a single, clear focus for the chain of command to execute the Companys strategic initiatives and business plans. available at http://www.sec.gov/Archives/edgar/data/800240/000119312509050893/ddef14a.htm

35

Appendix B: Top 20 U.S. Industries with Highest Import and Export Tariff Rates, 1986-1988

Four-digit SIC

Industry description Average tariffs

Import tariff 3021 Rubber and Plastics Footwear 36.06% 0182 Food Crops Grown Under Cover 33.40% 2342 Brassieres, Girdles, and Allied Garments 29.13% 2326 Men's and Boys' Work Clothing 28.88% 2075 Soybean Oil Mills 22.49% 2321 Men's and Boys' Shirts, Except Work Shirts 21.90% 2325 Men's and Boys' Separate Trousers and Slacks 21.06% 2331 Women's, Misses', and Juniors' Blouses and Shirts 20.86% 2335 Women's, Misses', and Juniors' Dresses 20.14% 3253 Ceramic Wall and Floor Tile 20.00% 2311 Men's and Boys' Suits, Coats, and Overcoats 19.97% 2111 Cigarettes 19.33% 2337 Women's, Misses', and Juniors' Suits, Skirts, and Coats 18.11% 2369 Girls', Children's, and Infants' Outerwear, Not Elsewhere Classified 18.10% 2252 Hosiery, Not Elsewhere Classified 16.81% 2231 Broadwoven Fabric Mills, Wool (Including Dyeing and Finishing) 16.53% 2381 Dress and Work Gloves, Except Knit and All-Leather 14.99% 2257 Weft Knit Fabric Mills 14.69% 3262 Vitreous China Table and Kitchen Articles 14.68% 3151 Leather Gloves and Mittens 14.56% Export tariff 2082 Malt Beverages 48.32% 2259 Knitting Mills, Not Elsewhere Classified 24.33% 3151 Leather Gloves and Mittens 24.33% 2381 Dress and Work Gloves, Except Knit and All-Leather 24.33% 2369 Girls', Children's, and Infants' Outerwear, Not Elsewhere Classified 24.26% 2335 Women's, Misses', and Juniors' Dresses 24.22% 2361 Girls', Children's, and Infants' Dresses, Blouses, and Shirts 24.10% 2331 Women's, Misses', and Juniors' Blouses and Shirts 23.73% 2386 Leather and Sheep-Lined Clothing 23.59% 2329 Men's and Boys' Clothing, Not Elsewhere Classified 23.51% 2321 Men's and Boys' Shirts, Except Work Shirts 23.43% 2311 Men's and Boys' Suits, Coats, and Overcoats 23.13% 2342 Brassieres, Girdles, and Allied Garments 22.99% 2391 Curtains and Draperies 22.75% 2253 Knit Outerwear Mills 22.69% 2337 Women's, Misses', and Juniors' Suits, Skirts, and Coats 22.20% 3021 Rubber and Plastics Footwear 22.01% 3142 House Slippers 22.01% 3143 Men's Footwear, Except Athletic 22.01% 3144 Women's Footwear, Except Athletic 22.01%

36

References Aboody, D., Lev, B., 2000. Information asymmetry, R&D, and insider gains. Journal of Finance

55, 2747-2766. Acemoglu, D., Aghion, P., Lelarge, C., Van Reenen, J., Zilibotti, F., 2007. Technology,

information, and the decentralization of the firm. Quarterly Journal of Economics 122, 1759-1799.

Adams, R., Almedia, H., Ferreira, D., 2005. Powerful CEOs and their impact on corporate

performance. Review of Financial Studies 18, 1403-1432. Adams, R., Hermalin, B., Weisbach, M., 2010. The role of boards of directors in corporate

governance: A conceptual framework and survey. Journal of Economic Literature 48, 58-107.

Agrawal, A., Chadha, S., 2005. Corporate governance and accounting scandals. Journal of Law

and Economics 48, 371-406. Alam, Z., Chen, M., Ciccotello, C., Ryan, H., 2012. Does the location of directors matter?

Information acquisition and board decisions. Journal of Financial and Quantitative Analysis, forthcoming.

Amore, M., Zaldokas, A., 2012. Corporate governance and international trade shocks. Working

paper. Anderson, R., Reeb, D., 2003. Founding-family ownership and firm performance: Evidence from

the S&P500. Journal of Finance 58, 1301-1328. Balakrishnan, K., Cohen, D., 2011. Product market competition and financial accounting

misreporting. Working paper. Berger, P., Ofek, E., Swary, I., 1996. Investor valuation of the abandonment option. Journal of

Financial Economics 42, 257-287. Bertrand, M., Mullainathan, S., 2003. Enjoying the quiet life? Corporate governance and

managerial preferences. Journal of Political Economy 111, 1043-1075. Bloom, N., Sadun, R., Van Reenen, J., 2010, Does product market competition lead firms to

decentralize. American Economic Review: Papers & Proceedings 100, 434-438. Bloom, N., Van Reenen, J., 2007. Measuring and explaining management practices across firms

and countries. Quarterly Journal of Economics 122, 1351-1408. Brander, J., 1991. Election polls, free trade, and th