mi

This is to certify that the
dissertation entitled

INSTITUTIONAL INVESTOR PREFERENCES AND FIRM
VALUE

presented by

 

Gwinyai T. Utete

LIBRARY
Michigan State
University

has been accepted towards fulfillment
of the requirements for the

 

 

 

PhD. degree in Finance

 

 

“

.4? km [Ml

 

 

Major ProfessorUSignature
071/01 /Jo~7

Date

MSU Is an affirmative-action, equal-opportunity employer

 

 

l

l .

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 p:/CIRC/DateDue.indd-p.1

 

 

 

 

INSTITUTIONAL INVESTOR PREFERENCES AND FIRM VALUE
By

Gwinyai T. Utete

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Finance

2007

ABSTRACT
INSTITUTIONAL INVESTOR PREFERENCES AND FIRM VALUE
By
Gwinyai T. Utete

This dissertation demonstrates the existence of different institutional investor
preferences for equity characteristics, and makes the link between these preferences and
firm value. We show that transient institutional investors (those that trade frequently with
a view to maximizing short term gains) possess superior information to other market
participants and actively seek out Situations in which they can exploit this informational
advantage. Their presence, particularly under conditions where firm-level information
quality is poor, is associated with both higher returns and higher subsequent firm values.

We also find that dedicated (long term) and quasi-indexing (passive with broad
holdings) institutional investors are attracted to firms that enable them to engage in
monitoring activities. Although there is some evidence to suggest that the arrival of
dedicated institutional investors enhances firm transparency, the presence of both
dedicated and quasi-indexer investors is of limited importance in determining overall
firm value when other features of the firm’s contracting environment (such as corporate
governance provisions, information precision and free cash flows) are fully considered.
These results are robust to both a fixed effects and an instrumental variables procedure
that mitigates the endogeneity problem. Therefore, the previously documented linear
relationship between institutional holdings and Tobin’s Q is almost entirely driven by the
trading actions of active investors who do not engage in monitoring in the conventional

SCI‘ISC.

ACKNOWLEDGMENTS

I would like to express my gratitude to the members of my dissertation
committee, Dr. Jun—Koo Kang, Dr. Edward Fee, Dr. Stephen Dimmock and Dr. Wei-Lin
Liu, for their insightful comments and guidance through this process. I would also like to
extend my thanks to the Department Chair, Dr. Geoffrey Booth, the Director of the PhD.
program, Dr. Charles Hadlock and Dr. Kirt Butler for all of their advice and assistance.
Finally, I would like to thank my colleagues for their contributions during our time

together in the Ph.D. program.

iii

TABLE OF CONTENTS

INSTITUTIONAL INVESTOR PREFERENCES AND FIRM VALUE

LIST OF TABLES ......................................................................................................... v
1 .1 Introduction ........................................................................................................ 1
1.2 Corporate Ownership and Firm Value ............ . ................................................... 2
1.3 Institutional Investor Types and Their Characteristics ...................................... 5
1.4 The Endogeneity of Ownership ......................................................................... 8
2.1 Hypothesis Development ................................................................................... 9
2.1.1 Institutional Demand for Equity Characteristics ................................................ 9
2.1.2 Herding Behavior of Institutional Investors and Informed Trade ................... 18
2.1.3 The Link Between Ownership Structure and Returns ..................................... 23
2.1.4 Conditional Interpretations of the Ownership - Firm Value Relationship ....... 26
3.1 Data and Descriptive Statistics ........................................................................ 29
4.1 Results .............................................................................................................. 35
4.1.1 Testing the Institutional Equity Preferences Hypothesis ................................. 35
4.1.2 Testing the Herding and Informed Trade Hypothesis ...................................... 38
4.1.3 Testing the Ownership Structure — Firm Value Hypothesis ............................ 43
4.1.4 Conditional Firm Value Tests .......................................................................... 46
5.1 Summary and Conclusions .............................................................................. 47

APPENDIX. TABULATED RESULTS ......................................................... 50

BIBLIOGRAPHY ............................................................................................ 91

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

Table 10

Table 1 1

Table 12

Table 13

Table 14

Table 15

Table 16

Table 17

Table 18

Table 19

Table 20

LIST OF TABLES

Summary Statistics ............................................................................... 51
Investment Objective and Investment Style ........................................ 55
Correlation Coefficients ....................................................................... 57
Aggregate Institutional Investor Preferences ....................................... 60
Corporate Insider Preferences .............................................................. 62
Equity Demand Functions: Transient Institutional Investors .............. 63

Equity Demand Functions: Quasi-Indexer Institutional Investors ....... 64

Equity Demand Functions: Dedicated Institutional Investors ............. 65
Changes in Institutional Ownership and Firm Transparency .............. 66
Aggregate Changes in Institutional Ownership ................................... 67
Changes in Insider Ownership ............................................................. 69
Portfolio Rebalancing: Transient Institutional Investors ..................... 71
Portfolio Rebalancing: Quasi-Indexer Institutional Investors ............. 73
Portfolio Rebalancing: Dedicated Institutional Investors .................... 75
Performance Attribution: Institutional Investors and Insiders ............. 77
Performance Attribution: Institutional Investors by Class ................... 79

Performance Attribution: Transient Investors: Growth vs. Value ....... 81
Ownership Structure and Finn Value .................................................. 84
Robustness Tests .................................................................................. 85

Conditional Specifications of the Firm Value - Ownership Relation ..87

1.1 Introduction

This dissertation examines the link between ownership structure and firm value
by focusing on the role of differences in institutional investment philosophy. In a
standard rational expectations framework, investors are only concerned with the expected
payoffs of the assets that they hold. Assuming a common information set, market clearing
prices can be derived from a given agent’s expectation of those payoffs and the filll
covariance matrix of payoffs with respect to all other assets in the economy. In reality
however, investors display considerable heterogeneity in terms of both the information
they possess and their preferences for certain characteristics in equities. The key
contribution of this study lies in how we use a more refined demarcation of investor types
to reveal how institutional ownership structure relates to firm value. We contend that in
the presence of information asymmetry, differing demand specifications are associated
with real effects on asset prices, and we test this conjecture in a multiple regression
framework.

Using prior empirical research as a guide, we develop three categories of stock
characteristics that are concordant with investor demand. Briefly, these are the risk-retum
relationship, the relative opacity of the firm’s operations to outside investors and features
that enable monitoring. We find that institutional investors differ quite dramatically in
their preferences for these characteristics, primarily because of differences in both their
investment horizons and ability to collect finn Specific information. Short term, high
t'umover (transient) institutional investors are found to be attracted to firms that perform
well (in both an accounting and financial sense) but have a high level of informational

uncertainty. Institutional investors with a longer term focus (quasi-indexer and dedicated)

are found to be attracted to firms with features that facilitate oversight of managerial
behavior. However, we find that only the presence of the transient class of institutional
investor is consistently associated with higher firm values.

The rest of the paper is organized as follows: in the remainder of this section, we
motivate the ownership-performance problem through a discussion of key contributions
in the area. In section 2, we develop our hypotheses and design appropriate empirical
tests. In section 3, we present our data sources and conditions for inclusion in the final
sample of firms. Section 4 provides results and section 5 concludes with an analysis of

the results and possibilities for future research.

1.2 Corporate Ownership and Firm Value

The relationship between corporate ownership and firm performance is one that
has attractedsubstantial interest since the pioneering work of Berle and Means (1933).
According to Berle and Means (1933), as a firm’s ownership structure becomes more
diffuse, shareholders’ ability to control management diminishes. The resulting Shift in
power allows managers to act in their own self interest potentially destroying Shareholder
value in the process. This rather straightforward argument is a variant of the classic
principal-agent problem. Berle and Means’ (1933) proposition went essentially
unchallenged and unmodified in the financial economics literature for over four decades.
Jensen and Meckling (1976) represented the first significant reexamination of the topic.
Jensen and Meckling (1976) propose equity ownership by management as a solution to
the agency problem. They argue that as managers’ shareholdings in their own companies

increase, their incentives become more aligned with those of outside shareholders. As a

result, they become apt to take actions that increase firm performance and consequently,
firm value. While this argument is intuitively appealing, it presents some uncomfortable
equilibrium implications. If firms can increase value by awarding the CEO more stock,
why do we observe a multiplicity of firm ownership structures empirically?

Demsetz and Lehn (1985) formalized this question by examining the
shareholdings and financial performance of 511 large US. firms over the period 1976-
1980. Their OLS regressions of return on equity on a variety of ownership measures fail
to detect any significant relationship between the two. They interpret this evidence as
indicative of firms always being at their optimum ownership structure. Much like the
Jensen and Meckling (1976) supposition, this argument, while intuitively attractive, also
suffers from a critical flaw in equilibrium. If firms naturally tend towards their optimum
shareholder structure then those that do not will eventually fail. However, in the long run,
one would expect to see all observed shareholding structures perform equally well.
Clearly this does not conform to empirical reality either]. Thus we have a conflict which
persists in the literature to this day.

Previous studies that examine the link between ownership structure and
performance tend to make a direct examination of how a particular type of ownership
relates to firm value. There is a common structure to the research question that is asked;
how does management ownership relate to firm value? Or, how does the ownership of
large block holders relate to firm value? These studies typically make fairly broad
assumptions about the motivations of potentially disparate groups of Shareholders. We
instead take the View that insiders and other shareholders are unique actors who differ in

their utility specifications, initial wealth, and ability to access markets.

 

' La Porta, Lopez—de-Silanes, Shleifer and Vishny (1998).

Amongst outside shareholders, there exist differing cost structures with respect to
firm specific information acquisition. Diamond’s (1984) model of financial
interrnediation is instructive. To wit: Given a firm with m outside investors and a
manager in control of a project whose terminal outcome is the random variable y, the

optimal contract specifies:

max¢(,) E5.[maxz€[0,)~.] j" — Z '— ¢(Z)]

(1)
subject to
z e arg maxzemfl y — z - ¢(z)
(2)
and
Ej.[arg maxzemfl y — z - ¢(z)] Z R (3)

Where y is the true realization of y, z 2 0 is the payment the manager makes to the
outside investors and R is the competitive rate Iof interest in the economy. <D(z)
corresponds to the non pecuniary penalty function that the firm’s security holders can
impose on the manager, given a reported value of 2. Under these conditions, there exist
incentives for intermediaries to develop specialized abilities at monitoring managerial
behavior. More formally, delegation of monitoring is optimal provided that the expected
deadweight penalty when there is no monitoring and the total cost of monitoring by all
investors both exceed the cost of having a specialized monitor:

K + D S m111[E'§[¢*(.T‘)],(m - IO] (4)

Where K is the direct cost of monitoring and D is the cost of delegation.

Casting institutional investors as the intermediary between firms and individuals,
it follows that the array of information a given institution collects will be the outcome of
a competitive process. The inputs to which are the demands of its clientele and the innate
abilities and comparative advantages it may possess with regards to information
acquisition. Therefore, in a competitive equilibrium, there can exist multiple classes of
intermediaries pursuing different strategies. By acknowledging these differences, the
predictions concerning the relevance of ownership structure with respect to firm value

change dramatically, and this is what we demonstrate empirically.

1.3 Institutional Investor Types and Their Characteristics

While there is no unambiguous theoretical relationship between shares demanded
and observable firm characteristics, prior empirical research has identified certain. .
patterns. For example, with regards to institutional investors, Gompers and Metrick
(2001) find that firm size has a concave relationship with institutional holdings.
Furthermore, they find that institutions express a preference for firms with high share
turnover and prices per share. Gompers and Metrick (2001) argue that prudence
considerations cause institutional investors to avoid small illiquid stocks. In terms of
historical return patterns, Gompers and Metrick (2001) find that the prior 11 month
returns have a negative association with future institutional holdings. They also find that
institutions exhibit a weak preference for high book to market stocks. As such there is
some evidence of institutional investors as a group preferring value over growth stocks.

Falkenstein (1996) and Del Guercio (1996) provide confirmatory evidence of Gompers

and Metrick’s (2001) findings. Additionally, Wahal and McConnell (1998) demonstrate
that institutional ownership is positively related to research and development intensity.

All of these studies use the standard classification scheme provided by the
Thomson Financial CDA/Spectrum database. The database categorizes firms’
institutional equity holdings into shares held by banks, insurance companies, investment
advisers and pensions and endowments. Bushee (1998, 2001), however, notes that these
categories mask tremendous heterogeneity in both investor behavior and investment
objective. The problem is further compounded by errors in the database itself. Indeed the
accompanying manual notes that “TYPECODE [has] serious classification errors in
recent years, such that the Other group is unrealistically large. Many Banks and
Independent Investment Advisors are improperly classified in the Other group in 1998
and beyond. For example, in the first quarter of 1999, the number of independent
investment advisors drops from over 1200 to about 200, while the Other group jumps
from roughly 100 to over 1300. TFN...[regrets] that [this] occurred, but they have no
plans to fix the problem.”

To mitigate the impact of these issues, Bushee (1998, 2001) develops a two tiered
institutional investor taxonomy based on observable characteristics of the portfolios that
each institution holds. For the first layer of the classification scheme, he identifies eight
variables that characterize prior trading behaviorz. Bushee calculates these eight variables

at the end of each calendar quarter for every institution on the Thomson Financial

 

2 These are: an institution’s quarterly portfolio turnover percentage, its quarterly portfolio turnover
percentage using only sales transactions, its percentage of holdings held continuously for two years, the
percentage of its portfolio firms that it holds for more than two years, its percentage of holdings that
comprise blocks of 5% or more, the percentage of portfolio firms held in blocks of 5% or more, its average
percentage ownership of its portfolio firms and the average investment size (measured in millions of
dollars) in its portfolio firms.

database and then computes annual averages. He then uses these year end averages as
inputs to a factor and k-means cluster analysis which condenses them into two measures.
These first is a measure related to the size of holdings. The second is a measure related to
the frequency of portfolio turnover. The k-means cluster analysis procedure produces
factors that explain the common variance between the two characteristics. Thus, firms in
each cluster are more similar to each other than firms in any other cluster.

Having done so, he is then able to classify institutions into three types: dedicated,
transient and quasi-indexer. Dedicated institutions tend to trade infrequently and hold
large positions in a few firms. Transient institutions are characterized by having small
ownership stakes in several firms and turning over their portfolios frequently. Quasi-
indexers hold large well diversified portfolios and do not trade frequently.

In the second tier of the Bushee3 taxonomy, he identifies fifteen stock
characteristics that relate to investment style. In similar fashion, he condenses these
variables into four factors that explain the common variance with respect to size
considerations (firm size, maturity and index membership), growth characteristics
(historical sales, earnings growth and firm risk), value characteristics (income investment
style features such as high book to market ratios, earnings to price ratios and dividend
yields) and fiduciary incentives (high stock ratings, steady earnings growth, lower
leverage and positive earnings). From these four factors, he classifies each institution into

the investment styles large value, large growth, small value and small growth. The terms

 

3 This procedure is detailed in Abarbanell, Bushee and Raedy (2003). The fifteen characteristics are: the
weighted average market capitalization, the percentage of holdings that are S&P 500 firms. the weighted
average time since the stock was listed, the weighted average price per share, the weighted average
earnings growth, the weighted average beta, the weighted average standard deviation of returns, the
weighted average earnings to price ratio, the weighted average dividend yield, the weighted average book
to price ratio, the percentage of holdings in firms with 5 consecutive years of earnings growth, the weighted
average S&P stock rating, the percentage of holdings in firms with positive earnings and the weighted
average debt to equity ratio.

large and small refer to the correlation with market capitalization, whilst value and
grth refer to preferences for the conventional interpretations of the terms value stocks
and growth stocks.

Because of the way in which the Bushee (1998, 2001) method exploits the data, it
serves as a more consistent characterization of what a fiduciary entity really does. We
therefore employ his taxonomy in all of our empirical tests. It is important to note that the
two levels of the Bushee system are not substitutes. The categories transient, quasi-
indexer and dedicated are relatively fixed and refer to a given institutions overall
investment philosophy. The categories large value, large growth, small value and small

growth refer to possibly temporary preferences for given stock characteristics.

1.4 The Endogeneity of Ownership

A major obstacle to drawing inference in ownership -— perfonnance studies is the
problem of endogeneity, which may arise in either of the familiar forms; omitted variable
bias and reverse causality. Therefore, for each test that we propose, we perform Fama-
Macbeth (1973) regressions as an exploratory procedure. Thereafter, we account for
unobserved heterogeneity by performing the same regressions using firm-level fixed
effects analysis in a panel data setup. However, as Zhou (2001) noted, the problem with
using fixed effects estimators in this context is that they rely on within firm variation to
identify the relationship between firm value and ownership. Considering that the cross-
sectional variation in ownership is far richer than that within each firm, using fixed

effects may lead to a rejection of the performance ownership relation even if one exists.

Furthermore the fixed effects method will not get around the reverse causality problem.
We therefore also implement an instrumental variables procedure.

The advantage in using an instrumental variables approach is that we can mitigate
both sources of endogeneity. However, the drawback lies in assessing the validity of the
instruments. Himmelberg, Hubbard and Palia (1999) suggest using the log of total assets,
its square and the standard deviation of residuals fiom a CAPM regression as instruments
for management ownership. Given that we have three different types of potential
endogenous measures of institutional ownership, we incorporate their instruments and
add four more of our own. Our final set includes the log of total assets, its square, the
variance of residuals from a market model regression, beta from a market model
regression, its square, and the error in analyst forecasts of EPS as of the previous fiscal

year and its square.

2.] Hypothesis Development

2.1.] Institutional Demand for Equity Characteristics

Although the Bushee (1998, 2001) categories are not legal definitions, they are a
useful guide to each institution’s corporate mandate because of the way in which they are
constructed. Transient institutions undertake investments that are intended to maximize
short term gains. We therefore expect them to be most sensitive to fluctuations in the
pattern of returns. To be truly effective in their trading activity, however, we also
anticipate that they will seek out situations in which they can gain a competitive
advantage over other investors in terms of information gathering. At the other end of the

Spectrum, dedicated institutions commit significant resources to a handful of firms and do

not sell off easily. Quasi-indexers are also substantially committed to their portfolio
firms, however, they do have greater flexibility than dedicated investors in terms of when
and how much they can sell. This suggests that dedicated and quasi-indexer institutions
should be more concerned with firm characteristics that facilitate monitoring and firm-
specific information acquisition. However, as quasi-indexers hold broader portfolios than
dedicated institutions, they face higher costs of oversight. We therefore expect them to be
attracted to firms in which it is easy to glean information to a greater extent than
dedicated investors.

