A321

fii
5

5‘2...
13!?
via.

1.

EEWAWW . , d
. zq.... . ., . ”Nam.

f9.
2:.

 

 

.LIBRARY
Michigan State
University

 

 

This is to certify that the
dissertation entitled

Essays on Institutional Ownership

presented by

William C. Gerken

has been accepted towards fulfillment
of the requirements for the

Ph.D. degree in Finance

 

 

 

21:45- 3

Major Professo Signature
“-/’8 /3'oo?
/ I

Date

MSU is an Afiinmtive Action/Equal Opportunity Employer

 

4..-.-._t_._._.-...._t_._.V_...._.-._._.-.—....-.-.-.—.-._.—.-.-._._.-.-.-.-.—.-.—.-.-.-.-.—.-.-._.--.-r.--.-.-A

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 Kthroj/Aoc8PresIClRC/DateDue.Indd

 

 

 

ESSAYS ON INSTITUTIONAL OWNERSHIP
By
William C. Gerken

A DISSERTATION

Submitted to
Michigan State Ui‘iiversit.__v
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Finance

2009

ABSTRACT
ESSAYS ON INSTITUTIONAL OWNERSHIP
By
William C. Gerken

The first chapter addresses the link l‘)etween the liquidity of a firm’s equity securities
and the ability of large shareholders to influence control of a firm. Using a sample
of US. outside blockholdings from 1994-2005, I examine whether liquidity influences
the creation of block holdings. Using an instrumental variable approach, I find that
liquidity increases the likelihood of block formation. Consistent with prior theory,
blockholders of more liquid securities take smaller stakes that do not precommit them
to monitor. I find evidence that. the threat. of exit from a. block can discipline managers
and that this threat. is more effective when liquidity is higher. While liquidity increases
exit from existing blocks, I find no evidence that. illiquidity forces blockholders to
actively monitor. Blockholders7 returns are consistent with liquidity facilitating costly
monitoring. and blockholder choose forms of monitoring that are more effective when
liquidity is higher.

In the second chapter. we empirically examine empltwee ownership of institutional
investment. management firms. \Ve show that empltwee ownership is common, and the
majority of firms in this industry are employee owned. The distribution of employt-‘e
ownership is consistent. with an optimal contracting equilibrium. It is more prevalent
when it is less costly, more efficient, and when alternative incentives are. less attrac-
tive. The level of employee ownership does not predict. risk—adjusted returns, also
consistent with an optimal contracting equilibrium. Finally. we show that employee
ownership predicts risk taking. Portfolios managed by employee owners have signifi-
cantly higher tracking errors. betas. and standard deviations even after controlling for

firm characteristics.

In the third chapter, we use mandatory disclosures by investment advisers to pre-
dict which firms have future incidences of fraud and other investment-related crime.
We find that internal polices that allow for more potential conflicts of interest are
associated with an increased level of future events. Internal monitoring and incentive
aligning mechanisms lead to lower levels of future events. The presence of sophis-
ticated clients is negatively related to the frequency of future events. Even after
accounting for all the above factors, a history of disciplinary actions against the firm
predicts future events. Overall, the required disclosure is useful for predicting events,
and the probability of events is positively correlated with permissive firm policies and

negatively correlated with internal and external monitoring.

DEDICATION

To my family. without whom this "(could not have been possible.

iv

ACKNOWLEDGMENT

I am deeply indebted to my dissertation committee members: J tin-Koo Kang
(chair), Steve Dimmock, Naveen Khanna, and Zoran Ivkovic their insightful comments,
helpful guidance, and unwavering support thorough the entire process. I would also
like to thank my co—author Jennifer Marietta-VVestberg for her assistance and knowl-
edge about the SEC. My journey has been made more enjoyable because of my fellow
students especially Ranadeb Chaudhuri, Gwinyai Utete, Bill Johnson, Iordanis Kara-
giannidis, and Neslihan Yilmaz. I would like to especially thank Ranadeb not only
putting up with my antics throughout our tenure, but also for being there through the
toughest times. Simiarily, the support I have received from my parents, brothers and
close friends has helped me perserve. Even though they have been physically far away,
I feel that I have grown so much closer during my stay here in Michigan. I would like
to give a special thanks to my wife. Dana, for putting up with all the long nights and

reminding me that there is so much more to life than work.

TABLE OF CONTENTS

List of Tables ............................................................ viii

Blockholder Ownership and Corporate Control: The Role of Liquid-

lty ......................................................................... 1
1.1 Hypothesis Development .......................... 4
1.1.1 Trade—off between liquidity and control .............. 6
1.1.2 Control through threat of exit .................... 9
1.2 Data ..................................... 10
1.2.1 Summary statistics of firm characteristics ............. 18
1.2.2 A measure of liquidity ....................... 21
1.3 Empirical results .............................. 23
1.3.1 Blockholder preferences ....................... 23
1.3.2 Liquidity and Preconnnitment ................... 26
1.3.3 Threat of exit ............................ 28
1.3.4 Loyalty, exit, or voice decision .................. 30
1.3.5 Blockholder Gains from Activism ................. 32
1.3.6 Blockholder choice of action and success ............. 34
1.4 Conclusion .................................. 35
1.5 Appendix .................................. 37
Employee Ownership of US. Institutional Investment Management
Firms ..................................................................... 48
2.1 The Institutional Investment i\lanagt-nnent Industry ........... 54
2.2 Theory and Hypotheses .......................... 54
2.2.1 The Costs of Employee Ownership ................ 55
2.2.2 The Benefits of Employee Ownership ............... 57
2.2.3 Employee Ownership and Performance .............. 59
2.2.4 Employee Ownership and Risk Taking .............. 59
2.3 Data ...................................... 61
2.3.1 Employee Ownership of Institutitmal Investment. .\lanagement
Firms ................................ 61
2.3.2 Institutional Inxr'estment l\lanagement Firms ........... 62
2.3.3 Portfolio Manager Ownership of Institutional Investment Man-
agement Firms ........................... 64
2.3 1 Product Performance ........................ 66
2.3.5 Comprehensiveness and Survival .................. 67

vi

2.4 Determinants of Institutional Investment

Management Firm Ownership ....................... 69
2.5 Portfolio Manager Ownership of Institutional
Investment l\~Ianagement Firms ...................... 71
2.6 Employee Ownership and Alpha ...................... 74
2.6.1 IIM Firm Employee Ownership and Alpha ............ 74
2.6.2 Portfolio Manager Ownership and Alpha ............. 76
2.7 Employee Ownership and Risk Taking .................. 77
2.8 Conclusion .................................. 78
2.9 Appendix .................................. 81
Fraud and Registered Investment Advisers ........................... 96
3.1 Registered investment advisers (RIAs) .................. 98
3.1.1 Conflicts ............................... 100
3.2 Operational risk ............................... 103
3.3 Data ..................................... 106
3.4 Empirical results .............................. 110
3.4.1 Prediction of fraud ......................... 110
3.4.2 Investor Reaction .......................... 115
3.4.3 RIA reaction ............................ 117
3.5 Conclusion .................................. 118
Bibliography .............................................................. 128

vii

1.1

1.2

1.3

1.4

1.6

1.7

1.8

1.9

2.1

2.2

IO
03

2.4

iv
01

2.6

LIST OF TABLES

Summary of block acquisitions ........... - ............
Summary statistics .............................
Principal component analysis .......................
Likelihood of being targeted by a blockholder ..............
Determinants of the size of initial blockholding ..............
Target Announcc‘n‘nent Cumulative Abnormal Returns (CARS) .....
Likelihood of Voice and Exit ........................
Blockholding returns ............................
Choice of blockholder action ........................
Summary Statistics .............................
Summary Statistics by Employee Ownership ...............
Summary Statistics by Portfolio I\Ianager Ownership ..........
Survival ...................................

Determinants of Institutional Investment Management Firm Employee
Ownership .................................

Determinants of Portfolio Managers Ownership of Institutional Invest.—
rnent Management Firms ..........................

viii

39

40

43

44

45

46

47

83

85

89

2.7

2.8

2.9

3.1

3.2

3.3

3.4

3.6

3.7

3.8

3.9

Employee Ownership and Alpha ...................... 90

Portfolio Manager Ownership and Alpha ................. 92
Employee Ownership and Tracking Error, Betas, and Standard Deviations 94

Summary .................................. 119
Consistency of RIA practices ........................ 120
DRP Summary ............................... 121
Probit .................................... 122
Probit by Firm Size ............................. 123
Negative binomial .............................. 124
Owner Probit ................................ 125
Flows .................................... 126
Key Person Turnover ............................ 127

Chapter 1

Blockholder Ownership and
Corporate Control: The Role of

Liquidity

This essay examines a sample of outside blockholdings in US. firms to determine
whether the liquidity of a firm’s equity affects the propensity of block shareholders to
engage in activism. Theoretical work such as Shleifer and Vishny (1986) suggests that
by purchasing a. significant block of shares a blockholder can overcome the free-rider
problem inherent. in widely dispersed shareholdings. The relative 1‘)aucity of inter—
vention by blockholders in the US. when compared to other countries like. Germany
or Japan has led several scholars to cite the higher liquidity of US. securities as an
obstacle to blockholder intervention. Their reasoning follows that higher liquidity low-
ers the cost of exiting the position (Bhide (1993)) or increases the potential benefits
from speculation (Kahn and Winston (1998)). These views neglect to consider why
blockholders Would rationally establish the block in the first place. In a theoretical

work, Maug (1998) counters that more liquid securities will attract more. blockholder

intervention because blocks become. cheaper to acquire and higher liquidity allows the
cost of intervention to be borne across more shareholders. The blocks are cheaper
in more liquid securities not just. because of lower transaction costs, but because the

higher liquidity allows blockholders not to preconnnit to monitoring.

As disagreement. exists among theorists regarding the relation of liquidity and
blockholder intervention. I examine liquidity’s on blockholder intervention empirically
using a newly constructed sample of blockholdings in S&P 1500 firms from 1994-2005.
The comprehensive nature of this sample also contributes to the literature on block-
holders as prior work has focused on either a particular type of blockholder (e. g. 1,902
hedge fund blockholdings in Clifford (2008)) or only activist. events (e.g. 244 activist
blockholdings in Bethel, Liebeskind. and Opler (1998)). The sample of 18,210 block-
holdings includes both active and passive filings from all outside blockholders. The
broad coverage of the sample is important as I find that. characteristics of the blocks
such as size and level of activism vary with the identity of the blockholder and that
certain types of blockholders tend to be more passive or active. Their tendency to
either engage or refrain from activism correlates with regulatory and business con-
straints. By using a well defined set of potential targets. the S&P 1500, I am able
to avoid self—selection issues that other papers that only study the characteristics of
observed blockholdings suffer.

W ith this sample. I investigate whether liquidity increases the likelihood of new
block formation. As liquidity of a firms equity and block stock holdings are. endoge-
nously detern‘iined. I establish causality of the relationship using the change in tick
size on major US. stock exchanges in 1997 and 2001 to help form an instrument. for
liquidity. I find that liquidity increases the probability of block formation in my sam-
ple. This result supports the theoretical claim that more liquid markets encourage

the formation of blocks. As blocks will only form when the benefits of monitoring

are higher than the cost. the result is consistent the conjecture in Maug (1998) that.
higher liquidity leads to a higher socially improving level of monitoring. though such a
conclusion is difficult. to support empirically without. observing the cost of monitoring
and losses to other stakeholders.

I then turn my attention to the set of observed blockholdings and examine the
determinants of the size of the blockholders stake. The model in Maug (1998) implies
that a blockholder will take smaller stakes in more liquid securities all else equal.
The blockholder’s decision to perform costly monitoring is private information, so
the higher liquidity allows the blockholder to engage in more informed trading with
liquidity traders. Therefore, blockholders have a greater potential to gain when they
are less precomnritted to monitoring as they can buy shares for a lower price that does
not fully incorporate, the bt‘rnefits of their monitoring activity. As expected, I find that
blockholders take a smaller initial position in more liquid securities.

Besides encouraging activism by making blockholdings more profitable, I find that
liquidity can enhance governance through the threat. of exit. If managerial compensa—
tion is sensitive to the share price then blockholders can encourage managers to engage
in share price maximizing behavior by threatening to sell their block, an event that.
would punish managers. Illiquid securities reduce the credibility of the threat to exit
since blockholders would receive lower prices for their shares. I show that firm value
is enhanced in situations when the threat to exit is most credible - when managerial
sensitivity to the stock price is high and when shares are liquid. This result contradicts
the suggestion that liquidity lnrrts gm'ernance.

I also look at existing blockholdings to see if illiquidity encourages blockholders
to be more active rrrorritors of nranagcmcnt. Though I find that increasing liquidity
increases the propensity for blockholders to exit. their position, I do not find any sup-

)ort for illl(. uiditv increasing the ')1'() )ensitv to erre‘zwe in activism. Instead of exitin r
l . I") t C") {'3 f3

or fighting management, blockholders of illiquid positions often choose a third option
and do nothing. As building blocks in less liquid firms provides less opportunity to
engage in beneficial monitoring and is more costly, investors will only do so when ex-
pected returns from doing so are higher. I find that block holding return measures are
increasing in illiquidity though blockholders with fewer constraints demand a smaller
illiquidity premium. Finally, I show that. the liquidity affects the choice of nicmitoring
action.

This paper contributes to the literature by conducting an empirical test of the
effect of liquidity on a blockholders decision to intervene. I provide evidence consistent
with theoretical models that predict that improved liquidity will enhance monitoring
by blockholders by permitting profitable action more often. I find no evidence that
illiquidity forces institutions to monitor when exit is costly. Instead, I find that many
blockholders are bound by regulatory restrictions or fiduciary responsibility and choose

not to engage in shareholder activism.

1.1 Hypothesis Development

Difficulties in contracting that arise from the separation of ownership and control as
stated in Jensen and Meckling (1976) provide small atmnistic shareholders with little
incentive to exert. control. They hear the full cost of monitoring to reduce agency
costs and receive only a small portion of the benefits of their actions. The existing
literature suggests that this free—rider problem can be overcome by the presence of a
large outside blocklmlder. For example. Shleifer and Vishny (1986) present a model
in which small minority shareholders in widely-held firms have little incentive to incur
monitoring costs because each would like to free-ride on the monitoring of the others.

but a blockholder can profitably take action if its stake. is large enough. Tln‘ouglmut

this paper, I use this definition of monitoring - an action by an outside sharel‘iolder
which can increase shareholder value relative to the value if the outside shareholder
takes no action. This is similar to the definition adopted in Maug (1998) and makes
no differentiation whether the action increases overall firm value or just expropriates
from other stakeholders in the firm. In practice, these actions can take a variety of
forms: engaging in conversaticms with management. starting proxy fights, “vote no"

campaigns, the threat of selling the block, and even hostile takeover attempts.

Blockholdings in public companies are commonplace around the world and are
found frequently even in the relatively more dispersed US public equity market.1
Using a sample of 1.500 companies,2 Dlugosz. Fahlenbrach, Compers, and Metrick
(2006) observe that the average firm in their sample has one outside shareholder that
controls between 14-18% of the outstanding equity. Despite of this finding regarding
the ubiquitous nature of blockholders, evidence of intervention by these blockl‘iolders
is mixed. Clearly some blockholders, for instance wealthy activist individual investors,
play an important role in gm'ernance.3 In recent decades. other types of instituticmal
investors. such as pension funds and hedge funds. have also attracted media and
academic attention for their activist actions (e.g. Smith (1996) studies a series of
activist. interventicms by CalPERS).

In contrast. Jensen (1989) notes that financial institutions and money management
firms. which control over a. third of all corporate equity in the United States. are. typi-

cally uninvolved in the major decisions and long—term strategies of the firms in which

 

IFaccio aml Lang (2002) finds that in Western countries 92% of firms have at. least one. shareholder
with at least 5% of voting rights.

QThc sample taken from the Investor Responsibility Research Center (IRRC) covers about 90
of the value of the NYSE. AMEX. and NASDAQ markets and covers a set of firms and years similar
to the sample used in this paper.

3For an example of recent. intervention by an individual blockholder. see (‘arl
Ichann's recent involvement. with Yahoo!: http://blogs .wsj .com/deals/2008/O7/21/
what—can-carl-icahn-do-for—yahoo-now/

c/_
//‘f

C}!

they invest. Furthermore. more involved actions such as seeking board representation
and engaging in proxy fights are rarer still. Jensen (1989) attributes this perceived
passivism to a host of populist laws and regulations approved in the wake of the Great
Depression, such as the Glass-Steagall Banking Act of 1933, the Securities Exchange
Act of 1933. the Securities Exchange Act of 1934. and the Investment Company Act
of 1940. Black (1990) suggests this passivity may be justified by the burden of legal
obstacles that hinder rational action in all but extreme cases. Another frequently cited
explanation for this lack of shareholder activism is that institutions would rather take
the “Wall Street walk“4 - a colloquialism that implies selling a poorly run stock is

much easier than dealing with management to try to in‘iprove the firm.

1.1.1 Trade-off between liquidity and control

Given that blockholders may choose to exit from their blockholding when costly
monitoring is needed. highly liquid markets may be a hindrance to effective corporate.
governance by permitting blockholders an easier and cheaper exit. This view fails to
consider that blockholders will rationally consider the liquidity of the security before
choosing to form the block. Recent. theoretical work. such as in the model presented
in Maug (1998). has countered that. more liquid markets may actually lead to more
monitoring by blockholders as blocks become. cheaper to form and liquidity allows the
cost of intervention to be shared with the liquidity traders.

In response to a perceived need for iniprovement in an organize-ition. Hirschman
(1970) suggests three possible outcomes: exit. voice. or loyalty. In the case of block
ownership. the blockholder can sell their shares (exit). engage in activism (voice). or

simply do nothing and maintain their position (loyalty). In this frmnework. a trade-

 

"l'l'he oft—cited ““3,“ street walk" or "Wall street rule" traces its origins to guidelines published
by the American Bankers Association in the Wills.

6

off occurs between exit and voice if the choice to remain loyal is not viable. Holding
all else constant, as the cost of exit is lowered. exit becomes preferable to voice.
Previous finance literature have suggested that. this relation is an important reason
for why the US. market displays so little large shareholder intervention - highly liquid
securities markets enable blockhtflders to cheaply dump under performing firms. Bhide
(1993) argues that a natural trade-off between stock liquidity and active. investing is
inevitable. Active shareholders could reduce agency problems by providing internal
monitoring, but the act of monitoring makes these investors informed and thereby
reduces the stock liquidity of their position owing to information asymmetry problems.
Conversely, stock liquidity discourages internal monitoring by reducing the cost. of exit
of unhappy shareholders. Bhide (1993) concludes that the public policy choices in the
US. that have provided a very liquid stock market may come at. the cost of foregoing

potentially valuable active investing.

The cost of monitoring may also play an important role in which monitoring activ-
ities blockholders choose. though inexpensive forms of monitoring may be ineffective.
As Black (1990) states, some institutions face legal barriers against accumulating the
size of the stake necessary to make value enhancing actions profitable. These legal
rules were often intended to protect to mutual fund investms. Their ultimate effect is
to render these blockholders inactive. Sin‘iilarly. Bolton and von Thadden (1998) argue
that the liquidity of the stock market. will reduce activism as such liquidity encour-
ages them to trade on private inforn‘iation. The incentive to speculate increases with
the blockl'ioltlecs infcn'mational advantage over other investors. which will be higher
in smaller more. opaque firms. These are typically the firms that are traditionally

thought to need monitoring by blockholders the most.

The above analysis does not take into account the blockholders decision to form

the block. Kahn and Winston (1998) and Maug (1998) show that liquid markets can

§

help large blockholders overcome the free-rider problem. In particular, Maug (1998)
presents a model in which the large stakeholder buys an initial position that is too
small in the sense that the capital gain on the initial position does not cover the cost
of monitoring. However. the ability to purchase shares on the open market at a price
that only partially reflects the blockholders monitoring efforts gives the blockholder
incentive to monitor. A larger toehold that would cover the cost of monitoring would
precommit the blockholder to monitor. and thus prices would reflect this precommit—
ment. By making the decision to monitor uncertain, the blockholder creates private
information from which it can engage in informed trading. The blockholder gains most
when other shareholders are most uncertain about whether the blockholder will mon-
itor. The ability to make greater gains allows the blockholder to intervene profitably
in situations with higher monitoring costs. This is the mechanism by which liquidity
can enhance monitoring by blockholders. Higher liquidity may lead to a socially im-
proving higher level of intervention (some stakeholders like managers with excessive

compensation may be worse off).

Liquidity allows the blockholder to share the costs of monitoring with the small
shareholder through informed trading with them overcoming the free-rider problem.
Since the blockholder"s decision to engage is costly and the blockholders initial stake
does not precommit. them to monitoring, the blockholder can make profits by making
the private decision to monitoring and then trading with the knowledge that their
decision to monitor will improve firm value. The price of the shares will partially
reflect the improvement in firm value that monitoring by the blockholder could provide.
Therefore, the blockholder can choose to intervene and then buy shares that only

partially reflect the full value of the blocklmlder‘s monitoring in'iproveinents.

In order for this to occur, the. (.lecision to monitor must be not deterininistic. One

plausible reason that. a blockholder would use a random strategy would be that the

8

improvements of the blockholder’s monitoring are known to all traders, but only the
expected cost of monitoring is known. Once the blockholder takes a toehold, it receives
a realization about the true cost of mcmitoring and then make its decision of whether
to monitor based on this realization that is known only to the blockholder. Therefore,
equity liquidity should enhance blockholders" ability to engage in costly monitoring
as liquidity allows informed trading to spread the cost of monitoring among liquidity
traders. As the cost to the blockl'iolder is lower, blockholdings will emerge in firms
where the cost of performing monitoring would prohibit profitable blockholdings if
the firm‘s equity was less liquid. Liquidity should also affect the choice of monitoring
action. In Maug (1998), the author shows that a blockholder should be more concerned
with effectiveness of the action than the cost in a. more liquid market since the cost
can be shared over the liquidity traders.

This gives me three testable implications. First, increasing liquidity should en-
courage the formation of blockholding ceteris paribus. For a given monitoring cost,
higher liquidity will allow the blockholder to spread more of that. cost to other passive
shareholders as liquidity increases. Second. following the same logic, conditional on a
block being formed, when liquidity is higher the initial stake taken by the blockholder
will be smaller all else equal. Third. blockholders should choose more effective. forms

of monitoring when the liquidity is higher.

1.1.2 Control through threat of exit

Jensen (1989) suggest. that institutional investors are "remarkably 1‘)owerless: they
have few options to express dissatisfaction with management other than to sell their
shares and Vote. with their feet”. As .-"\dmati and Pfleiderer (2098) points out. exit.
through the "V'Vall Street walk" is not. necessarily an alternative to activism. The threat

of exit. may itself be a form of corporate governance. While managers might prefer

9

frequent turnover by institutional investors to large active investors that desire to serve
on the boards to monitor and correct managers’ mistakes. managers would really prefer
locked-in passive investors who do not sell their shares. If the liquidation of large block
holdings has an adverse effect on the stock price, then managers who have much of
their compensation tied to the share price either through stock or option holdings are
credibly threatened by the possibility of exit by these blockholders. While this may
not be monitoring in the conventional sense, the presence of the large blockholders
can significantly improve firm value by encouraging managers to enhance shareholder
value.

This leads to another testable implication. As transaction costs impose a cost to
exit, the effectiveness of a large shareholder’s threat to exit is increasing in market liq-
uidity. In the Adrnati and Pfleiderer (2008) model, the discipline effect of a potential
exit on the managers decision is increasing in the interaction of liquidity of the large
shareholder’s position and the fraction of managerial compensation tied to stock per-
formance. In Edmans (2008), the author presents a model in which privately informed
blockholders remain even when exit is viable as a way to over myopic investment by
management. In either model, the ability to exit enhances the value created by the
blockholder. Therefore. I expect to see firm value enhanced when blockholders buy
stakes in firms they can credibility use this threat against - firms with high levels of

liquidity and also high managerial compensation sensitivity to the share price.

1.2 Data

The initial sample consists of block share acquisitions of Strl’ 1500 firms by outside.
blockholders between 1994 and 2005. Prior work has focused on either on a. particular

type of blockholder or only activist events. Bethel. Liebeskind, and Opler (1998)

1 0

survey activism by all types of blockholders in Fortune 500 companies. Several recent
papers study US. hedge fund activism using Schedule 13D filings. For the period
1998 to 2005, Clifford (2008) studies 1.902 sets of block acquisitions (both active and
passive) by hedge funds, focusing on the stock price reaction and changes in operating
performance. Using a sample of 194 Schedule 13D filings from 2003 to 2005, Khein
and Zur (2009) examine entrepreneurial activists (both hedge funds and non—hedge
funds). but focus on confrontational activism ignoring passive filings. With a sample
of 1,059 Schedule 13D filings from 2001 to 2006. Bray, Jiang. Partnoy, and Thomas
(2008) find that hedge fund activists are typically successful in the majority of their

activist attempts.

I limit my sample to 8851’ 1500 firms for two reasons. The first is a. data constraint.
I need information on managerial stock ownership which I obtain from Standard and
Poor"s Executive Compensation Database (Execucomp) for some of my empirical tests.
The second is that I need a well-defined population, so that. I can also observe which
firms do not. have blockholdings. While the S&P 1500 represents 87 percent. of the total
US. equity market capitalization, the sample selection may limit the applicability of
some of the results to other samples. Using a more extensive sample, Cadman. Klasa,
and l\*Iatsunaga (2007) find Execucon'ip firms rely more heavily on aggregate financial
performance measures, such as earnings and stock returns to determine CEO cash
compensation. As the stock incentive effect is integral in order for the threat of exit
to provide discipline, this threat. may be less credible in iniii-Execuemnp firms.

“'hen a person or group of persons acquires beneficial ownership, that. person must.
file a Schedule 13D with the SEC. Beneficial mvnership is defined by the Securities and
Exchange commission (SEC) as voting power or investment power (direct. or indirect.
power to sell the security) of more than 5% of a. voting class of a company’s equity.

Schedule 13Ds must. be filed with the SEC within 10 days of an entity obtaining 5%

11

or more of any class of a company’s securities. Alternatively, the filer can submit
the short-form, Schedule 13G, which is intended for passive investments. By filing
the Schedule 13G, the filer (i.e. l')lockholder) cedes the right to effect. or influence
the control of the target.5 The penalties engaging in control purposes after filing a
Schedule 13G can include losing the right to vote any stock in excess of 5%, loss of
profits and even criminal sanctions.6 Filers must update Schedule 13D upon changes
in the position, while filers of Schedule 13G must update their holdings only once a
year. I use the required subsequent filings (Schedule 13D\A or 13G\A) to determine
the post-acquisition changes in holdings.

To construct my sample, I obtain 407,809 Schedule 13D and 13C filings and their
amendments which have S8513 1500 firms as targets. These filings are available on
the EDGAR website7 for the years 1994 through 2005. The 407,809 individual filings
correspond to 20.684 target-blockholder pairs and give a time-series evolution of each
blockholding. I define the holding period as the period from initiation of the block
until the blockholder reports a shareholding less than 5% or is no longer required to
report (i.e. when holdings drop below 5%). In cases in which multiple blockholders file
together on the same Schedule 13D, I consider only the lead filer. This choice should

not affect inferences since the group members should share the same incentives.

In my study, I focus on outside block ownership and do not include managerial
and employee stock ownership since managers and employees may have additional
economic interests other than their interest. as shareholders. For example, ownership

by managers mayr have conflicting influences on firm value and agency costs. Man-

 

E’Though passive filers may be eligible to file the Schedule 1358. the Schedule 13D is the default
filing. Since a filer has to petition the SEC to file as a passive investm', filers that do not. choose to
do this will file a Schedule 13D even if they have no intentions of engaging in activist activities.

“For an example of a legal case in which an investor failed to disclose a control purpose as required
see Gulf f3 Western Industries. Inc. 1!. Great Atlantic 63 Pacific Tea Company. Inc.

7http://www.sec.gov/edgar.shtm1

12

agers may value consuming perquisites or keeping their job even when they should
be replaced at the cost of other stakeholders, particularly shareholders. Jensen and
Meckling (1976) propose that ownership by managers can help align incentives and
reduce agency costs. Ernpirically, Morck, Shleifer, and Vishny (1988) find that while
small levels of managerial ownership reduce agency costs, high levels of managerial

ownership can serve to entrench management and reduce firm value.

Similarly for rank and file employees. the relation between ownership and firm
value is not clear. Ownership by rank and file employees could better motivate. and
align interests. Kim and Ouimet (2008) show that small employee share ownership
plans (ESOPs) may increase firm value, while large (i.e. greater than 5%) ESOPS do
not increase firm value. A large ownership stake by employees may allow them to
extract unearned benefits at the expense of other stakeholders. Consistent with this
explanation, Faleye, l\r‘Iehrotra, and Morck (2006) document lowered investment, poor
performance and decreased firm value in firms with large ESOPs. As the interests
of managerial and employee block ownership are ambiguous, I exclude them from my

analysis and focus only on outside block ownership.

