 

 

 

 

 

 

..

 
 

 

 

 

 

 

I--<‘;Q

,..u

nu...»

 

,u‘ ._-.. .

l 4:1'

 

 

 
 
 
 
 
 
 
 

 

”2““ .LleRY '
20.: 3 Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

INVESTMENT, ACQUISITIONS, AND FINANCIAL
CONSTRAINTS

presented by

JOSHUA ROBERT PIERCE

has been accepted towards fulﬁllment
of the requirements for the

Doctoral degree in Finance

 

 

W All/ﬂex

 

Major Professor’s Signature
5/ l/ / Zoo 7

Date

 

Doctoral Dissertation

MSU is an afﬁrmative-action, equal-opportunity employer

 

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

 

DATE DUE DATE DUE DATE DUE

M M M

'WW ‘7 -‘ luVV‘Y

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 p:/CIRCIDateDue indd-p.1

INVESTMENT, ACQUISITIONS, AND FINANCIAL CONSTRAINTS
By

Joshua Robert Pierce

A DISSERTATION
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Department of Finance

2007

 

ABSTRACT
INVESTMENT, ACQUISITIONS, AND FINANCIAL CONSTRAINTS
By

Joshua Robert Pierce

In this dissertation we examine the determinants of investment at the ﬁrm level. We
investigate the impact of cash ﬂow on investment, ownership on investment, and ﬁnally
the classiﬁcation and correlation of qualitative measures of ﬁnancial constraint status
with quantitative accounting variables. In the ﬁrst essay we use a unique, hand-collected
dataset, and ﬁnd a signiﬁcant positive relation between a ﬁrm’s U.S. advertising
spending and its contemporaneous foreign cash flow. This relation holds even after
controlling for factors that should be related to the optimal level of US. advertising and
is stronger for subsets of ﬁrms that we expect to be relatively more ﬁnancially
constrained including younger ﬁrms, ﬁrms that pay low dividends, highly levered ﬁrms,
ﬁrms with low credit ratings, and ﬁrms that do not hedge. Our evidence supports the
important hypothesis that there is a causal and economically substantial link between a
ﬁrm’s cash ﬂow and its investment decisions, even for intangible investments such as
advertising. In the second essay we study a sample of 555 brands that experience an
ownership change and ﬁnd that new owners Often sharply increase or decrease
advertising spending on the acquired brand, with large cuts being particularly common.
Increased private ownership of a brand is associated with a signiﬁcant downward shift in
advertising relative to other deals. Buyers tend to cut advertising in existing brands that
closely overlap with purchased brands, but do not cut spending on non-overlapping

brands. Combined buyer-seller announcement returns are positively related to some

 

measures of post-acquisition downward shifts in advertising spending. Acquired brands
do not on average experience signiﬁcant losses in market share, even when advertising
spending is revised downwards. Our evidence is consistent with the hypothesis that the
identity and characteristics of an asset's owner are important determinants of investment
policy for the asset. In the ﬁnal essay we examine the annual reports of 400 randomly
chosen companies between 1995-2004 to investigate both the existence and implications
of ﬁnancial constraints based on qualitative and quantitative measures. We extend the
work of Kaplan and Zingales (1997) and use over 1900 ﬁrm-year Observations to
construct indicators of ﬁnancial constraints based on the management discussion and
analysis section of annual reports and quantitative measures constructed from accounting
data on COMPUSTAT. We use these measures to investigate how stable ﬁnancial
constraint classiﬁcations are over time and how robust they are to sample size issues.
Further, we are able to ascertain whether the correlations between qualitative and
quantitative measures of ﬁnancial constraints have changed over time and what is driving
this change. Finally, we re-investigate the sensitivity of investment to cash ﬂow using an
updated and more heterogeneous sample. We ﬁnd that (1) the distribution of ﬁnancial
constraint status has changed over time (2) the correlation between quantitative factors
and qualitative ﬁnancial constraint classiﬁcations has changed and (3) the relation
between investment and cash ﬂow has decreased over time and is a questionable measure

of ﬁnancial constraint status when cash ﬂow and investment are positively correlated.

 

Copyright by
JOSHUA ROBERT PIERCE
2007

 

To Renee and Willow

 

ACKNOWLEDGEMENTS

My deepest appreciation goes to my wife Renee. She was undoubtedly the largest
factor in the completion of my degree. I could not have done this without her
overwhelming love and support. I am also indebted to my parents Bob and Sharon Pierce
who instilled in me a work ethic that allowed me to complete my degree. My brother
Brian provided much needed humor and support throughout the process. Of course,
without my inquisitiveness in the stock market from my grandfather Robert E. Pierce, I
would have never gone into ﬁnance. I thank him from the bottom of my heart.

I thank the members of my dissertation committee Ted Fee (co-chair), Charlie
Hadlock (co-chair), J un-Koo Kang, and Mike Mazzeo. I could not have had better
chairmen. Words cannot express my gratitude towards Ted and Charlie. I especially
thank Jun-Koo for admitting me into the doctoral program and for his never ending
encouragement.

Fellow graduate students Hoontaek Seo, Hyun-Seung Na, Kung-Shing Chang,
and Jin-Mo Kim provided unwavering support and assistance and I would not have been

able to complete this dissertation without them.

vi

 

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................... x

ESSAY 1. INVESTMENT, FINANCING CONSTRAINTS, AND INTERNAL
CAPITAL MARKETS: EVIDENCE FROM THE ADVERTISING
EXPENDITURES OF MULTINATIONAL FIRMS

1 .1 Introduction .................................................................................................. 1
1.2 Related Literature and Empirical Strategy .................................................. 4
1.2.1 Financial Constraints and Investment ...................... 4
1.2.2 Internal Capital Markets .......................................... 5
1.2.3 Advertising as an Investment ................................... 8
1.2.4 Empirical Strategy ................................................... 9
1.3 Data and Sample Selection ........................................................................ 11
1.3.1 Sample Selection ................................................... 11
1.3.2 Summary Statistics on Advertising ....................... 13
1.3.3 Foreign and Domestic Cash Flows ........................ 15
1.4 Models of Advertising Investment and Cash Flow ................................... 16
1.4.1 Firm Advertising and Firm Cash Flow .................. 16
1.4.2 US. Advertising and Foreign Cash Flow .............. 19
1 .4.3 Robustness ............................................................. 22
1.5 Financial Constraint Measures and the
Advertising-Cash Flow Relation ............................................................... 35
1.5.1 Measures of Financial Constraints ......................... 36
1.5.2 Empirical Analysis ................................................. 37
1.5.3 Hedging .................................................................. 39
1.6 Conclusion ................................................................................................. 40
APPENDIX 1.TABLES OF ESSAY 1 ................................................................. 43

ESSAY 2. WHAT HAPPENS IN ACQUSITIONS? EVIDENCE FROM
BRAND OWNERSHIP CHANGES AND ADVERTISING

2. 1 Introduction ................................................................................................ 49

vii

 

2.2 Motivation and Empirical Strategy ............................................................ 53
2.2.1 Value Creation and the Market
for Corporate Assets .............................................. 53
2.2.2 Sources of Value Changes in Acquisitions ............ 55
2.2.3 Investment Policy and Changes in Control ........... 57
2.2.4 Owner Type and Acquisition Activity ................... 59
2.2.5 Advertising as an Investment ................................. 61
2.2.6 Empirical Strategy ................................................. 62
2.3 Data and Sample Selection ........................................................................ 63
2.3.1 Identifying the Sample ........................................... 63
2.3.2 Characteristics of Buyers, Sellers, and
Transactions ........................................................... 67
2.4 Advertising and Market Share Changes Associated with
Brand Transfers ......................................................................................... 68
2.4.1 Variability in Advertising Policy Around
Ownership Changes ............................................... 69
2.4.2 Owner Type and Post-Acquisition Advertising ..... 72
2.4.3 Firm Characteristics and Advertising Changes ..... 77
2.4.4 Buyer Advertising and Brand Overlap .................. 80
2.4.5 Acquisition Returns and Advertising Behavior ..... 83
2.4.6 Acquisitions, Advertising, and Market Share ........ 87
2.5 Conclusion ................................................................................................. 90
APPENDIX 2.DATA CONSTRUCTION F OR ESSAY 2 ................................... 94
APPENDIX 3.TABLES OF ESSAY 2 ................................................................. 97
ESSAY 3. QUALITATIVE MEASURES OF FINANCIAL CONSTRAINTS:
A CLOSER EXAMINATION
3A Inuoducﬁon .............................................................................................. 109
3.2 Related Literature .................................................................................... 111
3.2.1 Financial Constraint Status and Investment ........ 111
3.3 Empirical Strategy, Data, and Sample Selection ..................................... 112
3.3.1 Empirical Strategy ............................................... 112
3.3.2 Data and Sample Selection .................................. 113
3.3.3 Creation of Financial Constraint Variables ......... 114
3.3.4 Summary Statistics of Financial Constraints ....... 115

viii

3.4 Models of Constraints and Investment—Cash Flow

Sensitivities .............................................................................................. l 16

3.4.1 Correlation of Qualitative Measures of Financial
Constraints and Quantitative Variables ............... 116

3.4.2 Investment-Cash Flow Sensitivities by

Financial Constraint Status .................................. 117
3.5 Conclusion ............................................................................................... I 19
APPENDIX 3.TABLES OF ESSAY 3 ............................................................... 121
REFERENCES .................................................................................................... 127

ix

 

Table 1.1

Table 1.2

Table 1.3

Table 1.4

Table 1.5

Table 2.1

Table 2.2

Table 2.3

Table 2.4

Table 2.5

Table 2.6

Table 2.7

Table 2.8

Table 2.9

Table 3.1

Table 3.2

Table 3.3

Table 3.4

LIST OF TABLES

Summary Statistics for Compustat Firms ...................................... 43
Summary Statistics for Top 1000 US. Advertisers ....................... 44

