Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
EISEVIER
Journal of Financial Economics 40 (1996) 185-211
Higher market valuation of companies with
a small board of directors
David Yermack
Stern School of’ Business, New York Universiiy, New York, NY 10012. USA
(Received November 1994; final version received July 1995)
Abstract
I present evidence consistent
with theories that small boards of directors are more
effective, Using Tobin’s Q as an approximation of market valuation, I find an inverse
association between board size and firm value in a sample of 452 large U.S. industrial
corporations between 1984 and 1991. The result is robust to numerous controls for
company size, industry membership, inside stock ownership, growth opportunities, and
alternative corporate governance structures. Companies with small boards also exhibit
more favorable values for financial ratios, and provide stronger CEO performance
incentives from compensation and the threat of dismissal.
Key words: Boards of directors; Corporate governance
JEL classification: G30; G32; K22
1. Introduction
A growing body of empirical research examines the structure and effectiveness
of corporate governance systems. An important insight from this literature is
that top managers’ decisions appear to be influenced by executive compensa-
tion, takeover threats, monitoring by boards of directors, and other control
I have benefited from helpful comments by Rita Maldonado-Bear, Wayne Mikkelson (the editor),
Eli Ofek, participants in the NYU finance seminar, and an anonymous referee. I thank Eli Ofek for
the use of company diversification data, and Andrei Shleifer for arranging financial support. Data
collection for this paper would not have been possible without the help of research assistants Jason
Barro and Melissa McSherry, the staff of the Cole Room at Harvard Business School’s Baker
Library, and more than 100 companies that kindly responded to data inquiries.
0304-405X/96/$15.00 0 1996 Elsevier Science S.A. All rights reserved
SSDI 0304405X9500844 5
186
D. YermacklJournal qf’Financia1 Economics 40 (19%) 1X5&2/l
mechanisms. I contribute to this literature by evaluating a proposal for limiting
the size of boards of directors in order to improve their effectiveness. My
evidence supports this proposal, as I find an inverse association between firm
value and board size in a panel of major U.S. companies.
Lipton and Lorsch (1992) state that ‘. . . the norms of behavior in most
boardrooms are dysfunctional’, because directors rarely criticize the policies of
top managers or hold candid discussions about corporate performance. Believ-
ing that these problems increase with the number of directors, Lipton and
Lorsch recommend limiting the membership of boards to ten people, with
a preferred size of eight or nine. The proposal amounts to a conjecture that even
if boards’ capacities for monitoring increase with board size, the benefits are
outweighed by such costs as slower decision-making, less-candid discussions of
managerial performance, and biases against risk-taking. Jensen (1993) takes up
this theme, pointing out the ‘great emphasis on politeness and courtesy at the
expense of truth and frankness in boardrooms’ and stating that ‘when boards get
beyond seven or eight people they are less likely to function effectively and are
easier for the CEO to control’.
Some evidence shows that reducing board size has become a priority for
institutional investors, dissident directors, and corporate raiders seeking to
improve troubled companies. Kini et al. (1995) present evidence that board size
shrinks after successful tender offers for under-performing firms. At American
Express, the outside director who in 1993 organized the removal of the com-
pany’s CEO cited the ‘unwieldy’ 19-person board as an obstacle to change,
stating that the ‘size of the board does make a difference’, according to Monks
and Minow (1995). Smaller boards have emerged recently during overhauls of
corporate governance at such prominent companies as General Motors, IBM,
Occidental Petroleum, Scott Paper, W.R. Grace, Time Warner, and Westing-
house Electric. Institutional investor pressure reportedly contributed to many
of these changes, such as the 1995 reduction in Grace’s board from 22 directors
to 12.
In a sample of 452 large U.S. public corporations observed over the period
1984 to 1991, I find an inverse relation between firm market value, as repre-
sented by Tobin’s Q, and the size of the board of directors. The association
appears in both cross-sectional analyses of the variation among firms and in
time-series analyses of the variation within individual companies. The negative
relation between board size and firm value attenuates as boards become large,
implying that the greatest incremental costs arise as boards grow from in size
from small to medium. The loss in firm value when boards grow from six to 12
members, for example, is estimated to be equal to the value lost when boards
grow from 12 to 24. Very few boards have fewer than six or more than 24
directors,
A range of additional evidence is consistent with the finding that companies
achieve the highest market value when boards are small. Several measures of
D. Yermack/Journal of Financial Economics 40 (I 996) I&- 211
187
operating efficiency and profitability are negatively related over time to board
size within firms. Smaller boards are more likely to dismiss CEOs following
periods of poor performance. Similarly, evidence shows that CEO compensation
exhibits greater sensitivity to performance in companies with small boards.
Stock returns for a sample of companies announcing significant changes in
board size show that investors react positively when boards shrink and nega-
tively when board size increases.
The inverse association between board size and firm value proves robust to
a variety of tests for alternative explanations. I introduce variables to control for
firm size, industry, board composition, inside stock ownership, the presence of
growth opportunities, diversification, company age, and different corporate
governance structures. None of these modifications changes the conclusion that
companies with small boards are valued more highly in the capital markets.
An alternative interpretation of the results is that board size arises from prior
company performance, with troubled firms adding directors to increase
monitoring capacity. I conduct a range of tests to obtain insight into the
direction of causation between board size and firm value. The tests show that
while the rate of director turnover increases following poor performance, board
size remains quite stable over time with little sensitivity to performance.
The remainder of this paper is organized as follows. Section 2 reviews prior
research on board structure and firm performance. Section 3 presents the main
result of an inverse association between board size and firm value, and illustrates
the result’s robustness to controls for firm size, growth opportunities, diversifica-
tion, board composition, and other variables. Section 4 presents evidence about
the direction of causation between board size and firm value. Section 5 provides
supporting evidence that smaller boards oversee managers more effectively,
showing that financial ratios and CEO incentives are stronger when boards are
small and that investors react favorably to large reductions in board size.
Section 6 concludes the paper.
2. Prior research on hoard structure
Criticisms and proposals for the reform of boards of directors have prolif-
erated in recent years. Monks and Minow (1995) present a lengthy summary of
this literature, which relies upon a premise that monitoring by the board can
improve the quality of managers’ decisions. Many commentators urge that
boards have a large fraction of outside directors, that directors own large
amounts of company stock, that CEOs have only limited power to set board
agendas and appoint new directors, and that rigorous CEO performance re-
views take place regularly. As noted above, limiting board size has begun
appearing on some agendas for reform, although only Lipton and Lorsch (1992)
and Jensen (1993) identify board size as a high priority.
188 D. YermacklJoumal of Financial Economics 40 (I 996) 185 21 I
Much empirical research has examined whether board structure is related to
company performance, but these studies have largely overlooked board size.
Instead, investigators have most frequently examined the importance of outside
directors and directors’ equity ownership.
Studying board composition, Hermalin and Weisbach (1991) find no relation
between firm performance and the fraction of outside directors. However, this
conclusion is not supported by Baysinger and Butler (1985), who find some
evidence that companies perform better if boards include more outsiders. Other
studies find that boards dominated by outsiders are more likely to behave in
shareholders’ interest. See, for example, Weisbach (1988) (CEO turnover), Byrd
and Hickman (1992) (tender offer bids), and Brickley, Coles, and Terry (1994)
(poison pill adoptions and control auctions). Rosenstein and Wyatt (1990) find
positive investor reactions to appointments of outside directors.
With respect to board stock ownership, Merck, Shleifer, and Vishny (1988)
find significant, though nonmonotonic, associations between different levels of
director stock ownership and Tobin’s Q, suggesting that some levels of board
stock ownership have systematic advantages. McConnell and Servaes (1990)
and Hermalin and Weisbach (1991) report similar results, while Bagnani,
Milonas, Saunders, and Travlos (1994) find that bondholder returns also exhibit
a nonmonotonic association with board stock ownership.
