P1254325873SXhGD

1
The Role of Advisory Services in Proxy
Voting By Cindy R.
Alexander Duane J. Seppi
Securities
Exchange Commission Carnegie Mellon
University Mark A.
Chen Chester S. Spatt
Georgia State University
Carnegie Mellon University European Financial
Management Symposium, Corporate Governance and
Control Judge Business School University of
Cambridge April 10, 2009

This research was conducted while the authors
were affiliated with the Office of Economic
Analysis, U.S. Securities and Exchange
Commission. The Securities and Exchange
Commission, as a matter of policy, disclaims
responsibility for any private publication or
statement by any of its employees. The views
expressed herein are those of the authors and do
not necessarily reflect the views of the
Commission, the Commissioners, or members of the
staff.
2
Motivation

• Most public U.S. companies use proxy voting to
  make a broad range of corporate decisions.
• 2005 about 300 billion shares voted on 2,293
  compensation plans, 3,300 auditor ratifications,
  and 35,283 individual director elections
• Agency problems can limit the effectiveness of
  proxy voting as a monitoring/disciplinary/decision
  making device
• Dispersed corporate ownership ? free-rider
  problems
• Delegated voting Institutions own gt 65 of U.S.
  voting securities but may follow the Wall Street
  rule
• Growing importance of third-party proxy
  advisory firms
• New rules on mutual fund proxy disclosure
• Investment advisers fiduciary duty vote in best
  interests of clients
• Market solution to agency problems or new agency
  problems?

3
Contributions of the Study (3)

• Empirically examine vote recommendations of
  Institutional Shareholder Services (ISS), the
  most prominent proxy advisor
• New data are from US proxy contests, 1992-2005
• - Board seats at stake, not merely advisory
• - Dissident distributes its own proxy materials,
  separately from management (as distinct from US
  shareholder proposals)
• Research questions, regarding proxy contests,
• Do proxy recommendations bring information to the
  market?
• If so, what type of information?
• - Prediction ?
• - Certification?
• 3) By what method can we disentangle the
  stock-price impacts of (a) prediction and (b)
  certification?

4
Empirical Findings

• Information content Find stock price impact of
  vote recommendations
• Average abnormal return not different from 0 for
  rec. window
• ISS recommends for dissident ? 3.76 abnormal
  return
• ISS recommends for management ? -0.56 abnormal
  return
• Prediction Vote recommendations are good
  statistical predictors of contest outcomes (who
  wins)
• Multivariate probit analysis allows us to control
  for other predictors, e.g., voting rules, contest
  characteristics, mgmt/dissident holdings
• Certification Vote recommendations convey
  information about the relative value of the
  dissident outcome (valuation if dissident wins)
• A structural model allows us to identify the
  stock-price effect of the recommendation as
  prediction and/or certification

5
Related Literature

• Sparse literature on proxy vote recommendations
• Recent literature on vote recommendations in
  non-contest settings
• Bethel and Gillan (2002) Morgan and Poulsen
  (2001) Maug and Rydqvist (2006) Cai, Garner and
  Walkling (2008) Choi, Fisch and Kahan (2008)
• Economic determinants (or predictors) of proxy
  contest outcomes
• Duvall and Austin (1965) Pound (1988, 1989)
• Proxy contests and shareholder value
• Dodd and Warner (1983), DeAngelo and DeAngelo
  (1989), Ikenberry and Lakonishok (1993), Mulherin
  and Poulsen (1998)

5
6
Background Institutional Shareholder Services
(ISS)

• ISS is the leading proxy advisor n.k.a.
  RiskMetrics
• Founded in 1985 by Robert A.G. Monks
• Goal was to promote responsible proxy voting by
  institutions and pension fund fiduciaries
• Proxy advisory services launched in 1986
• Most business is from sale of subscription
  services research reports, recommendations, and
  other voting services. As of June 2006
• Research coverage on 35,196 companies worldwide,
  incl. Russell 3000
• 1,667 institutional clients with assets totaling
  over 25 trillion clients include Fidelity,
  Janus, T. Rowe Price, etc.
• Other businesses
• Proprietary Corporate Governance Index
• Corporate consulting business about 15 of
  revenue
• ISS is a registered investment advisor ?
  Regulation FD applies

