Journal of Financial Economics: K.J. Martijn Cremers, Lubomir P. Litov, Simone M. Sepe

Journal of Financial Economics 126 (2017) 422–444
Contents lists available at ScienceDirect
Journal of Financial Economics

journal homepage: www.elsevier.com/locate/jfec
Staggered boards and long-term firm value, revisitedR,RR

K.J. Martijn Cremers a,∗, Lubomir P. Litov b, Simone M. Sepe c,d
a
University of Notre Dame, 264 Mendoza College of Business, Notre Dame, IN 46556, United States
b
Price College of Business, University of Oklahoma, Norman, OK 73069, United States
c
University of Arizona James E. Rogers College of Law 1201 E. Speedway Tucson, Arizona 85721, United States
d
Institute for Advanced Study in Toulouse, Fondation Jean-Jacques Laffont, Toulouse School of Economics, 21 Allée de Brienne, 31015
Toulouse Cedex 6, France
a r t i c l e i n f o a b s t r a c t
Article history: This paper revisits the staggered board debate focusing on the long-term association of
Received 9 March 2016 firm value with changes in board structure. We find no evidence that staggered board
Revised 13 October 2016
changes are negatively related to firm value. However, we find a positive relation for firms
Accepted 24 October 2016
engaged in innovation and where stakeholder relationships matter more. This suggests that
Available online 18 August 2017
staggered boards promote value creation for some firms by committing the firm to un-
JEL classification: dertaking long-term projects and bonding it to the relationship-specific investments of its
G32 stakeholders. Our results are robust to matching procedures and an exogenous change in
G34 Massachusetts corporate law that mandated staggered boards.
K22 © 2017 Elsevier B.V. All rights reserved.
Keywords:
Staggered board
Firm value
Stakeholder relationships
Innovation
1. Introduction
R
The results in the Online Appendix are available at http://ssrn.com/ Whether staggered (or classified) boards are a desirable
abstract=2866677. governance arrangement for publicly traded firms is the
RR
We thank seminar participants at Columbia University, Arizona State
subject of a long-standing debate. Unlike a unitary board,
University, Emory University, Economic University of Vienna, Harvard Uni-
versity, Northwestern University, Toulouse School of Economics, Univer- where all directors stand for reelection each year, in a
sity of Arizona, Brigham Young University, University of Southern Cali-
fornia, University of Iowa, University of Nebraska, University of Kansas,
University of Oklahoma, Texas Tech University, University of Virginia, Uni- gelo, Isil Erel, Dirk Jenter, Olubumni Faleye, Chitru Fernando, Allen Fer-
versity of TexasSan Antonio, University of Illinois at Urbana-Champaign, rell, Jesse Fried, Oliver Hart, Gerard Hertig, Gregg Jarrell, Kose John, Dalida
as well as participants at the 2014 American Law and Economics Associa- Kadyrzhanova, Louis Kaplow, Jonathan Karpoff, Reinier Kraakmann, Chris
tion Meeting, the 2014 IDC Herzlyia Conference at Tel Aviv University, the Lamoureaux, Saura Masconale, William Megginson, Alan Schwartz, Laura
2014 Ackerman Conference on Corporate Governance at Bar-Ilan Univer- Starks, Neal Stoughton, Oren Sussman, Michael Roberts, Jean Tirole, Lucy
sity, the 2014 University of Delaware Corporate Governance Symposium, White, and Joseph Zechner. We also thank Lucian Bebchuk for providing
the 2014 European Summer Symposium in Economic Theory, the 2014 us with access to his data on charter-based and bylaws-based staggered
Financial Management Association meeting, the 2014 Society for Empir- board classifications, Dirk Jenter for providing access to his CEO turnover
ical Legal Studies meeting at the University of California in Berkeley, the data, and David Larcker for providing access to the data errors for stag-
2015 Midwestern Finance Association meeting, the 2015 Finance Caval- gered boards in the IRRC volumes documented in Larcker, Reiss, and Xiao
cade, the 2015 Western Finance Association meeting, and the 2016 Amer- (2015).
∗
ican Finance Association meeting. For useful comments and discussions, Corresponding author.
we thank Yakov Amihud, Thomas Bates, Stefan Bechtold, Bernard Black, E-mail addresses: mcremers@nd.edu (K.J.M. Cremers), litov@ou.edu
Patrick Bolton, Jeff Coles, Jacque Cremer, Vicente Cunat, Harry DeAn- (L.P. Litov), sms234@email.arizona.edu (S.M. Sepe).
http://dx.doi.org/10.1016/j.jfineco.2017.08.003
0304-405X/© 2017 Elsevier B.V. All rights reserved.
K.J.M. Cremers et al. / Journal of Financial Economics 126 (2017) 422–444 423
staggered board directors are typically grouped into three between changes in board structure and long-term firm
different classes serving staggered three-year terms, with value is that staggered boards, like other corporate ar-
only one class of directors standing for reelection each rangements, are not randomly assigned (Adams, Hermalin
year. As this requires challengers to win at least two elec- and Weisbach, 2010). We try to mitigate these endogeneity
tion cycles to replace a majority of the board and, hence, concerns in the following four ways.
to endure a costly delay before gaining voting control, a First, we consider three selection mechanisms related
staggered board protects directors from market discipline. to: (1) the market for corporate control, where antici-
Conventional wisdom holds that such protection of- pated future takeovers affect valuations; (2) entrenchment,
fered by a staggered board may operate to entrench direc- where changes in manager-board relationships provide ex-
tors and managers, thereby encouraging shirking, empire planatory power for changes in board structure: and (3) re-
building, and/or private benefits extraction (Manne, 1965; verse causality, where ex ante low (high) firm valuations
Jensen, 1988, 1993). Consistent with this view, several em- provide explanatory power for the adoption (removal) of
pirical studies show that in the cross-section, firms with a a staggered board. We find no evidence for a selection
staggered board tend to have lower firm value as measured bias related to merger and acquisition (M&A) activity or
by Tobin’s Q (Bebchuk and Cohen, 2005; Faleye, 2007; Be- entrenchment, but show that firms with lower value are
bchuk, Cohen and Ferrell, 2009). substantially more likely to adopt a staggered board. Re-
A contrasting view is that the adoption of staggered verse causality thus helps provide explanatory power for
boards could potentially contribute to firm value, in two the sign reversal between our time series results and the
ways. First, a staggered board could help mitigate man- cross-sectional results in most of the literature. While the
agers’ incentives to overinvest in short-term projects if in- value of firms staggering up tends to increase afterwards,
vestors are uninformed or myopic (Stein, 1988, 1989). Sec- this increase is insufficient to fully erode the value dif-
ond, a staggered board could serve to bond a manager’s ference with other firms in the same industry, generating
commitment to the relationship-specific investments made the negative cross-sectional association between firm value
by the firm’s stakeholders, reducing the likelihood that and staggered boards. Conversely, we find no significant
the firm’s business strategy is changed via a takeover and association between lagged firm value and the decision to
thus lowering the risk that takeovers will impose costs de-stagger.
on stakeholders (Knoeber, 1986; Laffont and Tirole, 1988; Second, we incorporate possible selection effects
Shleifer and Summers, 1988). Consistent with this “bond- through the creation of multiple matched samples based
ing hypothesis,” more recent studies report evidence that on different matching procedures. In each matched sample,
firms may benefit from a staggered board if they have each firm with a changing board structure in a given year
strong stakeholder relationships (Cen, Dasgupta and Sen, is matched to a firm with the same ex ante board structure
2016; Johnson, Karpoff and Yi, 2015, 2016). and similar observable characteristics that relate to board
Motivated in part by this more recent evidence, we re- structure, but which did not change its board structure in
visit the staggered board debate using a longer sample pe- that year. The matched samples confirm the positive (neg-
riod (1978–2015) than the existing empirical literature and ative) relation between the adoption (removal) of a stag-
with a new focus on the long-term association of firm gered board and firm value.
value with changes in board structure (i.e., decisions to Third, we employ the dynamic generalized method
adopt or remove a staggered board). Our main findings of moments (GMM) estimator proposed by Arellano and
are as follows. We find no evidence that staggered boards Bover (1995) and Blundell and Bond (1998). As explained
have a strong or persistently negative association with firm by Wintoki, Linck and Netter, (2012), this methodology es-
value. Rather, in more innovative firms, or where stake- timates a simultaneous system in which firm value, board
holder investments are more relevant (e.g., with a large structure, and other key corporate characteristics are all
customer or in a strategic alliance), adopting (removing) a endogenous and dynamically interrelated. As this method
staggered board is associated with an increase (decrease) relies on strong assumptions, these results are presented
in long-term firm value. For example, the adoption (re- as a robustness check rather than as showing causality. Us-
moval) of a staggered board is associated with an increase ing a system where we can reject that the instruments are
(decrease) in firm value, as proxied by Tobin’s Q, of 5.3% weak and that accounts for unobservable heterogeneity us-
for firms with a large customer, and has an insignificant ing firm fixed effects, the dynamic GMM results also show
association for firms without a large customer. Further, our a positive (negative) relation between adopting (removing)
results are driven by the second half of our time period a staggered board and firm value.
(i.e., the 1996–2015 sub-period) and generally insignificant Fourth, we conduct a long-term event study exploit-
in the first half (i.e., the 1978–1995 sub-period). Overall, ing plausibly exogenous variation in board structure due
these results suggest that staggered boards have heteroge- to changes in Massachusetts corporate law. In 1990, Mas-
neous effects across firms and time, providing no support sachusetts made staggered boards “quasi-mandatory” by
for the entrenchment view but also making it difficult to requiring firms incorporated in the state to adopt a stag-
draw any one-size-fits-all inference about the relation be- gered board by default and making it difficult to opt out of
tween staggered boards and firm value. this requirement. We compare the value of Massachusetts
Our results are robust to various methodologies, includ- firms in the few years before and after this legal change
ing regressions of changes in Tobin’s Q on changes in board in a matched sample of firms, where the control firms
structure and a stock portfolio return approach. However, are incorporated outside of Massachusetts but have a sim-
a primary challenge in interpreting the empirical relation ilar size, are in the same industry, and have the same
424 K.J.M. Cremers et al. / Journal of Financial Economics 126 (2017) 422–444
board structure as the Massachusetts firms. After the le- sure may have been the cause of increased de-staggering
gal change, the value of the Massachusetts firms increased during the 20 0 0s, with this pressure being at least in part
more than the value of their control firms, but with no dif- motivated by the belief that staggered boards lower firm
ference between Massachusetts firms with and without a value (Subramanian, 2014).
staggered board prior to the law change. Therefore, while In conclusion, our paper makes two main contributions
these results do not provide direct evidence that mandat- to the literature. First, it provides an improvement over
ing the adoption of a staggered board increases firm value, the identification strategies used in prior studies, as most
they are also clearly inconsistent with the hypothesis that of these studies examined the relation between staggered
having a staggered board in place lowers firm value. boards and long-term firm value by relying exclusively on
These results notwithstanding, we acknowledge that cross-sectional results (Bebchuk and Cohen, 2005; Faleye,
our attempts to mitigate the endogenous choice of a stag- 2007). Compared to these studies, we employ a more com-
gered board have significant limitations. As an alternative prehensive sample, firm fixed effects in pooled panels, re-
to resolving endogeneity, we examine two economic chan- gressions of changes in value on changes in board struc-
nels through which a staggered board could be associated ture, several different matching procedures, dynamic GMM
with an increase in long-term firm value, i.e., the myopic estimation, and exogenous variation in Massachusetts cor-
market hypothesis and the bonding hypothesis. As outlined porate law.
above, under the myopic markets hypothesis, a staggered Second, we show that changes in firm value around
board helps shield management from the pressure of my- changes in board structure are inconsistent with the en-
opic investors. According to the bonding hypothesis, a stag- trenchment view of staggered boards. Rather, our results
gered board provides an efficient commitment device to- point to a heterogeneous relationship between board struc-
wards the firm-specific investments of a firm’s stakehold- ture and performance, suggesting that for different subsets
ers, such as top employees, large customers, suppliers, and of firms, staggered boards could contribute to firm value
strategic alliance partners. by preventing inefficient takeovers and/or serving to bond
Our results suggest that the role of staggered boards a firm’s commitment to the firm’s long-term stakehold-
differs across firms in a way that both economic chan- ers. Consistent with a large body of literature (see Adams
nels could play a role, although overall the results seem et al., 2010, for a review; as well as Barry and Hatfield,
more consistent with the bonding hypothesis. We find that 2012; Ahn and Shrestha, 2013; and Ge, Tanlu and Zhang,
the adoption (removal) of a staggered board has a pos- 2016), the heterogeneous relation of board structure with
itive (negative) association with firm value among firms performance thus indicates that a one-size-fits-all view of
with stronger stakeholder relationships, such as firms with board structure is not supported by the data.
large customers, productive employees, and in strategic al-
liances. We similarly find that the adoption (removal) of 2. Data and descriptive statistics
staggered boards has a more positive (negative) associa-
tion with firm value among firms whose projects require 2.1. Data
longer-term investments and are likely harder to value by
outside investors, such as firms with more investments in Our data come from several sources, with the main data
innovation and intangibles. Conversely, using proxies that sample covering the time period 1978–2015. Data avail-
other research has linked to myopic investment behavior ability varies with the different sources. We obtain data
(Bushee, 1998; Cremers, Pareek and Sautner, 2016), we find for the key independent variable of our study, Staggered
no evidence that the association between staggered boards board, from two main sources, covering a total number of
and firm value is stronger for firms with more short-term 3,076 firms. For the time period 1990–2015, as in prior
institutional investors. studies on the value impact of staggered boards (Bebchuk
Nonetheless, the decline in firm value after de- and Cohen, 2005; Faleye, 2007; Masulis, Wang and Xie,
staggering remains challenging to interpret, under either 2007), we use the corporate governance data set main-
channel. De-staggering may be due to outside shareholder tained by RiskMetrics, which acquired the Investor Respon-
pressure, to which firms with worse prospects could be sibility Research Center (IRRC).1 For the time period 1978–
more likely to give in. As a result, the subsequent decline 1989, we use data from Cremers and Ferrell (2014), who
in value could reflect selection rather than causality. Al- hand-collected information on firm-level corporate gover-
ternatively, the bonding channel suggests that firms that nance provisions for these years, including information on
agree to de-stagger their board signal that the firm expects
fewer rents (i.e., lower future surpluses) from its long-term 1
During the period 1990–2006, IRRC published volumes in the fol-
stakeholder relationships, which would explain the subse- lowing years: 1990, 1993, 1995, 1998, 20 0 0, 20 02, 20 04, and 20 06. We
quent decline in firm value. hand-checked the data on staggered boards in all missing years in the
1994–2006 time period using proxy statements from the Security and
Further, while our paper focuses on examining the
Exchange Committee’s ‘EDGAR’ website, in order to capture the precise
long-term effects of staggered boards on firm value rather timing of changes in board structure. We start our hand checks in 1994
than the question of why firms adopt or remove a stag- because electronic records on the SEC’s website are only available since
gered board, the above results suggest at least a partial an- that year. Furthermore, Larcker et al. generously shared the data errors
swer to that question. On the one hand, some firms appear for staggered boards in the IRRC volumes documented in Larcker, Reiss,
and Xiao (2015). Incorporating these errors results in changing 115 ob-
to stagger up to secure the benefits arising from long-term
servations, and eliminating 11 removals and seven adoptions of staggered
investments and stronger shareholder bonding (Johnson boards, while introducing eight other removals and 14 other adoptions of
et al., 2015, 2016). On the other hand, shareholder pres- staggered boards. These changes do not significantly affect our results.
the same provisions tracked by the IRRC for the period fairly stable ratio of firms with a staggered board, hovering
1990–2015 and, in particular, staggered boards. at around 60%. After 2006, the ratio of firms with a stag-
Since our main focus is on the value relevance of staggered board steadily declines, until reaching about 35% in
gered boards, the main dependent variable in our analy- 2015.
sis is firm value. Consistent with many prior studies in- Fig. 2 shows the percentage of firms with a staggered
vestigating the relation between governance arrangements board over time. We demonstrate the dynamics of stag-
and firm value (Demsetz and Lehn, 1985; Morck, Shleifer gering up and staggering down within six different co-
and Vishny, 1988; Lang and Stultz, 1994; Yermack, 1996; horts of firms through time, where no new firms enter
Daines, 2001; Gompers, Ishii and Metrick, 2003), we mea- each cohort subsequently, while firms drop out of the co-
sure firm value using Tobin’s Q (Q), defined as the ra- horts due to M&A, privatizations, bankruptcies, and other
tio of market-to-book value of assets (Fama and French, de-listing events. The six cohorts of firms are as follows:
1992) from Compustat data. As an additional measure of (1) firms with a staggered board in 1978, (2) firms without
changes to firm value, we use the stock returns surround- a staggered board in 1978, (3) firms with a staggered board
ing changes in board structure, obtaining stock return data in 1990, (4) firms without a staggered board in 1990, (5)
from the Center for Research in Security Prices (CRSP) firms with a staggered board in 20 0 0, and (6) firms with-
database (see Section 3.2). In our analysis of (in) voluntary out a staggered board in 20 0 0.
Chief Executive Officer (CEO) turnover (see Section 3.3.2), Among the 195 firms in our sample with a staggered
we employ CEO turnover data from Jenter and Kanaan board in 1978, only a few de-staggered before 2005, as
(2015) over the time period 1993–2001 for all ExecuComp nearly 93% remain staggered in 2004 (out of the firms still
firms. in the sample). Starting in 2005, a large number of firms in
We provide brief definitions of all the controls and the this cohort de-staggered, with only about 36% of the sur-
interaction variables in Table 1. Our default controls are viving 42 firms remaining staggered in 2015. Conversely,
Ln(Assets), Delaware incorporation, return on assets (ROA), among the 684 firms without a staggered board in 1978,
CAPX/Assets, research and development expenditures over almost half staggered-up by 1989 and 55% had a stag-
total sales (R&D/ Sales), and Industry M&A volume. The gered board in 2004, after which many firms in this cohort
last control is used to verify the existence of an antic- de-staggered, leaving only 24% of the cohort sample (146
ipation effect of future takeover activity (Edmans, Gold- firms) that survives with a staggered board in 2015. Com-
stein and Jiang, 2012). Our extended set of controls also paring the 1990 and 20 0 0 cohorts to the 1978 cohort, we
includes G-Index and Insider ownership, to replicate more observe analogous trends. In particular, among the firms
closely Bebchuk and Cohen (2005). G-Index is a composite with a staggered board in 1990 and 20 0 0, almost all re-
of 24 provisions that measures the strength of sharehold- mained staggered until 2005 and began to increasingly de-
ers rights, with a higher score indicating weaker share- stagger afterward. We note that when a firm staggers up,
holder rights. In computing G-Index, we remove Staggered it typically takes a while before it decides to de-stagger.
board (as in Bebchuk and Cohen, 2005) and Poison pill, as Lastly, in the 1995–2002 time period that has been the
we separately include these two provisions. We obtain G- focus of most prior studies on staggered boards (Bebchuk
Index data from Cremers and Ferrell (2014) for 1978–1989 and Cohen, 2005; Faleye, 2007; Bebchuk et al., 2009), there
and the RiskMetrics data set (formerly IRRC) for 1990– is little time variation in board structure.
2006.2 Insider ownership data reduce our sample size, as
Compact Disclosure (data source until 2006) primarily cov- 3. Results
ers NYSE and Amex firms before 1995.
Table 2 presents the descriptive statistics of all the vari- 3.1. Staggered boards and firm value: cross-sectional and
ables.3 Averaged across all firm-year observations, about time series results
52% of firms have a staggered board. The average Q in our
sample is 1.61 with a standard deviation of 0.89.4 In this section, we consider both the cross-sectional
and time series association between staggered boards and
2.2. Staggering and de-staggering long-term firm value. In particular, for the time series as-
sociation, we show results for both pooled panel Q regres-
Fig. 1 presents the annual percentage of firms with a sions with firm fixed effects and changes-in-Q regressions
staggered board in our sample from 1978 to 2015. We ob- on changes in board structure. For all tables, we show t-
serve substantial time variation. This period is character- statistics based on robust standard errors clustered by firm.
ized by a slow trend of staggering up, which rapidly accel- The motivation for employing standard errors clustered by
erates in 1984 to 1992. In the 1992–2006 period, there is a firm is to incorporate the correlation of regression residu-
als across time for a given firm, which is particularly im-
portant for variables with little time variation (Petersen,
2
As a caveat, Larcker, Reiss, and Xiao (2015) document some cod- 2009). In Table 3 only, we also provide the t-statistic based
ing errors in the G-Index provisions. We use the G-Index primarily to on robust standard errors that are not clustered, where t-
more closely replicate past literature using the uncorrected G-Index data, statistics are considerably smaller when we cluster by firm.
and only incorporate coding corrections from Larcker, Reiss, and Xiao
Table 3 presents results for the pooled panel Q regres-
(2015) pertaining to the staggered board.
3
Pairwise correlations are provided in Online Appendix Table A.1. sion with year and four-digit Standard Industrial Classi-
4
The averages of the control variables are similar across the sample of fication (SIC) industry fixed effects in the regression for
firms with and without staggered boards (see Online Appendix Table A.2). columns 1–3 and year and firm fixed effects in the regres-
Table 1
Definitions.
Dependent variables Definition
Tobin’s Q[t] Market value of assets (total assets – book equity + market equity) divided by book value of assets.
Calculation follows Fama and French (1992). Source of data is Compustat annual data file.
Monthly returns on long (short) Monthly return of a portfolio created by stocks that stagger up (down) their boards. Portfolio is created by
portfolio “6m12” including all stocks of firms that have (de)staggered their board for 12 months, starting six months before
the fiscal year-end of the year in which the company has reported its board being (de-)staggered for the
first time. Returns are either equally or value weighted.
Monthly returns on long (short) Analogous to “6m12,” but now including all stocks of firms that have (de)staggered their board for 12
portfolio “12m12” months, starting 12 months before the fiscal year-end of the year in which the company has reported its
board being (de)staggered.
Monthly returns on long (short) Analogous to “12m12,” but now including all stocks of firms that have (de)staggered their board for 24
portfolio “12m24” months.
(Forced) CEO turnover[t] Defined as one if there is an (in) voluntary CEO departure in the Jenter and Kanaan (2015) data file; zero
otherwise. Data are available for 1993–2001.
Independent variables
CAPX/Assets[t] Capital Expenditure[t] /Total Assets[t] .