Lipton and Lorsch (1992), Jensen (1993), and other advocates of small boards
contend that board size affects corporate governance independent of other
board attributes. As noted above, their arguments focus on the productivity
losses that arise when work groups grow large, an insight borrowed from
organizational behavior research such as Steiner (1972) and Hackman (1990).
According to Jensen (1993), ‘. . as groups increase in size they become less
effective because the coordination and process problems overwhelm the advant-
ages from having more people to draw on’.
Empirical research on the importance of board size is thin. Holthausen and
Larcker (1993a, b) consider board size among a range of variables that might
influence executive compensation and company performance. Holthausen and
Larcker (1993a) present results indicating a positive association between board
size and the value of CEO compensation. Holthausen and Larcker (1993b) fail
to find consistent evidence of an association between board size and company
performance.
A clear problem in studying board size is that the number of directors might
arise endogenously as a function of other variables, such as company size,
performance, or the CEO’s preferences. Along these lines, the managerial quality
hypothesis of Byrd and Hickman (1992) argues that high-caliber CEOs may
‘ . . . dress up their firms’ boards with independent directors’ to please share-
holders with an illusion of active monitoring; a similar argument could be made
about the willingness of good CEOs to surround themselves with small boards.
Because many intangible forces of this type might influence board size, we
D. Yemack/Journal of Financial Economics 40 (1996) 185-211
189
cannot accept at face value an association between board size and firm value
without considering alternative explanations. I investigate many of these pos-
sible explanations in Section 3.3, after presenting my main finding of a negative
relation between firm value and board size.
3. Board size and firm value
The main hypothesis of this paper is that firm value depends on the quality of
monitoring and decision-making by the board of directors, and that the board’s
size represents an important determinant of its performance. Below I estimate
a straightforward model of the relation between firm value and board size.
I follow the methods of several recent related studies, such as Merck, Shleifer,
and Vishny (1988), Hermalin and Weisbach (1991), and Lang and Stulz (1994)
by regressing a set of explanatory variables against an estimate of Tobin’s Q,
which measures the ratio of a firm’s market value divided by the replacement
cost of its assets. I include controls for such variables as firm size, industry
membership, board composition, and past company performance. After presen-
ting the main result, I illustrate its robustness to a variety of alternative
specifications and evaluate whether alternative theories can account for the
observed inverse relation between board size and firm value.
3.1. Data description
My analysis uses a panel of firms drawn from the annual
Forbes
magazine
rankings of the 500 largest U.S. public corporations based on sales, total assets,
market capitalization, and net income. I use a sample selection rule that requires
each company to qualify for any of these
Forbes
lists during at least four years of
the eight-year period between 1984 and 1991. I also require each company to have
four or more consecutive fiscal years of stock market and financial statement data
between 1984 and 1991. The four-year requirement represents an attempt to
balance two sampling issues: collecting several observations for each company so
that econometric panel data techniques can be used, and limiting survivorship
bias by allowing companies to enter and exit the panel over time. 1 omit utility and
financial companies because of concerns that government regulation leads to
different, more limited roles for their boards of directors. I obtain a final sample of
3,438 observations for 452 companies across eight years.’
’ Three sample firms began 1984 as public companies, remained public at least until 1988, were
delisted in going-private transactions, and became public again by 1991. In these cases, the data set
does not include the firms’ second incarnations. Approximately 20 firms changed the timing of their
fiscal years during the sample period; in these cases, ‘flow’ variables, such as sales, are normalized to
12-month equivalents for the transition fiscal years which were not 12 months long.
190
D. YermacklJoumal of Financial Economics 40 (1996) 185-21 I
Table 1 presents characteristics of the board of directors for the sample
observations, including mean and median values for key variables, and sample
correlations of other board attributes with board size. Board sizes range be-
tween four and 34 for sample firms, with a mean of 12.25 and a median of 12.
The sample correlations between board size and other variables do not give
a consistent indication of whether smaller boards should be expected to monitor
top managers more effectively than larger boards. While numerous studies have
found positive links between firm performance and the presence of independent
or expert board members, these types of directors are less likely to appear on
small boards. Board size is negatively correlated with the fraction of directors
who are corporate insiders or who have conflicts-of-interest due to their ‘gray’
status (signifying directors who are relatives of company officers and those who
benefit from personal business ties to the firm), and positively correlated with
the fraction of directors who serve as CEOs of other firms. Further, smaller
boards are more likely to have CEOs who either founded the company or
belong to the founding family, a quality that Johnson et al. (1985) found was
associated with low firm value. However, directors on small boards tend to have
greater levels of stock ownership and are more likely to receive performance-
based director fees in the form of stock options. Each of these characteristics
suggests that small boards may have better incentives to monitor, though
Merck, Shleifer, and Vishny (1988) and others have found that board equity
ownership has ambiguous associations with firm value. Small boards are also
likelier to include active monitors in the form of major stockholder-directors
and non-CEO chairmen. Smaller boards have lower rates of director turnover
than other boards, probably due to the better performance of companies with
small boards documented herein.
From the evidence in Table 1, I conclude that number of directors is but one
of many board attributes that might contribute to firm value, and that the
complex associations between board size and other variables do not suggest
clearly whether firms with small boards should have high or low market values.
3.2. Regression analysis
To investigate whether board size has a significant association with firm
value, I estimate least-squares regressions, using Tobin’s Q as the dependent
variable and board size as one of many explanatory variables. The dependent
variable, measured for each company at the close of each fiscal year ending in
calendar 1984 through 1991, is defined as
Tobin’s Q =
Market value of assets
Replacement cost of assets ’
D. YermackjJoumal ofFinancial Economics 40 (1996) 185-211
191
Table 1
Board of directors characteristics
Descriptive statistics for characteristics of boards of directors. The sample consists of 3,438 annual
observations for 452 companies between 1984 and 1991. Companies are included in the sample if
they are ranked by
Forbes
magazine as one of the 500 largest U.S. public corporations at least four
times during the eight-year sample period. Utility and financial companies are excluded. The table
presents the mean, median, and standard deviation for each variable, as well as Pearson sample
correlation coefficients between the board-size variable and all others.
Board size represents the number of members of the board of directors as of the annual meeting date
during each fiscal year. The percentage of inside directors is the fraction of board members who are
current or former officers of each company. Gray directors are those who have substantial business
relationships with the company, either personally or through their main employers, and also
relatives of corporate officers. Outside directors are those who have neither inside nor gray status.
Director fees include annual retainers and fees paid for regular and special board meetings during
the fiscal year. The dummy variable for director stock option plan equals one if the company has
a plan in place for awarding stock options to outside directors. Director turnover is the fraction of
board members who leave before the next annual meeting. The CEO-from-founding-family dummy
variable equals I if the CEO is from a family which either founded the company or acquired control
during a takeover. The dummy variable for 5% stockholder-directors equals 1 if one or more
members of the board beneficially own at least 5% of the common stock or serve as representatives
of an outside 5% holder (not including employee stock ownership plans).