Regression Models Of Investment, Total Advertising,
and Firm Cash Flow ....................................................................... 45

Regression Models of US. Advertising and Foreign Cash Flow..46

Regression Models of US. Advertising for Sub-samples

Group by Firm Characteristics ....................................................... 47
Sample Characteristics ................................................................... 97
Characteristics of Buyers and Sellers ............................................ 98
Details of Ownership Change ........................................................ 99
Variability in Advertising and Brand Ownership Changes ......... 100
Advertising Changes and Type of Owner ................................... 101

Advertising Changes and Relative Characteristics of
Buyer and Seller .......................................................................... 103

Advertising Changes in Buyer’s Brands and
Purchased Brands ......................................................................... 105

Acquisition Announcement Returns and
Advertising Behavior ................................................................... 106

Market Share Changes following Brand Ownership Changes 108
Sample Summary Statistics ......................................................... 121

Summary Statistics for Annual Financial
Constraint Classiﬁcation .............................................................. 122

Annual Financial Constraint Status ............................................. 123

Ordered Logit for Predictability of Financial
Constraint Status .......................................................................... 124

 

Table 3.5 Investment-Cash Flow Sensitivity by Financial
Constraint Status .......................................................................... 125

xi

 

ESSAY l. INVESTMENT, FINANCING CONSTRAINTS, AND INTERNAL
CAPITAL MARKETS: EVIDENCE FROM THE ADVERTISING
EXPENDITURES OF MULTINATIONAL FIRMS

1.1 Introduction

A large literature in economics and ﬁnance explores the role of ﬁnancial
constraints in the investment activities of ﬁrms (see Hubbard (1998) and Stein (2003)).
This is a fundamental issue, since many of the capital market frictions studied by
ﬁnancial economists have the potential to inﬂuence ﬁrms’ investment decisions, thus
possibly playing a substantive role in resource allocation. Despite the importance of this
issue, received evidence is far from conclusive.l

A common approach to studying the impact of ﬁnancial constraints on
investment, Often associated with Fazzari, Hubbard, and Petersen (1988), is to examine
the relation between a ﬁrm’s investment and its cash ﬂow. However, several studies,
most notably Kaplan and Zingales (1997), have criticized this approach and argued that
investment-cash ﬂow sensitivities are often uninformative with regard to the presence of
ﬁnancial constraints.2 These papers argue that a ﬁrrn’s cash ﬂow may be correlated with
a ﬁrm’s investment opportunities in a manner that could generate the empirical patterns
that researchers have identiﬁed in the data.

A small strand Of research has attempted to overcome this methodological

problem by identifying variations in cash flow that are independent of investment

 

' Chapter 1 and Chapter 2 of this dissertation are based on joint work with Charles J. Hadlock and C.
Edward Fee.

2 See, for example, C1eary(1999), Kaplan and Zingales (2000), Erickson and Whited (2000), Alti (2003),
and Moyen (2004). For a defense of the investment-cash ﬂow approach, see Fazzari, Hubbard, and
Petersen (2000).

opportunities.3 This approach has been exploited in small samples by Blanchard, Lopez-
de-Silanes, and Shleifer (1994) and Lamont (1997). More recently, Rauh (2006) uses this
approach in a large sample context. He identiﬁes exogenous variations in cash flow
arising from pension funding requirements and ﬁnds that these variations have a
surprisingly large impact on corporate investment.

In the spirit of this strand of research, we present large sample evidence on the
role of ﬁnancial constraints on a special type of investment, namely advertising. We use
a unique, hand-collected dataset on US. advertising expenditures by multinational ﬁrms,
and study the role Of foreign cash ﬂows in explaining U.S. advertising. After properly
controlling for US. investment opportunities, if ﬁrms are unconstrained, we would
expect to observe no relation between foreign cash ﬂows and the amount a ﬁrm
advertises in the US. However, if ﬁrms are constrained, we would expect variations in
foreign cash ﬂow to be transmitted to US. advertising decisions. Our key identifying
assumption is that advertising is a local investment in nature, and thus cash ﬂows from
one locality should not be positively related to advertising elsewhere for an unconstrained
ﬁrm. Consistent with this assumption, we ﬁnd that foreign cash ﬂows are not positively
related to a ﬁrm’s subsequent cash ﬂow growth in the US.

In our sample of over 2,600 ﬁrm-years, we ﬁnd that foreign cash ﬂows do have a
signiﬁcant effect on US. advertising. This result is robust to a variety of speciﬁcation
choices. Our estimates imply that an increase of one dollar in a ﬁrm’s foreign cash ﬂow

results in an increase of around $033 in spending by the ﬁrm in the US. advertising

 

3 Researchers have also attempted to test for the presence of ﬁnancial constraints on investment by looking
at other implications arising from the presence of constraints including Euler equation implications [Whited
(1992)] and implications concerning ﬁrms’ cash management policies [Almeida, Campello, and Weisbach
(2004)].

 

outlets that we track. Since advertising is generally small relative to ﬁxed investment, we
show via some calibration exercises that this estimate is consistent with a relatively large
effect of exogenous changes in cash ﬂow on total ﬁrm investment, on the order of $.30 in
increased total spending for each dollar of internal cash ﬂow.

After presenting our sample-wide estimates, we proceed by grouping observations
into sub-samples based on ﬁnancial characteristics that may be associated with ﬁnancial
constraints. Using several different a priori measures of constraints, we ﬁnd that ﬁrms
that are assigned to the more constrained groups generally do display a higher sensitivity
of US. advertising investment to foreign cash ﬂow. In addition, consistent with F root,
Scharfstein, and Stein (1993), we ﬁnd that the effect of foreign cash ﬂow on advertising
investment is smaller for ﬁrms that are engaged in currency hedging activities. We
interpret these ﬁndings as providing evidence concerning which ﬁrm characteristics are
associated with a ﬁrm’s ﬁnancially constrained status. However, these results can also be
viewed as providing independent conﬁrmation of the validity of our empirical approach.

Taken as a whole, the evidence we present echoes the important ﬁndings of Rauh
(2006) that cash ﬂow does have an effect on investment in large samples, an effect that is
magniﬁed in sub-samples where ﬁnancial constraints are expected to be magniﬁed. We
extend Rauh’s (2006) ﬁndings by providing evidence that the impact of ﬁnancing
constraints is not restricted to ﬁxed capital investment, but also extends to intangible
investments such as advertising.4 Our results also add to the literature supporting the
notion that ﬁrms have active internal capital markets in which they move funds across

activities and regions. Moreover, our results are consistent with the hypothesis that some

 

‘ See Almeida and Campello (2005) for an analysis of the role of asset tangibility in the relation between
investment and cash ﬂow.

ﬁrms use hedging to minimize the internal capital market transmission of geographic
shocks to investment.

The rest of the paper is organized as follows. In section 1.2 we ﬁrst discuss the
relevant literatures on ﬁnancing constraints, internal capital markets, and advertising, and
then outline our empirical strategy. In section 1.3 we discuss our data and sample
selection procedures. In section 1.4 we present our main analysis of advertising and cash
ﬂows. In section 1.5 we explore how the advertising-cash ﬂow relation varies across
sub—samples grouped by variables related to ﬁnancial constraints and hedging activity.

Section 1.6 concludes.

1.2 Related literature and empirical strategy

1.2.] Financial constraints and investment

A large part of corporate ﬁnance research is devoted to understanding the various
agency and information problems that result in distortions in ﬁrms' investment and
ﬁnancing decisions. As Stein (2003) discusses in his survey of this research, many of the
theorized frictions that arise in the governance and ﬁnancing of ﬁrms can result in an
outcome where internal and external funds are not viewed as perfect substitutes.5
Consequently, the supply of internal funds may have a causal impact on ﬁrms’

investment choices.

 

5 The implications of imperfect substitutability depend on the underlying frictions generating it. For
example, if managers tend to invest internal funds in poor projects because of incentive problems or
behavioral biases, ﬁnancial slack should be discouraged. However, if ﬁrms frequently pass up good
projects because of problems raising external capital, ﬁnancial slack can be valuable. For some relevant
tests, see Hadlock (1998).

While many authors have studied the relation between investment and cash ﬂow,
if other control variables (e.g., Tobin’s Q) do not adequately control for investment
opportunities, any reported relationship between investment and cash ﬂow may arise
because cash ﬂow is proxying for the quality of investment opportunities rather than
eased ﬁnancial constraints. Moreover, as Alti (2003) has shown, there are reasons to
expect the correlation between cash ﬂow and investment opportunities to vary with ﬁrm
characteristics. Thus, reported variations in investment-cash ﬂow sensitivities across
different types of ﬁrms may be consistent with ﬁrms acting in an unconstrained manner
(see also Moyen (2004)).

The most direct solution to this problem is to identify variations in cash ﬂow that
are not correlated with investment Opportunities. This is a challenge, Since cash ﬂows
from operations are typically driven by the success of a ﬁrm’s business model, and
therefore are often closely related to the proﬁtability of future investment. One approach,
exploited by Blanchard, Lopez-de-Silanes, and Shleifer (1994) and Rauh (2006), is to
identify unique events or institutional features that sever the link between cash ﬂow and
investment opportunities. An alternative approach, suggested by the work of Lamont
(1997), is to exploit the presence internal capital markets. This latter approach is the one

we adopt.

