The Influence of Institutional Investors on Myopic R&D Investment Behavior

Author(s): Brian J. Bushee

Source: The Accounting Review , Jul., 1998, Vol. 73, No. 3 (Jul., 1998), pp. 305-333
Published by: American Accounting Association

Stable URL: https://www.jstor.org/stable/248542

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

American Accounting Association is collaborating with JSTOR to digitize, preserve and extend
access to The Accounting Review

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
THE ACCOUNTING REVIEW
Vol. 73, No. 3
July 1998
pp. 305-333

The Influence of Institutional

Investors on Myopic R&D
Investment Behavior
Brian J. Bushee
Harvard University

ABSTRACT: This paper examines whether institutional investors create or re-

duce incentives for corporate managers to reduce investment in research and
development (R&D) to meet short-term earnings goals. Many critics argue that
the frequent trading and short-term focus of institutional investors encourages
managers to engage in such myopic investment behavior. Others argue that
the large stockholdings and sophistication of institutions allow managers to
focus on long-term value rather than on short-term earnings. I examine these
competing views by testing whether institutional ownership affects R&D
spending for firms that could reverse a decline in earnings with a reduction in
R&D. The results indicate that managers are less likely to cut R&D to reverse
an earnings decline when institutional ownership is high, implying that insti-
tutions are sophisticated investors who typically serve a monitoring role in
reducing pressures for myopic behavior. However, I find that a large proportion
of ownership by institutions that have high portfolio turnover and engage in
momentum trading significantly increases the probability that managers re-
duce R&D to reverse an earnings decline. These results indicate that high
turnover and momentum trading by institutional investors encourages myopic
investment behavior when such institutional investors have extremely high lev-
els of ownership in a firm; otherwise, institutional ownership serves to reduce
pressures on managers for myopic investment behavior.

This paper is based on my dissertation and was formerly entitled "Institutional Investors, Long-Term Invest-
ment, and Earnings Management." I would like to thank my committee members, Jeff Abarbanell, Doug Skinner,
Gautam Kaul and Phil Howrey for their invaluable guidance and support. I also thank the following for their helpf
comments and suggestions: Greg Miller, Ilia Dichev, Michael Eames, Paul Healy, Amy Hutton, Marco Tansiti,
Marilyn Johnson, Bjorn Jorgensen, Russell Lundholm, Marlene Plumlee, Nancy Rothbard, Jim Wahlen, two anon-
ymous reviewers, and workshop participants at the University of California at Berkeley, the University of Chicag
Columbia University, Cornell University, Dartmouth University, Duke University, Harvard University, the Universi
of Michigan, the University of North Carolina, Northwestern University, the University of Pennsylvania, the Uni
versity of Rochester, the University of Texas at Austin, Washington University, and the 1998 Midwest AAM
Meeting. I gratefully acknowledge the financial support of the Coopers & Lybrand Foundation, the National Do
toral Fellowship Program, and the William A. Paton Scholarship Fund.
Editor's note: This paper is the winner of the 1998 AAA Competitive Manuscript Award.
Submitted January 1998.
Accepted February 1998.

305

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 306 The Accounting Review, July 1998

Key Words: Institutional investors, R&D, Managerial myopia, Earnings

management.

Data Availability: All data are available from public sources.

I. INTRODUCTION
O ver the past decade, many critics have expressed concerns that myopic inv
behavior by U.S. corporate managers has caused U.S. firms to fall behind German
and Japanese firms in terms of competitiveness and technological development
(Jacobs 1991, 7-8; Porter 1992, 5-6; Laverty 1996, 825). Myopic investment behavior (or
"managerial myopia") refers to underinvestment in long-term, intangible projects such as
research and development, advertising, and employee training for the purposes of meeting
short-term goals (Porter 1992, 3-5). In this paper, I examine managerial myopia with
respect to research and development (R&D) investment because lagging U.S. investments
in R&D have triggered much of the concern over managerial myopia and because prior
research provides some evidence that managers use R&D to meet earnings goals (Jacobs
1991, 3; Dechow and Sloan 1991). U.S. accounting rules require that R&D expenditures
be immediately and fully expensed. Critics argue that this accounting treatment creates the
managerial myopia problem because the U.S. capital markets have a short-term focus that
creates pressure on managers to sacrifice R&D to maintain short-term earnings growth
(Drucker 1986; Jacobs 1991, 36; Porter 1992, 24).
A widely cited source of this pressure is the frequent trading and fragmented ownership
of institutional investors.' By acting as "traders" rather than "owners," institutional inves-
tors allegedly place excessive focus on short-term developments, leading managers to fear
that an earnings disappointment will trigger large-scale institutional investor selling and
result in a temporary undervaluation of the firm's stock price (Graves and Waddock 1990,
76-77; Jacobs 1991, 37-38; Porter 1992, 43-46). However, this view of institutional in-
vestor behavior is not universal. Others argue that the large stockholdings and sophistication
of institutional investors allow them to monitor and discipline managers, ensuring that
managers choose investment levels to maximize long-run value rather than to meet short-
term earnings goals (Monks and Minow 1995, chap. 2; Dobrzynski 1993). This paper tests
these two competing views by examining whether institutional investors create or reduce
pressures on managers to reduce investment in R&D to meet short-term earnings goals.
Myopic investment behavior is a type of earnings management that is most likely to
happen when managers face a trade-off between meeting earnings targets and maintaining
R&D investment. To capture this context, I select a sample of firms with pre-R&D earnings
that are below the prior year's level, but by an amount that could be reversed by reducing
R&D. For these firms, I test whether the level of institutional ownership affects the likeli-
hood that managers cut R&D to reverse the expected earnings decline, controlling for other
motivations to cut R&D (e.g., changes in the firm's set of profitable R&D opportunities).
The results indicate that managers are more likely to reduce R&D expenses, and hence
boost earnings toward last year's level, when institutional ownership is low. This result

As defined by the SEC in Rule 13-f, institutional investors are entities such as bank trusts, insurance companies,
mutual funds, and pension funds that invest funds on the behalf of others and manage at least $100 million in
equity. Entities such as arbitrageurs, brokerage houses, and companies holding stock for their own portfolio (as
opposed to in their pension funds) are not considered institutional investors by the SEC and are not required to
disclose their equity investments under Rule 13-f (Wines 1990).

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 307

suggests that institutional investors are sophisticated investors who

itoring role in reducing incentives to manage earnings with cuts in
The absence of evidence that institutional ownership, as a who
investment behavior does not preclude the possibility that predominant
groups of institutions could increase incentives for managerial myo
into groups based on their past investment behavior to test whethe
(long) investment horizons encourage (discourage) myopic investme
indicate that predominant ownership by "transient" institutions-wh
turnover and engage in momentum trading strategies (i.e., buy (se
earnings news)-significantly increases the likelihood that manage
earnings. This result supports the assertion that heavy institutional tra
earnings leads to myopic investment behavior by corporate manage
tutions exhibiting such behavior have concentrated ownership posi
These findings contribute to the literature on the relation betw
and its investment in R&D. Prior research in this area tests for syst
in R&D by examining the cross-sectional relation between the level
R&D-to-sales) and the level of institutional holdings. The evidence o
though largely supporting a positive association between R&D inten
of institutional holdings (Graves 1988; Baysinger et al. 1991; Hans
and McConnell 1997). I extend this line of inquiry by testing for
investment in R&D in years when managers face a trade-off betw
and meeting earnings targets. In doing so, I present evidence that
institutional ownership and changes in R&D differs depending on the f
performance.
This paper also contributes to recent research focusing on the influence of institutional
investors on earnings management. Rajgopal and Venkatachalam (1997) present evidence
consistent with greater institutional ownership reducing the incidence of a lower-cost form
of earnings management: discretionary accruals. I focus on a more costly means of earnings
management, R&D cuts, which have real implications for long-term value and, therefore,
are of greater concern to equity holders like institutional investors. In a sample of the top
100 R&D spenders among NYSE/AMEX manufacturing firms, Bange and De Bondt (1997)
find that greater institutional ownership is negatively associated with unexpected changes
in R&D that reduce the anticipated gap between analysts' earnings forecasts and reported
net income. I provide evidence that this relation holds for a broader set of firms and that
the relation specifically holds for firms that reduce R&D to increase earnings. I also extend
this research by providing evidence that ownership by different groups of institutional
investors has differential influences on the use of R&D to manage earnings.
Finally, this paper makes two methodological contributions. First, prior research ex-
amining the use of R&D to manage earnings generally relies on simple proxies, such as
industry changes in R&D, for nondiscretionary changes in R&D (Dechow and Sloan 1991;
Holthausen et al. 1995). I use prior work on the determinants of R&D spending to develop
a more extensive model for the firm's set of profitable R&D opportunities and the costs of
reducing R&D to manage earnings, leaving an unexplained portion that is more likely to
capture the changes in R&D resulting from earnings management. Second, this study in-
troduces a methodology by which institutional investors can be parsimoniously classified
into groups based on past portfolio behavior (e.g., trading frequency, level of diversifica-
tion). By classifying directly on such characteristics, this approach creates more homoge-
neous groups of institutions and allows for richer predictions than the classification of

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 308 The Accounting Review, July 1998

institutions by type (e.g., pension fun

(Lang and McNichols 1997).
The next section develops hypotheses by examining arguments for how the beh
of institutional investors influences managerial incentives to engage in myopic R&D
vestment behavior. Section III presents the methodology used in the empirical tests
IV describes the sample and presents descriptive statistics. Section V reports results
empirical tests for institutional ownership as a whole, and Section VI reports result
ownership by different groups of institutions. Section VII presents a summary and
conclusions.