Mean
Median
Board
size
Board composition
Inside directors
Gray directors
Outside directors
Directors who are CEOs of other firms
Gompensation and turnover
Director fees (1991 dollars)
Director stock option plan (dummy variable)
Director turnover 1% of board per year)
Governance structure and
stock
ownership
CEO from founding family (dummy variable)
Non-CEO chairman of board (dummy
variable)
Presence of 5% stockholder-director
(dummy variable)
Director and officer stock ownership
(% of common)
12.25 12
0.36 0.33
0.10 0.08
0.54 0.56
0.14 0.13
$29,539 $29,601
0.09 0
8.3% 7.1%
0.24 0
0.17 0
0.24 0
9.1%
2.8%
Correlation
with
Std. dev. board size
0.43
1.00
0.16
- 0.09
0.12
- 0.16
0.19
0.17
0.12
0.14
$10,657
0.33
0.29
- 0.09
10.2%
0.15
0.42
- 0.26
0.38
- 0.06
0.43
- 0.14
14.3%
- 0.27
Last column: All correlations with board size are significant at the 1% level:
192
D. Yemack/Journal
of
Financial Economics 40 (I 9%) 185.~ 21 I
I estimate the market value of assets by adding together estimated values of the
components of total liabilities and stockholders’ equity. The market value of
common stock is obtained directly from the CRSP database. I estimate the
market value of preferred stock by taking the ratio of preferred dividends over
the prevailing yield on Moody’s index of high-grade industrial preferred stocks.
The market value of long-term debt is estimated from a recursive algorithm that
infers the maturity structure of each firm’s debt and takes account of changes in
the prevailing yield on Moody’s index of A-rated industrial bonds. I assume
other liabilities have market value equal to book value. The replacement costs of
inventories and fixed assets are estimated by recursive algorithms that take
account of inflation, real depreciation rates, capital expenditures, and the
method of inventory valuation used by each company. Other assets are assumed
to have market value equal to book value. The recursive methods for valuing
debt, inventory, and fixed assets closely follow those of Perfect and Wiles (1994)
in their qpw estimator of Tobin’s Q.’
Fig. 1 illustrates mean and median values of Tobin’s Q for companies sorted
by board size. The number of directors for each company was obtained from
proxy statements for firms’ annual meetings, which usually occur in the fifth or
sixth month of each fiscal year. Mean and median Tobin’s Q values decline
almost monotonically over the range of board sizes. For companies with
between four and eight directors, mean Q values range between 1.5 and 2, while
the mean Q value falls to slightly above 1 for companies with 20 or more
directors.
In addition to board size, my regressions include controls for other variables
that I expect either to affect Tobin’s Q directly or to affect each board’s
incentives and ability to monitor managers.
A company’s profitability has a significant impact upon its market value, so
I include return on assets (ROA) in the regression model as an explanatory
variable. I calculate
ROA
as operating income divided by total assets (measured
at the start of each year) and compound the ratio continuously. The regression
model includes
ROA
for the most recent year and two years of lagged values.
In addition to current and past profitability, many theorists including Myers
(1977) and Smith and Watts (1992) argue that firm value depends on future
investment opportunities. Like others, I use the ratio of capital expenditures
over sales as a proxy for investment opportunities. Below, I consider whether
other possible measures of investment opportunities lead to differences in the
model’s estimates.
During the 1984-91 period of this study, diversified firms were valued less
highly in the capital markets than stand-alone businesses, as shown by Lang and
ZThe lone difference between my methodology and the
q
pw
estimator of Perfect and Wiles (1994) is
that my estimate of the replacement cost of property, plant, and equipment uses a slightly different
method for estimating real rates of economic depreciation and cost-reducing technical progress.
D. YermacklJoumal of Financial Economics 40 (1996) 185-211
193
2.5 5-
2.0
-.__
-_._._. --’
piizi-
0.5 -
0.0
I I t I1 ,/,,,I,,, [ ,,/
4 5
6 7 8 9
10 I1 12 13 14 15 16 17 18 19 20 21+
Board size
Fig. 1. Board size and Tobin’s Q: Sample means and medians.
Sample means and medians of Tobin’s Q for different sizes of boards of directors. The sample
consists of 3,438 annual observations for 452 firms between 1984 and 1991. Companies are included
in the sample if they are ranked by
Forbes
magazine as one of the 500 largest U.S. public
corporations at least four times during the eight-year sample period. Utility and financial companies
are excluded. Data for board size is gathered from proxy statements filed by companies near the start
of each fiscal year. Tobin’s Q is estimated at the end of each fiscal year as
Market value oj
assets/Replacement
cost of
assets.
The estimation of Q follows the ypw specification of Perfect and
Wiles (1994), which is described more fully in the text.
Stulz (1994) and Berger and Ofek (1995). Moreover, diversified companies are
likely to have larger boards, because many boards grow in size when companies
make acquisitions and because boards of conglomerates may seek outside
expertise for a greater number of industries. To control for diversification,
I include a variable that counts the number of business segments for which firms
report audited financial statement data in each fiscal year’s annual report.
As discussed in Section 2, many investigators have suggested that boards with
high stock ownership and a majority of outside directors monitor managers
more effectively. I include measures of these two variables in the regression
model: the percentage of common stock owned by directors and officers and the
percentage of outside directors on each company’s board. Outside directors
exclude current and former officers of the firm, and nonemployee directors who
194
D. Yermack/Journal of Financial Economics 40 (19%) IX5 III
have personal or business relationships with the company. These directors, often
referred to as ‘gray’, include such groups as lawyers, bankers, consultants, major
suppliers, and relatives of corporate officers.
I control for firm size with the log of total capital, measured in millions of
1991 dollars. Total capital equals the market value of equity at the end of the
year, plus the estimated values of long-term debt and preferred stock, calculated
as described above. Below, I consider alternative measures of company size in
Section 3.3.2.
Finally, I include in the regressions dummy variables for individual years and
two-digit SIC industries. I use a log specification for the board-size variable,
based upon the convex association between board size and market value
suggested by Fig. 1. I obtained the financial statement data used in regressions
from Compustat, except for a handful of observations for which data were hand-
collected. Data for board size, board composition, and inside stock ownership
were obtained from annual meeting proxy statements.
Because unobservable characteristics are likely to affect each company’s
market value, I estimate both ordinary least squares (OLS) regressions and
fixed-effects models. The OLS model includes two-digit SIC dummy variables
that allow a different intercept for firms in each industry, while the fixed-effects
estimator assigns a unique intercept to each company. Hausman and Taylor
(1981) state that the fixed-effects framework represents a common, unbiased
method of controlling for omitted variables in a panel data set. Table 2 presents
coefficient estimates for the OLS and fixed-effects models, with White (1980)
robust standard errors accompanying the OLS estimates.
The regression estimates for both models show an inverse and significant
association between firm value and board size. This downward slope is consis-
tent with an interpretation that coordination, communication, and decision-
making problems increasingly hinder board performance when the number of
directors increases. Further, the convex relation implied by the log form of the
board-size variable suggests that costs accumulate at a decreasing rate as board
size grows. A convex relation also emerges from estimates of different functional
forms, including piecewise linear models and regressions of Tobin’s Q against
board size and board size squared. For clarity, in the remainder of the paper,
I concentrate on results using the board-size log, and I generally report fixed-
effects estimates to control for unobservable company characteristics.
The fixed-effects estimate for the board-size log coefficient of - 0.337
implies that Tobin’s Q falls by about 0.23 if board size doubles and by about
0.13 if board size rises 50%. Expanding an eight-person board by one member
implies a reduction in Q of about 0.04, while adding one director to a 15-
person board implies a fall in Q of about 0.02. These changes in walue are
economicallysignificant. Since the median firm in my sample has a market
value of just under $2.6 billion (equity and long-term debt, in 1991 dollars),
and since most firms’ values of Tobin’s Q are close to one (see Fig. l), a change
D. Yermack/Journal of Financial Economics 40 (1996) 185-211
195
Table 2
Regression coefficient estimates: Board size and market valuation
Regression coefficient estimates of the association between Tobin’s Q and the number of directors
sitting on company boards. The sample consists of 3,438 annual observations for 452 firms
between 1984 and 1991. Companies are included in the sample if they are ranked by
Forbes
magazine as one of the 500 largest U.S. public corporations at least four times during the
eight-year sample period. Utility and financial companies are excluded. The dependent variable is
an estimate of Tobin’s Q at the end of each fiscal year. The log of board size is the natural log of the
number of directors sitting on each company’s board as of the annual meeting date each year, as
reported in company proxy statements. The first column presents OLS estimates with White
(1980) robust standard errors. The second column presents estimates from a fixed-effects model,
which assigns a unique intercept to each company.