Delaware incorporation[t] Indicator variable if the company is incorporated in Delaware.
Excess returns[t] Annual returns for each firm at the fiscal year end date net of market return for the same period. Data for
stock returns are from CRSP. Data for market returns are from Ken French’s data library. This variable is
then winsorized at 2.5% in each tail of its distribution.
G-Index (minus staggered board)[t] Sum of 23 (i.e., 24 when excluding staggered board) governance provision indicators in the corporate
charter or bylaws introduced by Gompers et al. (2003).
Insider ownership[t] The insider ownership in year t is the percentage of shares owned by insiders from all shares. Collected
from Compact Disclosure for 1986–2006. We supplement these data with ownership by the top five
officers of the firm from ExecuComp for 2007–2015.
Ln(Age) [t]
Natural logarithm of firm age, calculated as the difference in year t and the first year the company
appeared in the CRSP database.
Ln(Assets) [t] Natural logarithm of total book assets in year t.
Industry M&A volume[t] The ratio of mergers and acquisitions’ dollar volume in SDC to the total market capitalization from CRSP
per Fama-French 49 industries. We only consider ordinary stocks and excludes American depositary
receipts (ADRs) and real estate investment trusts (REITs). We only include SDC transactions that are
completed and where buyer achieves control of the target.
Poison pill[t] Antitakeover provision obtained from the Cremers and Ferrell (2014) database for 1978–1989, and from
RiskMetrics for 1990–2015.
R&D/ Sales[t] R&D[t] / Sales[t] .
ROA[t] Earnings before interest, taxes, depreciation and amortization over total assets (EBITDA[t] /Total assets[t] ).
Staggered board[t] Indicator variable equal to one (zero otherwise) if the board is staggered in year t. Data are obtain from
Cremers and Ferrell (2014) for 1978–1989, and from Risk Metrics, SharkRepellent.net, hand collection, and
incorporating the data corrections from Larcker et al., (2015) for 1990–2015.
࢞ Staggered board[t- 1 , t] Indicator variable equal to one if the board adopted a staggered board in year t and equal to minus-one if
the board removed a staggered board in year t, and equal to zero otherwise.
Staggered board adopted[t- 1 , t] Indicator variable equal to one if the board adopted a staggered board in year t, and equal to zero
otherwise.
Staggered board removed[t- 1 , t] Indicator variable equal to one if the board removed a staggered board in year t, and equal to zero
otherwise.
System GMM instruments
SSB /SALL Percentage of total sales in an industry-year by firms with a staggered board. Industry is defined as
three-digit SIC code.
SCON /SALL Percentage of total sales in an industry-year by conglomerate firms. Industry is defined as three-digit SIC
code.
SSTAGGER /SALL Percentage of total sales in an industry-year by firms that staggered their board in the past two years.
Industry is defined as three-digit SIC code.
SDESTAGGER /SALL Percentage of total sales in an industry-year by firms that repealed their staggered board in the past two
years. Industry is defined as three-digit SIC code.
Interacted variables
Institutional holding duration[t] Average length of time (in quarters) that the firm’s institutional owners of its stock have owned the
equities in their 13F holdings reports from Thomson Reuters (available through Wharton Research Data
Services), weighted by the institution’s ownership in the firm’s stock.
Percent transient institutions[t] Percentage of the stock’s institutional ownership that is ‘transient’ as classified in Bushee (1998) using
data available from his website. Transient institutions have relatively high portfolio turnover and are
relatively well diversified.
Firm sales[t] Ln(Sales) in year t.
Ln[(Total assets[t] - Net
PP&E[t] )/Total assets[t] ].
Ranked patent citation count[t] Annually ranked patent citation count. Data are available for 1978–2003. Citations are calculated following
Hall, Jaffe and Trajtenberg (2002). Source is the NBER U.S. Patent Citations data file. We divide the ranked
patent citation count by 10 0 0.
(continued on next page)
Table 1 (continued)
Research quotient[t] A firm-specific output elasticity of R&D, representing the percentage change in revenues for 1% change in
R&D, as introduced by Knott (2008). Source of data for 1978–2010 is WRDS and from Anne-Marie Knott’s
website for 2011–2015.
Contract specificity Industry-level fraction of the inputs that are sold on an organized exchange in the Nunn (2007) data file,
available for 1997 only, see http://scholar.harvard.edu/nunn/pages/data-0.
Labor productivity Data on output per hour from the Bureau of Labor Statistics, see http://www.bls.gov/bls/productivity.htm.
Available for 400 selected industries in manufacturing, mining, utilities, wholesale and retail trade, and
services.
Large customer Indicator variable equal to one if there is at least one customer accounting for at least 10% of the
consolidated sales of the firm in that fiscal year. The source of the data is the Compustat Customer
Segments database. Data are available since 1985.
Strategic alliance Indicator variable equal to one if the firm is in a strategic alliance. We only include strategic alliances with
up to three partners. Source of data is SDC Strategic Alliances. Data are available since 1985.
Table 2
Descriptive statistics for main dependent and independent variables.
This table presents sample descriptive statistics for the main dependent and independent variables, as well as the
interacted variables. All continuous variables are winsorized at 2.5% in both tails. See Table 1 for variable definitions.
Dependent variables: Mean Median St. dev. Min Max Obs.
Q [t] 1.613 1.307 0.889 0.728 4.729 34,476

࢞ Q[t- 1 , t] 0.003 0.012 0.449 −1.836 1.580 31,134
Independent variables: Mean Median St. dev. Min Max Obs.
CAPX/Assets[t] 0.059 0.047 0.046 0 0.198 34,476

Delaware incorporation[t] 0.558 1 0.497 0 1 34,476
G-Index (minus stagg. board)[t] 7.680 8 3.200 1 18 23,525
Insider ownership[t] 0.070 0.03 0.100 0 1 21,216
Ln(Age) [t] 2.870 3 0.980 0 4.450 27,754
Ln(Assets) [t] 7.351 7.238 1.578 4.584 11.171 34,476
Industry M&A volume[t] 0.028 0.012 0.049 0 0.359 34,476
R&D/Sales[t] 0.029 0 0.056 0 0.231 34,476
ROA[t] 0.139 0.135 0.079 −0.041 0.324 34,476
Staggered board[t] 0.524 1 0.499 0 1 34,476
Interacted variables: Mean Median St. dev. Min Max Obs.
Institutional holding duration 6.846 6.818 0.990 2.226 10.445 28,298