To complete the ownership picture, we also consider the role of insiders. There
are many reasons to suggest that insiders should exhibit different demand for stocks in
their own companies as compared to institutions. Firstly, a significant proportion of
insider wealth is determined by labor income. This has the effect of dampening the
demand for their own stocks as they attempt to hedge against future bad states;
Institutions, of course, do not face this issue. Secondly, insiders may face more financial
barriers in capital markets. We typically think of them as being more wealth constrained
and consequently likely to face greater borrowing restrictions than institutions.
Furthermore, legal considerations cause them to have short selling constraints in their
own stocks (SEC rule 16(b))4. These considerations suggest that they maximize a
different objective function to the large, comparatively less constrained financial
institution. Thirdly, insiders are likely to be more information privileged relative to
institutions. Although insider trading regulations limit what can be done with this

information (SEC rules 10(b-5)5 and l6(b)6), it would be naive to think that it is of no

 

I Although it should be noted that 13(i) institutions also incur some restrictions with respect to short sales.
3 Acting on information that is not publicly available.

10

value and has no bearing on insider behavior. For instance, Seyhun (1986) and
Lakonishok and Lee (2001) demonstrate that insiders possess some ability to predict
future returns. We therefore posit that insiders will be attracted to firm characteristics that
obscure managerial action7. The reasons for this are twofold. Firstly, the more difficult it
is to verify a managerial decision, the lower the probability of involuntary turnover in the
event of a bad outcome. Secondly, even if an insider does not intend to willfully
manipulate company resources to their own benefit, self interest will always make them

prefer a situation in which this is possible. We therefore hypothesize the following:

HI Investors exhibit heterogeneous preferences for security characteristics which
exist because of the number of constraints that they face. Transient institutions
face the least constraints and are most concerned with the properties of returns.
Dedicated and quasi-indexer institutions are concerned with the cost of
overseeing managerial behavior. Insiders are primarily motivated by concerns

about personal wealth.

In order to measure preferences for firm characteristics, we define generic share
demand functions for all four investor clienteles. Our choice of right hand variables is

motivated by the regressors used in similar studies and the desire to avoid simultaneity

 

6 Engaging in roundtrip trades in under 6 months.

7 With respect to the determinants of managerial shareholdings, we turn to Holdemess, Kroszner and
Sheehan (1999) as a guide. They find that firm size and leverage are negatively related to managerial
ownership. The finding for size is reflective of wealth constraints whilst they argue that the leverage result
is consistent with Stulz (1988). Stulz (1988) demonstrates that managers can use leverage to increase their
voting rights for a given level of equity ownership. The other important finding in Holdemess, Kroszner
and Sheehan is that volatility as proxied by the residual from the market model regression has a non—linear
effect on inside ownership. The coefficient on the level term is positive and the squared term is negative for
the 1995 cross section of US. firms. Interestingly, they can not replicate this finding for the earlier (1935)
cross-section. The authors interpret this as demonstrative of managers being more willing to commit
significant personal equity towards their firms because of the wide variety of hedging products that have
existed in financial markets in more recent times.

11

bias. To take a complete view of the firm’s environment, we assign each of our control
variables to one of three categories. The first group of variables deals with the
relationship between risk and returns (size, book to market, beta and previous excess
returns). The second group deals with characteristics that determine the relative opacity
of the firm to outside investors (R&D, advertising, free cash flows, capital expenditures
and market model residual variance). The third group, which is closely related to the
second, deals with firm conditions that permit monitoring and discipline from the
external market (leverage and the entrenchment index score). Naturally, there is some
degree of overlap in their interpretation as proxies. However, we find that by proceeding
in this manner, we are able to uncover some very compelling differences in the way in
which institutional investors allocate their portfolios. Furthermore, these differences are a
very plausible antecedent to the causal interpretation that we discuss later in the paper.

The complete regression specification for H. 1 is as follows:

0wnership%,-, = a0 + a1 log(ASSETS),-, + 0’2 log(B / M)” + a3BETA,-,_1
+a4EXRET3 —12,-, + a5 R0,, + a6AD Vi, + a7 EBITAl-t + agCAPEXi,
+a9RESID VARi,_1 + alOLE VERA GE“ + a] IEINDEXl-t + 5i: (5)

log(ASSE T S). Firm size is measured as the log of total assets at the end of fiscal
year t. For insiders, the impact of firm size may be purely related to the manager’s budget
constraint. Larger firms require more resources to achieve a certain level of ownership.
Size may also be a proxy for information asymmetry and the scope for moral hazard.
Larger firms may have less information asymmetry because of the degree of scrutiny they
receive from capital markets and rating agencies. We therefore predict a negative

relationship between firm size and insider holdings. From the perspective of institutional

12

investors, we propose that firm size will be a desirable quality for those institutions that
tend to engage in monitoring. The costs of monitoring decrease as firm size increases
because it becomes a shared undertaking. Furthermore, size can act as a proxy for
prudence and liquidity considerations. We therefore expect a positive coefficient on firm
size for dedicated and quasi-indexer institutions. In contrast, transient institutions do not
engage in any meaningful long term monitoring because of the rapid turnover of their
portfolios. We therefore propose that these investors will seek out situations in which
they can use their superior information gathering capabilities to extract gains. As a result,
we conjecture a negative coefficient on the size variable for transient institutions.
log(B/M). We take the log of the book to market ratio at the end of fiscal year t.
As proposed by Gompers and Metrick (2001), it is possible that institutional investors
may be attracted to characteristics that have been shown to be associated with historical
returns e.g. high book to market stocks earning higher returns than low book to market
stocksg. However, a priori, the direction of the relationship is not clear-cut. If the book to
market ratio represents a trading rule through which one can earn superior returns, the
association should be positive across all types of institutional investor. If however, one
interprets the book to market ratio as a rationally priced risk factor, then the relationship
could be negative for some types. If high book to market firms are on average more
distressed than low book to market firms, transient institutions might display an aversion
towards that characteristic despite the potential for superior returns because of the
mechanism through which those returns must be achieved. By virtue of the small Size of
their holdings and their Short horizon, transient institutions have little incentive to

monitor this type of firm and enforce long term operational changes. Furthermore, if

 

8 Fama and French (1992, 1993).

13

growth (low book to market) firms have less precise information about their future
earnings, then transient institutions should herd towards those firms because of their
ability to exploit the information differential between themselves and the rest of the
market. Therefore, we predict a negative coefficient on the book to market ratio in the
transient regression. On the other hand, dedicated and quasi-indexer institutions (that by
design become intimately knowledgeable about their target firms) might be willing to
take on these distressed firms with a view to improving them. Therefore, we predict a
positive coefficient for both quasi-indexer and dedicated institutional investors.

BETA. The beta coefficient from a market model regression is estimated. Pressure
from their clients may cause transient institutions to be drawn to more “glamorous”
stocks. We therefore predict a positive coefficient for beta in their demand function. We
predict that dedicated institutions will not use beta as a primary criterion in making their
assessment of the firms in which to invest and thus do not make a formal prediction with
regards to their preference. Quasi-indexers will tend to weakly avoid both high and low
beta stocks because they essentially hold the market portfolio. Consistent with Demsetz
and Villalonga (2001 ), we posit a negative coefficient in the insider holdings regression.

EXRET3-12. We measure cumulative excess stock returns relative to the CRSP
equal weighted index over a 9 month period. The terminal date for measurement is the
beginning of the quarter over which we measure our ownership variables. This variable is
intended to demonstrate if different investor classes use historical stock price
performance to make future trades. We predict a positive coefficient in the transient
regression. If so, this lends credence to our contention that they possess superior

information which they are able to act upon. The long run focus of dedicated institutions

14

makes it unlikely that this factor will affect their holdings of stocks. Quasi-indexers may
also be somewhat constrained in their ability to implement momentum style trading
strategies.

R&D, ADV and CAPEX. We collect R&D expenditures at the end of each fiscal
year and divide them by assets. Firms which do not report R&D expense in Compustat
are assigned a value of zero. We collect advertising expenditures at the end of each fiscal
year and divide them by assets. In like fashion, firms which do not report advertising
expense in Compustat are assigned a value of zero. We collect capital expenditures at the
end of the fiscal year and scale by assets. Following Himmelberg, Hubbard and Palia
(1999), we consider the preceding three variables to be indicative of the scope for
managerial discretion. R&D and advertising are examples of ‘soft’ investments because
they do not directly produce a tangible good. Capitalexpenditure on the other hand
produces an observable outcome and is thus less easy to manipulate. We therefore expect
. that firms with a greater emphasis towards intangible assets will be more prone to ,
managerial indiscretion. Although R&D and advertising are both ‘Soft’, we argue that
advertising may be ‘softer’ than R&D. We acknowledge that both types of investment are
subject to managerial malfeasance; however, in the case of R&D, it is arguable that the
link from investment to outcome is more clearly defined. At a cost, outside investors can
familiarize themselves with the technologies employed and assess the usefulness of a
project in a more quantitative way. The success of an advertising campaign on the other
hand is determined by a variety of factors including trends, culture, location and
historical market positioning. All of these factors are arguably more difficult to place in a

present value type framework. AS has been our theme so far, we expect that insiders will

15

prefer the characteristic that gives them greater discretion (i.e. the softer investment) and
institutions will avoid it to the degree that it runs counter to their investment objective.
Therefore we expect a positive coefficient on advertising in the management regression.
Bushee (2001) demonstrates that transient institutions may pressure management to cut
R&D expenses in order to meet earnings targets. Therefore we expect transient
institutions to be attracted to firms with high R&D because of the earnings cushion that it
provides.

EBI T A. We compute earnings before interest and taxes divided by assets in fiscal
year t as our measure of free cash flows. Jensen (1986) and Lang, Stulz and Walkling
(1991) find that firms with high free cash flows suffer more severe agency problems.
From the perspective of a self-interested manager, a high level of free cash flow may be a
desirable characteristic. However, free cash flows also may be indicative of good
fundamentals. Institutions may therefore be attracted to firms with high EBITA. We
include EBITA as a control variable.

RESIDVAR. The residual variance from the market model regression.
Idiosyncratic risk may have some association with management’s desire to own shares in
their own companies. Demsetz and Lehn (1985) argue that firm specific risk is a proxy
for the scope for managerial action, thereby making it a desirable quantity. If so, firm
specific risk should have a positive coefficient in the managerial ownership regression.
When we subdivide institutions according to type, they should differ in their preferences
for firm specific risk according to their ability to collect and act on superior information.
Therefore we propose that firm specific risk will have a positive effect on transient

investor holdings. On the other hand, in the level, dedicated and quasi-indexer institutions

16

will avoid firms with high firm specific risk because it increases the cost of oversight.
The coefficient should be more negative in the case of quasi-indexers because they face a
higher monitoring cost than dedicated investors.

LE VERAGE. We calculate leverage as the fiscal year end sum of Short term and
long term debt scaled by assets. Leverage is a proxy for the degree of monitoring
intensity because creditors provide an alternate mechanism by which managers can be
disciplined. Leverage should therefore have a negative effect on institutional ownership
as it lightens the burden of monitoring, particularly for those institutions that need
assistance. Therefore, given that quasi-indexers face the highest cost of information
acquisition, their holdings should have a negative association with the leverage variable.
As transient institutions do not engage in monitoring in the conventional sense, we do not
posit a role for this variable in their demand function. At the other end of the spectrum, ‘
dedicated institutions should not need assistance monitoring their target firms because of
their narrow focus. We therefore predict that they will also not consider leverage in
determining their optimal holdings of a given stock. If managers need to signal their
commitment to firm performance by owning equity, leverage decreases the size of the
ownership stake required. It is for this reason that we expect leverage to have a negative
effect on insider demand for equities.

E-INDEX. The entrenchment index score". This variable controls for differences
in governance environment. It is comprised of Six govemance provisions that empower
management, or, alternatively, discourage takeovers. These are: staggered boards, limits
to shareholder amendments of bylaws, supermajority requirements for mergers,

supermajority requirements for charter amendments, poison pills and golden parachutes.

 

9 For further discussion. we refer the reader to Bebchuk. Cohen and Ferrell (2004).

17

2.1.2 Herding Behavior of Institutional Investors and Informed Trade

H.I posits that differential preferences for equity characteristics are a
consequence of the nature of investment activities that a given institution undertakes. We
therefore seek to establish whether these investment activities have an effect on the target
firms themselves. If institutions are engaged in the acquisition and processing of firm
specific information, such activity could generate positive spillover effects to other
market participants by making prices more informative. Within our framework, this
effect should be most apparent in firms that experience an increase in holdings by
dedicated institutional investors. It is arguable, however, that if transients are also
proficient acquirers of information (as postulated in AU), the same effect should hold
true for them. Innovations in financial markets are easily duplicated, therefore whatever
advantage 3 given transient investor has in a given stock cannot persist indefinitely.
Consequently, a buildup of transient blocks should also, ultimately, reduce the opacity of
a firm. Finally, if quasi-indexer institutions free ride on the monitoring activities of
others, then their trades Should be minimally associated with firm transparency. In

summary, we contend that:

H.2A The accumulation of proficient information gathering shareholder blocks

enhances the transparency of a firm.

Morck, Yeung and Yu (2000) and Dumev, Morck and Yeung (2004) make the
case that the R2 statistic from a market model regression proxies for the lack of price

informativeness in a given stock. More recently, however, Kelly (2005), Griffin, Kelly,

18

and Nadari (2006) and Ashbaugh-Skaife, Gassen, and LaFond (2006) demonstrate that
this argument does not comport with the established market microstructure findings of
Shiller (1981), LeRoy and Porter (1981) and West (1988). Nor is it consistent with the
empirical finding that mandated reductions in trading cost structure are associated with
subsequent price stability, as per Jones and Seguin (1997) and Bessembinder and Rath
(2002)

Kelly’s (2005) study provides direct evidence that both higher trade costs and
higher illiquidity are associated with a lower market model R2. Additionally, he finds that
larger firms, older firms and firms with less binding short-sale constraints have higher
market model R2 statistics. Following this line of literature, the current study contains
tests to determine whether increases in shareholdings by each of the Bushee (1998, 2001)

classes of institutional investor are associated with subsequent changes in R2. The R2

statistic is obtained by running annual regressions of daily individual stock returns
against the market and the returns on the firm’s 2-digit SIC industry excluding the firm

itself, as follows:

Rit = a0 + IBM/(77,17 RMKT,t + .51,vo,z‘¢tt RIND,1'¢I‘I + 5'" (6)

Having collected R2 for each firm, we apply a logistic transformation and regress the

transformed variable against changes in institutional ownership, the market value of
equity, the number of years since the firm was first listed in CRSP, the number of

analysts providing estimates of the firm’s EPS in l/B/E/S and Amihud’s (2002) measure

19

of illiquidity. The Amihud (2002) illiquidity measure is defined as the annual average of

daily returns scaled by daily volume:

T
1 W l
ILL] -=-——x —“-
Q” D Elma]

 

1' r=1 (7)
The final regression specification is, therefore, as follows:
R2 3
In 2 = a0 + Z ark A0wnership%k,-,_1 +a4SIZE,-,_1
+a5AGEl-H +a6Nt/MEST,,_l +a71LL1Q,.,_1 (8)

The level demand equations proposed in H. 1 identify static relationships. We now
extend our model to incorporate the dynamics of trade. Beyond the ‘fundamental’
relationships proposed in H], some trading decisions will be made in response to the
actions of others i.e. herding behavior. Broadly. herding arises out of the desire to avoid a
negative outcome. For instance, institutional shareholders may prefer to act with the
crowd (even in the face of negative signals) in order to avoid punishment from clients in
the event that their contrarian decision does not pay off. Under conditions in which some
participants have more refined information than others, relatively uninformed traders may
mimic the trades of the informed in the hopes of earning superior returns.

In a study of US. mutual fund holdings over the period 1974-1994, Wermers
(1999) finds that while overall herding by mutual funds is fairly lowm, herding in

situations where information is imprecise is far more pronounced. Specifically, he finds

 

'0 Roughly 3 more funds out of every 100 trade on the buy Side than would be expected if trading were
completely random.

20

excessive herding by investment vehicles that characterize themselves as growth funds1 I.
Furthermore, sell-side herding is concentrated in small stocks'2 across all investment
style categories.

In this paper, we posit the existence of an information cascade. The management
team should possess the most accurate information about the current state of their
company by virtue of the fact that they observe operations on a continuous basis.
Thereafter, we believe that transient institutions will be the second most informed about
the future prospects of a firm. In general, transient institutional investors will be able to
overcome the noise (relative to the management team) in their signal through the volume
of trade that they are able to commit to a firm”. Furthermore, Ke and Petroni (2004)
demonstrate that transient investors communicate directly and indirectly with
management and appear to be able to forecast earnings declines at least one quarter
before they take place, unlike dedicated and quasi-indexer types. However, dedicated
institutions Should also be very well informed because of their role as long term monitors
of firm performance. Finally, quasi-indexer institutions should be the least informed
because of the breadth of their holdings.

Empirically, we predict that dedicated institutions will not follow the trades of
other investors. Primarily, this is because their investment mandate requires them to
evaluate stocks based on characteristics beyond the short term holdings of others. By
comparison, quasi-indexers have a need to acquire subsidized monitoring. We therefore
predict that they will be the class most affected by other investors’ trades. In particular, if

dedicated institutions are more proficient monitors, the quasi—indexer class should tend to

 

" Investments in firms with uncertain future cash flows.
" This is also a proxy for information asymmetry.
" We explore this issue more thoroughly in H3.

21

follow them into stocks. In some sense free-riding on the efforts of the dedicated. If this
explanation holds and entrenchment has any validity, we also expect to see management
tend to reduce their holdings in firms in which monitoring type institutions accumulate
blocks. Under these conditions, insiders would no longer need to hold as many shares to
credibly signal their commitment to outside investors.

The Ke and Petroni (2004) finding and the Bushee (2001) finding with respect to
R&D expendituresI4 suggest that transient institutions act independently and have
superior information gathering capabilities. However, as proposed in H. 2A, the
accumulation of shares by monitoring institutions may portend a change in the opacity of
the firm. We therefore conjecture that transients will tend to herd out of securities in
which long term investors have bought significant blocks. We summarize our predictions

about the dynamics of trade in H. 23.