The mixed empirical evidence of the effectiveness of outside blockholder activism
is not that surprising considering all blockholders do not face the same set of con-
straints. The ability to take advantage of liquidity may only hold for certain segments
of blockholders. Cronqvist and Fahlenbrach (2008) show that. blockholders are not a
hon'iogcneous group. Some blockholders appear to influence corporate behavior while
others seem to passively seek their preferred behavior. One explanation is that. some
entities, such as hedge funds, have few restricticms and can pursue whatever policy
their Inamigers see. fit. while other entities face binding institutional constraints. Even
for a single entity. the act of acquisition of shares above certain ownership levels may

impose constraints. For example. the Exchange Act Section 10(1)) requires that. block-

13

holders that own more than 10% of a share class report their sales and purcl‘iases
every month and forfeit. profits made from "round trip" transactions. This effectively

reduces the short-term liquidity of the position.

Certain histituticmal investors face a variety of regulatory barriers and potential
conflicts of interest that make active monitoring difficult, if not. impossible in many
cases. Legal or regulatory restraints may prevent some regulated financial firms from
accumulating the necessary size block that. makes monitoring cost effective. For in-
stance, a diversified fund, as defined in the. Investment. Company Act. of 1940, may
hold no more than 5% in any one comI.)any, and not more than 10% of any firm’s
outstanding shares. These constraints are binding for many investors. An investment
by the Fidelity Magellan mutual fund of only 0.05% of its portfolio is sufficient to
buy the maximum 5% ownership stake in the smallest Stk'P 1:300 firm. Biolase Tech.
Inc. (as of August 21, 2008). Likewise, conflicts of interest may exist when mutual
funds consider activism against current or potential clients. Davis and Kim (2006) use
proxy voting to show that mutual fund companies are less likely to vote against those
firms with which they have a business relaticm. Simila‘irly. pension funds are typically
bound by ERISA or “prudent man” regulation. This forces pension funds to only hold
prudent securities limiting their investment opportunity set. Also, “prudent investor”
rules require high levels of diversification. Given the constraints to holdings, these
financial blockholders may find exiting or remaining 1')assive more attractive than try-
ing to acquire. a large enough stake in the firm or forming a coalition of like—minded
sharelmlders to cover the costs of performing monitoring.

Like financial blockholders. non-financial (_)1‘_)erating companies may establish block-
holding in other firms. which I will call corporate blockholdings for the sake of brevity.
A large (and somewhat inconclusive) literature exists on the merits of diversification

strategies by such firms. Corporate bloc‘kholders may also seek other benefits when cs—

14

tablishing a blockholding. In a. sample of over 10,000 customer-supplier relationships,
Fee. Hadlock, and Thomas (2006) studies a firm’s decision to invest in trading part-
ners. They find equity stakes can often help overcome contractual incomple'tem‘iss and
also help provide quasi-inside financing to ease financial constraints of trading part-
ners. The presence of these intense trading relationships between firms may mitigate
the incentive to provide discipline. Kang and Kim (2006) show that the relatedness
of the acquirer and the target is an important determinate of blockholder interven-
tion. They find relatedness negatively impacts action as blockholders do not want. to
damage business ties through heavy—handed governance. Borokhovich, Brunarski, and
Parrino (2006) find that outside blockholders who do not have current or potential
business connections to a firm are perceived to be better monitors of management than
outside blockholders with such connections. Though corporate blockholders face these
conflicts of interest, they are typically free from ownership level restrictions unlike fi-
nancial blockholders. Corporate blockholders can and frequently do exercise control
through complete corporate control. Partial stakes are often a precursor to takeover
attempts. Kyle and Vila (1991) suggest that liquidity enables the formation of a toe-
hold stake necessary for profitable hostile takeovers. Overall, corporations may face
lesser regulatory constraints than financial firms7 but business relationships between

firms may limit aggressive monitoring activity.

Activist. investors such as hedge funds and individuals are typically free from the
regulatory barriers and conflicts of interest. that limit activism by financial firms and
corporations. Recently, hedge fund activism has been a hot topic both in the media
and acaden'iic literature. Unlike mutual funds. hedge funds can take much larger
undiversified positions since they are. not. subject. to the Investment Act of 1040 that
stifles activism by mutual funds. Bray. Jiang, Partnoy, and Thomas (2008) note

that hedge fund managers typically have strong i1‘1ce1‘itives to generate returns and

often require investors to "lock-up" funds for long periods of time allowing greater
flexibility in trading. \Vhile the academic literature tymcally focuses on hedge funds
as a special type of activist, the clizu‘acteristics attributed to them are not unlike
those of wealthy individual investors. Entrepreneurial investors, such as Carl Ichann,
Ronald Pereleman, George Soros. and \Varren Buffet t, can and frequently do acquire

blockholdings and sometimes engage in activism.

I combine both individual and hedge fund entities in the Individual/partnership
category for a variety of reasons. First. there is no generally accepted definition of a
hedge fund.8 Since the main issue of this paper is to examine the effect of liquidity
on governance by blockholders, lumping individuals with hedge funds is natural since
both face a similar lack of constraints on their ability to engage in activism. Among
the distinguishing features of hedge funds mentioned in prior literature are highly
incentivized managers. lack of regulation. ability to take concentrated undiversified
positions, and the use of derivatives and leverage. Clearly, most. wealthy individual
investors have extremely similar features. Khein and Zur (2009) also note that both
hedge funds and activist individuals are both relatively free from regulatory controls
of the Security Act of 1933, the Securities and Exchange Act of 1934, and most

importantly the Investment. Company Act of 1940.

Other categories of blocklmlders such as church plans and endowment are harder to
classify cleanly into any of the aforementioned categories. To cover blockholders that
do not natural fall into the financial. corporate and individual / partnership categories,
I create a category called other. On one hand. these blockholders may be exempt. from
the legal restrictions that apply to financial blockholder and the conflicts of interest

that emerge in corporate blockholdings. However. these entities may have many other

 

880e, http://www.sec.g()v/spotlight/herlgt‘funds/hetlge—vaughn.litm for a variety of opinions and
definitions.

16

self-imposed or social oriented constraints.

I exclude from my block formation sample filings from trusts. estates and foun-
dations that represent the passing of an already established block from one owner to
another.9 Similarly, I exclude filings reporting ownership in a. new company which
was formed from an existing company in which the filer had a blocklmlding (e.g.. a
merger or spin-off).

I use the Compustat Execucomp database to find directors and executives of target
firms. I then compare these with the filing to eliminate insider blockholdings from the
sample. The use of the Execucomp dataset limits my sample to the years 1994—2005
and coverage of firms in the S&P 1500.10 Since blockholdings must be reported
only if the filer owns at least. 5% of the target, this restriction limits the amount of
l)lt’)(‘-l<liolding observed compared to the set. of all pul‘,)licly traded firms since my sample
is biased toward large cap firms. To eliminate ESOPs and other profit sharing plans. I
examine the selflreported filer classification on the Schedule 13D and Schedule 13G. I
eliminate blockholdings in cases when the ESOP invests in the firm which employs the
participants of the ESOP. (I do retain ESOP investments in other “outside" firms).

One important caveat is that I only observe equity ownership. I may miss control
actions when an outside shareholder uses empty voting. Empty voting occurs when
investors borrow shares by short selling for the primary purpose of voting on corpo-
rate, matters. The practice creates a. larger control position relative to the economic
position. Hu and Black (200(5) notes that derivative positions are not fully disclosed

in these filings. To the extent that this 1,)ractice occurs in lieu of using equity. the

 

9 These positions frequently result from deaths and divorces and are almost exclusively for invest.-
ment purpose only.

”)Teehnically, I use the intersection of my blockholt‘lings data. and Execucomp, but. I refer to
the sample as the S&P 1500 for brevity. Execucomp covers the Ski) 1500 (excluding ADRs) plus
companies that were once part of the 1500 plus companies removal from the. index that are still
trading. and some client. requests. All told Execucomp contains over 2500 ('(‘impanies, both active
and inactive.

17

results will under report the amount of activism.

1.2.1 Summary statistics of firm characteristics

On Schedule 13D, Item 14 of the cover page asks the filer to self-classify the “Type
of reporting person.” Item 12 reports a. similar classification on the Schedule 13C.
The SEC provides thirteen possible classifications. To make analysis tractable, I
classify all filers into one of four types based on the similarity of the constraints
they face: financial, corporate, individual/ partnership, and other. I classify any filer
that reports as either a broker—dealer, bank, insurance cmnpany. investment company,
or investment adviser as a financial blockholder. As “type of reporting person" is
self—reported, I add an additional screen. Since one major feature that I am trying
to capture is the presence of regulatory restrictions. I collect all filings required under
the Investment Company Act of 1940 during the sample period from the EDGAR
website and classify any blockholder that files these forms as a financial blockholder.
Using the response to Item 14, I classify any filer that reports as either a corporation
or parent holding company as a corporate blockholder. This group consists of both
private and publicly traded companies. Again using the response to Item 14, I label
any filer that reports as either a individual or partnership as an individual / partnership
blockholder. This group contains venture capital funds, hedge funds, private equity
and wealthy private investors. Since I employed a CIK match to identify firms that
are subject to the Investment Company Act of 1940. the individual / partnership group
correctly contains hedge funds sponsored by large financial institutions that are legally
separate and therefore not subject to the same legal restrictions. In the final group. I
include those entities which are difficult to group with the prior three. The final group

includes filers that. report as employee benefit plan (excluding ESOP that represt—znt

18

employees of the target firm), endowment fund, savings association, church plan, and

other.

In Table 1.1 Panel A. I breakdown the filings by blockholder type. Financial block-
holders are by far the common blockholder type comprising nearly two thirds of the
sample. As the previous literature suggests, financial blockholders are predominantly
passive blockholders — only filing a. Schedule 13D 3.24% of the time. As a group they
have the smallest initial position size, 7.46% and rarely exceed a 20% ownership stake
in the target firm. This is consistent with the legal and structural ctmstraiuts they
face. Corporate blockholders comprise around 10% of the sample. They have a higher
degree of activism, filing a Schedule 13D 28.2% of the time and often take large initial
positions as a third of the initial blockholding exceed 10% ownership of the target. A
close reading of Item 4 “Purpose of Transaction” shows that many of these positions
are taken as a toehold in a. merger agreement. Corporate blockholders are unique in
that they can create synergies through cross—ownership with the target company that
other blockholders may not. be able to realize. As expected, the highest occurrence of
activism is seen among the individual/ partnership blockholders. While much of the
existing hedge fund literature has focused on activist filings with Clifford (2008) being
the notable exception. over half the filings by individual / partnership blockholders are
passive. This suggests that previous studies of activist hedge funds may not capture
the entire role of hedge funds as blockholders. Overall, the ownership patterns of the
other category most closely resembles the financial group, which is not surrn'ising since
the main constrz-tints governing the other group include ERISA and “prudent. investor”
rules. These findings are. consistent with the differences in constraints faced by each

group.

In Table 1.1 Panel B. the difference in blockholdings between active (Schedule

13D) and passive (Schedule 13G) filings becomes evident. The initial stake in an

19

active Schedule 13D filing is nearly twice the size of the stake in a passive Schedule
13G filing. Interestingly, Schedule 13D filings typically have a. longer holding length
(30.75 months vs. 23.12 for passive).11 This finding is interesting since hedge funds
and other activist investors are often accused of short—termism by critics. Despite
the low frequency of activism by financial l)lockholders, they still account for 22.49%
of the active filings due to the sheer number of financial blockholders. I create an
indicator variable that equals one if the blockholder subsequently increases the size of
their holding above their initial filing ownership level and zero if they keep the same
size stake or decrease it from the initial level. The probability of the position being
increased is slightly lower for Schedule 13D filings; however, the increase size of the
ownership position is greater for Schedule 13D filings. This is suggestive of larger
but. less anticipated increases in positions for the active filings consistent with Maug
(1998).12

Table 1.2 presents the summary statistics of the 2,456 unique targets of block
share acquisitions from 1994—2005. To control for other firm characteristics that could
impact the decision to intervene, I use return on assets as a measure of Performance (a
proxy for the need of monitoring as Kang and Shivdasani (1995) argues that activism
is more valuable when the firm is performing poorly), firm size as the logarithm of the
market capitalization (a. proxy for public monitoring as in Merton (1987)), Leverage
(a proxy for monitoring by debt holders which may lower monitoring costs for equity
holders). and G—Index (a. proxy for the level. of shareholders‘ rights defined in Gompers,
Ishii. and Metrick (2003)). Using Compustat, I obtain accounting data for all firms in
my sample years. I calculate RCA and leverage for all Compustat firms and winsorize

at the 1% and 99% levels. I then calculate industry adjusted values for RCA and

 

11In this table, I exclude tlmse Schedule 13D filings that. are not closed out by the end of the sample
period, so this may dcm'nward bias my duration figure.
12Since initial positions below 5"}. cannot be. observed. this result must be interpreted with caution.

20

leverage using the Compustat universe.

As a metric of managerial share price sensitivity, I calculate the Incentive Ratio
as suggested in Bergstresser and Philippon (2006). The metric is defined as the ratio
of (0.01 * Price * (Shares + Options) denoted ON EPCT over (ONEPCT + Salary +
Bonus). The ONEPCT corresponds to the increase in manager’s wealth caused by a
one percent increase in the stock price. The incentive ratio captures the relative value
of short-term compensation.

Several features in Table 1.2 are worth noting. Industry—adjusted ROA is negative
for active block targets and nearly zero for passive block targets suggesting that active
targets have more severe performance issues relative to industry peers. Active target
firms are also smaller and have relatively higher debt levels compared to industry
peers. One must keep in mind that these statistics are conditioned on being targeted

by a blockholder.

1.2.2 A measure of liquidity

Though there are many definitions of liquidity, the term typically refers to an
asset’s ability to be easily converted through an act of buying or selling without
causing a significant movement in the price and with minimum loss of value. The
existing literature conventionally expresses liquidity proxies in terms of trading cost
or price impact, which are decreasing in liquidity.

Using correlations with intra day measures, Goyenko, Holden, and Trzcinka (2009)
find that the Amihud ratio from Amihud (2002) does a good job of capturing the

price in'ipact. The Amihud illicuiidity ratio is the monthly average (using daily data)

 

of 1000 * (fl [hater/21 / (Dollar Trudi n g Volume). Another commonly used measure of
trading cost. is the effective cost. The effective cost of trading is usually estimated

from transaction-level trade and quote data. On the buy side. the effective cost is

21

the execution price less the midpoint of the prevailing bid and ask quotes (and the
opposite for the sell. side). Hasbrouck (2009) proposes a Gibbs estimate of effective
cost that is based on daily closing prices. In a broad sample of US. firms over a
sample period similar to the one I use (1993—2005), an annual Gibbs estimate based
on daily data achieves a correlation of 0.905 with the intra day TAQ value. of effective
cost.13 The effective cost contains both the trade-related temporary and permanent
(price impact) components of the price change. As a third alternative, I use the
proportion of zero return days in a month which I call the Z-Indear. Lesmond, Ogden,
and Trzcinka (1999) provides a theoretical justification for using this as a proxy of
liquidity. Given an arbitrary inforn‘iative signal, 17, a market participant will only trade
if the transactions costs are less than I or otherwise there will be no trade and hence
zero return. As transaction costs increase, for the same set of informative signals,
zero returns will be ol‘)served at a higher frequency. I also use the monthly dollar
volume of trading, which is increasing in liquidity, and the monthly bid—ask spread
as two additional liquidity metrics. In the sample, these measures are highly but
not perfectly correlated suggesting that each measure may capture slightly different

components of liquidity.

Since the measures are not directly comparable, I use a standardized version of
each illiquidity measure in all subsequent tables. In Table 1.3 Panel A, I report the
yearly averages for the five measures of liquidity. For expositional clarity and to
capture the commonality between the various liquidity measures, I use a principal
component analysis to create a single liquidity metric”. I use a methodology similar

to the analysis in Korajczyk and Sadka (2008). The results are in Table 1.3 Panel

 

131 thank Dr. Hasbrouck for providing the Gibbs estimates on his website, http://pages .stern.
nyu.edu/~jhasbrou/Research/GibbsEstimatesQOO6/LiquidityestimatesQOOS.htm

1‘fIn a previous version of this paper, I run all the. tests with the Amihud, Gibbs and Z-Index
measures individually and find similar results.

22

B. I only retain the first principal component as it explains 42% of the variance of
the five measures and the incremental value of the addition components in negligible.
The loadings of the first component incorporate all five measures fairly evenly, but the

signs make the intrepretation of the component as increasing in illiquidity.

1 .3 Empirical results

1.3.1 Blockholder preferences

I predict increasing liquidity should encourage the formation of blockholding ceteris
paribus. The blockholder can recoup the cost of monitoring through capital gains to its
initial blockholdings and also through informed trading with liquidity traders, since
only the blockholder know if it will monitor. If the blockholder did precommit to
monitor. then the other shareholders would only sell at a price that reflects the full
value of monitoring, thereby reducing the blockholders incentives to monitor. Some
talue improving monitoring opportunities will be lost because other shareholders will
free ride. As the blockholder receives more of the benefits of monitoring, blockholdings
will occur more often. all else equal. For fixed monitoring costs, higher liquidity will
allow the blockholder to spread more of that cost to other passive shareholders as
liquidity increases.

Observing such a relationship in the data is problematic. Monitoring cost is unob-
servable, and even if a good proxy could be found. mcmitoring cost may vary depending
on the characteristics of the blockholder. Another major concern is the endogeneity
lwtween block ownership and liquidity. Hefiin and Shaw (2000) attempt to look at the
effect of the presence of blmrkholders on liquidity and find that both inside and exter-
nal blockholders decrease s1.1l.)sequent liquidity. In equilibrium, blockholders will set

their holding levels according to the level of liquidity in the market. Block formation

23

are also self-selected in the sense that they are only observed in the cases in which the

blockholder believes the potential benefits of owning the block exceed the cost.

To get around these issues and in order to estimate how liquidity affects the propen-
sity for block formation, I appeal to an instrumental variables approach using a set
of exogenous shocks to liquidity to identify the effect on block formation. To mea—
sure the liquidity shocks, I use an indicator variable for the sixteenths and decimal

tick size regimes on the US. stock exchanges. I interact these with price

 

since lower
priced stocks should be more. affected by the shift in tick size. On June 2, 1997, the
NASDAQ pricing switched from eighths to sixteenths. The NYSE followed on June
24, 1997. The. NYSE switched stock pricing from eighths to decimals starting with
7 firms in August 2000 and all firms by January 29. 2001. The NASDAQ began to
switch to decimals tick sizes in March of 2001. The SEC ordered all US. stock markets
to convert to decimal pricing by April 9th, 2001. Chakravarty, Wood, and van Hess
(2004) find quoted and effective bid—ask spreads on the NYSE declined significantly
following decimalization. Furfine (2003) documents that though decimalization leads
to smaller spreads it also lead to lower depth therefore causing a. theoretically ambigu-
ous change to market liquidity. From his enmirical work using price impact measures,
he concludes that actively traded stocks (like the ones used in my study) generally

experienced an increase in liquidity following decimalization. Thus, different tick size

1

price

 

regimes interacted with should meet the criteria of a good instrumental variable
as the instrument is significantly correlated with liquidity and there. is not. a plausible

reason to believe that. block formation (or the unol‘iserved cost. to monitor) should be

correlated with the instruments except through liquidity.

To implement the. test. I collect a. sample of 157.703 firm-month observe-itions for
all the S&P 1500 firms between 1994 and 2005 that have liquidity data. available. I

create an indicator variable that is equal to one if a new block filing occurs in the firm-

24

month and zero otherwise, and a set of instruments as described earlier.

 

'e is also
included as an instrument so the interaction terms have the correct interpretation. In
the reduced form equation, I regress the liquidity proxies on the instruments (and the
exogenous control variables in the second specification). In the structural equation, I
use the indicator variable for new block filing as the dependent variable. In order to
control for other potentially influential target characteristics, I include log(market cap)
and leverage as proxies for monitoring by other stakeholders and industry-adjusted
ROA as a proxy for the need for monitoring. I estimate the instrumental variable
probit model using maximum likelihood estimation. Standard errors are clustered by

firm.

Table, 1.4 reports the results of the estimation. In the first column of each speci-
fication, Table 1.4 shows the reduced form model estimates for liquidity. In all three
specifications. the illiquidity measure has a negative and statistically significant coef-
ficient for the decimalization interaction variable. The. eighths regime interaction is

negative for all three Spt‘.("lfi(3all()l‘ls and statistically significant. once the controls are

1

price

 

included. As exj')e(;-ted, is positive and statistically significant in all specifica-
tions. The R2 for the reduced form model are relatively high and increase from 24%
for the liqui<lity-only model to 60% for the full model. This further supports that
interactions are a reasonable choice for an instrumental rariable. The estimates of
the structural model are reported in the right. column of each specification in Table
1.4. In the first sptx-ification, I only include the liquidity variables. The coefficient
on Illiquidity in the structure model is negative. but economically and statistically
significant. However. once I include the control variables in the second specification,
the llliquidity variable become statistically significant. The economic significance. is
large as well. A one standard deviation in Illiquidity more than doubles the rate of

block formation. These levels of economic and statistical significance remind the third

[\3
Ci]

specification where IU add a control for managerial ownership. For all three speci-
fications, I reject the hypothesis that the error terms in the structural equation and
the. reduced-form equation for the liquidity are uncorrelated, suggesting endogeneity
was a legitimate concern. The information asymmetry may arise when a blockholder
acquires a position as the blockholder becomes more informed about the potential cost
to monitor the firm. This information asymmetry could manifest in reduced liquidity.

I also find that. the coefficients of the control variables are consistent with the ex-
ante predictions. The negative coefficient in the second and third stage on Performance
implies that block formation is more likely when the firm is perforn'iing poorly. Poor
firms may have. more opportunity for improvement, and the gains to monitoring may
be greater when firms are under performing their peers. The coefficient on log(market
cap) is negative and strongly statistically significant across all three specifications.
This result is likely due to how the sample was constructed. Since I only observe
blockholding when the holding exceeds 5% of the firm, the same dollar value ownership
position in a small firm will not be ol_)sei'val:)le in a larger firm. I also observe that
block formation is more likely when Leverage is higher. This is consistent with the
interpretation that debt holders may help encourage monitoring by sharing monitoring
costs. There is no evidence that the G—Index is related to block formation.

In summary, the analysis of block formation supports the hypothesis that blocks
are more likely to form when li(.1ui(_lity is higher. The results also suggest that the
instrumental variable methodology was valid and necessary as reduced tick sizes im-

pacted liquidity and block formation and liquidity are endogenously determined.

1.3.2 Liquidity and Precommitment

According to the model in Maug (19.98) in an equililn'ium market. large sharehold—

ers will buy an initial stake that will not precommit. them to monitoring. That is,

26

the capital gain on their initial position due to monitoring is insufficient to cover the
cost of monitoring. However. the blockholder can gain two ways: capital gains to its
blockholdings and informed trading with liquidity traders. VV hen liquidity is higher,
the blockholder prefers to buy a smaller stake because it can do so at a lower price
because shares do not fully reflect the increase in value of the blockholder’s monitor-
ing. If the blockholder did precommit to monitor, then the other shareholders would
only sell at a price that reflects the full value of monitoring, thereby reducing the
blockholders incentives to monitor. In more liquid markets, the blockholder can profit.
from additional shares purchased from the other shareholders. Conditional on a block
being acquired, as liquidity decreases a larger initial stake is needed to be purchased

so that the blockholder can capture capital gains on the initial stake.

To test this empirically, I create a sample of 18,210 blockholder acquisitions in
which the blockholder obtains a 5% or greater stake in any S&P 1500 firm between
1994 and 2005. I expect that all else equal, a blockholder will take a smaller initial
stake in a. more liquid firm. I run a regression to predict the size of the blockholding
conditional on liquidity and control variables. The dependent variable is the ownership
percentage as recorded in the initial l:)lock.holding filing (i.e. Schedule 13D or Schedule
13G). As I did before, to control for other potentially influential target characteristics,
I include ROA, log(market cap), leverage, and the G—Index. To avoid issues with
simultaneity, I use the values of the illiquidity measures and control variables one year
before block acquisition. I also include an indicator variable for an active (Schedule
13D) filing. Since active filings will be more costly for blockholders, the size of the

position should be larger, ceteris 1‘)aribus.

I estimate the model using an instrumental variable Tobit model censored at 5%
and 100% and report the results in Table 1.5. As before, the rethiced form model is

consistent with the reduction of tick sizes increasing liquidity. In each of the specifi-

27

cations, I find a positive and significant coefficient on Illiquidity indicating that the
blockholder buys a sn‘ialler stake in a more liquid market. To understand the economic
significance, I find that a. one standard deviation decrease in Illiquidity corresponds
to a 1.03 to 1.37% reduction in initial ownership stake. These results support the hy-
pothesis that the size of the blockholding will decrease with liquidity as blockholders
rely more on capital gains on their initial position in illiquid markets and on informed

trading in liquid markets.

1.3.3 Threat of exit

In the. model presented in Admati and Pfieiderer (2008). a large shareholder can
reduce agency costs by discipling managers through the threat of exit. This implies
that the stock price should react favorably to a partially unanticipated acquisition by
a large block holder, though this prediction is not unique to Admati and Pfleiderer
(2008). Admati and Pfleiderer (2008) show that the effectiveness of the threat to exit
as a discipline action increases when the large shareholder can more credibility exit
its position. If the security is highly illiquid. then the threat to exit is not credible
as the blockholder does not want to suffer a large loss in value to exit its position.
In those situations where the threat is credible. the blockholder can influence man-
agerial behavior to be more aligned with shareholders by threatening the managers.
The amount of influence should be proportion to the size and credibility of the threat.
Therefore. the (‘lisciplining effect of a potential. exit on the managers decision is ill-
creasing in the interaction of the positions liquidity and the importance of i‘nanagerial
compensation tied to stock performance. The blockholder threat is credible even in a
very liquid market because the market will react to the information conveyed by the

blockholder leaving the position not because of the price impact of the exit trade.

28

In equilibrium, the current price reflects the expected firm value based on existing
public information. A positive announcement return indicates that the market believes
that the new blockholder will make the firm more valuable. If the threat of exit
can encourage managers to increase shareholder value, then abnormal returns should
increase when this threat is more credible, that is, when market liquidity is high and
managers are highly sensitive to share prices. To test this empirically, I compute the
abnormal return using the standard event—study methodology. I obtain my estimates
of the market model by using 200 trading days of return data beginning 220 days
before and ending 21 days before the announcement of the block share purchase. I
use the CRSP value weighted return as the market return. I sum daily abnormal
returns to get the cumulative abnormal return CAR(t1,t2) from day t1 before the
announcement date of the block share purchase to day t2 after the announcement

date.

I used the file date listed on Schedule 13D or Schedule 13G. Mikkelson and Ruback
(1985) note that the Schedule 13D only needs to filed within 10 days of crossing the
5% threshold, and this causes difficulty in determining the correct event date. The
problem is worse for Schedule 13G which only needs to be filed within 45 days of
crossing the 5% threshold. The acquiring blockholder also may spread its purchases
over several weeks triggering a filing only when it. crosses the 5% threshold. This will
reduce the power of my tests as I cannot percisely identify when knowledge of this
new blockholding becomes public. To the extent that this information falls outside
the window, the tests will be less likely to find a significant result.

I use a sample of 12868 block formations that. also have Execucomp data since I
need this data to construct the threat of exit proxy. I regress the abnormal return
experienced when the blockholder acquires the block on the Illiquidity measure, the

II‘le—‘llthC Ratio and the interaction of the two. I include the Illiquidity measure and

29

Incentive Ratio as separate terms to control for other explanations that do not involve
the threat of exit such as price pressure.

Table 1.6 presents the announcement returns estimates for four event windows: —1
to 1. -5 to 5, —10 to 10 and -20 to 20 days around the file date. Since I am using
an Illiquidity measure. I interpret a negative coefficient on the interaction term as
evidence of the threat of exit increases firm value. In Table 1.6 columns (1) to (4), I
observe that the results are generally consistent with the implications of the Admati
and Pflciderer (2008) model. I find that the coefficient on the interaction of Incentive
Ratio and Illiquidity is negative suggesting that in cases where managers have highly
aligned incentives illiquidity reduces the validity of the threat to exit. The effect is
modest for all cases ranging from llbp for the shortest. window to 38bp for the longest
window. In columns (5) - (8), I repeat the earlier test adding a dummy variable for
Active blockholdings. The magnitude of this variable is between 1 and 4%. These
results suggest that liquid markets allow blockholders to influence managerial behavior
to be more aligned with shareholders by threatening to exit blockholdings, but the

effect is much smaller than armouncing a specific activist agenda.

1.3.4 Loyalty, exit, or voice decision

The prior empirical tests only speak to the creation of new blocks. When reevalu-
ating their existing positions. blockholders may respond differently to liquidity. Using
the Hirschman (1970) nomenclature reaction to a declining firm. a blockholder can
remain loyal. exit. or exercise voice. Once a. block is formed. liquidity could hinder
activism (voice) as liquidity allows exit from the block to be a cheaper solution than
actively monitoring. To test whether increased liquidity encourages existing block—
holders to take the. exit via “W'all Street. walk" rather than exercise voice and take

an active stand against management. I create a sample of passive blockholdings (i.e.

30

blockholdings where the initial filing was a Schedule 13G). On Schedule 13D. the
blockholder is required to report if the blockholder has previously filed a statement
on Schedule 13G to report the acquisition which is the subject of this Schedule 13D.
I use this entry to create an indicator for a switch to activism. So for each month,
every existing passive blockholding has three possible outcomes. The position can
remain passive. the "loyal” case. Alternatively, the blockholder can file a Schedule
13D declaring its activist intentions thereby exercising “voice”. The other possible
outcome is the blockholder can simply “exit” and sell the blockholding.