II. HYPOTHESIS DEVELOPMENT

Short-Term-Oriented Institutional Investor Behavior and Managerial Myopia
The first hypothesis tests the view that the short-term focus of institutional invest
trading creates incentives for myopic investment behavior. This view argues that manag
have incentives to reduce R&D in order to avoid an earnings disappointment that would
trigger large-scale institutional investor selling and lead to a temporary misvaluation of the
firm's stock price (Graves and Waddock 1990, 76-77; Jacobs 1991, 37-38; Porter 1992,
43-46). This hypothesis requires that (1) managers have incentives to avoid temporary
misvaluations and (2) institutional investor trading is sensitive to current earnings news and
can cause a temporary misvaluation.
Prior research identifies two conditions necessary to create incentives for myopic in-
vestment behavior to avoid temporary misvaluations (Stein 1988, 1989). First, managers
must place enough weight on current stock price so that they would not be willing to "ride
out" a temporary misvaluation. Such concern over current stock price could be driven by
(1) stock-based compensation and the need to make liquidity trades, (2) near-term equity
funding requirements, (3) the threat that a raider will exploit a temporary undervaluation,
or (4) the time horizons of influential investors (Froot, Perold, and Stein 1992, 50-55).
Second, managers must believe there is a potential for misvaluation based on current earn-
ings, either through market overreactions to unmanaged earnings declines or through market
underreactions to reductions in valuable R&D that artificially boost earnings. Support for
this condition is provided by Bernard et al. (1993), who cite anecdotal evidence that man-
agers believe the market is capable of overreacting to earnings news.
The presence of institutional investors increases the perceived potential for misvaluation
if they trade heavily based on current earnings news and if such trading causes misvalua-
tions. Prior research provides evidence supporting both of these conditions. Lang and
McNichols (1997) find that institutional trading is sensitive to earnings news, even after
controlling for return movements. Other studies report that increased institutional ownership
is associated with higher trading volume and stock return volatility around quarterly earn-
ings announcements, as well as temporary undervaluations after spin-offs (Kim et al. 1997;
Potter 1992; Brown and Brooke 1993).2 Such heavy institutional investor trading in response
to reported earnings could result from several market frictions. First, institutions subject to
strict fiduciary responsibilities have incentives to sell stocks with declining earnings because
fund sponsors and the courts use earnings as an objective criterion to judge the prudence
of an investment (Badrinath et al. 1989). Second, institutional investors sometimes use

2 Eames (1997) reports that annual changes in institutional ownership do not affect earnings response coefficients
(ERCs) in a manner consistent with institutional trading leading to earnings-based misvaluations. However, in
using annual data, his study does not capture the effect of institutional trading on ERCs in the days or weeks
surrounding earnings announcements, when trading-based misvaluations are most likely to occur.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 309

current earnings as a "value proxy" in their trading decisions

metry surrounding the quality of R&D spending (Froot, Pero
Porter 1992, 43). This asymmetry could arise if institutions h
periods and focus on predicting near-term price movements in
prospects (Froot, Scharfstein, and Stein 1992; Porter 1992, 43).
diversified portfolios with numerous stocks and small stakes i
term prospects of each is less cost-effective (Froot, Perold, and
43).
These arguments for the extreme sensitivity of institutional investor trading to current
earnings news support the hypothesis that, relative to individual investors, institutional in-
vestors create a short-term-oriented environment that encourages managers to manipulate
earnings through investment cuts:3

Hia: Ceteris paribus, a higher percentage of institutional holdings in a firm increases

the likelihood that its manager cuts R&D to meet short-term earnings goals.

Sophisticated Institutional Investor Behavior and Managerial Myopia

The competing hypothesis tests the view that the sophistication and large stockholdings
of institutional investors remove incentives for myopic investment behavior by providing a
higher degree of monitoring of managerial behavior. This monitoring can occur either ex-
plicitly, through governance activities, or implicitly, through information gathering and cor-
rectly pricing the impact of managerial decisions. Institutions that invest in firms with the
intention of holding substantial ownership blocks over a long horizon have strong incentives
to incur the cost of explicitly monitoring managers and ensure the firm does not cut prof-
itable R&D to meet short-term earnings goals (Dobrzynski 1993; Monks and Minow 1995,
chap. 2). The positive impact of such institutional monitoring is documented by Opler and
Sokobin (1997), who present evidence that poorly performing firms targeted by the Council
of Institutional Investors for shareholder activism have substantially improved profitability,
greater asset divestitures, and reduced acquisition activity in subsequent years.
Sophisticated institutional investors can implicitly monitor managerial investment be-
havior by gathering information concerning the quality of R&D, thereby reducing the op-
portunities of managers to prevent a negative market reaction to an expected earnings
decline through cuts in R&D. Hand (1990) uses institutional ownership to proxy for the
type of sophisticated investor that would not be "functionally fixated" on earnings (i.e.,
focusing only on reported earnings and being misled by firms' accounting choices).
Schipper (1989, 98) argues that:

Concentrated user groups with substantial financial sophistication, material sums at

stake, and no contractual friction to inhibit their behavior are, for example, likely can
didates for undoing earnings management.

Considering the size of their investments and their use of buy-side analysts, institutional
investors could fit this example and serve to remove incentives for myopic investment
behavior.

I Following Lang and McNichols (1997, 31), I assume that the other major category of investors in firms is
individual investors. Other possible ownership groups include insiders, arbitrageurs and corporations making
strategic investments. Agrawal and Knoeber (1996) find that the level of institutional holdings in a firm is not
significantly associated with the level of insider holdings and is weakly negatively associated with large (greater
than 5 percent) blockholders (which would include strategic corporate investments). Thus, institutional ownership
likely does not proxy for the level of ownership by these types of investors.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
310 The Accounting Review, July 1998

This view of the sophistication of institut

pothesis that, relative to individual investors,
that reduces managerial incentives to sacrifi

Hib: Ceteris paribus a higher percentage of

the likelihood that its manager cuts R&D

The Influence of Different Groups of Instit

Hypotheses la and lb concern the impact o
institutional investors differ in their behav
groups of institutions influence myopic investm
finding for overall institutional ownership.
importance of investor relation activities wi
consulting firms offering institutional investor
managers not only know their investor bas
behavior of institutions and sometimes actively
tors (Anand 1991; Elgin 1992).4 Lang and Mc
their type (e.g., investment advisers, bank trus
ences in portfolio turnover and earnings-bas
there can be substantive differences in trad
institutions, I classify institutional investors in
institutional investor behavior that have been
to manipulate earnings. This approach classi
dedicated and quasi-indexer-based on their p
folio turnover, diversification, and momentum
Porter (1992, 42-46) argues that pressures
by "transient" institutional owners, who ho
frequently in and out of stocks, generally
current earnings. The short-term focus of tran
group to sell out of a firm with disappointin
agers to avoid earnings disappointments. Thi

H2a: Ceteris Paribus, if a firm's institution

institutions that are transient, then its ma
short-term earnings goals.

In contrast, Porter (1992, 46-49) argues th

ascribes to Germany and Japan) alleviate pre
their large, long-term holdings, which are co
tives to monitor managers and to rely on i
managers' performance. The monitoring role
hypothesis:

I For example, the consulting firm Georgeson Co. analyzes the portfolios of "almost 1,500 domestic and foreign
institutions" in order to characterize a client's investor base and assist managers in "screening out" institutions
with high turnover, index, or quantitative strategies in favor of institutions with long investment horizons and
holdings in companies with similar characteristics to the client (Elgin 1992).
5 This classification method does not proxy for classifications based on type, investment styles, or institutional
size. Institutions in the transient, quasi-indexer and dedicated groups all exhibit significant heterogeneity across
these other possible classification schemes. Because of this heterogeneity, I find no significant differences in
results of the myopic behavior tests across groups of institutions formed by these other methods.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 311

H2b: Ceteris paribus, if a firm's institutional ownership is

institutions that are dedicated, then its manager is less likely to cut R&D to meet
short-term earnings goals.

The third group of institutional investors, "quasi-indexers," use indexing or buy-and-

hold strategies that are characterized by high diversification and low portfolio turnover.
Porter (1992, 42-46) claims that the presence of quasi-indexers exacerbates incentives for
myopic investment behavior because their passive, fragmented ownership leads them to
gather little information on the company and provides little incentive to monitor managers.
This type of institutional ownership, in effect, abdicates its potential influence over man-
agers to other types of active investors who could pressure managers into focusing on short-
run earnings goals. However, Monks and Minow (1995, chap. 2) espouse the opposite
viewpoint that institutions that are unable to sell because of their indexing strategies have
strong incentives to monitor management to ensure it is acting in the long-term interests
of the firm. Because of these contrasting views, I do not make a formal hypothesis about
the influence of quasi-indexers, but include them in the analysis for descriptive purposes.

III. METHODOLOGY
Research Design
To test whether managers reduce R&D to meet earnings goals, I assume that managers
use last year's earnings per share as their earnings target. The business press uses this
benchmark to analyze current earnings performance and prior research provides evidence
that managers have incentives to avoid decreases in earnings (Burgstahler and Dichev 1997;
Barth et al. 1997).6 I divide the sample into three subsamples based on the relationship
between pre-tax, pre-R&D earnings changes and prior levels of R&D.7 I test for myopic
investment behavior in a sample of firms, which I call the "Small Decrease" (SD) sample,
whose earnings before R&D and taxes have declined relative to the prior year, but by an
amount that can be reversed by a reduction in R&D. This specification assumes that man-
agers have unbiased expectations of pre-tax, pre-R&D earnings early enough in the fiscal
year to make decisions to reduce R&D expenditures. Burgstahler and Dichev (1997) find
that firms with small earnings decreases are underrepresented in the distribution of observed
earnings changes, indicating that firms below but near last year's earnings are managing
earnings upward. Baber et al. (1991) report that such firms increase R&D by less, on
average, than firms with both increases or large decreases in earnings before investment,
which they cite as evidence that managers of firms with small earnings declines use R&D
cuts to increase earnings.
For comparison purposes, I also create "Increase" (IN) and "Large Decrease" (LD)
samples. The IN sample contains all firms with positive changes in pre-tax, pre-R&D earn-
ings; these firms could maintain last year's R&D and still have an increase in pre-tax
earnings. The LD sample consists of firms with declines in pre-tax, pre-R&D earnings
greater than the amount of prior year's R&D; these firms could eliminate R&D spending