H. 28 Dedicated institutions make trading decisions that are independent of the othrr
institutional investor classes. Quasi-indexer institutions incorporate the trades of
investors with superior monitoring abilities in making their trading decisions.
Transient institutions rely on their own research. but will avoid situations in

which their informational advantage is diminished.

To test this conjecture, we include the full set of control variables from the level

demand functions in H], which leads us to the following expression15 :

 

'4 Transient investors increase the probability of myopic R&D decisions.

'5 The subscript k references the three other investor classes relative to the class of interest. For instance, if
k=l denotes transient institutions then k=2, 3 and 4 correspond to the management team, quasi-indexers
and dedicated institutions respectively.

22

4

AOwnership%kit = a0 + Z aJ-A0wnership%
j=k

+a6 log(B / M),~,_1 + a7BETAi,_1+ agEXRET3 — 12,-,_1 + agRDI-t_1

+a10ADVl-,_l +(Z11EBIT it—l + 0312CAPEX”_1 + a13RESIDVAR1-Fl

+a14LE VERA GE” + a15EINDEX i, + 8,, (9)

j”‘1 + a5 log(ASSETS),-,_1

Note that each of the control variables in the regression is expressed as of the previous
fiscal year. Therefore, this regression is designed to answer the following question:
contingent on observing the change in holdings of all other classes of investors last year,

how does the class of interest respond this year?

2.1.3 The Link Between Ownership Structure and Returns

The implication of H. 1 and H.2 is that there is not perfect agreement in the market
for firm securities. This disagreement arises because of differential information. and
differential preferences for firm characteristics, leading, to some degree I of market
segmentation. As we discussed in the introduction, managers face certain portfolio
restrictions because of the structure of their compensation packages, and institutions are
affected by prudence and liquidity concerns. Even in an equilibrium framework such as
Merton’s (1987) ICAPM, investors may express preferences for stocks that allow them to
hedge uncertainty about future consumption-investment opportunities. Either way, the
implication is that there exist determinants of the demand for stocks that are not
accounted for in a standard CAPM framework. If this is correct, then we Should expect to
see that reflected in the returns accruing to stocks which possess the characteristics that

produced the disparities in demand in the first place. We posit transient institutions as the

23

marginal price setter in the economy because of the aggregate size of their holdings and
volume of shares that they can move at a single time.

Wermers (1999) performs a variety of future returns tests and finds that mutual
fund herding does contain some predictive content for long term future performance. He
finds that those stocks that are heavily bought by funds outperform those that are heavily
sold for six months following the execution of the trade. After six months, the
performance disparity declines but does not reverse. Furthermore, he finds that a zero
investment portfolio strategy that buys stocks that experience the highest level of buy
side herding and sells stocks that experience the highest level of sell side herding
generates positive, size adjusted abnormal returns of 13% during the same quarter as
portfolio formation and 4% over the following two quarters.

Nofsinger and Sias’ (1999) definition of a herd differs from Wermers (1999) in
that they are only concerned with the absolute change in the percentage of shares owned
by institutions. When looking at contemporaneous returns, they find that those firms that
experience the largest decrease in institutional ownership over the herding year earn
abnormal returns of -l3.12% whereas those firms that experience the largest increase in
institutional ownership earn abnormal returns of 18.38%. Furthermore, when they extend
their return horizon to 12 and 24 months after the herding year, they are unable to find
any evidence of return reversals. The authors interpret their collected evidence as being
consistent with institutions being both positive feedback traders and traders who purchase
undervalued and sell overvalued stocks.

To the extent that transient investors are less constrained relative to the other two

classes, they will be the most capable of trading in response to fluctuations in return

24

patterns. Since we are making the conjecture that transient institutional investors are the
marginal price setter, we conclude that transient holdings Should be associated with

higher subsequent values of Q. We summarize this discussion in hypothesis H3:

H3 Trading by transient institutional investors is associated with positive returns.
The presence of transient institutions is associated with higher subsequent values

of Tobin ’s Q.

In order to test the first component of H3, we regress individual monthly excess
stock returns on both the previous and contemporaneous level of ownership by each of
the four investor classes in separate specifications. As we already have a model in mind”),
we also include a test with a squared term in insider holdings. Finally, we partition
transient holdings by investment style and perform the same regressions according to
whether the institutions are categorized as value or growth type investors. In each
regression specification, we include the control variables proposed in Brennan, Chordia

and Subrahmanyam (1998). Variable definitions are provided in Table 7:

3
Ri, = a0 + Z ak Instz'tutioIIO/ok” +a4MI-Ii, + aSMflg + a6SIZEi, + 6278 / M it
k=l
+a8NASDUMi, +a9PRICEl-t + alONYDVOLi, +a| INASDVOLI-t + aIZYLDit
+al3RET2—3i,+a]4RET4-6,-,+a15RET7—12i, (10)

In order to test the second component of H3, we estimate the expression below”:

 

'6 McConnell and Servaes (1990, 1995. 2005)
'7 The subscript k=l.2.3 denotes transient. quasi-indexer and dedicated institution respectively.

25

3
Q" = or“ + Z aklnstitution‘yoki, + a4MHl-t + a5MH3 + a6RDi, + (17 AD Viz
k=l
+a8LEVERAGEi, +a9 log(ASSETS),-, (11)

As a robustness check, we also consider the role of investment style with
respect to transient institutions. If transient institutions tend to tilt their portfolios towards
grth (which would tend to have higher Q’s) versus value stocks18 then one might
expect to find a relationship between the two variables as a function of investment style.

We therefore perform tests of Q against subcategories of transient ownership.

2.1.4 Conditional Interpretations of the Ownership — Firm Value Relationship

To this point, we have been working under the assumption that transient
institutions possess more refined information about expected future cash flows and are
more able to act on that informationthan either dedicated or quasi—indexer institutions.
We also propose that managers have a preference for firms in which it is harder for their
actions to be understood by the external market. Two types of institution are primarily
concerned with the ability of management to act in ways that destroy long term firm
value (quasi-indexers and dedicated), whilst transient institutions are more concerned
with short term considerations such as recent stock price performance and liquidity. In
order to complete the picture, we therefore consider three different scenarios that impact
the ability of management to act and the ability of the external market to process the
information about a firm. These are corporate govemance, the expectation of the market

(as proxied by the dispersion of analyst forecasts) and the volatility of cash flows.

 

'8 The results in Table 2 for the early part of the sample are suggestive of this.

26

Corporate governance rules provide the framework through which managers can
act and also through which they can be disciplined. Bebchuk, Cohen and Ferrell (2004)
construct an index of six corporate governance provisions that proxy for the strictness of
a company’s corporate charter. They find that higher scores of the entrenchment index
(which corresponds to worse govemance) are associated with lower values for Q. These
results seem to suggest that governance matters. We expect to find similar results in that
good governance should be associated with higher firm valuation. However, we also
intend to make the case that a real source of discipline in the external market lies in the
pressure of transient investors. Since they are the closest to the management team'9 and
most capable of generating real effects through their trading decisions”, we argue that
they will continue to have an effect on firm value whereas the threat posed by the other
two classes will be subsumed by whatever governance provisions the firm already has in

place. Therefore, it is our contention that:

H4 The role of transient institutions in relation to Q is most pronounced under

circumstances in which information about future earnings is imprecise.

Notice that expressing H4 in terms of transient institutions does not lead to any
loss of generality. An alternative formulation could easily express it from the perspective
of quasi-indexer and/or dedicated institutions. In other words focusing on whether there
are conditional circumstances under which the effect of monitoring institutions becomes
apparent. Both formulations are answered in the tests that we propose. To test H.4, we

initially add the entrenchment index score as an independent variable to the Q regression.

 

'° Ke and Petroni (2004).
20 For instance by abandoning the stock en masse.

27

Thereafter, to isolate any structural differences related to governance scores, we partition
our data set into halves based on the level of the entrenchment index score. We then
interact each ownership variable with the governance index half to which the firm

belongs and re—run the Q regression.

3 3
Q” : a0 + Z ark Institution% kit + Z ,Bk InstitutionO/o kit x E -—INDEX _1

k=l k=l
+a4MHi, + asMHg ‘I' a6RDit + a7ADVit + a8LEVERA GEN
+09 log(ASSETS),-, (12)

Investment analysts are an important intermediary between firms and individual
investors. Their research and recommendations serve as a useful proxy for the market’s
overall expectation of a firm’s future cash flows. Consequently, when there is
considerable disagreement about future performance by professional analysts, it is
reasonable to assume that the same holds true in the market at large. Following the same
procedure used in equation (12), we partition our sample by the level of forecast

dispersion and interact this dummy variable with the level of ownership.

3 3
Q” = a0 + Z ak Institution‘Voki, + Z flkbzstitution‘Voki, x F —DIS.P_1

k=l k=1
+a9 log(ASSETS),, (13)

The third set of conditional tests that we perform considers the role of cash flow
volatility. We posit persistence in the time series patterns of cash flow volatility.
Therefore, it should be harder to reliably forecast the future cash flows of firms with a

higher existing level of cash flow volatility. If transient institutions are better informed

28

than other institutional investor classes, they should perform well under conditions in
which cash flow volatility is high. We therefore anticipate that the interactions of
transient holdings and cash flow volatility should have a positive effect on Q.

3 3

Q” = a0 + Z ak Institution%kit + z ,Bk Institution% kit x CASH VOL _1
k=l k=l

+a4MH,, + aSMHg + a420,, + 0:7 AD Vi, + aSLE VERA 013,,

+a9 log(ASSETS),-, (14)

To take the information gathering hypothesis further, we perform subtests where
we partition the ownership by transients according to investment style. If transient
institutions excel when the information differential between themselves and the rest of
the market is high, then investment style should have an effect on Q. Smaller firms and
firms with growth characteristics have less precise information about expected fiiture
cash flows. Therefore, there should be a positive association between the presence of
transient institutions pursuing a growth style and firm value. This test also serves as a
robustness check. If value style transient institutions are found to have a positive effect

on Q, then we can reasonably conclude that the proposed relationship is not purely driven

by portfolio strategies that overweight growth (presumably high Q) stocks.

3.1 Data and Descriptive Statistics

We initially construct our dataset by selecting all firms for which we can obtain
complete corporate ownership data. Institutional holdings are obtained from the Thomson
Financial Securities Ownership Database, which itself is derived from records of SEC

form 13(f). Form 13(f) must be filed quarterly by all investment managers who exercise

29

control over at least $100 million in securities. With respect to common stocks, form
13(f) contains all positions greater than 10,000 shares or $200,000.

In order to reconcile the annual insider ownership data with the quarterly
institutional data, we use a two month window in merging the two databases. In other
words, if a firm has a fiscal year ending that does not exactly match a quarterly
institutional filing date, we use the holdings of institutions from the previous quarter,
provided that the previous quarter is less than two months prior. We acknowledge that
this reduces the accuracy of our ownership measures. However, we find that only a few
observations are off by 60 days in an annual regression framework. Furthermore, the
results remain the same when we re-run the analyses using only the firms which coincide
perfectly in ownership reporting dates.

Having. collected a complete panel of ownership data, we then label each
institution according to the categories defined in Bushee (1998, 2001). Around 60% of
the institutions tracked by Thomson Financial are'defined as quasi-indexers. Transients
make up the second largest group, with around 6% being counted as dedicated.

On average, transient institutions turn over their portfolios almost three times as
much as dedicated and quasi-indexer institutions. The quarterly turnover percentages are
8.91%, 7.65% and 23.84% for dedicated, quasi-indexer and transient investors
respectively. Transient institutions only hold around 44% of their portfolio firms for two
years or more whilst dedicated and quasi-indexer investors hold around 70% of their
firms for long periods (more than 8 consecutive calendar quarters). Dedicated institutions
hold much larger blocks (average of 4.1% of the target firm vs. 0.5% for quasi-indexers

and 0.8% for transients) and much larger fimrs (average market capitalization of $24.3

billion vs. $6.9 billion for quasi-indexers and $7.9 for transients) than either of the other
two classes of investor. For more detail on the properties of the taxonomy we refer the
reader to Bushee (2001).

Our insider ownership variables are taken from the Standard and Poor’s Executive
Compensation Database (EXECUCOMP) for the fiscal years 1992 to 2003. The data
cover annual holdings by insiders in firms listed on the NYSE, NASDAQ and AMEX.
The Securities and Exchange Commission’s (SEC) defines an insider as an officer,
director or 10% principal stockholder who owns either 1,000 shares or at least $25,000 in
equity securities”. The EXECUCOMP database does not include the holdings of 10%
principal holders who are neither officers nor directors. As a result, it uses a stricter
definition of insider i.e. only those who are actually affiliated with the company. Because
of this artifact, we use the terms insider holdings and managerial ownership
interchangeably.

In addition to having data on ownership, we require that our firms have
accounting data on Compustat and market data on CRSP for the same years that we have
data on their shareholdings. Tobin’s Q is winsorized at the 5% and 95% level to prevent
the results from being dominated by outliers. Finally, we require that the firms are
followed by at least three analysts throughout the sample period. We obtain measures of
analyst forecast error and analyst forecast dispersion from the I/B/E/S database. Forecast
error is measured as the absolute value of the difference between the last mean estimate

of EPS and the actual fiscal year end realized value of EPS. Forecast dispersion is

 

2' Insider holdings are collected from SEC forms 3 and 4. Form 3 is the initial filing related to the
acquisition of equity securities by an insider. It must be filed within 10 days of the purchase. Subsequent
changes in ownership by insiders are reported on Form 4. Form 4 must be filed within the first 10 days of
the month after which the transaction occurred.

31

computed as the standard deviation of the last forecast of EPS for a given fiscal year.
These selection criteria produce an unbalanced panel of 2,609 firm years. Table 1 gives
summary statistics for the main variables used in the study.

The results in Table 1 show that the sample consists of fairly large, well
performing stocks. Average Tobin’s Q exceeds 2 for seven of the twelve years in our
study. The average beta is around 1.30 throughout the sample period. Average firm size
measured by both market capitalization and total assets increases dramatically (85.70%
and 54.97% respectively) over the sample period. The leverage ratio is fairly constant
around 20% for all twelve years. Expenditures on R&D and advertising remain close
throughout the sample.

Average ownership by insiders displays a steady decline from a 6.84% in 1992 to
a low of 4.31 % in 2003. On the institutional Side, the percentage of shares held by
transient investors becomes increasingly important in the later years of the sample; Their
holdings average 9.41% in 1992 and increase to 23.8% by the end of the sample with a
peak of 36.41 % in 2001. Quasi-indexers do not change their holdings by much,
consistently accounting for around one third of firm ownership. Finally, dedicated
investors increase their average holdings over the first six years (from 12.72% in 1992 to
19.85% in 1997) and then experience a sharp decline (1.08 % in 1998) before recovering
towards the end of the sample (10.44% in 2003).

In Table 2, we investigate the distribution of institutional ownership by
investment style. Table 2 shows a joint characterization of all the institutions on the
Thomson Financial database according to both layers of the Bushee (1998, 2001)

taxonomy. For instance, the first row shows the number and percentage of dedicated

32

institutions that are classified as following a large growth, large value, small growth and
small value trading strategy respectively in the year 1992.

In the first half of the sample, transient institutions appear to exhibit a deliberate
bias towards stocks in the growth category. For four out of the first six years, allocations
to small growth and large growth stocks come in first and second place in terms of
portfolio weight. While the importance of the large growth category fades away in later
years, the importance of the small growth category to transient investors continues into
2003. From 1998 onwards, the small value category becomes increasingly important as
over 30% of transients are routinely characterized as following a small value strategy.

1998 also represents a remarkable turning point in the trading strategy of
dedicated institutions. They appear to abandon the large value and large grth
categories altogether and begin to aggressively pursue a small value approach. Notice
also that this shift coincides with a sharp decline in thetotal number of institutions
characterized as dedicated (from 157 in 1997 to only 23in 1998). This corresponds
closely to the decline in the overall share of market value held by dedicated institutions
shown in Table 1.

As one might expect, the holdings by quasi-indexers exhibit the most stability.
Around 60%-70% of quasi—indexers are either large growth or large value institutions in
every year. Since quasi-indexers essentially hold the market portfolio, we would expect
them to have a bias towards larger stocks and, moreover, that the proportions they select
would change very gradually.

Table 3 presents some preliminary results on the preferences that the different

classes of investors have for stock characteristics. Holdings by transient institutional

33

investors are significantly positively correlated with prior returns, systematic risk as
measured by beta, and firm specific risk as measured by the residual variance from the
market model regression. On the other hand, the book to market ratio exhibits a negative
correlation with transient ownership. This seems to suggest that transient institutions herd
towards “glamour” stocks. They also display a negative correlation with stocks with
heavy long term investment as measured by capital expenditures.

In terms of discretionary investments, the correlation with R&D is positive whilst
advertising is negative. Transient ownership is also associated with higher levels of cash
flow volatility. In contrast, quasi-indexers, dedicated investors and insiders exhibit
significant negative correlation with this variable. Finally, transient holdings exhibit
negative correlation with measures of analyst uncertainty i.e. forecast error and forecast
dispersion. The holdings of the dedicated class of investors are only weakly correlated
with measures of systematic risk. However, they do show a highly significant negative
correlation with residual variance. In terms of discretionary spending, advertising 15
positively correlated with the dedicated class. Availability of free cash flows is also
shown to have some importance whilst forecast dispersion has a positive association with
dedicated ownership.

The directions and Significance of the correlations with the quasi-indexer class
mirror the dedicated class fairly closely (typically in a more pronounced manner both in
magnitude and statistical significance) with one important exception. Quasi-indexer
holdings are positively correlated with entrenchment index scores. Recall that higher

index scores are associated with worse governance.

34

Management holdings are significantly negatively correlated with size, beta,
leverage, residual variance, R&D and the entrenchment index. Their holdings are
positively associated with free cash flows and capital expenditures. Although Table 3 is a
simple correlation matrix it is nonetheless suggestive of the preferences we propose in
H]. We explore the interpretations of these associations more fully in the multiple

regression framework of section 4.1.

4.1 Results
4.1.1 Testing the Institutional Equity Preferences Hypothesis

The first specification aims to undercover the forms of the demand functions that
different groups of investors have for stocks. In order to test hypothesis H], both Fama-
Macbeth (1973) and fixed effects regressions in which the percentage of shares held by
each class of investor is the dependent variable are performed. Tables 4 through 8 present
the restilts with the respect to H. .I . The Fania-Macbeth procedure involves estimating
each equation in H] for each year (1992-2003) and then calculating the mean and time
series standard deviation of the 12 estimates of the coefficients. The fixed effects
procedure occurs at the firm level. We structure our regressions in a stepwise fashion
based on the three groups of stock characteristics proposed in H. .l .