Since I have three possible outcomes, I employ a multivariate logit regression to
analyze the determinants of these three outcomes. I make the loyalty or do nothing
case the omitted or base outcome. If liquid markets hinder activism, then I should see
a positive relation between illiquidity and activism. That is, if illiquidity encourages
blockholders to be more active in monitoring because they are stuck in their position,
I should observe a positive coefficient of illiquidity in the “voice” outcome. If liquidity
lowers the cost of exit, then I should see a negative coefficient on the illiquidity mea-
sures for the “exit” outcome. I also consider other factors that may affect. the choice of
exit, voice or loyalty. Demsetz and Lelm (1985a) argue that blockholders as long-term
investors have strong incentives to monitor management. So I include the length of the
time in months that the block has been held as an ex[.)lanatory variable.15 Therefore,
I include the blockholdings cumulative stock market performance to date.

Table 1.7 presents the results of the multivariate logit estimation. Consistent with
the theory of the trade-off between liquidity and control, higher liquidity (or lower
Illiquidity) increases the probability of exit. However. the evidence of a relationship

between “voice" and liquidity is mixed. In the first specification without controls. the

 

1" I do this instead of the total holdinor )e.riod of the block. since the total holdino‘ )eriod is not

b 5
known prior to termination of the blockholding. This avoids any spurious inferences that could be
caused by including information that is not in the information set at each block—month observation.

31

coefficient is positive and statistically significant, which is consistent with the argu-
ment. in Bhide (1993). However, the sign of the coefficient. flips and loses statistical
significance in the second s1_)ecification when controls are added. This suggests that
even if blockholders are locked into a blockholding position they will choose to remain
passive rather than engage in monitoring. The implication is that there is little ev-
idence of a trade-off between liquidity and control. Even if liquidity leads to more
exit, a blockholder who is unable to exit due to low liquidity is unlikely to engage in
monitoring. One reasonable explanation follows from Table 1.1. Passive blockholders
tend to be the more constrained financial blockholders who have. regulatory restric-
tions that inhibit. their ability to be effective active monitors. Following the results of
the previous section, financial blockholders may be more inclined to exert. discipline
by exit when faced with restrictitms on engaging on more traditional forms of mon-
itoring. I find this to be the case as corporate and especially individual/partnership
blockholders are more likely toflshift to activism. Consistent with presence of wealth
constraints on building a significant position in the largest of firms and presence of
other public forms of monitoring for well known firms, blockholdings in larger targets
are more likely to stay remain passive as the coefficients of log(market. cap) is negative

for both “voice” and “exit".

1.3.5 Blockholder Gains from Activism

The free—rider problem mentioned in Shleifer and Vishny (1986) suggests that a
blockholder will only intervene when the amount of benefits from their stake exceeds
the cost of monitoring. Unfortunately, the cost of monitoring is difficult to observe.
I make the natural assumption that in equilibrium, blockholders will only act.i\-'ely
monitor when the expected con1pensati(.)n they receive from the action exceeds the

expected cost. of taking the action. I use the. rerdized values to proxy for the expected

32

compensation for the blockholder. This allows me to compare the returns to active

and passive blockholding as well as among types of blockholders.

I calculate the holding period returns for each of the blocks using the closing CRSP
prices on the file date of the block announcement and the file date when the block
drops below 5% ownership. As holding lengths vary considerably. I also find market-
adjusted returns by netting the returns of the position against an equal size investment
in the market during the holding period. One important caveat is that these returns
do not. necessarily reflect the profits the bl(.)ckholders actually make. I do not. have .
actual prices for their trades and these figures do not take into account the possibility
that the blockholder held a sub 5% stake before the filing date (or retained a sub 5%
ownership stake post-blockholding). In the most illiquid securities, 3. blockholding will
only be observed if the high cost of monitoring can be compensated with equally high
returns. As liquidity increases, blockholders can monitor in more marginally beneficial
situations, so therefore I expect a positive relation between illiquidity and blockholder
returns. A more constrained blockholder will be limited in the size of its position
and possible actions it can take and will demand a higher premium. Conversely,
unconstrained blockholders, such as corporations and individual / partnerships will find

ormortunities profitable even will a relatively lower liquidity premium.

Columns (1) — (3) in Table 1.8 contain the holding period return, while columns
(4) and ( 5) contain the market adjusted returns. Despite the mentioned caveats, these
results are consistent with the pt‘)sitive relation of illiquidity and returns. In column
(1), I find that Illiquidity is 1.)ositivcly related to the blockholders holding period
return. Though the relation of illiquidity and returns is not an original finding. the
size of the effect on returns due to illiquidity appears to be too high to be attributed
only to trading costs. In column (2). I find that individual/partnerships have the

highest. returns. The results imply that i11dividual/partnerships can engage in more

33

costly monitoring and also the greater ability of these blockholders to take risk in the
absence of constraints that bind the other types of blockholders. In column (3), I
interact Illiquidity with a dummy for low constraints (i.e. non-financial blockholders).
I find that Illiquidity is still positive, but the interaction term is negative suggesting
that blockholders with fewer constraints can accept a smaller illiquidity premium.
Columns (4) and (5) repeat the earlier specifications with the market-adjusted returns.
All implications remain the same. Given that Kang and Shivdasani (1995) suggests
that poorly performing firms have the most to gain from monitoring, I partition the
sample on industry—adjusted ROA and again test if the relationship hold for below-

median performing firms. I find this is the case.

1.3.6 Blockholder choice of action and success

Blockholders often make very general or boilerplate demands in the “Item 4. Fur-
pose of Transaction” section of the Schedule 13D filing. \Nhile this is done to keep
open a wide host of potential remedies for poor firm performance, this also makes
empirical measurement of blockholder success of achieving their stated purpose diffi-
cult. I thereby focus on two specific types of purposes that have very clearly defined
objectives: mergers and CEO turnover. An example text is listed in the Appendix.
First, I identify all blockholdings in which the blockholder states that the objective of
the blockholding is to support or oppose a merger of the target with another firm. I
then use Factiva and the CRSP delisting data to determine whether the blockholders
objective was accomplished. Similarly, I examine the staten‘ient of purpose for calls
for resignation of the CEO and use Factiva and the Compustat Execucomp database
to determine whether the CEO was removed during the blockholding period. Overall,
I identify 1,297 cases and blockholders are successful in gaining their stated objective

in over half the cases.

34

Maug (1998) argues that blockholders should favor the more effective means of
i‘nonitoring in firms with highly liquid securities since cost is less of a factor since it can
be spread across the. plentiful liquidity traders. Conversely, when the firm's security is
less liquid, cost becomes a relatively more important factor than effectiveness. To test
this prediction, I examine the outcomes of the 1,297 cases in which I can determine
success. I consider three possible forms of activism: mergers, CEO turnover, and other
which contains all other classifial’fle purposes. In Table 1.9, I employ a mnltinomial
logit regression to determine whether liquidity plays a role in the choice of activism.
I find that Illiquidity is negatiVely related to mergers in both the. liquidity-only and
full specification. If mergers are the more effective means of controlling a firm (a
plausible assumption). then this is evidence in support of the prediction. I find that.
in merger the blockholder is slightly less likely to be successful (that. is the outcome
is the same as what the blockholder supported) than in CEO turnover. However,
this comparison suffers from sample selection is not necessarily a true comparison of
effectiveness. Without more detail on costs, I cannot definitively say that mergers are
more effective and more costly than other forms of activism. but the ccmjecture seems
reasmiable. If one accepts that conjecture, then the findings in Table 1.9 support. the

assertion in Maug (1998).

1.4 Conclusion

In this paper, I empirically examine the role of the liquidity of a. firm‘s equity
in allowing blockholders influence control over targets. By using the reduction in
tick sizes of the stock exchanges as an ii‘istrmmmt to break the endogeneity between
block formation and liquidity. I find liquidity 1_)ositively influences the creation of

blockholdings. Also consistent with theory. l.)l(._)ckholders take smaller stakes that do

35

not precommit them to monitor firms with more liquid securities. I also find that
blockholders can use liquidity to influence management. I find evidence that the
threat of exit. from a block can improve firm value and that this threat is more effective
when liquidity is higher. I fail to find evidence of a trade-off between liquidity and
control. While liquidity increases exit from existing blocks. I find no evidence that
illiquidity forces blockholders to monitor. Blockholders’ returns are consistent with
liquidity facilitating costly monitoring. I also the choice of blockholder activism and
examine success rates. I find evidence consistent with the Maug (1998) prediction that
blockholders will prefer more. effective methods of control like takeovers over lower cost
methods. Overall. increased liquidity, such as was seen after the reduction of the tick .

sizes improves blockholders ability to monitor firms.

3f}

1.5 Appendix

Instructions for filing Schedule 13D:

Item 4. Purpose of Transaction

State the purpose or purposes of the acquisition of securities of the issuer.
Describe any plans or proposals which the reporting persons may have
which relate to or would result in:

a. The acquisition by any person of additional securities of the issuer, or
the disposition of securities of the issuer;

b. An extraordinary corporate transaction, such as a merger, reorganiza-
tion or liquidation, involving the issuer or any of its subsidiaries;

c. A sale or transfer of a material amount of assets of the issuer or any of
its subsidiaries;

d. Any change in the present board of directors or management of the
issuer, including any plans or proposals to change the number or term of
directors or to fill any existing vacancies on the board;

e. Any material change in the present capitalization or dividend policy of
the issuer;

f. Any other material change in the issuer’s business or corporate struc-
ture, including but not limited to, if the issuer is a registered closed-end
investment company, any plans or proposals to make any changes in its in-
vestment policy for which a vote is required by Section 13 of the Investment
Company Act of 1940;

g. Changes in the issuer’s charter. bylaws or instruments corresponding
thereto or other actions which may impede the acquisition of control of
the issuer by any person;

h. Causing a. class of securities of the issuer to be delisted from a. national
securities exchange or to cease to be authorized to be quoted in an inter—
dealer quotatirm system of a registered national securities association;

i. A class of equity securities of the issuer becoming eligible for termination
of registration pursuant to Section 12(g)(-z’l) of the Act; or

j. Any action similar to any of those enumerated above.

Below is an example from a Schedule 1.3D filed on September 4th. 1996 by Clover
Capital i\lanageme1it for its investment in California Microwave, Inc.

37

Item 4. Purpose of Transaction. The subject securities were acquired and
continue to be held by the Reporting Persons for investment purposes.
Each reserves the right to acquire or dispose of the subject securities. On
August 28, 1996 representatives of Clover met with the Chief Executive
Officer and Chief Financial Officer of the Company and recommended that
the Chief Executive Officer be replaced because in Clover’s opinion (1) the
Company 8 credibility in the investment community has suffered because
of earnings forecasts made by the Chief Executive Officer that have failed
to materialize, (2) the Company 8 performance has been and continues
to be inferior to the Company 8 competitors, (3) certain of the Company
s fundamental business strategies are flawed and (4) the Company has
not provided a credible strategy as to how it will maximize shareholder
value. In the future, Clover may present. specific business strategies to the
Board of Directors or otherwise provide advice regarding the business of
the Company and possible ways to maximize shareholder value. Except
as set forth in this Item 4, the Reporting Persons have no present plan or
proposals that relate to or that would result in any of the actions specified
in clauses (a) through (j) of Item 4 of Schedule 13D of the Securities
Exchange Act of 1934.

38

 

 

 

 

 

 

 

 

www.mm camow $5.3 om.w H\cnmd 9x53: SEES 33.42 133%
mhdm $34.2 $256 mmqia §w5wm fimmfim Kawfim at: a>£o<
Ndmm 3&3de vabfi EN snowman gamed 959:» manna: airman
2:55; 033m A 953 A mm? :osmxca 9532552
.9: 95:22 ESE; 3:5 BEE \Escmficfi 3m.:i.50 Erozcih om? 035.2%

309:5 >5 >..§::::m waxy—o: xuoflm “m Ezcm
ox.m..m :5ch 15:}: omnw $3.5 $3.2 1:5.
coda “Kafm Kkndm worm 1.3de cmmd .239 .mt—E—rfictzo 539 £923.; .rztc
No.3. 9.536 fivwfim wnd 055.3... wmfié mazzéicaza $152315
Qamm REV: Kmmhm 8.: cavmwm mum; 3545.50
swam. $09.0 $8.2 owns oxsvmh 3%.: :Einm
$555: as?
flags sea A $2 A songs 9:
mass—OE 15:: 1.5:: 73:; eczemaom LET. iazcm .8~€-:E.EE me 23%

 

m9? $335333 .3 552255 9:33; #335 ”4. 7.5.an

 

326$sz mo mmoaasm - a. :8: mfifioaeu 5m mzoflosbmfi Unmm Sm 52:22:; 2: .53 .3253: 5.58::
0:76: a 3 sawiafim mmeimza mammmzommfi 59a $23 £533.. 23. .Q2 Bfioozom 2: :o $3.893 mm Humbug i: we .E..:::.....:::E 2:
RE :33: mczmba i fix: megs? 529283 93 $23 2:55 2:on A3258 5mg: 0: :omfimom 2: .25 @5553 925 1.995.; a
8:2: as: moot ho exaum can £5 a 339.: 3293003 2: 3:: 30:32: cam mfi 25888 EEOJVEOE BE .22: 855 $5 vi mix:
UmEEHBmv mm S .Ec: mm x003 a 3.39: we $255: of mm :3ch 9583i #3565303 mi 13 «53.5%: mewzfisosma 9:23:33
2: com: @953 mm was. :ofiamoa N835 “NEH. ADM; @Etmfiom :0 NH Eon: HOV Q2 Ezthom do 3 :5: 8 95:5sz $2334.53 2:
:0 women E 3?...3Easkbesfiaiesfl was .853ch .32: mo uo< EEQEOD “:25ng 8 goofinzfiv $323.5.“ SE mazoiixscf
.3 :ossofimmflu 2a. .5322253 wcgsvoe 22 \3 Q2 33?: 2: :o ~5me 5:13 3.52% 035 mEH .m<0Qm .3 cigars; .n:
mfiSEEEEs :2: EB mEE Um; can QM: mzéezom OWm Eoc ESEwEGom 4.53 .5 225% 2: :2:an _ .nocm «:5 2:: 2.5.5.5
:2: 33...; Loaf. k2,: E .231. $533.3.“ .5 New. c. 7.5.5:: #2: .CE 2: 4.575 E mzitiiooc .rzzEC—EE away: .t. 33:31. tats: 2:.

mnomfimmzvom x903 mo enemas—Sm
HA 2an

39

 

 

 

 

m; 2:. a: 22 3:: mm; 82 :2 22 22. 32...: 2:: 93.1%.;
:3 8.0 a 2; 3:: o. a; S: a S; :52 33-0
imam- :5- :53- Soc- 3%: 83. mg: 2:: 25.3- ES- Eng: 39.3.3
5%.: :3 33 $3 $8: 2.3 £3 3: £3 £3 EN: 25255?
......fl.2 :3 25.0- 33. ES: .33.. Sod- ES 25¢. 83. 22: 95.55.5110:
$3-9 9:32 5:132 532 Z 5:102 932 Z :5:sz 562 Z

wo:o$§fl 130% 9250:. mfimmwm

 

.18mi1:-%£m~.:1:: v.5 13.95500 50.: $33.81, .593 :< Afifiom + 51.1% + BUAMZOV
8.5 HOLMZO 13251 7:830 + 85:3 * moth * 8.8 m0 03.3 1:: mm 8093 9:511an 1:m Smmpcmwhum E mm 1.25.91 m:
2.1: 1:11:35 .3325 mo 35:: :00: + 333E111 mm 12:111 31: mwfgfl :00: 1322.8: 53:15 2: m: umgmamm Aways»:
13wm§<<flr§mm mm 12501 Emma: :0 53m: 139:1: 53:15 93 mm 8:35chng .ufim:9:oO 3 13.5%: we. :ozdxzxfiacu
.1329: :00: Q5 EEEmwOH 9: mm 38 “SISEmS .Amoomv :2:qu 1:1. :an £89.50 5 188mm: .3. V815 3:5:950 2: m:
$15-D 638:me t 52 2:12am 9: :o v :5: 3 @2095: 23 1:: 12a 3:1158 mo :93 13:3 :0 1.94:: 2:.va 1:: mfiwfii
0:: £093 95% 5320 H .oom 1:: 3.3 50315 m:o§mmsvow 8.3.» :83 ENE :0 3093 WD 0:615 31m mo @7125: 29:% 2E

mfiumfidum huwaasm
NA 23:8

40

 

 

 

 

 

 

 

 

 

Raw Va: QNH Mtcm gmfiv 035:?» m0 :OSpOQOum
:3 go 2.0 :1 :0 $32520
mg 80. Ed 95 2:0- gammy?
”3 33 and- So 3:. mags/.360
20. Ed no mod $5 $23
£6. and- 2:0- $3 8.3 3&0
3.0 mg- mg. mg moo 255%
:5 fig Em EN 3
5:0:09000 Emmczflm ”m 702.0%
23. 3.0- 34:- :5- 2.0- 30. 30. one 2.0 a: of E: 22533022.3.5
8.3- mg. mg- 30 mg 3.0 m3 Sc 8.0- 3.? and- ”we- aéamvmfi
”no 2: :5 :0 3o :0 2.0- :0- mac- :5- and- :3. 0515.530
5.0- Ed- mad- 8.9- mg- 23. 2.? woo 3.3 o: m: a: ”$200
3.0- was- 33. :3 3o :0 8.0 :0 mg mg mg mg 3.20
and- mod- mod 30 so 3o- 80 mod owe .80 one 2.9 3554
mcom wcom mesa moon Hoom coca amafi wamfi baofi mmafi mamfi vmmfi
wofin—SE‘Q/ .mfifimdvzm mo mwflom QSFH n< Ecxmnw
01.00%

x0005 003030. 23 mo 0m0$>0 330005 2: m“ Qqaeafimnfix mmmo 50b 33300. 955.5 m0 0§30> .3200 $5005 0%. mm 0533/
00:0Q .Soomv #0005303 80¢ @5005 m0 $00 0350mm m0 302500 mnfiU 93 mm 33.0 A3005 wfivmuoa 0:0 5 was: 553 0.8a
m0 00500005 9: .00 002900 .fl 305$ .A0~.:30> 9:00ng .00:0Qv\_:.5§m~_\/ * 0:3 m0 A330 fifiv wEmzv 0m0.§§ .3502: 2:
- Amcomv ~:::E< E m: 10:23. E ~5fi§v~ .3057. mi: E 01.: H 3:53: w: 7.0.5.535 «5.: 2: .3 .Efig»: km??? 0.: 1.209: < ESL

 

mmmbmam 30:09:00 3&0:th
m4 2an

41

Table 1.4
Likelihood of being targeted by a blockholder

This table reports the estimates of an instrumental variables probit model. The sample
consists of 195,984 firm-month observation of 88:? 1500 firms between 1994 and 2005. The
dependent variable, Block, equals one when a new block is formed in a particular firm-month
and zero otherwise. Illiquidity is the first principal component of the liquidity variables
as defined in Table 1.3. Performance is the industry adjusted return on assets defined as
EBITDA/ (lagged assets). log(market cap) is the logarithm of book market capitalization
as reported by Compustat. Leverage is the industry adjusted book leverage ratio defined
as debt / (debt + book equity of equity). G-Indezr is the Governance index as reported in
Gompers, Ishii, and Metrick (2003). Constants are included in the model but not reported
for brevity. Standard errors are clustered by target firm. The symbols *, ** and **" denote
significance at the 10%, 5% and 1% levels, respectively.

 

 

Spec 1 Spec 2 Spec 3
Illiquidity Block Illiquidity Block Illiquidity Block
Decilnalization —5.53*** —4.25"* -3.59*“
* (l/P) (3.18) (3.09) (2.36)
Eighths * (l/P) —4.29 -12.73*** -13.90***
(0.86) (6.74) (9.84)
(l/Price) 12.01“ 1801*“ 1839*”
(2.31) (9.90) (14.46)
Illiquidity -0.00 -0.05*” -0.05***
(0.01) (3.70) (3.48)
Performance. —().08 -0.29*** -0.18 —0.21**
(0.38) (3.95) (0.70) (2.65)
log(market cap) -0.44”* -0.10*“ -0.58’”* -0.08"*
(20.50) (11.37) (18.38) (6.67)
Leverage 0.54“M 015*” 0.52“” 0.09”
(6.55) (4.13) (5.26) (2.08)
G-Index 0.01 0.00 0.03“M 0.00
(1.37) (0.10) (3.42) (0.67)
Mgr Own % 0.02“M —0.01*”
(9.46) (6.00)
Observations 157.763 157 765 131.174 131.171 69.577 69.577
R2 0.24 0.57 0.00
\Vald test 8.01 14.98 7.87
\th > \‘3 0.0046 0.0001 0.0050

 

Table 1.5
Determinants of the size of initial blockholding

This table reports the estimates from a Tobit model where the dependent variable is the
ownership percentage as recorded in the initial blockholding filing (i.e. Schedule 13D or
Schedule 13G), the lower bound is 5% and upper bound is 100%. The initial sample of
14,027 blockholder acquisitions in which the filer that obtains a 5% or greater stake in any
S&P 1500 firm between 1994 and 2005. Illiquidity is the first principal component of
the. liquidity variables as defined in Table 1.3. Active means that the blockholder states
that it pursues action by the management of the target as reported on the Schedule 13D.
The actions range from discussing business strategy to a hostile takeover attempt. (See the
Appendix for SEC instructions for reporting Item 4 - Purpose of Transaction) Performance
is the industry adjusted return on assets defined as EBITDA/ (lagged assets). log(market
cap) is the logarithm of book market capitalization as reported by Compustat. Leverage is
the industry adjusted book leverage ratio defined as debt / (debt + book equity of equity). G-
Indez is the Governance index as reported in Gompers, Ishii, and Metrick (2003). Standard
errors are clustered by target firm. T-statistics are in parentheses. Constants are included
but not reported. The symbols *, ** and **" denote significance at. the 10%. 5% and 1%
levels, respectively.

 

 

 

Spec 1 Spec 2
Illiquidity Size Illiquidity Size
Illiquidity 1.01 1.366
(2.75)*** (2.26)**
Decimalization * (l/P) -2.47 —3.220
(2.65)*** (6.05)***
Eighths * (l/P) -4.93 -1.627
(1.04) (0.75)
(1 /Price) 7.30 2.77.5
(1.45) (1.13)
Active 0.164 6.393
(3.31)*** (13.48)***
Performance -0.445 0.199
(1.90)* (0.29)
log(market cap) -0.651 0.906
(25.65)*** (2.20)**
Leverage 0.812 -0. 430
(7.41)*** (0.71)
G-Index 0.015 -0.104
(2.00)“ (4.()4)***
\Observations 14.027 14.027 9.546 9.546

 

43

 

weed

wood

 

25¢ 25.: See :5: See :5: E.
3&2 3.2 93.2 mewfi Sea 3.2 95.2 woman massassflc
:2: :9: 25.8 80.3 as: 34.3 fine sea

was as- 2.0. 8.: Be Se 9:. 23 33.25
:23: 3...:va 1.3.3 1.34.3

m2. 8s E Q: ..._E_.é.
32.5.3 £33 .1993 as: :Lfle 32s. a 123.3 gene

as- Ed. 2d. no.0- and- a3- 2:- :5. 3V .. 3
figs $25.3 2;: 8.0.3 isms 2%.: :5: 9a.:

and. m2- 8:. mod. 42. as- was- 3.9- E c.3285
38.5 icon 2%.: as: 1.23.3 tend e4: read

was- :3- Ed- 2.0- go. 3.5- :e- was- 3 56:52:.

SENES 813-526 3356 2445.0 258-450 82.21.50 335.0 2.755

 

..».E>Eu.ime.~ £75.: ex; :2: RE. (53 2:
fit. eibtomEmfi 323: .1; was .1. .... £0357? ESL .memezéoemm E was murfifixmé. .Amzzom + .mfixem + HULmZCV .35 HUQMZC
tissue AmsosaO + 3.5sz * 82m * 5.8 m0 035 9: me $003 £59:an Use Emmmtmwemm E 30. Ezcec E 2%: as.\§e.e.:~
.553 Eviea was me. 532 wen—aways e3? ammo 2: cm: H .935 2036833 #53 2: 95ch .936 S @338 E35 52?: we as:
mazes: com ms? #255 noise: e5 eueEsmm H .3505 ”Boreas on... ma? £532 BEE: $.58an 05 8.59:8 fl .Ecousuexm
E 63.5%: flew QEmSEpo 3:93:55 gm: $5 meow was «6.3 seesfime 9535863 chasm x003 www.mH we mumwmzoo 295?. 3H.

AmmaflUV mafia—mm _wauo:£< o>$fl§550 unoaeunsonndw uewumfl
win 2an

44

Table 1.7
Likelihood of Voice and Exit

The sample consists of 278,987 blockholder-month observations in which the blockholder ini-
tially files a Schedule 13C. The dependent variable equals “voice” if the blockholder switches
to a Schedule 13D filing, “exit” if the blockholder reports a ownership level less than 5%.
The omitted case is if the blockholder continues to be passive (filing the required Schedule
13G every year). Cumulative return is the stock return since the initial blockholder filing.
Age of position is the number of months since the initial blockholder filing. Change in size is
the change in percentage of shares of the target owned by the blmtkholder from the initial fil-
ing. The classification of blockholders into Financial, Corporate, and Individual/Partnership
is based on the blockholder response to Item 14 on Schedule 13D (or Item 12 on Schedule
13G). Standard errors are clustered by target firm. Constants are included in the model but
not reported for brevity. The symbols *, ** and *** denote significance at. the 10%, 5% and
1% levels, respectively.

 

 

Spec 1 Spec 2
Voice Exit Voice Exit
Illiquidity 0.16*** -0.06*** -0.068 -0.139***
(2.79) (6.49) (0.67) (10.00)
Cumulative return -0.029 -0.020
(0.41) (1.20)
Age of position (months) 0.015 -0.001
(1.60) (0.73)
Change in size ‘70 0.036 -0.0428***
(0.73) (5.56)
Corporate 1.056* 0708* * *
(1.83) (12.3)
Individual/ partnership 1688* * * 0.269** *
(4.23) (2.69)
Performance -4.139*** -0.378**
(4.88) (2.17)
log(market cap) —0.308** -0.146***
(2.51) (10.5)
Leverage —0.181 0136*
(0.33) (1.93)
G—Index -0.041 -0.002
(0.56) (0.37)
Observations 278.987 278.987 198.647 198,647
Pseudo R2 0.001 0.001 0.012 0.012

 

45

 

25.0
ES
118.3

omd

5.:
:3
“Shawna
3.9-
$22.3
.23

flood
3mg:
***Aw©.mC
cad

isms
8.0-

:35
mod-

*meflv
So

83
32:
figs
So
82:
No.3-
A53
:3-
***Awm.wv
a:

wood
95.:
:22.me
mmd

$5.3
.2;

£33
2.3-

:me3
no.3

mood
32:
***:m.mv
wed
A38
Hmdu
$.78
mad-
***Amm.wv
DNA”

need
893
***A©H.omv
and

***:h.wv
cod

NE

253$»;er

fizfiéav

Miuzxfim

~5C¢SECD

3517:: 93(12: FEEH

Tvzmfihfiisblza “Z

322::x:c.Y:cZ * 2:62.72:

EUEEE

 

g

g

E

5

E

E

 

 

IZ~.~C .HCCL

39$? 2; ~53

:bzi ~r512::ngsaz

I.»

2.22;.“ 72.5; 35.1.:

533.5%?“ £75.: N; fizz ”KB. .33 E: p: 3.525537.
£23:me NEH APE fifibfi KS Uohflmio Em ESE EE‘EEm .02: mo 54‘ €251.35 2: >3 $5.551 .3:
wagogxooi 0.6 Egatmccuzoz A02 Estonom so NH Em: .Hov Om; Escoaom no 3 Eu: 3 95592 $305303 2: :0 ~53; .£.
as.£m§§fidn~§d§32€£ was £95er90 ”38:65,“ 9:: 98392003 mo :OEmoEmmfio 23L .Uevgufi PS :8» a3; 2: 5.5 9:8:r5fai
mfifimmmv: firs .055 Eco 6 5:38 5 .532 @2th 95:0: wepmsgw $4.55 2: mw 293.23, Eofizeauv 2: .n 1:5 w £5515
:H .532 @02va wEEon >8“ o5 fl 223:3 ”EmwchmU of pm; .3538 5 55m, >575 305 €03 m5 :2?» 3% 3a of U21.
fizz:52:32:: €33 2: .3 33. 3: E: :3 .552; Ammo $2.53... 2: mix: #55 2: we 4:3 .2; £25.: fists: $522; it _.;::~.::... H

mafia—emu wfigoaxoflm

m4 mzmfi

46

Table 1.9
Choice of blockholder action

The sample consists of 1,297 blockholdings in which the blockholder initially files a Schedule
13D requesting a change in company action. I use a multivariate logit model in which the
dependent variable equals “CEO” if the filer requests a change in chief executive officer,
“merger” if the filer requests a completion or rejection of a proposed merger, and “other” if
neither (the omitted case). Performance is the industry adjusted return on assets defined
as EBITDA / (lagged assets). log( market cap) is the logarithm of book market capitalization
as reported by Compustat. Leverage is the industry adjusted book leverage ratio defined
as debt / (debt + book equity of equity). G-Inder is the Governance index as reported in
Gompers, Ishii, and Metrick (2003). The symbols *, ** and *** denote significance at the

10%, 5% and 1% levels, respectively.