Percent transient institutions 0.222 0.196 0.138 0 0.923 28,298
Ln(Intangible / Total assets[t] ) −0.487 −0.323 0.483 −3.876 0 34,229
Ranked Patent citation count[t] 0.385 0.385 0 0.385 0.385 34,476
Firm sales[t] 7.270 7.193 1.527 −2.937 13.089 34,460
Research quotient[t] 0.116 0.112 0.049 −0.015 0.275 14,847
Contract specificity[t] 0.913 0.967 0.131 0.352 0.998 10,478
Labor productivity[t] 1.472 1.104 0.615 0.825 2.904 27,715
Large customer[t] 0.333 0 0.471 0 1 29,107
Strategic alliance[t] 0.214 0 0.410 0 1 27,782
sion for columns 4–6 for the full time period (1978–2015), We consider the time series evidence in column 4 in
as well as the first and second half of the sample (i.e., Table 3 by using the same pooled panel Tobin’s Q regres-
the 1978–1995 and 1996–2015 sub-periods, respectively). sions as for column 1 but now with firm rather than indus-
Separating these sub-periods is particularly interesting due try fixed effects. Including firm fixed effects is equivalent
to the many changes in corporate governance, ownership, to removing the time-invariant component in Q, Staggered
and takeover markets that occurred during these periods, board, and all controls, reducing the potential bias result-
which likely affected the role of staggered boards. ing from omitted time-invariant variables at the firm level.
In column 1 using our full 1978–2015 sample, the cross- Once we include firm fixed effects, we are essentially com-
sectional association of Staggered board and Q is negative paring the average firm value before versus after a change
and both statistically and economically significant, suggest- in board structure. We find a statistically significant pos-
ing that firms with a staggered board have a firm value itive time series association between Staggered board and
that is 2.1% (= −0.034/1.61) lower than firms without a Q.5 For the full sample in column 4, the adoption (removal)
staggered board. We thus find a negative cross-sectional
association like Bebchuk and Cohen (2005), but one with a
considerably lower economic magnitude. We find similarly 5
The firm fixed effects results in Table 3 exploit only the “within firm”
negative coefficient estimates of Staggered board in our two variation using firm fixed effects. As a robustness check, Online Appendix
Table A.3 reports the “between firm” coefficients of Staggered board, only
sub-periods, but the negative coefficient is only statistically
exploiting cross-sectional variation and ignoring time series changes in
significant (at 5% confidence level) for the 1978–1995 pe- board structure within firms. The “between firm” coefficients are similar
riod and becomes insignificant for the 1996–2015 period. to the results in Table 3, confirming that those pooled panel regression
Fig. 1. Percentage of firms with a staggered board. The figure shows the percentage of firms with a staggered board in our 1978–2015 sample period.
Excluded from the sample are firms that have dual class shares. See Table 2 for sample descriptive characteristics.
of a staggered board is associated with an increase (de- that the adoption (removal) of a staggered board is associ-
crease) in Q of 3.2% (=0.051/1.61), with a t-statistic of 2.06 ated with an increase (decrease) in Q of 5.5% (i.e., dividing
using standard errors clustered by firm.6 the coefficient of 0.098 by the average Q in the sample,
The positive long-term time series association between which is 1.77).
staggered boards and Q is driven by the second half of the Many possible explanations exist for these changes in
sample. As shown in column 5 in Table 3, the coefficient of the association between firm value and board structure
Staggered board is positive but insignificant for the 1978– across time. In general, these results suggest that the stag-
1995 period, while column 6 shows a coefficient of 0.098 gered board structure became increasingly relevant over
with a t-statistic of 2.73 when we use clustering by firm time. In particular, Cremers and Ferrell (2014) show that
for the 1996–2015 period. The economic magnitude of the board control, and more generally shareholder rights, be-
latter effect seems meaningful, as this coefficient indicates came more important after the judicial validation of the
poison pill by the Delaware Supreme Court in 1985. This
is because with a poison pill in place, or the ability to
results with industry fixed effects essentially capture cross-sectional vari- adopt a poison pill on short notice referred to as a “shadow
ation. Online Appendix Table A.3 also shows the variation decomposition pill” or a pill “on the shelf,” see Coates (20 0 0), hostile
of Q, indicating that the cross-sectional (between firm) variation in Q is takeovers became more difficult, as gaining board control
considerably larger than its time series variation.
6 and having the pill removed became the only route avail-
Online Appendix Table A.4 presents results for the time period used
in Bebchuk and Cohen (20 05), 1995–20 02, with their additional controls. able to a prospective bidder.7 Further, it is also likely that
Overall, the cross-sectional results are consistent with Bebchuk and Co-
hen (2005), although they report stronger economic and statistical signif-
7
icance (though using robust standard errors that do not seem to be clus- Consistent with this, in Online Appendix Table A.4 and using firm
tered), and using two-digit rather than four-digit SIC codes as we do. Re- fixed effects, we find that Staggered Board has an insignificant coefficient
sults with two-digit industry SIC codes are similar in magnitude to those for 1978–1985 (column (4)), while it has a positive and significant co-
presented in Bebchuk and Cohen (2005) (see Online Appendix Table A.5). efficient for 1986–2015 (column (5)). However, many other changes over
Our results, however, remain the same using either three-digit SIC codes the 1980s and 1990s may have played a role next to or with the poi-
or the Fama-French 49 industry groups. son pill, such as the emergence of institutional investors and increasingly
Fig. 2. Cohort analysis for staggering up and de-staggering. The figure shows the percentage of firms with a staggered board each year for six cohorts of
firms: (1) firms with a staggered board in 1978 (SB in 1978), (2) firms without a staggered board in 1978 (No SB in 1978), (3) firms with a staggered board
in 1990 (SB in 1990), (4) firms without a staggered board in 1990 (No SB in 1990), (5) firms with a staggered board in 20 0 0 (SB in 20 0 0), and (6) firms
without a staggered board in 20 0 0 (No SB in 20 0 0). The figure shows the annual percentage with a staggered board within each cohort as a percentage of
those firms that remain in our sample that year until the last year in our data, 2015.
Table 3
Firm value and staggered boards.
This table presents annual pooled panel Q regressions on Staggered board with industry (columns 1–3) or firm (columns 4–6) fixed effects. All specifica-
tions include year dummies and the following control variables: Staggered board[t- 1 ] , Ln(Assets)[t- 1 ] , Delaware incorporation[t- 1 ] , ROA[t- 1 ] , CAPX/Assets[t- 1 ] , R&D/
Sales[t- 1 ] , and Industry M&A volume[t- 1 ] . The analysis includes the following sub-periods: 1978–2015, 1978–1995, and 1996–2015. All variables are defined
in Table 1. T-statistics (in their absolute value) are based on robust standard errors clustered by firm and presented in parentheses below the coefficients.
Statistical significance of the coefficients is indicated at the 1%, 5%, and 10% levels by ∗∗∗ , ∗∗ , and ∗ , respectively. For the key independent variable, Staggered
board[t- 1 ] , we also show the t-statistics based on robust standard errors that are not clustered, reported in brackets.
Dep. variable: Q[t]
Variables (1) (2) (3) (4) (5) (6)
Period: 1978–2015 1978–1995 1996–2015 1978–2015 1978–1995 1996–2015

Fixed Effects: Industry + year fixed effects Firm + year fixed effects
Staggered board[t- 1 ] −0.034∗∗ −0.046∗∗ −0.036 0.051∗∗ 0.023 0.098∗∗∗

(firm cluster) (2.04) (2.39) (1.61) (2.06) (0.73) (2.73)
[no cluster] [4.34] [4.65] [3.18] [4.32] [1.58] [5.16]
Ln (Assets)[t- 1 ] −0.031∗∗∗ −0.033∗∗∗ −0.031∗∗∗ −0.222∗∗∗ −0.104∗∗∗ −0.352∗∗∗
(4.35) (3.83) (3.34) (13.14) (4.33) (15.08)
Delaware incorporation[t- 1 ] INcorporaIncoincorporation[t- 1 ] 0.02 0.023 0.014
(1.07) (1.09) (0.55)
ROA[t- 1 ] 5.338∗∗∗ 4.043∗∗∗ 5.825∗∗∗ 3.189∗∗∗ 2.039∗∗∗ 2.81∗∗∗
(34.17) (21.51) (30.11) (21.62) (12.53) (15.72)
CAPX/Assets[t- 1 ] −0.376 −0.493∗ 0.315 0.057 0.102 0.299
(1.68) (1.83) (1.00) (0.34) (0.58) (1.22)
R&D/ Sales[t- 1 ] 4.168∗∗∗ 4.168∗∗∗ 4.323∗∗∗ 1.274∗∗ 2.302∗ −0.288
(11.88) (5.98) (11.13) (2.46) (1.92) (0.50)
Industry M&A volume[t- 1 ] −0.231∗∗∗ 0.117 −0.32∗∗∗ −0.235∗∗∗ −0.047 −0.252∗∗∗
(3.03) (1.11) (3.46) (3.40) (0.47) (3.03)
# of firms in regression 3,076 1,581 2,415 3,076 1,581 2,415

N 34,476 13,022 21,454 34,476 13,022 21,454
Adj. R2 0.510 0.542 0.495 0.710 0.758 0.727
investors became more aware over time of the availabil- vide mixed and statistically fairly weak results (Jarrell and
ity of shadow pills, which could similarly explain why the Poulsen, 1987; Mahoney and Mahoney, 1993; Faleye, 2007;
deterrence effect of staggered boards gradually increased Guo, Kruse and Nohel, 2008, 2014). Following prior stud-
(Coates, 20 0 0).8 ies (Gompers et al., 2003; Bebchuk et al., 2009; Cremers
Next, we investigate the time series dimension of the and Ferrell, 2014), Table 5 reports the abnormal returns of
association between firm value and staggered boards by equally weighted portfolios buying (selling) stocks of firms
regressing changes in Q on changes in Staggered board. We around the time they stagger up (down).9 We consider
calculate the change in firm value at the end of the fis- portfolios that use staggered board information at the time
cal year when the board change occurred to the firm value this was public information, as well as portfolios that are
one, two, and three years later. The results in Panel A of constructed with perfect foresight of subsequent changes
Table 4 confirm that firm value, as proxied by Q, increases in board structure, and various lengths of time to hold the
following the adoption of a staggered board and decreases stocks.
following a decision to de-stagger. The coefficients on the Overall, the results of our long-term event study are
change in Staggered board in columns 1–3 show that the consistent with our results using Q, although the abnormal
increase (decrease) in value after staggering up (down) oc- return estimates are quite noisy due to the limited num-
curs gradually, rather than all in the first year. This sug- ber of stocks in each portfolio (on average 13–23 stocks,
gests that market participants need some time to fully depending on how long we keep stocks in the portfolio).
learn and process the changed prospects of the firm that In the long portfolio, the stocks of firms are purchased be-
occur after the change in board structure. Columns 4–6 fore they stagger up and tend to exhibit positive abnormal
show that the changes in board structure are not associ- performance that is statistically significant using the mar-
ated with changes in Q in the first half of the sample (i.e., ket model, insignificant using the three-factor model, and
1978–1995), while columns 7–9 show a strongly positive marginally significant using the four-factor model. In the
association in the second half of the sample (i.e., 1996– short portfolio, the stocks of firms are purchased before
2015), consistent with the pooled panel regressions with they de-stagger and generally have negative alphas that are
firm fixed effects in Table 3. consistently statistically insignificant. The long–short port-
Given the trends in staggering up and down over our folio consistently has positive abnormal returns with a sta-
full time period (i.e., 1978–2015), it is a useful robustness tistical significance that again depends on the factor model
check to distinguish between adoptions versus removals and portfolio construction that we use.
of staggered boards. The dummy variable Staggered board Combined with our change-in-Q regressions in
adopted equals one in the year that the firm adopts a stag- Tables 3 and 4, these results suggest that investors
gered board and zero otherwise. The dummy variable Stag- learn about prospective changes to board structure over
gered board removed equals one in the year that the firm time, rather than primarily at the proxy filing date or
removes a staggered board and zero otherwise. The results the annual meeting date. This apparent gradual learning
are presented in Panel B of Table 4. As shown in column also makes it more difficult to interpret short-term event
3 for the change in Q after three years and using the full studies of abnormal returns concerning staggered boards
time period, the coefficients on adoptions and removals around the proxy filing or the annual meeting dates.
of staggered boards are of similar magnitude but oppos-
ing signs, with a coefficient for Staggered board adopted of 3.3. Three specific selection mechanisms
0.112 (t-stat. =3.50) and Staggered board removed of −0.142
(t-stat. =2.86). As before, these results are driven by the As board structure is an endogenous choice (Adams
1996–2015 time period (columns 7–9) and are insignificant et al., 2010), a crucial challenge affecting our results is that
in the 1978–1996 time period (columns 4–6). endogeneity can lead to selection effects, where firms with
particular characteristics are more likely to adopt or repeal
3.2. Portfolio approach a staggered board. This includes the possibility that firm
performance drives changes in corporate governance rather
As a robustness check to our Q analysis, we con- than the other way around, or both could be related to fac-
duct a long-term stock return event study around changes tors that are omitted or difficult to observe, such as disrup-
in board structure that provides complementary evidence tive innovations or complex competition dynamics. While
to existing short-term event studies, which tend to pro- employing changes in firm value or using firm fixed effects
helps to ameliorate the endogeneity bias, these method-
ologies leave open the possibility of dynamic endogeneity
independent boards after the reform of NYSE listing standards and the (Wintoki et al., 2012; Hoechle, Schmid, Walter, and Yer-
introduction of the Sarbanes-Oxley Act. A more detailed investigation of mack, 2012). In order to mitigate these concerns, in this
these changes and their relevance for staggered boards falls outside the
section we examine three different selection mechanisms:
scope of this paper.
8 (1) the market for corporate control; (2) the possibility of
We also checked whether the association between changes in board
structure and firm value are strongly affected by the technology sector entrenchment or board capture; and (3) reverse causality.
valuation increase and decrease in 1999–2001, or by the recent financial In Sections 3.4–3.6, we employ several different matching
crisis. We also find that our results are robust to removing 1999–2001 or
20 07–20 09 from the sample (or 20 05–20 07 when we consider changes in
9
Q), as well as to adding higher-order fixed effects where we interact four- Results for value-weighted portfolios are reported in Panel A of On-
digit SIC codes and year fixed effects in a specification with firm fixed line Appendix Table A.7, and further robustness checks are provided in
effects (see Online Appendix Table A.6). Panel B of Online Appendix Table A.7.
Table 4
Changes in firm value and changes in staggered boards.
Panels A and B presents pooled panel first difference regressions with the dependent variable being the change in Q from t to t + 1 in columns 1, 4 and
7 (i.e., ࢞ Q[ t , t +1] ), the change in Q from t to t + 2 in columns 2, 5 and 8 (i.e., ࢞ Q[ t , t+ 2] ), and the change in Q from t to t + 3 in columns 3, 6 and 9 (i.e., ࢞
Q[ t , t +3] ). The dependent variables have been demeaned with their annual cross-sectional averages. As independent variables, we include the following: ࢞
Staggered board[t- 1 ,t] (in Panel A), ࢞ Ln(Assets)[t- 1 ,t] , ࢞ ROA[t- 1 ,t] , ࢞ CAPX/Assets[t- 1 ,t] , ࢞ R&D/Sales[t- 1 ,t] , and ࢞ Industry M&A volume[t- 1 ,t] . In Panel B, Staggered
board removed[t- 1 ,t] and Staggered board adopted[t- 1 ,t] are used instead of ࢞ Staggered board[t- 1 ,t] . Sample period is 1978-2015 in columns 1–3, 1978–1995 in
columns 4–6, and 1996–2015 in columns 7-9. Standard errors are clustered at the firm level. Results are robust to an adjustment to the standard errors for
autocorrelation as in Newey-West (where the adjustment includes up to six lags). T-statistics (in their absolute value) are based on robust standard errors
and presented in parentheses below the coefficients. Statistical significance of the coefficients is indicated at the 1%, 5%, and 10% levels by ∗∗∗ , ∗∗ , and ∗ ,
respectively. All control variables are defined in Table 1. Included but not shown are industry fixed effects using the Fama-French 49 industry definitions.
Our sample for column 1 includes 386 cases of staggering up and 309 cases of staggering down.
Panel A:
Dep. variable: ࢞ Q[t, t+ 1 ] ࢞ Q[t, t+ 2 ] ࢞ Q[t, t+ 3 ] ࢞ Q[t, t+ 1 ] ࢞ Q[t, t+ 2 ] ࢞ Q[t, t+ 3 ] ࢞ Q[t, t+ 1 ] ࢞ Q[t, t+ 2 ] ࢞ Q[t, t+ 3 ]
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9)
1978 – 2015 1978 – 1995 1996 – 2015
࢞ Staggered board[t- 1 , t] 0.0201 0.0767∗∗∗ 0.123∗∗∗ −0.00167 −0.0178 −0.0185 0.00138 0.0728∗ 0.135∗∗∗
(1.36) (3.01) (4.26) (0.09) (0.57) (0.55) (0.06) (1.82) (2.91)
࢞ Ln(Assets)[t- 1 , t] −0.303∗∗∗ −0.545∗∗∗ −0.594∗∗∗ −0.144∗∗∗ −0.268∗∗∗ −0.316∗∗∗ −0.366∗∗∗ −0.666∗∗∗ −0.733∗∗∗
(14.65) (17.33) (15.82) (5.86) (7.81) (7.25) (13.96) (16.47) (14.79)
࢞ ROA[t- 1 , t] −0.373∗∗∗ −0.844∗∗∗ −1.289∗∗∗ −0.143 −0.202 −0.451∗∗∗ −0.474∗∗∗ −1.206∗∗∗ −1.857∗∗∗
(4.44) (8.58) (10.83) (1.59) (1.57) (3.30) (3.98) (8.51) (10.33)
࢞ CAPX/Assets[t- 1 , t] −0.902∗∗∗ −0.930∗∗∗ −1.200∗∗∗ −0.393∗∗∗ −0.487∗∗∗ −0.718∗∗∗ −1.404∗∗∗ −1.559∗∗∗ −2.027∗∗∗
(6.47) (5.35) (6.65) (2.98) (3.59) (4.38) (6.14) (4.98) (6.32)
࢞ R&D/ Sales[t- 1 , t] 0.674 0.374 0.167 1.898 1.159 1.815 0.459 0.0226 −0.387
(1.61) (0.74) (0.28) (1.57) (0.86) (1.20) (0.99) (0.04) (0.59)
࢞ Industry M&A volume[t- 1 ,t] −0.0503 −0.107∗∗∗ −0.161∗∗∗ 0.0934∗ −0.107∗ 0.0427 −0.0708∗∗ −0.104∗∗ −0.195∗∗∗
(1.59) (2.87) (3.99) (1.78) (1.87) (0.71) (2.00) (2.47) (4.11)
N 31,049 27,017 24,146 11,435 11,014 10,638 19,614 16,003 13,508