The model includes control variables for company performance (return on assets in the current
year and two lags), firm size (the log of total capital, in 1991 dollars), growth opportunities (capital
expenditures over sales), diversification (the number of business segments for which financial
statement data is reported), board composition (the percentage of outside directors on the board),
and inside stock ownership (director and officer beneficial ownership, in percent).
ROA
equals
operating income over total assets (start of year) and is compounded continuously. Total capital
equals the market value of common stock at the end of the year, plus estimates of the market
values of long-term debt and preferred stock. Both models include dummy variables for years, and
the OLS model includes two-digit SIC industry dummy variables.
Dependent variable:
Tobin’s Q
Variable
OLS
estimates
Fixed-effects
estimates
Log of board size
Return on assets (current year)
Return on assets (prior year)
Return on assets (two years prior)
Firm size’(log of total capital)
Capital expenditures/SaIes
Number of business segments
Board composition (% outside directors)
Officer and director stock ownership (%)
Sample size
F-statistic
(P-value)
R-squared
Significant at 1% (***), 5% (**), and 10% (*) levels.
- 0.428*** - 0.337***
(0.043) (0.056)
3.856***
2.048***
(0.403) (0.147)
0.502
- 0.093
(0.536) (0.151)
1.039** 0.450***
(0.45 1) (0.130)
0.119*** 0.413***
(0.012) (0.020)
-0.116
0.176
(0.209)
(0.123)
- 0.042*** - 0.049***
(0.006)
(0.009)
- 0.213*** 0.172*
(0.067) (0.088)
0.279*** 0.310***
(0.096) (0.108)
3,400 3,400
68.1 2.7
(0.00) (0.00)
0.5459 0.3021
196
D. YmnackfJournal of Financial Economics 40 (1’861 1X5--21 I
in Q of 0.01 reduces firm value by about l%, or approximately $25 million for
the median firm.
While the results suggest a monotonic relation between smaller board size
and higher firm value, we should be cautious about concluding that the associ-
ation holds at very small levels of board size. This conjecture is difficult to test,
since few companies in my sample have board size below six (only 87 observa-
tions out of 3,438), and every board has at least four members. Fig. l’s display of
mean and median values of Tobin’s Q suggests that no consistent association
between board size and firm value exists over the lowest range of board sizes, as
the values of Q decline steadily only after board size grows beyond seven.
Re-estimating Table 2’s OLS regression for the subset of observations with
board size of seven or less yields a negative but insignificant estimate for the
board-size log coefficient.
Coefficient estimates for other variables in Table 2 are generally significant in
the expected direction. Current and past levels of profitability, measured by
return on assets, have positive associations with Tobin’s Q. Diversified firms
appear to be valued less highly than other companies. Board stock ownership
has a positive association with firm value. The effect of the board composition
variable is ambiguous and appears sensitive to the inclusion of firm effects in the
model. Capital expenditures over sales, the variable I use to measure investment
opportunities, does not have significant coefficient estimates, though in the
fixed-effects model the estimate is positive as expected, with a p-value of 0.15.
Firm size, when measured by the log of the market value of total capital, is
positively associated by construction with Tobin’s Q. Other measures of firm
size are discussed below.
The finding of an inverse association between board size and firm value
appears insensitive to the method of estimation. A between-firms estimator,
which considers only cross-sectional averages of the variables across companies,
produces an estimate of - 0.396 for the board-size log (significant at the 1%
level). A random-effects model, which represents a minimum-variance weighted
average of the within- and between-firms estimators (Hausman and Taylor,
198 l), yields an estimate of - 0.450 for the board-size coefficient @-value below
1%). An OLS estimate based on first differences of the variables produces an
estimate of - 0.239 for the board-size log (p-value below 5%). Year-by-year
cross-sectional estimates of the OLS model in Table 2 yield negative and
significant coefficients for the board-size log variable in every year. While the
estimate is closest to zero in 1990 and 1991, it is difficult to spot a time trend, as
the two most negative estimates occur for 1987 and 1989.
3.3. Further controls
Several plausible explanations could account for the negative association
between board size and firm value. Regression results in Table 2 are consistent
D. YermacklJournal of Financial Economics 40 (1996) 185-211
197
with the hypotheses of Lipton and Lorsch (1992) and Jensen (1993), that small
boards operate more effectively, but board size could be associated with other
corporate attributes that affect firm value. Moreover, my results imply, some-
what counterintuitively, that many companies have bypassed a simple, inexpen-
sive way of improving corporate performance. For these reasons, I conduct
additional tests of the robustness of my basic finding and discuss the results in
the following sections.
3.3. I. Growth opportunities
Tobin’s Q is an ambiguous measure of value-added by management, since the
Q- ratio can also capture the value of future investment opportunities. While the
model controls for growth opportunities by using capital expenditures over
sales, Smith and Watts (1992) and other authors have used additional proxies,
including research and development (R&D) expense over sales, depreciation
expense over sales, the earnings-price ratio, and the variance of common stock
returns. I re-estimate the fixed-effects model of Table 2 with each of these growth
opportunity measures substituted for capital expenditures over sales, but the
key coefficient on the board-size log exhibits virtually no change. Board size
itself does not appear to have any systematic association with the presence of
growth opportunities; regressions of the five growth-opportunity variables
against the board-size log yield only one statistically significant estimate and no
consistent pattern of signs.
3.3.2. Firm size
Company size, board size, and firm value can be correlated in complicated
ways, so 1 check the robustness of my results to different definitions and
functional forms of the firm size variable. I estimate a total of 12 fixed-effects
models, based upon three different measures of size: total capital (market value
of equity and long-term debt), total assets (book value), and net sales during the
prior fiscal year. I use four different specifications for each size variable: log
terms, linear terms, linear and squared terms, and linear, squared, and cubed
terms. The estimated coefficient for the board-size log is negative and significant
in every specification (nine estimates have p-values below l%, two have p-values
of 3%, and one has a p-value of 8%).
3.3.3. Active monitors
The correlations in Table 1 suggest that small boards have a higher incidence
of active monitors. I re-estimate the model after adding dummy variables for the
presence of a non-CEO chairman, a non-CEO company president who also sits
on the board, and a 5% stockholder-director. The fixed-effects estimate for the
board-size log remains virtually unchanged. Of the new variables, only the
dummy for non-CEO chairman has an interesting estimate: The coefficient is
198
D. YermacklJournal of Financial Economics 40 (1996) 185 -21 I
0.039 and significant at the 11% level, implying that firms are valued more
highly when the CEO and chairman positions are separated. Numerous alterna-
tive specifications of the stockholder-director variable do not change the main
conclusion.
3.3.4. Close ownership structures
Table 1 suggests that small boards are more common in companies controlled
by founding families. These firms could have tighter ownership structures and
turn over assets more slowly, leading to low book values of assets, which imply
high values of Tobin’s Q. I add to the model a dummy variable equal to one if
the CEO belongs to the company’s founding family, but the fixed-effects esti-
mate for the board-size log exhibits little change. The founding-family;variable
has a negative and significant estimate of - 0.11, suggesting that firm valde
increases once the founding family surrenders control. Company-age could also
be closely associated with ownership structure, since firms probably become
more widely held over time. I re-estimate the model with a variable equal to the
number of years since each firm’s original incorporation; again, the board-size
estimate remains nearly the same as in Table 2.