6 I chose not to use analysts' forecasts as the earnings target because this benchmark would (1) bias the sample
toward firms with greater levels of institutional ownership (O'Brien and Bhushan 1990), (2) require an estimate
of the analysts' expectation of R&D to get expected pre-R&D earnings, and (3) create a specification problem
if analysts anticipate earnings management, as suggested by the findings of Burgstahler and Eames (1998).
7 I use pre-tax earnings, rather than after-tax earnings, to make the results comparable to prior research (Baber et
al. 1991) and to avoid introducing error in the estimation of marginal tax rates to obtain after-tax R&D. As a
check, I estimated after-tax R&D using three tax rate assumptions: (1) the statutory rate, (2) the effective tax
rate, and (3) the effective tax rate adjusted for the estimated R&D credit. The correlations between sample
assignments across these three methods and the before-tax method were all over 0.85.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 312 The Accounting Review, July 1998

and still report a decrease in pre-tax earnings. Because managers of IN and LD firms do
not have the same incentives to reduce R&D to meet the earnings target, the relation
between institutional ownership and R&D cuts in these samples should differ from the
relation in the SD sample if the relation is driven by institutional investor pressure for or
against myopic investment behavior.
The dependent variable I use to test for myopic investment behavior is an indicator
variable (CUTRD) that equals one if the firm cuts R&D relative to the prior year, and zero
if the firm maintains or increases R&D. I use an indicator variable rather than the magnitude
of the change in R&D because theory does not address how the level of institutional
ownership should affect the magnitude of increases in R&D, and because the magnitude
of decreases in R&D should be based on the amount of the earnings decline. Thus, the
magnitude of the change in R&D is unlikely to be a linear function of the level of insti-
tutional ownership.8
To test hypotheses la and lb on the influence of institutional ownership on myopic
investment behavior, I use a logit model to regress CUTRD on the level of the firm's
institutional ownership and a set of control variables (discussed below) that proxy for
changes in the R&D investment opportunity set, differences in the costs of cutting R&D,
and proxies for other earnings management incentives:

Prob(CUTRDI = 1) = F(o + PIPCRDi + I2CIRDi + I3CGDPi + I4TOBQi

+ I5CCAPXj + I6CSALESj + I7SIZEj + I8DISTj + I9LEVi
+ I3OFCFi + I11PIHj) (1)

where F(-) is the logistic c.d.f.

CUTRD = 1 if R&D is cut relative to the prior year, 0 otherwise;

PCRD = prior year's change in R&D;
CIRD = change in industry R&D-to-sales ratio, defined using 4-digit SIC codes;
CGDP = change in gross domestic product;
TOBQ = Tobin's q (the sum of market value of common equity, book value of
preferred equity, long-term debt, and short-term debt, divided by total
assets);
CCAPX = change in capital expenditures;
CSALES = change in sales;
SIZE = market value of equity;
DIST = distance from earnings goal relative to prior year's R&D, defined as the
change in pre-tax, pre-R&D earnings divided by prior year's R&D;
LEV = leverage (debt-to-assets);
FCF = free cash flow (cash flow from operations less capital expenditures, scaled
by net tangible assets);
PIH = percentage of institutional holdings.

To test hypotheses 2a and 2b on the effect of different groups of institutions on myopic

8 I also considered using an indicator variable for whether the firm successfully reversed the earnings decline with
R&D cuts. I chose not to use this variable because I could not apply the model out of the SD sample as a
robustness check. I reran the analysis with this indicator variable in the SD sample and the results for the
institutional holdings variable were unchanged.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 Bushee-Institutional Investors and Myopic R&D Behavior 313

investment behavior, I first restrict the sample to firms with at least f

ownership to ensure that institutional ownership is likely to be influen
I add indicator variables for the composition of institutional owner
indicator variable for each group (transient, quasi-indexer, dedicat
firm's institutional ownership is dominated by this group, and zer
fication more powerfully identifies firms in which a given group is
(compared to the percentage ownership by each group) and avoids
centage ownership by the three groups is highly correlated. This ap
lowing regression:

Prob(CUTRDi - 1) = F(t + I31PCRDi + I32CIRDi + I33CGDPi + I34TOBQi

+ I35CCAPXN + 36CSALESi + I37SIZEi + I38DISTi + I39LEVi
+ I31QFCFi + I11PIHi + I12DQ5(Transient)i
+ 1313DQ5(Quasi-Indexer)i + I314DQ5(Dedicated)i) (2)

where DQ5(GROUPk)A = 1 if firm i is in the top quintile of proportional ownership by

group k, and zero otherwise; GROUPk = Transient, Quasi-Indexer, Dedicated.

Control Variables
To control for motivations to cut R&D that are unrelated to myopic investment behavior
incentives, I include proxies for the R&D investment opportunity set (i.e., the set of av
able positive NPV projects in which the firm could invest). These proxies are based on t
expectations model for the level of R&D intensity developed in Berger (1993). Except
Tobin's q, I take first differences in his variables to obtain a model for the change in R
Each of the variables is defined to have a negative expected relation with the decision
cut R&D (see table 1 for variable definitions).9
The prior year's change in R&D (PCRD) proxies for changes in the firm's R&D op-
portunity set over time (Berger 1993). If a firm's R&D opportunity set has been declining
(increasing) over time, a decline (increase) in R&D last year makes a firm more (less) likely
to cut R&D this year.'0 The change in industry R&D intensity (CIRD) captures changes in
the R&D opportunity set within the firm's industry and changes in the level of R&D
spending needed to stay competitive within the industry (Berger 1993). Both explanations
suggest that firms in industries with increasing (decreasing) R&D are expected to be less
(more) likely to cut R&D. Changes in the Gross Domestic Product (CGDP) measure growth
in the overall economy and proxy for increases in the level of technological progress in
the economy (Berger 1993). In years when CGDP is high (low), firms are expected to have
more (fewer) profitable R&D opportunities and thus be less (more) likely to reduce R&D.
Berger (1993) includes Tobin's q (TOBQ) to capture the marginal benefit-to-cost ratio of
undertaking new investment. As a proxy for the unobservable marginal TOBQ, Berger
(1993) uses the average TOBQ, defined as the market value of the firm's equity and debt,

I I drop two variables from his model: a measure of funds available for investment and an R&D tax credit usability
variable. I drop the funds available variable because it is pre-R&D earnings plus depreciation, which is similar
to what I use to partition the sample. I drop the credit usability variable because it does not explain cuts in
R&D and it requires three years of data on deferred tax liability, thereby reducing the sample size. Reed and
Plumlee (1998) find that tax motivations only lead to cuts in R&D when the firm's earnings are substantially
above the prior year's level.
10 An alternative explanation is that if the firm reduced R&D in the prior year, it becomes more costly to cut R&D
again, indicating a positive relation between PCRD and CUTRD.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 314 The Accounting Review, July 1998

TABLE 1
Definitions of Variables

Variable Definition Expected Signa

Indicator for cut in R&D (CUTRD) 1 if (RDt-RDt l) < 0,

0 otherwise

Prior change in R&D (PCRD) ln(RDt- )-ln(RDt-2)

Change in industry R&D intensity ln(IRD /ISALESt)-ln(IRDt-1/ISALEStj) -

(CIRD)

Change in capital expenditures ln(CAPXt)-ln(CAPXt- )

(CCAPX)

Change in sales (CSALES) ln(SALESt)-ln(SALESt-1)

Tobin's q (TOBQ) TOBQ,

Change in gross domestic product ln(GDPt)-ln(GDPt- )

(CGDP)

Firm size (SIZE) ln(MVEt)

Distance from earnings goal (EBTRDt-EBTRDt_ 0)/RDt_ I +

(DIST)

Leverage (LEV) (LTDt+STDt)/ASSETSt +

Free cash flow (FCF) (CFO -Average CAPXtlj tt_)/CAtj

Percentage of institutional holdings ISHARESt/TSHARESt +/-
(PIH)

Indicator for predominant 1 if PIH(GROUPk)/PIH in top quintile of +/-

ownership by group k distribution,
(DQ5(GROUPk)), k = quasi- 0 otherwise
indexer, transient, dedicated

RDt = Research and development per share (46/54)b

IRDt = Total research and development for all firms in same 4-digit SIC code, excluding
ISALESt = Total sales for all firms in same 4-digit SIC code, excluding the firm
CAPXt = Capital expenditures per share (30/54)
SALESt = Sales per share (12/54)
TOBQt = Tobin's q [(Market value of equity + preferred stock + total debt)/total
+ 130+9+34)/6]
GDPt = Gross domestic product
MVEt = Market value of equity (199*25)
EBTRDt = Earnings per share before taxes and R&D ((170+46)/54)
LTDt = Long-term debt (34)
STDt = Short-term debt (9)
ASSETSt = Total assets (6)
CFOt = Cash from operations (18-A4+A5+A1-A34+ 14)
CAt = Current assets (4)
ISHARESt = Total shares held by institutional investors
TSHARESt = Total shares outstanding
PIH(GROUPk) = Percentage institutional ownership by group k, k = Quasi-indexer, transient, dedicat
a This column lists each variable's expected relation to CUTRD.
b Compustat numbers are in parentheses.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 315

divided by the book value of its assets. Because of the accountin

measure is similar to the market-to-book ratio in that both are
the future value of R&D spending. Based on these interpretation
TOBQ have more (fewer) valuable R&D opportunities and face a higher (lower) cost of
reducing R&D for myopic reasons."
I include additional variables not found in Berger's (1993) model to control for other
R&D opportunity set measures, for the costs of cutting R&D to manage earnings, and for
other incentives to manage earnings. The change in capital expenditures (CCAPX) proxies
for reduced funds available for investment and/or entry into a more mature, lower invest-
ment stage of the firm's life cycle (Perry and Grinaker 1994). Both explanations yield an
expected negative relation with CUTRD. The change in sales (CSALES) is not included in
Berger's (1993) model because he is estimating the R&D-to-sales ratio. I include CSALES
to proxy for firm growth and funds available for R&D investment, as well as to capture
the fact that R&D budgets are often based on sales (Berger 1993). For each of these reasons
a negative relation with the decision to cut R&D is expected.
A size variable (SIZE), the log of market value of equity, proxies for two possible
effects, both of which lead to a negative expected relation with CUTRD. First, SIZE proxies
for the amount of information available about the firm (Wiedman 1996). Larger (smaller)
firms have richer (poorer) information environments that should reduce (increase) oppor-
tunities for successful earnings management with R&D. Second, SIZE proxies for the like-
lihood the firm faces cash constraints (Jalilvand and Harris 1984). Smaller firms are more
likely to suffer cash flow shortages that force them to reduce R&D.'2
The distance of the expected earnings from the earnings target (DIST) measures the
percentage of R&D that would need to be cut to obtain a positive earnings change. As a
firm's expected earnings get farther from (closer to) the earnings goal, the cost of cutting
R&D to manage earnings increases (decreases). This variable is defined as the change in
pre-tax, pre-R&D earnings divided by prior year's R&D. For firms in the SD sample, this
variable ranges between 0 and -1. A firm with a value of DIST equal to -0.8 (-0.2)
would need to reduce its R&D by 80 percent (20 percent) relative to the prior year to
reverse the decline in earnings. Firms with high (less negative) values of DIST have the
ability to meet their earnings target through less severe cuts in R&D (e.g., reductions in
the number of prototypes or test runs and/or elimination of basic research projects); whereas
firms with lower (more negative) values of DIST need to make more substantial cuts in