Adjusted R2 0.017 0.031 0.032 0.009 0.012 0.013 0.021 0.041 0.047
Panel B:
Dep. Variable: ࢞ Q[t, t+ 1 ] ࢞ Q[t, t+ 2 ] ࢞ Q[t, t+ 3 ] ࢞ Q[t, t+ 1 ] ࢞ Q[t, t+ 2 ] ࢞ Q[t, t+ 3 ] ࢞ Q[t, t+ 1 ] ࢞ Q[t, t+ 2 ] ࢞ Q[t, t+ 3 ]
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9)
1978 – 2015 1978 – 1995 1996 – 2015
Staggered board removed[t- 1 , t] −0.012 −0.087∗∗ −0.142∗∗∗ 0.033 0.171 0.127 0.002 −0.068∗ −0.105∗∗
(0.53) (2.22) (2.86) (0.34) (1.02) (0.96) (0.10) (1.73) (2.04)
Staggered board adopted[t- 1 , t] 0.029 0.070∗∗ 0.112∗∗∗ 0.001 −0.005 −0.010 0.019 0.089 0.207∗∗∗
(1.55) (2.23) (3.50) (0.05) (0.16) (0.28) (0.45) (0.98) (2.82)
࢞ Ln(Assets)[t- 1 , t] −0.303∗∗∗ −0.545∗∗∗ −0.594∗∗∗ −0.144∗∗∗ −0.267∗∗∗ −0.316∗∗∗ −0.366∗∗∗ -0.666∗∗∗ −0.732∗∗∗
(14.64) (17.34) (15.82) (5.85) (7.79) (7.24) (13.95) (16.47) (14.76)
࢞ ROA[t- 1 , t] −0.374∗∗∗ −0.844∗∗∗ −1.288∗∗∗ −0.144 −0.204 −0.453∗∗∗ −0.474∗∗∗ −1.206∗∗∗ −1.856∗∗∗
(4.44) (8.58) (10.83) (1.60) (1.59) (3.31) (3.98) (8.51) (10.32)
࢞ CAPX/Assets[t- 1 , t] −0.902∗∗∗ −0.930∗∗∗ −1.200∗∗∗ −0.393∗∗∗ −0.489∗∗∗ −0.719∗∗∗ −1.404∗∗∗ −1.559∗∗∗ −2.027∗∗∗
(6.48) (5.35) (6.64) (2.98) (3.60) (4.39) (6.14) (4.98) (6.32)
࢞ R&D/ Sales[t- 1 , t] 0.674 0.374 0.167 1.898 1.159 1.816 0.459 0.023 −0.383
(1.61) (0.74) (0.28) (1.57) (0.86) (1.20) (0.99) (0.04) (0.59)
࢞ Industry M&A volume[t- 1 ,t] −0.050 −0.107∗∗∗ −0.161∗∗∗ 0.093∗ −0.107∗ 0.043 −0.071∗∗ −0.104∗∗ −0.195∗∗∗
(1.59) (2.87) (3.99) (1.78) (1.87) (0.71) (2.00) (2.47) (4.12)
N 31,049 27,017 24,146 11,435 11,014 10,638 19,614 16,003 13,508

Adjusted R2 0.017 0.031 0.032 0.009 0.012 0.013 0.021 0.041 0.047
procedures, dynamic GMM estimation, and exogenous vari- convince shareholders to adopt a staggered board in order
ation in Massachusetts corporate law. to negotiate from a stronger position with potential acquir-
ers, which could provide explanatory power for the sub-
3.3.1. The market for corporate control sequent increase in firm value. Reversely, the decrease in
Staggered boards could relate to firm value through firm value upon de-staggering could be due to an antici-
takeovers, as positive (negative) changes in firm value after pated takeover not occurring.
adopting (removing) a staggered board may be partly due However, Q regressions that include interactions with
to an anticipation effect of future takeovers that may or different proxies for M&A intensity provide no evidence
may not materialize (Song and Walkling, 20 0 0; Cremers, consistent with this possibility (see Online Appendix Ta-
Nair and John, 2008; Edmans et al., 2012). In particular, ble A.8). These results are consistent with Edmans et al.
boards could opportunistically use takeover anticipation to (2012), who find that the “trigger effect” of future takeover
Table 5
Portfolio analysis.
This table presents abnormal returns of equally weighted monthly portfolios of firms that have staggered up (in the long
portfolio) and firms that have de-staggered (in the short portfolio) around board staggering and de-staggering events in our
sample of firms during the time period 1978–2015. The long (short) portfolios are composed as follows. For portfolio 6m12
(12m12), we include all stocks of firms that have (de)staggered their boards starting six (12) months before the fiscal year-end
of the year in which the firm has reported its board being (de)staggered for the first time, and hold these stocks for 12 months.
Portfolio 12m24 is constructed like 12m12 portfolio except that stocks are held for 24 months. We use three models: the four-
factor Carhart (1997) model (i.e., momentum, high minus low book-to-market (HML), small minus big (SMB), and market return),
the three-factor Fama-French model (i.e., HML, SMB, and market return), and the market model (i.e., including only the market
return). For each model, we present the returns to the long portfolio, short portfolio, and long minus short portfolio. T-statistics
(in their absolute value) are based on robust standard errors and presented in parentheses below the coefficients. Statistical
significance of the coefficients is indicated at the 1%, 5%, and 10% levels by ∗∗∗ , ∗∗ , and ∗ , respectively. The number of stocks in
the long and short portfolios is averaged across all months.
Four-factor model Three-factor model Market factor model
Portfolio “6m12”
Long Short Long–Short Long Short Long–Short Long Short Long–Short
∗ ∗∗
Alpha (monthly) 0.449 0.145 0.361 0.415 0.063 0.41 0.652 0.175 0.442
(1.68) (0.56) (0.87) (1.53) (0.25) (1.00) (2.26) (0.66) (1.08)
Average # firms 12 19.1 − 12 19.1 − 12 19.1 −
N 355 250 231 355 250 231 355 250 231

Adj. R2 0.562 0.641 0.019 0.562 0.626 0.017 0.491 0.583 0.004
Alpha (monthly) 0.581∗ −0.184 1.086∗∗ 0.455 −0.296 1.136∗∗ 0.598∗∗ −0.175 1.082∗∗
(1.79) (0.77) (2.12) (1.39) (1.25) (2.32) (1.97) (0.72) (2.40)
Average # firms 12 18.9 − 12 18.9 − 12 18.9 −
N 353 256 235 353 256 235 353 256 235

Adj. R2 0.448 0.632 0.006 0.443 0.628 0.009 0.402 0.598 0.007
∗ ∗∗
Alpha (monthly) 0.30 0.045 0.273 0.202 −0.053 0.259 0.424 0.035 0.311
(1.69) (0.23) (0.98) (1.12) (0.29) (0.96) (2.18) (0.18) (1.18)
Average # firms 21.5 28.6 – 21.5 28.6 – 21.5 28.6 –
N 438 397 391 438 397 391 438 397 391

Adj. R2 0.627 0.639 0.006 0.621 0.634 0.008 0.553 0.608 0.001
activity seems to dominate the “anticipation effect,” mean- (down) if the board has good (bad) private information
ing that low firm value tends to attract more takeover that is not yet incorporated into the stock price. Stagger-
activity rather than (the expectation of) more takeover ing up in the case of anticipated long-term value increases
activity resulting in higher firm value. These results are may thus involve a potential trade-off between protecting
also consistent with Bates, Becher and Lemmon (2008), the higher long-term value from being expropriated in a
who find that firms with a staggered board have similar short-term takeover and the potential cost of entrenching
takeover outcomes as firms with unitary boards. For ex- future directors. De-staggering in the case of anticipated
ample, they show that firms with a staggered board, once bad news may facilitate an acquisition before the bad news
targeted, have both a similar likelihood to be acquired and is fully realized or may increase shareholder power to re-
similar bid announcement returns, concluding that “the move directors in the near term.
economic effect of bid deterrence on the value of the firm By construction, this selection mechanism is hard to
is quite small. Overall, the evidence is inconsistent with (dis)prove, as it involves private information. However, the
the conventional wisdom that board classification is an resulting changes in board structure seem inconsistent
anti-takeover device that facilitates managerial entrench- with the interests of currently entrenched insiders who
ment (p. 656).” aim to prioritize their own interests over shareholder in-
terests. If directors want to adopt a staggered board against
3.3.2. Private information of (entrenched?) boards shareholder interests, then the right circumstances to do so
A further selection mechanism consistent with our re- would seem to be either when directors have private in-
sults is that firms could be more likely to stagger up formation of future bad news or right after (rather than
Table 6
Firm value and staggered boards: reverse causality tests.
This table presents regression results for the adoption (columns 1–3) and removal (columns 4–6) of a staggered
board as a function of the valuation of the firm (as captured by Q[t- 1 ] ) plus other characteristics. The 1978–2015 sample
for columns 4–6 (1–3) includes all firms that do (not) have a staggered board up until (and including) the year in which
they remove (adopt) the staggered board if there is any such change, and are dropped from the sample afterwards.
We use the Cox proportional hazard model (see Greene, 20 0 0) and report the marginal effects using robust standard
errors clustered at firm level (after standardizing the continuous variables to have zero mean and unit variance). The
model includes the following control variables: Q[t- 1 ] , Ln(Assets)[t- 1 ] , Delaware incorporation[t- 1 ] , ROA[t- 1 ] , CAPX/Assets[t- 1 ] ,
R&D/ Sales[t- 1 ] , and Industry M&A volume[t- 1 ] . T-statistics (in their absolute value) are based on robust standard errors
clustered by firm and presented in parentheses below the coefficients. Statistical significance of the coefficients is
indicated at the 1%, 5%, and 10% levels by ∗∗∗ , ∗∗ , and ∗ , respectively. All variables are defined in Table 1.
Cox models Cox models

Dep. variable: Pr (Stagger in period t) Pr (De-stagger in period t)
1978–2015 1978–1995 1996–2015 1978–2015 1978–1995 1996–2015
Variables (1) (2) (3) (4) (5) (6)
Q[t- 1 ] −0.844∗∗∗ −0.347∗∗∗ −0.196 −0.12 0.246 −0.127

(7.59) (4.54) (1.00) (1.48) (1.16) (1.55)
Ln(Assets) [t- 1 ] 0.106∗ 0.054 0.122 0.769∗∗∗ 0.343 0.806∗∗∗
(1.70) (0.89) (0.97) (9.99) (0.88) (12.17)
Delaware incorporation[t- 1 ] −0.24∗∗∗ 0.022 −0.625∗∗∗ 0.005 -0.083 0.063
(2.17) (0.18) (2.52) (0.04) (0.15) (0.56)
ROA[t- 1 ] 0.36∗∗∗ 0.227∗∗∗ −0.07 0.10 −0.658 0.142
(4.80) (2.77) (0.36) (1.16) (1.51) (1.80)
CAPX/Assets[t- 1 ] 0.122∗∗∗ 0.156∗∗∗ −0.061 0.073 0.245 0.021
(2.66) (2.69) (0.51) (1.53) (1.36) (0.43)
R&D/ Sales[t- 1 ] −0.083 −0.077 −0.196 0.015 −0.539 0.077
(1.18) (1.36) (1.14) (0.22) (1.62) (1.11)
Industry M&A volume[t- 1 ] 0.009 0.051 0.175∗ −0.155 0.065 −0.177
(0.14) (1.03) (1.75) (1.59) (0.32) (1.66)
# of firms in regression 1,683 1,003 1,035 1,513 607 1,476

N 15,661 7,376 8,406 14,587 4,275 12,338
Pseudo R2 0.038 0.012 0.024 0.038 0.043 0.042
before) good news is released that increases the stock until they adopt (remove) a staggered board.10 In each
price. Indeed, if directors are entrenched and only willing specification, all variables are standardized to have zero
to give up staggered boards under strong shareholder pres- mean and unit variance so that the presented marginal
sure, we would expect boards to be more likely to give in likelihoods can directly indicate the economic magnitude
to such pressure after, rather than before, the realization of a standard deviation change in the independent variable.
of privately anticipated bad news, in the hope that the bad As shown in column 1 in Table 6, the coefficient of
news will not materialize. lagged Q indicates that a standard deviation increase in
Setting aside the question of whether selection based firm value is associated with a decrease of 84% in the
on private information is consistent with the entrenchment probability of staggering up (t-stat. =7.59). This suggests
view of staggered boards, we empirically verify that our that the choice of staggering up is primarily made by
results do not seem driven by more entrenched boards as firms with a relatively low valuation before they adopt a
proxied by a high G-Index score or a poison pill (see Online staggered board, and helps reconcile the negative cross-
Appendix Table A.8). Further, using CEO turnover data from sectional association to our evidence of a positive time se-
Jenter and Kanaan (2015), we show in Online Appendix Ta- ries association. Indeed, the financial value of firms that
ble A.9 that having a staggered board is unrelated to the staggered up in our sample increased after staggering up,
likelihood to fire the CEO, which again seems inconsistent but is insufficient to erode the pre-existing value differ-
with staggered boards promoting entrenchment. ence with the other firms in their industry. Results using a
random probit model are similar for the decision to adopt
3.3.3. Reverse causality a staggered board for the full time period, although with
Another possible self-selection mechanism is reverse reduced economic magnitude.11 Concerning de-staggering
causality. In this case, having a relatively low firm value
induces firms to adopt a staggered board (rather than a
staggered board causing a low firm value), thereby pro- 10
Online Appendix Table A.10 shows the analogous results using probit
viding explanatory power for the cross-sectional result that models, where the marginal likelihood of lagged Q is negative but only
firms with staggered boards tend to have low firm values. significant in the period 1978–1995.
11
Table 6 presents the results of Cox proportional hazard Our reverse causality conjecture is confirmed when we add the one-
year lagged Q as an additional control to the pooled panel regressions
model (Greene, 20 0 0, 20 04) regressions explaining the with industry fixed effects regressions of Table 3, presented in Online
adoption (removal) of a staggered board, including in the Appendix Table A.4 in column 6. If reverse causality affects the cross-
sample all firms that do not (do) have a staggered board sectional results, we expect the identified negative impact to become con-
decisions, columns 3 and 4 of Table 6 show statistically in- nificant differences (even at the 10% level) across the two
significant results, indicating that firm value does not reli- groups in any of the matched samples.
ably predict de-staggering, which is inconsistent with the The matched sample results in Table 7 are quite simi-
reverse causality argument. lar to the full sample results in Tables 4 and 5. As shown
in Panel A, the coefficient of Staggered board from the
3.4. Changes in Q using matched samples pooled panel regressions with firm fixed effects and the
standard controls is statistically significant at the 5% level
Motivated by the results in Section 3.3, we next incor- for all four matching methods. Further, the results in Panel
porate possible selection effects through the creation of a B again confirm our main result that the financial value
matched sample for all firm-years that have either stag- of firms that adopt (remove) a staggered board goes up
gered up or staggered down (i.e., the treatment firms), (down), as the coefficient on ࢞Staggered board remains
where the control firms in the matched sample do not positive and statistically significant across all specifications
change their staggered board structure in the year that the and matching procedures.
treatment firms change their board structure (the event
year). In creating various matched samples, our aim is to 3.5. System GMM estimator
select control firms with similar firm- and industry-level
characteristics as the treatment firms, in order to increase Another possibility is that changes in board structure
the likelihood that unobserved endogeneity similarly af- might reflect a dynamic endogeneity. We consider this
fects both groups. by applying the dynamic system GMM estimator pro-
We consider four different methods to assign con- posed by Arellano and Bover (1995) and Blundell and Bond
trol firms to the treatment firms: (1) propensity score (1998) and implemented in a corporate finance setting by
matching using the probit models for staggering and Wintoki et al. (2012), among others. This methodology es-
de-staggering decisions, respectively; (2) propensity score timates a system in which firm value, board structure, and
matching using the Cox proportional hazard model of other key corporate characteristics are all jointly endoge-
columns 1 and 4 of Table 6 for staggering and de- nous and dynamically interrelated. Indeed, our previous
staggering decisions, respectively; (3) nearest-neighbor probit and Cox regression results indicate that staggering
matching (Abadie and Imbens, 2006) using the main char- up decisions are strongly related to past performance and
acteristics that are significantly related to staggering and other firm- and industry-level characteristics. As a result,
de-staggering decisions in Table 6, namely, lagged Q, prof- it is useful to show that our results are robust to im-
itability, capital expenditures, and R&D; and (4) radius plementing a methodology that explicitly incorporates dy-
matching, where treated firms are matched to one or more namic and simultaneous endogeneity, as well as unobserv-
control firms having a propensity score within a prede- able heterogeneity.
fined radius (Dehejia and Wahba, 2002).12 In all of these The disadvantages of the GMM approach are that re-
methods, for each treatment firm changing its board struc- sults may be sensitive to the choice of instruments, and,
ture, we choose a control firm in the same Fama-French 49 more importantly, that weak instruments could bias the
industry group, although results are similar if we do not results. We try to mitigate these concerns by using a vari-
require industries to be the same. ety of specifications with different sets of instruments for
For the various matched samples, we rerun the firm which statistical tests reject that the particular set of in-
fixed effects pooled panel regressions we used for col- struments is weak. Another limitation is that this method-
umn 1 of Table 4 (reporting the results in Panel A of ology effectively considers only the variation in board
Table 7) and the change-in-Q regressions for Table 5 (re- structure that is related to variation in observable charac-
porting the results in Panel B of Table 7). For the pooled teristics (whereas many changes in board structure may be
panel regressions, we only use observations for the two driven by variables that are hard to observe), such that we
years preceding and the two years following the event year present these results mainly as a robustness check rather
in which the treatment firms change their board struc- than as showing causality.
ture. While doing so limits the number of observations, Our implementation closely follows Hoechle et al.
it also decreases the likelihood of other changes rendering (2012), who apply the dynamic system GMM estimator to
treatment and control firms less comparable.13 Further, the model the diversification discount. First, we choose the set
treatment and control firms have similar characteristics, as of endogenous variables to include: (i) firm value (as prox-
none of the variables have economically or statistically sig- ied by Q, and its first and second lag); (ii) having a stag-
gered board; (iii) profitability (as proxied by ROA); (iv) cap-
ital expenditures (CAPX/Assets); and (v) research and de-
siderably weaker once we control for lagged firm value. Consistent with
velopment expenditures (R&D/Sales), where the last three
this, the cross-sectional association of Staggered board and firm value be-
comes insignificant once we control for lagged firm value.
variables are included because they help predict changes
12
We use the probit models in columns 1 and 4 in Online Appendix to board structure (see Table 6). Each endogenous vari-
Table A.10. Following Guo and Masulis (2015), we use a radius (or caliper) able is modeled in the level equation as depending on
of 0.15. Our results are robust to an alternative radius of 0.10. A benefit of (i) the first difference of the fourth lag of the variable at
the radius matching is that it allows for the use of additional control firm
hand and the other endogenous variables (which within
matches.
13
Results are similar if we use the three years preceding and following
the system are “predetermined” and therefore constitute
the event year, or if we use one year preceding and following the event valid instruments), (ii) the standard pre-determined con-
year, or if we simply use all available observations. trols (namely, year fixed effects, Delaware incorporation,
Table 7
Firm value and staggered boards: firm fixed effects in matched samples.
The results in Panel A are from an annual pooled panel Q regression on Staggered board with firm and year dummies and
control variables, as in Table 3. We use different matching procedures for the sample for each column. The sample includes
treated firms (i.e., firm that either stagger up or de-stagger) and their matched peers, in a sample starting at t-2 and ending
t + 2 years around the event year. Panel B presents pooled panel first difference regressions with the dependent variable being
the change in Q from t to t+ 1 in Column 1 (i.e., ࢞ Q[ t , t +1] ), the change in Q from t to t+ 2 in Column 2 (i.e., ࢞ Q[ t , t +2] ), etc.
The dependent variables have been demeaned with their annual cross-sectional averages. We use matched samples that consist
of treatment firms that change their board structure in year t and control firms that do not change their board structure in year
t (and have the same board structure before the change). We consider four different methods to construct matched samples. In
both samples, we include the following control variables: Staggered board[t- 1 ] , Ln (Assets)[t- 1 ] , Delaware incorporation[t- 1 ] , ROA[t- 1 ] ,
CAPX/Assets[t- 1 ] , R&D/ Sales[t- 1 ] , and Industry M&A volume[t- 1 ] . The time period is 1978–2015. All variables are defined in Table 1.
Included but not shown are industry fixed effects using the Fama-French 49 industry definitions. T-statistics (in their absolute
value) are based on robust standard errors clustered by firm and presented in parentheses below the coefficients. Statistical
significance of the coefficients is indicated at the 1%, 5%, and 10% levels by ∗∗∗ , ∗∗ , and ∗ , respectively.
Panel A:
Dep. variable: Q[t]
Variables (1) (2) (3) (4)
Matching procedure: Propensity score based Propensity score Nearest-neighbor Radius