1.2. 2 Internal capital markets
The basic insight of Lamont (1997) is that a cash ﬂow shock that affects one part
of a ﬁrm may be transmitted to other parts of the ﬁrm, even if the investment

opportunities of these other parts of the ﬁrm are not affected by the shock. He exploits

 

this approach in a small sample of industrially diversiﬁed ﬁrms. Peck and Rosengren
(1997) use this approach for a somewhat larger sample of geographically diversiﬁed
ﬁrms of a specialized type (Japanese banks).6

In a large sample context, Shin and Stulz (1998) use Lamont’s (1997) empirical
framework and ﬁnd that a division’s investment spending is related to the cash ﬂows of a
ﬁrm’s other divisions. However, this type of analysis can be problematic in a large
sample, because there are frequently economic linkages within a ﬁrm that can induce a
correlation between cash ﬂows in one part of the ﬁrm and investment opportunities
elsewhere. For example, one division may sell a signiﬁcant fraction of its output to
another. In fact, Chevalier (2004) presents evidence indicating that the Shin and Stulz
(1998) ﬁndings may in fact be driven by a correlation between cash ﬂows in one division
and investment opportunities elsewhere in the ﬁrm.

On way to circumvent this problem is to identify a type of investment spending
for which variations in operating cash flows in other parts of the ﬁrm should be
exogenous. Local advertising and foreign cash ﬂows seems to ﬁt this description well.
Absent ﬁnancial constraints, there is little reason to believe that selling more goods in
Asia should affect how much a ﬁrm advertises in a Detroit newspaper, particularly once
we control for how the ﬁrm is performing in the US. Any beneﬁts of advertising in the
US. are likely to remain in the US. Thus, for an unconstrained ﬁrm, the US. advertising
decision should be independent of performance overseas. However, if a ﬁrm is

ﬁnancially constrained, cash flows that are generated overseas have the potential to affect

 

6 See Houston, James, and Marcus (1997) and Campello (2002) for evidence on the role of internal capital
markets in US. bank's investment policies.

 

investment decisions throughout the ﬁrm via the internal capital market, including
advertising in the US. This observation forms the motivation for our empirical tests.

For our tests to have power, it must be the case that ﬁrms actively move capital
across geographic regions. Prior work by Desai, Foley, and Hines (2004) and Desai,
Foley, and Forbes (2005) suggests that this type of capital movement often does occur.
However, ﬁ‘ictions induced by the tax system or regulatory restrictions may at times
hinder the active movement of capital. For example, a ﬁrm with a high level of foreign
cash ﬂow may choose not to use these funds for domestic investment activities because
of the tax costs associated with repatriating foreign earnings. Such a ﬁrm may still
display a positive sensitivity of domestic advertising investment to foreign cash ﬂow, as
the ﬁrm may ﬁnd that it needs to inject less cash into its foreign operations when they are
performing relatively well. With these considerations in mind, one can view our
investigation as a joint test Of (1) the presence of ﬁnancial constraints at the ﬁrm level
and (2) non-trivial levels of intra-ﬁrm capital mobility across geographic regions.

Firms do have some control over the variability in their foreign cash ﬂows
through their hedging decisions. In particular, cash ﬂow variability arising from currency
changes and foreign macroeconomic events can be partially hedged via the use of the
appropriate derivatives positions. If ﬁrms are actively hedging, the analysis of F root,
Scharfstein, and Stein (1993) would suggest that ﬁrms are more likely to be smoothing
cash ﬂows in a manner that minimizes the likelihood of a mismatch between investment

needs and available internal funds. Consequently, even when cash ﬂows are measured

net of hedging activity, the effect of foreign cash ﬂows on US. advertising may be

diminished] We will examine this possibility below.

1.2.3 Advertising as an investment

The topic of advertising has a long history in the ﬁelds Of economics and
marketing. Bagwell (2005) provides a thorough survey of this research. Since
advertising is not a very tangible input into the proﬁt generating process, it is not
surprising that there has been signiﬁcant debate on the beneﬁts of advertising.8 The most
sensible starting point, from our viewpoint, is to take the perspective of Schmalensee
(1976). He asserts that proﬁt maximizing ﬁrms should be expected to equate the
perceived marginal beneﬁts of advertising with the marginal costs. Since many ﬁrms
advertise heavily, the perceived beneﬁts to the ﬁrm must be non-trivial.

Under this perspective, advertising is like any other investment good. F inns
spend on advertising today in expectation of a ﬂow of beneﬁts in the future. The beneﬁts
of advertising may include ﬁxture increased sales levels or higher prices. Since the goal
of advertising is typically to shift the demand curve for a product, we would expect local
advertising decisions to be inﬂuenced by local demand conditions.

One important dimension where advertising may differ from physical investment

is in its durability. Researchers in advertising have long debated the timing of the ﬂow of

 

7 This argument assumes that the hedging decision is exogenous. If more constrained ﬁrms are more likely
to hedge, our tests would be biased against ﬁnding evidence in support of the hypothesized relation.
Allayannis and Mozumdar (2000) and Deshmukh and Vogt (2005) study hedging and ﬁrm-wide
investment-cash ﬂow sensitivities.

8 In a well-known article, Bass (1969) writes, “There is no more difficult, complex, or controversial
problem in marketing than measuring the inﬂuence of advertising on sales.” Bemdt (1991) discusses much
of the subsequent econometric evidence and interpretation problems arising from a lack of exogenous
variation in advertising.

beneﬁts that arise from advertising, with a wide range of reported estimates.9 For our
purposes, a precise estimate of the durability of advertising investment is not crucial.
However, in our empirical design, we do need to account for the possible feedback from
advertising into contemporaneous sales arising from the potential for a relatively quick
payback for this type of investment.

For our study to be informative, only some minimal assumptions regarding
advertising need to be true. The precise timing and size of the beneﬁts of advertising do
not need to be determined. What is necessary is for managers to make advertising
decisions in a similar manner to other investments. In particular, we need to assume that
managers conduct some sort of cost-beneﬁt marginal analysis in their advertising
expenditure decisions. Just as with physical investment, managerial agency problems or
capital market imperfections may tilt ﬁrm’s choices away from the ﬁrst-best. If
managers view internal and external capital as imperfect substitutes in their spending
decisions, exogenous variations in cash ﬂow have the potential to affect all investment
decisions, including advertising. If we impose the added assumption that ﬁrms scale all
investments up or down approximately proportionally when their sources of internal
funding vary, we can use our estimates on advertising-cash ﬂow sensitivities to gauge the
approximate impact of cash ﬂow variations on a ﬁrm’s other, more tangible, investment

activities.

1.2.4 Empirical strategy

 

9 Researchers have examined distributed lag models in which sales is a function of advertising (e. g., Peles
(1971), Landes and Rosenﬁeld (1994)), and models where Tobin’s Q is a function of past advertising (e.g.,
Ross (1983), Hirschey and Weygandt (1985)).

 

Our basic empirical strategy is to estimate models explaining U.S. advertising as a
function of foreign cash ﬂows and a set of control variables. The models we estimate are
most closely related to the models of Shin and Stulz (1998). However, our analysis is
quite distinct from theirs in that we organize cash ﬂows by geographic area rather than by
industrial activity, and we construct dependent variables based on advertising
expenditures rather than ﬁxed investment. In our modeling we experiment with several
different controls for US. investment opportunities, and we account for ﬁrm-speciﬁc
effects in all of our estimated models.

To better understand advertising spending, we begin by estimating empirical
models explaining total ﬁrm advertising as a function of variables that are often used to
explain ﬁxed investment (e.g., Tobin’s Q, total ﬁrm cash ﬂow). This provides an initial
picture of how advertising spending compares to ﬁxed capital investment. We then
proceed to estimate models of US. advertising as a function Of foreign cash ﬂows. After
estimating these models, we examine the robustness of our ﬁndings to a variety of
speciﬁcation choices and alternative explanations.

After establishing a link between US. advertising and foreign cash ﬂows, we
consider variations in the strength of this relation across ﬁrms sorted by ﬁnancial
characteristics and hedging activity. If one accepts our methodology as a meaningful
indicator of ﬁnancial constraints, this analysis can be viewed as an examination of what
types of ﬁrm characteristics are related to ﬁnancial constraints. This is our preferred
interpretation of this analysis. However, if one believes on a priori grounds that these

ﬁnancial characteristics (or at least a subset of them) are valid indicators of ﬁnancial

lO

constraints, this analysis can be viewed as an independent check on the validity of our

methodology.

1.3 Data and sample selection

1.3.1 Sample selection

Our chief source of advertising data is the publication Ad $ Summary published
by Competitive Media Reporting/Taylor Nelson Sofres Company (CMR/TNS) and its
predecessors. This data is relied on heavily by ﬁrms in their marketing intelligence.
CMR/TNS collects advertising information through electronic monitoring and revenue
reports from a large number of US. media outlets (major radio and television stations,
national newspapers, etc.). They report information on advertising spending aggregated
at the company and brand level. Several prior academic research studies use CMR/TNS
advertising data in their analysis (e. g., Boyer and Lancaster (1986), Iizuka (2004)).

The annual issue of Ad $ Summary includes a report of advertising during the
calendar year for the top 1000 advertisers in the US. market. Given the high costs of
hand coding this data and some practical issues regarding how the data is reported for
ﬁrms below the top 1000, it was not feasible for us to collect systematic data on ﬁrms
outside of the top 1000 list. Since the top 1000 ﬁrms are organizations who spend
heavily on advertising, our power to detect the presence of ﬁnancial constraints in
advertising decisions should be maximized by studying this group. However, since we
are concerned about biases that may arise by excluding observations where advertising

drops sharply, whenever a ﬁrm drops out of the top 1000 list in Ad 8 Summary we make

11

sure also to record its advertising spending in the subsequent year from a companion
publication (Company Brand $) issued by the same publisher.