I use the level of Tobin's q rather than the change in Tobin's q because the latter measure is highly correlated
with changes in the market value of equity. Since such changes can be driven by factors other than changes in
the R&D opportunity set (such as market reactions to the decision to cut R&D), the construct validity of the
levels measure is likely to be higher. If I replace TOBQ with the change in Tobin's q, the coefficient on the
latter is positive and significant, which is opposite of what is expected and likely reflects the construct validity
problem. I also replaced TOBQ with the market-to-book ratio and found similar results. Neither of the changes
affects the sign or significance of the institutional holdings variables.
12 SIZE could also proxy for two effects that lead to a positive relation with CUTRD. First, smaller firms generally
have larger managerial holdings, which reduces incentives to manage earnings based on capital market pressures
(Warfield et al. 1995). Second, smaller firms tend to have higher (lower) percentages of new product research
(basic research) in their R&D budgets (Mansfield 1981). To the extent that it is more (less) costly to cut R&D
for new products (basic research), smaller firms have less flexibility to cut R&D to manage earnings
(Schneiderman 1991). In both these cases, SIZE is a fairly weak proxy for the underlying effect, making a
positive relation between SIZE and CUTRD less likely.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 316 The Accounting Review, July 1998

R&D (e.g., personnel and equipment) to meet their earnings goals. Thus, a positive relation
between DIST and CUTRD is expected.'3
The firm's leverage ratio (LEV) captures potential debt covenant incentives to manage
earnings (Duke and Hunt 1990). A positive sign is expected if meeting debt covenants
motivates earnings management. Higher LEV could also indicate fewer growth opportuni-
ties for the firm (Myers 1984), in which case a positive sign would also be expected. Finally,
I include a measure of free cash flow (FCF), defined as cash flow from operations less
capital expenditures, scaled by net tangible assets, to proxy for possible near-term financing
requirements. Firms with substantially negative free cash flows have a greater need to raise
equity in the near-term, and thus have incentives to boost earnings and reduce the chance
of an undervaluation (Dechow et al. 1996). This variable also proxies for reduced funds
available for investment, which might trigger reductions in R&D. Either explanation indi-
cates that a negative relation with CUTRD is expected.'4

Institutional Ownership Variables

I measure the percentage of institutional holdings (PIH) as total shares held by insti-
tutions divided by total shares outstanding. PIH is calculated at the end of the calendar
quarter in which the firm's third fiscal quarter ends. This date is chosen to measure insti-
tutional ownership in the middle of the second half of the fiscal year, when I assume the
manager likely has an accurate expectation of annual earnings and begins to consider my-
opic investment behavior decisions.'5
To obtain variables on the percentage of ownership by different groups of institutions,
I first use factor analysis and cluster analysis to assign institutions into groups based on
their past investment behavior (see section VI for details of this procedure). Then, I calculate
the proportion of ownership held by each group of institutions for each firm (e.g.,
PIH(GROUPk)A/PIHj), where the groups are transient, dedicated, and quasi-indexer. I ran
firms into quintiles based on the distribution of this proportion for each group and create
an indicator variable (DQ5(GROUPk)A) that equals one if the firm is in the top quintile of
proportional ownership by group k, and zero otherwise.6

13 The DIST variable captures other incentives outside of the SD sample. In the IN sample, DIST is always positive,
and large values indicate a large positive change in pre-R&D earnings relative to prior R&D. If managers have
incentives to increase R&D to dampen earnings growth, a positive sign should result. In the LD sample, DIST
is always less than - 1, and small values indicate a large negative change in pre-R&D earnings relative to prior
R&D. If managers have incentives to increase R&D to take a "big bath" in earnings during the year, a negative
relation should result.
14 I chose not to include a proxy for bonus motivations because Holthausen et al. (1995) provide evidence that
R&D is not managed for bonus reasons and because hand collection of a bonus plan proxy would significantly
reduce the sample size.
'5 To obtain some descriptive evidence on this assumption, I examined quarterly R&D disclosures on Compustat.
Compustat reports quarterly R&D for years after 1989, but quarterly R&D disclosures are voluntary and are
only reported by approximately 40 percent of the sample firms. For firms in the SD sample, the mean seasonal
percentage changes in R&D by quarter are 10.8 percent for the first quarter, 8.7 percent for the second, 7.3
percent for the third, and 4.8 percent for the fourth. For firms in the IN (LD) sample, the comparable changes
are 12.1 percent, 12.8 percent, 14.6 percent and 14.8 percent (12.6 percent, 10.7 percent, 5.6 percent, and 7.0
percent), respectively. This evidence indicates that some SD sample firms are able to reduce R&D late in the
fiscal year.
16 I use quintiles because they yield extreme proportional differences while still maintaining a reasonable number
of observations in the top quintile. I performed the ranking, and the subsequent analyses, using treciles and
quartiles. The results are similar, though less significant, using the less extreme fractiles. I also formed the
quintiles after ranking on SIZE and PIH to control for any association between extreme ownership by groups
and those variables. The results were insensitive to this approach.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 317

IV. SAMPLE AND DESCRIPTIVE STATISTICS

The sample includes all firm-years between 1983 and 1994 with available da
sample period is restricted by the availability of institutional holdings data and t
ment of two years of data from which to calculate portfolio characteristics of in
I collect institutional holdings data from the University of Michigan Spectrum
which contains all 13-f filings between September 30, 1981 and December 31, 1
cording to SEC Rule 13-f, all institutions managing more than $100 million in e
file a quarterly report listing all equity holdings that are greater than 10,000
$200,000 in market value. Thus, for each firm, total institutional holdings is def
sum of all end-of-calendar-quarter holdings by fund managers filing 13-fs on t
gather total shares outstanding from the 1995 CRSP Daily NYSE, AMEX and NA
files, and GDP data from the 1995 Economic Report of the President. All other
are obtained from the 1995 Compustat PST, Full Coverage and Research files.
Panel A of table 2 shows the criteria used to select the sample. I exclude fir
observations for which: (1) current or prior year R&D data is missing, (2) the ratio of
R&D-to-sales is less than 1 percent (i.e., R&D is not a significant portion of earnings),

TABLE 2
Sample Sizes

Panel A: Sample Selection Criteria

Criteria Firm-Years

Compustat firm-years with nonmissing R&D, 30,324

1983-1994

Less: Missing prior year R&D data (3,754)

R&D/sales less than 1% (6,135)
Fewer than three other firms in industry (131)
Missing data to calculate independent variables (4,165)
Missing Spectrum institutional holdings data (2,195)
Final sample 13,944

Panel B: Number of Obs

(IN) and Large Decrease (LD) Sampler

Samples

SD Sample IN Sample LD Sample Total

Cut R&Db 852 (34.5%) 2,190 (25.7%) 1,041 (35.2%) 4083

Increase R&D 1,617 (65.5%) 6,330 (74.3%) 1,914 (64.8%) 9861

Total 2,469 8,520 2,955 13,944

a The SD sample consists of all firm-y

of all firm-years for which (EBTRD
(EBTRDt-EBTRDt_1) < -RDt-1. EBTR
b'The Cut (Increase) sample consists
(nonnegative).

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 318 The Accounting Review, July 1998

(3) fewer than three other firms are availab

industry R&D intensity, (4) data needed to c
equal to zero (which prevents the calculation
and (5) institutional ownership data is not a
of 13,994 firm-years.
Panel B of table 2 summarizes the number of observations in the SD, LD and IN
samples, and the number of firm-years for which R&D was cut in each case.'7 The table
shows a significant percentage of firm-year observations with R&D cuts in each sample,
and that the number of firm-year observations with R&D cuts in the SD and LD samples
are roughly equivalent. It also shows that a majority of firm-year observations in the SD
sample do not cut R&D and, hence, do not boost earnings through R&D cuts.18 These
findings indicate that managerial myopia incentives do not completely explain differences
in R&D investment behavior across the samples. This observation motivates the inclusion
of the controls for opportunity set differences, other earnings management proxies, and
institutional ownership variables to explain cross-sectional differences in the decision to cut
R&D.
Panel A of table 3 presents means and standard deviations for the variables use
empirical tests for the SD, IN and LD samples. The percentage change in R&D (C
significantly less positive in the SD sample than in either of the other two sampl
sistent with the results of Baber et al. (1991). For most of the other variables, the
in the SD sample lie between the means in the other two samples, as might be e
given the differences in earnings performance across samples.
Panel B of table 3 compares the means of these variables for firm-year obse
with R&D cuts and R&D increases in the SD sample. These comparisons provide u
evidence on myopic investment behavior. The significant differences for PCRD,
TOBQ, CCAPX and CSALES indicate that reduced opportunity sets significantly
reductions in R&D. Firms that cut R&D in the SD sample are smaller and more h
leveraged, as expected based on the arguments in the prior section. The percent
institutional holdings is significantly lower for firms in the SD sample that cut
dicating that higher levels of institutional ownership make managers less likely to
cuts to manage earnings. However, the correlation between PIH and SIZE is high (r
= 0.57); thus, the relation between institutional holdings and the decision to cut R&D must
be examined conditional on SIZE.'9

17 I examined the sample by industry and by year to check for any significant clustering of observations. The
majority of the sample firms are in manufacturing, pharmaceuticals/chemicals, and software development in-
dustries. The proportion of firms in each industry that cut R&D when in the SD sample does not deviate greatly
from the proportion of total observations in the industry. The percentage of firms cutting R&D by year ranges
from 23 percent (in 1984) to 41 percent (in 1986 and 1987). Other than these three years, there does not appear
to be significant time clustering in R&D reductions.
18 Because the sample design ignores other methods of managing earnings (e.g., accrual manipulation), these
findings likely understate the amount of earnings management that occurs in firms with small declines in
earnings. Some firms with expected declines in earnings might have managed earnings using other, less costly
means to get into the IN sample. Thus, the SD sample is potentially biased toward firms that lack the discretion
in their reporting or economic decisions to boost earnings, thereby reducing the power of the tests.
Other than this correlation, the only pairwise correlations larger than 0.20 between independent variables are
the correlations between TOBQ and FCF (r = -.34), CSALES and CCAPX (r = 0.27), and TOBQ and PCRD
(r = 0.24). Thus, the independent variables appear to be proxying for different effects. To assess the potential
impact of more complex forms of multicollinearity, I examined variance inflation factors (VIF) for each variable.
VIFs are based on the R2 from regressing each independent variable on all other independent variables. A VIF
of I indicates no multicollinearity and a VIF in excess of 10 indicates harmful multicollinearity (Kennedy 1992,
183). All of the VIFs were less than 1.7, indicating that multicollinearity is not a problem in this sample.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 319