Given the problem of unobserved heterogeneity, and limitations in the length of
the time series, we believe that the firm-level fixed effects are a more appropriate
technique. The Fama-Macbeth results Should therefore be considered as a preliminary
investigation of the underlying patterns that we hypothesize. However, as this study

builds on Gompers and Metrick (2001), the Fama-Macbeth (1973) regressions allow for

35

comparability. Nonetheless, in discussing the results below, the fixed effects coefficient
estimates will be accorded prominence over the Fama-Macbeth (1973) coefficient
estimates.

The results in Table 4 Show that firm size, prior excess returns and market model
residual variance have a statistically significant positive effect on aggregate institutional
ownership. With respect to insiders, firm size, R&D expenditure and entrenchment index
score are consistently associated with lower levels of management ownership. On the
other hand, advertising expense has a statistically significant positive effect. As
hypothesized, the finding with respect to firm size is probably related to the wealth
constraints faced by individual managers. In terms of potential manipulation of corporate
resources, we find that management’s preference for firms in which advertising is high
but research and development expense is low is consistent with the “softer investment”
explanation proposed in H. I . In other words, investment in advertising may provide more
amenity potential22 to management than R&D. The negative effect of the governance
index is likely reflective of the fact that a firm with more entrenching provisions would
require a lower level of holdings to insulate management from external discipline.

When institutional investors are separated according to type, some important
differences emerge. In general, there is considerable support for H]. Table 6 shows that
transient institutional investors hold stocks that have experienced good recent excess
returns, have lower book to market ratios and a high residual variance. This suggests that
while they are positive feedback traders, they are also attracted to firms that are harder to

place within a valuation framework. The coefficient on the book to market may be

 

22 The use of corporate resources to pursue objectives that enhance (often non-monetary) individual utility
but have no beneficial spillover effects to outsiders. See Demsetz and Lehn (1985).

36

indicative of differences in investment styles. Higher book to market stocks or value
stocks are less “glamorous” than growth stocks. Such stocks may not fit the criteria
demanded by transients, who turn over their portfolios frequently and have to justify this
frequent activity to the clients on whose behalf they trade.

In Table 7, the quasi-indexer group is found to take larger positions in stocks that
have experienced poor returns in the prior 9 months. This is indicative of them being
negative feedback traders. At the other extreme, in Table 8, dedicated institutions are
completely indifferent to recent stock price performance. Given the long term focus of
these types of institutions, this result is not surprising. Since they are invested in their
target firms for the long haul, they are less inclined to respond to temporary fluctuations
in market performance. In terms of the information acquisition specifications, dedicated
institutions display a strong aversion to firms with high unsystematic risk as measured by
the variance of the residual. from the market model regression. However, quasi-indexers
avoid these firms to an even greater extent than dedicated institutions (the negative-
coefficient is more than twice as large in the F ama-Macbeth specification).

Modern portfolio theory implies that this unsystematic risk component should be
diversifiable and should not impact the decision on whether to invest in a particular stock.
The finding of a strong negative relationship therefore suggests that the dedicated and
quasi-indexer portfolios do not span the market. This corroborates our argument that
those institutions that are inclined to perform long term monitoring tend to avoid those
situations in which it is expected to be most costly. The fact that this aversion is stronger
in quasi-indexers is suggestive of a differing cost structure between the two groups.

Given that they hold more diversified portfolios than the dedicated class, their marginal

37

cost of monitoring is likely higher and thus they are affected by this consideration to a
greater extent. This contention is also borne out in the coefficients on the leverage ratio.
Leverage negatively affects quasi-indexer holdings, but not dedicated ownership. This is
to be expected if leverage is a form of subsidized monitoring that lessens the size of the
stake required for effective oversight.

Both transients and quasi-indexers are attracted to firms with high EBITA. While
EBITA is conditionally associated with the free cash flow problem, it is also indicative of
good fundamentals; therefore, it is not surprising that both groups find this to be a
desirable quality. It is interesting to contrast this finding with the earlier finding that
dedicated institutions are unconcerned with good historical market performance and that

quasi-indexers even appear to avoid it to some extent.

4.1.2 Testing the Herding and Informed Trade Hypothesis

With respect to the common control variables, the results displayed in Table 9 are

broadly consistent with Kelly (2005). Firm size and age are found to have a positive

assocratron wrth R . Grven these variables extensrve use as proxres for Information

asymmetry23 in corporate finance research, this lends credence to the interpretation of R2

as an inverse measure of firm opacity. On average, the four specifications explain over
32.4% of the variation in the log-transformed coefficient of determination.

In terms of the unique right hand side variables, the parsimonious Specification in
the first column of Table 9 suggests that changes in transient institutional ownership have

limited association with subsequent firm transparency. In the second and third columns, it

 

23 Himmelberg, Hubbard and Palia (1999)-

38

appears as though increases in quasi-indexer ownership negatively forecast transparency,
whilst dedicated accumulations predict higher future transparency. With the exception of
the dedicated class, these findings change in the full specification. The coefficient on the
change in transient ownership remains positive, but becomes strongly statistically
significant. Meanwhile the change in quasi-indexer variable loses its statistical
significance and reverses sign. The dedicated result is strongly persistent, retaining its
statistical significance, magnitude and direction. The differences between the full and
parsimonious results are suggestive of omitted variable bias.

With reference to H.ZB, column 4 of Table 9 provides evidence in support of the
spillover effect from information gathering. The information gathering activities of
dedicated investors enhance firm transparency as reflected by the improved fit of the
market model regression. In contrast, because quasi-indexers can not observe their target
firms with the same degree of scrutiny, their trading actions do not have an appreciable
marginal effect on the quality of information produced. The positive transient coefficient
is considerably smaller in magnitude than the dedicated. coefficient“. This is also
consistent with HZB in that the transient information gathering process ultimately
becomes available to the market at large, despite their preference for firm opacity. The
preference for opacity and the probable bias25 in the nature of the firm specific
information that transient investors collect also explains the smaller economic impact of
their actions when contrasted with the dedicated class.

The tests in section 4.1.1 reveal a positive association between lagged returns and

institutional holdings and also provide new insight into how investor heterogeneity

 

34 . . . . . . .
The effect of changes In dedicated ownership 15 over four times as large as changes In transrent.

7'5 . .

"' A greater focus on short term consrderatrons.

39

affects this relationship. Transient institutions prefer good recent performance, quasi-
indexers prefer poor recent performance and dedicated institutions and insiders are
indifferent to previous returns. Because of the summing up constraint with respect to
shares held, we could only examine one type of ownership at a time. Placing other
ownership classes as right hand side variables would lead to spurious regression resultszo.

In order to understand how the different types of ownership interact with each
other, we perform a series of linear Granger causality tests of changes in ownership on
lagged changes in ownership. We compute the change in holdings over the course of each
fiscal year for each class of investor. Notice that using first differences and their lags
causes us to drop the first two time periods (1992 and 1993). Tables 10 and 11 present
the results of performing this procedure for all institutions as a group and corporate
insiders. Tables 11 through 14 Show the results for decomposed institutional ownership.
For each of the three types of institution, two specifications of their change in demand

functions are considered; a parsimonious versron that relates their trades to insiders only

and a full specification in which the trades of all classes are included. The gains in
adjusted R2 in the full system suggest that this is the more appropriate one from which to

draw inference.

The first two columns of Table 11 indicate the difficulty in generating reliable
predictors of portfolio rebalancing by individual investors based on firm (accounting and
financial) characteristics. Only the excess returns variable is significant at conventional
levels. The dynamics with respect to other investors, Shown in columns 3 and 4 of Table

II, reveal a rich set of interactions however. Lagged purchases by both quasi-indexers

 

3" For instance. a firm with 99% dedicated ownership by definition has low ownership by all other classes.

40

and dedicated institutions forecast future sales by the management team. If the demand
functions are interpreted literally, this suggests that the arrival of longer horizon investors
produces a change in environment that may make it unfavorable or unnecessary for
managers to maintain large positions.

Two explanations are consistent with this result. A heavy build up of long term
block-holders could signal increased monitoring intensity. The transparency result with
respect to dedicated investors in Table 9, and the entrenchment index result with respect
to quasi-indexers in Table 7 corroborate this proposition. In the face of increased external
scrutiny, insiders become less able to manipulate corporate resources for personal gain.
The reduced scope for malfeasance would have the effect of increasing the importance of
portfolio diversification considerations for an insider who is over weighted in their own
company’s stock. One would therefore expect such an insider to reduce their holdings in
the firm. Alternatively, under the assumption of incentive alignment, it could be the case
that managers no longer need to hold as many shares (to signal their commitment to
outside shareholders) when monitoring institutions arrive.

H.2A proposes that the arrival of monitoring institutions has consequences for the
information asymmetry of a firm. Firms that attract the attention of dedicated institutions
become more transparent to the market. HI and by extension, H. 28, suggest that
transients Should avoid these situations. This is shown to be the case in Table 12. The
accumulation of dedicated blocks forecasts sales by transients in the following year. The
lack of statistical significance on the quasi-indexer variable combined with the result that

quasi-indexers do not increase firm transparency provides further support for H. 28.

41

All three types of institution tend to engage in trading reversals. In other words,
the coefficients on their own lagged changes are negative. We interpret this as meaning
that they all trade to achieve a certain outcome and once it has been achieved they
abandon the stock. In the case of transients, the explanation appears to be fairly
straightforward. Per the evidence presented thus far, they herd towards stocks with good
recent market performance (positive significant in both the level and change regressions)
but have opaque or imprecise information about their operating environment. Following
their arrival, these stocks continue to perform well, at which point they sell and repeat the
process elsewhere”.

In contrast, the case of dedicated and quasi-indexers reveals a different dynamic.
As discussed earlier, the anival of dedicated and subsequently, quasi-indexer investors
appears to reflect a change in operating environment. It is therefore possible that the
quasi-indexers use the dedicated class to identify stocks and act as the initial proponents
of increased monitoring. Once the dedicated class. has established itself, quasi-indexers
attempt to free-ride on the improvement generated. As such, the lagged dedicated change
variable enters positively in the lagged quasi-indexer regression. This is consistent with
the information cascade proposed earlier. Indeed, there is even a role for prudent man
considerations here. The presence of dedicated investors may serve as a form of
certification, thereby insulating quasi-indexers from negative repercussions in the event
of an unsuccessful investment decision. With respect to the other control variables, the
direction and statistical Significance remains the same in both the parsimonious and full

specifications.

 

27 With their exit being hastened by dedicated investor attention.

42

4.1.3 Testing the Ownership Structure - Firm Value Hypothesis

Hypothesis H.3 posits a role for transient institutions as the marginal price setter
in the market. In order to test this conjecture, we use the Brennan, Chordia and
Subrahmanyam (1998) method of performance attribution and present the results in
tables 15 through 17. By regressing monthly excess individual stock returns on both
contemporaneous and lagged institutional holdings, we find that the contemporaneous
measure is significant. However, this positive relationship is entirely driven by the effect
of transient institutions. Furthermore, the effect is particularly concentrated amongst
those transient institutions that pursue growth trading strategies. The insider ownership
regression lacks statistical significance at conventional levels, nonetheless, it is
interesting to note that insider ownershipiha’s a positive coefficient Whilst squared insider
ownership has a negative coefficient. The relevance of this finding will become more
apparent when we perform a test of the McConnell and Servaes (1990. 1995) model in
the next section.

The lack of significance in the lag ownership terms indicates that this relationship
does not represent an unambiguous trading rule through which one can earn superior
returns. Given that the highest frequency data available based on 13(t) filings is quarterly,
this is to be expected. However, the new finding here is that only transient institutions
matter in establishing price levels. Furthermore, transient institutions excel in situations
under which information is most imprecise i.e. those transient institutions that track

growth firms.

43

Turning to the firm value relationship, we use the McConnell and Servaes (1990,
1995, 2005) model because it is widely recognized in the performance — ownership
literature and has been extensively tested by researchers on both sides of the debate. The
first two papers in the McConnell and Servaes series use single year cross sections. The
third paper uses a first differencing panel data approach. We attempt to replicate the
McConnell and Servaes (1990, 1995) findings using all of the years in our sample in both
a Fama-Macbeth (1973) and panel data setup. For each year, regressions of Q on
management holdings, management holdings squared, institutional holdings, R&D,
advertising, leverage and firm size are performed. If the patterns first identified by
McConnell and Servaes (1990) are persistent, they should be detectable throughout the
sample years. The chosen panel data methods are both fixed effects and instrumental
variables. To instrument for ownership, we use the log of total assets, its square, beta
from the market model regression, its square, the residual variance from the market
model regression, the I/B/E/S reported analyst IforecaSt'error of EPS as of the end of the
previous fiscal year and its square.

Column 1 of Table 18 essentially replicates the findings of McConnell and
Servaes (1990, 1995). The coefficients on the insider holdings terms confirm the
characteristic inverted ‘U’ Shaped relationship between ownership structure and firm
value. However, when ownership is modeled as endogenous in column 2, this effect
disappears. Aggregate institutional ownership enters linearly in a positive and significant
way in both the Fama-Macbeth (1973) and fixed effects specifications. Research and
development expenditures and advertising are both positively associated with firm value.

The coefficient on leverage shows that firms with more debt have lower Q values. Higher

44

leverage may be an indicator of financial distress, the costs of which are reflected in
lower firm valuation. Finally, as expected, the entrenchment index score is shown to be a
negative predictor for Q indicating that firms with more manager friendly rules have
lower valuations.

Without controlling for governance, it appears as though the holdings of quasi-
indexers and transient institutions weakly offset each other (quasi-indexer significance at
the 10% level in the Fama-Macbeth (1973) regression), while the dedicated class has no
effect on Q. When we control for governance, the quasi-indexer variable loses statistical
significance, leaving only the transient coefficient positive. Furthermore, this offsetting
effect can only be found in a correlation sense (Fama-Macbeth (1973) regression). In
contrast, the dominance of the transient coefficient persists in all specifications. Upon
performing the instrumental variables procedure, Shown in column 1 of Table 19, the
overall pattern still obtains. Transient holdings enter the Q regression in a positive
fashion. Although the economic significance of sornecf the. ownership. coefficients
becomes inflated, the most egregious cases (dedicated and management ownership) are

statistically insignificant. The problems raised about the difficulties in finding good
instruments is reflected in the decline in the goodness of fit (R2 falls to around 7%).

However, the consistency of the instrumental variables approach with both the Fama-
Macbeth (1973) and fixed effects regressions is tentatively suggestive of causality.

A natural question at this point is whether the results presented are partially a
function of investment style. If transient institutions are attracted to lower book to market
stocks, one might expect to see their ownership being mechanically associated with

higher Q’s. To address this issue, Table 19 also contains a regression in which transient

45

ownership is subdivided by preferences for growth vs. value characteristics. In this case a
“success” is defined as a positive coefficient on either of the value categories. We find
that the presence of transients following a large value investment strategy is also

associated with higher Q’s.

4.1.4 Conditional Firm Value Tests

As informative as these results are, they are not a sufficient basis on which to
completely dismiss the role of the dedicated and quasi-indexer classes. We therefore
consider the conditional aspects of the relationship between Q and our measures of
ownership. Three conditional variations of the Q regression are considered. Firstly, how
the relationship might vary depending on the strength of the govemance environment.
Secondly, how the relationship may be affected by the general expectations of the
market, and finally, how the relationship may vary according-to the volatility of cash
flows. '

To implement the first test, the sample is divided in half according to the level of
the entrenchment index score and each firm’s grouping is interacted with the level of its
institutional ownership variables. For the second conditional relationship, the institutional
ownership variables are interacted with membership in the top half by analyst forecast
dispersion. Finally, to assess differential effects of cash flow volatility, the ownership
variables are interacted with membership in the top half of the sample by standard
deviation of EBITA over the previous five years. Table 20 presents results. In all three

conditional specifications, the level of transient holdings remains significantly positive.

46

The only specification in which the interaction terms are significant is the cash
flow one. The effect of transients is incrementally higher when cash flow volatility is
high. We view these results as further support for the importance of transient institutions
as the marginal price setter in the economy. Interestingly, the dedicated interaction is also
positive. Nonetheless, the overall picture remains consistent. The level of dedicated and
quasi-indexer ownership is not associated with firm value in a systematic fashion. Even
when we allow ownership to have differential effects based on deleterious corporate
governance provisions, those institutions that we typically think of as monitors, have little

predictable effect on firm value.

5.1 Summary and Conclusions

The major innovation in this paper lies in the way that we are able to analyze the
importance of multiple ownership structure configurations. We use more comprehensive
data than previous studies in this area in terms of the number of years and the types of
ownership considered. By taking into account differences in demand for equity
characteristics, we feel that we approach something closer to the reality of how securities
markets operate. We find that transient institutional investors display a preference for
firms that have good market adjusted performance and imprecise information concerning
their expected future cash flows. On the other hand, dedicated and quasi-indexer
institutions tend to avoid stocks with high firm specific risk and imprecise information.
This suggests that these two classes of investor are more concerned with firm
characteristics that facilitate monitoring. We also consider the role of insiders and find

that they exhibit some attraction to features that obscure managerial behavior. For

47

instance, they express a strong preference for discretionary expenditures in the form of
advertising.

Having isolated the different demand specifications, we then investigate whether
they have any bearing on firm value. We demonstrate that the presence of transient
institutional investors is strongly contemporaneously associated with higher returns and
as a consequence, with higher subsequent firm values (as measured by Tobin’s Q).
Although we do find evidence that suggests that the arrival of dedicated institutions
enhances firm transparency, neither of the two types of institutions with longer
investment horizons (dedicated and quasi-indexer) are systematically associated with
either returns or firm value. The pervasive result of this essay is that transient institutions
have some inherent advantage over other classes of investors in terms of locating
profitable investment opportunities.

We therefore conduct a series of conditional tests to determine if transient
preferences with regards to information precision facilitate this advantage. We confirm
that their preference for conditions under which firm specific information is imprecise or
hard to process has a relationship with firm value. In particular, transient institutional
ownership is positively associated with firm value when cash flow volatility is high and
when they pursue an investment strategy that favors stocks with high background noise
about future earnings. This provides further support for the contention that transient
institutions possess superior ability in terms of collecting and processing the information
about the firms in which they invest.