 

 

 

Spec 1 Spec 2
CEO Mergers CEO Mergers
Illiquidity 0.02 -0.13*** 0.16 -0.129**
(0.13) (3.62) (0.56) (2.14)
Performance 2.67 0.83
(0.68) (1.05)
log(market cap) -0.04 -0.07
(0.11) (0.99)
Leverage 1984* -0.45
(1.67) (1.36)
G-Index -0.01 -0.02
(0.10) (0.62)
Constant -4.24*** -().31*** -3.96 0.35
(13.26) (5.39) (1.32] [0.64]
Observations 1.297 832
Pseudo R2 0.01 0.012
Success rate 75% 50% 73% 65%

 

47

Chapter 2

Employee Ownership of U.S.
Institutional Investment

Management Firms

48

Institutional investment management firms are the single largest category of in-
vestors in the U.S. equity market. These firms manage portfolios on behalf of insti-
tutional clients such as pension funds, university endowments, and charitable founda-
tions. As of December 2005, institutional investment management firms (11M firms)
managed $5.8 trillion worth of U.S. equities, almost a. trillion dollars more than the
second largest category of investors, mutual fundsl. Despite the massive amount of
wealth controlled by these firms, surprisingly little academic research has focused on
this industry.

Although IIM firms all offer the same primary service, delegated portfolio man-
agement, many different organizational structures coexist within the industry. One
structure which varies widely is employee ownership2. Slightly over half of the firms in
this industry are wholly employee owned, 29% have no employee ownership, and the
rest are partially employee owned. In this paper, we address the questions: Why do so
many different employee ownership structures coexist among firms providing similar
services? Does the variation in employee ownership structures predict performance or
investment behavior?

\Ve take an optimal contracting approach to answering these questions. We view
employee ownership as one tool IIM firms use to reduce agency problems with their
employees. Of course, Il.\l firms can use other incentives to motivate their en'iployees.
They will select employee ownership only when the benefit of employee ownership
outweighs the cost relative to alternative incentives. We identify variables measuring

these costs and benefits and analyze the determinants of employee ownership.

 

1There is some double (bunting between these two categories of investors. Many mutual funds out.—
source portfolio management to IIM firms. For example, Vanguard markets the Vanguard \Vellington
Fund. Stock selection for this fund is outsourced to \Vellington h'lanagement Company. The Van—
guard fund is counted as part of mutual fund assets. and these assets are counted again as part. of
\Vellington’s assets.

2Throughout this paper we use the term ownership to refer to ownership of investment manage~
ment firms. We do not use this term to describe the holder of portfolio securities.

49

\Ve begin by examining aggregate employee ownership at the firm level. Consistent
with an optimal contracting equilibrium, employee ownership is lower when it provides
less benefit, which we measure in several ways. First, when many people contribute
to the success of the firm the free rider problem is greater and the incentive effect. of
employee ownership is weaker. \Ne find strong support for this argument. En’iployee
ownership is decreasing in the number of professional employees, the number of invest-
ment styles offered by the firm, and the amount of assets under management. Second,
one benefit of employee ownership is that it reduces the need for firms to monitor
employees, but this benefit is small when the cost of monitoring is low. Consistent
with this idea, employee ownership is lower in firms with a large proportion of indexed

funds.

Our next step is to look within firms, and examine several factors which predict
individual employee ownership. First, we find that portfolio managers who manage a.
large proportion of their firms’ assets and portfolios, and thus generate a large pro-
portion of their firms’ profits. have higher ownership. Second, when an employee has
multiple roles within the. firm, ownership will be an attractive incentive, because it. re.-
wards the value maximizing allocatitm of effort across tasks. Consistent with this idea,
we find that portfolio managers who are also firm executives have significantly higher
ownership. Third, en’iployee ownership creates a. strong link between the firm and the
employee. This is less costly when there is low uncertainty about employees’ quality.
we find that portfolio managers with longer tenure have higher ownership. Fourth,
ownership provides a. coordinating incentive that encourages cooperation within firms.
W’e find that portfolio managers wlmse investment. style overlaps with their firms” dom-
inant style have higher mvnership. which we interpret. as due to the benefits of creating
an incentive to share information and methodologies. These results hold even after

including firm "fixed effects.

If firms and employees optimally allocate ownership, then employee ownership will
not predict performance. If employee ownership predicted performance. then firms
would alter their ownership structure and clients would alter their investment flows to
eliminate the outperformance. However, because ownership rarely changes shocks to
the economic environment may have resulted in suboptimal ownership structures. This
does not appear to be the case. Consistent with equilibrium, we find no relationship
between firm level employee ownership and risk adjusted performance. After control-
ling for firm characteristics, there is no evidence that portfolio managers’ ownership

of HM firms predicts performam1e.

Employee ownership alters firms’ risk taking incentives in two ways. First, em-
ployee ownership reduces risk sharing decreasing employees’ incentive to take risk.
Second, employee owners are less likely to be terminated, reducing their career con-
cerns and increasing their ability to bear risk. We test which of these effects dominates
by regressing portfolio risk on employee ownership of IIM firms. The results Show that
portfolios managed by employee owned firms, and portfolios managed by individual
employee owners, have higher tracking errors, betas, and standard deviations. Even
after including firm fixed effects, rmrtfolios managed by employee owners have higher

risk.

Our work contributes to the literature on agency problems in delegated portfo-
lio management. Most of the existing literature has focused on the agency problem
between portfolio management firms and investors. For example. Almazan, Brown,
Carlson, and Chapman (2004) show that explicit investment. restrictions are. a sub-
stitute to other control mechanisms used to protect. investors. Chen. Goldstein, and
.liang (2008). Del Guercio, Dann. and Partch (2003), Khorana, Tufano. and Wedge
(2007), Meschke (2006), and Tufano and Sevick (1997) show that mutual funds boards'

characteristics explain fee setting and restructiiring decisions. Deli (2002) empirically

examines mutual fund contracts and finds that variation in fund fees is consistent with

rational contracting.

While there have been many studies of the agency problem between investors
and portfolio management firms, there has been far less research on agency problems
within portfolio nn-inagement firm. Clearly these agency problems are linked. To
minimize agency problems with investors. portfolio management firms must control
their employees. The earliest studies of agency problems within portfolio management
firms, Chevalier and Ellison (1999) and Khorana (1996) focus on the role of career
concerns in aligning e1111')loyees' interests with the firm. Specifically, they show that.
poor performance leads to termination. Gervais. Lynch, and Musto (2005) derive a
model explaining these. empirical findings. They assume that mutual fund families are
better informed about portfolio managers (uiality, and they can credibly signal their

information to investors by terminating some portfolio managers.

In addition to disciplining portfolio managers through termination, investment
management firms can reward en'iployees for good performance. Kempf and Ruenzi
(2008) find that relative performance within mutual fund family results in risk shift.-
ing, which they attribute to competition among portfolio managers for promotion and
access to resources. Farnsworth and Taylor (200(5) survey portfolio 111anagers about
compensation. They show that performance based bonuses are widespread. Interest-
ingly, bonuses are usm—illy discretionary rather than formula based, and investment
performance is not the prinmry determinaiit. Khorana. Servaes, and Wedge (2007)
show that mutual fund manages who invest. in their own funds have positive risk-
adjusted 1')erformance.

Our paper continues this line of research on agency problems within portfolio
management firms. \Ve make several novel ("outributions. Most importantly. this is

the first study of the role of employee ownership in portfolio manageu‘ient companies.

Employee ownership is related to firms' economic structure and investment behavior,
and is widespread but it has not. received any prior academic attention. This is
also the first study of agency issues within the institutional investment management

industry.

More generally, our results contribute to the literature on employee ownership and
its role in controlling agency problems. There is an old debate in the finance literature
on the effect of employee ownership on performance. McConnell and Servaes ( 1990)
and Merck, Shleifer, and Vishny (1988) argue that employee ownership of publicly
traded corporations has an observable effect on firm value. On the other side, Demsetz
(1983), Demsetz and Lehn (1985b), and Himmelberg, Hubbard, and Palia (1999)
argue that competitive pressure forces firms to optimally allocate ownership, and that
employee ownership varies depending on its costs and benefits. Further, they believe
that any observed relationship between ownership structure and firm value is the result
of an omitted variable bias caused by failing to include factors which affect ownership.
Our results are consistent. with the optimal allocation of ownership within firms, and
we do not find a relationship between employee ownership and performance. Our
sample includes a large number of private firms, and so we observe a much greater

range of en‘iployee ownership than previous studies.

The final contribution of our paper is that we provide a. detailed description of the
structure and organization of institutional investment. nn-magement firms. The few
prior studies of HM firms have. primarily focused on performance persistence. Busse,
Goyal. and VVahal (2007). Christopherson, Person. and Classman (1998). Coggiu,
Falmzzi, and Rahman (1993) and Lakonishok, Shleifer, and Vishny (1992) all find
evidence of HM firm performance persistence. Del Guercio and Tkac (2002) focus on
fund flows rather than performance persistence, and show that the performance fund
flow relationship is linear for lll\l firms. \Vith our focus on employee ownership we

r

03

provide a far more. detailed description of the structure IIM firms than prior papers.

2.1 The Institutional Investment Management In-
dustry

Like mutual funds, institutional investment. management (HM) firms provide del-
egated portfolio management services to their clients. However, IIl\I firms differ from
mutual funds in several important ways. First, mutual funds directly own their port-
folios. and mutual fund shares represent claims on these portfolios. IIM firms provide
security selection services, but typically their clients directly own the securities. Sec-
ond, IIl\=‘I firms have large minimum im'estmcnts, and their clients are institutional
investors and wealthy individuals. Third, unlike mutual funds, the portfolios do not
have a board of directors to protect investors’ interests, and there currently exists
little SEC regulation.

All IIM firms offer their clients portfolio management products. whicl’i represent
security selection services in a specified investment style. Frequently, within a product
each clienth assets are held in separate accounts. A product’s performance is a. value-
weighted composite of the constituent accounts. Each account in a composite will
hold the same portfolio. subject to some variation resulting from differences such as

social responsibility screens and diversification restrictions.

2.2 Theory and Hypotheses

Ownership structures vary widely in the. institutional investment management i11-
dustrv. Our U‘oal is to understand whv this \vl'ariation occurs. and its relation to in-
u ('5 t... .

vestment behavior. To examine these questions. we use an optimal cmitracting per-

spective. We view employee ownership as one tool IIM firms use to align employees’
interests with the firm. In equilibrium, IIM firms and employees will jointly determine
ownership by trading off the costs and benefits, while also considering alternatives.
The IIM industry is highly competitive for two reasons. First, the barriers to entry
are low. Second, clients can withdraw funds under management. The combination
of these two factors creates strong product market competition. Fama and Jensen
(1983a) argue that firms with the lowest cost structure will be the ones to survive
competition. Failure to efficiently solve agency problems will result in higher prices or
worse performance, and eventually firm failure. To survive, IIM firms must. optimally

allocate ownership.

2.2.1 The Costs of Employee Ownership

Fama and Jensen (1983b) and Demsetz and Lehn (1985b) state that employee
ownership is costly because it requires risk-averse employees to hold undiversified
portfolios. As a result, equity is worth substantially less to employees than to di—
versified outsiders. Eirij;)l(1)yee ownership may also distort incentives and encourage
employee owners to invest in low risk projects to reduce their personal risk, even if
these investments are inferior. Empirically this implies that, all else equal, firms with
lower risk will have higher employee ownership3. As many of the firms in our sample
are private, we do not directly observe firms' equity price volatility, and we do not
directly observe profits. However. profits are a. function of assets managed4, so the
volatility of a firms assets under management will be strongly correlated with firm

risk.

 

3Another important consitleration is that emplm’ees' ('mttside wealth should influence their will-
ingness to own their employer. Unfortunately we cannot observe employees wealth.

1l\lost firms are coimwusated as a percentage of assets under management. Many firms also offer
clients the option of paying partially through incentive fees.

Ci“!
(4’!

Employee ownership provides a strong incentive when a small number of individ-
uals control the key decisions that determine firm performance. However, Fama and
Jensen (1983a) argue that when decision making is dispersed throughout the firm,
the incentive effect of ownership is diluted, and free riding will occur. As decision
making becomes increasingly dispersed throughout the. firm, individual specific incen-
tives such as salary and bonuses, become relatively more efficient. Thus we expect.
employee ownership to decrease as the number of business segments, employees, funds,

and breadth of products offered increases.

Employee ownership reduces or eliminates a firm’s ability to terminate an em-
ployee. This is not entirely a bad thing. Chevalier and Ellison (1999) Show that fear
of termination causes portfolio managers to herd. By reducing the probability of ter-
mination, en'iployee ownership helps to create the correct ex ante incentives. However,
ex post, once a firm has acquired additional information about an employee’s skill,
the option to terminate is valuable. Because employee ownership makes termination
costly or impossible, it reduces the firms options. When the firm is certain that an
empkryee's quality is high. it is less costly to give up the option to terminate the em-
ployee. Since the firm learns about employee quality over time, employee ownership

should be higher for employees with long tenure.

Lakonishok, Shleifer, and V ishny (1992) show that the HM industry is composed
of two segments: a small set of large firms which offer generic products and compete
by offering low costs and stability, and a large number of small boutique firms offering
specialized niche products. Economies of scale are more important for generic low
cost products. Since high employee ownership limits firms ability to raise. exterue-tl
capital, we expect that. employee. ownership will be low for large firms offering generic
products and high for small specialized firms.

r

.36

2.2.2 The Benefits of Employee Ownership

The most obvious benefit of employee ownership is that it creates an incentive
to exert effort. Of course there are other incentives, such as bonuses. profit. sharing,
and career concerns. \V-r expect employee ownersl‘iip to be high when the benefits of
employee ownership are high and the costs are low relative to alternatives.

Relative to other incentives, an advantage of employee ownership is that it correctly
aligns employees’ incentives with the firm. This will be especially important when
employees have multiple roles within the firm, and the correct allocation of effort
across roles depends on employees information. This information asymmetry will
weaken other incentives. Bonuses will be inefficient because the firm does not know
how the employee should allocate their effort, and profit sharing will cause employees
to trade long term value creation for short term profits. Fama and Jensen (1983a)
argue that when decision rights and decision control are held by a single individual it
is optimal for them to have ownership. As a result, we expect. employees who manage
multiple portfolios or who manage the firm to have higher employee ownership.

In the previous subsection, we stated that. employee ownership is a diluted incentive
because its value depends on the actions of all employees. However, this can be
beneficial if the firm needs to coordinate employees actions. There is evidence that
cooperation within portfolio management firms is important. Farnsworth and Taylor
(2006) show that firm perforn'iance has a larger effect on bonuses than individual
performance in HM firms. Pomorski (2008) shows that. information sharing between
mutual funds within a family is impm'tant, and that. information sharing is higher when
funds have similar styles. Since information sharing is more valuable for products with
similar styles. we hypothesize that that employee ownership should be higher when a.
firms product offerings are. concentrated in a narrow range of investment styles.

Fama (1980) argues that career concerns will align employee interests with the firm.

5?

However, Holmstrom (1982) shows that when effort is unobserved and output is noisy,
career concerns will usually fail to fully align incentives. For example, career concerns
diminish close to retirement resulting in reduced effort. Morrison and Wilhelm (2004)
develop a model of partnerships, which closely resemble many of the HM firms in our
sample. In their model, firms benefit when senior employees mentor young employees
to transfer soft skills. However, because mentoring is unverifiable and noncontractible,
senior employees will underinvest in mentoring. Employee ownership is a solution to
this problem, as senior employees can sell their equity to the younger employees at
retirement5. The price paid at retirement will depend on the retirees’ prior investment
in mentoring. Because employee ownership can be sold at retirement, it provides an
incentive for older employees. Empirically this implies that longer tenure employees

will have higher ownership.

Employee ownership is the only incentive that does not require external monitor-
ing. This implies that ermiloyee ownership should be high when there are large benefits
to aligning incentives and external monitoring is difficult. Chen, Goldstein, and Jiang
(2008) and Cren‘lers, Driessen, Maenhout, and Weinbaum (2006) argue that monitor-
ing is more valuable and more difficult when products hold risky assets, actively trade,
and have high turnover. Thus, we. expect to see higher employee ownership for firms
i'nanaging portfolios with these characteristics. Cremers, Driessen, Maenhout, and
\N’einbaum (2006) also argue that there are economies of scale in portfolio monitorine,
which suggests that employee ownership will be higher for firms with less assets under

111anagement.

”Ai‘iecdotally, we are told it is common for retiring employee owners to sell their equity to junior
ermiloyees. We thank Steven M. Levitt of Park Sutton Advisers for helpful discussions on this point.

2.2.3 Employee Ownership and Performance

Does employee or portfolio manager ownership reliably predicts performance? In—
tuitively it seems that employee ownership and skill should be p(;)sitively related. Port-
folio managers who know their skill is high will form their own firms and existing firms
will offer their most skilled employees ownership. However. this does not necessarily

imply an observable relationship between employee ownership and alpha.

If employee ownership is optimally determined in equilibrium, two forces will pre-
vent employee ownership from predicting performance. First, if employee ownership
caused outperformance competition would cause firms ”to alter their ownership struc-
tures until the outperformance was eliminated. Second, clients select firms based on
expected net..-of-fee alpha. Even if employee ownership predicts, but does not cause
outperformance, it would affect fund flows. Berk and Green (2004) show that if there
are decreasing marginal returns to scale in portfolio management, then in equilibrium
clients will allocate money to firms with predictably positive alphas until expected
alpha is zero. Given that Chen, Hone. Huang, and Kubik (2004) and Pollet and W il-
son (“2008) show that mutual funds have diminishing marginal returns to scale, this
suggests that fund flows will eliminate performance predictal’nlity. However, employee
rarely changes and fund flt.)ws may not fully eliminate predictability and so we test if

employee. ownership predicts performance.

2.2.4 Employee Ownership and Risk Taking

Fama and Jensen (1083b) argue that risk-averse enmloyee owners will choose to
reduce firm risk, because of their undiversified holdings. However, employee ownership

reduces or eliminates firms ability to terminate employers. which (flecreases employ-

59

ees’ career coneernsG. Prior studies show that mutual fund managers’ career concerns
affect portfolio risk. Chevalier and Ellison (1999) show that younger managers. whose
termination-performance relationship is stronger, take on less unsysternatic risk and
hold more conventional portfolios. Khorana (2001) shows that following poor perfor-
mance, portfolio managers increase portfolio risk prior to termination. These papers
suggest that employee ownership will affect risk taking, but the direction of the re-
lationship is unclear. Ownership will increase employees’ rewards from positive out-
comes. But for negative outcomes the ownership has two competing effects: employee
owners will suffer direct losses if their firms’ products underperform, but they have

lower career concerns. Whether higher potential rewards and lmver career concerns

outweigh potential capital losses is an empirical question.

There are two types of investment risk affecting IIM firms: asset price volatility
and tracking error. Since fees are based on a percentage of assets managed, revenue
will fluctuate along with asset prices. Firms can control this risk by managing their
portfolios’ betas and standard deviations. Fund flows are heavily influenced by perfor—
mance relative to a. benchmark as shown by Del Guercio and Tkac (2002) and James
and Karceski (2000) show that institutional funds have. a linear 1)erft)rmance—flow rela-

7

tionship and underperforming a benchmark results in significant outflow-rs. Portfolio

managers can reduce the risk of outflows by tracking the benclnnark closely.

 

6In our san'iple. we find a. very strong negative relationship between employee ownership and
terminatitm. for both key personnel and port folio managers. Results are available from the authors
upon request.

7Del Guercio and Tkac (2002) and James and Karceski (2006) show that the perforniance— fund
flow relationship is linear for IIM firms’ products. We find a similar result in our sample.

60

2.3 Data

We use two datasets in this study: a. panel of Form ADVs8 filed with the SEC, and
the PSN Database produced by Informa Investment Solutions. All IIM firms with at
least $25 million in assets under management are required to file Form ADV9. Firms
must file Form ADV at least annually and more frequently if there are material changes
to the. firm, including changes to owners controlling more than 5% of the firm. “7e
have a panel of all Form ADV filings from 200010 through 2006, including the filings
of defunct firms. This panel should be comprehensive and survival bias free, because
firms are legally required to file Form ADV.

The PSN database11 is designed for plan sponsors and consultants to identify
potential asset managers. It contains information on investment performance as well
as firm and portfolio characteristics. Although the PSN Database begins in 1979, we

use only the portion that overlaps with our Form ADV data from 2000—2006.

2.3.1 Employee Ownership of Institutional Investment Man-
agement Firms
\Ve obtain information on employee ownership of HM firms using informatimi from

SEC Form ADV. Schedule A of Form ADV requires each firm to list all direct owners

with a stake greater than 33% as well as all executive officers and directors regard-

 

“Active IIM firms’ most recent Form ADV filings are available at. : http://www.adviserinfo.
sec.gov/IAPD/Content/Search/iapd_0rgSearch.aspx.

UIntentional misstatement, deliberate omission, or failure to file Form ADV is a federal crime. In
practice, criminal prosecution is rare and firms are brought into compliance by the threat of legal
action.

“’VVe have all FOl‘lll ADV filings from January 1, 2000. However. firms can file as little as once per
year. and so we use Form ADV information from the beginning of 2001 to be certain our sample is
complete.

llBerzins and Trzcinka (2005) and Del Guercio and Tkac (2002) use Mobius Group's :\l—search
database. In 2000 Informa Investments purchased and integrated the Mobius database into the PSN
Database.

61

less of ownership. Each owner is required to list their title or status within the firm.
Schedule B identifies indirect ownership, which is common as many employees own
equity through layers of trusts and holding companies. Both schedules report own-
ership by categories rather than exact percentages. \V'e impute ownership using an
algorithm described in the Appendix. Because non-executive owners with less than
5% ownership are not required to report, we do not observe ownership stakes below
5% for non—executives. However, we will observe employee ownership that represents
meaningful control rights over firms’ operations.

Employee ownership of firms is common in the institutional investment manage-
ment. industry. Table 1 Panel A shows 72.07:. of firms have employee ownership greater
than zero. 'We include three measures of employee ownership: the. largest position,
the sum of the three largest positions, and total employee ownership. The summary
statistics are calculated conditional on employee ownership greater than zero. Clearly,
employee ownership is concentrated. The average largest position is 56.8%, the average
top three positions is 78.2%, and average total employee ownership is 89.5%.

\l’e also look within firms, and measure individual portfolio managers" ownership.
By combining Form ADV data with portfolio manager names from the PSN database
to identify portfolio managers who are also employee owners. Table 1 Panel A shows
that 17.5% of the products in the sample are managed by portfolio managers with
at least a. 5% ownership stake in their firm. Conditional on non-zero ownership, the

average portfolio managers ownership is 52.1%.

2.3.2 Institutional Investment Management Firms

Table 1 Panel B shows there are 1.118 firms in the intersection of the. Form ADV
sample and PSN database. Table '2 Panel A shows the firms divided into four cate—

gories: zero employee ownership. minority employee owned. majority employee owned.

62

and wholly employee owned. The majority of firms are wholly employee owned and
28.8% have no employee ownership. l\‘lajority employee ownership is about twice as
common as minority employee ownership.

Table 2 Panel A shows that firms with zero employee ownership have far more
professional employees and manage more separate products. l\r-‘Iinority employee owned
firms have moderately more professional employees and products than majority and
wholly employee owned firms.

From Form ADV we. (_)bse.1've if firms have additional business segments engaged in
the following business activities: broker-dealer, registered representative of a broker
dealer, commodity trading, real estate, insurance, banking, and other financial prod-
ucts. we calculate the variable, Other Business Segments, as the sum of the number
of additional business segments. Index (76 is the percentage of the firms’ assets under
management in index products. Table 1 Panel A shows that very few firms offer iii-
dex products. This segment of the market is dominated by a few large firms. Firm
Portfolio Turnover is the value weighted annual turnover across a firm’s products.

We measure the homogeneity of a firm’s products with the variable Style Herfind-
ahl. This is the sum of the squared percentage of total assets under management
invested in each equity style. We use 12 equity style categories based on four size
categories: all, large, mid and small, and three style categories: value, growth, and
core. The average Style Herfindahl is 0.89 indicating that most firms focus on a narrow
sector of the equity market. Style Herfindahl is higher for employee owned firms.

Form ADV requires firms to list additional services provided to portfolio clients
from the following list: financial planning, pension consulting, selection of other ad-
visers, publications of periodicals or newsletters. security rating or pricing services,
market timing services, and other. N on—Portfolio Services is the sum of the additional

services prm'ided to portfolio clients. The majority of firms in our sample do not

63

provide any additional services.

Average assets under management is $17.8 billion but this figure is highly skewed,
median assets under management is only $1.2 billion. Table 2 Panel A shows that
firms with zero employee ownership are much larger than the other firms. Minority
employee owned firms are considerably larger than majority or wholly employee owned
firms.

Equity is the largest component of assets under management. and more than half
the firms have only equity products. Employee owned firms are more focused on
equity products. Table 2 Panel A shows that employee owned firms manage fewer
international products, more small cap products, and marginally fewer core equity

products.

2.3.3 Portfolio Manager Ownership of Institutional Invest-

ment Management Firms

In the combined PSN and ADV sample, we obserye all portfolio managers who
own at least 5% of their firm. Table 1 Panel C shmvs there are 3,118 distinct. portfolio
managers in our sample, who on average manage 1.9 products.

Key Person is an indicatm' variable that equals one if a portfolio manager is also
an executive officer of the firm. Table 1 Panel C shows that. 18% of portfolio managers
are also executive officers. Table 3 Panel A shows that portfolio managers who are
also executive officers are more likely to have an equity stake in their firm.

We include two variables to measure the importance of a portfolio manger within
the firm. Proportion of Products Managed is the. number of products managed by a.
portfolio manager divided by the total number of products offered by their firm. Pro-

portion of F irm’s Assets Managed is the total value of assets controlled by a portfolio

64

manager divided by the total value of assets managed by their firm. Table 3 Panel A
shows that. both of these variables are higher when the portfolio manager is an owner.
Tenure is the number of years the portfolio manager has been at the firm. Table 1
Panel C shows that on average portfolio managers have been at. their current firm for
10 years. Table 3 Panel A shows that portfolio manz-igers with an ownership stake

have longer tenure than non-owners

PM Index is the proportion of assets managed by a portfolio manager in index
products. Very few portfolio managers control index products. PM Turnover is the
portfolio managers’ value weighted average turnover across the products they manage.
Table 3 Panel A shows that portfolio managers with an ownership stake have lower

t.1_1rno\-'er.

We. include two variables that measure the similarity between a. portfolio managers
products and their firms products. PM Style Complement is the percentage of the
firm‘s total assets under management in the same equity style as portfolio Iiianagers
products. For example, if a portfolio manager controlled a single small cap value port-
folio, and 35% of the firms assets under management were invested in small cap value
products, this variable would be 0.35. For portfolio managers with multiple products
it is the value weighted average across their products. P:\1 Asset Class Complement
is calculated in the. same way, but measures the asset class overlap between firms and
portfolio managers" products. Portfolio managers with an ownership stake in their

firm have higher values of both variables.

Table 3 Panel A shows that pcn‘tfolio managers with an ownership stake usually
are more focused on equity, but the differences in equity style are not large. These

results are generally consistent with the firm level findings.

65

2.3.4 Product Performance

The product returns reported in the PSN Database are a composite of returns on
clients’ accounts. Accounts within the same product. can have different returns for
a variety of reasons, such as social responsibility screens and diversification criteria.
A products” return is a value-weighted average of all an IIM firm’s accounts with a
similar investment style”. The SEC checks reported returns during random audits of
IIM firms. Table 1 Panel D shows summary statistics of the returns reported to the
PSN Database. The mean monthly return is 0.72%. To risk adjust returns, we use

two variations of the Carhart (1997) four factor rnodell3.

Bit = (ii + v‘lfli(R./l[f — th) + fiQiSJl/[Bt + [.337inllth + 4,134Z‘PR12t + (it (2.1)

In the first version, denoted Forward Carhart Alpha we estimate the Carhart model
over the 24 months following the measurement of ownership. We also estimate a. one

period Carhart alpha as:
(lit 2 Return” — [ti-llijylt — th) +j321'81'lth +1332ijl-[Lt +f3421PR12tl (2.2)

where the coefficients are estimated using data from the previous 24 months i.e. t-
24 to t0. As a robustness check, and because it is common practice in industry,
we include benchmark adjusted abnormal returns”. We calculate this benchmark,

denoted as Russell alpha. as the geometric mean return on the fund over the 24 months

 

”PM the rules governing the calculation of composite returns see http://wwwgipsstandards.
org/.

13Factor returns are from Ken French's webpagc.

”we assign each fund to one of 12 size/style groups. There. are three styles: core. growth, and
value, coupled with four size groups: large. mid, small. and all. Vt’e use the appropriate Russell
size/style index except for large core, where we use the S&P 500 index. as the PSN reports that
the S&P 500 is the most widely used index for this group. Russell Indexes are the most common
benchmark for all other groups.

66

after measuring ownership, minus the geometric mean return on the appropriate style
matched index. Table 2 Panel C shows that the returns and alphas are similar for
firms with different. levels of employee ownership, and Table 3 Panel B shows returns

and alphas are similar regardless of portfolio managers’ ownership of their II.\‘I firm.