For each organization listed in the top 1000, we search through the Compustat
ﬁles to match the organization to its Compustat record. Many of the listed advertisers
cannot be matched to Compustat because they are private ﬁrms, foreign ﬁrms, nonproﬁts,
or governmental bodies. Following standard practice, we exclude regulated utilities (SIC
codes 4900-4949), non-U.S. based ﬁrms, and ﬁnancial ﬁrms (SIC codes 6000-6999)
from the sample. All other observations that can be matched to a Compustat record from
1984 to 2002 are retained. We begin the sample period in 1984, since this is the ﬁrst year
that foreign income ﬁgures are available on Compustat.

The advertising ﬁgures reported in Ad 8 Summary are broken down into several
mutually exclusive categories. Ad $ Summary does not survey all US. media outlets in
each category, but it does generally survey the largest and most important outlets in the
category. In the ﬁrst sample year there were six tracked categories, and gradually over
the subsequent period several new categories were added. To maintain a consistent data
series over time, we record the advertising numbers for the six categories that were
reported for the entire sample period (magazines, network television, spot television,
network radio, outdoor advertising, and newspaper supplements). We then sum the
spending ﬁgures for these six media categories to create a variable that represents an
annual measure of total US. advertising in tracked outlets. While we will sometimes
refer to this as (tracked) U.S. advertising in what follows, it is worth noting here that it
represents only a subset of all US advertising. Presumably other domestic advertising

scales roughly proportionately with the tracked ﬁgure that we calculate.

12

1. 3.2 Summary statistics on advertising

To gauge the overall magnitude of advertising by public ﬁrms, we ﬁrst present in
Table 1 some summary statistics on total advertising expenditure (and other variables) for
the set of all Compustat ﬁrms with book assets in excess of $10 million. Since ﬁrms are
not required to report advertising spending if it is immaterial, we assign all missing
observations on advertising a value of 0. All variables used in this paper are winsorized

at the 1% and 99% points unless otherwise indicated.

(“‘3‘

For the Compustat universe as a whole, the ﬁgures in Panel A indicate that the

 

median ﬁrm has advertising expenditures of 0. This is not surprising, as one would
expect advertising to be unimportant for many types of economic activities. If we look at
mean levels of spending normalized by book assets, we see that mean advertising
spending is about one third of R&D spending (1.55% vs. 4.56%), and about a ﬁfth of the
level of capital expenditures (1.55% vs. 8.28%). When we calculate the mean ratios of
advertising to R&D and advertising to capital expenditures, these ratios are .53 and .46
respectively. These ﬁgures indicate that advertising attracts a non-trivial fraction of a
ﬁrm’s investment dollars, but certainly on average is less important than R&D or ﬁxed
capital investment.

In Panel B Of Table 1 we present the same summary statistics, but restrict
attention to ﬁrms that report at least some advertising spending (approximately 1/3 of the
Compustat universe). For this group, the mean level Of advertising normalized by assets
actually exceeds R&D spending (4.85% vs. 4.30%), and it is not too different from the

ﬁgure for capital expenditures (8.23%). Summary statistics in Panel B on the ratios of a

13

ﬁrm’s advertising to R&D and to capital expenditures also indicate that, for this group,
advertising attracts a signiﬁcant fraction of a ﬁrm’s investment dollars. Thus, conditional
on doing some material advertising, it appears that the advertising spending decision is an
important one.

In Table 2 we present summary statistics for the sample of ﬁrms in the Ad 3
Summary top 1000 list. Since our subsequent analysis requires data on foreign cash
ﬂows (described more below), these statistics are calculated over the set of observations
where foreign income on Compustat is non-zero.10 As we would expect, the ﬁgures in
Panel A conﬁrm that the top 1000 list includes heavy advertisers. The mean level of total
advertising normalized by assets is almost as large as the ﬁgure for capital expenditures
(6.20% vs. 7.20%). Certainly, for these ﬁrms, advertising is an important category in the
allocation of investment dollars.

To gauge how much of a ﬁrm’s advertising is being captured in the CMR/TNS
data source, we report in Table 2 the mean and median ratio of tracked U.S. advertising
to total Compustat advertising. These statistics are .503 and .355 respectively, indicating
that the CMR/I'N S data capture a substantial part of, but certainly not all of, a typical
ﬁrm’s advertising spending. The difference between the tracked U.S. advertising ﬁgures
and the total advertising ﬁgures should reﬂect both advertising spending in the US. in
non-tracked media outlets, and all advertising spending in foreign markets. The simple
correlation between tracked U.S. advertising normalized by assets and total ﬁrm
advertising normalized by assets is .714, indicating that there is a close relation between

the two. We would expect this high degree of correlation if ﬁrms’ advertising

 

'° Approximately 47% of all observations are cases with non-zero foreign income.

14

expenditures in non-tracked US media outlets scale approximately proportionally with

their advertising in the tracked U.S. outlets.

1.3.3 Foreign and domestic cash ﬂows

Beginning in 1984, the F ASB mandated that ﬁrms report separately pre-tax
income from their foreign and domestic operations. By subtracting off measures of
foreign and domestic taxes, we are able to construct measures of foreign and domestic
proﬁts.ll To gauge the importance of foreign operations to sample ﬁrms, we report in
Panel B of Table 2 summary statistics on foreign and domestic proﬁts for our sample
ﬁrms. The mean (median) level of foreign proﬁts normalized by assets of .023 (.015) is
approximately half (one third) of the domestic proﬁts ﬁgure of .047 (.045). Consistent
with these ﬁgures, the mean (median) ratio of foreign proﬁts to total proﬁts is .295
(.240). These ﬁgures indicate that foreign operations are an important source of proﬁts
for the typical sample ﬁrm. Consequently, variations in foreign proﬁts have the potential
to alleviate or exacerbate ﬁnancial constraints.

Turning to the relation between domestic and foreign proﬁts, we report in Table 2
that the correlation between domestic and foreign proﬁts (both normalized by assets) is
.161 (.197) measured in levels (changes). This indicates that there is a common

component to the success of a ﬁrm’s domestic and foreign operations, but also a large

 

" Properties of foreign earnings have been studied in the accounting literature (e. g., Bodnar and Weintrop
(1997)). Some income information is also available from the Compustat geographic segment ﬁles.
However, this data is more frequently missing and often does not permit separation into US. and non-U.S
components (e.g., segments listed as “North and Latin America”). We do use geographic segment data in
some of our robustness checks below.

15

independent component. The overall correlation between changes in the sum of a ﬁrm’s
foreign and domestic proﬁts and changes in total ﬁrm cash ﬂow as it is traditionally
deﬁned (income plus depreciation) is .956, indicating that these geographically based
proﬁt ﬁgures do correspond quite closely to the ﬁgures used in prior studies of
investment and cash ﬂow.

The one component of cash ﬂow that is not available separately for domestic and
foreign operations is depreciation expense. Consequently, in our initial analysis below,
we treat income as equivalent to cash ﬂow (i.e., we do not add back depreciation).
However, in the robustness checks, we explore the effect of alternative allocation rules in
assigning the ﬁrm’s depreciation charges to domestic and foreign operations. Since we
suspect that most of the within-ﬁrm time variation in cash ﬂows is likely to arise from
variation in income rather than variation in depreciation charges, the effect of
depreciation adjustments on our estimates should be limited. Moreover, the presence of
this measurement error in our cash ﬂow variable arising from depreciation charges
should, if anything, tend to bias downwards our estimates of the effect of foreign cash
ﬂows on US. advertising investment. We explore this more below.

A special issue in creating cash ﬂow measures to explain advertising arises from
the fact that advertising expenditures are expensed in the calculation of a ﬁrm’s income.
Thus, the dependent variable (advertising) has effectively been subtracted off from an
explanatory variable of interest (cash ﬂow). This is not an issue in studies of ﬁxed
capital investment, since capital expenditures are not expensed, and thus cash ﬂow is
measured before any subtractions for ﬁxed investment. To adjust our data accordingly,

in what follows we add the tracked U.S. advertising expense back to US income in our

16

 

calculation of US. cash ﬂow. This effectively creates a domestic cash ﬂow measure that
is measured independent of the ultimate investment choice being considered (tracked
U.S. advertising). From a theoretical perspective, this would appear more appropriate
than not adding back the advertising ﬁgures. Moreover, as a practical matter, this is a
conservative choice, since our results are even stronger if we do not make this
adjustment. We discuss some alternative treatments of this issue in our robustness checks

below.

1.4 Models of advertising investment and cash ﬂow

1.4.1 Firm advertising and ﬁrm cash flow

Since the determinants of advertising investment have not been widely explored,
we ﬁrst present a brief multivariate analysis of the relation between advertising and
variables that have been associated in prior research with capital investment decisions.
For comparison purposes, we also present corresponding results for ﬁxed capital
investment. As has been widely discussed in the literature and is noted above, these
estimates cannot be regarded as causal. They are simply meant to provide a sense as to
patterns in total advertising spending at the ﬁrm level. We use the same sample period as
in our later analysis (1984-2002), and include all US. based Compustat listed ﬁrms with
book assets in excess of $10 million except those in the ﬁnancial and regulated utility
industries.

In column 1 of Table 3 we present a traditional investment-cash ﬂow regression

where the dependent variable is capital expenditures (normalized by assets) and the

17

independent variables are ﬁrm cash ﬂow (normalized by assets) and Tobin’s Q. Firm
ﬁxed-effects and year effects are included in the estimated model, but these estimates are
not reported. The estimates in column 1 are similar to what has been reported by others
for similar samples (e.g., Baker, Wurgler, and Stein (2003)). The coefficient of .069 on
cash ﬂow is highly signiﬁcant, and could be indicating either the widespread presence of
ﬁnancial constraints, or a correlation between cash ﬂow and investment Opportunities (or
both).