This result is particularly interesting because of the limited significance of long

term monitoring institutions (dedicated and quasi-indexers) with respect to firm value,

48

even under the conditional specifications that explicitly include governance. This
suggests that the conventional wisdom about the role of monitoring by outside investors
needs extension. The activity of transient institutions should also have some influence on
the nature of contracts formed between managers and outside investors. We feel that this
represents a compelling area for further investigation. The patterns we identify are robust
to both a fixed effects and an instrumental variables technique that models ownership as
endogenous. In addition to confirming the strength of our findings, we also take this as a

cue to further examine the role of causality.

49

APPENDIX
TABULATED RESULTS

50

Table 1
Summary Statistics

Table 1 contains descriptive statistics of the key variables used to test our hypotheses. Tobin’s Q is
measured as the market value of equity (Compustat item 199 x Compustat item 25) plus the book value of
assets (Compustat item 6) minus the book value of common equity including deferred taxes (Compustat
items 60 and 74) all divided by the book value of assets. BETA and RESIDVAR are the beta and residual
variance estimates from a market model regression with a terminal date one year before the current fiscal
year. For inclusion in this estimation procedure, we require that a firm has at least 110 days of returns data
on CRSP prior to the terminal date. MH%, TRA%, QIX% and DED% refer to the fraction of outstanding
shares held by management, transient, quasi-indexer and dedicated institutional shareholders respectively at
the end of the fiscal year. LEVERAGE is the ratio of long term debt plus short term debt to total assets at
the end of the fiscal year. B/M is the book value of common equity divided by market value of common
equity at the end of the fiscal year. RD and ADV refer to fiscal year end research and development and
advertising expense scaled by total assets respectively. Firms which do not report R&D or advertising in
Compustat are assigned a value of 0. F -DISP is the I/B/E/S reported standard deviation of analyst forecast
estimates of the current fiscal year’s earnings per share (EPS). F-ERROR is the difference between the
I/B/E/S calculated mean estimate of EPS and the realized value of EPS. When there are multiple
observations of both of these variables for a given fiscal year, we use the one taken closest to the end of the
fiscal year. CASHVOL is the standard deviation of the previous 5 years earnings before interest and taxes.
E-INDEX is the fiscal year end entrenchment index score from Bebchuk, Cohen and Ferrell (2004). For
years in which no entrenchment index score was reported, we use the previous year’s value. The index
ranges from 0 to 6 with one point being added for each governance provision that favors management.

51

Table 1

 

Panel A (1992 - 1997

 

 

Year 1992 1993 1994 1995 1996 1997
Q Mean 1.865 2.039 1.896 1.981 1.995 2.014
Std dev 0.930 0.918 0.885 0.887 0.946 0.910
N 95 180 198 193 182 164
Market Mean 2027 1802 1871 1724 2285 2793
capitalization
($ Millions) Std dev 5665 4743 5060 2751 4127 4283
N 95 180 198 193 182 164
BETA Mean 1.330 1.307 1.276 1.299 1.353 1.121
Std dev 0.578 0.724 0.769 0.616 0.879 0.561
N 95 180 198 193 182 164
RESIDVAR Mean 0.022 0.022 0.021 0.021 0.021 0.022
Std dev 0.007 0.007 0.008 0.007 0.007 0.085
N 95 180 198 193 182 164
MH% Mean 6.84 7.15 6.49 6.84 6.09 6.05
Std dev 9.62 10.39 9.95 8.87 8.32 7.98
N 95 180 198 193 182 164
TRA% Mean 9.41 10.44 8.95 11.88 10.48 11.48
Std dev 6.23 8.41 7.54 9.57 7.56 7.59
N 95 180 198 193 182 164
QIX% Mean 33.18 31.06 29.68 27.52 26.05 26.92
Std dev 10.44 10.60 9.35 8.74 8.69 8.47
N 95 180 198 193 182 164
DED% Mean 12.72 13.69 17.91 17.70 18.95 19.85
Std dev 8.99 8.61 8.83 9.44 9.26 9.78
N 95 180 198 193 182 164
Total assets Mean 1670 1231 1491 1422 1932 2328
($ Millions) Std dev 2414 1986 2266 2019 4124 4183
N 95 180 198 193 182 164
LEVERAGE Mean 0.209 0.189 0.207 0.193 0.219 0.224
Std dev 0.169 0.175 0.171 0.165 0.192 0.165
N 95 180 198 193 180 161
B/M Mean 0.521 0.450 0.496 0.468 0.474 0.437
Std dev 0.286 0.212 0.252 0.242 0.256 0.213
N 95 179 198 193 180 163
R&D Mean 0.034 0.038 0.036 0.037 0.040 0.034
Std dev 0.052 0.057 0.060 0.055 0.062 0.057
N 90 167 198 175 168 149

 

 

52

Table 1 (cont’d)

 

 

 

 

 

 

Year 1992 1993 1994 1995 1996 1997
ADV Mean 0.021 0.024 0.014 0.009 0.011 0.008
Std dev 0.034 0.051 0.037 0.044 0.049 0.023
N 51 115 170 187 178 163
F-DISP Mean 0.040 0.049 0.050 0.053 0.053 0.043
Std dev 0.044 0.073 0.079 0.1 12 0.108 0.056
N 95 180 198 193 182 164
F-ERROR Mean 0.083 0.068 0.068 0.064 0.099 0.073
Std dev 0.134 0.138 0.123 0.143 0.344 0.173
N 95 180 198 193 182 164
CASHVOL Mean 0.033 0.037 0.033 0.035 0.037 0.034
Std dev 0.026 0.032 0.027 0.029 0.037 0.028
N 94 176 195 182 176 162
E-INDEX Mean 2.12 2.19 2.29 2.19 2.23 2.19
Std dev 1.43 1.47 1.44 1.33 1.35 1.39
N 95 180 198 193 182 164
Panel B (1998 - 2003
Year 1998 1999 2000 2001 2002 2003
Q Mean 2.098 2.01 2.03 2.15 1.91 2.04
Std dev 1.147 1.22 1.18 1.19 1.05 0.93
N 261 253 242 214 308 319
Market Mean 2450 2967 2678 3074 3395 3764
capitalization ‘
($ Millions) Std dev 5869 7888 4662 4754 17133 15972
N 261 253 242 214 308 319
BETA Mean 1.377 1.250 1.294 1.01 1.55 1.74
Std dev 0.675 0.479 0.685 0.873 1.01 0.84
N 261 253 242 214 308 319
RESIDVAR Mean 0.024 0.028 0.031 0.037 0.035 0.031
Std dev 0.009 0.009 0.010 0.013 0.014 0.013
N 261 253 242 214 308 319
MH% Mean 6.77 6.68 6.08 4.89 4.99 4.31
Std dev 9.91 9.51 8.99 7.84 8.46 6.62
N 261 253 242 214 308 319
TRA% Mean 17.14 20.94 18.86 36.41 33.82 23.80
Std dev 8.89 9.84 10.35 13.68 12.41 9.13
N 261 253 242 214 308 319
QIX% Mean 40.20 31.11 31.93 22.74 26.18 38.43
Std dev 13.04 10.55 10.95 6.01 7.56 9.85
N 261 253 242 214 308 319

 

 

53

Table l (cont’d)

 

 

Year 1998 1999 2000 2001 2002 2003
DED% Mean 1.08 8.98 9.41 10.70 1 1.61 10.44
Std dev 3.32 8.10 8.08 7.91 8.11 7.39
N 261 253 242 214 308 319
Total assets Mean 1733 2284 2172 2548 2236 2588
($ Millions) Std dev 3433 5188 4972 3743 5627 6022
N 261 253 242 214 308 319
LEVERAGE Mean 0.249 0.567 0.240 0.241 0.201 0.197
Std dev 0.206 0.192 0.178 0.180 0.181 0.167
N 258 252 241 212 307 318
B/M Mean 0.465 0.519 0.517 0.472 0.535 0.450
Std dev 0.269 0.304 0.299 0.278 0.292 0.226
N 254 247 238 212 305 318
R&D Mean 0.051 0.031 0.034 0.040 0.044 0.039
Std dev 0.093 0.052 0.057 0.067 0.063 0.053
N 244 239 229 204 303 314
ADV Mean 0.090 0.009 0.010 0.008 0.008 0.009
Std dev 0.023 0.029 0.031 0.020 0.019 0.025
N 258 251 241 213 306 312
F-DISP Mean 0.044 0.036 0.036 0.040 0.030 0.026
Std dev 0.086 0.062 0.053 0.071 0.063 0.040
N 261 253 242 214 308 319
F-ERROR Mean 0.099 0.103 0.090 0.106 0.063 0.059
Std dev 0.226 0.458 0.172 0.282 0.151 0.132
N 261 253 242 212 307 318
CASHVOL Mean 0.044 0.042 0.040 0.047 0.058 0.050
Std dev 0.048 0.043 0.034 0.045 0.055 0.048
N 238 239 230 211 287 306
E-INDEX Mean 2.069 2.21 2.29 2.34 2.30 2.41
Std dev 1.322 1.32 1.29 1.26 1.26 1.28
N 261 253 242 214 308 319

 

 

54

Table 2
Investment Objective and Investment Style

Table 2 shows the distribution of investment Styles according to institutional investor type. DED, QIX and
TRA refer to the number of institutions classified as dedicated, quasi-indexer and transient on the Thomson
Financial 13(f) database for the years 1992-2003. LGR, LVA, SGR and SVA refer to the investment styles
large growth, large value, small growth and small value determined by the methodology of Abarbanell,
Bushee and Raedy (2003). The percentages in parentheses show the relative proportion of investors
following a given investment Style within a given class of institutional shareholders each year.

55

Table 2

 

 

Year Type LGR LVA SGR SVA
1992 DED 11 (10.2%) 10 (9.3%) 63 (58.3%) 24 (22.22%)
QIX 174 (23.80%) 283 (38.71%) 153 (20.93%) 121 (16.55%)
TRA 29 (17.16%) 29 (17.16%) 104 (61.54%) 36 (21.30%)
1993 DED 11 (9.6%) 12 (10.5%) 66 (57.9%) 25 (21.93%)
QIX 165 (21.94%) 291 (38.69%) 175 (23.27%) 121 (16.09%)
TRA 46 (21.30%) 31 (14.35%) 109 (50.46%) 30 (13.89%)
1994 DED 8 (6.2%) 12 (9.2%) 77 (59.2%) 33 (25.39%)
QIX 183 (23.17%) 305 (38.61%) 179 (22.65%) 123 (15.57%)
TRA 45 (23.68%) 27 (14.21%) 96 (50.53%) 22 (11.58%)
1995 DED 8 (6.3%) 12 (9.4%) 75 (59.1%) 32 (25.20%)
QIX 189 (23.95%) 310 (39.29%) 149 (18.89%) 141 (17.87%)
TRA 57 (22.35%) 39 (15.29%) 117 (45.88%) 42 (16.47%)
1996 DED 10 (7.5%) 8 (6.0%) 76 (57.1%) 39 (29.32%)
QIX 211 (25.18%) 316 (37.71%) 176 (21.00%) 135 (16.11%)
TRA 43 (16.41%) 53 (20.23%) 122 (46.56%) 44 (16.79%)
1997 DED 13 (8.3%) 7 (4.5%) 96 (60.9%) 41 (26.28%)
QIX 240 (27.03%) 319 (35.92%) 165 (18.58%) 164 (18.47%)
TRA 71 (24.83%) 41 (14.33%) 126 (44.06%) 48 (16.78%)
1998 DED O (0.0%) 0 (0.0%) 6 (26.1%) 17 (73.91%)
QIX 284 (36.59%) 214 (27.58%) 97 (12.50%) 181 (23.33%)
TRA 89 (16.89%) 102 (19.36%) 159 (30.17%) 177 (33.58%)
1999 DED 1 (1.8%) 2 (3.6%)? g 16 (29.1%) 36 (65.46%)
QIX 335 (38.55%) 267 (30.73%) 81 (9.32%) 186 (21.40%)
TRA 121 (19.45%) 91 (14.63%) 197 (31.67%) 213 (34.24%)
2000 DED 2 (3.1%) 4 (6.3%) 16 (25.0%) 42 (65.63%)
QIX 344 (38.78%) 275 (31.00%) 104 (11.73%) 164 (18.49%)
TRA 114 (15.81%) 148 (20.53%) 232 (32.18%) 227 (31.48%)
2001 DED 3 (5.4%) 1 (1.8%) 16 (28.6%) 36 (64.29%)
QIX 309 (45.71%) 216 (31.95%) 45 (6.66%) 106 (15.68%)
TRA 188 (18.65%) 218 (21.63%) 276 (27.38%) 326 (32.34%)
2002 DED 6 (9.2%) 3 (4.6%) 25 (38.4%) 31 (47.69%)
QIX 337 (41.50%) 253 (31.16% 82 (10.10%) 140 (17.24%)
TRA 144 (17.14%) 169 (20.12% 253 (30.12% 274 (32.62%)
2003 DED 6 (11.1%) 3 (5.6%) 20 (37.0%) 25 (46.30%)
QIX 369 (39.81%) 303 (32.69%) 75 (8.09%) 180 (19.42%)
TRA 119 (18.36%) 106 (16.36% 202 (31.17%) 221 (34.11%)

 

 

56

Table 3
Correlation Coefficients

Table 3 reports the Pearson correlation coefficients of the key variables used in this study. Values in
boldface indicate statistical significance at the 5% level or lower. MH%, TRA%, QIX% and DED% refer to
the fraction of outstanding shares held by management, transient, quasi-indexer and dedicated institutional
shareholders respectively at the end of the fiscal year. ASSETS references total assets measured at the end
of the current fiscal year. B/M is the book value of common equity divided by market value of common
equity at the end of the current fiscal year. BETA and RESIDVAR are the beta and residual variance
estimates from a market model regression with a terminal date one year before the current fiscal year. For
inclusion in this estimation procedure, we require that a firm has at least 110 days of returns data on CRSP
prior to the terminal date. EXRET3-12is the cumulative excess return relative to the CRSP value weighted
index over the previous 12 months. LEVERAGE is the ratio of long term debt plus short term debt to total
assets at the end of the fiscal year. RD, ADV, EBITA and CAPEX refer to fiscal year end research and
development expense, advertising expense, earnings before interest and taxes and capital expenditures
scaled by total assets respectively. Firms which do not report R&D or advertising in Compustat are
assigned a value of 0. CASHVOL is the standard deviation of the previous 5 years earnings before interest
and taxes. F-DISP is the I/B/E/S reported standard deviation of analyst forecast estimates of the current
fiscal year’s earnings per share (EPS). F-ERROR is the difference between the I/B/E/S calculated mean
estimate of EPS and the realized value of EPS. When there are multiple observations of both of these
variables for a given fiscal year, we use the one taken closest to the end of the fiscal year. E-INDEX is the
fiscal year end entrenchment index score from Bebchuk, Cohen and Ferrell (2004). For years in which no
entrenchment index score was reported, we use the previous year’s value. The index ranges from 0 to 6
with one point being added for each governance provision that favors management.

57

Table 3

 

 

 

Panel A

Variable TRA% DED% QIX% MH% ASSETS B/M
TRA% 1.00

DED% -0.16 1.00

QIX% -0.05 -0.18 1.00

MH% -0.19 —0.13 -0.21 1.00

ASSETS 0.04 0.06 0.12 -0.11 1.00

B/M -0.14 0.01 0.03 0.02 0.05 1.00
BETA 0.16 -0.03 -0.11 -0.14 -0.14 -0.10
EXRETIZ 0.14 0.03 0.00 -0.00 0.04 -0.33
LEVERAGE -0.04 -0.01 -0.01 -0.04 0.32 0.20
RESIDVAR 0.42 -0.16 -0.24 -0.04 -0.23 -0.05
R&D 0.09 -0.03 -0.13 -0.16 -0.30 -0.18
ADV -0.05 0.04 -0.01 0.02 -0.02 -0.06
EBITA «0.03 0.06 0.13 . 0.09 0.00 -0.36
CASH V01. 0.19 -0.07 -0.16 -0.07 41.26 -0.08
CAPEX -0.08 -0.04 -0.04 0.09 -0.09 0.03
F-DISP -0.12 0.06 -0.03 -0.05 0.00 0.22
F-ERROR -0.04 -0.02 -0.06 -0.05 -0.07 0.14
E-INDEX -0.00 0.00 0.15 -0.22 0.09 0.17

 

 

58

Table 3 (cont’d)

 

 

 

 

 

 

5:23;; BETA EXRET] 2 LEVERAGE RESIDVAR R&D ADV
BETA 1.00

EXRET12 -0.03 1.00

LEVERAGE -0.16 -0.03 1.00

RESIDVAR 0.63 0.01 -0.09 1.00

R&D 0.46 -0.06 -0.27 0.43 1.00

ADV -0.03 -0.04 -0.05 -0.04 0.01 1.00
EBITA -0.28 0.13 -0.17 -0.37 -0.33 0.04
CASHVOL 0.41 -0.07 -0.03 0.49 0.47 0.07
CAPEX -0.06 -0.04 0.05 -0.09 -0.09 0.02
F—DISP -0.04 -0.11 0.16 -0.02 0.04 0.04
F-ERROR 0.01 -0.10 0.09 0.06 0.09 0.00
E-INDEX -0.18 -0.02 0.12 -0.16 -0.13 0.03
Panel C . .
V ariable EBITA CASHVOL CAPEX ' F-DISP F-ERROR E-IN DEX .
EBITA , 1 .00 ‘
CASHVOL -0.42 1.00

CAPEX 0.04 0.02 1 .00

F-DISP -0.29 0.19 0.06 1.00

F-ERROR -0.25 0.1 8 0.00 0.47 1.00

E-INDEX -0.02 -0.13 -0.08 0.03 0.03 1.00

 

 

59

Table 4
Aggregate Institutional Investor Preferences

Table 4 presents the results of a sequence of Fama-Macbeth (1973) and firm level fixed effects regressions
in which the dependent variable is the fraction of shares outstanding held by institutional investors (IH%).
F-M denotes Fama-Macbeth (1973) and FE denotes fixed effects coefficients. The F-M estimates are the
time series average coefficients for 12 cross sectional regressions for each year from 1992-2003. Fama-
Macbeth (1973) t-statistics are reported in parentheses. The fixed effects regressions include intercept terms
and year dummies but these are not reported. Fixed effects standard errors are reported in parentheses. For
both methods, we proceed stepwise, adding variables that proxy for the following three groups of
considerations: the relationship between risk and returns, the quality of information surrounding the firm’s
operations and the ease with which the firm can be monitored by outside investors. Firm size is measured
as the log of total assets at the end of the current fiscal year. Log(B/M) is the log of the book value of
common equity divided by market value of common equity at the end of the current fiscal year. BETA and
RESIDVAR are the beta and residual variance estimates from a market model regression with a terminal
date one year before the current fiscal year. For inclusion in this estimation procedure, we require that a
firm has at least 110 days of returns data on CRSP prior to the terminal date. EXRET3-12 is a security’s
cumulative excess return relative to the CRSP equal weighted index over the 9 months beginning 12
months prior to the measurement of the ownership variable and ending at the beginning of the current
(ownership measurement) quarter. LEVERAGE is the ratio of long term debt plus short term debt to total
assets. RD, ADV, EBITA and CAPEX refer to research and development expense, advertising expense,
earnings before interest and taxes and capital expenditures scaled by total assets respectively. E-INDEX is
the entrenchment index score taken from Bebchuk, Cohen and Ferrell (2004). The index ranges from 0 to 6
with one point being added for each governance provision that favors management. With the exception of
BETA and RESIDVAR (which are lagged by one fiscal year), all control variables are measured as of the
end of the current fiscal year. ***, ** and * refer to statistical significance at the 1%, 5% and 10% levels
respectively.