2.3.5 Comprehensiveness and Survival

The fact that participation in the PSN database is voluntary may create two prob-
lems: selection bias and survival bias. Selection bias will occur if firms’ decision to
participate in the PSN database is correlated with characteristics of interest. To ex-
amine this issue, we compare the PSN data with statistics from the Conference Board
(2007) report. We take the percentage of the total U.S. equity market managed by
all institutions and subtract off mutual fund and hedge fund holdings. The remainder
is IIM firms’ holdings. and direct. stock ownership by insurance companies, pension
funds, and endmvments. PSN firms manage 90% of this remainder. Because the
remainder contains direct ownership by other institutions, it should be larger than
the value of funds managed in the PSN database. Since the unexplained remained is
relatively small, it is suggestive that the PSN dataset is reasonably comprehensive.

Because particij'mtion is voluntary the PSN dataset may contain a survival bias.
There are three forms of survival bias: backfilling, liquidation bias, and non-reporting.
Liquidation bias occurs if the terminal returns reported for a fund do not include the
terminal returns from dissolving the fund. Non—reporting bias occurs when a. firm
strategically ceases reporting following poor performance. As discussed in Busse,
Goyal, and \Vahal (2007). the PSN database has not. permitted backfilling since 1994.
We use only post 1901 data. in our sample and so backfilling is not an issue. In their
study, Busse, Goyal. and VVahal (2007) examine the PSN database and conclude that

it does not have a. meaningful survival bias.

6

‘sl

Table 4 Panel A shows summary statistics of armualized firm level survival. On
average 3.5% of firms cease reporting each year. We divide disappearing firms into
two categories: cease filing Form ADV, and continue to file. The majority of firms
that exit the PSN dataset also cease reporting to the SEC, suggesting that they have
genuinely not. survived. I-Imvever. each year 1.3% of firms exit the PSN database. while
continuing to file Form ADV. It is possible these firms cease managing institutional
money but continue other activities requiring them to file Form ADV, so 1.3% is an
upper bound on the firm level survival bias. Even if some surviving firms exit the PSN
database, this will only bias our results if exit is correlated with employee ownership.

The differences in survival across ownership categories are not statistically significant.

Table 4 Panel B shows summary statistics of annualized product level survival.
Product exit, is lower than firm exit because large firms have more products and higher
survival rates. In an average year, 1.8% of products exit the PSN database. Slightly
over half of these cases occur when the. firm managing the product exits the PSN
database, but 0.7% of products exit while the managing firm continues to report other
products to PSN. Unfortum-rtely there is no way to determine if the firm has genuinely
closed the product or if the product. still exists and is not reported. The non-survival
of products in this dataset is lower than that of mutual funds reported in Carhart
(1997). suggesting that non-sm‘vival in this dataset. is relatively low. Once again, the
differences in survival across ownership categories are not. statistically significant.

Studies examining performance persistence are concerned about. survival because
performance has a strong negative correlation with survival. As a result, survival bias
causes researchers to overestimate persistence. Our focus is employee ownership and
so our concern is whether there is a relationship between employee ownership and
survival. To test this relationship. we regress non-survival on employee ownership

using a random effects panel probit model. The dependent variable equals one if it

68

is the last period the product reports. The results in Table 4 Panel C show that the
relationship between employee ownership and non-survival is not. significant. Given
the absence of backfilling, the similarity between non-survival of HM firms mutual
funds, and the fact that ownership and non-survival are uncorrelated, we conclude it

is unlikely that survival bias affects our results.

2.4 Determinants of Institutional Investment
Management Firm Ownership

Table 5 examines the determinants of NM firm employee ownership. We use three
definitions of employee ('iwnership: the largest position, the sum of the three largest po—
sitions, and total emj')loyee ownership. Because employee ownership is bound between
0% and 100% we use a. random effects panel Tobit model.

The logarithm of the number of professional employees is negative and statistically
significant. in all specifications. This result is consistent. with the idea that when
many employees contribute toward the value of the firm. ownership‘s incentive effect.
is diluted. Since each employee receives a smaller benefit from their effort. free riding
occurs. For all models. a one standard (l("\'1‘(1l1()11 decrease in the number of professional
employees implies an increase in employee ownership of greater than 20% relative to
the mean.

The logarithm of the number of products and the Other Business Segments index
are also included primarily as measures of firm focus. Concentrated employee owner-
ship is lower when there are many products. as the number of products grows firms
either disperse equity across more. employees. or avoid employee ownership entirely.
Other Business Segments has a significant and positive relationship with the largest

employee ownm'ship position, but is insignificant in the other two specifications.

69

There is a clear negative relationship between employee ownership and the per-
cent age of a firm's assets under management in index products. There are several very
good reasons for this. First, the cost of monitoring index products is very low. Index
funds are simple and trz‘msparent, the portfolio managers task is clearly defined, and
performance is easy to evaluate. When the costs of alterimtives to employee ownership
are low, employee ownership will be low. Second, the transparency and simplicity of
index funds implies very strong product market competition, reducing the need for
other incentives. Third, there appear to be large economies of scale for index funds.
This segment of the market is dominated by a relatively small number of large insti-

tutions.

Chen, Goldstein, and Jiang (“2008) and Cremers, Driessen, Maenliout, and VVein-
baum (2006) argue that portfolio turnover is related to the need for external moni—
toring. High turnover implies both greater portfolio manager discretion and higher
external monitoring costs. suggesting that employee owned firms will have a competi-
tive advantage in high portfolio turnover strategies. However, the data do not support.
this argument. There is a significant negative relationship between employee owner-
ship and 1.)ortfolio turnover in the first column. and no relationship in the remaining

two columns.

\Ve include two \v'ariables that measure the scope of the firms’ operations. Style
Herfindahl is a l’lerfindahl Index of each firm's investment styles. Non-Port folio Ser-
vices measures the number of additional services that the firm offers to its portfolio
clients. Each v:-u'ial')le is weakly significant in one specification, but. overall these vari-
ables have little significance. The logarithm of total assets under mane-igement. is highly
significant. Firms managing more money have significantly lower employee ownership.
This effect is economically large. If the logarithm of total assets under management

(-lecreases by one standard deviation it implies an increase in employee ownership of

70

between 15%1—35‘70 relative to the mean. This negative relationship is unsurprising.
First. one way a firm becomes large is by taking in a large amount of external equity.
Second, the larger the firm the more difficult it is for any one individual employee
to have a large. impact. on firm performance. Third. Almazan, Brown. Carlson, and
Chapman (2001) and Cremers. Driessen, .\laenhout. and \Veinbaum (2006) find evi-
dence of economies of scale in mutual fund families monitoring of funds, suggesting
that the relative costs of alternatives to employee ownership decrease in firm size.
The remaining variables all control for the type of investment products the firm
offers. The results show that broad based employee ownership, shown in column three
is associated with equity investment. But there is no relationship for closely held
firms. Possibly this suggests that. it. is easier for employee owners to bear undiversified
firm risk when this risk is spread across many employees. The clearest result is that
employee owned firms invest far less in core equity products, and specialize in either

value or growth products.

2.5 Portfolio Manager Ownership of Institutional

Investment Management Firms

In this section, we examine which employees within IIM firms have ownership.
Because of data limitations we limit our focus to portfolio managers. If a portfolio
manager controls multiple. products they are aggregated. resulting in one observation
per portfolio manager per year. The first column of Table (i is estimated using a
random effects panel Tobit model. The second column is estimated using a linear

_ . . f'
panel regression model wrth firm level fixed cffectsl').

 

l‘i’l‘arainetric panel Tobit models with fixed effects are not, consistent. Semiparametric models for
panel Tobit fixed effects are available only when the fixed effect is for the unit of observation. Since
we estimate fixed effects at the firm level. and not the portfolio mam-iger level, the semiparametric
models are not applicable.

71

Key Person equals one if the employee is a key person as defined in Form ADVIG.
Key Persons have responsibility and control over the entire firm. Since they affect
overall firm profits. the incentive effect of equity is not diluted. Further. employees who
are portfolio managers and Key Persons have multiple roles within the firm. It would
be extremely difficult to write a compensation contract specifying the allocation of
effort between portfolio management. and firm nn-magement. Equity ownership solves
this problem, and rewards the correct allocation of effort. The Key Person variable
is highly significant even with firm fixed effects included. In the Tobit model, the
results suggest that Key Persons have ownership stakes about 16% higher relative to
the mean. When fixed effects are included the implied effect is larger. implying a key
person has an ownership stake 75% higher relative to the mean.

We include two closely related variables: the proportion of the firms’ total assets
under management controlled by the portfolio manager, and the proportion of the
firm’s total products controlled by the portfolio manager. When either variable is high
the portfolio manager has a large effect on overall firm profitability. As predicted, the
results show significant positive relationships between both variables and ownership.
The implied effect. of these variables is relatively modest in the panel Tobit, a one
standard deviaticm decrease change in these variables results in decree-ises in portfolio
manager ownership of 2.5% and 0.5% relative to the mean. However. the implied
effect is much larger in the firm level fixed effects regressions. For both variables,
a. one standard deviation decrease is associated with more than a 25% decrease in
portfolio manager ownership relative. to the mean.

There is a strong negative relationship between the logarithm of firm total assets

under management and portfolio manager ownership in the Tobit regression. This is

 

lbForm ADV defines key persons as: Chief Executive Officer. Chief Financial Officer. Chief ()pera—
tions Officer. Chief Legal Officer. Chief Cmnpliance Officer. director, and any other individuals with
similar status or functions.

consistent with economies of scale in portfolio monitoring. The results state that a
one standard deviation change in a firm’s assets under management is associated with
3%-4% higher portfolio manager ownership. Once firm fixed effects are. included this
variable is not significant.

Portfolio mangers with long tenure have significantly higher ownership. There are
many reasons to expect this result. First, it is less costly for the firm to eliminate
its option to terminate long—term employees as there is less uncertainty about ability.
Second, employee ownership makes it significantly more costly for the firm to terminate
a portfolio manager, increasing tenure. Third, portfolio managers with high tenure
likely have greater wealth, increasing their capacity to bear the risk of a large position
in their employer. Fint-illy, skill or some other third variable may drive both tenure
and ownership. In the fixed effects regression a one standard deviation decrease in

tenure is associated with a 15% decrease in employee ownership relative to the mean.

W'e include PM Index % and PM Portfolio Turnover as measures of the cost of
external monitoring. The results are significant and negative for both variables in
the Tobit regression. The result for index fund management is consistent with our
hypothesis that ownership is lower when there is less need for monitoring. However,
the negative result for portfolio turnover was not predicted. Once firm level fixed
effects are included, neither variable is significant.

We include two "\yr'ariables to measure the overlap between portfolio managers’ prod-
ucts and their firms’ products. The results show a strong positive relationship between
portfolio manager ownersl'iip and the ccmiplement of their equity style with their firm.
The firm fixed effects regression suggests that a. portfolio manager whose style comple-
ment. is one standard deviation below the mean will have 9% less ownership relative
to the mean. These results are consistent with Pomorski (“2008), who shows that there

are greater benefits to sharing information when portfolio styles overlap. Portfolio

73

manager ownership is higher in firms which manage primarily ecuiity. However, once
firms’ asset class focus is controlled for with fixed effects. portfolio managers with both
pure equity funds and balanced funds are more likely to be IIM firm owners.

We include controls for the asset class and investment style of the portfolio man-
agers’ products. The panel Tobit regression small cap portfolio managers have higher
ownership. Once firm fixed effects are included neither style or market. cap is related

to portfolio manager ownership.
2.6 Employee Ownership and Alpha

Beginning in this section we limit our sample to U.S., actively managed, equity
products. Most prior empirical results in the managed funds literature are for equity
products, and so restricting our sample allows for greater comparability with the
existing literature and established benchmarks.

We test if employee ownership predicts performance using both firm level employee
ownership and portfolio manager ownership. For each product, we estimate alpha us—
ing the three benclunarking nrethods discussed in subsection 3.4. We use two methods
to test for significance: pooled OLS regressions with standard errors clustered by prod-
uct and size as reconmiemled by Petersen (2009), and Fama—hIacBeth regressions. We
also forrrr equally weighted portfolios of products based on firm and portfolio manager
ownership and estinrate the Carhart (1.007) alpha. we include the logaritlnns of firm
assets under management and product. assets under nranagement as control variz-rbles

following the results of Chen. Hong. Huang, and Kubik (2004).

2.6.1 IIM Firm Employee Ownership and Alpha

Because several authm's have argued there is a. non-linear relationship between em-

ployer-r ownership and performance for publicly traded companies i.e. Morck, Shleifer.

T-l

and Vishny (1988) and l\/‘IcConnell and Servaes (1990). we measure firm level employee
ownership with a series of indicator variables”: Minority, Majority, and Wholly Em-

ployee Owned.

Table 7 Panel A shows some of the coefficients on employee ownership are, signifi-
cant. But it is very difficult. to argue that there is a consistent or meaningful pattern
of significance. ()f the. alpha estimates. forward Carhart alpha is the most precisely es-
timated. Employee ownership and forward Carhart. alpha are not significantly related
in the clustered regression and one coefficient is significantly negative in the Fama-
MacBeth regressions. For the other alphas there is some positive significance between
employee ownership and alpha, but exactly which ownership level is significant varies
across the specifications. Perhaps most striking is the small size of the estimated
coefficients. Most coefficients represent only a few basis points per month and so in

addition to sporadic statistical significance there is little economic significance.

There are three sets of portfolio regression results in Table 7 Panel B. The first
row shows the alpha from Carhart regressions run on firm employee ownership sorted
portfolios. The alphas are insignificant for all pm'tfolios. Because the results in Panel
A Show that firm and product assets under management predict alpha, we. perform
two-way portfolio sort. First, we divide products into two categories depending on
whether the managing firms’ total assets under management are above or below the
median. For small firms none of the alphas are significant. For large firnrs all of the
employee ownership sorted portfolios have alphas significantly different. from zero, but
not significantly different. from the zero employee ownership portfolio. Second, we
divided products into two categories depending on whether there product total assets

are above or below the median. There are no significant difference in alpha across

 

1‘If we estimate this relationship using the percentage of the HM firrrr owned by employees or a
quadratic specification. instead of indicator variables. the results are not significant.

product total asset and employee ownership sorted portfolios.

2.6.2 Portfolio Manager Ownership and Alpha

Since portfolio performance ultimately depends upon the portfolios manager’s ac-
tions, we examine the relationship between performance and portfolio manager IIM
firm ownership. The results in Table 8 Panel A do not show a clear relationship
between portfolio manager ownership and performance. The coefficients are insignifi-
cant in five of the six specifications. The Fama-MacBeth regression using the forward
Carhart alpha. shows a significant negative relationship with portfolio manager own—
ership. However, given that the most reasonable ex ante prediction was for a. positive
coefficient and only one of six specifications is significant, we interpret these results
as failing to shot" a n‘reaningful relationship between alpha and portfolio manager

ownership.

The portfolio regression results in Table 8 Panel B show a marginally significant
positive alpha for products managed by non—owners and no significant result. for the
products managed by employee owners. The long-short portfolio alpha is significantly
negative. The alpha of the large firm/ zero portfolio manager ownership portfolio is
significantly positive. After performing a two way sort by portfolio managt‘tr IIM firm
ownersl‘rip and firm total assets under management there are no significant differences
in alpha between portfolios. Similz-rrly, after performing a two-way way sort. with prod—
uct total assets, there. are no significant. differences in alpha. between portfolios. These
results suggest that any significe-rnce between portfolio manager Ill\f firm (ms’nership
and alpha is driven by the correlation between portfolio manager IIM firm ownership

and firnr and product size.

76

2.7 Employee Ownership and Risk Taking

Employee ownership has two effects 011 risk incentives. First, employee owners
have a large undiversified stake in their employer, which will create an incentive to
reduce. firm risk. Second, employee owners reap all of the gains from risk taking, and
have lower termination risk. To examine the tradeoff between these. considerations. we
regress portfolio risk measures on employee ownership. We measure portfolio risk with
three variables: tracking error, beta, and portfolio standard deviations. We measure
tracking error as the standard deviation of the difference between a product’s return
and the benclunark return over the 24 months subsequent to measuring ownership.
Table 9 Panel A shows the results of regressions of tracking error on employee own-
ership. We include controls for firm and product. size as well as a set of indicator
variables for the style and market cap of the products holdings. In the first three

columns the t—statistics are based on standard errors clustered by product and time.

The first. colunm shows the results of regressing tracking error on firm employee
ownership. There is a significant positive relationship between firru employee owner-
ship and tracking error. The average tracking error of products managed by wholly
employee owned firms is higher by 0.1% per month than products offered by firms

with no employee ownership.

In columns two and t hree. the portfolio managers’ Il.\l firm ownership is included as
an independent variable. Portfolio manager ownership is significant. and the coefficient
is twice the size of the coefficient. on firm enrployee ownership. These results provide
strong support. for the hypothesis that portfolio manager o\\-'nership reduces career

concerns sufficiently to affect investment behavior.

The last. column of Table 9 Panel A contains results from a panel regression with

firm fixed effects. After controlling for firm level fixed effects. the effect of portfolio

K1
‘1

manager IIM firm ownership is smaller, but the statistically significance is much higher.
Even within a firm, products managed by employee owners have higher tracking error
than products managed by non-owners. This strongly supports the notion that there
is a positive relationship between ownership and risk taking.

Table 9 Panel B shows the results of regressing betas and portfolio standard de-
viations on employee ownership. In the first two columns. the dependent variable is
portfolio beta and in the third and fourth columns the dependent variable is portfolio
standard deviation. Both are estimated over the 24 months after measuring owner-
ship. Firm level employee ownership does not significantly predict betas, but it does
have a positive relationship with portfolio standard deviations. The portfolio manager
results include firm level fixed effects. and find significant positive coefficient for both
betas and portfolio standard deviations. The portfolio manager results include firm
level fixed effects. These results strongly suggest that employee ownership is positively
associated risk taking.

The causal interpretation of these results is that employee ownership reduces career
concerns, resulting in greater risk taking. However, there are alternative explanations.
The reverse causality explanation is that firms grant ownership as a reward for taking
risk. Given that employee ownership rarely changes in our sample this seems unlikely.
Another alternative is that both portfolio risk and employee ownership are driven by
portfolio managers’ risk aversion. Individuals with low risk aversion are more likely

to form their own firms and manage riskier portfolios.

2.8 Conclusion

Eriipltiiyee mvnership of HM firms is common. and there is large variaticm in own-

ership structures across firms. In this paper. we 1')rovide the first. empirical analysis of

78

IIM firm employee ownership. We View employee ownership as one tool that IIM firms
use to control the agency problem between firms and employees, and we argue that in
equilibrium firms should be driven to optimal ownership structures by market. com-
petition. This implies that employee ownership should. vary cross-sectionally based
on firm characteristics measuring the costs and benefits of employee ownership. How—
ever, in equilibrium there should not be an observable relationship between employee
ownership and performance.

We begin our empirical tests by analyzing the determinants of employee ownership
at the firm level. Then. we look within firms, and test which portfolio managers
have an ownership position in their employer. Our results are broadly consistent with
an optimal contracting equilibrium. Employee ownersl‘iip is higher when its value is
greater. Within firms. we find that portfolio managers who are also firm executives,
who manage multiple products, or who manage a large proportion of their firms’ assets,
have significantly higher ownership.

Next we test if employee ownersl‘iip predicts performance. We fail to find a con—
sistent significant relationship between firm level employee ownership and alpha. We
interpret this result as consistent with an e(_1uilibrium in which firms and employees
allocate ownership optimally. and clients allocate funds correctly given the observable
characteristics of firms and products.

Finally, we test if employee ownership is related to risk taking. Vi-"e show that
employee owned firms‘ products have significantly higher tracking errors and standard
deviations. \IVithin firms. products managed by einpltiiyee owners have significantly
higher tracking errors. betas. and standard deviations than products managed by
ll(’)11-()\\'Il(—'?I‘S. The portfolio nr-amager results hold even after including firm fixed effects.

While there is a. huge body of academic work examining agency conflicts between

portfolio management. firms and their clients. the agency problem between portfolio

79

management firms and their employees has received far less attention. This is the first
study to examine employee ownership of portfolio management firms as a. means of
controlling this agency problem. Overall, our results are consistent with an optimal
contracting equilibrium. in which firms and employees efficiently trade off the costs

and benefits of employee ownership.

80

2.9 Appendix

The firm is defined as the investment adviser or “separately identifiable department
or division” (SID) of a bank. Each firm may have. one or multiple products. Firms
are matched from the SEC Form ADV data to the PSN dataset using a name match

and are verified using a combination of city, state and assets under management.

The employee ownership variable captures the amount. of t he investment firm itself
(i.e. not the assets under management) that is owned by employees of the firm. Our
data source for investment firm ownership is SEC Form ADV. If a firm files multiple
times within a month, we retain the latest filing in that month. The form must be
filed annually and "other than annually” if Items I (Identifying Information). 3 (Form
of Organization), 9 (Custody) or 11 (Disclosure Information) become inaccurate or
Items 4 (Successions), 8 (Participation or Interest in Client Transaction) or 10 (Control
Persons) become materially inaccurate. Schedule A contains information about direct
owners and executive officers. Each CEO, CFO, COO, CLO (Chief Legal Office), CCO
(Chief Corntfliance Officer). director must be reported. Each shareholder with a direct
ownership of greater than 5%, all general partners, and these limited partners and
members that have right to receive upon dissolution or have contributed more than
5% of the capital must report ownership on Schedule A. On Schedule B, all indirect
owners that have a. 25% interest in any entity listed in Schedule A are recorded.
Using Schedule B we find the true controlling ownership stake of each entity listed
in Schedule A. Based on the field “Title or Status“, we define whether each entity
is an enmloyee or non-em})loyee. On Schedule A, the ownership is classified into 6
groups: “NA - Less than 5%", “A -5‘% but than “JO/t”. “B 40% but less than 25%",
"C 25% but less than 50%", “D 50% but less than 75%", “E—75‘7t1: or more". To

construct. a single value for each ownership stake, we apply the following algorithm.

81

We sum the number in each ownership group. Starting at “IS—75% or more”, we build
an upper and lower constraint based on the sum of each of the other groups except
“El-75% or more” multiplied by the maximum and mininnnn possible value for each of
the other groups. We then take the midpoint of the maximum and minimum possible
value as the. value for any entity classified as “Ii-75% or more”. We then construct
the constraints for “D 50% but less than 75%" again as above using an upper and
lower constraint based on the sum of each multiplied by the maximum and minimum
possible value for each group except now we omit “D 50% but less than 75%” and
use the value for “E-75% or more” as both the minimum and maximum constraint
for “E-75% or more”. Again, we take the midpoint of the constraints as the value
for “D 50% but less than 75%”. We proceed recursively until we obtain values for
each group, finishing with the smallest ownership group. We verify the validity of the
results of the algorithm by ensuring that each calculated group value falls within the
prescribed range and that the values of all the stakes in a single firm sum to 100%.
For the small number that do not, we correct these entries by hand. (E.g. there are

% or more” ownership stake is

reporting errors where a. single individual with “E—75
listed multiple times for multiple positions: CEO, CCO) We then sum the ownership

stake associated with employees.

82

Table 2.1
Summary Statistics

Panel A shows summary statistics of employee ownership. Panel B shows summary stat is-
tics of firm level variables. Panel C shows summary statistics for portfolio manager level
variables. In Panels A through C each observation is included once per year. Panel D shows
product level summary statistics with monthly observations.

 

Panel A: Ownership

 

 

 

 

Mean SD 25th% Median 75th%
Firms with Employee Ownership 72.6%
Avg. Largest Position > 0 56.8% 29.3 32.5 55 82.5
Avg. Top Three Positions > 0 78.2% 28 62.5 92.5 100
Avg. Employee Ownership > 0% 89.5% 23.2 100 100 100
Products with PM Ownership 17.5%
Avg. PM Ownership > 0 52.1% 33.3 21.3 50.0 82.5
Panel B: Institutional Investment 1\~’Ianagement Firms

Mean SD 25th% Median 75th%
Total # of Firms 1,118 '
Avg. # of Firms per period 843
# of Professional Employees 50 176.8 6 13 29
# of Products 5.06 7.54 1 3 5
Other Business Segments 0.19 0.49 0 0 0
Index % 2.4% 12.9 0.0 0.0 0.0
Firm Portfolio Turnover 65.5% 64.5 26.2 46.7 82.7
Style Herfindahl 0.89 0.19 0.87 1 1
Non—Portfolio Services 0.61 0.77 0 0 1
Firm Total Assets $31 17,783 76.110 270 1,222 5.810
Percent Equity 72.8% 37.0 47.7 100.0 100.0
Percent International 4.3% 17.5 0.0 0.0 0.0
Percent Core 32.7% 41.9 0.0 0.1 79.6
Small Cap % 16.8% 31.9 0.0 0.0 14.6

 

83

Table 2.1 continued
Panel C: Portfolio 1\‘Ianagers

 

 

 

 

Mean SD 25th% Median 75th%
Total # of PMs 3,118
# Products l\lanaged 1.9 2.1 1 1 2
Key Person 0.18 0.38 0 0 0
Proportion of Products Managed 32.3% 34.8 5.9 16.7 50.0
Proportion of Assets Managed 24.7% 32.9 1.3 7.1 37.8
Tenure 10.5 7.9 5 9 14.7
PM Index 0.0% 6.9% 0.0 0.0 0.0
PM Turnover 53.7% 70.9 3.9 34.0 75.0
PM Style Complement 35.4% 40.6 0.0 14.2 74.8
PM Asset Class Complement 53.9% 39.4 9.0 55.9 99.2
PM Percent Equity 70.2% 44.3 0.0 100.0 100.0
PM Percent Core 22.2% 38.7 0.0 0.0 33.3
PM Percent Small Cap 23.3% 40.9 0.0 0.0 21.7

Panel D: Investment Products

Mean SD 25th% Median 75th%
Total # of Products 3.605
Product Total Assets $.\I 1.350 3.868 48.6 251.8 1.016
Unadjusted Returns 0.72% 4.56 -1.92 0.96 3.55
One Period Carhart Alphas 0.05% 1.87 -0.86 0.01 0.91
Forward Carhart Alphas 0.04% 0.48 -0.19 0.02 0.27
Russell Alphas 0.24% 2.48 -0.85 0.15 1.25
Trackil’lg Error 1.90 1.22 1.09 1.04 2.50

 

84

Table 2.2
Summary Statistics by Employee Ownership

This table shows pooled averages of the variables used in this paper. Each column shows
pooled averages for firms that fall into a specific ownership category. Panel A shows summary
statistics for firm level variables. Each firm is observed annually. Panel B contains product
level variables. Each product. is observed monthly.

 

Panel A: Institutional Investment Management Firms

 

 

 

 

 

 

Zero Minority Maj ority 100%
% of Firms in Category 28.8% 5.5 10.9 54.7
# Professional Employees 116.1 23.3 13.6 17.7
# of Products 9.4 4.8 3.2 3.6
Other Business Segments 0.29 0.16 0.20 0.15
Index % 4.8% 6.3 1.6 1.9
Firm Portfolio Tiu'nover 73.0% 70.4 68.6 60.7
Style Herfindahl 0.80 0.87 0.93 0.93
Non—Portfolio Services 0.72 0.52 0.53 0.59
Firm Total Assets $M 55,577 18,254 2,728 4,573
Percent Equity 74.1% 85.4 91.7 83.6
Percent International 13.0% 9.4 8.8 5.3
Percent Core 36.1% 34.2 32.8 30.2
Small Cap % 15.6% 21.7 26.6 19.7

Panel B: Products

Zero Minority Majority 10(_)%
Product Total Assets SM 1,914 936 810 898
Unadjusted Returns 0.74% 0.65 0.82 0.70
One Period Carhart Alphas 0.04% 0.06 -0.01 0.06
Forward Carhart Alphas 0.06% 0.03 0.01 0.02
Russell Alphas 0.25% 0.28 0.18 0.25
Tracking Error 1.8% 2.1 1.9 2.0

 

Table 2.3
Summary Statistics by Portfolio Manager Ownership

This table shows pooled averages of portfolio and product level variables used in this paper
for two ownership categories: products managed by employees with no employee ownership,
and products managed by employee owners. Panel A contains summary statistics at the
portfolio manager level. Each portfolio manager is observed annually. Panel B contains
summary statistics at the product level. Each product is observed monthly.

 

Panel A: Portfolio I\<Ianager Level
Zero PM Ownership > 0

Key Person 10.40% 55.7
Proportion of Products Managed 24.5% 71.4
Proportion of Assets Managed 11.8% 52.7
Firm Total Assets $3M 107,718 3,030
Tenure 9 .8 13.0
PM Index 0.1% 0.0
PM Turnover 56.1% 43.9
PM Style Complement 0.32 0.51
PM Asset Class Complement 0.52 0.61
PM Percent Equity 68.7% 76.6
PM Percent Core 21.6% 24.9
PM Percent Small Cap 12.8% 14.6

 

 

Panel B: Product Level
Zero PM Ownership > 0

 

Product Total Assets $M 1,997 657
Unadjusted Returns 0.64% 0.60
One Period Carhart Alphas 0.11% 0.10
Forward Carhart Alphas 0.12% 0.06
Russell Alphas 0.26% 0.24-
Tracking Error 1.90% 2.0

 

86

Table 2.4
Survival

This table shows annual product. survival of firms and products in the PSNT database.
Panel A shows summary statistics of firm survival for different. categories of employee
ownership. Firms are considered to disappear if they no longer file SEC Form ADV. If
a firm continues to file Form ADV but. no longer reports to the SEC it is identified as
ceasing to report. Panel B shows summary statistics of product survival for different
categories of employee ownership. Panel C shows the result of random effect panel
probit regressions where the dependent variable equals one if it is the firm or products’
last period. Constants are included but not reported. The symbols *, ** and ***
denote significance at the 10%, 5% and 1% levels, respectively.