In column 2 we present corresponding estimates with advertising expenditures
(normalized by assets) as the dependent variable. 12 The cash ﬂow coefficient in this
regression is smaller than in the ﬁxed investment regression (.019 vs. .069), but is still
highly signiﬁcant. The smaller coefficient on cash ﬂow in this model is likely to reﬂect
the presence of many ﬁrms that do little or no advertising. In column 3 we restrict
attention to ﬁrms that have reported at least some positive advertising either as of the
observation year or at some point prior to the Observation year. In this model, the
coefﬁcient on cash ﬂow is substantially larger in magnitude (.035 vs. .019) and is highly
signiﬁcant. In column 4, we restrict attention to ﬁrms in our list of top 1000 advertisers.
In this model, the coefficient on cash ﬂow grows substantially in magnitude to .171.

The estimates in Table 3 indicate that advertising is correlated with the same
factors that are associated with capital investment decisions, namely ﬁrm cash ﬂow,
Tobin’s Q (in some models), and lagged sales growth. For sub-samples of ﬁrms that do
more advertising, the estimated relation between advertising and ﬁrm cash ﬂow is larger.

Overall, the ﬁndings in Table 3 are broadly consistent with the idea that the advertising

 

”Following the argument we outline above, we add total advertising back to cash flow in our regression
models of advertising (column 2-4 of Table 3), but not in our model of ﬁxed investment (column 1).

18

spending decision shares some important similarities with the ﬁxed investment decision.
We now turn to investigating whether exogenous variations in cash ﬂow play a role in

these decisions by exploiting data on the location of both cash ﬂows and spending.

1. 4.2 US. Advertising and Foreign Cash Flow

We model U.S. advertising as a ﬁmction of domestic cash ﬂow, Tobin’s Q, and
foreign cash ﬂow. The controls for US. cash ﬂow and Tobin’s Q are intended to control
for US. investment opportunities. Thus, the coefficient on foreign cash ﬂows should
provide us with information on the role of variations in cash ﬂow on advertising
investment, holding constant the relevant investment opportunity set.

A few special issues arise in this empirical modeling. First, as discussed above, in
our calculation of domestic cash ﬂow we add the tracked U.S. advertising expenses back
to domestic income since these expenditures, which we are seeking to explain, are
subtracted out (i.e., expensed) in the calculation of domestic income. Second, Tobin’s Q
will be a noisy measure of domestic investment opportunities, since it measures the
ﬁrm’s investment prospects both domestically and abroad. An alternative, suggested by
the work of Shin and Stulz (1998), is to use the median Tobin’s Q of the purely domestic
ﬁrms that are in the same industry as a sample ﬁrm. We prefer to use a ﬁrm’s own Q,
since this measure uses ﬁrm speciﬁc information and the majority of most ﬁrms’ incomes
are derived domestically. However, in our robustness checks below, we explore some
potential alternatives suggested by the work of Shin and Stulz (1998).

A ﬁnal modeling issue that arises is identifying the correct estimation framework.

In models of capital expenditures, most authors rely on a ﬁxed-effects estimator.

l9

However, a preliminary analysis of the advertising data suggests extreme serial
correlation in advertising spending.l3 In these settings, Wooldridge (2002) argues that
estimating a model in ﬁrst-differences is generally a more efficient estimation procedure.
Consequently, in what follows we emphasize ﬁrst-difference estimates. However, as an
alternative, we also report results for ﬁxed-effects estimators that allow for, and estimate,
an underlying AR(1) structure for the residuals using the Baltagi and Wu (1999)
procedure. '4

We present our baseline model explaining tracked U.S. advertising in column 1 of
Table 4 (estimated in ﬁrst differences). The coefﬁcient on US. cash ﬂow of .055 is
positive and highly signiﬁcant (tr-8.01). For the usual reasons, the correct interpretation
of this coefﬁcient is unclear. The coefﬁcient on Tobin’s Q is small (.001), but positive
and highly signiﬁcant, suggesting that the Q-theory of investment has some relevance for
explaining advertising spending. Most importantly for our purposes, the coefficient on
foreign cash ﬂows of .033 is positive and signiﬁcant (t=2.40). As we argue above, a
positive sign on this coefﬁcient provides evidence that the availability of internal cash
plays a role in a ﬁrm’s advertising investment decisions. This is the key ﬁnding in our
paper, and thus we present a battery of relevant robustness checks and extensions below.

Before proceeding, we consider the magnitude of the estimated effect. Our earlier

summary statistics indicate that the advertising we track in our dependent variable

 

’3 For example, when we estimate ﬁxed-effects models and then regress residuals against lagged residuals,
the coefﬁcient on lagged residuals is large (around .75), with a very high t-statistic. When we estimate
models in ﬁrst-differences, the analogous coefﬁcient on lagged residuals is close to 0 and insigniﬁcant.

'4 We thank Jeff Wooldridge for clarifying these issues for us. Our sample is not large enough to ignore
estimation efficiency. Since the data suggest that the ﬁrst difference estimator appropriately controls for
serial correlation, the estimates in Tables 4 and 5 are robust only to heteroskedasticity. However, the
results are qualitatively unchanged if we cluster the ﬁrst—difference standard errors by ﬁrm to account for
any arbitrary residual serial correlation.

20

represents 35.5% of the median ﬁrm’s total advertising spending. Thus, if all forms of
advertising scale up proportionately with increases in internal cash ﬂow, the coefﬁcient
estimate of .033 for tracked U.S. advertising would imply a total increase of (.033/.355) =
$093 in total advertising for each additional dollar of cash ﬂow. Pushing these
calculations further, the median sample ﬁrm has a ratio of total advertising to capital
expenditures of .441. Thus, if capital expenditures also scale proportionally with
advertising, we would expect an increase of .093/.441 = $211 in capital expenditures.
Adding the two ﬁgures would imply a total spending increase of approximately $.30.
These ﬁgures, while smaller than the estimates reported by Rauh (2006), would suggest
an overall important economic effect of cash ﬂow variation on investment activities.

In column 2 of Table 4 we present estimates for the same model, but where we
use a ﬁxed-effects estimation framework with an assumed AR(1) disturbance for the
error term. The signs and signiﬁcance levels of these estimates are similar to our ﬁrst
difference estimates. The coefﬁcient on foreign cash ﬂow in this column of .024 is
slightly smaller than the ﬁrst-difference estimate, but it remains signiﬁcant (t=2.06) and
is again suggestive of a reasonably large effect of cash ﬂow on total investment.

Since some ﬁrms in our sample are much heavier advertisers than others, we
might expect to observe heterogeneity in the size of the effect of cash ﬂow on advertising
spending. To investigate, we separate the sample into big advertisers and small
advertisers based on whether the ﬁrm’s lagged U.S. advertising exceeds the sample
median. Using the ﬁrst-difference estimator, we present coefﬁcient estimates for big
(small) advertisers in column 3 (column 4) of Table 4. While the foreign cash ﬂow

coefﬁcient is signiﬁcant for both groups, the magnitude of the coefﬁcient is substantially

21

larger for the big advertisers compared to the small advertisers (.052 vs. .008). The
difference between these two coefﬁcients is signiﬁcant (t=2.04). This suggests that the
effect of cash ﬂow variation on investment activity (in this case advertising) is
proportional to the intensity of the activity.

In columns 5 and 6 we again estimate models for big and small advertisers, but
using the ﬁxed-effects AR(1) estimation procedure. Similar to the ﬁrst-difference
estimates, the coefﬁcient on foreign cash ﬂow is signiﬁcant for both groups, but
substantially larger for the big advertisers compared to the small advertisers (.046 vs.
.009). However, in these models the difference between the two coefﬁcients is not quite
signiﬁcant (t=1.62). Certainly the character of these results is similar to the ﬁrst-
difference estimation in that they suggest a larger role for cash ﬂow in the sub-sample of

ﬁrms where one would expect a relatively larger effect.15

1.4.3 Robustness

Our main ﬁnding above is that foreign cash ﬂow is related to US advertising
even after we control for US advertising investment opportunities via Tobin’s Q and
US. cash ﬂows. In this subsection we consider a host of robustness checks concerning
this ﬁnding. To conserve space, the estimates underlying these robustness checks are not

reported in the tables but are available from the authors upon request.

 

’5 To consider a measure of intensity that is independent of a ﬁrm’s choices, we also experimented with
using a ﬁrm’s (4-digit) industry advertising intensity. When we interact this variable with foreign cash
ﬂow, the coefﬁcients are positive and signiﬁcant at the 10% level (5% level) using the ﬁrst-difference
(ﬁxed-effects AR(1)) estimation framework. These ﬁndings are broadly consistent with the reported
results on big and small advertisers.

22

Alternative controls for US. investment opportunities

In the models of Table 4, we include Tobin’s Q as a control for a ﬁrm’s U.S.
advertising investment opportunities. This variable will be a noisy measure of domestic
investment opportunities, since it will also include information regarding a ﬁrm’s foreign
Operations. As an alternative, we could identify measures that are speciﬁc only to a
ﬁrm’s US. investment opportunities. To accomplish this, we follow the corresponding
analysis in Shin and Stulz (1998) and calculate the median Tobin’s Q over the set of all
ﬁrm’s with no reported foreign income in the same 2-digit industry as the sample ﬁrm.
We then estimate models 1 and 2 of Table 4 with this variable used in place of the ﬁrm’s
Tobin’s Q. When we estimate this model, the coefﬁcient on foreign cash ﬂow maintains
its positive sign and increases slightly in magnitude and signiﬁcance. While the estimate
on industry Q is positive, it is smaller than what we report in Table 4 and is insigniﬁcant.
This indicates that a ﬁrm’s Tobin’s Q is a better predictor of subsequent U.S. advertising
than the industry-based pure U.S. measure.