on probit based on Cox matching matching
Staggered board[t- 1 ] 0.067∗∗ 0.066∗∗ 0.045∗∗ 0.079∗∗

(firm cluster) (2.98) (2.44) (2.09) (3.13)
Ln (Assets)[t- 1 ] −0.175∗∗∗ −0.202 −0.192∗∗∗ −0.227∗∗∗
(4.67) (3.26) (5.27) (4.88)
ROA[t- 1 ] 1.690∗∗∗ 1.858 1.851∗∗∗ 2.296∗∗∗
(5.97) (4.64) (6.39) (7.65)
CAPX/Assets[t- 1 ] 0.176 0.060 0.095 0.250
(0.47) (0.11) (0.28) (0.75)
R&D/ Sales[t- 1 ] 0.357 −0.250 −0.383 1.268
(0.34) (0.20) (0.45) (0.80)
Industry M&A volume[t- 1 ] −0.016 0.102 −0.110 −0.035
(0.13) (0.52) (0.75) (0.19)
N 5,845 3,164 5,923 6,081

Adjusted R2 0.794 0.771 0.786 0.750
Panel B:
Matching method Dep. variable: ࢞ Q[t, t+ 1 ] ࢞ Q[t, t+ 2 ] ࢞ Q[t, t+ 3 ]
(1) (2) (3)

Propensity score based on probit likelihood
࢞ Staggered board[t- 1 , t] 0.029∗∗ 0.041∗∗ 0.057∗∗

(2.16) (2.25) (2.32)
Propensity score based on Cox proportional hazard
࢞ Staggered board[t- 1 , t] 0.025∗ 0.039∗∗ 0.052∗∗

(1.89) (2.13) (2.10)
Nearest-neighbor match

(2.46) (2.39) (2.30)
Radius matching using propensity score based on probit likelihood

(1.95) (2.19) (2.11)
Ln(Assets), and Industry M&A), plus (iii) a set of exoge- M&A), plus the first differences of the exogenous instru-
nous instruments, described in more detail below and in ments described below.
Table 1. We use up to four exogenous industry-level instruments
Second, we take the first difference (in the differenced (plus their squared values) in our various specifications,
equation of the GMM system) of all endogenous vari- each of which relate to changes in board structure and
ables in order to control for unobserved heterogeneity plausibly only affect firm value through changes in board
and reduce potential bias from omitted variables. We in- structure. These instruments include the percentage of to-
strument such endogenous variables with the fourth lag tal sales in the firm’s industry for firms that (1) have a
of the variable at hand and the other endogenous vari- staggered board, (2) staggered their board in the past two
ables, as well as the first differences of the pre-determined years, (3) de-staggered their board in the past two years,
controls (Delaware incorporation, Ln(Assets), and Industry and (4) are conglomerates. The first three instruments ex-
Table 8
Firm value and staggered boards in a dynamic GMM framework.
This table presents results for annual pooled system GMM Q regressions on Staggered board and
control variables, allowing for two lags of the dependent variable. In this system GMM estima-
tor, we assume that all control variables are endogenous, except for Delaware incorporation[t- 1 ] ,
Ln(Assets)[t- 1 ] , and Industry M&A volume[t- 1 ] . The data sample is 1978–2015. The instruments used
in the system GMM estimation are listed in Online Appendix Table A.8, which also presents the
first stage results. The AR(1) and AR(2) are tests for first-order and second-order serial correlation in
the first-differenced residuals, under the null of no serial correlation. The Hansen test of over iden-
tification is under the null that all instruments are valid. The Diff-in-Hansen test of exogeneity is
under the null that instruments used for the equations in levels are exogenous (Roodman, 2009). All
variables are defined in Table 1. T-statistics (in their absolute value) are based on robust standard
errors and presented in parentheses below the coefficients. Statistical significance of the coefficients
is indicated at the 1%, 5%, and 10% levels by ∗∗∗ , ∗∗ , and ∗ , respectively.
Dep. variable: Q[t]
Variables (1) (2) (3) (4)

∗∗∗ ∗∗∗ ∗∗∗
Q[t- 1 ] 1.048 1.049 1.029 1.031∗∗∗
(4.36) (4.50) (4.25) (4.47)
Q[t-2] −0.134 −0.141 −0.129 −0.133
(0.87) (0.94) (0.84) (0.91)
Staggered board[t- 1 ] 0.041∗∗∗ 0.036∗∗∗ 0.052∗∗∗ 0.048∗∗∗
(3.04) (2.69) (2.99) (2.78)
Ln (Assets)[t- 1 ] 0.006 0.005 0.005 0.005
(0.97) (0.84) (0.79) (0.78)
Delaware incorporation[t- 1 ] −0.021∗∗ −0.018∗ −0.018∗ −0.016∗
(2.13) (1.90) (1.77) (1.66)
ROA[t- 1 ] −0.523 −0.424 −0.436 −0.366
(0.84) (0.68) (0.70) (0.59)
CAPX/Assets[t- 1 ] −0.176 −0.192 −0.03 −0.117
(0.18) (0.21) (0.03) (0.13)
R&D/ Sales[t- 1 ] 2.001∗∗∗ 1.877∗∗ 1.949∗∗∗ 1.877∗∗
(2.60) (2.44) (2.61) (2.51)
Industry M&A volume[t- 1 ] −0.211∗∗ −0.212∗∗ −0.209∗∗ −0.209∗∗
(2.14) (2.17) (2.14) (2.19)
N 25,644 25,644 25,644 25,644

Number of firms 2,581 2,581 2,581 2,581
F-stat (p-value) 363.783 (0.0) 369.265 (0.0) 8872.137 (0.0) 8696.713 (0.0)
p-value for AR(1) 0.003 0.002 0.003 0.002
p-value for AR(2) 0.309 0.28 0.332 0.299
Hansen J (p-value) 6.64 (0.156) 11.03 (0.087) 10.83 (0.119) 10.78 (0.095)
Hansen J test diff. p-value –– 0.345 – 0.281
ploit the empirical finding that changes in board structure 3.6. Changes in the corporate law of Massachusetts
tend to cluster across industries and time. The fourth in-
strument is motivated by the observation that conglomer- On April 17, 1990, Massachusetts introduced new stag-
ate firms are less likely to stagger up because their larger gered board rules in Massachusetts State Bill HB 5640. Un-
size means that they are less likely to be takeover targets der the new rules, Massachusetts-incorporated firms are
(Cremers et al., 2008). required to have a staggered board, which can only be
The coefficients of the GMM system’s equation for time removed (starting after January 1992) with a board vote
variation in firm value are presented in Table 8. 14 While or, alternatively, the vote of a qualified supermajority of
lagged Q is strongly statistically significant, twice-lagged Q shareholders.15 This legislative change produced two rele-
is insignificant, which is consistent with twice-lagged en- vant effects. First, all firms incorporated in Massachusetts
dogenous variables being used as valid instruments. Most had a staggered board for the mandatory initial period
importantly for our purposes, the coefficients of Staggered of 18 months. Second, after January 1992, the provision
board across the various specifications are consistently pos- of a staggered board became a “sticky default” or quasi-
itive, similar in magnitude to those reported in Tables 3 mandatory rule (Ayres, 2012) for Massachusetts firms, as
and 4, and strongly statistically significant. If our instru- switching to a unitary board regime was made consider-
ments are valid, this indicates that our main result is ro- ably more difficult by the new rules.16
bust to incorporating simultaneous dynamic endogeneity
and unobservable heterogeneity.
15
Mass. Ann. Law Sch. 156B, Section 50A (Law Co-op 1996), see
Subramanian (2002, p. 1859).
14 16
Online Appendix Table A.11 shows the first stage results in Panels A Daines, Li, and Wang (2016) also consider the association between
and B, and diagnostic test results in Panel C. Panels D and E show the cor- staggered boards and firm value in the context of the 1990 Massachusetts
relation matrix of the instruments in the level and differenced equations, law change, finding results consistent with the inference that staggered
respectively. boards do not, in general, decrease firm value.
Our analysis focuses on how the value of firms incor- We add a triple interaction of MA incorporated x Post
porated in Massachusetts changed before versus after this 1990 with a dummy indicating that the firm had a stag-
law went into effect, relative to similar firms not incor- gered board in 1989 (SB in 1989). This triple interaction has
porated in Massachusetts. We construct a sample of Mas- an insignificant coefficient in column 2, which shows that
sachusetts firms with the following procedure. We first the law change has a similarly positive association with
identify Massachusetts firms from the January 1989 Com- firm value for firms in Massachusetts with and without
pact Disclosure disc and hand-collect data from SEC form a staggered board prior to the new law. This result does
DEF 14A, coding their board structure in the year be- not support the notion that forcing a switch to staggered
fore the new rules went into effect. We then construct a boards generally increases firm value. Rather, it raises the
matched sample of Massachusetts firms and control firms possibility that Massachusetts firms in general—both with
from our main data set that are not incorporated in Mas- and without a staggered board prior to the law—were af-
sachusetts. fected by some contemporaneous effect after 1989 that did
A significant limitation is that the number of firms in- not affect the control firms. Therefore, on the one hand,
corporated in Massachusetts is fairly small. For our basic the results in column 2 strongly contradict the view that
matched sample, we require that the relative difference in staggered boards are detrimental to firm value or that hav-
Q and log assets between the Massachusetts firm and the ing a staggered board is generally a sign of an entrenched
control firm is minimized and at most 30% (above or be- board. On the other hand, it does not provide strong ev-
low), as well as that both firms have an identical board idence that mandating staggered boards on average helps
structure (i.e., unitary or staggered board). Under these increase firm value.
constraints, we obtain a basic sample of 79 firms incor- Next, we employ firm rather than industry fixed effects.
porated in Massachusetts, out of which 55 have a unitary The results in columns 3 and 4 remain quite similar, sug-
board and 24 have a staggered board. For robustness, we gesting that they are not driven by unobserved firm char-
also use a more restrictive matched sample, for which we acteristics that do not vary much during 1988–1992. For
require that the relative difference in Q and log assets be- example, the coefficient on MA incorporated x Post 1990 in
tween the Massachusetts and control firms is at most 20% column 3 equals 0.200 (t-stat. =2.19), which indicates that
(above or below) and that both the Massachusetts firms Massachusetts firms with a mandatory staggered board in
and control firms have no missing data in the five-year pe- 1991 and 1992 increased more in value than their con-
riod (1988–1992) surrounding the new Massachusetts rules trol firms. Finally, the results using the restricted matched
(so that we can estimate the change in value over this pe- sample in columns 5–8 are similar to those for the basic
riod more accurately). This restricted matched sample con- matched sample.
sists of 68 Massachusetts firms, of which 44 have a unitary
board and 24 have a staggered board.17 3.7. Other evidence on causality from the literature
The results for pooled panel Q regressions using the
basic matched sample are presented in columns 1–4 of In this section, we consider four closely related stud-
Table 9, and using the restricted matched sample in ies, each addressing the endogeneity problem in the asso-
columns 5–8. We regress Tobin’s Q on a dummy indicat- ciation between firm value and staggered boards in a dif-
ing whether the firm is incorporated in Massachusetts (MA ferent way. Overall, the results of these studies are mixed,
incorporated), the interaction of MA incorporated with a as two of them report evidence suggesting that staggered
dummy indicating the period after the law change (i.e., boards improve firm value, while the other two indicate
for 1991 and 1992, Post 1990), a set of standard controls, the opposite.
as well as year and industry fixed effects using four-digit Larcker, Ormazabal, and Taylor (2011, p. 432) consider
SIC codes and present the results in column 1. The coeffi- the stock market reaction to 18 events concerning corpo-
cient of the interaction MA incorporated x Post 1990 equals rate governance regulations in 20 07–20 09, which they ar-
0.191 (t-stat. = 1.98). This result indicates that firms that gue collectively “represent an exogenous shock to equilib-
were incorporated in Massachusetts (and thus had a quasi- rium governance practices.” They show a negative stock re-
mandatory staggered board in 1991 and 1992) had a larger turn around proposal announcements for proxy access re-
firm value than their control firms after the law change. form, including a proposal to eliminate staggered boards,
As the average Q in this sample equals 1.53, the coeffi- observing that their results are “inconsistent with the mar-
cient suggests that firms incorporated in Massachusetts in- ket viewing the elimination of staggered boards as value
creased about 10% in value in the two years following the increasing. If anything, the results suggest the opposite.
implementation of the law, relative to their control firms. The elimination of the option to have a staggered board
is value decreasing” (p. 433).
Cunat, Gine and Guadalupe (2012) consider both short-
17
Online Appendix Table A.15, Panel C, shows the number of firms
term abnormal returns and long-term changes in perfor-
at various stages of the sample construction, for both samples. Panel D
shows descriptive statistics for the restricted matched sample, which are mance in a study focusing on close votes on shareholder-
very similar for the basic matched sample. As shown in Panel E, the Mas- sponsored proposals, describing the passing of such pro-
sachusetts firms have economically similar Tobin’s Q, size, and profitabil- posals as “akin to an independent random event (it
ity as their controls before the change in Massachusetts’ corporate law. is ‘locally’ exogenous) and therefore uncorrelated with
Statistically, the differences are significant for CAPX/Assets and R&D/Assets,
but these differences seem economically minor, while the difference for Q
firm characteristics” (p. 1944). Among other results, they
is statistically significant for the basic matched sample, and insignificant show significant long-term performance improvements af-
for the restricted matched sample. ter shareholders narrowly vote in favor of shareholder-
Table 9
Firm value and staggered boards – Massachusetts incorporated firms.
This table presents results for annual pooled panel Q regressions relating firm value to the enactment of the Massachusetts staggered board law (i.e., Mas-
sachusetts State Bill HB5640) in 1990. We include firms incorporated in the state of Massachusetts and their matched control firms of non-Massachusetts
incorporated firms in our main database selected as the closest match based on Q, logarithm of asset size, and board structure (staggered or unitary board)
as of 1988 and 1989. MA incorporated takes the value of one if the company is incorporated in the state of Massachusetts (and zero otherwise), and Post
1990, equals one if the observation is after the passage of the Massachusetts staggered board bill, and zero otherwise. SB in 1989 is an indicator variable
that equals one if the firm had a staggered board in 1989. All specifications include time fixed effects, while the regressions for columns 1–2 and 5–6 in-
clude industry fixed effects using four-digit SIC code industry groups, and columns 3–4 and 7–8 include firm fixed effects. All columns also include control
variables Ln (Assets)[t- 1 ] , ROA[t- 1 ] , CAPX/Assets[t- 1 ] , R&D/ Sales[t- 1 ] , and Industry M&A volume[t- 1 ] . The sample includes two years before and two years after
the year of the law’s passage, i.e., 1988–1992. We use the basic matched sample for columns 1–4, and the restricted matched sample with more stringent
matching criteria for columns 5–8. All variables are defined in Table 1. T-statistics (in their absolute value) are based on robust standard errors clustered
by firm and presented in parentheses below the coefficients. Statistical significance of the coefficients is indicated at the 1%, 5%, and 10% levels by ∗∗∗ , ∗∗ ,
and ∗ , respectively.
Dep. variable: Q[t]
Basic (unbalanced) matched sample Restricted (balanced) matched sample