 

Panel A: Firm Survival Summary
Zero l\"linority Majority 100% All

 

 

Survive 96.0% 96.9 96.7 96.7 96.5
Firm Ceases to Exist 2.6% 2.6 2.4 2.0 2.2
Firm Ceases to Report 1.5% 0.5 0.9 1.3 1.3

 

 

Panel B: Product. Survival Summary
Zero Minority l\iajority 100% All

 

 

Survive 98.5% 98.2 98.1 98.1 98.2
Product and Firm Disappear 0.7% 0.6 1.1 1.3 1.0
Product Disappears 0.8% 1.2 0.9 0.6 0.7

 

 

Panel C: Panel Probit Regressions of Survival

 

 

 

 

Firm Survival Product Survival
Firm Ceases Product and Only Product
Disappears Reporting Firm Disappear Disappears
Employee Ownership % 0.0001 ~0.017
(0.08) (1.63)
PM Ownership % -0.003 -0.001
(1.01) (0.07)
Ln(Firm Total Assets) 0.10 -0.860 0.159 —0.285
(-0.23) (4.73)*** (6.36)*** (1.06)
Ln(Product Total As- -0.293 0.083
sets)
(-14.())*** (-0.39)
Lagged Alpha. -0.613 0.100
(-8.97)*** (-0.27)
# Observations 5.081 5.081 222.721 222.721

 

Table 2.5
Determinants of Institutional Investment Management Firm
Employee Ownership

This table shows the results of random effect panel Tobit. regressions where the dependent
variable is IIM firm employee ownership. There is one observation per IIM firm per year.
In column one, employee ownership is the single largest employee ownership position. In
column two, employee ownership is the three largest positions. In column three, employee
ownership is the sum of all employee ownership positions. Constants are included but. not
reported. The symbols *, ** and *** denote significance at. the 10%, 5% and 1% levels,
respectively.

 

Largest Three Largest Total Employee

 

Position Positions Ownership
Ln(# Professional Employees) -8.278 -9.602 -12.561
(3.32)*** (5.30)*** (4.48)***
Ln(# of Products) -5.722 -8.468 0.957
(3.11)*** (4.43)*** (0.39)
Other Business Segments 4.923 0.101 -3.041
(2.11)** (0.04) (0.73)
Index % —12.046 -44.420 -23.672
(2.36)** (6.25)*** (2.11)**
Firm Portfolio Turnover -0.061 -0.018 -0.006
(4.01)*** (1.13) (0.23)
Style Herfindahl —5.516 14.228 -1.378
(1.08) (2.00)** (0.16)
Non—Portfolio Services -1.223 -1.622 3.618
(0.29) (1.20) (1.77)*
Ln(Firm Total Assets) -7.090 -7.675 -4.464
(9.24)*** (7.89)*** (3.03)***
Equity % -2.140 0.883 24.005
(0.39) (0.20) (3.33)***
Percmit International 1.734 2.726 -5.454
(0.35) (0.57) (0.79)
Percent Core -12.191 -9.053 -27.77'
(2.88)*** (3.48)*** (7.42)***
Percent Small Cap 3.053 3.064 —3.947
(0.74) (0.94) (0.90)
Pseudo R2 0249 0.300 0.231
Number of Observations 2.609 2.609 2.609

 

Table 2.6
Determinants of Portfolio Managers’ Ownership of

Institutional Investment Management Firms
This table shows the results of regressions where the dependent. variable is portfolio manager
ownership of their IIM firm. There is one observation per portfolio manager per year. In the
first column, the coefficients are estimated using a random effects panel Tobit model. In the
second column, the results are estimated using a panel regression with firm level fixed effects.
Constants are included but not reported. The symbols *, ** and *** denote significance at
the 10%, 5% and 1% levels, respectively.

 

Random Effects Firm Fixed Effects

 

Tobit Regression

Key Person 16.150 6.516
(17.54)*** (17.76)***

Proportion of Products IVIanaged 40.131 7.052
(23.06)*** (6.80)***

Proportion of Assets IVIanaged 4.679 7.855
(3.42)*** (12.99)***

Ln(Firm Total Assets) -6.752 -0.075

(25.40)*** (0.26)

Tenure 0.612 0.157
(11.98)*** (8.87)***

PM Index -18.053 1.111

(2.86)*** (0.60)

PM Portfolio Turnover -0.017 0.000

(2.52)** (0.06)

PM Style Complement 10.551 1.992
(7.66)*** (4.31)***

PM Asset Class Complement. 7.405 -2.193
(4.51)*** (4.34)***

PM Equity % —8.279 0.413

(5.58)*** (1.13)

PM Percent Core. 0.561 0.229

(0.49) (0.64)

PM Percent Small Cap 6.070 0.283

(4.54)*** (0.78)

Firm Fixed Effects No Yes
Time Effects Yes Yes
(Pseudo) R2 0.356 0.359
Number of Observations 11.283 11,283

 

89

 

 

week: 332 3:2: $22 332 £19.: 2555530 a
3.: 3o :53. 25s 33 33.3 a:
:34: is: :13: E: 8.0.: is:
55. was- macs- was- :93. so? 2%»... 33. 335253
3.80.: E»: 1:24: 38.: .28.: 3.3::
an; :8 So: as: 23.0 53 Eta”: 3% 25:3
is: ism: E»: as: 3.8:: 2::
god. 32. was. ago 30.0 25s were: 3:25: 33:3
3.: 5...: 13.63.: :53 95.3 25.3
meg- $3.0- once- 83.- mass. so: Ego a....,.3_._:sm £2.52
3.3.: E: so: .95.: 1.3:: :3:
5.2. secs 89o ES :50 :3 Base 3,35: 4.2.2252.
::E< £334 £83.50 S334 £53.80 £33. £334 fic:.§0 27:44 8.23.30
:ommflm Uomponm 0C0 Uhdggorm :Qmmzm UOCOQ 0C0 VEBKWA

 

flames/7:820 US$300

m:c_m.r.w:mmm 6R5 SHE “< 350

 

 

€3.89: .00: :5 c.3533 Be. 3552.50
.3815 60.50.9553: 2: :0 5:3,: :32: 3,5950% 23 #8:: 52:9: vm 58: 23 .815 £238 :88 259503.: 9535:: of :euBfiE
0.598%: 2: 3: 8:3: :wmmsm 2:32 583 :23: E83: 8:5 3:252 553 wemmfl 00 835.5 of 2:5,: 5:2: Eats...
05 E 550: $30.: 35:3 oi we US$598 Be. 927:: fimieU :28: 0:0 52:20:30 :0 30:85.32: 2: “0:530:30 :28: 23:2:
am of $.20 #35:: Cam: €84.30 mi “36: Max:538 2w mange. 9:350 thrash .5353 "5:: 2:3 3 $335.23 $5.53 2353?.
2:? 305.6%»: @200: 05:: £:2£2w3 naemowgeafim “£5389: 0.5 wEm: 8:855? $8 o>> €2.50 8.3350 >535»
...$Eo.€2 .3225: ”832.23 “386:: 5:5 9:20:30 883:8 358:: £5 423?. mbuswoa #56362: 2. 32:21..» 8:25:33:
2: 8.53.5 33mm: 23 we: < Essa .QEmS:Bo mohOEEo :56 2: :0 SEE: 038:8: @3823: we $33: 2: P52? 23:... mEH.

@334 9:: 3:98:30 oohofimam
EN 23:8

90

 

5:: :0: 8:: :3: 01::

 

HOG: OH: mac dyad H3: fi~_n:< :AZCEU 00:73:0w sag;

:2: an: 3:: at: 0.:

m:.:. ::.: ::.: ::.: m5: 234. ....§_._:0 2:55 :35:

39.3 :Acmdv £393 .13ch Ammav

3.: :0: 0.: :0: 3.: 2:2 002.30 :2: 0:3

3...: 2.0:: Gm: $2: :0:

3:- 3.: E: 8.: 5.: 2:2. 22.2.0 5:: :3: 1

30:: 3:: 3:: $0: 3:: 9

8:- S: E: 0:: ::.: 2:2 0.22.2.0
_:.:0:m-m:04 00:30 >225) ..mtSQZ 3:254 0:0N sexton 30:00:30

 

0::m0m 00:.5m 0:009:03 “m 3:.an
... . .. . .. . . 0f. . x D .. - . .. .. .... .-. _ . _
30.50300: $0.7; :5: 1:: :.\:m .32 0f :0 00:00ch? 00:00 .13.. 1:0 in * £0€F§ BEL £03003 ..£:E 35.75.50
00.0150 00: :02: 0:: 303003 ESE 00:30 00.5350 >203? w:0_ 0:02:03 0 :0 0:3: 05 0.50:7. i.:.;m-u:04: _:::_:0
0:5. ...»..3>500000: 0E0. 3:003 :0 ::c :02005 0:: 0: 02:00: 00: .2030. 0:0 .090: .3300: 0:02:03 00:5: 03:0: 2:
5 9:08:30 00...».0350 “0:30 2: .0 £0.00 0:: :0 00:28 00:93:03 mix: 0:20:0Q0: 0:00:00 :5: x33? 2...: m 3:0,:

00:55:00 EN 0309

 

 

mnmroi $1.062 wovfio mono: mvoom: $3.9. 22.55.9830 kw
:oo Hoo moo moooo Hoooo ooo.o mm
3:: :30 12.: 2a.: :3: 5.3
ooo.o omfio Hmoo owmo wmoo ooo.o 0:32:00
***C:o.mv *Lmfimv *umeoov *Lhoév 87C ***Awm:3
mmoo- mmoo- vmoo- foo- ooo.o- goo- Amuumm< :50: 92:00:32:
:58 :2: 13:3 :58 :33 "33::
:oo ooo.o o._o.o ooo.o ooo.o o:o.o $324 :39: :Ziri
:59 :29 $an :58 :33 363
ooo.o ooo.o ooo.o ooo.o ooo.o ooo.o- “N. 9:20:30 :5
23:: . .232 fimfiao 232 ”22:50 232 232 22:50 232 $2220
:0mm:m :03?» 0:0 $835,: :023: :020& 0:0 $33.5":

 

 

:EmmeO
m:0:mm0:m0m 70.22: 9250?: ”d. :25:

:Emuag.2::m

 

 

30:85:20: £3.20: XL :25 0.0m. xxx: 2: a: 00::0:::m:m 30:0: 32 1::
J. 20:25? 27: .303003 ESE QEmbiao 09:01:25 0.5x 0.5:? :25 m_:_0~.;.0.,.0.:: 3.20:: :02r50 00.3350 2.20:3 WE: A.0::A0L::,~A,02: d. :0
:25.q 6:: .0w.::, fizzflt

**

27:: 23 250:0 :fi0:m-m:0:: 5:300 27o 59:95:20: 02m 02:00.5 :0 :5: 35:02: 2:: 0,: 0.25.30: 0.8. .
0:03:02: wofiom 03:0: 0:: :: .QEEQESO 0020350 .Swm:2: 0:09.95: :0 20:: 2: :0 “00:200. 20:02.5: 9:»: 2057.05,: 05:20
:5: 22:3: 20: m 3:45: 20:5 :EE:0:0: 0:0 :0 2:30: :02: 2.30800w 0:: 6:: 23:2: vm 38: 0:0 :25 2:30.: :22: 2.52:00w
35300:: 2:: :0230: 00:08::0 0:: 0.5 20:90. :0nm:m £530: 588 @030: 28.550 22:3 @523: :32: vuwws: :0 5300.3 2:
2:5: 5:9: 28.50 2: E 550.: 3:00: mnwa: 2: 0: @03835 0.5 20:30 022250 :25: 0:0 @3225»: :0 2:222:32:
2: 295020: @020: 5:08 wm 23 :25 3:9: fiooi ta:$0 0:0 mam: #0322520 0:». 22:3: 93:20 02.05.20,: .5508: :5:
2:3 2: @8323 20:0 @328? 5:: 25600ng @200: :5: .2563:me :Emumz-2::m “$0052: 03.: mix: 02:3:2w5 .73
03 :i: 2: :2: 5 9:20:30 “mSwfiHE: 0:038: 2 .05 9:20:30 2n: 6:3: M33005 12:03:05 .2 032.3». 9201:0901
2: 0:2,?» 3:50.: 2: 0:: < 35$ d322§0 .Swm:w:~ 0:398: :0 22:3: 0:80.20: $5223.: :0 £170.: 2: 1.30:1. .033 2::

2:34 6:: 3:20:30 :mwwgmz ozotuom
wd 03:8

 

 

an: $9.3 :630

r5.0. 00.0 2.0 5:34. 0.20.30 0,935.5 3.0.34
3:: 89.3 3...:

00.0- 00.0 #00 337$ €20.26 Sizzm 2.32m
20.3 850v #32098

3.0 :20 2.0 .5002 0,542.0 EEK 93...;
ans $58 E:

N00- 30 00.0 37:4. ,0.:.:_.:U :ZE :E:m
.3253 See 1%.:

000. N00 0H.0 37:4. £22030

...:im-w:aq 0:15:30 2m GEN A..::.t.5m 5:33:33

 

£15.30 035m 233.5% ”m 355
032328 0N @319

93

Table 2.9
Employee Ownership and Tracking Error, Betas, and

Standard Deviations

Panel A shows the results of regressing the next 24 months’ tracking error on employee
ownership. The first three columns show pooled regressions with standard errors clustered
by product and time. The last column is a panel regression with firm fixed effects. Panel B
shows the result of regressing the next 24 months’ betas and portfolio standard deviations
on employee ownership. For both variables, the first column contains results from pooled
regressions with standard errors clustered by product and time, and the second column

 

 

contains results from panel regressions with firm level fixed effects. The symbols *, ** and
“* denote significance at the 10%, 5'70 and 1% levels, respectively.
Panel A: Tracking Error
Employee Ownership ‘70 0.0010 0.0010
(3.16)*** (2.38)**
PM Ownership ‘70 0.002 0.002 0.001
(1.96)** (2.68)*** (4.03)***
Ln(Firm Total Assets) -0.041 -0.035 -0.044 0.012
(4.19)** (3.35)*** (4.27)*** (1.16)
Ln(Product Total Assets) -0.074 —0.074 -0.073 —0.082
(7.21)** (7.24)*** (7.14)*** (50.91)***
Style and Market Cap Effects Yes Yes Yes Yes
Firm Fixed Effects No Yes No Yes
Time Effects No Yes No Yes
Constant 2.274 2.210 2.323 2.534
(20.88)*** (19.17)*** (20.36)*** (26.88)***
R2 0.10 0.10 0.10 0.19
# Observations 97.540 97.540 97,540 97.540

 

94

 

Table 2.9 continued
Panel B: Betas and Standard Deviations

 

Portfolio Betas

Portfolio Std. Deviations

 

Employee Ownership (7C 0.0001
(0.70)
PM Ownership %
Ln(Firm Total Assets) 0.003
(1.71)*
Ln(Product Total Assets) 0.007
(3.48)***
Style and l\-‘Iarket Cap Ef- Yes
fects
Firm Fixed Effects No
Time Effects No
Constant 0.816
(39.61)***
R2 0.07
# Observations 97.540

0.0002
(2.47)**
0.020
(7.52)***
0.010
(23.95)***
Yes

Yes
Yes
0.654
(27.43)***
0.05
97,540

 

0.001
(1.09)**

0.013
(1.45)
-0020

(100*

Yes

No
No
3.477
(25.51)***
0.18
97.540

0.002
(7.71)***
0.087
(7.64)***
0.001
(0.51)
Yes

Yes
Yes
4.149
(40.82)***
0.58
97.540

 

Chapter 3

Fraud and Registered Investment

Advisers

The recent publicized cases of fraud by investment advisers has bought much at-
tention to fraud detection. On December 11, 2008, the Securities and Exchange Com-
mission (SEC) charged Bernard “Bernie” Madoff and his investment firm. Bernard L.
Madoff Invest incnt. Securities LLC, with securities fraud for a multi-hillion dollar Ponzi

1

scheme that he pe.r1_)etrated on advisory clients of his firm. Estimates of client. losses
(included fal'n‘icated gains) amounted to nearly $65 billion.2 Only two month later,
on February 17. 2009, the SEC charged Robert Allen Stanford and three of his compa-
nies for orchestrating a fi‘z-i.uc.lulmit., multi-billion dollar investment. scheme.3 Given the
massive task of overseeing over ten thousand individual investment. advisers. federal
regulators could benefit by knowing which firm attrilmtes, such as potential conflicts

of interest, make firms more likely to commit fraudulent actions. The SEC tries to

examine advisers with a high risk profile on a three-year cvcle and has no set cycles

 

1http://w\vw.sec.g()v/news/press/2008/20t18—293.]itin

2Amir Efrati and Robert Frank “Madoff Set to Plead Gnillv to 11 Felonies “a“ Street Journal.
“’ednesdav. March 11. “2009

'3http://www.sec.gov/news/press/2()09/2t109-26. ht m

96

for others. A better prediction model would also benefit regulators by quantifying the

potential for trouble.

The inx-estment advisers in the. U.S. provide advice to over 20 million clients and
have discretionary control of over $32 trillion in assets. Despite this large sum, reg-
istered investment advisers (RIAs) have very few restrictions imposed by federal law.
Instead, the SEC relies on mandatory disclose of potential conflict of interests. so that
clients can make an informed decision. The recent incidences of fraud by investment
advisers raise the question of whether these disclosures are valuable. Using these
mandatory disclosures by investment advisers, we are able to predict which firms have
future incidences of fraud and other investment—related crime. We find that. conflicts
of interest are associated with an increased level of fraud. Internal monitoring and
aligned incentives lead to 1(_)wer frequency of fraud. The presence of sophisticated
clients is negatively related to the frequency of fraud. Even after accounting for all
the above factors, a history of disciplinary actions against the firm predicts fraud.
Overall, the required disclosure is useful for predicting fraud, and the probability of
events is positively correlated with permissive firm policies and negatively correlated

with internal and external monitoring.

The ability to avoid losses resulting from inadequate or failed internal processes,
people and systems is important for investors using advisers in the United States. The
primary federal law that regulates investment advisers is Investment Advisers Act of
1940. Unlike many countries, in the United States, the federal securities laws do
not mandate a minimum level of experience, specific qualifications, or accreditations.
The laws also do not prohibit advisers from having substantial conflicts of interest.
that. could impact their objectivity. Instead. federal law only outlines the disclosures
that advisers must. provide and leaves the responsibility for selecting the advisers to

investors.

97

we also examine investors reaction to events. Investors appear to only withdraw
funds from the firm when the offense is by investment adviser firm owner. A rational
manager will fire an employee only when the costs of retaining the employee exceed
the benefits. If the employee is also an owner, it may be more difficult to terminate
the employee. Consistent with this view, we find that key persons that commit fraud
or other crimes are fired more often when they are not an owner.

The biggest contribution of this paper is to show that the disclosure mandated by
the SEC is useful in predicting future events, even if we cannot say with certainty
that this is economically efficient since we do not observe the cost of disclosure or
expected losses from operational events. Our work contributes to the larger literature
of prediction of fraud and operational events in other financial companies such as in
hedge funds: Liang (2003), Bollen and Pool (2008), Brown, Goetzmann, Liang, and
Schwarz (2008), banks: Chernobai, Jorion, and Yu (2009), and mutual funds: Cici,
Gibson, and Moussawi (2006) and Nobel, Wang, and Zheng (2006). Our work also
contributes to the literature on firm and market reaction after a scandal in mutual
funds: VVellman and Houge (2005), Choi and Kahan (2007) and public companies:
Agrawal, Jaffe, and Karpoff (1999), Karpoff and'Lott, Jr. (1993), Karpoff, Lee, and
Martin (2008), Fich and Shivdasani (2007), N iehaus and Roth (1999), and Srinivasan
(2005)

3.1 Registered investment advisers (RIAs)

An investment adviser receives compensation for providing advice about individual
securities or managing portfolio of securities for clients. The Investment Advisers Act
of 1940 requires any investment. adviser that manages $25 million or more in client

assets to register with the SEC. while those with less assets under management must.

98

register with the state of their principal place of business. The Office of Investment
Adviser/ Investment Company Examinations reports the existence of over 10.600 IAs
4

with total assets of $32.3 trillion dollars and nearly 20 million clients.

Common examples of investment advisers include pension fund managers, mutual
fund families, and trust fund managers. Individuals, partnerships, or certain corpora-

. I"
tions may also be reglstered under the Act.‘)

W'hile passed at a similar time as the Investment Company Act. of 1940, the Invest.-
ment Advisers Act of 1940 covers a. related but broader set of investment firms. For
example, the investment adviser, Fidelity Management and Research Company (cov-
ered by the Investment Advisers Act of 1940), advises the Fidelity family of mutual

funds (covered by the Investment Company Act of 1940).

Section 203(1))(3) of the Advisers Act exempts from registration investment ad-
visers that during the preceding 12 months have had fewer than 15 clients, do not
advise an investment company registered under the Investment Company Act of 1940,
as amended, nor “hold thmnselves out to the public." as investment advisers. Many
hedge funds use this exemption to avoid registration. A rule passed by the SEC re-
quired hedge fund managers to register by F eln'uary 1, 2006, but this rule was reversed
by the U.S. Court of Appeals for the District of Columbiz—r on June 23, 2006. Further-

more. an investment adviser can be part of a firm that serves many different types of

 

4These values are repm'ted as of October 2nd. 2007. Source: http://www.sec.gov/about/
offices/ocie/ocie_offices.shtml

'5 The definition in the Investn‘icnt Advisers Act of 1940 includes any person or business who, for
compensation. engages in the business of advising others. either directly or through publications or
writings. as to the value of securities or as to the advisability of investing in. purchasing. or selling
securities. or who, for compensation and as part of a regular business. issues reports concerning
securities. The definition also explicitly exempts banks; a lawyer, accountant, engineer, or teacher
whose performance of such services is solely incidental to the. practice of his profession: any broker
or dealer whose performance of such services is solely incidental to the conduct of his business as a.
broker or dealer: and a publisher of any bona fide newspaper. news magazine or business or financial
publication. A “two year lockup" provision is included so that venture capital and private equity
firms are. excluded from registering as an RIA.

99

clients or conducts other lines of business such as insurance or banking. A firm that
qualifies under the Investment Adviser Act of 1940 must. register even if investment.
advice is not its primary business.

With the variety of business models and ways to generate revenue, potential con-
fiicts of interest may arise within the RIA firm. In addition to managing portfolios
for clients, RIAs may provide a broader range of financial planning services such as
insurance, tax, and estate-planning. Other RIAs may provide pension consulting, se-
lection of other advisers, publication of newsletters, security rating, and market timing
services. The investment adviser may be part of a firm that also engages in business
as a broker-dealer, insurance broker, and/ or bank. In these cases, an opportunity for
self-dealing may arise. By law an investment. adviser is considered to be acting in a
fiduciary capacity on behalf of clients with a. higher standard of disclosure and due
care, a commitment to disclose, minimize and resolve conflicts of interest than would
be found in a traditional securities brokerage environment. Registration also requires

that firms adopt a code of ethics.

3.1.1 Conflicts

While an RIA has fiduciary duty to investors, a number of potential conflicts of
interest. can arise depending on the way the RIA is structured. The conflicts can be
mitigated by external monitoring by clients who can leave the firm if they disapprove
of firm behavior. Internal governance mechanisms can align incentives and make fraud
less attractive. The. presence of a. conflict is not necessarily a bad thing; likewise, more
internal governance is not always better. In equilibrium. the added cost due to the
potential for wrong doing owing to a particular business practice must be balanced
out by the benefits of that practice in a competitive industry. Of course, there is no

guaramtee that drew is an equililjirium relationship in the investment adviser industry.

100

For example, the practice of soft dollar brokerage has offsetting costs and benefits.
In the advisory business, a common practice is to direct client trades to a particular
brokerage that may have a higher commission than other brokerages in return for
credits that. can used for proprietary research or other benefits. Since RIAs have a
fiduciary obligation to their clients. those. RIAs that engage. in this practice argue.
that the extra commissions that they pay are fairly compensated through the research
and other products they receive. However. the lack of transparency makes this hard
to observe. Bogle (2009) argues that this practice should be discontinued as the
lack of transparency and accountability subjects clients to abuse by unscrupulous
advisers. Others such as Horan and .lolmsen (2008) defend the practice. They find that
premium commissions are related positively to risk adjusted performance suggesting
soft dollar usage benefits investors. Overall, while the question of whether soft dollars
arrangements benefit clients is still an open one, we predict a positive relationship

with the likelihood of a fraud event.

RIAs can recommend securities in which they have an ownership interest, serve as
an underwriter. or have. any other sales interest. This creates a potential conflict of
interest between the firms’ and the clients’ interests. \Vhile the partiality of the firm
may be questioned in these cases, refusing to advise to clients about these securities

would harm investors by limiting their investment. opportunity set.

RIAs may retain custody of clients’ assets or choose to use an outside service agent.
Keeping custody in-house allows a one-stop shop that produces all the documentation
and holds the money to control costs and provide fast access to assets. However,
the arrangement removes any tl'iird-party checks. In her t.esti1non\-' before the Senate
Committee. on Banking, Housing And Urban Affairs on Ii/2T/2tlflfl. SEC Chairman
Mary Schapiro suggested that the Madoff fraud Would likely have. been discovered

sooner if stricter rules had been in effect governing instances in which RIAs take

101

custody of client assets.

Likewise, a tight affiliation with a. broker/dealer can open up similar issues. While
cost-savings and expediency of service may be gained by in-house broker-dealer ar-
rangements, the lack of outside verification can permit fraud to go undiscovered for
longer periods of time. Affiliations with other financial businesses can also increase

potential conflicts of interest and create opportunities for fraud.

Overall, we predict that the use of practices that introduce conflicts of interest.
or avoid third-party monitoring (such as interest. in client transactions, soft dollars,
internal custody of assets, close affiliation with a broker/ dealer and other affiliations)
will open up the firm to a. higher likelihood of a. fraud event.

External monitors are a potential solution to mitigate operational risk. Sophisti-
cated, powerful investors may be better able to demand relevant information and spot
trouble than less sophisticated investors. Unsophisticated clients may underestimate
the amount of operational risk that a firnr permits. Alternatively, smaller clients may
not have the market power to demand the disclosures that would allow the client to
precisely determine operational risk.

The RIA firm may also use internal controls to mitigate operational risk within
the firm. By having a high level of employee ownership, the individual employees are
more closely tied to the firm, so that reputational penalties may be more effective.
The bonding of employee ownership works in two ways. By tightly tying the employee
to the firm, ownership increases the employees share of the negative effects the firnr
incurs due. to fraud and reduces the profitalnlity of the fraud. Likewise with non-
employee owners. the effect of fraud can be mitigated by promptly firing the employee.
Also, compensation that is tied to performance can help mitigate. wasteful practict-ws
like frivolous use of soft dollars since the managers coinpensation is tightly tied to the

client’s performance. In addition to aligning incentives of employees with both the

1 02

firm and clients. the firm can put in place a more stringent internal compliance system,
such as separating the role of chief compliance officer.6 This may allow the firm to
catch potential problems early on and then fix them. Of course, since the disclosed
data only contains events that the SEC discovers, a better compliance department

may actual increase the probability that fraud is discovered.

3.2 Operational risk

The Basel Committee defines operational risk as: “The risk of loss resulting from
inadeqtu—rte or failed internal processes, people and systems or from external events.”
In this paper, we use a similar definition focusing on the risk caused by employees
and affiliates of the firm, but. excluding operational risks from external events such as
terrorism or natural disasters.

In their paper on operational risk in the hedge fund industry, Brown, Goetzmann.
Liang. and Schwarz (2008) construct a measure of operational risk, the w score, by
canonical analysis using the hedge fund (late-il'x-rse, TASS, and disclosures from Form
ADV. They find no relation between hedge fund investors fund flows and the. a; score.
Brown, Goetzmann, Liang. and Schwarz (2008) also report firm factors like high lever-
age and concentrated ownership are associated with a past history of events.

In the banking industry, Chernobai, Jorion, and Yu (2009) find that operational
losses are related to firm factors such as firm size, volatility. increasing leverage, the
number of employees, and profitability as well as the n1acroeconornic environment.
The firms suffering from operational losses also tend to be more complex and have.

fewer auditor's on the board.

 

“In the case of Bernard l.. Madoff Investment Securities, LLC, Peter Madoff is listed as both the
chief compliance officer and the director of trading.

103

A major caveat in interpreting the literature is that committing a fraudulent act
is related to but distinct from being caught committing a fraudulent act. This dis-
tinction may be meaningful. Operational risk by definition is impossible to observe.
The risk factors can vary according to the internal controls put in place by the firm,
external monitoring and incentives of the employees. However, even a low-risk firm
may experience an event by chance. The observation of operational losses can also
depend on the degree of scrutiny that is placed on the firm. If two firms have equal
operational risk, but one firm is monitored more tightly then we are more likely to
observe. an operational risk event in that firm even if the true risk is the same. Failure
to detect an event early can also impact the magnitude of the loss as was the case
in the Madoff Ponzi scheme. Ultimately, a. documented fraud event are a product. of

both an actual fraudulent event. and the detection of the event.