V. RESULTS FOR INSTITUTIONAL OWNERSHIP AS A WHOLE

Main Results
Table 4 presents results for the logit regression of the indicator for cuts in R&D
(CUTRD) on the control variables and the percentage of institutional ownership (PIH) (see
equation (1)). The first column presents coefficients and p-values for the SD sample. Con-
sistent with the univariate tests, most of the opportunity set variables (PCRD, CIRD,

TABLE 3
Descriptive Statistics

Panel A: Descriptive Statistics for Small Decrease (SD), Increase (IN), and Large Decrease (LD)
Samples

Sample Differences in Means

Variable SD IN LD (SD)-(IN) (SD)-(LD)

CUTRDb Mean 0.345 0.257 0.352 0.088** -0.007

Std. dev. 0.475 0.437 0.478

CRD Mean 0.049 0.146 0.121 -0.097** -0.072**

Std. dev. 0.330 0.383 0.469

PCRD Mean 0.167 0.150 0.157 0.017* 0.010

Std. dev. 0.349 0.418 0.506

CIRD Mean 0.110 0.110 0.091 0.000 0.019*

Std. dev. 0.257 0.250 0.283

CGDP Mean 0.027 0.030 0.027 -0.003** 0.000

Std. dev. 0.017 0.017 0.018

TOBQ Mean 1.536 1.856 1.378 -0.320** 0.158**

Std. dev. 1.741 1.896 1.593

CCAPX Mean -0.019 0.106 -0.108 -0.125** 0.089**

Std. dev. 0.784 0.824 0.895

CSALES Mean 0.027 0.181 -0.068 -0.154** 0.095**

Std. dev. 0.243 0.271 0.315

SIZE Mean 4.300 4.725 3.724 -0.425** 0.576**

Std. dev. 2.173 2.190 2.041

DIST Mean -0.417 2.279 -4.949 -2.696** 4.532**

Std. dev. 0.283 9.335 12.921

LEV Mean 0.184 0.180 0.251 0.004 -0.067**

Std. dev. 0.178 0.172 0.227

FCF Mean -0.156 -0.054 -0.257 -0.102** 0.101**

Std. dev. 0.417 0.453 0.525

PIH Mean 0.233 0.264 0.190 -0.031** 0.043**

Std. dev. 0.208 0.228 0.192

N 2469 8520 2955

(Continued on next page)

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 320 The Accounting Review, July 1998

TABLE 3 (Continued)

Panel B: Descriptive Statistics for SD Sample by Cut or Increase in R&D

Differences in
Samples Means

Variable Cut Increase (Cut)-(Increase) Expected Signd

CRD Mean -0.248 0.206 -0.454**

Std. dev. 0.314 0.208

PCRD Mean 0.124 0.190 -0.066**

Std. dev. 0.367 0.337

CIRD Mean 0.079 0.127 -0.048**

Std. dev. 0.256 0.256

CGDP Mean 0.026 0.027 -0.001

Std. dev. 0.016 0.017

TOBQ Mean 1.420 1.598 -0.178**

Std. dev. 1.759 1.729

CCAPX Mean -0.225 0.089 -0.314**

Std. dev. 0.794 0.756

CSALES Mean -0.035 0.060 -0.095**

Std. dev. 0.249 0.233

SIZE Mean 3.938 4.491 -0.557**

Std. dev. 2.242 2.111

DIST Mean -0.424 -0.414 -0.010 +

Std. dev. 0.281 0.284

LEV Mean 0.202 0.174 0.028** +

Std. dev. 0.205 0.161

FCF Mean -0.156 -0.156 0.000

Std. dev. 0.414 0.418

PIH Mean 0.201 0.250 -0.049** +/-

Std. dev. 0.198 0.212

N 852 1617

* Significantly different
** Significantly differen
a The SD sample consist
of all firm-years for w
(EBTRD,-EBTRD,_1) < -
b CRD is the log change
cThe Cut (Increase) sam
(nonnegative).
d This column lists each variable's expected relation to CUTRD.

CCAPX, CSALES) are significantly associated with the probability of cutting R&D in the
direction anticipated. The coefficient on SIZE indicates that smaller firms are significantly
more likely to cut R&D, consistent with the explanations that small firms are more cash

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05: Jan 1976 12:34:56 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 321

TABLE 4
Logit Regression of Indicator for Cut in R&D (CUTRD) on Control Variables
Institutional Holdings

Prob(CUTRDi = 1) = F(a + 3JPCRDi + P2CIRDj + P3CGDPi + P4TOBQj

+ 35CCAPX, + 36CSALESi + 37SIZER + 38DIST, + 9LEV,
+ 3J0FCF, + 311PIHi)

SD Samplea IN Sample LD Sample

Expected Coefficient Marginal Coefficient Marginal Coefficient Marginal
Variable Sign (p-value) Effectsc (p-value) Effects (p-value) Effects
Interceptb -0.158 0.372 0.071
(0.323) (0.000) (0.596)

PCRD - -0.448 -0.029 -0.335 -0.018 -0.249 -0.020

(0.001) (0.000) (0.005)

CIRD - -0.677 -0.027 -0.706 -0.020 -0.600 -0.024

(0.000) (0.000) (0.000)
CGDP - 0.399 0.001 -3.398 -0.005 -6.138 -0.020
(0.882) (0.036) (0.008)

TOBQ - -0.020 -0.004 0.018 0.004 -0.066 -0.001

(0.497) (0.300) (0.044)

CCAPX - -0.443 -0.075 -0.404 -0.054 -0.219 -0.044

(0.000) (0.000) (0.000)

CSALES - -1.464 -0.056 -2.161 -0.082 -1.526 -0.087

(0.000) (0.000) (0.000)

SIZE - -0.054 -0.030 -0.226 -0.117 -0.055 -0.031

(0.031) (0.000) (0.033)

DIST +d 0.072 0.007 -0.036 -0.012 0.061 0.043

(0.649) (0.000) (0.000)

LEV + 0.782 0.041 0.769 0.030 0.613 0.038

(0.002) (0.000) (0.002)

FCF - 0.064 0.004 -0.349 -0.015 -0.076 -0.005

(0.595) (0.000) (0.479)
PIH +/- -0.928 -0.069 -0.030 -0.002 -0.412 -0.026
(0.001) (0.859) (0.144)

pseudo-R2 0.122 0.220 0.121

N 2469 8520 2955

a The SD sample consists of a

of all firm-years for which
(EBTRDt-EBTRDt_) < -RDt-.
b All variables are defined in
The marginal effect represen
independent variable over the
dThe expected sign for DIST o
managers of firms in the IN

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 322 The Accounting Review, July 1998

constrained and/or have greater information asymmetries which allow them to use R&D
to manage earnings. The coefficient on LEV is significantly positive, consistent with earn-
ings management based on debt covenant incentives and/or reduced growth opportunities.
Neither free cash flow (FCF) nor the distance from the earnings goal (DIST) variable is
significant.
Turning to the main result, the coefficient on the PIH variable is significantly negative.
This result supports Hlb, suggesting that institutional investors are sophisticated investors
who serve a monitoring role, relative to individual investors, in reducing the incentives of
managers to reverse a decline in earnings through cuts in R&D. This finding does not
support the concern, expressed in Hla, that institutional investors cause underinvestment in
R&D based on their expected reactions to disappointing current earnings news.
The second column of table 4 presents the estimated impact of a change in each
independent variable across its interquartile range on the probability of cutting R&D. To
obtain this impact, I multiply the marginal effect of a one-unit change in the independent
variable (which is calculated as the estimated coefficient times the logit density function
evaluated at the sample means of the independent variables) by the interquartile range for
each variable. A change in institutional holdings across the interquartile range (from 5
percent to 38 prcent) reduces the probability that a manager cuts R&D by 6.9 percent. This
effect appears economically significant, given that about 30 percent of all firms cut R&D
in any given year. Only the CCAPX variable has a larger estimated impact in this sample
than the PIH variable.
A possible explanation for the negative relation between institutional ownership and
the decision to cut R&D is that institutions prefer to invest in more innovative firms that
are unlikely to cut R&D in any circumstance. If this were the case, the negative relation
should persist outside of the SD sample. Columns 3-6 of table 4 present results for the
logit model in the IN and LD samples. The percentage of institutional holdings has no
significant influence on the decision to cut R&D out of the SD sample. These results suggest
that institutional ownership exerts a greater influence on R&D investment decisions when
managers face a trade-off between maintaining R&D and increasing earnings.
Columns 3-6 of table 4 also indicate that the coefficients on the opportunity set controls
retain their sign and significance in the IN and LD samples, indicating that these variables
explain R&D changes equally well across firm-years with different earnings performance.
The coefficients on the SIZE and LEV variables also maintain their significance regardless
of the realization of current year earnings. Thus, these variables are likely capturing op-
portunity set determinants of R&D spending rather than proxying for earnings management
motivations. Finally, the coefficient on the DIST variable is significant in the IN and LD
samples. In the IN sample, the negative coefficient on DIST indicates that firm-years with
the largest (smallest) positive changes in earnings relative to prior R&D expense are less
(more) likely to cut R&D, which is consistent with the best performers dampening their
earnings growth through increased R&D spending. In the LD sample, the positive coeffi-
cient on DIST suggests that extremely poor performers are less likely to cut R&D. This
indicates that these firms either take a "big bath" with increased R&D spending because
they are far below their earnings target or that these are firms in an early high-investment-
low-earnings stage of their life cycle.