7
Data

• Initial contest sample
• Identify all contested solicitations (DEFC14A)
  from SEC EDGAR, 1992-2005
• Eliminate duplicates, erroneous filings, and
  non-CRSP companies
• ? 342 separate contests for CRSP-listed
  companies from 1992-2005
• 2. Recommendations sample use several sources
  for ISS recs.
• ISS archives research reports
• Investext, LexisNexis additional research
  reports
• News databases (Factiva, LexisNexis)
  stories/releases cover ISS rec.
• 236 recommendations for distinct voting episodes
• 3. Contest characteristics read through news
  articles, SEC filings
• Initial filing date was a PREC14A filed?
• What is dissident group asking for?
• (e.g., board seats, corporate governance reform,
  or sale of co.)
• Ownership and governance, voting rules, dissident
  characteristics

8
Data

• Contest outcomes Factiva/LexisNexis news
  articles
• Eventual outcome of contests
• Earliest news date of announced outcome
• Settlement by vote vs. private settlement
• Final sample Board-related contests (170)
• ? 170 recommendations with available data
• Classification of recommendations and outcomes (
  0 or 1)
• a. ISS Rec. pro-dissident if
• ISS supports at least one director
• b. Outcome dissident win if
• Dissident wins one or more board seats
• Negotiated settlements classified as dissident
  victories.

9
Recommendations, by Dissident Type

By Whether Board Seats are Involved By Whether Board Seats are Involved By Whether Board Seats are Involved By Direction of Recommendation By Direction of Recommendation
Dissident Type Not Board-Related Board-Related Board- Related Recs. for Diss. Recs. for Mgmt.

Investment company 8 70 89.7 35 43
Corporation (not incl. investment co.) 17 42 71.2 35 24
Individual shareholder activist 4 29 87.9 15 18
Labor union 3 2 40.0 4 1
Current officer or director 1 4 80.0 4 1
Former officer or director 1 18 94.7 3 16
Individual activist former officer or director 1 4 80.0 3 2
Other shareholder group 3 29 90.6 11 21
Total 38 198 83.9 110 126
Sources Dow Jones/Factiva, LexisNexis News, SEC
filings
10
Key Contest Dates
Ex Proxy contest involving Brantley Capital Corp
(BBDC)
Annual meeting 9/17/2002
Contest resolution 9/20/2002
Initial filing 7/3/2002
ISS rec
9/10/2002
9/9/2002
Philip Goldstein files preliminary controversy
proxy PREC14A targeting Brantley Capital
Shareholder meeting votes cast
Co. announces preliminary outcome of election
insurgent nominees defeated by vote
First news article reporting ISS recommendation
in favor of mgmt. board nominees
Date of ISS report publication
11
I. Stock Price Reaction Around Key Contest Dates
Market-model cumulative abnormal returns
-25,1 around earliest filing date of dissident proxy -25,1 around earliest filing date of dissident proxy -1,1 around date of public resolution of contest -1,1 around date of public resolution of contest 25 days before earliest filing date through 1 day after contest resolution 25 days before earliest filing date through 1 day after contest resolution
N Mean T-stat Mean T-stat Mean T-stat
All contests 170 10.48 7.29 0.35 0.71 17.26 6.27
By company size
Large (Assets gt median) 87 9.53 5.66 0.66 1.13 16.43 5.00
Small (Assets lt median) 83 11.48 4.86 0.02 0.03 18.12 4.06
By time period
Pre-FD 69 13.88 6.39 0.03 0.04 24.43 5.89
Interim period 52 7.32 2.50 0.28 0.28 14.21 2.55
Post-NPX rule 49 9.05 3.73 0.87 1.08 10.39 2.22
By outcome
Dissident win 78 10.09 4.75 2.02 2.89 14.91 3.50
Management win 92 10.82 5.53 -1.07 -1.56 19.24 5.38
By ISS recommendation
For dissident 75 8.54 4.18 1.43 2.05 17.37 4.50
For management 95 12.01 5.98 -0.50 -0.74 17.16 4.44