Variables (1) (2) (3) (4) (5) (6) (7) (8)
MA incorporated x Post 1990 0.191∗∗ 0.204∗ 0.200∗∗ 0.218∗ 0.150∗ 0.150∗ 0.17∗∗ 0.175∗
(1.98) (1.72) (2.19) (1.84) (1.92) (1.68) (2.12) (1.77)
MA incorporated −0.064 −0.088 0.130 0.114
(0.37) (0.42) (0.69) (0.52)
SB in 1989 x MA incorporated x Post 1990 −0.045 −0.056 −0.001 −0.014
(0.31) (0.37) (0.01) (0.10)
SB in 1989 x MA incorporated 0.074 0.045
(0.31) (0.17)
SB in 1989 x Post 1990 −0.047 −0.015
(0.27) (0.08)
Ln(Assets)[t- 1 ] −0.039 −0.04 −0.015 −0.013 −0.021 −0.021 0.034 0.036
(0.70) (0.71) (0.06) (0.06) (0.36) (0.36) (0.17) (0.19)
ROA[t- 1 ] 2.011∗∗∗ 2.013∗∗∗ 0.416 0.445 2.344∗∗∗ 2.35∗∗∗ 1.179 1.188
(3.06) (3.16) (0.54) (0.57) (3.63) (3.70) (1.52) (1.5)
CAPX/Assets[t- 1 ] 0.371 0.344 1.944 1.948 −0.988 −1.024 0.823 0.825
(0.35) (0.32) (1.64) (1.65) (1.11) (1.15) (0.96) (0.95)
R&D/ Sales[t- 1 ] 2.305 2.319 0.303 0.404 4.548 4.478 2.019 2.057
(0.78) (0.82) (0.12) (0.15) (1.47) (1.55) (1.03) (1.04)
Industry M&A volume[t- 1 ] −0.338 −0.332 −0.184 −0.181 −0.019 −0.015 0.016 0.014
(0.86) (0.86) (0.53) (0.52) (0.06) (0.05) (0.06) (0.05)
N 552 552 555 555 452 452 452 452

Adjusted R2 0.609 0.607 0.816 0.816 0.642 0.639 0.848 0.847
Fixed effects Year, Year, Year, Year, Year, Year, Year, Year,
Industry Industry Firm Firm Industry Industry Firm Firm
sponsored proposals to reduce antitakeover provisions, in- for firms more generally, and (3) shareholder-sponsored
cluding proposals to remove a staggered board. However, proposals are generally not binding, since dismantling a
they do not separate proposals concerning staggered board staggered board commonly requires both shareholder and
removals from those pertaining to the removal of other an- board approval. As a result, the market reaction after
titakeover provisions. close votes of shareholder-sponsored proposals likely re-
In Online Appendix Table A.12, we replicate the main Q flects market learning about a combination of factors, in-
results of Cunat et al. (2012), and show that once we sepa- cluding the potential repeal of a staggered board, existence
rate out proposals to remove different antitakeover mea- of shareholder pressure, and disagreement among share-
sures, we find no statistically significant evidence of in- holders.
creases in Q following narrow shareholder approval of pro- Cohen and Wang (2013) study two systematic events
posals to remove staggered boards. Further, results using related to staggered boards that involved Air Products’ at-
close votes on shareholder-sponsored proposals need to be tempt to acquire Airgas in 2010. The first is an Octo-
interpreted with caution, for three reasons: (1) these re- ber 2010 ruling by the Delaware Chancery Court, which
sults are arguably driven by relatively few observations,18 temporarily weakened the insulation force of staggered
(2) corporations where these votes are close may be dif- boards. The second is a subsequent ruling by the Delaware
ferent from other firms and this may introduce selection Supreme Court in November 2010, which reversed the
effects when considering the relevance of these results Chancery Court’s decision. Combining the two-day abnor-
mal returns of these events, Cohen and Wang (2013) report
evidence (statistically significant at 10%) that is consistent
with staggered boards causing a lower firm value. How-
18
Defining “close votes” as shareholder votes of 49–51%, there are 31
ever, while we can replicate the main results as reported
close votes for proposals to remove a staggered board (out of 445 total)
and 45 for proposals to remove other antitakeover provisions (out of 881
in their paper, our replications in Online Appendix Table
total). A.13 suggest that these results are not robust to either en-
larging the sample by adding more data through hand col- In order to test these hypotheses, we present results
lection or using different industry fixed effects.19 using four sets of proxies, although the related channels
Lastly, Johnson et al. (2015) follow Coates (2001), who are non-exclusive and difficult to separate empirically.20 In
argued that the identity of an initial public offering (IPO)’s particular, the first two sets of proxies, namely, for (1) the
law firm helps to explain variation in takeover defenses. prevalence of long-term innovation and (2) firms with as-
Relatedly, Johnson et al. employed the identity of an sets with more asymmetric information or greater com-
IPO firm’s law firm as an instrument for shareholder plexity, are meant to test both hypotheses simultaneously.
rights provisions in corporate charters and bylaws, hand- The third and fourth sets of proxies, namely, for (3) the
collecting this information for a large set of IPO firms for likelihood that the firm is subject to market pressure from
1997–2005. Using the IPO law firm instrument, they find short-term investors and (4) the importance of particular
that firm value at the IPO is positively related to having a stakeholder relationships, are chosen to separate the my-
staggered board for firms with stronger stakeholder rela- opic and bonding hypotheses.
tionships. While this particular instrument seems applica- At the same time, we acknowledge that the decline in
ble only to IPO firms rather than the more mature firms firm value after de-staggering remains challenging to inter-
in our sample, it is noteworthy that using a different set pret, under either channel. Under the myopic market hy-
of firms and employing a novel instrument yields cross- pothesis, de-staggering may be due to pressure from out-
sectional results that are consistent with our time series side, myopic investors. Firms with worse prospects could
results. This is also consistent with Johnson et al. (2016), be more likely to give in, such that the subsequent decline
who find that staggered boards have a positive effect on in value would then reflect selection rather than causal-
firm value in young (rather than just IPO) firms with im- ity. Under the bonding channel, it is similarly possible that
portant stakeholder relationships. firms that agree to de-stagger their board signal that the
firm expects lower future surpluses from its stakeholder
3.8. The channels through which staggered boards relate to relationships, which would explain the decline in firm
firm value value after de-staggering (rather than the de-staggering it-
self causing the decrease in firm value).
An alternative to resolving the endogeneity of board
structure is to investigate two channels through which 3.8.1. Research and development investments
staggered boards could be associated with long-term firm Under the myopic markets hypothesis, the pressure on
value. First, Stein (1988, 1989) hypothesizes that markets management towards myopic investment decisions seems
may be myopic, such that shareholders with a short hori- most important for firms with projects that require long-
zon may mistakenly seek to replace management or agree term investments, such as investments in R&D (Bushee,
to a suboptimal takeover, especially in firms with signif- 1998; Chan, Lakonishok and Sougiannis, 2001; Eberhart,
icant asymmetric information. In response to this mar- Maxwell and Siddique, 2004). As a result, under the my-
ket pressure from short-term investors, management may opic markets hypothesis, one would expect firms with sig-
develop incentives towards myopic investment behavior. nificant R&D investments to benefit more from a staggered
Staggered boards could help prevent this market ineffi- board. At the same time, the commitment device a stag-
ciency by reducing market pressure. gered board provides toward the firm’s stakeholders under
Second, Knoeber (1986), Shleifer and Summers (1988), the bonding hypothesis could also matter more for such
and Johnson et al. (2015, 2016) consider the bonding hy- firms. As innovation often requires firm-specific invest-
pothesis, which pertains to firms with important relation- ments by top employees, suppliers, customers, or strategic
ships with stakeholders such as top employees and large alliance partners, a staggered board could be useful to pre-
customers. As these stakeholders make firm-specific in- vent the ex post expropriation of the stakeholders’ firm-
vestments in their relationship with the firm, they risk specific investments in firms more engaged in R&D invest-
expropriation if the firm is taken over. In this case, stag- ments. Therefore, our three proxies for the importance of
gered boards could efficiently serve as a commitment de- R&D are useful to test both hypotheses: (1) R&D/Sales, the
vice towards more stable stakeholder relationships, lower- intensity of corporate expenditures on research and de-
ing the cost of contracting with these stakeholders, and fa- velopment activities; (2) R&D quotient, a measure of the
cilitating investments in value-creating projects (see Arlen, firm-specific output elasticity of R&D proposed by Knott
2007; and Cremers, Nair and Peyer, 2008, for further dis- (2008); and (3) Ranked patent citation count from the Na-
cussion). tional Bureau of Economic Research (NBER) U.S. Patent Ci-
tations data file, measuring the relative importance of the
19
The sensitivity of these results to changing the sample construction firm’s innovative research assets.
and industry fixed effects is perhaps unexpected, as the sample in Cohen
and Wang (2013) is fairly small, with 77 treatment and 62 control firms,
20
and considering the noise associated with two-day abnormal returns. On the one hand, firms in which specific investments by stakehold-
Amihud and Stoyanov (2017) find that the Cohen and Wang (2013) re- ers are more important may also be subject to more pressure from my-
sults become insignificant after they exclude from the sample stocks with opic investors. For example, stakeholder relationships are more likely to
a price below $1, market capitalization below $10 million or firms that be characterized by asymmetric information, such that investors may be
are delisted and traded only over-the-counter. On the other hand, Cohen relatively less informed at firms with stronger stakeholder relationships.
and Wang’s (2017) response counters that their results, especially using On the other hand, longer-term projects that are subject to pressure from
expanded samples, are robust to the variations in sample and methodol- myopic investors are more likely to require firm-specific investments by
ogy considered by Amihud and Stoyanov (2017). stakeholders.
Table 10
Firm value and staggered boards: interactions with investments in innovation.
This table presents the results of a time series analysis as in Table 3 that includes interactions with vari-
ables that capture investments in innovation. We include the following control variables in the regression: Ln
(Assets)[t- 1 ] , Delaware incorporation[t- 1 ] , ROA[t- 1 ] , CAPX/Assets[t- 1 ] , and R&D/Sales[t- 1 ] , and Industry M&A volume[t- 1 ] ,
which we do not show for brevity (unless a variable is being interacted with Staggered board[t- 1 ] ). The inter-
acted variables include the following: R&D/ Sales[t- 1 ] , Ranked patent citation count[t- 1 ] , and Research quotient[t- 1 ] .
The sample period is 1978–2015, except columns 3 and 4, which sample period is 1978–2007. Estimation is
using pooled panel Tobin’s Q[t] regressions. We include year and firm fixed effects. All interaction and control
variables are defined in Table 1. T-statistics (in their absolute value) are based on robust standard errors clus-
tered by firm and presented in parentheses below the coefficients. Statistical significance of the coefficients is
indicated at the 1%, 5%, and 10% levels by ∗∗∗ , ∗∗ , and ∗ , respectively.
Dep. Variable: Q[t]
Variables (1) (2) (3) (4) (5) (6)
Staggered board[t- 1 ] 0.112∗∗ 0.051∗∗ 0.044 0.041 0.045 0.046

(2.39) (2.06) (1.14) (1.07) (0.98) (1.00)
R&D/ Sales[t- 1 ] 0.644 1.274∗∗
(1.00) (2.46)
Ranked patent citation count[t- 1 ] 0.016 0.127∗∗
(0.18) (2.02)
Research quotient[t- 1 ] 0.236 0.601∗∗∗
(0.92) (2.79)
∗
R&D/ Salest- 1 1.202∗
Staggered board[t- 1 ] (1.85)
Ranked patent citation count[t- 1 ] ∗ 0.204∗
Research quotient[t- 1 ] ∗ 0.849∗∗
N 34,476 34,476 17,778 17,778 14,847 14,847