4. Past performance and current board size
The analysis above shows that companies with small boards of directors
attain higher values in the capital markets than do their counterparts with large
boards. However, we might interpret these findings in two ways: Small boards
could contribute to better performance, or companies might adjust board sizein
response to past performance. If companies expand their boards in the after-
math of poor performance, the causation of the board size-firm value relation
may run in the opposite direction from the Lipton-Lorsch (1992) and Jensen
(1993) hypotheses.
Prior studies by Hermalin and Weisbach (1988) and Gilson (1990) examine
the interplay between company performance and changesinboards of directors.
Hermalin and Weisbach (1988) find that poor performaglce lead&to both more
departures of board members and more appointments to the board. While
Hermalin and Weisbach are silent on the net ieffect.’ of these two forces, they
estimate similar magnitudes for each, suggesting&at director turnover increases
after poor performance, but board size does not. Gilson (MO), in a study limited
to financially distressed companies, also finds that boarukrtatuwiwer increases after
poor performance; moreover, Gilson finds that firms reduce board sizes during
distressed periods. If this pattern held for all companies, one would observe
smaller boards in low-valued companies, contrary to the fmdihgs-of this paper.
However, it is not clear that Gilson’s results apply to the majority of companies,
since the median firm in his sample is small and performs very poorly.
D. YermacklJoumal of Financial Economics 40 (1996) 185-211
199
I examine the question of causation by estimating regression models. of the
association between past performance and changes in board size. Following
Hermalin and Weisbach (1988), I estimate maximum-likelihood Poisson models
of the number of directors leaving and joining each company’s board each year.
Table 3
Regression coefficient estimates: Effect of company performance on director appointments, depar-
tures, and changes in board size
Coefficient estimates for regression models of changes in board size. The first two columns present
maximum-likelihood Poisson models of the number of directors joining and leaving each company’s
board. The third column presents OLS estimates of the net change in board size, equal to director
additions minus director departures. All variables are measured annually, based upon the member-
ship of companies’ boards as reported in proxy statements for annual shareholder meetings.
The sample consists of 3,438 annual observations for 452 firms between P&land 1991. Companies
are included in the sample if they are ranked by
Forbes
magazine as one of the*500 largest U.S. public
corporations at least four times during the eight-year sample period. Utility and financial companies
are excluded. The key explanatory variable for all three-models is the firm’s abnormal stock return
during the fiscal year, defined as the raw return minus the return predicted by the CAPM. Other
explanatory variables are similar to those used by Hermalin and Weisbach (19@0. CEO at
retirement age is a dummy variable equal to one if the company’s CEO is between the ages of 62 and
66. New CEO is a dummy equal to one if the CEO has four years of tenure or less. All models include
two-digit SIC industry dummy variables. The text discusses assumptions used in calculating
abnormal stock returns. Each coefficient estimate appears with robust standard errors.
Dependent variable
Director
appointments
Director
departures
Change in
board size
Estimation Poisson ML
Estimate
Poisson ML
Estimate
OLS
Estimate
Abnormal stock return
(fiscal year)
Abnormal stock return
(prior fiscal year)
Change in firm size
[log(sales)]
CEO at retirement age
(dummy for ages 62 to 66)
New CEO
(dummy for 4 years’ tenure or less)
SIC industry dummies
Sample size
F-statistic
(P-value)
R-squared
- 0.238**
(0.093)
- 0.087
(0.088)
0.419***
(0.163)
0.129**
(0.054)
0.148***
(0.048)
2-digit
2,943
- 0.238***
(0.084)
- 0.126
(0.084)
- 0.347*
(0.185)
0.133**
(0.055)
0.273***
(0.050)
2-digit
2,943
- 0.005
(0.106)
0.039
(0.091)
0.770***
(0.198)
- 0.015
(0.062)
- 0.156***
(0.055)
2-digit
2,943
1.9
(0.00)
0.0313
Significant at 1% (***), 5% (**), and 10% (*) levels.
200 D. Yermnck/Journal of Financial Economics 40 (I 996) I&--21 I
As a further check, I estimate a least squares model in which the dependent
variable is equal to the total annual change in board size (director additions
minus departures). Key explanatory variables for all models are abnormal stock
returns, the change in company size (measured by the log of net sales), and
dummy variables for whether the company’s CEO is nearing retirement (be-
tween the ages of 62 and 66) or new (appointed within the last four years).
Abnormal stock returns equal the firm’s raw stock return during the fiscal year,
minus the return predicted by the capital asset pricing model (CAPM). The
CAPM calculations use fi estimates calculated over the last 120 trading days of
the prior fiscal year, and risk-free rates equal to the yield on ten-year U.S.
Treasury bonds.
Table 3 displays the results. The models provide no evidence that boards
either expand or contract in response to performance. Like Hermalin and
Weisbach, I find that poor performance is associated with higher levels of both
director appointments and departures. The effects are similar, suggesting that
more directors are replaced when companies perform poorly, but that total
board size does not change. The same conclusion emerges from the OLS model
of the net annual change in board size. I obtain qualitatively similar results from
re-estimating all three equations, using seven-year aggregate totals of all vari-
ables between 1984 and 1991 for each company. The OLS model does show that
boards tend to grow larger in response to changes in company size, as opposed
to performance, but the effect appears weak; a firm almost needs to quadruple in
size before it can be expected to add one director.
I conduct additional tests of whether the link between board size and firm
value can be attributed to adjustments in board size due to past performance.
I re-estimate the basic model of firm value and board size in an instrumental
variables framework, using lagged values of the board-size log as instruments for
the current value. The new estimates are virtually indistinguishable from the
original model’s OLS estimate in Table 2. I also regress Tobin’s Q against long
lags of board size and compare the results with regressions of current levels of
board size against long lags of Tobin’s Q. I use a fixed-effects framework similar
to that in Table 2. I find that up to three years’ lagged values of board size have
significant associations with subsequent values of Tobin’s Q, while no corres-
ponding association exists between lagged vaiues of Tobin’s Q and subsequent
values of board size. I conclude that the evidence supports the interpretation
that past board size influences current firm value, rather than the opposite - that
past performance influences current choices of board size.
5. Additional evidence of small boards’ effectiveness
Further evidence about the performance of small boards is reflected in
patterns of company operating performance, CEO turnover, and executive
D. YermacklJournal of Financial Economics 40 (1996) 185-21 I
201
compensation, as well as shareholder reactions to board size changes. The
analysis in the following sections illustrates that key financial ratios exhibit an
inverse association with board size, and that CEO incentives - from compensa-
tion and the threat of dismissal - diminish in strength as board size increases.
I also identify six sample companies that announce significant reductions in
board size for corporate governance reasons. These firms realize positive abnor-
mal stock returns around the announcement date, while a parallel sample of
four companies announcing large expansions in board size realize negative
abnormal returns.
5.1. Board size and jinancial ratios
If corporate governance becomes less effective as board size increases, I expect
lower profitability in companies with large boards, and I also expect less efficient
use of assets. I estimate fixed-effects models of board size and three key financial
ratios: sales over assets, return on assets, and return on sales. I define return on
assets as operating income over total assets at the start of the year, and return on
sales as operating income over net sales, and compound both ratios continu-
ously. I regress all three variables against the board-size log and control
variables similar to those in the model for Tobin’s Q: firm size (the log of total
capital), board stock ownership, board composition, the number of business
segments, and dummy variables for individual years. I do not include controls
for investment opportunities, since it is not obvious why their presence should
influence current operating performance.
Table 4 presents the fixed-effects estimates. Consistent with the finding for
Tobin’s Q, all three dependent variables have negative and significant associ-
ations with the board-size log. Companies with large boards appear to use assets
less efficiently and earn lower profits.
However, the favorable evidence of an association between board size and
financial ratios does not extend to models using less aggregate measures of
profitability and efficiency. I estimate three further regressions that have de-
pendent variables equal to sales per employee; the cost of goods sold over sales;
and selling, general, and administrative expenses over sales. The board-size log
coefficient is not significant in any of these models when they are estimated in
a fixed-effects framework.