Regulators may not play an important role if market participants have alternative
mechanisms to combat fraud. Chang and Evans (2007) argue that while corporate
fraud can impose significant costs if left unchecked, evidence shows market mechanisms
discipline much bad behavior. The autlmrs conclude the benefits of criminalization

must. be balanced with the reduction of socially efficient. risk taking behavior.

In a study of the mutual fund scandals, Choi and Kahan (2007) find market-based
penalties for mutual funds provide substantial incentives to adopt an organizational
structure that reduces the likelihood of scandals. However, they also observe that.
investors do not withdraw their assets after scandals that do not harm the investors
(e.g. when fund managers use insider trading to benefit. investors). In their work on
market. timing in international funds, Goetzmann, Ivkovic, and Rouwenhorst (2001)
find that. very limited exploitation of these opportunities suggesting that either the
funds effectively curtail day traders or few investors are aware. of these ormortunities.

In addition to discipline from clients, firms may be. disciplined by the providers

104

of capital. The firm may lose value because of operational events through losses to
intangible assets such as reputation. Karpoff and Lott, Jr. (1993) and Karpoff, Lee,
and Martin (2008) show that the negative effect. for firms that commit criminal fraud
comes primarily from the reputational penalty. Only a small fraction (6.5% in the
Karpoff and Lott, Jr. (1003) study) is directly due. to legal fees and fines. Even after
accounting for changes due to reporting the correct earnings, the majority of the

announcement loss in firm value is due to reputational penalties.

Market reactions or reputation penalties will only occur if the operational risk
event conveys new information about the firm. The occurrence of an operational risk
event may cause clients to update their beliefs about the unobservable operational
risk. However, an occurrence of an operational risk event does not necessarily mean
risk is increased. The firm can also update its beliefs about the operation risk and
engage in efforts to lower that risk. In Agrawal, J affe, and Karpoff (1999), the authors
find that managerial turnover is higher after a fraud event; however, after controlling
for firm characteristics fraud events, they find that these events do not increase the

benefits to managerial turnover.

All activities that. the investment adviser engages in generate some level of opera-
tional risk. For example, the RIA may take legal control of the client’s assets to enable
it. to easily manage the funds. However, this control makes it possible for the RIA to
steal the funds. If the market is competitive, firms should only engage in activities
when benefit more than offsets the cost caused by increased operation risk. The RIA
industry is highly competitive for several reasons. The barriers to entry are very low.
The cost. to file and set up a. new firm is under $1000, as very little physical capital
is needed. The basic service is very fungible: a client can easily remove money and
start using a competitor. This feature of the industry creates strong product market

competition. Fama and Jensen (1983a) argue that firms with the lowest. cost structure

105

will be ones to survive. Failure to mitigate operational risk subject to the costs of
mitigation will result in uncompetitive fees and/or perforn’iance. To survive, RIAs

must optimally mitigate operational risk.

3.3 Data

To disclose the potential conflicts of interest. presence of external monitors and
internal measures, the SEC requires all registered investment advisers to file a Form
ADV annually with the SEC.7 This filing must be updated upon material change -
including the occurrence of a fraud charge, so that. clients and potential clients have
current available information about the RIA.8

We obtain data for our study from the Form ADV filing required for all RIAS for
the years 2001 to 2006. All investment advisers with at least $25 million in assets
under management are required to file the Form ADV. Our panel data includes all
initial filings and amendments, including the filings of now defunct firms. As firms
are legally required to file the Form ADV and we have all filings, the dataset. should
be (:oml‘n'ehensive and survival bias free. A criminal DRP must be filed if a “person
associated with an investment adviser”9 has been charged or convicted of or plead
guilty or nolo contendere (“no contest”) in a domestic, foreign, or military court. to

a. felony or misdemeanor involving: investments or an investment—related business, or

 

7This form is available at http://www.adviserinfo.sec.gov/IAPD/Content/Search/iapd_
0rgSearch.aspx.

8 The form must be filed annually and “other than annually” if Items 1 (Identifying Information).
3 (Form of Organization), 9 (Custody) or 11 (Disclosure Information) become inaccurate or Items
4 (Successions), 8 (Participation or Interest. in Client. Transaction) or 10 (Control Persons) become
materially inaccurate.

U The Investment. Advisers Act of 1010 defines this term as any partner. officer, or director of such
investment adviser (or any person performing similar fin‘ictions), or any person directly or indirectly
controlling or controlled by such investment adviser, including any employee of such investment
adviser. except that. for persons associated with an investment adviser whose functions are clerical
or ministerial.

106

any fraud, false statements, or omissions, wrongful taking of property, bribery, perjury,

forgery, counterfeiting, extortion, or a conspiracy to commit. any of these offenses.

Amended Form ADV filings can occur as frequently as multiple times in a single
day. To create an annual panel data set, we use the current filing as of August Blst
of each year.10 All explanatory variables are measured as of August 3lst. We then
collect information on disclosure reporting pages (DRPS) filed September lst to August
31st of the next. year. We have 51,397 firm-years representing 13,579 unique RIAs.
The number of unique RIAs exceeds the number of active RIAs reported by the OIA
because our sample includes defunct firms. Firm-years with at least one DRP filing

L

are called an ‘event year”, and those without a DRP filing are called a. “clean year”.

In Table 3.1, we observe that RIAs come in a wide array of sizes. While the
median assets under management (AUM) is $100 Million, the mean AUM is over $2.4
billion. In Table 3.1, Panels B and C show the structure of the industry with many
small advisers and a handful of large advisers. Also noticeable is that the number and
frequency of reported events is much higher for larger firms. Firm policies are, vary
greatly among investment advisers. We examine several of these policies in Table 3.2.
Interest. in Client Transaction is a binary variable that takes the value of one if the firm
recommends securities in which it has an ownership interest, serves as an underwriter
and / or has any other sales interest. Soft Dollars is a binary variable equal to one if the
firm receives research, other products or services other than execution from a broker-
dealer or a third party in connection with client transactions. Custody of Assets
is a binary variable equal to one if the firm retains custody of clients’ cash and/ or
securities. Broker/ Dealer is a binary variable that takes the value of one if the firm

reports an affiliation with a bri)ker/dealer. Other Affiliation is a binary variable that

 

l“\\'e choose August 31st to maximize the number of annual observations since our data set of
ADV filings ends in September of 2006. “"e have DRP filings through ‘2007.

107

takes the value of one if the firm reports an affiliation with an investment company,
other investment adviser, bank, insurance company or other financial company. Small
Client Focus is a binary variable that takes the value of one if the reported percent
of individual (non-high net worth) clients exceeds 50%. Separate chief compliance
officer (CCO) is a binary variable that takes the value of one if the person reported
on the Schedule A filing has no job title other than CCO. The CCO of a company is
the officer primarily responsible for overseeng and managing compliance issues within
an organization. Performance-Based compensation is a binary variable that takes the
value of one if the. firm reports that it is compensated based on performance. History
of Violations is a binary variable equal to one if the firm has to file a DRP for a fraud
event during the last 10 years. The history can be removed if the responsible party
is no longer affiliated with the firm or more than 10 years has elapsed. In the first
column of Table 3.2, we observe that firm policies that may cause a conflict of interest.
are common but not ubiquitous. The last variable, History of Violations, is relatively
uncommon. This is not surprising since reputation is very important in the industry.
The policies are fairly stable over time as shown in column two. Over ninety percent

of firms have the same policy in the current. year as they had in the past year.

In Table 3.3 Panel A, we examine the difference in firm policies and fraud. We
split the sample into firm—years with zero DRP filings (clean firms) and firm-years
with one or more DRP filings (fraud firms). 11 Consistent with the view that firms
with more potential conflicts of interest will have more operational risk, we observe
a positive and significant differences for all four internal conflict variables. For clean
firms, less than a third of firms recommend securities to clients in which they have

an economic interest. in nearly tln'ee—fourths of fraud firms, the firm engages in this

 

11To be precise since our data is a panel, a firm may appear in both samples though in different
years.

108

practice. Similarly, the use of soft dollars is 16% higher in event years. The difference
in internal custody of assets is nearly 40% higher in fraud firms. The presence of other
business activities within the firm or affiliations with related firms with these activities
also varies significantly between clean and fraud firms. Broker/Dealer and Other
Affiliations are found in 80% and 90% of fraud firms. Both values are significantly

higher than in clean firms.

In clean firms, only 22% of firms have a primarly low-net worth individual client
base, while the figure is 43%- of firms for fraud firms. The prescence of a separate chief
compliance officer (CCO) is slightly higher in fraud firms. Also, Performance-Based
compensation does not meaningfully vary between the two groups. The most dramatic
results while only 0.87:. of firms have a prior history of an event for clean firms, 27.5%

of firms have a prior history of events for fraud firms.

While these are only unconditional averages and should not be interpreted causally,
there is a strong connection between permissive internal polices and future events.
Similarly, unsophisticated clients are more likely to be present during event years.
Interestingly, the relation between a dedicated CCO and event is positive. One expla-
nation is that a dedicated CCO may be more likely to find fraud causing more. events.
However, the correlation could be spurious is the number of events increases with size

of the firm and larger firms are more able to afford a dedicated CCO.

In Table 3.3 Panel B. we examine the frequency of events by year. Although there
is no strong pattern in the trend, we include year dummies in all subsequent analysis
to guard again faulty inference that could be caused by changing in monitoring by

regulators over time.

109

3.4 Empirical results

3.4.1 Prediction of fraud

()ur main research goal is to predict which firms will suffer an operation event
based on observable firm characteristics. The 1;)urpose is two-fold. First, an accurate
prediction model is useful to investors and regulators to identify firms that have fea—
tures that are consistent with a higher incidence of operational risk events. Second, we
would like to determine a causal relation between firm cl‘iaracteristics and incidences of
fraud events. While our research design limits the possibility of reverse causation, we
still need to be concerned about the possible endogeneity of our explanatory variable
and event variable. We observe the act of getting caught, not the fraud event itself, so
in that our explanatory variable are. correlated with the probability of getting caught
conditional on committing a fraud the inferences will shift. Still, we can answer an in-
teresting question: do certain practices increase the chance of the firm getting caught
for fraud.

We use. the filing of a criminal disclosure reporting page (DRP) as the operational
risk event. Vt’e cmistruct a panel of RIA firm-year observations and then measure
the number of DRP filings over the next year. In our main set. of tests, we estimate
a probit. model using an indicator that equals one when there is one or more DRP
filings over the subsequent year and zero otherwise. We report the results in Table 3.4.
In the first column. we include. five variables that indicate. different types of intermil
conflicts.

\Ye find that having an Interest in Client Transactions. use of Soft. Dollars, Custody
of Assets, affiliation with a. Broker/Dealer. and Other Affiliations all have. statistically
significmrt positive relations with the incidence of DRP filings. This is consistent with

the view that given a. greater latitude to commit fraud. more fraud will be. cormnitted.

110

 

In all specifications in Table 3.4, we control for the number of employees using a series
of dummies related to the range reported on Item 5a of the Form ADV: 1-5, 6—10, 11-
50, 51—250, 251-500, 501-1,000 and More than 1,000. We also control for firm size, firm
age and year effects. As expected. the point estimate of the employee effect increases
as the number of employees increase, and a F—test of the combined significance. of the.
number of employee variables is strongly significant for all specifications.

In the second column of Table 3.4, we include two proxies for external monitoring
that are measures of the client type of the RIA: a dummy if the primary client type
is individual investors12 and average account size. Neither variable is statistically
significant in this specification though the point estimates are the predicted signs.

In the third specification. we examine three measure of internal monitoring. Em-
ployee ownership is derived from the Schedule A and B filings as described in Dirn-
mock, Gerken, and Marietta—Westberg (2009). Employee ownership has a strong and
significant negative effect as expected. Performance-Based compensation has a nega-
tive effect as expected, and a separate CCO is related to a slightly higher incidence
of fraud even after controlling for firm size. However. neither of these variables are
statistically significant.

In the fourth model. we include all explaimtory variables for internal conflict,
external monitoring and internal monitoring. While the model still has the same in-
terpretation as a predictive model, the interpretation as a causal model shifts since
the some of the explanatory variables are determined by the RIA firm, while others
are. set by clients and can be considered an outcome of RIA firms' choices. Overall,
the results are remarkably similar to the inferences from the prior models. The statis-

tical significance remains for all of internal confiict variables except Other Affiliates.

 

')r _ . . . . . . ..., . . ,

l'“I‘lns classification excludes high-net worth investors who have over $150,000 In assets invested
with the RIA or over $1.5 million in total net worth, as well as institutional investors that aggregate
assets from individuals such as mutual and pension funds.

111

Interestingly, the average account size variable is now significant suggesting that after
accounting for the potential conflicts and internal firm monitoring, larger clients are

associated with a lower incidence of fraud.

In the fifth specification, we also add a control for a history of violations. History
of \V'iolations is a binary variable that takes the value of one if the firm reports “yes" to
any question on Item 11 (Disclosure Information). This variable is strongly significant
indicating a prior history of events is a strong predictor of future events. Again, the
interpretation of the other variables shifts as their effect is now conditional on having a
past ewmt. However, the statistical and economical significance of the other explana-
tory variables remains essentially the same. In the sixth specification, we employee.
a random effects probit model to control for unobserved firm-specific cliaracteristics
to address an omitted variable concern. Even with this specification, we still have
qualitatively similar results that are still statistically significant.

Overall, the results are consistent with our predictions. When an RIA firm chooses
practices that permit greater freedom to engage in fraudulent practice, more fraud
is observed. Sophisticated clients invest with firms that have a lower subsequent
incidence of fraud. Also, firms that put strong internal monitoring and incentives to
mitigate. fraud see lower subsequent rates of fraud.

To test the predictive accuracy of our models. I use the Hanssen—Kuipers score
as discussed in Granger and Pesaran (2000). The authors show that the Hanssen-
Kuipers score can be interpreted as average economic value when the payoff ratio is

(onstant over time and equal to the unconditional forecast probability. The score is

 

' fruc )osifil’c. . .
calculated by H A = H + F, where H = . . 1 . , . . 1s the lnt rate
' ' true posz/H'('+julsc positudc
'(zlsc nc( (Mice
and F 2 f '1

. . is the false—alarm r'rte. Usin ‘ the full )robit
qus-e neg/(1t:m9+fruc negative ' ‘ ( g I

model from the last column in Table 3.4, the hit rate. is 20.4% and the false-alarm

rate is ().2(i"<,~ yielding a Halissen—I\'uipers score of 20.2%. \Vc can reject that this value

112

 

equals zero at the one—percent level. A score of zero indicates no skill. This method
is preferable over an accuracy measure based only on the number correct since such a
method is heavily influenced by the most common category and very few firms have
events. For example, a naive forecast of that no firm has a event would score yield a
Hanssen-Kuipers score of zero, but have a 99.6% accuracy. Also, our full-model has a
statistically significantly higher score than a model that. only considers the historical

events, which has a Hanssen—Kuiper’s score of 14.8%.

Anecdotally, our model also fares well. Using the most recent filing of the Bernard
L. Madoff Investment Securities Form ADV available online (dated 01/07/2008), we
find that the projected value. operation risk places the firm at the 97th“ percentile
of all firms in our sample. The firm’s custody of its clients assets, close association
with a broker/dealer and history of violations13 are all factors for the high score.
Another firm that has recent media attention is Stanford Capital Management. Like
Bernard L. Madoff Investment Securities, Stanford has custody of its clients assets,
close association with a broker / dealer and a history of violations14 as well as various

interests in client transactions resulting in a 99th percentile score.

Given the strong significance of the number of employees and incidence of fraud,

 

13According to the Investment Adviser Registration Depository (IARD), the firm had two DRPs on
file. On 07/06/2005, the NASD alleged the firm failed to display immediately customer limit orders
in N asdaq securities in its public quotation. Without admitting or denying the allegations, the firm
consented to the described sanctions and to the. entry of findings, was censured and fined $7,000.00.
On 02/26/2007, the firm admitted violations of limit order display and limit order protection. VVith-
out admitting or denying the allegations, the firm consented to the sanctions. was censured and fined
$8,500.00.

14011 04 /'12/ 2007, the firm held customer funds without making required reserve computations
and to make deposits into a special reserve bank account for the exclusive benefit of the customers.
The findings stated that the firm failed to establish and maintain a supervisory system reasonably
designed to achieve compliance with applicable securities laws, regulations and NASD rules in that
it. failed to provide each of its branch offices with copies of its written supervisory procedures or an
equivalent document regarding the timely processing of customer checks. The findings also stated
that the firm conducted a securities business while failing to maintain its required minimum net
capital. \Vithout admitting or denying the findings, Stanford Group Company consented to the
described sanctions and to the entry of findings, therefore. the firm was censured and fined $20,000.

113

we run two other checks to ensure that our inferences are not due to the differences
in the size (number of employees) of firms. First, in Table 3.5, we. again use the
same specifications as in Table 3.4 with one change. Instead of including number of
employee dummy varialfles, we split the sample into small and large firms, with small
defined as less than 50 employees. Overall, the results are ctmsistent with our previous
estimates. The differences in the sample levels of the explanatory variable can account
for the difference in statistical estimates. For example, all large firms in our sample

have at least one affiliated business so we cannot estimate an effect for that variable.

For our second robustness check, we account for the fact that. a firm can have
multiple DRP filings for different events in a single year by estimating a count. model
where the depei’ident variable is the number of DRPs filed in a single year. we use
a negative binomial instead of a Poisson model due to overdispersion of DRPs (the
variance is greater than the mean - thus Poisson model is inefficient with downward
biased standard errors). We also use the same sets of explanatory regressors as in
Table 3.4. ‘We report our results in Table 3.6. Again, our point estimates are qual—
itatively siniiilar and all of the explanatory variables except for that Broker/ Dealer
remains significant in the full specification (5) and Other Affiliates liiecmnes statisti-
cally significant... Overall, our robustness tests confirm that our earlier results are not

due solely to difference in firm size or clustering of DRP filings in a. single firm.

In Table 3.7. we perform a. similar analysis to predict DRP filed for owners of the
RIA firm. We find that we get qualitatively similar results, although the statistical
significance is diminished for the conflict of interest variables. Employee ownership
remains strongly negative even though there is a positive mechanical relationship

lgn‘étween employee ownership and an owner DRP filing.

114

3.4.2 Investor Reaction

\Vhile prediction of fraud events is clearly important to regulators, do these events
matter to investors? One way to answer this question is to look at investors react
to these events. Choi and Kahan (2007) find a negative outflow by fund investors in
response to the annoimcement of mutual fund timing scandals. Investor reaction also
varied by the degree the scandal negatixcly impacted the fund investors. VVellman
and Houge (2005) find that. other funds in same family also suffer outflows suggesting
that. investors update their beliefs about risk at the firm level. Interestingly, using a
sample of hedge funds, Brown. Goetzmann. Liang, and Schwarz (2008) find no relation
between their risk measure from Brown. Goetzmann, Liang, and Schwarz (2009) and
flows. There results may differ because their sample only contains a single cross-section
of ADV data at the end of their sample period and thus can only tell whether a firm
has had a reportable incident during the last 10 years, but not the number or timing
of the incidents. Also. their sample consists of only hedge funds. They interpret the
findings as hedge fund investors either disregard the ADV or already have the disclosed
information from other sources. H(.)wever, their data does not allow them to rule out
other plausible explanations such as: investors react immediately or the firm could
change after events (e.g. altering policies and/or firing employees).

Unfortnnately, the required data on the Form ADV is insufficient to back out
investor flow information for RlAs. “/0 therefore use data. on portfolio holdings from
the required SEC 13F filings. One limitation of this approach is that this restriction
changes our sample from all Rl.-'~\s with over $25 million in AUM to those firms that
have at least $100 million in eligible securities. However, this reduced sample contains
the. majority of firms with a DRP filing.

The Spectrum 13F institutional investor holding database includes all long equity

positions on securities that trade on an ()Xt‘:liaiigc or NASDAQ. closed-end funds.

115

 

some equity options, warrants, and some convertible debt, but no information on
short positions or derivatives. We match the Spectrum 13F to Investment Adviser
Public Disclose (IAPD) data using a name match and verify our matches using asset

under management and place of business that is given in both the 13F and ADV
Assets” —(Assetsiqt_ 1*Retu'r'nsiqt_1)

 

filin 1‘s. W’e c'rlcul-tted flows using Flow- 2
f“ ( ( b lat A88€3t8i t—I

as suggested in Sirri and Tufano (1998).

In Table 3.8, we report regressions of flows using the combined data set. Total DRP
is the total number of DRPs filed in the year in which flows are calculated. Owner
DRP is number of DRPs filed in which the reported person is an owner of the firm.
Prior Return is the return for the prior year. Log(Portfolio Value) is the value of the
reported assets in the 13F filing. Four style dummies, large-growth, large-value, small—
growth, and small-value are calculated as in Bushee (1998). In the first specification,
we see a negative, but economically small and statistically insignificant effect of a
DRP on firm flows. However, in the secmid specification, we see that. an owner DRP
has an ectmomically and statistically significant effect of roughly -30%. This negative
effect of owner DRPs is robust. to inclusion of overall firm DRPs, returns, employee,
and style dummies. The evidence is consistent. with investors having little reaction to

overall DRPs, but only reacting when the event. is associated with a firm owner.

One. reason that our results differ from Brown. Goetzmann. Liang, and Schwarz
(2008) may be that data limitations force the authors to use a. risk measure based
on a ten year aggregate 1119? sure of events. Therefore, they only observe factors that
correlate with a. history of events and not the events directly. Another reason is that

we study all RIAs. while their study focuses on hedge funds.

116

 

3.4.3 RIA reaction

Prior work on firm reaction to fraud finds mixed evidence on employee retention
after a. fraud event generally implying negative job market penalty for being associated
with a fraud. Srinivasan (2005) find directors experience significant labor market
penalties. Niehaus and Roth (1099) finds CEO turnover around securities class actions
lawsuits, and this effect is larger for successful lawsuits. Farber (2005) finds a positive
association between fraud detection and subsequent improvements in board quality
and audit committee activity. However, Agrawal, .laffe, and Karpoff (1999) find little
evidence of turnover of management or senior directors after the revelation of fraud
after control for firm char-attteristics. Fich and Shivdasani (2007) also finds that outside
directors do not face abnormal turnover after the rmrelation of f and, but subsequently
hold less other board seats.

In this section, we investigate one possible explanation for the results in the prior
section. While a firm can quickly rid itself of DRP filing by firing the offender, a. firm
may be less likely to do so if the person is an owner.

One limitation is that while we see all firm employee who report. a DRP, we do
not all the complete set of employees who never report a DRP. V’Ve do however have
information on all key persons of the firm. A key person is an executive officer, director
or owner with influence over firm policy. \V'e can use this information to (‘letermine
when a key person leaves the firm, but unfortunately we do not know why they leave.
This leaves us with a relative small sample of 87 key people.

Using the key person data found on Schedule A. we. construct a. sample of all
key persons who file a DRP. In Table 3.9 Panel A, we examine the unconditional
probability of departure from the firm. The probability of leaving the firm is much
higher for non-owners than owners. Also. key persons who file a. DRP are. more likely

to leave for both owners and non—owners. However, the increase in likelihood for non-

117

owners is 5.95% versus an increase of only 0.62% for owners. This is consistent with
non-owners being more likely to leave the firm after a DRP. In Table 3.9 Panel B, we
use a. probit model where the dependent variable is one if the key person leaves the
firm (we cannot separate voluntary departure from forced). we find a strong negative
effect of ownership on leaving the. firm even if we include year and key person fixed-
effects. Since owners are more likely to be retained after a DRP violation, this can

explain why investors react more strongly to owner DRPS than non-owner DRPs.

3.5 Conclusion

In this paper, we find evidence that the mandatory disclosures required by the
SEC of RIA firms are useful in predicting future fraud and other criminal behavior by
RIA firm employees. The firms choice of engaging in practices that. produce potential
conflicts of interest is related to an increased probability of future fraud events. Con-
versely, internal monitoring, incentive alignment, and sophisticated external monitors
reduce the probability of fraud. Investors only react strongly to these violations when
the offender is a firm owner. We find that owners are less likely to leave the firm after
being charged with fraud. Together this suggests that firms can mitigate the impact

of fraud events by firing offenders and investors react. accordingly.

We leave to future work extending our analysis to other regulatory violations such
as regulator," and civil actions. While criminal behavior is clearly important. it is
interesting to see if other types of events, such as regulatory or civil actions can be
predicted by firm characteristics. Also, it may be interesting to see if the iriarket

[)E‘ll'tlC’l])fllltS take these actions as seriously as they do criminal ones.

118

 

Table 3. 1

Summary
This table presents characteristics of the 13,579 registered investment adviser firms that filed
Form ADV with the SEC from 2001 to 2006. Panel A contains the size, number of accounts,
age, and employee ownership of the firm. Panel B summarizes the disclosure reporting page
(DRP) filings by number of employees. Panel C summarizes the disclosure reporting page
(DRP) filings by total assets under management.

 

Panel A: Firm Characteristics

 

 

 

 

 

 

 

 

Mean SD 25th IVIedian 75th
Number of firms 13,579
AUM ($Million) 2,841 19,952 38.3 100 453
Accounts 111,669 11,441,819 8 82 292
Average Account Size ($Thousand) 82,212 839,312 324 1,148 20,782
Firm Age (years) 9.05 ' 8.8 2.25 6.15 13.9
Employee Ownership % 66.7% 44.6% 0% 100% 100%

Panel B: Rate by Number of Employees
Employees DRPS Total Firm-Years DRPs per Year
1-5 31 25.087 0.001
6—10 18 9,671 0.002
11-50 41 11,144 0.004
51-250 44 3,498 0.013
251.500 ' 21 583 0.036
500-1000 13 432 0.030
1000+ 641 564 1.137
All 809 50,979 0.016
Panel C: Rate by Discretionary Assets Under lVIanagement

Discretionary AUM DRPS Total Firm-Years DRPs per Year
Less than 1M 68 9,872 0.007
1M-101\'I 1 1.684 0.001
10l\»I-10(');\I 31 17.634 0.002
1001\I—1B 74 14.734 0.005
lB-10B 158 5,510 0.029
100B-1008 328 1.659 0.198
1008+ 149 304 0.490
All 809 51.397 0.016

 

119

Table 3.2
Consistency of RIA practices

This table reports the frequency of certain RIA practices over subsequent firm-years. Interest
in Client Transactions is a binary variable that equals one if the firm recommend securities in
which it has an ownership interest, serves as an underwriter or has any other sales interest.
Soft Dollars is a binary variable that equals one the firm receive research, other products,
or services other than execution from a broker-dealer or a third party in connection with
client securities transactions Custody of Assets is a binary variable that equals one the
firm has custody of clients’ cash and/or securities. Performance—Based compensation is a
binary variable that equals one if the firm provides compensation based on performance.
Broker/ Dealer is a binary variable that equals one if the firm reports an affiliation with a
Broker/ Dealer. Other Affiliation is a binary variable that equals one if the firm reports an
affiliation with an investment company, other investment adviser, bank, insurance company
or other financial company. Small Client Focus is a binary variable that equals one if the
reported percent of individual (non-high net worth) clients exceeds 50%. Separate CCO is
a binary variable that equals one if the person reported on the Schedule A filing has no job
title other than CCO. History of Violations is a binary variable that equals one if the firm
reports “yes” to any question on Item 11 (Disclosure Information).

 

 

Total Same Remove Add
Interest in Client Transactions 31.6% 96.2% 1.4% 2.5%
Soft Dollars 58.9% 96.0% 1.6% 2.4%
Custody of Assets 26.1% 94.4% 1.4% 4.2%
Performance-Based compensation 25.5% 97.7% 1.0% 1.9%
Broker / Dealer 40.5% 96.9% 2.0% 1.7%
Other Affiliation 56.1% 95.9% 1.9% 2.7%
Small Client Focus 23.9% 97.1% 1.5% 2.0%
Separate CCO 16.0% 91.9% 0.8% 7.9%
History of Violations 0.8% 99.7% 0.2% 0.2%

 

120

Table 3.3
DRP Summary

Panel A reports the relative frequency of certain RIA practices depending on whether the firm
has a fraud incidence during the year. The frequency of each practice among all firm-years
is tabulated in the first column. The frequency of each practice among all firm-years that do
not report a DRP is tabulated in the second column. The frequency of each practice among
all firm-years that report at least one initial DRP is tabulated in the third column. The
difference between the second and third columns is reported in the fourth column. Fisher’s
exact test is used to determined statistical significance of the difference. The symbols *, **
and *** denote significance at the 10%, 5% and 1% levels, respectively. Panel B reports the

frequency of DRP filings by year.

 

Panel A: Internal policies versus DRP filings

 

 

 

 

Total NO DRP 2 I DRP Difference.
Interest in Client Transactions 31.0% 30.9% 74.6% 43.7%?”
Soft Dollars 58.2% 58.1% 74.1% 16.0%”*
Custody of Assets 25.7% 25.6% 64.6% 39.0%***
Performance—Based compensation 26.6% 26.5% 29.6% 3.1%
Broker / Dealer 39. 2% 39.0% 85.7% 46.7%" *
Other Affiliation 54.4% 54.3% 92.1% 37.8%..”*
Small Client Focus 22.8% 22.7% 42.9% 20.1%*”
Separate CCO 14.7% 14.6% 31.2% 16.6%”‘**
History of Violations _ 0.8% 0.7% 27.5% 26.8%"”

Panel B: Initial DRPS by year
2002 2003 2004 2005 2006 Total

 

Initial DR.PS

62

98

202

198 809

 

Table 3.4
Probit

This table shows the results of pooled probit regressions where the dependent variable is a
binary variable. that equals one if an initial DRP is filed in a firm-year. # of Employees is
set of dunnnies for each range of employees given in Form ADV Item 5: 1—5, 6-10, 11-50,
51-250, 251-500, 501-1,000 and More than 1,000. For brevity, we include but do not report.
these dummies are included (omitting 1-5). We include a set of year dummies. Standard
errors are clustered by firm in (1)—(5). Random firm effects are included in (6). Constants
are included in the model but not reported for brevity. The symbols *, ** and *** denote
significance at the 10%. 5% and 1% levels, respectively.