Similar to Shin and Stulz (1998), we also experiment with using a ﬁrm’s lagged
US. income growth rate as a measure of US. advertising investment opportunities.
When we include this variable in place of Tobin’s Q in models 1 and 2 of Table 4, it is
positive but not signiﬁcant. Most importantly, for our purposes, the coefﬁcient on
foreign cash ﬂow remains positive and again increases slightly in magnitude and
signiﬁcance. Thus, it appears that our ﬁndings are not sensitive to how we measure U.S.

growth opportunities.

23

Alternative deﬁnitions of cash ﬂow

As we discuss above, depreciation ﬁgures pertaining separately to domestic and
foreign income are not available. Consequently, the cash ﬂow variables we use above do
not have the relevant depreciation ﬁgures added back to them. To make an adjustment
for depreciation, we experiment with allocating depreciation expenses to a ﬁrm’s US.
and foreign operations. In particular, we assume that the fraction of the ﬁrm’s assets that
are allocated to domestic (foreign) income is given by the fraction of the ﬁrm’s pre-tax
income that is derived from domestic (foreign) sources. After allocating depreciation
expense in this manner, we add the corresponding depreciation charges back to US. and
foreign cash ﬂow and estimate the speciﬁcations in models 1 and 2 of Table 4 using this
alternative deﬁnition of cash ﬂow. The results with this modiﬁcation are very similar to
what we report in Table 4. In particular, the role of foreign cash ﬂow in explaining U.S.
advertising remains positive and signiﬁcant.

As we discuss earlier, our domestic cash ﬂow variable is constructed by adding
domestic advertising expenses back to domestic income. We construct the variable in
this way so that, in keeping with the prior literature on investment and cash ﬂow, we can
explain domestic (tracked) advertising investment as a function of the funds available for
this investment. However, if advertising has an immediate impact on a ﬁrm’s sales, then
part of the domestic cash ﬂow variable may include the feedback effect of domestic
advertising on domestic sales. To consider this possibility, we experiment with not
adding tracked domestic advertising spending back to domestic income when
constructing the domestic cash ﬂow ﬁgure. This is the appropriate treatment if each

dollar of domestic advertising immediately creates an additional dollar of domestic sales,

24

since the revenues associated with the advertising from this feedback effect are then
effectively eliminated by expensing advertising in the calculation of domestic cash ﬂow.
Clearly this is the other extreme from our (implicitly) maintained assumption above that
current advertising has no contemporaneous effect on sales.

If we estimate models 1 and 2 of Table 4 using this alternative deﬁnition of
domestic cash ﬂow, the coefﬁcient on foreign cash ﬂow becomes larger in magnitude and
more signiﬁcant. Thus, if we assume that advertising has a rapid effect on sales, our
results appear to be even stronger. In addition, if we add only a fraction of a ﬁrm’s
domestic advertising expense back to cash ﬂow (reﬂecting an assumption that each dollar
of domestic advertising results in a contemporaneous increase in domestic sales of $.25,
$.50, or $.75) the results are as strong as or stronger than what we report in Table 4.

Thus, our results appear fairly robust to any feedback effect of advertising on sales.
Since the dynamics of how advertising affects sales is an unsettled issue, we prefer to rely
on the conservative approach used above where we implicitly assume a negligible

immediate impact of advertising on sales. ’6

Sample selection concerns

The sample we choose is conditioned on a ﬁrm appearing in the list of top 1000
US. advertisers. In many ways this is a natural selection criterion, since advertising is
most interesting to study in a population of ﬁrms with nontrivial advertising

expenditures. However, ﬁrms that are on the border of being included in this list will

 

'6 We also experimented with adding the ﬁrm’s total Compustat advertising to domestic cash ﬂow and,
alternatively, to both domestic and foreign cash ﬂow (using a pro-rata income-based rule). The results with
these modiﬁcations are also quite similar to what we report in Table 4.

25

tend to appear more frequently (less frequently) in our sample in years where their
advertising is unexpectedly large (small). This could create a negative correlation
between the error term and the explanatory variables that appear positively related to
advertising (e.g., cash ﬂow and Q) for these borderline ﬁrms, likely biasing downwards
our estimates. In our modeling above we partially attenuate this problem by collecting
and including in our sample advertising ﬁgures for the ﬁrst year after a ﬁrm drops out of
the top 1000 list. This data collection is quite cumbersome, since it requires collecting
information from an alternative data source (Company Brand $) from the same publisher
but reported in a different format.

To assure that these sample selection issues have no substantive effect on our
results, we estimate our model on a sub-sample of ﬁrms where the effect of sample entry
and exit should be unimportant. Speciﬁcally, we include ﬁrms only after they have been
in the top 1000 for three consecutive years. We use advertising data going back to 1980
to determine the number of years a ﬁrm has been listed in the top 1000. Once a ﬁrm
drops out of the top 1000, we include the ﬁrm in the ﬁrst year of their exit and never let
them re-enter the sample. These procedures assure that we should eliminate the
borderline ﬁrms that enter or re-enter the sample when advertising is particularly high.
When we estimate models 1 and 2 of Table 4 on this subset, the results are qualitatively
unchanged from what we report above. Thus, our results do not appear to be sensitive to

these sample selection concerns.

26

Miscellaneous speciﬁcation and data issues

The ﬁnancial data we use pertains to ﬁscal years while the advertising data we use
is reported on a calendar year basis. While in most cases ﬁscal years and calendar years
are the same (approximately 2/3 of our sample), for ﬁrms with non-December ﬁscal year
endings there will not be a complete overlap between ﬁscal years and calendar years. To
assure that these timing issues have no substantive effect on our results, we restrict our
sample to observations where the ﬁrm has a December ﬁscal year ending. Using this
sub-sample, we obtain similar estimates to what we report above for models 1 and 2 of
Table 4. We also experiment with matching the advertising data to the accounting data
for the ﬁscal year that ends on or before the end date of the advertising data calendar
year, effectively lagging the accounting data for ﬁrms with ﬁscal year endings in January
through May. This alteration again has no substantive effect on the results in models 1
and 2 of Table 4. In particular, we continue to ﬁnd a signiﬁcant relation between foreign
cash ﬂows and US. advertising investment.

In our analysis above, we follow the fairly standard practice in this literature of
winsorizing all variables at the 1% and 99% points. This is intended to lessen the impact
of outliers, which can often be quite large when using ratios based on accounting data. If
we do not winsorize any variables, the estimated coefficient on foreign cash ﬂows in
models 1 and 2 of Table 4 become larger in magnitude and the t-statistics increase
slightly. If we do not winsorize the dependent variable but continue to winsorize the
explanatory variables, the coefﬁcients on foreign cash ﬂow fall slightly in signiﬁcance to
the 7% level in both models 1 and 2 of Table 4 (t=1.88 and 1.91 respectively). Taken as

a whole, it appears that our conclusions concerning the role of foreign cash ﬂows in US.

27

advertising are fairly robust to the treatment of winsorizing. We prefer to rely on the
results with winsorization of all variables, since these are likely to be more informative
given the skewed nature of the accounting variables.

The speciﬁcations in Table 4 assume that the effect of cash ﬂow on investment
occurs in the same annual period (i.e., the effect is contemporaneous). This is in keeping
with the related previous literature. However, it is possible that there is also a delayed
effect so that increased cash ﬂow in any given year results in increased spending both in
the current year and the subsequent year. To investigate, in unreported results we added
to models 1 and 2 of Table 4 lagged versions of all three explanatory variables — domestic
cash ﬂow, foreign cash ﬂow, and start-of-period Q. In both models the coefﬁcient on
lagged domestic cash ﬂow was positive and signiﬁcant (although much smaller than the
contemporaneous coefﬁcient), while the coefficients on lagged Q and lagged foreign cash
ﬂow were insigniﬁcant. In these models the estimated coefficients on the
contemporaneous variables maintain the same signs, signiﬁcance levels, and approximate
magnitudes as we report in the table. This evidence demonstrates that our ﬁndings are
robust to including lags and also indicates that the effect of cash ﬂow on advertising

investment is fairly immediate.

Alternative explanations

The checks we discuss above indicate that there is a fairly robust relation between

a ﬁrm’s foreign cash ﬂows and its domestic advertising investment decision. While it is

28

hard to envision a link between foreign cash ﬂows and domestic advertising that is

independent of a ﬁnancial constraints story, we do consider here some possibilities.l7

I_ax Explanations

One possible explanation for our ﬁndings arises from the possibility that ﬁrms use
the leeway provided by the accounting system to shift domestic earnings to foreign
operations for tax purposes. If ﬁrms behave in this manner, foreign earnings may contain
some incremental information regarding a ﬁrrn’s US. operations that could impact their
optimal advertising decision. To investigate, we construct a dummy variable that
indicates whether a ﬁrm’s domestic tax rate exceeds its foreign tax rate. Following the
approach of Hines (1996), we assume the ﬁrm’s foreign (domestic) tax rate is equal to its
foreign (domestic) taxes divided by foreign (domestic) pre-tax income. Domestic taxes
are assumed to equal total taxes less foreign taxes. For each annual observation, this tax
rate variable is calculated using income from the year prior to the observation year.