Robustness Tests
Another possible explanation for the negative relation between institutional owner
and the decision to cut R&D in the SD sample is that, based on past appearances by fi
in this sample, institutional investors invest in firms that are less likely to cut R&D
faced with a small decline in earnings. To test this possibility, I examine whether firm

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 323

made repeat occurrences in the SD sample drive the results in

of the firm-year observations in the SD sample are also in the s
an additional 25 percent are in the SD sample an additional two
into firms that made multiple appearances in the SD sample an
single appearance, and run the regression in equation (1) in each partition. The coefficient
on PIH is negative and significant at the 0.05 level in both partitions (not reported). Thus,
the results in table 4 are not driven by firms that repeatedly fall in the SD sample.
Because institutional ownership is positively associated with the level of analyst fol-
lowing (O'Brien and Bhushan 1990), another possible explanation for the results in table
4 is that managers are less likely to cut R&D given a large analyst following. Analyst
following proxies for a richer information environment for the firm and less uncertainty
about the quality of investment spending. As an additional control variable, I added the
number of analysts following the firm, defined as the number of analysts issuing a forecast
of annual earnings on the I/BIEIS database at the time that institutional holdings are
collected.20 The coefficient on this variable was insignificant, and the sign and significance
of the PIH coefficient was unaffected by the inclusion of this variable.
I tested whether the results are sensitive to the definition of the SD sample, specifically
the assumptions that prior year's earnings is the target and that prior year's R&D is the
maximum amount of R&D that could be cut to meet the target. I modified the tests to (1)
specify the lower bound of the SD sample to be pre-R&D earnings changes equal to one-
half, one-third, one-fourth, one-fifth, and one-tenth of prior year's R&D, (2) define the
earnings target using a random walk with drift model, and (3) use the Hunt et al. (1996)
time trend algorithm for determining the earnings target. In each case, the coefficient of
the PIH variable is significant at the 0.01 level in the SD sample and insignificant in the
other samples.2'
I added additional variables to control for potential differences in the costs of cutting
R&D across firms. Some R&D projects, such as basic research, are easier for managers to
justify dropping because of the lack of immediate benefits and highly uncertain future
payoffs (Schneiderman 1991). Since no public data are available regarding the composition
of R&D expenditures of individual firms, I use proxies for differences in the average com-
position of R&D across industries based on survey results from Mansfield (1981). I include
indicator variables for firms in industries in which R&D has a higher component of (1)
basic research, (2) long-term projects, (3) new projects, and (4) risky projects. As an alter-
native approach, I add an indicator variable for firms in "high-tech" industries, as defined
by Chan et al. (1990). They find that the market reacts positively to R&D increases for
high-tech firms, but not for "low-tech" firms, indicating that firms in low-tech industries
generally have less valuable R&D and would be more likely to cut R&D to meet earnings
goals. None of the additional control variables are significant, likely because they are poor
proxies for the costs of cutting R&D for an individual firm. The inclusion of these proxies
does not affect the significance of the PIH variable.
I also added additional variables to control for other differences in ability or incentives
to maintain earnings growth. I reran the tests adding an indicator variable for firms with at
least five years of increasing earnings. Barth et al. (1997) find that such firms have strong

20 I would like to thank I/B/E/S International Inc. for providing this data.
21 I also tested the results with the SD sample based on quarterly earnings performance and quarterly R&D expenses
for a subsample of firms voluntarily reporting quarterly R&D on Compustat. In this specification, the coefficient
on the PIH variable in the SD sample was negative, but not significant. The lack of significant results likely
stems from lower power due to a smaller sample (the sample for these tests have less than 400 observations)
and to the fact that firms voluntarily reporting quarterly R&D have significantly higher levels of institutional
ownership than firms not reporting quarterly R&D, reducing the dispersion in the PIH variable.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 324 The Accounting Review, July 1998

incentives to maintain their earnings gr

multiple. I also included future values o
sales, and change in pre-R&D earnings to
The only variable with a significant coeff
in R&D. This variable had a negative coe
the test year had smaller changes in R&
also did not affect the significance of the PIH variable.
Finally, I tested whether the decision to cut R&D is a function of the change (or
expected change) in institutional holdings. I reran the analysis adding the contemporaneous
and future changes in institutional holdings as independent variables. I examined future
changes up to five years subsequent to the test year. The coefficients on these variables
were insignificant in the SD sample, indicating that managers' decisions to cut or increase
R&D in this sample do not change the opinions of institutions to the extent that they
significantly increase or decrease their holdings.

VI. RESULTS FOR OWNERSHIP

BY DIFFERENT GROUPS OF INSTITUTIONS
Classification of Institutional Investors
This section describes how I use factor analysis and cluster analysis to classify insti-
tutional investors into groups based on their past investment behavior. Using prior research
and guides on institutional ownership, I constructed nine variables that describe the past
investment behavior of institutional investors (Baysinger et al. 1991; Porter 1992; Carson
Group 1995). I use four variables to proxy for the level of portfolio diversification of each
institution (see table 5 for definitions). The level of portfolio concentration (CONC) mea-
sures the average percentage of an institution's total equity holdings invested in each port-
folio firm. The average percentage holding variable (APH) measures the average size of
the institution's ownership position in its portfolio firms. Two variables capture how much
of the institution's total equity is invested in firms for which the institution has an influential
ownership position. LBPH is the percentage of the institution's equity that is invested in
firms where it has greater than 5 percent ownership and HERF is a Herfindahl measure of
concentration calculated using the squared percentage ownership in each portfolio firm.
I include two variables to measure the degree of portfolio turnover of the institution.
Portfolio turnover (PT) measures the average absolute change in the institution's ownership
positions over a quarter, scaled by the change in total equity of the institution. The relative
stability of the institution's holdings in its portfolio firms (STAB) measures the percentage
of the institution's total equity invested in firms that it has continuously held for the prior
two years.
Finally, I include three variables to measure the institution's trading sensitivity to cur-
rent earnings. The first variable, CETS1, measures the interaction between changes in an
institution's holdings in a firm over a quarter and that firm's seasonal change in quarterly
earnings announced during the quarter. The value of CETS1 will be high (low) if the
institution invests more heavily in firms with larger positive (negative) earnings changes
and divests more heavily of firms with negative (positive) earnings. Since this measure
assumes a linear relation between the magnitude of the change in holdings and the earnings
change, I use two additional measures that allow for nonlinear relations. The second vari-
able, CETS2, measures the difference between the average earnings change of firms in
which the institution increased its holdings and the average earnings change of firms in

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 325

TABLE 5
Institutional Investor Characteristics"

Variable Definition

Portfolio concentration (CONC) E wkt/NSTK,

Average percentage holding (APH) (I Wkt PHkt)/I Wkt

Percent held in large blocks (LBPH) (I Wkt LBkt)/I Wkt

Herfindahl measure of concentration (HERF) ln (E PHkt2)

Stability of holdings (percent held for two

years) (STAB) (I Wkt LTkt)/I Wkt

Portfolio turnover (PT) E|iv WktI/(a Wkt + E Wkt-1)

Trading sensitivity to current earnings
(CETS1) (Ed Wkt RWEkt)/1IA WktI

Average earnings change of firms bought vs.

firms sold (CETS2) Avg. (RWEktIA Wkt > 0)-Avg. (RWEktIA Wkt < 0)
Change in holdings in firms with positive
earnings vs. firms with negative earnings [(Ed wktIRWEkt > 0)-(HA wktIRWEkt < 0)]/
(CETS3) 1IA WktI

NSTK1 = number of stocks owned by institutio

Wkt = portfolio weight (shares held times stoc
Awkt = Wkt- Wkt I;
PHkt = percentage of total shares in firm k held by institution at end of quarter t;
LBkt = 1 if PHkt > 0.05, 0 otherwise;
LTkt = 1 if institution held firm k continuously for prior eight quarters, 0 otherwise;
RWEkt = seasonal random walk change in quarterly earnings per share of firm k for quarter t (deflated by sa
for quarter t-4);
a The characteristics are calculated at the end of each calendar quarter for every institution on the Spectrum databa
The quarterly values are averaged over all quarters available for the calendar year to get end-of-year averag
values of each characteristic for each institution. These average annual values are used in the subsequent fact
and cluster analyses.

which the institution reduced its holdings. The third variable, CETS3, measures the differ
ence between the total change in holdings of firms with positive earnings changes and th
total change in holdings of firms with negative earnings changes.
Many of these portfolio characteristics are highly correlated with each other, making
it difficult to draw conclusions based on a single characteristic or to include multiple char
acteristics in the same analysis. To mitigate this problem, I perform principal factor analysis
with an oblique rotation to identify and interpret common factors.22 This procedure reduces

22 The principal factor approach is appropriate when the goal of the analysis is exploratory rather than confirmator
and is adopted in a similar context by Dechow et al. (1996). I chose an oblique rotation over an orthogon
(varimax) rotation because it does not rely on the assumption that the factors are independent. However,
varimax rotation yields virtually identical results.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 326 The Accounting Review, July 1998

the dimensionality of the data by combin

simonious set of linear combinations th
teristics. Once common factors are ident
in the subsequent cluster analysis.
Panel A of table 6 presents the results of the factor analysis, which produced three
common factors.23 The BLOCK factor captures the average size of an institution's stake in
its portfolio firms. Institutions with high (low) BLOCK scores have portfolios characterized
by larger (smaller) average investments in their portfolio firms. The PTURN factor measures
the degree of portfolio turnover. Institutions with high (low) PTURN scores trade more
(less) frequently and are less (more) likely to hold any given firm in their portfolio contin-
uously for two years. The MOMEN factor captures the trading sensitivity to current earn-
ings news. Institutions with high MOMEN scores are momentum traders who tend to in-
crease (decrease) their holdings in stocks with positive (negative) current earnings news;
whereas institutions with low MOMEN scores are contrarian traders who tend to increase
(decrease) their holdings in stocks with negative (positive) current earnings news.
I perform k-means cluster analysis on the factor scores to obtain the final separation
of institutions into groups. Cluster analysis is a technique for grouping observations into
clusters so that institutions are more similar to institutions in the same cluster than they are
to institutions in other clusters.24 There are no standard objective criteria for choosing the
number of clusters. Hair et al. (1995) suggest computing a number of different cluster
solutions, within a reasonable prespecified range, and deciding among the alternatives based
on a priori criteria, theoretical considerations, or practical judgment. My goal was to find
cluster solutions without disproportionate numbers of observations in some clusters (e.g.,
one cluster with 90 percent of the observations and the rest with the remaining 10 percent)
that closely match the theoretical reasons for splitting institutions into groups.
Panel B of table 6 shows the results of the cluster analysis. The three-cluster solution
yields groups matching the definitions of transient, dedicated, and quasi-indexer institutions.
Based on the mean factor scores for each cluster, "transient" institutions have the highest
turnover (PTURN) and highest use of momentum strategies (MOMEN), along with rela-
tively high diversification (small BLOCK). "Dedicated" institutions have high concentra-
tion, low turnover, and almost no trading sensitivity to current earnings (average MOMEN
near zero). "Quasi-indexers" exhibit high diversification and low turnover, consistent with
index-type, buy-and-hold behavior. Quasi-indexers also exhibit contrarian-trading tenden-
cies (low MOMEN), which are consistent with most buy-and-hold value strategies.
The quasi-indexers are the largest class of institutional investors, accounting for 70
percent of the sample. Since only about 10 percent of institutions have stated index strat-
egies (Carson Group 1995), the majority of the quasi-indexer group does not explicitly
index, but follow longer-term buy-and-hold strategies. The predominance of this group is
likely due to (1) the fact that the transient group captures only the most aggressive mo-
mentum traders, which comprise a relatively small portion of institutional strategies (Carson
Group 1995); and (2) the aggregation of all individual funds within a fund group (e.g., all