5.2. Board size and CEO turnover
Selecting, evaluating, and dismissing a company’s top managers represents
a central responsibility of boards of directors. If large board size contributes to
behavioral norms that inhibit candid discussions of managerial performance, as
argued by Lipton and Lorsch (1992) and Jensen (1993), we should expect weaker
links between performance and turnover for CEOs in companies with large boards.
202
D. YermackjJournal cf Financial Economics 40 (1996) IX5 21 I
Table 4
Fixed-effects estimates: Board size and financial ratios
Coefficient estimates for fixed-effects regression models of financial ratios. Standard errors appears
below each estimate. The sample consists of 3,438 annual observations for 452 firms between 1984
and 1991. Companies are included in the sample if they are ranked by Forbes magazine as one of the
500 largest U.S. public corporations at least four times during the eight-year sample period. Utility
and financial companies are excluded.
The dependent variables are three standard measures of operating efficiency and profitability. ROS
and ROA are based on operating income and are compounded continuously. The main explanatory
variable for each model is the natural log of board size, which represents the number of directors
sitting on each company’s board as of the annual meeting for each fiscal year. In addition to the
other controls listed, each model includes dummy variables for individual years.
Dependent variable
Sale/
assets
Estimate
Return on
assets
Estimate
Return on
sales
Estimate
Log of board size - 0.163**
(0.079)
Board composition (% outside directors) - 0.187
(0.126)
Officer and director stock ownership (%) 0.287*
(0.153)
Number of business segments 0.006
(0.013)
Firm size (log of total capital) 0.083***
(0.026)
Sample size
3,428
F-statistic 23.6
(P-value)
(0.W
R-squared 0.0647
Significant at 1% (***), 5% (**), and 10% (*) levels.
- 0.037*** - 0.025***
(0.008)
(0.006)
0.005 - 0.009
(0.012)
(0.009)
0.022 - 0.029***
(0.015) (0.011)
- 0.0026** - 0.0020**
(0.0013) (O.ooo9)
0.045*** 0.031***
(0.003) (0.002)
3,425
3,425
66.3 44.3
(0.W (0.W
0.1627 0.1148
I estimate a probit model of CEO turnover, similar to the model of Warner,
Watts, and Wruck.(l988). The (0, 1) dependent variable equals one if the CEO
leaves his position before the end of the current fiscal year or during the first half
of the subsequent:‘fiscal year. The main explanatory variable is the firm’s
cumulative abnormal stock return over the current fiscal year and two prior
years. The use of a relatively long interval for the abnormal stock return is
consistent with other studies. For example, Warner et al. (1988) use four years of
performance data in their turnover model. I construct the abnormal stock
returns described in Section 4. I include controls for CEO age (which should be
positively related to turnover) and CEO stock ownership (which should aI%&
turnover negatively, if high stock ownership represents a form of CEO entrench-
ment). I also include dummy variables for industries and CEO ages 44,65, and
D. YermacklJournal @Financial Economics 40 (19%) IKS- 21 I
203
Table 5
Probit coefficient estimates: Board size and CEO dismissal incentives
Regression coefficient estimates for a binary probit model of CEO turnover. The dependent variable
equals one if a CEO leaves his position during the second half of the current fiscal year or the first
half of the subsequent fiscal year. The main explanatory variable is the firm’s cumulative abnormal
stock return for the current year and two prior years. The abnormal stock return is defined as the
raw stock return minus the return predicted by the CAPM. The sample consists of 3,438 annual
observations for 452 firms between 1984 and 1991. Companies are included in the sample if they are
ranked by Forbes magazine as one of the 500 largest U.S. public corporations at least four times
during the eight-year sample period. Utility and financial companies are excluded. Coefficient
estimates appear with White (1982) robust standard errors. Missing values occur for companies that
do not have a three-year history of common stock returns.
To illustrate how CEO incentives from the threat of dismissal are affected by different characteristics
of boards of directors, the second column presents estimates for a model that includes an interaction
term between the abnormal stock return and the log of board size. Both models also include
variables for the fraction of common stock owned by the CEO, CEO age, dummy variables for CEO
ages 64, 65, and 66, and dummies for one-digit SIC industries.
Dependent variable: CEO leaves
position (0, I)
Estimate
Estimate
CEO age
CEO stock ownership (%)
Cumulative abnormal stock return
(current year and two prior years)
Interaction term:
Abnormal stock return x log (board size)
Sample size
0.043***
(0.006)
- 2.022***
(0.755)
- 0.358***
(0.071)
3,305
0.043***
(0.006)
- 2.027***
(0.746)
- 1.496***
(0.581)
0.471**
(0.239)
3,305
Significant at 1% (***). 5% (**), and 10% (*) levels.
66, since much of the CEO turnover at these ages represents regular planned
retirements. Table 5 presents coefficient estimates with White (1982) robust
standard errors. In the first column, variables for abnormal stock performance,
CEO age, and CEO stock ownership all have expected signs with high signifi-
cance.
To test for the importance of board size in the link between performance and
CEO turnover, I add to the model an interaction term equal to the abnormal
stock return times the board-size log. As shown in the second column of Table 5,
the augmented model has a more negative estimated association between
performance and turnover, coupled with a positive and significant estimate for
the interaction term. I interpret these estimates as evidence that smaller boards
are more likely to dismiss CEOs for poor performance, and that this threat of
dismissal declines as board size increases.
204
D. YermacklJournal of Financial Economics 40 (1996) 185~ 21 I
A sampling issue arises when considering whether to exclude CEOs who have
served less than three years, since the main explanatory variable is a three-year
abnormal stock return. Even if part of this return occurred under a prior CEO, it
is likely that boards will hold new CEOs responsible for several years of past
performance, since most are promoted from within the firm. Deleting from the
sample CEOs in their first and second years (19% of the observations) only
enhances the strength of the conclusions. The estimates change little, and the
t-statistics increase for both the abnormal stock return and its interaction with
the board-size log (results not reported).
The analysis of CEO turnover in Weisbach (1988) shows that boards domin-
ated by outside directors are more likely to dismiss CEOs for underperfor-
mance. I control for this effect by adding an interaction term between the
abnormal stock return and the fraction of outside directors on each board. This
variable’s estimate is insignificant. I obtain similar results for an interaction term
between board stock ownership (excluding the CEO’s shares) and the abnormal
stock return. When all three interaction terms appear in the same model, only
board size has a significant association with the strength of CEO dismissal
incentives; when board size is dropped and the other two interaction terms
remain in the model, neither is significant. I conclude that CEO dismissal
incentives weaken as board size increases, and I am unable to reach similar
conclusions about the importance of the other board attributes.
5.3. Board size and CEO compensation incentives
Establishing managerial incentives through compensation contracts repre-
sents another important board responsibility. I study the interplay between
CEOs’ compensation incentives and board size to see whether small boards
perform this task more effectively. I use Jensen and Murphy’s (1990) model for
studying the pay-performance sensitivity of CEO compensation. The authors
define pay-performance sensitivity as the dollar change in CEO compensation
per dollar change in stockholder wealth; they estimate this coefficient with linear
regressions, using first differences of annual variables. Following their approach,
I regress one-year changes in CEO salary plus bonus (obtained from corporate
proxy statements) against one-year changes in stockholder wealth for each
firm-year in the sample.3 To capture the importance of board size in the
pay-performance relation, I repeat the approach followed in my model of CEO
3 All dollar values are adjusted for inflation. The change in stockholder wealth is calculated as the
return to common stockholders during the fiscal year times market capitalization at the start of the
year, all as reported by CRSP. For firms with more than one class ofcommon stock, I add changes in
the value of all classes. The analysis omits CEOs who do not serve 12-month fiscal years. Missing
values occur during CEO transition periods, since first differences cannot be calculated.