If our results are explained by foreign income containing “hidden” US income,
we would expect this effect to be largest for ﬁrms with a relatively higher domestic tax
rate and therefore a stronger incentive to move earnings elsewhere. However, when we

interact the foreign cash ﬂow variable with a dummy variable indicating whether the

 

17Ideally we would like to instrument for foreign cash ﬂows using exogenous variables such as currency
changes. However, given the coarse nature of our income data (i.e., it is not available at the country level),
we could not identify a variable of this type that was sufﬁciently strongly related to foreign cash ﬂows to
serve as a meaningful instrument. The weak relation between currency changes and foreign income has
been noted by Lee and Suh (2004).

29

ﬁrm’s domestic tax rate exceeds its foreign tax rate, the interaction terms in model 1 and
model 2 of Table 4 are positive and insigniﬁcant. The interaction terms remain
insigniﬁcant if we also add to these speciﬁcations the high domestic tax dummy variable.

Thus, there is little evidence that tax shifting opportunities are driving our ﬁndings.

Cost allocation discretion

It is possible that a ﬁrm’s geographically based earnings ﬁgures are noisy
indicators of cash ﬂow performance because of arbitrary non-economic rules regarding
how a ﬁrrn’s costs are allocated across units. For example, if a ﬁrm is having a poor year
domestically, there may be organizational incentives to try to allocate a disproportionate
share of the ﬁrm’s costs in that year to the ﬁrm’s foreign activities. To evaluate whether
this type of behavior generates our ﬁndings, in unreported results we experiment with
using foreign sales as an instrument for foreign cash ﬂow in the baseline model of
column 1 in Table 4.’8 When we estimate this model using two-stage least squares, the
coefﬁcient on foreign cash ﬂow in the domestic advertising (i.e., second- stage)
regression is positive and highly signiﬁcant (F362). Thus, it appears that our results
hold even if we restrict attention to foreign cash ﬂow variation that arises from changes

in foreign sales, indicating that cost shifting is not the driving force behind our ﬁndings.

Earnings Management

 

'8 Since the estimated model is in changes, we use change in foreign sales as an instrument for change in
foreign cash ﬂows. Sales data is only available on the Compustat geographic segment tapes.
Consequently, sample sizes are reduced when we undertake this instrumental variables analysis. We retain
in the IV analysis only ﬁrms where one segment is listed as the U.S. and all other segments are identiﬁed
with areas that do not include the US. The sum of sales for these other segments is deﬁned to be foreign
sales.

30

Since intangible investments such as advertising and R&D are expensed for
ﬁnancial reporting purposes, it is possible that ﬁrms out these expenditures in an effort to
increase reported earnings when the ﬁrm is in danger of not meeting the market's
expectations. This behavior could generate a causal relation between advertising and cash
ﬂow, but for reasons that are very different from traditional thinking on how ﬁnancial
constraints are believed to affect investment decisions. To investigate, we identify ﬁrms
that are likely to be relatively more affected by earnings management incentives and
proceed to investigate whether they display a heightened sensitivity of advertising to cash
ﬂow.

Our ﬁrst approach is to identify observations where a ﬁrm just barely met or
missed analyst earnings expectations.19 Using the I/B/E/S database, we identify these
ﬁrms by looking at the difference between a ﬁrm's actual earnings per share ﬁgure for a
year and the consensus analyst estimate for this ﬁgure as of the start of the ﬁscal year.
We create dummy variables indicating whether the ﬁrm's earnings numbers were (a)
above the estimate by $.01 or less, (b) above the estimate by $.02 or less, (c) above or
below the estimate by $.01 or less, (d) above or below the estimate by $.02 or less (this
latter category encompasses approximately 20.46% of the sample). We select these
dummy variables (one at a time) and add the dummy variable and its interaction with
domestic cash ﬂow and foreign cash ﬂow to model 1 of Table 4. In all cases the
estimated coefﬁcients on the added variables are insigniﬁcant, while the coefﬁcient on

foreign cash ﬂow remains positive and signiﬁcant. These conclusions are qualitatively

 

’9 There is a large literature that attempts to identify earnings management by studying cases where there is
a small difference between actual and expected earnings. See Degeorge, Patel, and Zeckhauser (1999) and
Brown (2001).

31

unchanged if we instead use dummy variables based on the difference between a ﬁrm's
actual earnings ﬁgures and the penultimate consensus analyst forecast for the ﬁscal year.
This investigation Offers no evidence that advertising-cash ﬂow sensitivities are
magniﬁed in cases where ﬁrms are under particular pressure to meet earnings numbers.
Our second approach for examining earnings management effects is to select
ﬁrms that have high levels of current discretionary accruals. We follow the exact
methodology of Teoh, Welch, and Wong (1998) and create a discretionary accruals
measure based on residuals from a regression model explaining current accruals.20
Following Teoh, Welch, and Wong (1998), we assign observations that are in the top
quartile of the sample based on this measure into the aggressive earnings manager
category. We create a dummy variable for this category and add the dummy variable and
its interaction with domestic cash ﬂow and foreign cash ﬂow to model 1 of Table 4.
When we estimate this model, the coefﬁcients on these added variables are small and
insigniﬁcant, while the coefficients on the other variables maintain the same signs and
signiﬁcance levels as reported in the table. This suggests once again that our results are

unlikely to be driven by cases where ﬁrms are aggressively managing earnings.

Potential linki between domestic and foreign markets

If ﬁrms earn a substantial part of their foreign income in countries that are
contiguous to the US. (e. g., Mexico or Canada), it may be that domestic advertising is
intended to build the ﬁrm’s brand in these foreign markets via spillover effects ﬁ'om

media outlets near borders. In this case, foreign cash ﬂow would not be exogenous in the

 

2° Identifying earnings management by the use of discretionary accruals is a common approach in the
accounting literature. For an early study of this type, see Jones (1991).

32

domestic advertising decision. To investigate, we examine the Compustat geographic
segment ﬁles and create a dummy variable named FAR which is intended to indicate that
the ﬁrm’s foreign Operations are concentrated in areas far from the US. market. This
variable assumes a value of 1 if the ﬁrm has foreign segments and none of these segments
include Mexico, Canada, or a Central American country. If the ﬁrm has foreign
segments that include these regions, the FAR variable is set equal to 0. For all other
ﬁrms, the variable is set equal to missing.”

When we add to model 1 (model 2) of Table 4 an interaction of foreign cash ﬂow
with the FAR variable, the estimated coefficient on the interaction is small, negative
(positive), and insigniﬁcant. If we add the FAR variable along with the interaction of
FAR with foreign cash ﬂows to these Table 4 models, the interaction terms are again
insigniﬁcant. Thus, we ﬁnd no convincing evidence that the relation between foreign
cash ﬂows and domestic advertising is related to a measure of the closeness of the ﬁrm’s
foreign cash ﬂows to domestic locations. This indicates that a channel though which
domestic advertising is intended to respond to foreign investment opportunities by
appealing to nearby foreign markets is an unlikely explanation for our main ﬁndings.

For a few types of products, it is possible that domestic advertising is intended to
appeal directly to purchasers in foreign markets (and thus might be responsive to foreign
performance). A leading example that has been suggested to us is the case of apparel,
where fashion trends in the US. may increase demand for the product overseas. The

other important case is the travel industry, as advertising in the US may affect purchase

 

2' Some ﬁrms are not on the geographic segment tapes. In addition, changes in segment reporting rules
make it difﬁcult to use this data to create consistent variables over time. Thus, we choose not to use
geographic segment data in our main analysis of the advertising-cash ﬂow relation presented earlier.

33

decisions made by individuals exposed to the advertising when they travel to foreign
markets. To investigate, we create a dummy variable that assumes a value of l for ﬁrms
in the clothing or travel industries [SIC codes 2300-2390, 3020-3021, 3100-3199, 4481,
4500-4581 , and 7010-7011]. When we interact this variable with the foreign cash ﬂow
variable in models 1 and 2 of Table 4, the estimated coefﬁcients on the interaction terms
are insigniﬁcant. Thus, it does not appear that our results are driven by industries where
there may be a connection between foreign performance and the optimal level of
domestic advertising.

Our maintained assumption is that foreign cash ﬂows provide no information
concerning the optimal level of US. advertising once U.S. cash ﬂows and Q have been
taken into account. Consequently, how a ﬁrm is doing overseas should have no marginal
effect on its decision to build up its brand in the US, unless ﬁnancial constraints are at
work. Notwithstanding our efforts above, one might still be concerned that we are
missing some link between foreign cash ﬂows and a ﬁrm’s proﬁt potential in the US.
that may cause it to increase its US. advertising.22

If a ﬁrm's foreign cash ﬂows are related to its domestic prospects, we would
expect this effect to be magniﬁed for ﬁrms with a particularly large foreign presence. For
these ﬁrms foreign cash ﬂows are more likely to be related the overall success of a ﬁrm's
underlying business strategy, and thus they have a greater potential to provide

incremental information on the ﬁrm's domestic opportunities. To investigate, we create a

 

22 We have investigated the possibility that a change in a ﬁrm's foreign cash ﬂow that has the same sign as
the change in the ﬁrm's domestic cash ﬂow may have a different impact on domestic advertising than other
foreign cash ﬂow changes. This could arise if there is something special about foreign proﬁt signals that
echo domestic proﬁt signals. For example, increases in both domestic and foreign cash ﬂows may indicate
a new product introduction that is accompanied by heavy advertising. We ﬁnd no evidence of a
signiﬁcantly enhanced role of foreign cash ﬂows in domestic advertising for these types of cash ﬂow
changes.