23 I used three criteria to determine the appropriate number of factors to keep: scree plots, minimum eigenvalue,
and proportion of variance explained. All three criteria produced consistent results.
24 The k-means method of cluster analysis starts with a set of cluster seeds and then assigns observations into
clusters based on the Euclidean distances between the observation and the means of the cluster seeds (which
change as additional observations are added). With large numbers of observations, the formation of clusters
using this method is insensitive to the selection of initial seeds and to the order of the observations (Hair et al.
1995).

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 327

TABLE 6
Factor Analysis and Cluster Analysis of Institutional Investor Portfolio Characteristics

Panel A: Factor Analysisa

Factors

Variable BLOCK PTURN MOMEN

APHb 0.818 0.019 -0.013

LBPH 0.798 -0.023 -0.004

HERF 0.338 0.063 0.073

CONC 0.327 -0.062 -0.009

PT 0.033 0.766 -0.003

STAB 0.036 -0.751 0.015

CETS1 0.022 0.004 0.727

CETS2 0.004 -0.009 0.722

CETS3 0.009 -0.015 0.053

Variance
Explained 43.2% 30.7% 26.1%

Panel B: Clusters Based on Factor Scoresc

Mean Factor Scoresd

Cluster N__ BLOCK PTURN MOMEN

Quasi-Indexers 6060 -0.153 -0.250 -0.272

Transient 2225 -0.075 0.337 0.688

Dedicated 319 3.303 -0.625 -0.078

Panel C: Persistence of Cluster Membership Across Time

Percent in Cluster-Year t+1 Percent in Cluster-Year t+3

Quasi- Quasi-
Year t Cluster Indexersf Transient Dedicated Indexers Transient Dedicated

Quasi- 80.2 19.2 0.6 78.5 20.5 1.0

Indexers

Transient 56.7 42.5 0.8 57.5 41.2 1.3

Dedicated 10.2 4.1 85.8 18.9 4.9 76.2

a Factors are estimated using principal factor an

b All variables are defined in table 5.
c Clusters are formed using k-means cluster ana
d The factor scores are generated from the fact
eN represents the number of institution-years i
'The percentages in each column represent the
year t+ 1 (or t+3) cluster, excluding institution

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 328 The Accounting Review, July 1998

of Fidelity's funds are listed in one holdings fig

tutions (4 percent) is not surprising given that
that limit the size of any stake they can take in
governance incentives that make diversification
now 1995, chap. 2). Overall, the relative numb
Porter's (1992, 8) argument that U.S. institution
ownership positions in firms, but rather are lik
and/or turnover these positions frequently.
Panel C of table 6 documents the persistence
stitutions classified as quasi-indexer and dedicat
the future, as over 80 percent (75 percent) of thes
one (three) years hence. On the other hand, the
another group in subsequent periods. In the nex
40 percent of transient institutions remain clas
in the quasi-indexer category in subsequent year
the transient classification is the MOMEN factor, w
is classified as transient. Thus, many institutions a
trading they do over time, indicating that such
persistent across time as might be expected.

Results for Predominant Ownership by Different Groups of Institutional Investors

Table 7 presents results for the logit regression of CUTRD on the control variables and
institutional holdings including the indicator variables for predominant ownership by tran-
sient, dedicated, and quasi-indexer institutions (see equation (2)). The restriction that firms
have at least 5 percent institutional ownership reduces the sample sizes relative to table 4
(see the last row of table 7). Column 1 presents results for the SD sample. The only change
in the significance of the control variables between table 4 and table 7 is that the coefficient
on SIZE is no longer significant. This lack of significance is caused by constraining the
sample to only include firms with greater than 5 percent institutional ownership (which are
primarily larger firms) and not due to the addition of the indicator variables for holdings
by group. Consistent with H2a, firms with an extremely high proportion of ownership by
transient institutions are significantly more likely to cut R&D.26 The coefficient on the
transient indicator variable is insignificant in both the IN and LD samples, indicating that
extreme transient ownership only influences the decision to cut R&D for firms with small
declines in earnings. This evidence supports the concern of Porter (1992), Jacobs (1991),
Graves and Waddock (1990), and others that transient institutional investor behavior leads
to myopic R&D investment behavior.
The results indicate that extreme proportions of ownership by dedicated institutions
and by quasi-indexers have no incremental impact on the likelihood of R&D cuts in any
of the subsamples. For dedicated institutions, the lack of any significant incremental impact
probably stems from the fact that any effect of dedicated ownership is difficult to detect
due to the limited number of cases in which dedicated institutions own majority shares of

25 Many institutions that maintain numerous funds, such as Fidelity and Dreyfus, are split between transient and
quasi-indexer categories across years in the sample. These switches in category likely occur when one set of
funds within the fund family has an exceptionally good year or receives a larger influx of new money, tilting
the average portfolio characteristics toward these funds.
26 The coefficient on the transient indicator variable is also significantly positive (p = 0.03) when defined using
quartiles, and insignificantly positive (p = 0.16) when defined using treciles.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 329

TABLE 7
Logit Regression of Indicator for Cut in R&D (CUTRD) on Control Variables and
Institutional Holdings By Group of Institution

Prob(CUTRDi = 1) = F[a + r3PCRDi + r2CIRDi + r3CGDPi + r4TOBQj

+ r5CCAPX, + r6CSALESi + r3,SIZEi + r3DIST, + LEV,

+ 3,0FCF, + ,,PIH, + P,2DQ5(Transient)i

+ P,3DQ5(Quasi-indexer)j + P,4DQ5(Dedicated)iJ
SD Samplea IN Sample LD Sample

Expected Coefficient Marginal Coefficient Marginal Coefficient Marginal

Variable Sign (p-value) Effectsc (p-value) Effects (p-value) Effects

Interceptb -0.184 0.595 0.207

(0.424) (0.000) (0.305)

PCRD - -0.484 -0.029 -0.629 -0.027 -0.379 -0.027

(0.008) (0.000) (0.004)

CIRD - -0.982 -0.037 -0.742 -0.019 -0.762 -0.030

(0.000) (0.000) (0.000)

CGDP - -0.791 -0.002 -4.877 -0.011 -3.332 -0.010

(0.820) (0.025) (0.286)

TOBQ - -0.099 -0.016 -0.037 -0.007 -0.065 -0.008

(0.068) (0.202) (0.256)

CCAPX - -0.521 -0.077 -0.425 -0.045 -0.216 -0.039

(0.000) (0.000) (0.002)

CSALES - -1.917 -0.062 -3.012 -0.092 -1.720 -0.084

(0.000) (0.000) (0.000)

SIZE - -0.031 -0.016 -0.212 -0.085 -0.071 -0.037

(0.390) (0.000) (0.045)

DIST +d 0.232 0.023 -0.025 -0.006 0.060 0.037

(0.221) (0.028) (0.000)

LEV + 0.863 0.041 0.681 0.022 1.058 0.060

(0.014) (0.001) (0.000)

FCF - 0.119 0.006 -0.461 -0.016 0.141 0.008

(0.557) (0.000) (0.426)

PIH +/- -0.946 -0.060 -0.096 -0.005 -0.269 -0.016

(0.016) (0.671) (0.655)

DQ5(Transient) + 0.365 0.007 -0.067 -0.001 -0.062 -0.001

(0.015) (0.464) (0.655)

DQ5(Quasi-indexer) +/- 0.166 0.003 -0.144 -0.002 0.030 0.001

(0.240) (0.116) (0.821)

(Continued on next page)

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 330 The Accounting Review, July 1998

TABLE 7 (Continued)

SD Samplea IN Sample LD Sample

Expected Coefficient Marginal Coefficient Marginal Coefficient Marginal

Variable Sign (p-value) Effectsc (p-value) Effects (p-value) Effects

DQ5(Dedicated) - 0.207 0.003 0.107 0.002 0.214 0.003

(0.167) (0.200) (0.137)

pseudo-R2 0.143 0.221 0.121

N 1826 6384 1995

a The SD sample consist

of all firm-years for w
-EBTRD,-l) < -RDt,-. EB
greater than 5 percent tot
bDQ5(GROUPk)i equals o
in firm i is in the top qu
c The marginal effect re
independent variable ove
dThe expected sign for D
managers of firms in t

firms.27 For quasi-i

relation between ex
which quasi-indexers are the largest group.