Adjusted R2 0.713 0.713 0.716 0.716 0.707 0.706
The results for interactions of the Staggered board 3.8.2. Informational complexity
dummy with each of these three proxies in pooled panel The second set of proxies that we use to verify the two
Q regressions with year and firm fixed effects plus our channels through which staggered boards could be associ-
standard controls are presented in Table 10. Consistent ated with long-term firm value consists of two “catch-all”
with both the myopic markets and the bonding channels, measures of the complexity of firm operations and asym-
these results show that changes in staggered board struc- metric information (Core, Holthausen and Larcker, 1999;
tures are more strongly related to firm value for firms Duru, Wang and Zhao, 2013). These measures include Firm
that are more engaged in research and development, have size (the log of total revenue) and Ln(Intangibles assets/Total
more patent citations, or have (arguably) more complex assets), where the latter is calculated as the log of one
operations. The interaction with R&D/Sales (column 1 of minus the ratio of the book value of plant, property, and
Table 10) has a positive, statistically, and economically sig- equipment (PP&E) over total assets. While both proxies
nificant coefficient. Firms whose R&D/Sales is one stan- could indicate firms that are more likely to be subject to
dard deviation higher than the mean experience a 4.2% market pressure from short-term investors, such proxies
(= 1.202∗ 0.056/1.6) higher level of Q after staggering up could also indicate firms where stakeholder relationships
relative to firms whose R&D/Sales is at the mean.21 Sim- are more important. Nevertheless, in general, these prox-
ilarly, firms that have a one standard deviation higher ies are harder to interpret and less directly related to both
Ranked patent citation count have a 2.9% (= 0.204∗ 0.23/1.6) hypotheses than other proxies we use, and so we mainly
higher Q if they stagger up compared to firms with mean show these additional tests as a robustness check.
(i.e., very low) patent counts (column 3). Further, the Re- The results in columns 1–4 of Table 11 indicate that
search quotient and Staggered board interaction has a pos- both firm size and intangibility capture an aspect of the
itive coefficient of 0.849 with a t-stat of 2.13 (column 5), heterogeneity in the association between staggered boards
which indicates that firms whose Research quotient is a and firm value and are broadly supportive of both the
standard deviation higher (than the mean) have a 2.6% (= myopic markets hypothesis and the bonding hypothesis.
0.849∗ 0.049/1.6) higher increase in Q after staggering up, For larger firms and firms with more intangible assets,
compared to firms with an average Research quotient. adoptions (removals) of staggered boards are associated
with larger increases (decreases) in firm value. Firms with
Ln(Intangible assets/Total assets) that is one standard devia-
tion higher than the mean present a 3% (= 0.10∗ 0.48/1.6)
21
higher Q if they stagger up relative to firms with aver-
We obtain this estimation by multiplying the coefficient of the inter-
age intangible assets (column 1). Firms with Firm sales that
acting variable (i.e., 1.202) by the standard deviation of R&D/Sales (0.056),
divided by the average Q (1.6) in the sample for column 1. is one standard deviation higher than the average tend to
Table 11
Firm value and staggered boards: interactions with operational complexity and institutional horizon.
This table presents the results of a time series analysis as in Table 3 that includes interactions with variables that capture operational
complexity and the presence of short-term institutional investors. We include the following control variables in the regression: Ln(Assets)[t- 1 ] ,
Delaware incorporation[t- 1 ] , ROA[t- 1 ] , CAPX/Assets[t- 1 ] , and R&D/Sales[t- 1 ] , and Industry M&A volume[t- 1 ] , which we do not show for brevity. The
interacted variables include the following: Intangible assets/ Total assets[t] , Firm size[t] , Institutional holding duration[t- 1 ] , and Percent transient
institutions[t- 1 ] . The sample period is 1978–2015 for columns 1–4 and 1982–2015 for columns 5–8 due to data availability. Estimation is using
pooled panel Tobin’s Q[t] regressions. We include year and firm fixed effects. All interaction and control variables are defined in Table 1. T-
statistics (in their absolute value) are based on robust standard errors clustered by firm and presented in parentheses below the coefficients.
Statistical significance of the coefficients is indicated at the 1%, 5%, and 10% levels by ∗∗∗ , ∗∗ , and ∗ , respectively.
Dep. variable: Q[t]
Variables (1) (2) (3) (4) (5) (6) (7) (8)
Staggered board[t- 1 ] 0.064∗∗ 0.052∗∗ 0.028 0.049∗∗ -0.183 0.093∗∗ 0.104∗∗∗ 0.086∗∗
(2.38) (2.08) (1.14) (1.97) (−1.29) (2.37) (2.61) (2.34)
Ln(Intangible assets/ Total assets[t- 1 ] ) 0.088∗∗ 0.136∗∗∗
(2.38) (4.32)
Firm sales[t- 1 ] −0.218∗∗∗ −0.201∗∗∗
(11.67) (12.07)
Institutional holding duration[t- 1 ] −0.045∗∗∗ −0.024∗∗
(3.12) (2.13)
Percent transient institutions[t- 1 ] 0.468∗∗∗ 0.420∗∗∗
(5.91) (7.41)
Ln(Intangible assets/Total assets[t- 1 ] ) ∗ 0.102∗∗∗
Firm sales[t- 1 ] ∗ 0.035∗∗
Institutional holding duration[t- 1 ] ∗ 0.040∗∗
Percent transient institutions[t- 1 ] ∗ −0.087
N 34,229 34,229 34,460 34,460 28,298 28,298 29,305 29,305

Adjusted R2 0.713 0.712 0.711 0.710 0.694 0.693 0.694 0.694
have a 3.3% (= 0.035∗ 1.5/1.6) higher Q if they stagger up Using these proxies, Bushee (1998) and Cremers et al.
relative to firms with average size sales (column 3). (2016) find that the presence of transient and short-term
institutional investors, respectively, is associated with my-
opic investment behavior, such as temporary cuts in R&D.
3.8.3. Short-term investors
Accordingly, we use both proxies to test the hypothe-
In order to separately test the myopic market hypoth-
sis that staggered boards can mitigate market pressure
esis, we use two proxies that capture the presence of
towards myopic investment behavior, examining whether
short-term institutional investors, adopting variables that
adoptions (removals) of staggered boards are more strongly
have been related to market pressure on management to-
related to increases (decreases) in value for firms with
wards myopic investment decisions. First, Institutional hold-
more short-term or transient institutional investors.
ing duration measures the average holding duration (across
We interact the Staggered board dummy with Institu-
their entire, aggregate U.S. public equity holdings) for the
tional holding duration and Percent transient Institutions in
institutions currently owning the stock, weighted by the
pooled panel Q regressions with year and firm fixed ef-
amount each institution owns. This proxy was introduced
fects plus our standard controls, with the results reported
in Cremers and Pareek (2015) and employs SEC Form 13F
in columns 5–8 of Table 11, respectively. The interaction of
institutional holdings data from Thomson Reuters. Second,
Staggered board with Institutional holding duration is posi-
Percent transient institutions measures the percentage of in-
tive and significant, such that changes in staggered board
stitutional ownership that can be classified as transient ac-
structure are more strongly related to firm value for firms
cording to Bushee (1998), i.e., institutions that are both
that have longer-term institutional owners. The interaction
well-diversified and have high turnover.22
coefficient of 0.040 suggests that firms whose Institutional
holding duration is one standard deviation higher than the
22
mean experience a 2.2% (= 0.04∗ 0.99/1.8) higher level of
The 13F data start in 1980, but we require at least two years of hold-
ings reports before calculating an institution’s holding duration over the Q after staggering up relative to firms whose Institutional
last five years or the percentage of institutional ownership that is tran- holding duration is at the mean.23 Further, the interaction
sient. The classifications for transient firms are available on Bushee’s web- of Staggered board with Percent transient institutions is neg-
site. Both variables are winsorized at 1%, which is especially important
for Percent transient institutions, which is calculated as the ratio of the
percentage of stocks held by transient institutions over the percentage of
23
stocks held by institutional investors. As shown in Table 2, the average In- We obtain this estimation by multiplying the coefficient of the in-
stitutional holding duration equals 6.8 quarters (with fat tails on both sides teracting variable (i.e., 0.04) by the standard deviation of Institutional
of its distribution), and the average Percent transient institutions equals Holding Duration (0.99), divided by the average Q (1.8) in the sample for
22%. column 1.
ative and statistically insignificant, showing that there is itive association between staggered boards and firm value
no evidence that staggered boards matter more for firms is stronger or only apparent for firms where stakeholder
with more transient institutions.24 Therefore, the evidence relationships appear to be more important.
on the myopic markets hypothesis is mixed, as the results The interaction between Staggered board and Large cus-
for firms with more R&D investments and more informa- tomer has a positive and both statistically and economi-
tional complexity are consistent with the hypothesis, while cally significant coefficient equal to 5.3% (0.083/1.6; t-stat.
the results on short-term investors are not. =2.80, see column 1). This indicates that decisions to adopt
or remove a staggered board are on average associated
3.8.4. Firm-specific stakeholder relationships with substantially larger increases and decreases, respec-
In order to specifically test the bonding hypothesis, we tively, in firm value for firms with a large customer.
lastly consider four different proxies that aim to capture The coefficient in column 3 from the Staggered board
the importance of stakeholder relationships. The first two interaction with Strategic alliance is 3.4% (i.e., 0.055/1.6; t-
proxies are at the firm level and time varying, and are both stat. =1.89, statistically significant at 10%) and shows that
adopted following Johnson et al. (2015). First, Large cus- the positive association between staggered boards and firm
tomer is a proxy for the importance of customers in creat- value is nearly twice as strong for firms engaged in a
ing financial value, measured as a dummy that equals one strategic alliance than for firms without such long-term
if the firm has at least one customer accounting for 10% or engagement. Next, the results in column 5 show that the
more of its sales according to the historic Compustat Seg- interaction between Staggered board and Labor productiv-
ment tapes (available for 1985–2015). Second, Strategic al- ity is also positive and both strongly statistically and eco-
liance measures whether the corporation has a long-term nomically significant. Economically, the coefficient implies
partnership with another business (Bodnaruk, Massa and that if a firm is in an industry where Labor productivity is
Simonov, 2013), and is an indicator variable equal to one if a standard deviation above the average, the adoption of a
the firm participates in one, and zero otherwise, obtained staggered board is associated with a 3.8% greater increase
from the Thomson Reuters SDC M&A database for 1978– in firm value.26
2015. We only include strategic alliances with up to three Lastly, the results in column 7 show that the Staggered
partners (representing 97% of all strategic alliances). Since board interaction with Contract specificity also has a pos-
data prior to 1985 are sparse, we start the Strategic alliance itive and statistically significant coefficient, which implies
sample in 1985. that if a firm is in an industry whose Contract specificity
The third and fourth proxies, Labor productivity and is a standard deviation above the average, the adoption of
Contract specificity, are at the industry level with limited a staggered board is associated with a 3.3% (0.41∗ 0.13/1.6,
and no time variation, respectively. The use of industry- t-stat. of 1.93) higher increase in firm value compared to
level proxies is a useful complement to firm-level proxies, firms in industries with average Contract specificity.27
where selection effects may be more severe. Labor produc-
tivity comes from the U.S. Bureau of Labor Statistics (using 3.8.5. Massachusetts firms
the four-digit SIC code level data) and is available for only In Online Appendix Table A.15, we show the analogous
a subset of firms’ industries. This proxy aims to measure results for all our proxies for the importance of invest-
which industries have a higher marginal product of labor ments in long-term projects such as R&D (in Panel A) and
and, hence, more firm-specific investments by the employ- firm-specific relationships with employees, customers, sup-
ees. Contract specificity is borrowed from Nunn (2007), and pliers, and in strategic alliances (in Panel B) for the re-
measures the fraction of inputs (i.e., products and services) stricted matched sample of Massachusetts and matched
that are not sold on an organized exchange or reference non-Massachusetts firms in 1988–1992 (see Section 3.6).
priced in a trade publication, and for which the market The results are generally consistent with the results in the
thus appears less complete.25 Industries with more con- full sample. In particular, Massachusetts firms that have
tract specificity require more firm-specific investments and higher R&D expenses, more patent citation counts, or have
commitment by suppliers. strategic alliances increased more in value following the
Table 12 presents the results of interacting the Staggered Massachusetts law change that rendered staggered boards
board dummy with each of these four stakeholder rela- quasi-mandatory. The results for the other proxies are sta-
tionship proxies in pooled panel Q regressions with year
and firm fixed effects plus our standard controls. These re-
sults support the bonding hypothesis, as generally the pos- 26
The economic significance of the interacted impact of Labor produc-
tivity and Staggered board on Q is calculated by dividing the coefficient
of 0.10, times the standard deviation of Labor productivity of 0.61, by the
24
We also control for the percentage of institutional ownership and re- sample average Q during 1978–2015 of 1.6. Results using the basic, bal-
port the results in columns 5–8 of Table 11. The results are similar if we anced sample are quite similar.
only use firms with at least 50% institutional ownership or the percent- 27
The results in Online Appendix Table A.14, Panels A and B suggest
age of total transient institutional ownership rather than the percentage that the nine different proxies for the importance of investments in long-
of institutional ownership that is transient. We find that both proxies are term projects and important stakeholder relationships generally have in-
unrelated to adoptions or removals of staggered boards, and that changes significant, inconsistent or weak associations with the decision to adopt
in board structure are unrelated to changes in either proxy (see Online (in Panel A) or remove (in Panel B) a staggered board. This confirms that
Appendix Table A.14, Panels C and D). having a low firm value is what matters most for the decision to adopt
25
Data for Contract specificity are made available at http://scholar. a staggered board. Nonetheless, once a firm has staggered up, firms with
harvard.edu/nunn/pages/data-0 for 1997 and for about a quarter of the more long-term investments and stronger stakeholder relationships ben-
industries in our sample (and is set as missing otherwise). efit the most from the decision to stagger up.
Table 12
Firm value and staggered boards: interactions of Staggered board with proxies for stakeholder commitment.
This table presents results for pooled panel Q regressions with firm and year fixed effects that includes interactions
with variables that proxy for stakeholder commitment. We include the following control variables in the regression: Ln
(Assets)[t- 1 ] , Delaware incorporation[t- 1 ] , ROA[t- 1 ] , CAPX/Assets[t- 1 ] , R&D/Sales[t- 1 ] , and Industry M&A volume[t- 1 ] which we do not
show for brevity (unless a variable is being interacted with Staggered board[t- 1 ] ). The interacted variables include the following:
Large customer[t- 1 ] , Strategic alliance[t- 1 ] , Labor productivity[t- 1 ] , and Contract specificity[t- 1 ] . The sample period is 1985–2015 for
columns 1–4 and 1997–2015 for columns 5 and 6. All variables are defined in Table 1. T-statistics (in their absolute value) are
based on robust standard errors clustered by firm and presented in parentheses below the coefficients. Statistical significance
of the coefficients is indicated at the 1%, 5%, and 10% levels by ∗∗∗ , ∗∗ , and ∗ , respectively.
Dep. variable: Q[t]
Variables (1) (2) (3) (4) (5) (6) (7) (8)
Staggered board[t- 1 ] 0.039 0.066∗∗ 0.058∗ 0.072∗∗ -0.078 0.076∗∗ -0.287∗ 0.091∗
(1.14) (2.12) (1.87) (2.19) (1.45) (2.33) (1.70) (1.81)
Large customer[t- 1 ] −0.093∗∗∗ −0.046∗∗∗
(3.65) (2.77)
Strategic alliance[t- 1 ] −0.031 −0.001
(1.51) (0.05)
Labor productivity[t- 1 ] −0.209∗∗∗ −0.161∗∗∗
(7.56) (6.85)
Contract specificity[t- 1 ] −0.971 −1.006
(1.49) (1.55)
Large customer[t- 1 ] ∗ 0.083∗∗∗
Strategic alliance[t- 1 ] ∗ 0.055∗
Labor productivity[t- 1 ] ∗ 0.099∗∗∗
Contract specificity[t- 1 ] ∗ 0.411∗
Table 3 controls included Yes Yes Yes Yes Yes Yes Yes Yes
N 29,107 29,107 27,782 27,782 27,715 27,715 10,478 10,478
Adjusted R2 0.713 0.713 0.713 0.713 0.714 0.713 0.664 0.664
tistically insignificant, which may be due to relatively little firm value is nuanced and heterogeneous. Our results thus
cross-sectional dispersion in this fairly small sample. challenge the “one-size-fits-all” policy that favors the an-
nual election of directors—and which is currently sup-
ported by many proxy advisory firms and other share-
4. Conclusion holder advocates—as inconsistent with the empirical evi-
dence on staggered boards.
In this paper, we revisit the association between stag-
gered boards and long-term firm value. Using a panel data
set for 1978–2015, we focus on long-term changes in firm References
value around the adoption (removal) of a staggered board.
Abadie, A., Imbens, G., 2006. Large sample properties of matching estima-
We find no support for the entrenchment view that stag- tors for average treatment effects. Econometrica 74, 235–267.
gered boards encourage shirking, empire building, or pri- Adams, R., Hermalin, B.E., Weisbach, M.S., 2010. The role of boards of di-
vate benefits extraction, as we uncover no evidence that rectors in corporate governance. A conceptual framework and survey.
Journal of Economic Literature 48 (1), 58–107.
changes in board structure have a strong or persistently Ahn, S., Shrestha, K., 2013. The differential effects of classified boards on
negative association with changes in firm value. Rather, we firm value. Journal of Banking and Finance 37, 3993–4013.
find that firm value increases (decreases) after firms adopt Amihud, Y., Stoyanov, S., 2017. Do staggered boards harm shareholders?
Journal of Financial Economics 123, 432–439.
(remove) a staggered board for certain subsets of firms, Arellano, M., Bover, O., 1995. Another look at the instrumental variable
namely, those engaged in long-term projects, with impor- estimation of error-component models. Journal of Econometrics 68,
tant stakeholder relationships, or that are more difficult to 29–51.
Arlen, J., 2007. Regulating post-bid embedded defenses: lessons from Ora-
value. cle versus PeopleSoft. Harvard Negotiation Law Review 12 (1), 71–102.
These results are most consistent with the bonding Ayres, I., 2012. Regulating opt-out: an economic theory of altering rules.
hypothesis that staggered boards can serve as an effi- Yale Law Journal 121, 2032–2405.
Barry, J.M., Hatfield, J.W., 2012. Pills and partisans: understanding
cient commitment device towards the firm’s stakehold-
takeover defenses. University of Pennsylvania Law Review 160,
ers. Laffont and Tirole (1988), Knoeber (1986), Shleifer and 633–713.
Summers (1988), and Johnson et al. (2015, 2016) examine Bates, T., Becher, D., Lemmon, M., 2008. Board classification and manage-
rial entrenchment: evidence from the market for corporate control.
the bonding hypothesis and suggest that takeover defenses
Journal of Financial Economics 87, 656–677.
such as the staggered board can lower the cost of contract- Bebchuk, L.A., Cohen, A., 2005. The costs of entrenched boards. Journal of
ing with the firm’s stakeholders and facilitate investments Financial Economics 78, 409–433.
in value-creating long-term projects. Bebchuk, L.A., Cohen, A., Ferrell, A., 2009. What matters in corporate gov-
ernance? Review of Financial Studies 22, 783–827.
Overall, our results are also consistent with the lit- Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in
erature that the relation between board structure and dynamic panel data models. Journal of Econometrics 87, 115–143.
Bodnaruk, A., Massa, M., Simonov, A., 2013. Alliances and corporate gov- Guo, L., Masulis, R., 2015. Board structure and monitoring: new evidence
ernance. Journal of Financial Economics 107, 671–693. from CEO turnovers. Review of Financial Studies 28, 2770–2811.
Bushee, B., 1998. The influence of institutional investors on myopic R&D Hall, B.H., Jaffe, A.B., Trajtenberg, M., 2002. The NBER patent citation data
investment behavior. The Accounting Review 73, 305–333. file: lessons, insights and methodological tools. In: Jaffe, A., Trajten-
Carhart, M.M., 1997. On persistence in mutual fund performance. Journal berg, M. (Eds.), Patents, Citations and Innovations. The MIT Press,
of Finance 52, 57–82. Cambridge, MA, pp. 403–460.
Cen, L., Dasgupta, S., Sen, R., 2016. Discipline or disruption? Stakeholder Hoechle, D., Schmid, M., Walter, I., Yermack, D., 2012. How much of the
relationships and the effect of takeover threat. Management Science diversification can be explained by poor corporate governance. Journal
62, 2820–2841. of Financial Economics 103, 41–60.
Chan, L.K.C., Lakonishok, J., Sougiannis, T., 2001. The stock market valua- Jarrell, G.E., Poulsen, A.B., 1987. Shark repellents and stock prices: the
tion of research and development expenditures. Journal of Finance 56 effects of antitakeover amendments since 1980. Journal of Financial
(6), 2431–2456. Economics 19, 127–168.
Coates, J.C., 20 0 0. Takeover defenses in the shadow of the pill: a critique Jensen, M.C., 1988. Takeovers: their causes and consequences. Journal of
of the scientific evidence. Texas Law Review 79 (2), 271–382. Economic Perspectives 2, 21–48.
Coates, J.C., 2001. Explaining variation in takeover defenses: blame the Jensen, M.C., 1993. The modern industrial revolution, exit, and the failure
lawyers. California Law Review 89, 1301–1421. of internal control systems. Journal of Finance 48, 831–880.
Cohen, A., Wang, C., 2013. How do staggered boards affect shareholder Jenter, D., Kanaan, F., 2015. CEO turnover and relative performance evalu-
value? Evidence from a natural experiment. Journal of Financial Eco- ation. Journal of Finance 70 (5), 2155–2184.
nomics 110, 627–641. Johnson, W., Karpoff, J., Yi, S., 2015. The bonding hypothesis of takeover
Cohen, A., Wang, C., 2017. Staggered boards and shareholder value: a reply defenses: evidence from IPO firms. Journal of Financial Economics 117,
to Amihud and Stoyanov. Journal of Financial Economics (in press). 307–332.
Core, J.E., Holthausen, R.W., Larcker, D.F., 1999. Corporate governance, Johnson, W., Karpoff, J., Yi, S., 2016. The lifecycle effects of firm takeover
chief executive officer compensation and firm performance. Journal of defenses, Working paper.
Financial Economics 51, 371–406. Knoeber, C.R., 1986. Golden parachutes, shark repellents, and hostile ten-
Cremers, K.J.M., Ferrell, A., 2014. Thirty years of shareholder rights and der offers. American Economic Review 76, 155–167.
firm valuation. Journal of Finance 69 (3), 1167–1196. Knott, A.M., 2008. R&D/returns causality: absorptive capacity or organiza-
Cremers, K.J.M., Nair, V., John, K., 2008. Takeovers and the cross-section of tional IQ. Management Science 54 (12), 2054–2067.
returns. Review of Financial Studies 22, 1409–1445. Laffont, J., Tirole, J., 1988. Repeated auctions of incentive contracts, invest-
Cremers, K.J.M., Nair, V., Peyer, U., 2008. Takeover defenses and compe- ment and bidding parity, with an application to takeovers. RAND Jour-
tition: the role of stakeholders. Journal of Empirical Legal Studies 5, nal of Economics 19, 516–537.
791–818. Lang, L., Stulz, R., 1994. Tobin’s Q, corporate diversification, and firm per-
Cremers, K.J.M., Pareek, A., 2015. Short-term trading and stock return formance. Journal of Political Economy 102, 1248–1280.
anomalies: momentum, reversal and share issuance. Review of Fi- Larcker, D., Ormazabal, G., Taylor, D.J., 2011. The market reaction to cor-
nance 19 (4), 1649–1701. porate governance regulation. Journal of Financial Economics 101,
Cremers, K.J.M., Pareek, A., Sautner, Z., 2016. Short-term investors, 431–448.
long-term investments, and firm value. University of Notre Dame, Rut- Larcker, D., Reiss, P.C., Xiao, Y., 2015. Corporate governance data and mea-
gers University and Frankfurt School of Finance and Management Un- sures revisited. Stanford University Graduate School of Business Un-
published working paper. published working paper.
Cunat, V., Gine, M., Guadalupe, M., 2012. The vote is cast: the effect of Mahoney, J.M., Mahoney, J.T., 1993. An empirical investigation of the effect
corporate governance on shareholder value. Journal of Finance 67, of corporate charter antitakeover amendments on stockholder wealth.
1943–1977. Strategic Management Journal 14, 17–32.
Daines, R.M., 2001. Does Delaware law improve firm value? Journal of Fi- Manne, H., 1965. Mergers and the market for corporate control. Journal of
nancial Economics 62, 525–558. Political Economy 75, 110–118.
Daines, R.M., Li, S.X., Wang, C., 2016. Can staggered boards improve value? Masulis, R., Wang, C., Xie, F., 2007. Corporate governance and acquirer re-
Evidence from the Massachusetts natural experiment. Stanford Law turns. Journal of Finance 62, 1851–1859.
School and Graduate School of Business, University of Southern Cali- Morck, R., Shleifer, A., Vishny, R., 1988. Management ownership and mar-
fornia and Harvard Business School Unpublished working paper. ket valuation: an empirical analysis. Journal of Financial Economics
Dehejia, R., Wahba, S., 2002. Propensity score-matching for non-ex- 20, 293–315.
perimental causal studies. Review of Economics and Statistics 84, Nunn, N., 2007. Relationship-specificity, incomplete contracts, and the
151–161. pattern of trade. Quarterly Journal of Economics 122, 569–600.
Demsetz, H., Lehn, K., 1985. The structure of corporate ownership: causes Petersen, M.A., 2009. Estimating standard errors in finance panel data
and consequences. Journal of Political Economy 93, 1155–1177. sets: comparing approaches. Review of Financial Studies 22, 435–480.
Duru, A., Wang, D., Zhao, Y., 2013. Staggered boards, corporate opacity and Roodman, D., 2009. How to do xtabond2: an introduction to difference
firm value. Journal of Banking and Finance 37, 341–360. and system GMM in Stata. The Stata Journal 9, 86–136.
Eberhart, A.C., Maxwell, W.F., Siddique, A.R., 2004. An examination of Shleifer, A., Summers, L., 1988. Breach of trust in hostile takeovers. In:
long-term abnormal stock returns and operating performance follow- Auerbach, A.J. (Ed.), Corporate Takeovers: Causes and Consequences.
ing R&D increases. Journal of Finance 59, 623–650. University of Chicago Press, Chicago, IL, pp. 33–68.
Edmans, A., Goldstein, I., Jiang, W., 2012. The real effects of financial Song, M., Walkling, R., 20 0 0. Abnormal returns to rivals of acquisition tar-
markets: the impact of prices on takeovers. Journal of Finance 67, gets: a test of the "acquisition probability hypothesis.". Journal of Fi-
933–971. nancial Economics 55, 143–171.
Faleye, O., 2007. Classified boards, firm value, and managerial entrench- Stein, J., 1988. Takeover threats and managerial myopia. Journal of Politi-
ment. Journal of Financial Economics 83, 501–529. cal Economy 96 (1), 61–80.
Fama, E., French, K., 1992. The cross-section of expected stock returns. Stein, J.C., 1989. Efficient capital markets, inefficient firms: a model of
Journal of Finance 47, 427–465. myopic corporate behavior. The Quarterly Journal of Economics 104,
Ge, W., Tanlu, L., Zhang, J.L., 2016. What are the consequences of board 655–669.
destaggering? Review of Accounting Studies 21, 808–858. Subramanian, G., 2002. The influence of antitakeover statutes on incor-
Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity poration choice: evidence of the “race” debate and antitakeover over-
prices. Quarterly Journal of Economics 118, 107–155. reaching. University of Pennsylvania Law Review 150 (6), 1795–1873.
Greene, W., 20 0 0. Econometric Analysis, Fourth Ed. Prentice Hall, Upper Subramanian, G., 2014. Delaware’s choice. Delaware Journal of Corporate
Saddle River, NJ. Law 39, 1–45.
Greene, W., 2004. The behavior of the fixed effects estimator in nonlinear Yermack, D., 1996. Higher market valuation of companies with a small
models. The Econometrics Journal 7, 98–119. board of directors. Journal of Financial Economics 40, 185–211.
Guo, R.J., Kruse, T.A., Nohel, T., 2008. Undoing the powerful anti-takeover Wintoki, M., Linck, J., Netter, J., 2012. Endogeneity and the dynamics of
force of staggered boards. Journal of Corporate Finance 14, 274–288. internal corporate governance. Journal of Financial Economics 105,
Guo, R.J., Kruse, T.A., Nohel, T., 2014. Activism and the shift to annual di- 581–606.
rector elections. Journal of Accounting and Finance 14, 83–103.