12
Contest Sample (1) All Recs. (2) Rec. for Mgmt. (3) Rec. for Dissident P-value of Test for Diff., (2)-(3)

All contests 1.35 (1.60) 170 -0.56 (-0.47) 95 3.76 (3.17) 75 0.009
Tender offer by the dissident -0.68 (-0.36) 31 -2.73 (-1.03) 16 1.50 (0.57) 15 0.081
Management win 1.42 (1.21) 92 0.09 (0.06) 56 3.48 (2.01) 36 0.161
Dissident win 1.26 (1.04) 78 -1.49 (-0.83) 39 4.02 (2.46) 39 0.015
Large firms (assets above sample median) 0.63 (0.63) 87 -0.29 (-0.21) 47 1.72 (1.17) 40 0.327
Small firms (assets below sample median) 2.09 (1.53) 83 -0.82 (-0.43) 48 6.09 (3.18) 35 0.008
Pre-FD (Jan. 1992 to Oct. 2000) 2.64 (2.04) 69 -0.30 (-0.17) 40 6.69 (3.57) 29 0.018
Interim Period (Nov. 2000 to Mar. 2003) 2.76 (1.64) 52 1.50 (0.61) 27 4.12 (1.80) 25 0.427
Post-NPX (Apr. 2003 to Dec. 2005) -1.98 (-1.39) 49 -2.92 (-1.45) 28 -0.72 (-0.37) 21 0.164
Filing date return above sample median 0.78 (0.59) 85 -1.58 (-0.89) 52 4.49 (2.36) 33 0.009
Filing date return below sample median 1.92 (1.83) 85 0.68 (0.47) 43 3.19 (2.12) 42 0.293
13
Summary of Findings Price Reaction to
Recommendations

• Recommendations for management do not seem to
  have significant price impact
• Recommendations that favor dissidents are
  associated with significant, positive stock price
  reactions. True across various subsamples and
  time periods.
• Board contests
• Small firms
• Filing date below sample median
• ? Findings broadly suggest information is being
  revealed about dissident strength and/or quality

14
II. What is the Nature of the Information?

• What accounts for the price impact of vote
  recommendations?
• At generic time t during a proxy contest
• pt markets perceived probability of a
  dissident victory
• Dt and Mt per-share outcome-contingent
  valuations given a dissident or management
  victory
• During the contest, between time tA-1 and time tA
  
• Proxy voting advice, A, becomes publicly known
• Advice either favors the dissident (A AD) or
  favors incumbent management (A AM)

15
A. Prediction Hypothesis
Prediction Hypothesis recommendation is a good
statistical predictor of outcome, by either
passive forecasting or active influence
Rec. Favors Dissident Rec. Favors Incumbent Total
Dissident Wins 49 (55.06) 45 (41.28) 94 (47.47)
Management Wins 40 (44.94) 64 (58.72) 104 (52.53)
Total 89 109 198
Pearson Chi-Squared Test p-value 0.054 p-value 0.054 p-value 0.054 p-value 0.054
16
Prediction Hypothesis Multivariate Probit
Analysis

• Control for other factors known to the market at
  the time of the recommendation
• Voting power
• Dissident ownership of voting shares held by
  the dissident shareholders excludes shares
  obtainable from options and other exercisable
  securities
• Management ownership of voting shares held by
  officers and insiders excludes shares
  obtainable from exercise and shares held by
  dissidents on the board
• Voting provisions (rules of the game)
• Cumulative Voting favors dissident in board
  elections
• Majority vote rule more than 50 needed to elect
  directors
• Form of Contest
• Special meeting vs. annual meeting
• Consent solicitation

17
Other Contest and Firm Characteristics

• Other Sources of Dissident or Management
  Advantage
• Professional Proxy Solicitors (e.g., Georgeson,
  Inc.)
• Number of days between contest initiation and
  resolution (cf. Pound (1988))
• Other Dissident and Contest Characteristics
• Outstanding takeover bid
• Shareholders of record proxy for size of
  uncommitted shareholder base
• Activist dissident (e.g., Carl Icahn, Philip
  Goldstein) may pursue broad agendas
• CEO Chairman, CEO tenure in office
• Misc. controls
• Firm size
• Volatility
• Institutional ownership
• Adjusted stock performance
• Industry, time effects