60

Table 4

 

 

 

 

 

Dependent Variable IH% IH% IH% IH% IH% IH%
Regression Type F-M FE F—M FE F -M FE
Risk-Return
Log(Assets) 0.015*** 0023*" 0.018*** 0.027*** 0.018** 0.028***
(4.84) (0.003) (3.96) (0.003) (3.66) (0.004)
Log(BM) —0.007 -0.012** 0.021 0.003 0.0182 -0.003
(-1.l8) (0.006) (1.40) (0.007) (1.11) (0.007)
BETA 0.017* 0016*" 0027* —0.007 0.029** -0.006
(1.48) (0.004) (1.93) (0.006) (2.10) (0.006)
EXRET3-12 0.059*** 0037*" 0.069*** 0.042*** 0.069*** 0.041 ***
(4.94) (0.008) (4.12) (0.009) (3.94) (0.008)
Information Asymmetry
R&D 0.232*** -0.014 0.215*** 0.012
(3.60) (0.07) (3.55) (0.074)
ADV 0.048 -0.019 0.048 -0.070
(0.35) (0.126) (0.21) (0.126)
EBITA 0355*” 0202*“ 0.359” 0208’“M
(3.45) (0.046) (3.52) (0.047)
("APEX -0.137 -0.219*** 0.111“ ' -0.l89**"‘
(-3.04) (0.006) {-2.20} (0.059)
RESIDVAR -1.748 2.715*** -l.687 2.862 *** -
(-l .62) (0.412) (-1.35) (0.411)
Monitoring
LEVERAGE -0.003 -0.049**
(-0.55) (0.024)
E-INDEX 0.008” 0.014***
(3.22) (0.003)
R2(%) 8.93 4.07 17.43 6.84 19.18 8.16

 

Table 5
Corporate Insider Preferences

Table 5 presents the results of a sequence of Fama-Macbeth (1973) and firm level fixed effects regressions
in which the dependent variable is the fraction of shares outstanding held by management (MH%). F-M
denotes Fama-Macbeth (1973) and FE denotes fixed effects type regressions. Fama-Macbeth (1973) t-
statistics and fixed effects standard errors are reported in parentheses. The independent variable definitions
are consistent with Table 4.

 

 

 

 

 

 

Dependent Variable MH% MH% MH% MH% MH% MH%
Regression Type F-M FE F-M FE F-M FE
Risk-Return
Log(Assets) -0.010*** -0.010*** -0.015*** —0.014*** -0.014*** -0.011***
(—9.21) (0.002) (-6.59) (0.002) (-4.69) (0.002)
Log(BM) 0.002 0.003 -0003 0.002 0.004 0008*
(0.23) (0.003) (-0.83) (0.003) (0.28) (0.003)
BETA -0.005* -0.007*** 0.000 0.004 -0.002 0.001
(-224) (0.002) (0.21) (0.003) (-0.87) (0.003)
EXRET3-12 0.003 0.002 -0.000 -0.001 0.003 -0.001
(0.28) (0.004) (—0.67) (0.004) (-027) (0.004)
Information Asymmetry
R&D -0.372*** -0.280*** -0.386*** -0.307***
I {-7.36) (0.036) (-742) (0.036)
I
Anv ’ 0.261 0.1529** 0.298“ 0.170***
(1.68) (0.063) (2.01) (0.036)
EBITA 0.008 0.034 -0.000 0.012
(0.44) (0.023) (0.33) (0.023)
CAPEX 0.056 0.120*** 0.011 0.083***
(0.96) (0.029) (0.16) (0.029)
Monitoring
RESIDVAR 1.109” 0.013 0.497" -0.132
(2.68) (0.206) (2.19) (0.200)
LEVERAGE -0.014 -0.022*
(-1.36) (0.011)
E-INDEX -0.016*** -0.016***
(-11.66) (0.001)
th‘Vo) 3.78 2.54 14.44 10.90 21.02 12.68

 

62

Table 6
Equity Demand Functions: Transient Institutional Investors

Table 6 presents the results of a sequence of Fama-Macbeth (1973) and firm level fixed effects regressions
in which the dependent variable is the fraction of shares outstanding held by transient institutional investors
(TRA%). F-M denotes Fama-Macbeth (1973) and FE denotes fixed effects type regressions. Fama-
Macbeth (1973) t-statistics and fixed effects standard errors are reported in parentheses. The independent
variable definitions are consistent with Table 4.

 

 

 

 

 

 

Dependent Variable TRA% TRA% TRA% TRA% TRA% TRA%
Regression Type F -M FE F -M FE F -M FE
Risk-Return
Log(Assets) -0.003 0.007*** 0.003 0.019*** 0.003 0.018***
(-0.81) (0.002) (0.86) (0.002) (0.77) (0.003)
Log(BM) -0.014*** -0.019*** -0.003** -0.010* -0.005** -0.013**
(-3.61) (0.004) (-2.67) (0.005) (-2.88) (0.005)
BETA 0.032*** 0.028*** 0.019“ -0.037*** 0.021*** -0.036***
(7.16) (0.003) (3.86) (0.004) (4.10) (0.004)
EXRET3-12 0.063*** 0.065*** 0.063*** 0.027*** 0.062*** 0.026***
(7.63) (0.006) (8.30) (0.006) (8.22) (0.006)
l_n_[ormation Asvmmrj 11
R&D 0.083“ -0.060 0089* -0.052
(2.11) (0.049) (1.85) (0.051)
ADV -0.094 -0.284*** -0.095 -0.310**
(-0.44) (0.087) (-0.54) (0.087)
EBITA 0140*" 0135*" 0.149*** 0.139***
(4.66) (0.032) (4.16) (0.033)
CAPEX 0.019 -0.078** 0.034 -0.060
(0.69) (0.041) (1.39) (0.041)
RESIDVAR 1.126** 6895*“ 1233*“ 6982*“
(2.47) (0.285) (2.77) (0.285)
Monitoring
LEVERAGE 0.016 -0.033**
(0.06) (0.016)
E-INDEX 0.002** 0.008***
(2.54) (0.002)
R208) 25.27 8.31 31.79 22.41 32.82 26.55

 

Equity Demand Functions: Quasi-Indexer Institutional Investors

Table 7

Table 7 presents the results of a sequence of Fama-Macbeth (1973) and firm level fixed effects regressions
in which the dependent variable is the fraction of shares outstanding held by quasi-indexer institutional
investors (QIX%). F-M denotes Fama-Macbeth (1973) and FE denotes fixed effects type regressions.
Farm-Macbeth (1973) t-statistics and fixed effects standard errors are reported in parentheses. The
independent variable definitions are consistent with Table 4.

 

 

 

 

Dependent Variable QIX% QIX% QIX% QIX% QIX% QIX%
Regression Type F -M FE F-M FE F—M FE
Risk-Return
Log(Assets) 0.012*** 0.011*** 0.013** 0.005" 0.011*** 0.006***
(3.86) (0.002) (3.92) (0.002) (4. 12) (0.002)
Log(BM) -0.001 0.004 0.017** 0.009* 0.016“ 0.006
(0.13) (0.004) (4.08) (0.004) (3.47) (0.004)
BETA -0.022*** -0.013*** -0.002 0014*“ -0.002 0014*"
(-5.28) (0.003) (-0.56) (0.003) (-0.60) (0.004)
EXRET3-12 -0.012* -0.015*** -0.005 -0.013** -0.010 -0.013**
(-1.75) (0.005) (-1.07) (0.006) (-1.18) (0.006)
Information Asymmfir
R&D 0.112** -0.033 0.058 -0.035
(2.21) (0.046) (1.53) (0.047)
ADV -0.016 0.036. -0.050 0.078
(0082) (0.079) (-1.51) (0.079)
EBITA 0209* 0.067** 0.190** 0.066“
(3.00) (0.029) (2.88) (0.029)
CAPEX —0.044** -0.111*** -0.014 -0.088**
(-2.40) (0.037) (-1.35) (0.037)
RESIDVAR -2.371*** -2.275*** -2.022*** -2.123***
{-3.86) (0.260) (-3.61) (0.259)
Monitoring
LEVERAGE -0.055*** —0.032**
(-3.79) (0.015)
E-IN'DEX 0.008*** 0.011***
(4.79) (0.002)
R2(%) 8.53 2.58 18.31 4.58 21.23 9.20

 

 

Equity Demand Functions: Dedicated Institutional Investors

Table 8

Table 8 presents the results of a sequence of Fania-Macbeth (1973) and firm level fixed effects regressions
in which the dependent variable is the fraction of shares outstanding held by dedicated institutional
investors (DED%). F-M denotes Fame-Macbeth (1973) and FE denotes fixed effects type regressions.
Fame-Macbeth (1973) t-statistics and fixed effects standard errors are reported in parentheses. The
independent variable definitions are consistent with Table 4.

 

 

 

 

Dependent Variable DED% DED% DED% DED% DED% DED%
Rggression Typ_e F-M FE F-M FE F-M FE
Risk-Return
Log(Assets) 0006*" 0005*" 0.006*** 0.004" 0.004 0.004M
(3.27) (0.002) (2.78) (0.002) (0.004) (0.002)
Log(BM) 0.006 0.002 0.005 0.002 0.005 0.003
(0.90) (0.003) (0.63) (0.004) (0.61) (0.004)
BETA 0.006 -0.003 0.012“ 0.011*** 0.012 0014*”
(1.05) (0.002) (1.65) (0.003) (1.67) (0.003)
EXRET3-12 0.005 —0.014 0.009 0.002 0.011 0.007
(0.50) (0.005) (0.81) (0.005) (0.76) (0.005)
lntormatio_n_A.!')'mmey‘y
R&D 0.075 0.109*** 0.083” 0.117***
(2.88) (0.041) (2.65) (0.042)
ADV 0175* 0.094 0203* 0.100
(2.10) (0.07) (2.10) (0.072)
EBITA —0.003 0.034 0.009 0.034
(-0.21) (0.026) (0.13) (0.027)
CAPEX -0.048 —0.031 -0.066* 0.053
(-1.76) (0.033) (-1.99) (0.003)
RESIDVAR -0.713* -1.663*** -0.866* -1.714*"‘*
(-1.79) (0.234) (—1.89) (0.234)
Monitoring
LEVERAGE 0.030 -0.000
(1.67) (0.001)
E-INDEX —0.003"‘ -0.002
(-1.76) (0.002)
mm 5.99 1.00 12.31 3.83 14.73 3.97

 

 

65

Table 9
Changes in Institutional Ownership and Firm Transparency

R2 are obtained from annual market model regressions augmented with an industry factor. Daily individual
stock returns are regressed against the CRSP value weighted index and the return on a stock’s 2-digit SIC

industry, excluding the stock itself. As R2 is bounded between 0 and 1, we apply the logistic transformation

of Morck, Yeung and Yu (2000). For each year, the transformed R statistic is regressed on the prior year’s
change in institutional ownership, the market value equity, the number of analysts providing annual EPS
estimates in I/B/E/S, Amihud’s 2002 illiquidity measure and the number of years since initial listing in
CRSP in a pooled time-series cross-section setup. ATRA%, AQIX%, ADED% refer to changes in holdings
by transient, quasi-indexer and dedicated institutional investors respectively. ILLIQ is multiplied by
1,000,000 as per Hasbrouck’s (2005) formulation.

 

 

Dependent Variable Logistic R2 Ifiisfic R2 Logistic R2 Logistic R2
Lag ATRA% 0.137 0.382***
(0.131) (0.141)
Lag AQ1X% -0.652*** 0.151
(0.142) (0169)
Lag ADED% 1.587*** 1728*"
(0.158) (0.183)
SIZE 0649*” 0.657*** O.651*** 0.844***
(0.017) (0.018) (0.017) (0.017)
NUMEST —0.004 -0.006 -0005 -0003
(0.003) (0.004) (0.003) (0.003)
ILLIQ 0.000 0.000 0.000 0.000
(0.000) (0.000: (0.000) (0.000)
AGE 0.002** 0002* 0.003" 0.003**
(0.001) (0.001) (0.001) (0.001)
R2(%) 31.72 31.95 32.87 32.98

 

 

66

Table 10
Aggregate Changes in Institutional Ownership

Table 10 presents the results of a sequence of Fame-Macbeth (1973) and firm level fixed effects
regressions in which the dependent variable is the annual change in the fraction of shares outstanding held
by institutional investors in the aggregate (MH%). To investigate the dynamics of trade, changes in
institutional ownership are related to their own one year lag and the one year lag of changes in insider
ownership (Lag AMH%). F-M denotes Fania-Macbeth (1973) and FE denotes fixed effects type
regressions. The F-VI estimates are the time series average coefficients for 10 cross sectional regressions
for each year from 1994-2003 (with the years 1992 and 1993 being dropped because of the lag structure of
the specification). Fama-Macbeth (1973) t-‘statistics are reported in parentheses. The fixed effects
regressions include intercept terms and year dummies but these are not reported. Fixed effects standard
errors are reported in parentheses. All control variables are measured as of the end of the previous fiscal
year. ***, ** and * refer to statistical significance at the 1%, 5% and 10% levels respectively.

67

Table 10

 

 

 

 

Dependent Variable MH% MH%
Regression Type F—M FE
Lag MH% 0187‘” -0.202***
(3.24) (0.022)
Lag AMH% -0.245* —0.052
(-1.79) (0.077)
Log(Assets) -0.009** -0.003
(-3.82) (0.003)
Log(BM) 0.010 -0.001
(1.21) (0.007)
BETA -0008 _0.030***
(-1.68) (0004)
EXRET3-12 0.084*** 0065*“
(8.54) (0.006)
R&D 0.076 0.059
(1.37) (0.048)
ADV 0.097M 0.087
(2.59) (0985)
EBITA 0.045 0.024
(0.17) (0.030)
CAPEX -0.016 ' -0.011**
(-1.23) (0.037)
RESIDVAR 0.069 1592*“
(0.12) (0.280)
LEVERAGE 0.011 -0.011
(0.42) (0.016)
E-lNDEX -0.003 -0.002
(-1.56) (0.02)
R2(%) 38.35 18.57

 

68

Table 11
Changes in Insider Ownership

Table 11 presents the results of a sequence of Fame-Macbeth (1973) and firm level fixed effects
regressions in which the dependent variable is the annual change in the fraction of shares outstanding held
by corporate insiders (AMH%). To investigate the dynamics of trade, changes in insider ownership are
related to their own one year lag and the one year lag of changes in holdings by institutional investors.
MH%, ATRA%, AQIX%, ADED% refer to changes in holdings by institutions in the aggregate, transient
institutional investors, quasi-indexer institutional investors and dedicated institutional investors
respectively. F-M denotes Fama-Macbeth (1973) and FE, denotes fixed effects type regressions. The F-M
estimates are the time series average coefficients for 10 cross sectional regressions for each year from
1994-2003 (with the years 1992 and 1993 being dropped because of the lag structure of the specification).
Fame-Macbeth (1973) t-statistics are reported in parentheses. The fixed effects regressions include
intercept terms and year dummies but these are not reported. Fixed effects standard errors are reported in
parentheses. All control variables are measured as of the end of the previous fiscal year. ***, ** and *
refer to statistical Significance at the 1%, 5% and 10% levels respectively.

69

Table 11

 

 

 

 

Dependent Variable AMH% AMH% AMH% AMH 0/8
Regression Type F-M FE F-M FE
Lag MH% -0.011 -0.012
(-1.48) (0.007)
Lag AMH% 0.046 -0.016 0.047 -0.021
(-0.15) (0.029) (-0.01) (0.029)
Lag ATRA% 0.004 -0.004
(1.04) (0.008)
Lag AQIX% ~0.030** -0.027***
(-3.12) (0.009)
Lag ADED% -0.016 -0.023**
(-1.57) (0.010)
Log(Assets) 0002* 0.001 0.0019* 0.001
(1.94) (0.001) (1.59) (0.000)
Log(BM) 0.001 0.001 0.0017 0.002
(0.61) (0.001) (0.80) (0.001)
BET A -0.003 -0.001 0003 -0.000
(-0.69) (0.001) (-0.73) (0.001)
EXRET3-12 -0.004* -0.005*** -0.004 -0.006***
(-1.83) . (0.002) (-1.59) (0.002)
R&D 0.003 0.027 -0.001 0.018
(0.75) (0.016) (0.56) (0.016)
ADV -0.018 0.002 -0.018 0.013
(-0.35) (0.028) {-0.39) (0.027)
EBITA -0.013 -0.018 -0.008 -0.017
(-0.60) (0.010) (-O.37) (0.010)
CAPEX -0.013 -0.010 -0.014 -0.006
(-1.31) (0.012) (-1.22) (0.012)
RESIDVAR 0.006 -0.096 -0.076 -0.084
(-0.84) (0.094) (-0.69) (0.094)
LEVERAGE 0.000 0.005 -0.000 0.004
(0.35) (0.005) (0.48) (0.005)
E-INDEX 0.000 0.000 0.000 0.000
(0.98) (0.000) (0.86) (0.000)
R2194) 12.46 7.24 14.53 11.19

 

70

Table 12
Portfolio Rebalancing: Transient Institutional Investors

Table 12 presents the results of a sequence of Fama-Macbeth (1973) and firm level fixed effects
regressions in which the dependent variable is the annual change in the fraction of shares outstanding held
by transient institutional investors (ATRA%). To investigate the dynamics of trade, changes in transient
ownership are related to their own one year lag and the one year lag of changes in holdings by all other
classes of investors. For expository purposes, we report results for both the parsimonious (own lags only)
and full (all institutional classes lagged) regression specifications. AMH%, AQIX%, ADED% refer to
changes in holdings by corporate insiders, quasi-indexer institutionalinvestors and dedicated institutional
investors respectively. F—M denotes Fama-Macbeth (1973) and FE denotes fixed effects type regressions.
The F-M estimates are the time series average coefficients for 10 cross sectional regressions for eachyear .
from 1994-2003 (with the years 1992 and 1993 being dropped because of the lag structure of the
specification). Fama-Macbeth (1973) t-statistics are reported in parentheses. The fixed effects regressions
include intercept terms and year dummies but these are not reported. Fixed effects standard errors are
reported in parentheses. All control variables are measured as of the end of the previous fiscal year. ***,
** and * refer to statistical significance at the 1%, 5% and 10% levels respectively.