 

 

(1) (‘3) (3) (4) (5) (6)

Interest in Transactions 0.159“ 0216*" 0.206" 0.220"
(2.10) (2.64) (2.52) (2.19)

Soft Dollar 0.171“ 0.189“ 0.185“ 0.219"
(2.06) (2.15) (2.17) (2.37)
Custody 0.147" 0.139“ 0.099 0.143
(1.90) (1.68) (1.19) (1.63)

Broker/Dealer 0.280'“ 0.210” 0.203M 0.263"
(3.36) (2.20) (2.10) (2.34)
Other Affiliates 0.228“ 0.146 0.151 0.170
(2.24) (1.35) (1.41) (1.28)
Small Client Focus 0.073 0.070 0.077 0.093
(0.92) (0.81) (0.91) (0.89)

Log(Avg. Account Size) -0.025 -0.044** 41.035” -0.034*
(1.49) (2.53) (2.09) (1.73)

Employee Ownership % -0.426*** -0.310*** —0.291*** -0.316***
(4.66) (2.76) (2.67) (2.71)
Separate CCO 0.056 0.030 0.023 0.015
(0.71) (0.39) (0.32) (0.16)

Performance-Based -0.041 -0.031 -0.043 -0.041
(0.47) (0.31) (0.46) (0.43)

History 0901"“ 0698"”
(6.41) (4.71)

Log(AU;\l) —0.027* 0.012 -0.020 0.000 -0.006 -0.012
(1.00) (0.00) (1.19) (0.02) (0.20) (0.51)
Log(l7irm Age) 0.038 0.025 0.039 0.027 0.024 0.027
(1.26) (0.84) (1.23) (0.80) (0.75) (0.75)

Observations 46.349 46.341 44.376 44.370 44.370 44.370

Pseudo R2 0.24s 0.220 0.252 0.275 0.300

 

122

 

Table 3.5
Probit by Firm Size

This table shows the results of pooled probit regressions where the dependent. variable is a
binary variable that equals one if an initial DRP is filed in a firm-year. The first column.
Small Firms, includes only RIA firms with 50 or less employees. The second column, Large
Firms, includes only RIA firms with greater than 50 employees. Constants are included in
the. model but not reported for brevity. Standard errors are clustered by firm. The symbols

*, ** and ”* denote significance at the 10%, 5"}; and 1% levels. respectively.

 

Small Firms Large Firms

 

Interest in Client Transactions 0.183” 0.274*
(1.96) (1.78)
Soft Dollar 0.075 0.357"
(0.73) (2.42)
Custody 0.116 0.265“
(1.18) (1.99)
Broker/Dealer 0269*” 0.146
(2.65) (0.61)
Other Affiliates 0.11.1
(1.03)
Small Client Focus 0.030 0.090
(0.24) (0.65)
Log(Avg. Account Size) -0.005 -0.105"*
(0.23) (3.62)
Employee Ownership % -0.300** —0.477*
(2.50) (1.76)
Separate CCO 0.016 0.034
(0.14) (0.31)
Performalice-Based 0.085 -0.321”
(0.70) (2.17)
Log(AUl\I) —0.022 0.051
(0.69) (1.59)
Log(I7irm Age) 0.007 0214”“
(0.18) (2.96)
Year Dummies YES YES
Observations 40.320 3.790
Pseudo H2 0.072 0.154

 

_H

Table 3.6
Negative binomial

This table shows the results of negative binomial regressions where the dependent variable
is the number of initial DRP filed in a firm-year. Standard errors are clustered by firm.

* tit

Constants are included in the model but not reported for brevity. The symbols , an(

.1 1*!!!

denote significance at the 10%, 5% and 1% levels, respectively.

 

 

(1) <2) <3) (4) (5)

Interest in Client Transactions 0.375 0.680” 0.677”
(1.41) (2.41) (2.40)

Soft Dollar 0.135 0.422 0489*
(0.46) (1.42) (1.71)
Custody 0.181 0.240 0.099
(0.67) (0.82) (0.35)
Broker / Dealer 0646* 0.4.81 0.450
(1.82) (1.19) (1.06)

Other Affiliates 1205*" 0686* 0.711*
(2.84) (1.71) (1.83)
Small Client Focus 0.221 0.314 0.244
(0.81) (1.20) (0.99)

Log(Avg. Account Size) —0.020 -0.099* -0.090*
(0.37) (1.67) (1.68)

Employee Ownership % -1.575*** -1.202*** 1170*"
(5.04) (3.05) (3.09)

Separate CCO 0.111 -0.026 -0.032
(0.51) (0.13) (0.17)
Performance—Based -0.063 0.006 0.065
(0.23) (0.018) (0.20)

History 2542*“
(7.35)
Log(AUM) 0.001 0.053 -0.030 0.037 0.013
(0.020) (0.85) (0.56) (0.60) (0.23)
Log(Firrn Age) 0.097 0.062 0.138 0.106 0.103
(0.93) (0.59) (1.33) (0.98) (0.98)
# of Employee Dummies YES YES YES YES YES
Year Dummies YES YES YES YES YES
Observations 46.349 46.341 44.376 44.370 44.370

 

124

Table 3.7
Owner Probit

This table shows the results of pooled probit regressions where the dependent variable is a
binary variable that equals one when an initial DRP is filed for a firm owner in a firm-year.
# of Employees is set of dummies for each range of employees given in Item 5: 1-5, 6-10,
11—50, 51—250, 251-500, 501-1,000 and More than 1,000. For brevity, we include but do not
report these dummies. We include a set of year dummies. Standard errors are clustered by
firm. Constants are included in the model but not reported for brevity. The symbols *, **
and *** denote significance at the. 10%, 5% and 1% levels, respectively.

 

 

(1) (2) (3) (4) (5)
Interest in Client Transactions 0.015 -0.187 -0.211
(0.097) (1.05) (1.13)
Soft Dollar 0.108 0.091 0.100
(0.64) (0.56) (0.64)
Custody -0.036 —0.037 —0.054
(0.21) (0.25) (0.35)
Broker / Dealer 0.123 0.069 0.066
(0.75) (0.37) (0.34)
Other Affiliates -0.085 -0.153 —0.154
(0.52) (0.82) (0.80)
Small Client. Focus 0.068 0.087 0.096
(0.41) (0.50) (0.56)
Log(Avg. Account Size) 0.024 -0.008 -0.005
(1.16) (0.36) (0.24)
Employee Ownership % -0.282 -0.371* -0.341*
(1.42) (1.74) (1.74)
Separate CCO -0.019 0.007 0.006
(0.10) (0.044) (0.032)
Performaiice-Based 0.222 0.311“ 0325*
(1.55) (1.75) (1.86)
History 0.826M
(2.38)
Log(AUM) -0.072*** -0.088*** —0.076*’” -0.077*** -0.083***
(3.05) (3.57) (3.67) (4.17) (4.66)
Log(Firm Age) 0.010 0.020 0.016 0.015 0.010
(0.16) (0.32) (0.23) (0.23) (0.15)
Observations 46.349 46.341 44,376 44.370 44.370
Pseudo 17") 0.0177 0.0157 0.0390 0.0498 0.0635

 

Ca"!

The dependent variable is Florel-‘t =

Table 3.8

Flows
Assets”—-Asscts,j‘t_1*Rcturnsm 1

 

. The sam-
Assetsi‘t_1

ple is RIA firms with available 13F data. # of Employees is a set of dummies for each
range of employees given in Item 5: 1-5, 6—10, 11-50, 51-250, 251-500, 501-1,000 and
More than 1,000. Style is a set of dummies for large cap growth, large cap value, small
cap value, and small cap growth. For brevity, we simply report YES if these sets of
dummies are included. Standard errors are clustered by firm and year. The symbols

***

, and *** denote significance at the 10%, 5% and 1% levels. respmtively.

 

 

(1) (‘2) (3) (4) (5) (6)

Total DRP -0.0001 0.001 0.001 0.0003
(0.042) (0.64) (0.65) (0.23)

Owner DRP -0.295*** -0.437"* —0.438*“‘ —0.407***
(3.95) (5.56) (5.61) (5.32)
Prior Return 0.252 0.253 0.252 0.253
(1.20) (1.20) (1.20) (1.20)

Log(Portfolio Value) -0.010* -0.010* -0.010‘ —0.018”
(1.66) (1.66) (1.07) (2.53)

Constant 0.050 0.050 0.281“M 0281*” 0282"” 0.392‘"
(0.67) (0.67) (2.64) (2.63) (2.65) (3.26)
#— of Employee Durn— NO NO NO NO NO YES

nnes

Style Dummies NO NO YES YES YES YES
Observations 7,199 7,199 7,199 7,199 7,199 7,190
R2 0.0000 0.0001 0.048 0.048 0.048 0.053

 

Table 3.9

Key Person Turnover

This table shows the likelihood of key persons leaving the firm after a DRP is filed. Panel
A shows unconditional averages. Panel B shows the results of probit regressions where the
dependent variable is one if the key person leaves the firm and zero otherwise. Owner is
a binary variable that equals one if the key person has an ownership stake in the firm.
log(Tenure) is the logarithm of the number of days since the person assumed their current
title with the firm. In column three, we estimate the model with a job (employee-firm pair)
fixed effect. The symbols *, ** and *** denote significance at the 10%. 5% and 1% levels,

 

 

 

 

 

 

 

 

 

respectively.
Panel A: Probability of leaving the firm
Non-owner Owner Difference
N0 DRP 11.91% 2.77% 9.14%”*
DRP 17.86% 3.39% 14.47%***
Difference 5.95% 0.62%
Panel B: Regression Analysis
(1) (2) (3)
Owner -0.900M -1.034** 4.066”
(2.06) (2.38) (2.21)
log(Tenure) «0.020 0.003 0.007
(0.13) (0.020) (0.036)
Constant ~0.907"* —0.963*** —0.971***
(2.97) (3.18) (2.93)
Year NO YES YES
Key Person Fixed Effect NO NO YES
Observations 87 87 87
Pseudo R3 0.1020 0.1231

 

127

BIBLIOGRAPHY

128

 

Admati, Anat R., and Paul Pfleiderer, 2008, The ”\Vall Street \Valk” and Shareholder
Activism: Exit as a Form of Voice, Review of Financial Studies forthcoming.

Agrawal, Anup, Jeffrey Jaffe, and Jonathan Karpoff, 1999, Management turnover and
governance changes following the revelation of fraud. J oarnal of Law and Economics
42, 309-342.

Almazan, Andres, Keith C. Brown, Murray Carlson, and David A. Chapman. 2004,
Why Constrain Your Mutual Fund l\AIanager?. Journal of Financial Economics 73,
289-321.

Amihud, Yakov, 2002. Illiquidity and stock returns: Cross-Section and time efffects,
Journal of Financial Markets 5, 31—56.

Bergstresser, Daniel B., and Thomas Philippon, 2006, CEO Incentives and Earnings
Managen'ient, Journal of Financial Economics 80, 511—529.

Berk, Jonathan B., and Richard C. Green, 2004, I\‘Iutual Fund Flows and Performance
in Rational Markets, Journal of Political Economy 112, 1269—1295.

Berzins, Janis, and Charles A. Trzcinka, 2005. Performance and Survival of Institu-
tional l\«'loney Managers, W’orking Paper, Indiana University.

Bethel, Jennifer, Julia Porter Liebeskind, and Tim Opler, 1998, Block Share Purchases
and Corporate Performance, The Journal of Finance 53, 605—634.

Bhide, Amar, 1993, The Hidden Costs of Stock Market. Liquidity, Journal of Financial
Economics 34, 3151.

Black, Bernard S., 1990, Shareholder Pa.ssi\-’ity Reexamined, llIic/zigan Law Review
89,811~892.

Bogle, John C., 2009. The End of ”Soft Dollars" 1’. Financial Analysts Journal 65,
48 53.

Bollen, Nicolas PB, and V ironika K. Pool, 2008, Conditional Return Smoothing in
the Hedge Fund Industry, Journal of Financial and Quantitative Analysis 43, 267—
298.

Bolton. Patrick. and Ernst-Ludwig von Thadden. 1998, Blocks. Liquidity and Corpo-
rate Control, Joarnal of Finance 53, 1-25.

Borokhovich, Kenneth A., Kelly R. Brunarski. and Robert Parrino. 2006. Variation in

the monitoring incentives of outside blockholders. Journal of La w and Economics
49. 651-680.

129

Brav, Alon, Wei Jiang, Frank Partnoy, and Randall Thomas, 2008, Hedge Fund Ac—
tivism, Corporate Governance, and Firm Performance, Journal of Finance 63, 1729—
1776.

Brown. Stephen J ., William N. Goetzmann, Bing Liang, and Christopher Schwarz,
2008, Mandatory Disclosure and Operational Risk: Evidence from Hedge Fund
Registration, Journal of Finance 63, 2785—2815.

Brown, Stephen J ., \Nilliam N. Goetzmann, Bing Liang, and Christopher Schwarz.
2009, Estimating Operational Risk for Hedge Funds: The w-score, Financial Ana-
lysts Journal 65, 43—53.

Bushee, Brian, 1998, The Influence of Institutional Investors on Myopic R&D Invest-
ment Behavior, Accounting Review 73, 305—333.

Busse, Jeffrey A., Amit Goya], and Sunil VVahal, 2007, Performance Persistence in
Institutional Investment l\lanagement, Working Paper, Emory University.

Cadman, Brian D., Sandy Klasa. and Steven R. i\'latsunaga. 2007, Evidence on
How CEO Compensation Functions Differ between Execucomp Firms and Non-
Execucomp Firms, Working Paper.

Carhart, Mark M., 1997, On Persistence in Mutual Fund Performance, Journal of
Finance 52, 57—82.

Chakravarty, Sugato, Robert W'ood. and Robert van Hess, 2004, Decimals and Liq-
uidity: A Study of the NYSE, Journal of Financial Research 27, 75—94.

Chang. Howard H., and David S. Evans, 2007, The Optimal Prosecution of C(iirporate
Fraud: An Error Cost Analysis, \Vorking Paper.

Chen. Joseph, Harrison Hong, Ming Huang, and Jeffrey D. Kubik, 2004, Does Fund
Size Erode Mutual Fund Performance? The Role of Liquidity and Organization,
American Economic Review 94, 1276*1302.

Chen. Qi. Itay Goldstein, and Wei Jiang, 2008. Directors Ownership in the U.S.
Mutual Fund Industry, Journal of Finance 63, 2629* 2677.

Clwrnobai, Anna, Philippe Jorion, and Fan Yu, 2009, The D(_>t.er1ninants of Opera.-
tional Risk in U.S. Financial Institutions, Working Paper.

Chevalier. Judith. and Glenn Ellison. 1999, Career Concerns of Mutual Fund Man-
agers. Quarterly Journal of Economics 114, 389 432.

Choi. Stephen J., and Marcel Kahan. 2007, The Market Penalty for Mutual Fund
Scandals, Boston University Law Review 87. 1021 1057.

130

Christopherson, Jon A., Wayne E. Person, and Debra A. Glassman, 1998, Conditioning
Manager Alphas on Economic Information: Another Look at the Persistence of
Performance, Review of Financial Studies 11, 111—142.

Cici, Gjergji, Scott Gibson, and Rabih l\t’Ioussawi, 2006, For Better or W'orse? Mu-
tual Funds in Side-by-Side Management Relationships with Hedge Funds, Working
Paper.

Clifford, Chris, 2008, Value Creation or Destruction? Hedge Funds as Shareholder
Activist, Journal of Corporate Finance 14, 3231336.

Coggin, T. Daniel, Frank J. Fabozzi, and Shafiqur Ralnnan, 1993, The Investment
Performance of U.S. Equity Pension Fund Managers: An Empirical Investigation,
Journal of Finance 48, 1039-1055.

Conference Board, 2007, Institutional Investment Report: Patterns of Institutional
Investment and Control in the United States, New York, NY.

Cremers, Martijn, Joost Driessen, Pascal l\*:Iaenhout, and David Weinbaum, 2006, Does
Skin in the Game Matter? Director Incentives and Governance in the Mutual Fund
Industry, Journal of Financial and Quantitative Analysis , Forthcoming.

Cronqvist, Henrik, and Rudiger Fahlenbrach, 2008, Large Shareholders and Corporate
Policies, Review of Financial Studies forthcoming.

Davis, Gerald E, and E. Han Kim, 2006, Business ties and proxy voting by mutual
funds, Journal of Financial Economics 85, 552—570.

Del Guercio, Diane, Larry Y. Dann, and M. Megan Partch, 2003, Governance and
Boards of Directors in Closed—End Im'estment Companies, Journal of Financial
Economics 69, 111~152.

Del Guercio, Diane, and Paula A. Tkac, 2002. The Determinants of the Flow of Funds
of Managed Portfolios: Mutual Funds vs. Pension Funds, Journal of Financial and
Quantitative Analysis 37, 523—557.

Deli, Daniel N., 2002, Mutual Fund Advisory Contracts: An Empirical Investigation,
Journal of Finance 57, 109~133.

Demsetz. Harold. 1983, The Structure of Ownership and the 'I‘hem'y of the Firm,
Journal of Law and Economics 26, 375 390.

Demsetz. Harold. and Kenneth Lehn. 1985a. The Structure of Cbrporate Ownership:
Causes and Consequences, Journal of Political Economy 93, 115-5 77.

Deinsetz. Harold. and Kenneth Lehn. 1985b. The Structure of Corporate Ownership:
Causes and Consequences, Journal of Political Economy 93. 1155771177.

131

Dimmock, Stephen G., William C. Gerken, and Jennifer Marietta—Westberg, 2009,
Employee Ownership of Institutional Investment Management Firms, Michigan
State University Working Paper.

Dlugosz, Jennifer, Riidiger Fahlenbrach, Paul Gompers, and Andrew Metrick, 2006,
Large blocks of stock: Prevalence, Size, and lV‘Ieasurement, Journal of Corporate
Finance 12, 594—618.

Edmans, Alex, 2008, Blockholder Trading. Market Efficiency and Managerial Myopia,
Journal of Finance forthcoming.

Faccio, Mara, and Larry Lang, 2002, The ultimate ownership of Western European
corporations, Journal of Financial Economics 65, 365—395.

Faleye, Olubunmi, Vikas Mehrotra, and Randall Merck, 2006, When Labor has a
Voice in Corporate Governance, Journal of Financial and Quantitative Analysis 41,
489—510.

Fania, Eugene F., 1980, Agency Problems and the Theory of the Firm, Journal of
Political Economy 88, 288—307.

Fania. Eugene F., and Michael C. Jensen, 1983a, Separation of Ownership and Control,
Journal of Law and Economics 26, 301—325.

Fania, Eugene F., and Michael C. Jensen, 1983b, Agency Problems and Residual
Control, Journal of Law and Economics 26, 327—349.

Farber, David B., 2005, Restoring Trust After Fraud: Does Corporate Governance
Matter?, The Accounting Review 80, 539— 561.

Farnsworth, Heber, and Jonathan Taylor, 2006, Evidence on the Compensation of
Portfolio Managers, Journal of Financial Research 29, 305—324.

Fee, C. Edward, Charles Hadlock, and Shawn Thomas, 2006, Corporate Equity Own-
ership and the Governance of Product l\=Iarket Relationships, Journal of Finance
61, 1217—1252.

Fich, Eliezer M ., and Anil Shivdasani, 2007, Financial fraud, director reputation, and
shareholder wealth, Journal of Financial Economics 86, 306—336.

Furfine, Craig, 2003. Decimalization and market. liquidity. Econmnic Perspectives 27,
1—12.

GCI‘V‘dlS, Simon. Anthony W'. Lynch. and David K. Musto, 2005. Fund Families as
Delegated Monitors of Money Managers. Review of Financial Stiulics 18. 1139—
1169.

Goctzmann, William, Zoran Ivkovic, and K. Geert Rmnvenhorst, 2001, Day Trading
International Mutual Funds: Evidence and Policy Solutions, Journal of Financial
and Quantitative Analysis 36, 287—309.

Gompers, Paul A., Joy L. Ishii. and Andrew Metrick, 2003, Corporate Governance
and Equity Prices, Quarterly Journal of Economics 118, 107 155.

Goyenko, Ruslan, Craig W. Holden, and Charles Trzcinka. 2009, Do l\leas1.1res of
Liquidity Measure Liquidityl, Journal of Financial Economics 92, 153—181.

Granger, Clive, and M. Hashem Pesaran. 2000, Economic and Statistical measures of
Forecast Accuracy, Journal of Forecasting 19, 537—560.

Hasbrouck, Joel, 2009, Trading costs and returns for US equities: Estimating Effective
Costs from Daily Data, Journal of Finance forthcoming.

Heflin. Frank, and Kenneth W'. Shaw, 2000, Blockholder Ownership and Market Liq-
uidity, Journal of Financial and Quantitative Analysis 35, 621-7 633.

Himmelberg, Charles P., R. Glenn Hubbard, and Darius Palia, 1999, Understanding
the Determinants of Managerial Ownership and the Link between Ownership and
Performance, Journal of Financial Economics 53, 353—384.

Hirschman, Albert, 1970, Exit. l‘bice and Loyalty: Responses to Decline in F irrns,
Organizations and States. (Harvard University Press).

Holmstrom, Bengt, 1982, l\lanagerial Incentive Problems - A Dynamic Perspective,
in Essays in Economics and Management in Honor of Lars ll-‘alzlbeck pp. 210— 235.
Swedish School of Economics. Helsinki.

Horan, Stephen M., and D. Bruce Johnsen, 2008, Can third-party payments benefit
the prinicipal? The case of soft-dollar brokerage, International Review of Law and
Economics 28, 56—77.

Hu, Henry TC, and Bernard Black, 2006, Empty voting and hidden (morphalfle)
ownership: Taxonomy, implications and reforms, Business Lawyer 61, 1011—1070.

James. Christopher, and Jason I’x'arceski, 2006, Investor Monitoring and Differences
in Mutual Fund Performance, Journal of Banking and Finance 30, 2787—2808.

Jensen. Michael C., 1989, Eclipse of the. Public Corporation. Harvard Business Review
67. 617-75.

Jensen. Michael C., and \Villiain H. Meckling. 1976. Theory of the firm: Managerial
behavior. agency costs. and mvnership structure, Journal of Firmncial Economics

3, 305 —- 360.

133

 

.
g .
17
I z.
*.
.. .
‘l
-
;’,.
‘7.
..r
..
-- .
.7.
7.?
’ ‘
M.
a n
.Y
. .
-
..
'u.‘
‘I-
:.N-
-.-q".
”.9
e:
.43‘
12
..
, -

"-"v‘l"l-
I u,‘ . u.
'.. '5

.
'-
...4 I.

3

I
t

'. ‘1':
‘.J~5.

.«f

 

Kalm, Charles, and Andrew Winston, 1998, Ownership Structure. Speculation and
Shareholder Intervention, Journal of Finance 53, 99—129.

Kang, Jun-Koo, and Jin-Mo Kim, 2006, Foreign Blockholders, Shareholder Rights and
Corporate Governance: Evidence from Partial Acquistions of U.S. Firms by Foreign
Firms, V’Vorking Paper.

Kang, Jun—Koo, and Anil Shivdasani, 1995, Firm performance, corporate governance
and top executive turnover in Japan, Journal of Financial Economics 38, 29—58.

Karpoff. Jonathan M., D. Scott Lee, and Gerald S. l\*Iartin, 2008, The Cost to Firms
of Cooking the Books, Journal of Financial and Quantitative Analysis 43, 581—612.

Karpoff, Jonathan M., and John R. Lott, Jr., 1993, The Reputational Penalty Firms
Bear from Committing Criminal Fraud, Journal of Law and Economics 36, 757—802.

Kempf, Alexander, and Stefan Ruenzi, 2008, Tournaments in l\*Iutual—Fund Families,
Review of Financial Studies 21, 1013—1036.

Khein. April, and Emanuel Zur, 2009, Entrepreneurial Shareholder Activism: Hedge
Funds and Other Private Investors, Journal of Finance 64, 187—229.

Khorana, Ajay, 1996, Top l\-lanagement Turnover: An Empirical Investigation of Mn—
tual Fund Managers, Journal of Financial Economics 40, 403—427.

Khorana, Ajay, 2001, Performance Changes following Top l\=’lanagement Turnover:
Evidence from Open-End Mutual Funds, Journal of Financial and Quantitative
Analysis 36, 371—393.

Khorana. Ajay, Henri Servaes, and Lei “Ridge, 2007, Portfolio Manager Ownership
and Fund Performance, Journal of Financial Economics 85, 179—204.

Klu'n'ana. Ajay, Peter Tufano, and Lei W'edge, 2007, Board Structure, Mergers and
Shareholder Wealth: A Study of the ;\-alutual Fund Industry, ..lournal of Financial
Economics 85, 571—598.

Kim. E. Han. and Paige Parker Ouimet, 2008, Employee Capitalism or Corporate
Socialism? Broad—Based Employee. Stock Ownership, Working Paper.

Korajczyk. Robert A.. and Ronnie Sadka, 2008, Pricing the cmnmonality across alter-
native measures of liquidity, Journal of Financial Empirical 87, 45 - 72.

Kyle. Albert. and Jean-Luc Vila. 1991. Noise Trading and Takeovers. RAND Journal
of Eco'nornic's 22, 54- 71.

Lakonishok, Josef. Andrei SlllOlll‘l'. and Robert Vt". Vislmy. 1992. The Structure and
Perforim'mce of the Money Management Industry. Brool'ings Paper: illicr'oeconornics
pp. 339 391.

134

Lesmond, David A., Joseph P. Ogden, and Charles Trzcinka, 1999, A New Estimate
of Transaction Costs, Review of Financial Studies 12, 1113—1141.

Liang, Bing, 2003, The accuracy of hedge fund returns. Journal of Portfolio Manage-
ment 29, 111—122.

Maug, Ernst, 1998, Large Shareholders as l\s’lonitors: Is There a. Trade-Off between
Liquidity and Control? , Journal of Finance 53, 65—98.

McConnell, John J ., and Henri Servaes, 1990, Additional Evidence on Equity Owner—
ship and Corporate Value, Journal of Financial Economics 27, 595—612.

Merton, Robert C., 1987, A Simple Model of Capital l\r‘larket Equilibrium with Incom-
plete Information, Journal of Finance 42, 483—510.

Meschke, J. Felix, 2006, An Empirical Examination of Mutual Fund Boards, Working
Paper, University of Minnesota.

Mikkelson, Wayne, and Richard S. Ruback, 1985, An Empirical Analysis of the Inter-
firm Equity Investment Process, Journal of Financial Economics 14, 523—553.

Morck, Randall, Andrei Shleifer, and Robert V ishny, 1988, Management Ownership
and Market Valuation: An Empirical Analysis, Journal of Financial Economics
20, 293—315.

Morrison, Alan, and William J. \Vilhelm, 2004, Partnership Firms, Reputation, and
Human Capital, American Economic Review 94, 1682—1692.

Niehaus, Greg, and Greg Roth, 1999, Insider Trading, Equity Issues, and CEO

Turnover in Firms Subject to Securities Class Action, Financial Arianagerncnt 28,
['5
52 72.

Nohel, Tom, Jay \Vang, and Lu Zheng, 2006, Side-By-Side Management of Hedge
Funds and Mutual Funds, Working Paper.

Petersen, Mitchell, 2009, Estimating Standard Errors in Finance Panel Data Sets:
Comparing Approaches, Review of Financial Studies 22, 432 e480.

l’ollet. Joshua M., and Mungo Wilson, 2008. How Does Size Affect Mutual Fund
Behavior? , Journal of F inancc , Fortl‘icoming.

Pomorski, Lukasz. 2008, Acting on the Most Valuable Informaticm: “Best Idea" Trades
of Mutual Fund l\lanagers. \Vorking Paper, University of Toronto.

Shleifer, Andrei, and Robert “7 Vishny, 1986., Large Shareholders and Corporate
Control. Journal of Political Economy 94, 461 88.

1 35

Sirri, Erik R., and Peter Tufano, 1998, Costly Search and Mutual Fund Flows, Journal
of Finance 53, 1589—1622.

Smith. Michael P., 1996. Shareholder Activism by Institutional Investors: Evidence
from CalPERS, The Journal of Finance 51, 227252.

Srinivasan, Suraj, 2005, Consequences of Financial Reporting Failure for Outside Di-
rectors: Evidence from Accounting Restatements and Audit Committee Members,
Journal of Accounting Research 43, 291—334.

Tufano, Peter, and Matthew Sevick, 1997, Board Structure and Fee-Setting in the
U.S. Mutual Fund Industry, Journal of Financial Economics 46, 321—355.

\Yellrnan, Jay W ., and Todd Houge, 2005, Fallout from the l\Iutual Fund Trading
Scandal, Journal of Business Ethics 62, 129-139.

136

M'lllllllllllllllflllllllllllllES