18
Probit Analysis
Independent Variable (1) (2) (3) (4)

RECD (1 rec. for dissident) 0.141 (1.89) 0.179 (2.17) 0.213 (2.16) 0.295 (2.75)
Ln(assets) 0.005 (0.22) 0.044 (1.74) 0.036 (1.03) -0.010 (-0.22)
Ln(Dissident ownership) 0.144 (3.12) 0.108 (1.86) 0.084 (1.31)
Ln(Management ownership) -0.094 (-2.31) -0.177 (-3.21) -0.250 (-3.83)
Cumulative voting 0.427 (2.74) 0.359 (2.27) 0.427 (2.43)
Majority needed to elect 0.030 (0.28) 0.018 (0.15) 0.065 (0.46)
Special meeting 0.244 (1.52) 0.230 (1.08) 0.255 (1.05)
Written consent solicitation 0.321 (1.90) 0.383 (2.34) 0.436 (2.47)
Ln(Institutional ownership) 0.135 (2.10)
Ln(volatility) 1.503 (0.41)
Ln(contest length) 0.002 (1.24) 0.002 (1.40)
Ln(shareholders) -0.018 (-0.44) -0.026 (-0.60)
(continued)
19
Probit Analysis (continued)

Independent Variable (1) (2) (3) (4)

Dissident hires solicitor 0.248 (1.62) 0.306 (1.87)
Incumbent hires solicitor -0.413 (-2.57) -0.454 (-2.76)
Activist dissident -0.066 (-0.49) -0.100 (-0.64)
Takeover bid by dissident -0.060 (-0.48) -0.103 (-0.74)
CEO chairman -0.244 (-2.37) -0.330 (-2.90)
Ln(CEO tenure) 0.214 (2.99) 0.280 (3.02)
Filing-date abnormal return -0.074 (-0.29)
Adjusted return, prior year 0.198 (1.37)
Number of observations 198 197 168 160
Pseudo-R2 0.081 0.181 0.291 0.370


20
B. Certification Hypothesis

• Key Idea derive a test for certification that
  accounts for structural relationships involving
  stock prices, outcome probabilities, and
  certification effects.
• Stock prices as expectations
• Recall
• Functional forms for certification
• gD(RECD) dD RECD dM (1 RECD)
• gM(RECD) µD RECD µM (1 RECD)
• Goal relate underlying parameters dD, dM, µD, µM
  to observable/
• estimable quantities.

21
Model of Pricing Price-Change at REC Conclusion
?D dD
RECD
D
?M µD
? PC
?PA
?D dM
M
RECM
?M µM
Contest End Date (learn who wins)
Recommendation Date (learn whether ISS favors D
or M)
22
The Certification Hypothesis

• Empirical implementation of regression equation
• where
• ? Use a 2-stage approach
• 1. Estimate dissident-win probabilities
  and via probit models that exclude and
  include, respectively, .
• 2. Estimate cross-sectional OLS regression
  equation using fitted values from Step 1 to
  obtain certification parameters
• µM, (dM µM), (µD µM), and (dD dM µD µM).

22
23
Testing for Certification Initial Results (Table
8)
.
Model 1
Model 2
Variable Model 1 Model 2

Intercept -0.003 (0.010) 0.013 -0.003 (0.010) 0.019
0.040 (0.019) 0.023
0.080 (0.033) 0.040

Observations 172 172
R2 0.027 0.037
Parentheses OLS standard
errors Square brackets GMM standard
errors
24
Further Tests for Certification (Table 9)
Panel A Panel A Panel A Panel A Panel A
Unrestricted Model Restricted Model (µD µM ) Restricted Model (dD dM)

OLS regression
R2 0.054 0.045 0.009
Standard error 0.117 0.118 0.120
of observations 172 172 172