D. YermackjJournal ofFinancial Economics 40 (1996) 185-211
205
Table 6
OLS estimates: Board size and CEO compensation incentives
OLS estimates for a model of CEO performance incentives from compensation. The dependent
variable is the annual change in the CEO’s salary and bonus. The main explanatory variable is the
annual change in stockholder wealth. To illustrate how performance incentives are affected as board
size increases, the model in the right column includes an interaction term between the change in
stockholder wealth and the log of board size.
The sample consists of 3,438 annual observations for 452 firms between 1984 and 1991. Companies
are included in the sample if they are ranked by
Forbes
magazine as one of the 500 largest U.S. public
corporations at least four times during the eight-year sample period. Utility and financial companies
are excluded. Coefficient estimates appear with White (1980) robust standard errors. A large number
of missing values occur due to the first-differencing framework (which eliminates one year of data)
and episodes of CEO turnover (each of which eliminate two first-difference observations).
Dependent variable:
Change in CEO’s salary + bonus
Estimate Estimate
Constant
Change in stockholder wealth (per $1,000)
Interaction term:
Change in stockholder wealth x log (board size)
Sample size
F-statistic
(P-value)
R-squared
39,348***
(I 1,842)
0.015**
(0.007)
2,412
8.0
(0.00)
0.0033
35,668***
(12,750)
0.139**
(0.065)
- 0.045**
(0.023)
2,412
6.6
(0.00)
0.0055
Significant at 1% (***), 5% (**), and 10% (*) levels.
turnover, adding to the model interaction terms between board size and the
change in stockholder wealth.
Table 6 presents the results. In the first column, I estimate the pay-perfor-
mance sensitivity of CEO salaries and bonuses as $0.015 per $1,000 change in
stockholder wealth, which is quite close to Jensen and Murphy’s (1990) estimate.
The second column in Table 5 presents estimates for a model that includes an
interaction term between the change in stockholder wealth and the board-size
log. The negative and significant coefficient for this interaction term, coupled
with the increased estimate for the stockholder wealth coefficient, suggests that
CEOs receive stronger compensation incentives in companies with smaller
boards. The results change little when the analysis is repeated using abnormal
changes in stockholder wealth, calculated with the CAPM assumptions used
above.
I add to the model interaction terms between the change in stockholder
wealth/percentage of outside directors and the change in stockholder
Companies decreasing
board size
Number of directors % outside directors
Before After Before After
Table 7
Reactions to announcements of significant board size changes
Investor reactions to announcements by ten sample companies of significant changes in board size. Companies were identified by searches of the Nexis
database and the Wall Street Journal Index. The analysis includes firms that announced changes in board size of at least four members on an event date in
1984 or later. Events are excluded if board size changes were related to mergers, CEO transitions, or corporate control contests.
Abnormal stock returns around the announcement dates are calculated by three methods: raw returns minus market returns, raw returns minus market
model expected returns, and excess returns extracted from the CRSP database. The market return used in calculations is the CRSP value-weighted index,
and market model parameters are estimated from one year of daily trading data prior to the event period. T-statistics for market model abnormal returns
are based on standardized prediction errors (Dodd and Warner, 1983).
Abnormal stock returns
1 day before to 1 day after announcement
Net-of-
Market CRSP
market model excess return
Chrysler (1991)
Control Data (1985)
Adolph Coors (1988)
W.R. Grace (1991)
P&in-Elmer (1991)
Time Warner (1992)
Mean
T-statistic
P-value
20 13 70% 77% 5.1%
5.4%
5.5%
18 14 61% 36% 5.9%
1.2% 6.40/o
10
5 0% 0%
0.6% 0.7% “.a.
33
26
36% 54% - 0.7%
- 1.8%)
0.3%
13 9 85% 89% 2.1%
2.1% I .6%
21
15
48%
67% 0.7% 0.2% 0.5%
--.
2.3%* 2.3%
2.9O/<,*
2.10
1.61 2.24
0.09
0.17 0.09
Companies increasing
board size
Cummins Engine (1985)
Lafarge (1985)
Squibb (1984)
Xerox (1987)
Number of directors
Before
After
14 18
10 16
17
22
14
18
% outside directors
Before After
.~~
50% SO%
30% 50%
35% 45%
50% 50%
Net-of-
market
~ 1.1%
- 1.9%
- 0.5%
- 1.5%
Market
model
- 0.1%
- 7.5%
- 0.3%
~ 1.3%
Mean
- 2.8% - 2.5%
T-statistic
- 1.59
- 1.44
P-value
0.21 0.25
Difference in means
T-statistic
P-value
5.0%**
4.8%*
2.46 2.17
0.04 0.06
CRSP
excess return
~ 1.2%
- 7.0%
- 1.2%
P
- 1.7%
;;:
a
- 2.8%
f:
h
- 1.94
<
s
0.15
2
e
5.6%**
2.95
%
5’
0.02
is
Difference in means
T-statistic
P-value
Significant at 1% (***), 5% (**), and 10% (*) levels.
Abnormal stock returns
e
10 days before to 10 days after announcement 2
17.6%*** 17.70/a*** 14.9%**
:.
2
3.71 3.41 3.31
0.01 0.01 3 0.01
2
s
9
208
D. YermackiJournal oj’Financia1 Economics 40 (1996) IX5 21 I
wealth/level of board stock ownership. The additional interaction terms allow
me to check whether these board attributes also affect the level of CEO
incentives. Neither variable has a statistically significant estimate, and a likeli-
hood ratio test shows that they do not have joint significance.
5.4. Reactions to signijcant changes in board size
Investor reactions to changes in board size should provide additional evid-
ence of links between board size and market valuation. Within my 452 firms,
I analyze those companies that have changed board size by at least four
directors at one time since 1984. I base my choice (changes of four or more as the
cutoff point) upon the relative lack of observations for greater changes and my
subjective judgment that board-size changes of two or three members are not
substantial. From my candidate sample of 195 board-size changes of four or
more, I eliminate those changes occurring gradually during the fiscal year and
those for which no event date could be identified in press reports or proxy
statements. I also exclude changes made in the aftermath of mergers and
acquisitions and changes that occur around CEO transitions or as tactical
moves in a corporate control contest. The resulting sample for analysis, listed in
Table 7, includes ten cases of companies changing board size by four or more on
one date, explicitly or apparently for corporate governance reasons.4 Six of these
companies reduced board size by four or more members, while four companies
increased board size by at least four.
I study company stock returns around announcement dates of board-size
changes. I obtain abnormal stock returns by three methods: subtracting the
return for the CRSP market-wide index, subtracting market model expected
returns (with CI and /3 parameters estimated from one year of trading data prior
to the event period), and extracting excess daily returns from CRSP’s
NYSE/AMEX file. I use CRSP’s value-weighted, dividend-inclusive market
index as the market return. Table 7 presents the abnormal stock return in the
three right-hand columns, covering an event period beginning one day before
each announcement and ending one day after.
Investors appear to welcome decreases in board size and to disapprove of
board expansions. Abnormal stock returns are positive around five of six
4For two of the ten companies, simultaneous announcements of other news might contaminate the
events for purposes of study. When Chrysler reduced the size of its board in 1991, the change
attracted attention because of the exclusion from the new board of United Auto Workers President
Owen Bieber. Time Warner’s 1992 reduction in board size occurred within days of the death of
co-CEO Steven Ross. In both cases, the companies denied any connection between the reductions in
board size and the other news. It is not clear that either event should have affected the firms’ market
values, as Bieber’s role in Chrysler was described as marginal and Ross’s illness had led to
a well-publicized withdrawal from day-to-day management six months before.