71

Table 12

 

 

 

Dependent Variable TRA% TRA% TRA% TRA%
Rggression Type F-M FE F -M FE
Lag AMH% -0.033 -0.020 -0.065 -0.031
(-0.30) (-0.008) (-0.32) (0.080)
Lag ATRA% -O.289*** -0.258*** -0.266*** -O.276***
(-9.58) (0.022) (-9.69) (0.025)
Lag AQIX% 0.047 -0.025
(1.82) (0.031)
Lag ADED% 0.015 -0.089***
(0.58) (0.032)
Log(Assets) -0.007 -0.0025 -0.007 -0.000
(-1.58) (0.003) (—1.51) (0.002)
Log(BM) 0011* -0.004 0.012 -0.000
' (1.93) (0.004) (2.19) (0.004)
BETA -0.007 -0.052*** -0.008 -0.051***
(-1.51) (0.004) (-1.53) (0.004)
FXRET3-12 0.096*** 0.090*** 0.092*** 0089*“
(8.63) (0.006) (8.32) (0.006)
R&D 0.098 0085* 0.095 0088*
~ (1.67) (0.049) 1 (1.56) (0.049)
ADVI 0.071 0.012 0.061" 0.002
(0.47) (-0.016) (0.17) (0.088)
EBITA -0.088*** -0.004 -0.070*"‘* -0.022
(-4.26) (0.032) (-4.18) (0.032)
CAPEX 0.009 0.033 -0.002 0.026
(-0.59) (0.037) (-0.82) (0.038)
RESIDVAR -0.422 2.234*** -0.319 2.219
(-1.40) (0.291) (-1.24) (0.290)
LEVERAGE 0.005 0.025 0.002 0.023
(-0.04) (0.017) (-0.27) (0.017)
E-INDEX -0.000 -0.003* 0.000 -0.003*
(-O.83) (0.002) (-0.71) (0.002)
(gm) 41.84 24.89 43.20 25.57

 

 

72

Table 13
Portfolio Rebalancing: Quasi-Indexer Institutional Investors

Table 13 presents the results of a sequence of Fama-Macbeth (1973) and frrm level fixed effects
regressions in which the dependent variable is the annual change in the fraction of shares outstanding held
by quasi-indexer institutional investors (AQ1X%). To investigate the dynamics of trade, changes in quasi-I
indexer oWnership are related to their own one year lag and the one year lag of changes in holdings by all
other classes of investors. For expository purposes, we report results for both the parsimonious (own lags
only) and full (all institutional classes lagged) regression specifications. AMH%, ATRA %, ADED% refer
to changes in holdings by corporate insiders, transient institutional investors and dedicated institutional
investors respectively. F-M denotes tamer—Macbeth (1973) and FE denotes fixed effects type regressions.
The .F-M estimates are the time series average coefficients for 10 cross sectional regressions for each year
from 1994-2003 (with the years 1992 and 1993 being dropped because of the lag structure of the
specification). Fama-Macbeth (1973) t-statistics are reported in parentheses. The fixed effects regressions
include intercept terms and year dummies but these are not reported. Fixed effects standard errors are
reported in parentheses. All control variables are measured as of the end of the previous fiscal year. ***,
** and * refer to statistical significance at the 1%, 5% and 10% levels respectively.

73

Table 13

 

 

 

 

Dependent Variable QIX% QIX% QIX% QIX%
Regression Type F -M FE F-M FE
Lag AMH% -0.288** 0134* -0.251** -0109
(-2.56) (0.087) (-227) (0.08)
Lag ATRA% 0.0172 0.032
(0.92) (0.026)
Lag AQIX% 0244*" -O.262*** -0.219*** -0.135***
(-4.86) (0.025) (4.09) (0.033)
Lag ADED% 0.030 0226*"
(0.47) (0.035)
Log(Assets) -0002 -0.007** -0002 -0007"
(-0.60) (0.003) (-0.70) (0.003)
Log(BM) 0.000 0.000 -0002 -0000
(0.05) (0.005) (—0.17) (0.005)
BETA 0.001 0.034*** -0001 0.033***
(0.07) (0.005) (—O.63) (0.005)
EXRE'1‘3—12 ' 0.0027 -0.012* 0.004 -0009
(0.84) (0.006) (0.56) (0.006)
R&D 0.029 -0019 -0037 0022*
(-009) (0.054) (-039) (0.053)
ADV -0.018 -0049 0.014 -0.085
(0.32) (0.096) (0.89) (0.094)
EBITA 0.042** -0050 0.041" -0.068
(2.68) (0.034) (2.84) (0.034)
CAPEX -0014 -0042 -0014 -0039
(-130) (0.042) (-117) (0.041)
RESIDVAR 0.029 -2.433*** 0.136 -1.984***
(0.52) (0.317) (0.85) (0.313)
LEVERAGE 0.013 -0.01 1 0.016 -0.008
(1.03) (0.018) (1.17) (0.018)
E-INDEX -0002 -0000 -0003 -0000
(—1.03) (0.002) (-100) (0.002)
R2(%) 20.29 10.98 23.11 12.63

 

74

Table 14
Portfolio Rebalancing: Dedicated Institutional Investors

Table 14 presents the results of a sequence of Fama-Macbeth (1973) and firm level fixed effects
regressions in which the dependent variable is the annual change in the fraction of shares outstanding held
by dedicated institutional investors (ADED%). To investigate the dynamics of trade, changes in dedicated
ownership are related to their own one year lag and the one year lag of changes in holdings by all other
classes of investors. For expository purposes, we report results for both the parsimonious (own lags only)
and full (all institutional classes lagged) regression specifications. AMH%, ATRA %, AQIX % refer to
changes in holdings by corporate insiders, transient institutional investors and quasi-indexer institutional
investors respectively. F-M denotes Fama-Macbeth (1973) and FE denotes fixed effects type regressions.
The F—M estimates are the time series average coefficients for 10 cross sectional regressions for each year
from 1994-2003 (with the years 1992 and 1993 being dropped because of the lag structure of the
specification). Fama-Macbeth (1973) t-statistics are reported in parentheses. The fixed effects regressions
include intercept terms and year dummies but these are not reported. Fixed effects standard errors are
reported in parentheses. All control variables are measured as of the end of the previous fiscal year. ***,
** and * refer to statistical significance at the 1%, 5% and 10% levels respectively.

75

Table 14

 

 

 

 

Dependent Variable DED% DED% DED% DED%
Regression Type F-M FE F-M FE
Lag AMH% -0010 0.095 0.001 0.102
(-034) (0.076) (-023) (0.077)
Lag ATRA% 0.015 0.041
(0.19) (0.024)
Lag AQIX°/o 0.042 -0017
(1.07) (0.030)
Lag ADED% —0.233*** -0.327*** -0.204*** -0.333***
(-5.06) (0.023) (—3.83) (0.031)
Log(Assets) 0.002 0.008*** 0.002 0.008***
(0.47) (0.003) (0.40) (0.003)
Log(BM) -0004 0.000 -0002 0.002
(-147) (0.004) (-122) (0.004)
BETA -0001 -0.014*** -0.001 -0.013**
(-013) (0.004) (-0.24) (0.004)
EXRET3-12 0.002 0.009 0.000 0.000
(0.29) (0.005) (0.45) (0.006)
R&D 0 024 -0031 0.025 -0029
(0.77) (0.048‘ (0.77) (0.047)
ADV 0.032 0.128 0.020 0.131
(0.73) (0.080) (0.63) (0.085)
EBITA 0081* 0.117*** 0085* 0.088***
(2.14) (0.030) (2.10) (0.031)
CAPEX 0.009 -0043 —0.01 1 -0.018
(0.29) (0.036) (0.22) (0.036)
RESIDVAR -0.081 1.618*** -0.060 1.590***
(-009) (0.279) (-027) (0.279)
LEVERAGE -0024 -0.037** -0019 -0.037**
(-1 .66) (0.016) (-142) (0.016)
E-INDEX 0.000 0.003 -0000 0.002
(0.29) (0.002) (0.11) (0.001)
R3(o/,) 21.20 13.02 23.65 13.19

 

76

Table 15
Performance Attribution: Institutional Investors and Corporate Insiders

Table 15 shows the time series average coefficients for 132 cross-sectional regressions for each month from
December 1993-December 2003. The dependent variable is individual monthly excess stock returns
relative to the CRSP equal weighted index (Rit). IH% and MH% refer to the fraction of shares owned by
institutions and corporate insiders respectively at the end of the current quarter. Lagged versions of all
institutional ownership measures refer to holdings at the end of the previous quarter. Fama-Macbeth (1973)
t-statistics are reported in parentheses. SIZE is the market capitalization as of the previous fiscal year. B/M
is the book to market ratio as of the previous fiscal year. NASDUM is an indicator equal to one if the firm
traded on the NASDAQ at the beginning of month t. PRICE is the reciprocal of price as of month t-2.
NYDVOL and NASDVOL are the dollar volume of trade in month t-2 for stocks that trade on the NYSE or
AMEX and NASDAQ respectively. YLD is the ratio of dividends to market capitalization as of the
previous fiscal year. RET2-3, RET4-6 and RET7-12 are the compounded gross returns over the months t-3
to t-2, t-6 to t-4 and t—lZ to t-7 respectively. ***, ** and * represent statistical significance at the 1%, 5%
and 10% levels respectively.

77

Table 15

 

 

 

Dependent Variable Rit Rit Rit Rit
IH% 0.032***
(3.06)
Lag IH% -0.005
(-O.51)
MH% 2.288
(1.09)
Lag MH% 0996
(-0.99)
MH%Z -20.681
(-1.08)
Lag MH%:Z 9.456
(1.01)
SIZE 0000* 0.000 0.000 0.001
(1.77) (0.94) (0.88) (0.64)
B/M 0018*” 0.019*** 0.021“ 0021*
(2.81) (2.98) (2.31) (1.98)
NASDU M 0.03] -0.024 -0.000 -0.003
(0.55) (—0.49) (—0.000) (-0.041)
PRICE 0.004 -0.000 -0.001 -0.001
(1.15) (-0.12) (-0.01) (-0.01)
NYDVOL -0.002 -0.002 -0.000 -0.000
(-l.18) (-1.06) (-1.12) (047)
NASDVOL -0.004 0.000 -0.007 0.00
(-0.95) (0.00) (-1.00) (0.00)
YLD -0.229* -0.261** -0.214** -0.214**
(-1.88) (-2.02) (-2.09) (-2.38)
RET2-3 0.005 -0.002 0.001 -0.001
(0.44) (-0.l3) (0.00) (0.01)
RET4-6 -0.002 0.001 -0.000 0.001
(-0.83) (0.56) (-000) (0.07)
RET7-12 -0.014 0034* -0.000 0.054
(—0.45) (1.77) (-0.00) (0.21)
112(0),) 21.70 22.17 8.38 8.99

 

78

Table 16
Performance Attribution: Institutional Investors by Class

Table 16 shows the time series average coefficients for 132 cross-sectional regressions for each month from
December 1993-December 2003. The dependent variable is individual monthly excess stock returns
relative to the CRSP equal weighted index (Rit). TRA%, QIX% and DED% refer to the fraction of shares
owned by transient, quasi-indexer and dedicated institutions respectively at the end of the current quarter.
Lagged versions of all institutional ownership measures refer to holdings at the end of the previous quarter.
Fania-Macbeth (1973) t-statistics are reported in parentheses. SIZE is the market capitalization as of the
previous fiscal year. B/M is the book to market‘ratio as of the previous fiscal year. NASDUM is an
indicator equal to one if the firm traded on the NASDAQ at the beginning of month t. PRICE is the
reciprocal of price as of month t—2. NYDVOL and NASDVOL are the dollar volume of trade in month t-2
for stocks that trade on the NYSE or AMEX and NASDAQ respectively. YLD is the ratio of dividends to
market capitalization as of the previous fiscal year. RET2-3, RET4-6 and RET7-12 are the compounded
gross returns over the months t-3 to t-2, t-6 to t-4 and t—lZ to t-7 respectively. ***, ** and * represent
statistical significance at the 1%, 5% and 10% levels respectively.

79

Table 16

 

 

 

Dependent Variable Rit Rit Rit Rit Rit Rit
TRA% 0.132***
(7.42)
Lag TRA 0.004
(0.23)
QIX% -0.011
(—0.78)
Lag QIX% -0001
(-007)
DED% 0036*
(1.69)
Lag DED% 0.009
(0.40)
SIZE 0.000*** 0.000 0.000 0.000 0.000 0.000
(3.55) (1.34) (0.52) (0.96) (0.58) (0.77)
B/M 0.017** 0.018** 0.018** 0.019*** 0.019*** 0.017***
(2.72) (2.92) (2.69) (2.81) (3.03) (2.72)
NASDUM 0.041 -0027 0.043 -0.030 0.037 -0.027
(0.71) (-0.61) (0.76) . (-O.62) (0.67) (-0.56)
PRICE 0.004 -0001 . 0.001 —0.001 0.002 -0.003
(1.16) (-027) (0.42) (—0.17) (0.49) (-0.45)
NYDVOL —0.005** -0.003 -0.001 -0.002 -0001 -0003
(—2.54) (—1.28) (-072) (~1.26) (—0.48) (-121)
NASDVOL -0.008 -000 —0.005 0.000 -0004 —0.001
(-1.55) (-013) (—0.97) (0.02) (-O.80) (-014)
YLD -0032 0222* -0.242* —0.270** —0. 179 -0.269**
(-0.28) (-1.93) (-195) (-2.26) (-157) (-2.36)
RET2-3 -0010 -0.009 0.004 -0004 0.006 -0009
(-0.86) (-0.66) (0.33) (—0.31) (0.47) (-054)
RET4-6 —0.002 0.002 -0002 0.002 —0.003 0.003
(-1.15) (0.85) (-0.80) (0.71) (-1.02) (0.83)
RET7-12 -0.025 0041* -0.006 0.047 -0021 0.058
(-1.04) (1.89) (-O.28) (1.56) (-0.61) (1.49)
R306) 22.59 22.54 21.20 21.97 21.47 21.91

 

80

Table 17
Performance Attribution: Transient Institutional Investment Strategies: Growth vs.
Value

Table 17 shows the time series average coefficients for 132 cross-sectional regressions for each month from
December 1993-December 2003. The dependent variable is individual monthly excess stock returns
relative to the CRSP equal weighted index. TRA-LVA%, TRA-LGR%, TRA-SVA% and TRA-SGR%
refer to the holdings of transient institutions that invest in large value, large growth, small value and small
growth firms at the end of the current quarter, respectively. The categorization of transient institutions is
determined by the method of Abarbanell, Bushee and Raedy (2003). Lagged versions of all institutional
ownership measures refer to holdings at the end of the previous quarter. Fama-Macbeth (1973) t-statistics
are reported in parentheses. SIZE is the market capitalization as of the previous fiscal year. B/M is the book
to market ratio as of the previous fiscal year. NASDUM is an indicator equal to one if the firm traded on
the NASDAQ at the beginning of month t. PRICE is the reciprocal of price as of month t-2. NYDVOL and
NASDVOL are the dollar volume of trade in month t-2 for stocks that trade on the NYSE or AMEX and
NASDAQ respectively. YLD is the ratio of dividends to market capitalization as of the previous fiscal year.
RET2-3, RET4-6 and RET7-12 are the compounded gross returns over the months t-3 to t-Z, t—6 to t-4 and
t-12 to t-7 respectively. ***, ** and * represent statistical significance at the 1%, 5% and 10% levels
respectively.

81

Table 17

 

 

 

 

Panel A
Dependent Variable Rit Rit Rit Rit
TRA-LGR% 0198*"
(2.89)
Lag TRA-LGR% 0.044
(0.70)
TRA-SGR% 0.218***
(9.31)
Lag TRA—SGR% 0.008
(0.30)
SIZE 0.000 0.000 0.000 0.000
(0.77) (1.31) (4.37) (1.60)
BM 0018*“ 0016*“ 0.022*** 0019*"
(2.88) (2.59) (3.32) (2.95)
NASDUM 0.037 -0.029 0.067 -0.025
(0.70) - (-O.63_) (1.03) (-0.53)
PRICE 0.003 -0.002 0.003 -0.000
(0.90) (—0.41) (1.09) (-0.03)
NYDVOL -0.002 -0.004 -0.004** -0.002
(-1.09) (-1.27) (-2.47) (-1.39)
NASDVOL -0.005 -0.001 -0.009* -0.001
(-1.06) (-O.30) (-1.71) (-0.15)
YLD -0. 168 -0.264** 0.126 —0.219*
(-1.47) (-2.43) (1.14) (-l.77)
RET2-3 0.005 -0.011 -0.016 -0008
(0.44) (-0.58) (-1.42) (-O.66)
RET4-6 -0.002 0.006 -0.003 0.001
(-0.79) (0.95) (-1.66) (0.53)
RET7-12 -0.008 0.074 -0.038 0.028
(-0.30) (1.46) (-l.48) (1.56)
R290) 21.38 22.05 22.77 22.66

 

82

Table 17 (cont’d).