Parameters
dD 0.082 (0.037) 0.056 0.058 (0.033) 0.050 0.007 (0.022) 0.038
dM -0.054 (0.026) 0.041 -0.034 (0.022) 0.035 0.007 (0.022) 0.038
µD -0.011 (0.035) 0.055 0.024 (0.024) 0.041 0.048 (0.041) 0.050
µM 0.051 (0.033) 0.053 0.024 (0.024) 0.041 0.004 (0.024) 0.053
25
Structural Tests for Certification (Table 9)
Panel B (p-values shown) Panel B (p-values shown) Panel B (p-values shown) Panel B (p-values shown) Panel B (p-values shown) Panel B (p-values shown) Panel B (p-values shown) Panel B (p-values shown) Panel B (p-values shown)
Unrestricted Model Unrestricted Model Restricted Model (µD µM ) Restricted Model (µD µM ) Restricted Model (dD dM) Restricted Model (dD dM)
OLS GMM OLS GMM OLS GMM
Hypothesis Tests
Dissident Certification
H0 dD dM H1 dD ? dM 0.003 0.048 0.006 0.024 -- --

Management Certification
H0 µD µM H1 µD ? µM 0.197 0.378 -- -- 0.438 0.485

Differential Certification
H0 dD - µD dM - µM H1 dD - µD ? dM - µM 0.021 0.131 -- -- -- --
26
Summary

• Our study provides empirical evidence on the role
  of 3rd-party advisory services in proxy contests.
• Focus on vote recommendations issued by leading
  proxy advisor Institutional Shareholder Services
  (ISS)
• Corporate proxy contests high-stakes settings in
  which votes may be pivotal and outcome
  uncertainty high.
• Voting advice seems to bring new information to
  market participants
• Significant abnormal returns around pro-dissident
  vote recs.
• Abnormally high stock price volatility
  surrounding news of a rec.
• Evidence suggests proxy advice plays a dual role
• Prediction effect vote recommendations are
  strong statistical predictors of contest outcomes
  even after controlling for other predictors
• Certification effect estimates from a simple
  structural model suggest vote recs. convey
  information about relative quality of dissidents
  vs. incumbents

27
(No Transcript)
28
(No Transcript)
29
The Certification Hypothesis

• By construction, ? has some convenient
  properties
• ? has zero mean
• ? is normally distributed
• ? is uncorrelated with ptA-1, RECD, and
  (ptA-1RECD)
• OLS regression gives unbiased estimates of
• µM, (dM µM), (µD µM), and (dD dM µD µM)

Thus, we have the basic regression equation
29
30
The Certification Hypothesis

• By construction, ? has some convenient
  properties
• ? has zero mean
• ? is normally distributed
• ? is uncorrelated with ptA-1, RECD, and
  (ptA-1RECD)
• OLS regression gives unbiased estimates of
• µM, (dM µM), (µD µM), and (dD dM µD µM)

Thus, we have the basic regression equation
30
31
Independent Variables
Variable Mean Median Std. Dev. Obs.
RECD ( recommendation favors diss.) 0.45 0 0.50 198
Dissident ownership () 8.97 7.89 8.23 197
Management ownership () 8.73 5.30 10.17 198
Cumulative voting 0.08 0 0.27 198
Majority needed to elect 0.22 0 0.41 198
Special meeting 0.06 0 0.24 198
Consent solicitation 0.06 0 0.24 198
Dissident hires proxy solicitor 0.84 1 0.37 196
Incumbent hires proxy solicitor 0.97 1 0.17 196
Contest length (days) 34.78 29 28.51 198
Shareholders of record 3,693.67 1,042 7,472.51 168
Takeover bid by dissident 0.19 0 0.39 198
CEO tenure (yrs.) 6.34 4 7.67 198
CEO is chairman 0.63 1 0.49 198
Dissident is activist 0.15 0 0.35 198
Firm size, assets (M) 1,766.46 240.48 5,771.4 198
Institutional ownership () 34.37 28.72 26.47 197
Adjusted return, prior year () -28.99 -27.80 41.69 190
Volatility, prior year () 53.31 44.01 30.24 190