D. YermacklJournal of Financial Economics 40 (1996) 185-211
209
announcements of board-size reductions and negative around all four an-
nouncements of board-size increases. The average magnitude of each effect is
approximately 2.5%. The difference in means between the two subgroups, which
equals approximately 5%, has statistical significance at levels between 2% and
6%, depending on which type of abnormal return is analyzed.
Inspection of the data suggests that some information leakage takes place
before public announcement dates, as abnormal daily returns cumulate gradu-
ally for approximately two weeks prior to announcements. If the event window
is widened to encompass ten trading days before until ten trading days after each
announcement, the results become dramatic, with a difference in mean abnor-
mal returns of approximately 15% (p-value below 1%).
Though this analysis involves a small sample, the ten observations show that
investors react favorably to significant reductions in board size and react
unfavorably to board expansions. It is not likely that the results reflect changes
in board composition; Table 7 shows that although three of the sample com-
panies that reduced their boards simultaneously increased the role of outside
directors (including Adolph Coors, according to its announcement), one company
did the opposite, and the remaining two essentially left board composition
unchanged. Two of four companies expanding their boards increased the fraction
of outside directors, but nevertheless experienced negative abnormal returns.
Despite shareholders’ apparently favorable reactions to announcements of
smaller board size, subsequent changes in firm value over long periods do not
provide further support for the hypothesis that smaller boards manage com-
panies better. I calculate average changes in the value of Tobin’s Q over one,
two, and three year periods following the announcements of board size changes
shown in Table 7. Average Q values exhibit a slight upward pattern for both the
six companies reducing their boards and the four firms that added directors.
Differences between the two groups are not statistically significant.
6. Summary and conclusions
This paper evaluates a recent proposal in the legal and finance literature for
reducing the size of corporate boards of directors. Lipton and Lorsch (1992) and
Jensen (1993) have criticized the performance of large boards, stating that
problems of poor communication and decision-making overwhelm the effec-
tiveness of such groups. I find evidence consistent with this theory. Using
a variety of regression models with data from 1984-91 for 452 large public
corporations, I find an inverse association between board size and firm value.
The association appears to have a convex shape, suggesting that the largest
fraction of lost value occurs as boards grow from small to medium size.
The basic result proves robust to a variety of controls for company size, the
presence of growth opportunities, and alternative corporate governance and
210
D. Yemnck~.Journal of Financial Economics 40 (1996) /KS 21 I
ownership structures. No evidence is consistent with conjectures that companies
change board size as a result of past performance.
A range of supporting evidence is consistent with the main finding of an
inverse association between board size and firm value. Financial ratios related
to profitability and operating efficiency appear to decline as board size grows.
CEO performance incentives provided by the board through compensation and
the threat of dismissal operate less strongly as board size increases. A small
group of sample companies that announce significant reductions in board size
realize substantial excess stock returns around the announcement dates, while
the opposite occurs for companies that announce board expansions.
References
Bagnani, Elizabeth Strock, Nikolaos T. Milonas, Anthony Saunders, and Nickolaos G. Travlos,
1994, Managers, owners, and the pricing of risky debt: An empirical analysis, Journal of Finance
49,453417.
Baysinger, Robert D. and Henry N. Butler, 1985, Corporate governance and the board of directors:
Performance effects of changes in board composition, Journal of Law, Economics and Organiza-
tion 1, 101-124.
Berger, Philip G. and Eli Ofek, 1995, Diversification’s effect on firm value, Journal of Financial
Economics 37, 39965.
Brickley, James A., Jeffrey L. Coles, and Rory L. Terry, 1994, Outside directors and the adoption of
poison pills, Journal of Financial Economics 35, 371-390.
Byrd, John W. and Kent A. Hickman, 1992, Do outside directors monitor managers?, Journal of
Financial Economics 32, 195-22 1.
Dodd, Peter and Jerold B. Warner, 1983, On corporate governance, Journal of Financial Economics
11,401438.
Gilson, Stuart, 1990, Bankruptcy, boards, banks, and blockholders: Evidence on changes in
corporate ownership and control when firms default, Journal of Financial Economics 27,
3555387.
Hackman, J. Richard, ed., 1990, Groups that work (Jossey-Bass, San Francisco, CA).
Hausman, Jerry A. and William E. Taylor, 1981, Pane1 data and unobservable individual effects,
Econometrica 49, 137771398.
Hermalin, Benjamin and Michael Weisbach, 1988, The determinants of board composition, Rand
Journal of Economics 19, 589-606.
Hermalin, Benjamin and Michael Weisbach, 1991, The effects of board composition and direct
incentives on firm performance, Financial Management 20, 101-l 12.
Holthausen, Robert W. and David F. Larcker, 1993a, Boards of directors, ownership structure and
CEO compensation, Unpublished manuscript (Wharton School, University of Pennsylvania,
Philadelphia, PA).
Holthausen, Robert W. and David F. Larcker, 1993b, Organizational structure and financial
performance, Unpublished manuscript (Wharton School, University of Pennsylvania, Philadel-
phia, PA).
Jensen, Michael C., 1993, The modern industrial revolution, exit, and the failure of internal control
systems, Journal of Finance 48, 831-880.
Jensen, Michael C. and Kevin J. Murphy, 1990,‘Performance pay and top-management incentives,
Journal of Political Economy 98. 2255264.
D. YermacklJoumal of Financial Economics 40 (1996) 185-211
211
Johnson, W. Bruce, Robert P. Magee, Nandu J. Nagarajan, and Harry A. Newman, 1985, An
analysis of the stock price reaction to sudden executive deaths, Journal of Accounting and
Economics 7, 15 l-l 74.
Kini, Omesh, William Kracaw, and Shehzad Mian, 1995, Corporate takeovers, firm performance,
and board composition, Journal of Corporate Finance 1, 3833412.
Lang, Larry H.P. and Rent: M. Stulz, 1994, Tobin’s Q, corporate diversification, and firm perfor-
mance, Journal of Political Economy 102, 124881280.
Lipton, Martin and Jay W. Lorsch, 1992, A modest proposal for improved corporate governance,
Business Lawyer 48, no. I, 59- 77.
McConnell, John J. and Henri Servaes, 1990, Additional evidence on equity ownership and
corporate value, Journal of Financial Economics 27, 5955612.
Monks, Robert A.G. and Nell Minow, 1995, Corporate governance (Basil Blackwell, Cambridge).
Msrck, Randall, Andrei Shleifer, and Robert Vishny, 1988, Management ownership and market
valuation: An empirical analysis, Journal of Financial Economics 20, 293-315.
Myers, Stewart C., 1977, Determinants of corporate borrowing, Journal of Financial Economics 5,
147.-175.
Perfect, Steven B. and Kenneth W. Wiles, 1994, Alternative constructions of Tobin’s Q: An empirical
comparison, Journal of Empirical Finance 1, 3 13-341.
Rosenstein, Stuart and Jetfrey G. Wyatt, 1990, Outside directors, board independence, and share-
holder wealth, Journal of Financial Economics 26, 1755191.
Smith, Clifford W., Jr. and Ross L. Watts, 1992, The investment opportunity set and corporate
financing, dividend, and compensation policies, Journal of Financial Economics 32, 263-292.
Steiner, I.D., 1972, Group process and productivity (Academic Press, New York, NY).
Warner, Jerold B., Ross L. Watts, and Karen H. Wruck, 1988, Stock prices and top management
changes, Journal of Financial Economics 20, 461492.
Weisbach, Michael, 1988, Outside directors and CEO turnover, Journal of Financial Economics 20,
43 I-460.
White, Halbert, 1980, A heteroskedasticity-consistent covariance matrix estimator and a direct test
for heteroskedasticity, Econometrica 48, 817-838.
White, Halbert, 1982, Maximum likelihood estimation of misspecified models, Econometrica 50,
I -26.