 

 

 

 

Panel B
Dependent Variable Rit Rit Rit Rit
TRA-LVA% 0100*
(1.88)
Lag TRA-LVA% 0.065
(0.99)
TRA-SVA% -0.062
(-0.99)
Lag TRA-SVA% -0.033
(-0.50)
SIZE 0.000 0.000 0.000 0.000
(1.16) (1.02) (0.22) (0.56)
B/M 0019*“ 0.016*** 0022*“ 0020*“
(3.08) (2.49) (3.28) (2.96)
NASDUM 0.042 -0.026 0.041 -0.032
(0.75) (-0.53) (0.69) (-0.63)
PRICE 0.003 -0.001 0.003 0.001
(0.80) (-0.25) (0.75) (0.18)
NYDVOL -0.002 —0003 -0001 -0.002
(-I.O7) (-1.I4) (-0.72) (-1.19)
NASDVOL -0.005 -0.000 -0.004 0.000
(-1 .09) (-0.14) (-0.93) (0.06)
YLD -0.234** -0.266** -0.245** -O.263**
(-2.19) (-2.36) (-2.23) (-2.42)
RET2-3 0.007 -0.005 0.002 -0.006
(0.58) (-0.32) (0.17) (-047)
RET4-6 -0.001 0.005 -0000 0.003
(-0.50) (1.07) (-0.35) (0.92)
RET7-12 0.001 0069* -0002 0046*
(0.06) (1.76) (006) (1.81)
R290) 21.19 22.23 21.25 22.00

 

83

Table 18
Ownership Structure and Firm Value

Table 18 presents the results of a sequence regressions in which the dependent variable is Tobin’s Q
(measured as the market value of equity (Compustat item 199 x Compustat item 25) plus the book value of
assets (Compustat item 6) minus the book value of common equity including deferred taxes (Compustat
items 60 and 74) all divided by the book value of assets). F-M and FE denote Fama-Macbeth (1973) and
fixed effects regression types respectively. The F-M estimates are the time series average coefficients for
12 cross sectional regressions for each year from 1992-2003. Fania-Macbeth (1973) t-statistics are reported

in parentheses. The fixed effects specifications report standard errors in parentheses.

 

 

 

 

Dependent Variable Q Q Q Q Q Q
Rggression Type F—M FE F—M FE F-M FE
MH% 1.451** 0.649 1.732*** 1.209** 1.023** 0900*
(2.84) (0.555) (3.15) (0.557) (2.29) (0.477)
MH%Z -4.388** -1214 -3.877* -1.642 -3.502* -1977
(-250) (1.400) (-207) (1.414) (-202) (1.17)
IH% 0.535** 0.496***
(4.57) (0.126)
TRA% 2.243*** 1.059*** 2.073** 1.021***
(5.27) (0.161) (4.88) (0.140)
QIX% 0512* 0214 0325 0.090
(-205) (0.200) (-1.41) (0.171)
DED% 0.033 0.011 0101 0.069
(0.07) (0.225) (023) (0.193)
RD 2.049** 2.532*** 2.034** 2.697*** 1.502** 2.695***
(2.76) (0.352) (3.47) (0.357) (2.19) (0.310)
ADV 2.451** 1.942*** 2.087** 1.854** 2.631** 1.983***
(2.43) (0.717) (2.15) (0.724) (2.70) (0.594)
LEVERAGE -I.682*** -1.698*** -1.735*** -l.817*** -1.638** -1.472***
(~6.88) (0.121) (—6.74) (0.122) (079) (0.103)
Log(Assets) -0.0164 0003 0017 0000 0023 0.026
(055) (0.020) (064) (0.02) (086) (0.017)
E-INDEX -0.151*** 0149*“ -O.138*** 0129***
(-8.81) (0.016) (-9.63) (0.014)
R2(%) 25.26 19.54 27.09 17.66 29.84 19.63

 

84

Table 19
Robustness Tests

Table 19 presents the results of a sequence regressions in which the dependent variable is Tobin’s Q
(measured as the market value of equity (Compustat item 199 x Compustat item 25) plus the book value of
assets (Compustat item 6) minus the book value of common equity including deferred taxes (Compustat
items 60 and 74) all divided by the book value of assets). F-M, FE and IV denote Fania-Macbeth (1973),
fixed effects and instrumental variables regression types respectively. The F-M estimates are the time series
average coefficients for 12 cross sectional regressions for each year from 19922003. Fama-Macbeth
(1973) t-statistics are reported in parentheses. The fixed effects and instrumental variables specifications
report standard errors in parentheses. The fixed effects and instrumental variables specifications report
standard errors in parentheses. As instruments for our endogenous ownership measures we use the log of
total assets, its square, beta from the market model regression, its square, the residual variance from the
market model regression, the I/B/E/S reported analyst forecast error of EPS as of the end of the previous
fiscal year and its square.

85

Table 19

 

 

 

Dependent Variable Q Q Q
Rggression Type IV F-M FE
MH% 29.503 0.443 0.086
(20.298) (1.33) (0.513)
MH%? -49.138 -2012 0.474
(58.640) (-134) (1.297)
TRA% 3.991***
(1.514)
TRA-LVA% 1.221 1.541***
(1.16) (0.515)
TRA-LGR% 3.403** 4.048***
(2.66) (0.579)
TRA-SVA% -11.64*** -4.592***
(—3.23) (0.381)
TRA-SGR% 3.659*** 3.912***
(6.07) (0.287)
QIX% 0780 0.134 0.186
(2.909) (0.57) (0.184)
DED% 11.307 0119 0623***
(6.745) (026) (0.209)
RD 2.639*** 0.628 1.364***
(0.745) (0.78) (0.333)
ADV 2.135 3.065** 2.429***
(1.369) (2.98) (0.664)
LEVERAGE -1.716*** -1.433*** -1.411***
(0.241) (-795) (0.114)
Log(Assets) -0.079*** -0.048**
(-2.99) (0.020)
E-INDEX 0139*** 0107*** 0102***
(0.030) (-5.86) (0.015)
R2(%) 7.17 40.74 20.59

 

 

Table 20
Conditional Specifications of the Firm Value — Ownership Relation

Table 20 presents the results of a sequence of Fama-Macbeth (F-M) and fixed effects (FE) regressions in
which Tobin’s Q is the dependent variable. MH%, TRA%, QIX% and DED% refer to the fraction of shares
owned by management, transient, quasi-indexer and dedicated institutions respectively at the end of the
current fiscal year. E-INDEX_1 is a dummy variable that takes on a value of I if a firm is in the top half of
the sample by entrenchment index score and 0 if it is not. CASHVOL_1 is a dummy variable that takes on
a value of 1 if a firm’s volatility of EBITA measured over the previous 5 years is in the top half of the
sample and 0 if it is not. F-DISP_1 that takes on a value of 1 if a firm is in the top half of the sample by
standard deviation of analyst estimates of EPS and 0 if it is not. We interact each of the three dummy
variables with our institutional ownership measures. With the exception of BETA, RESIDVAR, and F-
DISP__1 (which are lagged by one fiscal year), all control variables are measured as of the end of the
current fiscal year. ***, ** and * refer to statistical significance at the 1%, 5% and 10% levels respectively.
All regressions include intercepts but these are not reported.

87

Table 20

 

 

 

 

Panel A
Dependent Variable Q Q
Rggression Type F-M FE
MH% 1.119** 0.745
(2.25) (0.554)
MH%2 -3.740* -l.198
(-195) (1.408)
TRA% 2.256*** 1.238***
(5.31) (0.209)
QIX% 0387 0161
(-131) (0.260)
DED% 0095 0263
(021) (0.289)
TRA% x E-INDEX_1 0497 0534*
(-1.39) (0.311)
QIX% x E-INDEX_1 0.339 ' 0.024
(0.96) (0.381)
DED% x E_1NDEx_1 0.583 0.472
(0.74) (0.441)
RD 1.541** 2.433***
' (2.44) (0.355)
ADV 2.781** 1.955***
(2.84) (0.717)
LEVERAGE -1 .643*** -1.729***
(-6.86) (0.124)
Log(Assets) —0.020 0.024
(—0-77) (0.020)
E-INDEX_1 0122** -0246
(-211) (0.169)
R2(%) 30.87 16.45

 

 

Table 20 (cont’d)

 

 

 

 

 

Panel B
Dependent Variable Q Q
Rggression Type F-M FE
MH% 0944* 0.344
(2.13) (0533)
MH%2 0222* -1179
(-1.93) (1.345)
TRA% 1.148 0804*"
(0.98) (0.217)
QIX% -0.116 -0.233
(020) (0.277)
DED% 1.009 0.315
(0.89) (0.328)
TRA% x F-DISP_I 0.934 -0.030
(0.83) (0.296)
QIX% x F_DISP_1 0.341 0.256
(0.65) (0.361)
DED% x F_DISP_1 -I.545 -0.472
(-1.30) (0.423)
RD 1.784“ 2.615***
(2.18) (0.343)
ADV 2.541*** 2.325***
(2.96) (0.689)
LEVERAGE -1.335*** -1.487****
(-5.89) (0.118)
Log(Assets) -0.002 0.017
00.09) (0.019)
E-INDEX -0.128*** -0138***
(-12.50) (0.015)
F-DISP_1 -0313** -0.527***
(-2.14) (0.158)
[fin/0) 39.69 17.91

 

Table 20 (cont’d)

 

 

 

 

Panel C
Dependent Variable Q Q
Regression Type F-M FE
MH% 1.089 0.570
(1.72) (0562)
MH%} -3.289 -1.489
(4.68) (1.409)
TRA% 0.425 0735“"I
(0.80) (0.241)
QIX% -0.146 -0.212
(-0.64) (0.274)
DED% 0655 -0.696**
(-1.27) (0.321)
TRA% x CASHVOL_I 2.742*** 0.500“
(3.78) (0.242)
QIX% x CASHVOL_1 0144 0.119
(-027) (0.386)
DED% x CASHVOLVI 1.089 1.265**
(1.65) (0.439)
RD 1.667" 2402*“
(2.78) (0.397)
ADV 2.742" 2376*“
(3.30) (0.754)
LEVERAGE -1.789*** -1.745
(-7.09) (0.125)
Log( Assets) -0.009 0.006
(039) (0.021)
E-INDEX -0129*** -0144***
(-l3.10) (0.016)
CASHVOL_1 0257* -O.4l3***
(—1.89) (0172)
R2(%) 34.2I 21,44

 

 

90

BIBLIOGRAPHY

91

Abarbanell, J .,Bushee, B.J., Raedy, J ., 2003. Institutional investor preferences and price
pressure: The case of corporate spin-offs. Journal of Business 76, 233-261.

Amihud, Y., 2002. Illiquidity and stock returns: cross section and time-series effects.
Journal of Financial Markets 5, 31-56.

Ashbaugh-Skaife, H., Gassen, J., LaFond, R., 2006. Does stock price synchronicity
represent firm-specific information? The international evidence. Working paper, MIT.

Bebchuk, L., Cohen, A., Ferrell, A., 2004. What matters in corporate governance?
Working paper, Harvard.

Berle, A.A., Means, G.C., 1932. The modern corporation and private property
(Macmillan, New York, NY).

Bessembinder, H., 2003. Quote-based competition and trade execution costs in NYSE-
listed stocks. Journal of Financial Economics 70, 385-422.

Brennan, M., Chordia, T., Subrahmanyam, A., 1998. Alternative factor specifications,
security characteristics and the cross section of expected returns. Journal of Financial
Economics 49, 345-374.

Bushee, B.J, 1998. The influence of institutional investors on myopic R&D investment
behavior. The Accounting Review 73, 305-333.

Bushee, B.J., 2001. Do institutional investors prefer near-tenn earnings over long run
value? Contemporary Accounting Research 18.

Carhart, M., 1997. On persistence in mutual fund performance. Journal of Finance 52,
57-82.

Cho, M.H., 1998. Ownership structure, investment, and the corporate value: an empirical
analysis. Journal of Financial Economics 47, 103-121.

Core, J.E., Larcker, DE, 2002. Performance consequences of mandatory increases in
executive stock ownership. Journal of Financial Economics 64, 317-340.

Coles, J.L., Lemmon, M.L., Meschke, J .F., 2003. Structural models and endogeneity in
corporate finance: The link between managerial ownership and corporate finance.

Working paper, Arizona State University.

Del Guercio, D., 1996. The distorting effect of the prudent-man laws on institutional
equity investments. Journal of Financial Economics 40, 31-62.

Demsetz, H., Lehn, K., 1985. The structure of corporate ownership: Causes and
consequences. Journal of Political Economy 93, 1155-1177.

92

Demsetz, H., Villalonga, B., 2001. Ownership structure and corporate performance.
Journal of Corporate Finance 7, 209-233.

Diamond, D.W., 1984. Financial intennediation and delegated monitoring. The Review
of Economic Studies 51, 393-414.

Dumev, A., Morck, R., Yeung, 3., 2004. Value-enhancing capital budgeting and firm-
specific stock return variation. Journal of Finance 59, 65-105.

Falkenstein, E.G., 1996. Preferences for stock characteristics as revealed by mutual fund
portfolio holdings. Journal of Finance 51, 111-135.

Fama, E.F., Macbeth, JD, 1973. Risk, return, and equilibrium: Empirical tests. Journal
of Political Economy 71, 607-636.

Fama, E.F., French, KR, 1992. The cross-section of expected stock returns. Journal of
Finance 47, 427-465.

Fama, E.F., French, KR, 1993. Common risk factors in the returns on stocks and bonds.
Journal of Financial Economics 33, 3-56.

Gompers, P., Ishii, J., Metrick, A., 2003. Corporate governance and equity prices. The
Quarterly Journal of Economics 118, 107-156.

Gompers, P., Metrick, A., 2001. Institutional investors and equity prices. The Quarterly
Journal of Economics 116, 229-260.

Griffin, J.M., Kelly, P.J., Nadari, F., 2006. Measurement and determinants of
international stock market efficiency. Working paper, University of Texas at Austin.

Hasbrouck, J ., 2005. Trading costs and returns for US equities: estimating effective costs
from daily data. Working paper, NYU.

Holdemess, C.G., Kroszner, R.S., Sheehan, DP, 1999. Were the good old days that
good? Changes in managerial stock ownership since the great depression. Journal of
Finance 54, 435-468.

Himmelberg, C.P., Hubbard, R.G., Palia, D., 1999. Understanding the determinants of
managerial ownership and the link between ownership and performance. Journal of

Financial Economics 53, 353-384.

Jensen, M.C., Meckling, W.H., 1976. Theory of the firm: Managerial behavior, agency
costs and ownership structure. Journal of Financial Economics 3, 305-360.

93

Jensen, M.C., 1986. Agency costs of free cash flow, corporate finance and takeovers.
American Economic Review 76, 323-339.

Jones, C., Seguin, P., 1997. Transaction costs and price volatility: Evidence from
commission deregulation. American Economic Review 87, 728-737.

Kang, J.-K., Stulz, R.M., 1997. Why is there a home bias? An analysis of foreign
portfolio ownership in Japan. Journal of Financial Economics 46, 3—28.

Ke, B., Petroni, K., 2004. How informed are actively trading institutional investors?
Evidence from their trading behavior before a break in a string of consecutive earnings
increases. Journal of Accounting Research 42, 895-927.

Kelly, P.J., 2005. Information efficiency and firm-specific return variation. Working
paper, University of South Florida.

Lakonishok, J ., Lee, I., 2001. Are insider trades informative? The Review of Financial
Studies 14, 79-11.

Lakonishok, J ., Shleifer, A., Vishny, R.W., 1992. The impact of institutional trading on
stock prices. Journal of Financial Economics 32, 23-43.

Lang, L.H.P., Stulz, R.M., Walkling, R.A., 1’991. Awtest of the free cash flow hypothesis:
The case of bidder returns. Journal of Financial Economics 29, 315-335.

LaPOrta, R.,‘ I.opez~de-Silanes, F., Shleifer,- A., Vishny, R.W., 1998. Law and finance. v
Journal of Political Economy 106, 1113-1155. -

LeRoy, S.F., Porter, RD, 1981. The present-value relation: tests based on implied
variance bounds. Econometrica 49, 555-574.

Merton, RC, 1987. A simple model of capital market equilibrium with incomplete
information. Journal of Finance 42, 483-510.

McConnell, J.J., Servaes, H., 1990. Additional evidence on equity ownership and
corporate value. Journal of Financial Economics 27, 595-612.

McConnell, J .J ., Servaes, H., 1995. Equity ownership and the two faces of debt. Journal
of Financial Economics 39, 131-157.

McConnell, J .J ., Servaes, H., Lins, K.V., 2005. Changes in equity ownership and changes
in the market value of the firm. Working paper, Purdue University.

Morck, R., Shleifer, A., Vishny, R.W., 1988. Management ownership and market
valuation: An empirical analysis. Journal of Financial Economics 20, 293-315.

94

Morck, R., Yeung, B., Yu, W., 2000. The information content of stock markets: why do
emerging markets have synchronous stock price movements? Journal of Financial
Economics 58, 215-260.

Myers, SC, 1977. Determinants of corporate borrowing. Journal of Financial Economics
5, 147-175.

Pound, J., 1988. Proxy contests and the efficiency of shareholder oversight. Journal of
Financial Economics 20, 237-265.

Nofsinger, J.R., Sias, R.W., 1999. Herding and feedback trading by institutional and
individual investors. Journal of Finance 54, 2263-2295.

Seyhun, H.N., 1986, Insiders’ profits, cost of trading, and market efficiency. Journal of
Financial Economics 16, 189-212.

Shiller, R. J., 1981. The use of volatility measures in assessing market efficiency. Journal
of Finance 36,291 3-04.

Stulz, R., 1988. Managerial control of voting rights: Financing policies and the market
for corporate control. Journal of Financial Economics 20, 25-54.

Wahal, S., McConnell, J., 1997. Do institutional investors exacerbate managerial
myopia? Working paper, Emory University.

Wermers,- R., 1999. Mutual fund herding and the impact on stock prices. Joumal of
Finance 54, 581-622. -

West, K.D., 1988. Dividend innovations and stock price volatility. Econometrica 56, 37-
61.

Zhou, X., 2001. Understanding the determinants of managerial ownership and the link
between ownership and performance: comment. Journal of Financial Economics 62, 559-
571.

95

l I

I

Iii

11.10.

I.