VII. SUMMARY AND CONCLUSIONS

This paper contributes to the ongoing debate about the short-term focus
agers and investors by providing evidence on the influence of institutional inv
investment behavior of managers facing incentives to cut R&D to maintain earning
The results indicate that managers are significantly less likely to cut R&D to
earnings decline when institutional ownership is high. This evidence supports t
relative to individual investors, the large stockholdings and sophistication of
investors allows them to monitor and discipline managers, ensuring that man
R&D levels to maximize long-run value rather than to meet short-term earn
However, given a significant level of institutional ownership, a high proportion of
by institutions exhibiting transient ownership characteristics (i.e., high portf
diversification, and momentum trading) significantly increases the probability tha
reduce R&D to boost earnings. This result supports the widely-argued view tha
oriented behavior by institutions creates pressures for managers to sacrifice
sake of higher current earnings, but only for a sample of firms with extreme
of transient ownership and not for the U.S. market, or for institutional investors
Some important caveats must be mentioned in these interpretations of the resu
the evidence on the relation between transient institutional ownership and th

27 Within the top quintile of proportional ownership by dedicated institutions, the mean holdi
institutions is less than 20 percent, compared to mean holdings of 94 percent (70 percent) f
(transient) institutions in the top quintile of their respective distributions. As a result, 21[25] fi
of 1 for both DQ5(Dedicated) and DQ5 (Quasi-indexer) [DQ5(Transient)] in the SD sample.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 Bushee-Institutional Investors and Myopic R&D Behavior 331

cut R&D must be interpreted with some caution due to the tim
bership in the transient group. This instability indicates that t
persist over time, but this fact works against a finding that past i
ences current managerial decisions. Second, the theory and the
ings are framed in terms of institutional investor influence on ma
R&D to manage earnings; but in an association test, it is always difficult to establish
causality. I address other explanations for the relation, such as self-selection by institutions
in the firms they choose, by examining various sample partitions, by comparing changes
in R&D to levels of institutional ownership, and by examining concurrent and future
changes in institutional ownership. The body of evidence produced by these approaches is
most consistent with institutional investors influencing managerial behavior rather than vice
versa. Finally, the process by which managers can manipulate R&D to meet earnings goals
is not widely understood, and is difficult to document with publicly available data. This
limitation makes it difficult to include specific controls for differences in the cost of cutting
R&D across firms, and ultimately makes it difficult to discern whether managers perma-
nently cut R&D or temporarily postpone it across fiscal periods. I try to overcome this
limitation by including control variables, such as indicators for the average composition of
R&D in the firm's industry and future values of R&D and earnings, but more detailed,
small sample research is needed to fully understand the mechanics of managerial myopia.
Given these caveats, this paper provides the first large-sample cross-sectional evidence
of institutional investor behavior contributing to the myopic investment problem in the U.S.
I find this evidence by focusing on a specific situation-reducing R&D to reverse an
earnings decline-and by partitioning institutions directly on the investment behavior ar-
gued to create incentives for myopic investment behavior. In doing so, this paper also
develops a new method for classifying institutional investors into groups based on past
investment behavior. This classification is applicable to other contexts, such as explaining
market reactions to earnings announcements, that will be explored in future research. Fi-
nally, this study motivates further research into the current trend toward more active investor
targeting and investor relations by firms as the evidence indicates that different compositions
of institutional owners affect managerial decisions.

REFERENCES
Agrawal, A. and C. Knoeber. 1996. Firm performance and mechanisms to control
between managers and shareholders. Journal of Financial and Quantitative Ana
ber): 377-397.
Anand, V. 1991. Big and small companies are marketing their stock. Investor's Daily (January
8.

Baber, W., P. Fairfield, and J. Haggard. 1991. The effect of concern about reported income on dis-
cretionary spending decisions: The case of research and development. The Accounting Review 66
(October): 818-829.
Badrinath, S., G. Gay, and J. Kale. 1989. Patterns of institutional investment, prudence, and
managerial "safety-net" hypothesis. Journal of Risk and Insurance 56 (December): 605-62
Bange, M., and W. De Bondt. 1997. R&D budgets and corporate earnings targets. Working pap
University of Oregon.
Barth, M., J. Elliott, and M. Finn. 1997. Market rewards associated with increasing earnings patt
Working paper, Stanford University.
Baysinger, B., R. Kosnik, and T. Turk. 1991. Effects of board and ownership structure on corp
R&D strategy. Academy of Management Journal 34 (March): 205-214.
Berger, P. 1993. Explicit and implicit tax effects of the R&D tax credit. Journal of Accounting
search 31 (Autumn): 131-171.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
 332 The Accounting Review, July 1998

Bernard, V., J. Thomas, and J. Abarbanell. 1993. How sophisticated is the market in interpreting
earnings news? Journal of Applied Corporate Finance: 54-63.
Brown, K., and B. Brooke. 1993. Institutional demand and security price pressure: The case of cor-
porate spin-offs. Financial Analysts Journal 49 (September/October): 53-62.
Burgstahler, D., and I. Dichev. 1997. Earnings management to avoid earnings decreases and losses.
Journal of Accounting & Economics 24 (December): 99-126.
, and M. Eames. 1998. Management of earnings and analyst forecasts. Working paper, Uni-
versity of Washington.
Carson Group. 1995. The 1995 International Who's Who of Institutions and Mutual Funds. New York,
NY: Carson Publications.
Chan, S., J. Martin, and J. Kensinger. 1990. Corporate research and development expenditures and
share value. Journal of Financial Economics 26 (August): 255-276.
Dechow, P., and R. Sloan. 1991. Executive incentives and the horizon problem: An empirical inves-
tigation. Journal of Accounting and Economics 14 (March): 51-89.
9 , and A. Sweeney. 1996. Causes and consequences of earnings manipulations: An
analysis of firms subject to enforcement actions by the SEC. Contemporary Accounting Research
13 (Spring): 1-36.
Dobrzynski, J. 1993. Relationship investing. Business Week 3309 (March 15): 68-75.
Drucker, P. 1986. A crisis of capitalism. Wall Street Journal (September 30): 32.
Duke, J., and H. Hunt. 1990. An empirical examination of debt covenant restrictions and accounting-
related debt proxies. Journal of Accounting and Economics 12 (January): 45-63.
Eames, M. 1997. Institutional investor myopia, ownership, earnings, and returns. Working paper, Santa
Clara University.
Elgin, P. 1992. Strategic pairings uncork blessings for investors, issuers. Corporate Cashflow 13
(September): 7-8.
Froot, K., A. Perold, and J. Stein. 1992. Shareholder trading practices and corporate investment
horizons. Journal of Applied Corporate Finance: 42-58.
, D. Scharfstein, and J. Stein. 1992. Herd on the street: Informational inefficiencies in a market
with short-term speculation. Journal of Finance 47 (September): 1461-1484.
Graves, S. 1988. Institutional ownership and corporate R&D in the computer industry. Academy of
Management Journal 31 (June): 417-428.
, and S. Waddock. 1990. Institutional ownership and control: Implications for long-term cor-
porate strategy. Academy of Management Executive 4 (February): 75-83.
Hair, J., Jr., R. Anderson, R. Tatham, and W. Black. 1995. Multivariate Data Analysis. 4th edition.
Englewood Cliffs, NJ: Prentice Hall.
Hand, J. 1990. A test of the extended functional fixation hypothesis. The Accounting Review 65
(October): 740-763.
Hansen, G., and C. Hill. 1991. Are institutional investors myopic? A time-series study of four
nology-driven industries. Strategic Management Journal 12 (January): 1-16.
Holthausen, R., D. Larcker, and R. Sloan. 1995. Annual bonus schemes and the manipulation o
earnings. Journal of Accounting and Economics 19 (February): 29-74.
Hunt, A., S. Moyer, and T. Shevlin. 1996. Managing interacting accounting measures to meet multiple
objectives: A study of LIFO firms. Journal of Accounting and Economics 21 (June): 339-374.
Jacobs, M. 1991. Short-Term America: The Causes and Cures of Our Business Myopia. Boston, MA:
Harvard Business School Press.
Jalilvand, A., and R. Harris. 1984. Corporate behavior in adjusting to capital structure and dividen
targets: An econometric study. Journal of Finance 39 (March): 127-145.
Kennedy, P. 1992. A Guide to Econometrics. 3d edition. Cambridge, MA: The MIT Press.
Kim, J., I. Krinsky, and J. Lee. 1997. Institutional holdings and trading volume reactions to quarter
earnings announcements. Journal of Accounting, Auditing, and Finance 12 (Winter): 1-14.
Lang, M., and M. McNichols. 1997. Institutional trading and corporate performance. Working paper
Stanford University.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms
Bushee-Institutional Investors and Myopic R&D Behavior 333

Laverty, K. 1996. Economic "short-termism": The debate, th

for management practice and research. Academy of Man
Mansfield, E. 1981. Composition of R and D expenditures: Re
and innovative output. Review of Economics and Statist
Monks, R., and N. Minow. 1995. Corporate Governance. Cam
Myers, S. 1984. The capital structure puzzle. Journal of Fi
O'Brien, P., and R. Bhushan. 1990. Analyst following and i
counting Research 28 (Supplement): 55-76.
Opler, T. and J. Sokobin. 1997. Does coordinated institutional shareholder activism work? An analysis
of the activities of the Council of Institutional Investors. Working paper, Ohio State University.
Perry, S., and R. Grinaker. 1994. Earnings expectations and discretionary research and development
spending. Accounting Horizons 8 (December): 43-51.
Porter, M. 1992. Capital Choices: Changing the Way America Invests in Industry. Boston, MA:
Council on Competitiveness/Harvard Business School.
Potter, G. 1992. Accounting earnings announcements, institutional investor concentration, and com-
mon stock returns. Journal of Accounting Research 30 (Spring): 146-155.
Rajgopal, S., and M. Venkatachalam. 1997. The role of institutional investors in corporate governance:
An empirical investigation. Working paper, University of Iowa.
Reed, M., and R. D. Plumlee. 1998. The interaction of earnings management and tax incentives: An
empirical analysis of R&D expenditures. Working paper, University of Cincinnati.
Schipper, K. 1989. Commentary on earnings management. Accounting Horizons 3 (December): 91-
102.
Schneiderman, H. 1991. Managing R&D: A perspective from the top. Sloan Management Review 32
(Summer): 53-58.
Stein, J. 1988. Takeover threats and managerial myopia. Journal of Political Economy 96 (Febru
61-80.
. 1989. Efficient capital markets, inefficient firms: A model of myopic corporate behavi
Quarterly Journal of Economics 104 (November): 655-669.
Wahal, S., and J. McConnell. 1997. Do institutional investors exacerbate managerial myopia? Wor
paper, Emory University.
Warfield, T., J. Wild, and K. Wild. 1995. Managerial ownership, accounting choices, and infor
tiveness of earnings. Journal of Accounting and Economics 20 (July): 61-91.
Wiedman, C. 1996. The relevance of characteristics of the information environment in the sele
of a proxy for the market's expectations for earnings: An extension of Brown, Richardson,
Schwager (1987). Journal of Accounting Research 34 (Autumn): 313-324.
Wines, R. 1990. 13F filings: How reliable are they? Directors & Boards 15 (Fall): 52-53.

This content downloaded from

14.139.157.21 on Tue, 31 Jan 2023 05:57:20 UTC
All use subject to https://about.jstor.org/terms