Journal of Financial Economics: K.J. Martijn Cremers, Lubomir P. Litov, Simone M. Sepe

Uploaded by

Copyright:

Available Formats

Journal of Financial Economics: K.J. Martijn Cremers, Lubomir P. Litov, Simone M. Sepe

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal of Financial Economics: K.J. Martijn Cremers, Lubomir P. Litov, Simone M. Sepe

Uploaded by

Copyright:

Available Formats

Journal of Financial Economics 126 (2017) 422–444

Contents lists available at ScienceDirect

Journal of Financial Economics

Staggered boards and long-term ﬁrm value, revisitedR,RR

Dependent variables Deﬁnition

CAPX/Assets[t] Capital Expenditure[t] /Total Assets[t] .

Dependent variables: Mean Median St. dev. Min Max Obs.

Q [t] 1.613 1.307 0.889 0.728 4.729 34,476

CAPX/Assets[t] 0.059 0.047 0.046 0 0.198 34,476

Institutional holding duration 6.846 6.818 0.990 2.226 10.445 28,298

Dep. variable: Q[t]

Variables (1) (2) (3) (4) (5) (6)

Period: 1978–2015 1978–1995 1996–2015 1978–2015 1978–1995 1996–2015

Staggered board[t- 1 ] −0.034∗∗ −0.046∗∗ −0.036 0.051∗∗ 0.023 0.098∗∗∗

# of ﬁrms in regression 3,076 1,581 2,415 3,076 1,581 2,415

1978 – 2015 1978 – 1995 1996 – 2015

N 31,049 27,017 24,146 11,435 11,014 10,638 19,614 16,003 13,508

1978 – 2015 1978 – 1995 1996 – 2015

N 31,049 27,017 24,146 11,435 11,014 10,638 19,614 16,003 13,508

Four-factor model Three-factor model Market factor model

Average # ﬁrms 12 19.1 − 12 19.1 − 12 19.1 −

N 355 250 231 355 250 231 355 250 231

Four-factor model Three-factor model Market factor model

Average # ﬁrms 12 18.9 − 12 18.9 − 12 18.9 −

N 353 256 235 353 256 235 353 256 235

Four-factor model Three-factor model Market factor model

Average # ﬁrms 21.5 28.6 – 21.5 28.6 – 21.5 28.6 –

N 438 397 391 438 397 391 438 397 391

Cox models Cox models

1978–2015 1978–1995 1996–2015 1978–2015 1978–1995 1996–2015

Variables (1) (2) (3) (4) (5) (6)

Q[t- 1 ] −0.844∗∗∗ −0.347∗∗∗ −0.196 −0.12 0.246 −0.127

# of ﬁrms in regression 1,683 1,003 1,035 1,513 607 1,476

Dep. variable: Q[t]

Variables (1) (2) (3) (4)

Matching procedure: Propensity score based Propensity score Nearest-neighbor Radius

Staggered board[t- 1 ] 0.067∗∗ 0.066∗∗ 0.045∗∗ 0.079∗∗

N 5,845 3,164 5,923 6,081

Matching method Dep. variable: ࢞ Q[t, t+ 1 ] ࢞ Q[t, t+ 2 ] ࢞ Q[t, t+ 3 ]

(1) (2) (3)

࢞ Staggered board[t- 1 , t] 0.029∗∗ 0.041∗∗ 0.057∗∗

࢞ Staggered board[t- 1 , t] 0.025∗ 0.039∗∗ 0.052∗∗

࢞ Staggered board[t- 1 , t] 0.033∗∗ 0.044∗∗ 0.057∗∗

࢞ Staggered board[t- 1 , t] 0.025∗∗ 0.040∗∗ 0.053∗∗

Dep. variable: Q[t]

Variables (1) (2) (3) (4)

N 25,644 25,644 25,644 25,644

Dep. variable: Q[t]

Basic (unbalanced) matched sample Restricted (balanced) matched sample

N 552 552 555 555 452 452 452 452

Dep. Variable: Q[t]

Variables (1) (2) (3) (4) (5) (6)

Staggered board[t- 1 ] 0.112∗∗ 0.051∗∗ 0.044 0.041 0.045 0.046

N 34,476 34,476 17,778 17,778 14,847 14,847

Dep. variable: Q[t]

Variables (1) (2) (3) (4) (5) (6) (7) (8)

N 34,229 34,229 34,460 34,460 28,298 28,298 29,305 29,305

Dep. variable: Q[t]

Variables (1) (2) (3) (4) (5) (6) (7) (8)