34

 

dummy variable indicating whether a ﬁrm derived more than 50% of its (lagged) total
proﬁts from foreign operations (corresponding approximately to the sample 75th
percentile). We then add this dummy variable and its interaction with foreign cash ﬂow
to models 1 and 2 of Table 4. The coefﬁcient on the interaction term is small and
positive in model 1, small and negative in model 2, and in both cases insigniﬁcant. Thus,
we ﬁnd no convincing evidence that the relation between foreign cash ﬂows and
domestic advertising is particularly large for ﬁrms with an enhanced foreign presence.
This suggests that our ﬁndings are not driven by a relation between foreign cash ﬂows
and domestic investment opportunities.

To further investigate the potential relation between foreign cash ﬂows and
domestic investment opportunities, we explore the link between a ﬁrm’s foreign cash
ﬂows and its subsequent proﬁt growth in the US. In particular, we regress the change in
a ﬁrm’s U.S. cash ﬂow (domestic income minus domestic taxes) between year t and t+3
against foreign cash ﬂow in year t, Q in year t, and year dummies. All cash ﬂow
variables are normalized by (beginning of period) year t assets. We then repeat this
procedure using a one year growth window and a ﬁve year growth window. The
coefﬁcient on foreign cash ﬂow in these regressions is negative in all cases, and typically
insigniﬁcant. If we allow for a ﬁrm-speciﬁc effect in these models by estimating the
regression equations in ﬁrst differences, the foreign cash ﬂow coefﬁcients are again
always negative. Thus, the data offer no evidence that domestic proﬁt growth is

positively related to foreign cash ﬂow performance.

1.5 Financial constraint measures and the advertising-cash flow relation

35

1.5.1 Measures of financial constraints

The results we present above establish that there is a relation between US.
advertising and foreign cash ﬂow. This ﬁnding appears robust to a variety of alternative
speciﬁcations. The results are exactly what we would expect if variations in foreign
income are transmitted through the internal capital market to a ﬁrm’s domestic operations
because of ﬁnancial constraints, and we uncover little support for alternative
explanations.

In this subsection we consider differences across ﬁrms in the US. advertising-
foreign cash ﬂow sensitivity (hereafter referred to as the advertising-cash ﬂow
sensitivity). Following the tradition in this literature, we group ﬁrms by characteristics
that have been associated with an increased likelihood of ﬁnancial constraints. We then
examine whether advertising-cash ﬂow sensitivities vary across these different groups. If
one is conﬁdent that advertising-cash ﬂow sensitivities measure ﬁnancial constraints, this
can be viewed as an exploration of which ﬁrm characteristics are associated with
ﬁnancial constraints. This is our view of the analysis. However, if one is conﬁdent that
the identiﬁed ﬁrm characteristics accurately sort ﬁrms by the severity of ﬁnancial
constraints, one can view this analysis as a further test of our interpretation of
advertising-cash ﬂow sensitivities.

Different measures of ﬁnancial constraints have been used by different authors.
We choose variants of measures suggested by two prominent recent studies in this area --
Ahneida, Campello, and Weisbach (2004) and Rauh (2006). In particular, we group

ﬁrms by age, size, dividend payout ratio, S&P credit rating, leverage, and cash

36

holdings.23 Many of these variables are endogenous, and it is not always clear what the
expected relation with ﬁnancial constraint status should be. For example, some authors
have argued that ﬁrms with lots of cash should be viewed as relatively unconstrained
because they have access to this source of funds. However, it is possible that these ﬁrms
are actually relatively more constrained, since their choice of a large buffer stock of cash
could indicate that they anticipate problems in raising external capital. Prior work
generally assumes that younger ﬁrms, smaller ﬁrms, highly levered ﬁrms, and ﬁrms with
low dividends, credit ratings, and cash balances are relatively more likely than others to

be ﬁnancially constrained. We search for evidence of this in our sample.

1.5.2 Empirical analysis

We select ﬁrms by whether the characteristic associated with ﬁnancial constraints
is in the top or bottom tercile, with the exception of credit ratings, which we discuss
ﬁirther below.24 We assign ﬁrms to groups based on start of year characteristics for stock
variables, and previous year characteristics for ﬂow variables. After assigning ﬁrms into
the relatively constrained and unconstrained groups, we estimate our regression models
explaining U.S. advertising for each group separately.

Our measure of age is the number of years since the ﬁrm’s ﬁrst Compustat record

with a non-missing stock price. Our measure of ﬁrm size is inﬂation adjusted book value

 

23 Rauh (2006) uses cash minus debt as a sorting criterion rather than cash and debt separately. However,
since cash and debt may not be perfect substitutes (see Acharya, Almeida, and Campello (2005)), we break
these characteristics out separately. Given that we are studying advertising intensive ﬁrms, we do not use
Rauh’s (2005) measure based on capital expenditures.

2‘ Almeida, Campello, and Weisbach (2004) and Rauh (2006) also sort ﬁrms by terciles. In a previous
draft we separated ﬁrms based on whether they were above or below the median. As one might expect,
these results were similar in character, but not quite as stark, as the results based on comparing top and
bottom terciles.

37

of assets. The dividend payout ratio is deﬁned as dividends over net income. Leverage is
deﬁned as short-terrn debt plus long-term debt divided by book assets. Cash holdings are
cash plus marketable securities divided by book assets. A ﬁrm’s S&P credit rating is
deﬁned to be the ﬁrm’s S&P senior debt rating. In the case of debt ratings, we separate
ﬁrms into rated ﬁrms with ratings above BBB+, rated ﬁrms rated at or below BBB+, and
unrated ﬁrms. We do not include the unrated ﬁrms with either of the other groups, since
we suspect that unrated ﬁrms include a mix of different types of ﬁrms. For example, this
group may include unconstrained ﬁrms that are not rated because they choose not to
borrow at a level that would to justify selling public debt, and constrained ﬁrms that
simply cannot access the public debt markets.

Regression results using the ﬁrst-difference estimator for sub-samples grouped by
these ﬁrm characteristics are reported in Table 5.25 The results are largely consistent with
our expectations. In particular, in all but one case (cash holdings) the point estimate on
the foreign-cash ﬂow coefficient is relatively larger for the group that we expect to be
more constrained - that is younger ﬁrms, smaller ﬁrms, ﬁrms with poor debt ratings,
ﬁrms that pay low dividends, and ﬁrms with high leverage. In addition, the coefﬁcient
on foreign cash ﬂow is signiﬁcant for most of the sub-samples that we label more
constrained, but is in most cases not signiﬁcant for the sub-samples that we label less
constrained. Moreover, for the age, dividend, leverage, and credit rating sorts, the
difference in the estimated foreign cash ﬂow coefficient across the sub-samples is

signiﬁcant at the 5%, 10%, 5% and 1% levels respectively.

 

2’ If we estimate the models in Table 5 with the ﬁxed-effects AR(1) estimator rather than the ﬁrst-
difference estimator, the results are similar in character to what we report in the table.

38

Taken as a whole, the evidence we present is broadly supportive of the notion that
ﬁrms that appear more likely to be constrained on a priori grounds tend to display a
higher sensitivity of advertising investment to exogenous variations in internal cash ﬂow.
Interestingly, the coefﬁcient on US. cash ﬂow in Table 5 is not unifome greater in the
constrained groups compared to the unconstrained groups. This echoes some of the
empirical ﬁndings of Rauh (2006) and theoretical arguments of Kaplan and Zingales
(1997), Alti (2003), and Moyen (2004) suggesting that the magnitude of the sensitivity of
investment to cash ﬂow is not necessarily an informative measure of ﬁnancial constraints

when cash ﬂow is correlated with investment Opportunities.

1.5.3 Hedging

The variation in a ﬁrm’s foreign cash ﬂows can be partially controlled by a ﬁrm
via their hedging policies. In particular, following the logic of Froot, Scharfstein, and
Stein (1993), ﬁrms may use currency derivatives to minimize their exposure to currency
changes that could impact their ability to fund promising investments. If this occurs,
variations in a ﬁrm’s net cash ﬂow may have a smaller impact on investment choices, as
the f'um is less likely to be in a region where there is a mismatch between funding
sources and funding needs.

To investigate, we gather information on sample ﬁrms’ hedging activity.
Collecting this data is expensive, as it requires hand collection from ﬁnancial statement
footnotes. Moreover, reliable information on hedging is limited in the earlier part of our
sample period because of limited disclosure requirements. As a practical way of getting

some hedging information, we examine all available IO-K ﬁlings and annual reports for

39

sample ﬁrms as of 1993 (the sample period midpoint). If a search of this document
contains any reference to the use of currency futures, forwards, options, or swaps, we
assign the ﬁrm to the hedging category (for all years). If the document reveals no such
reference, the ﬁrm is placed in the non-hedging category (for all years). Firms for which
no 10-K or annual report is available in 1993 have the hedging variable set equal to
missing. While this is a coarse assignment procedure, it provides a practical way to get a
sense as to which ﬁrms are more heavily involved in hedging their foreign cash ﬂow risk.
In models (13) and (14) of Table 5 we report estimates for sub-samples grouped
into non-hedgers and hedgers respectively. The coefficient on foreign cash ﬂow is
positive and highly signiﬁcant for the non-hedgers, while it is positive, smaller, but still
signiﬁcant for the hedgers. The estimated coefﬁcients are different at the 5% level
(t=2.46) across the two groups. These ﬁndings suggest that the investment policies of
ﬁrms that hedge are relatively less impacted by exogenous variations in net cash ﬂow.
This is consistent with the idea that hedging strategies allow these ﬁrms to decrease the
impact Of ﬁnancial constraints by minimizing the likelihood of a mismatch between

investment needs and internal cash ﬂow.

1.6 Conclusion

Using a unique, hand-collected dataset on advertising expenditures by

multinational ﬁrms, we uncover a signiﬁcant relation between a ﬁrm's U.S. advertising

spending and their foreign cash ﬂow, even after controlling for US. investment

opportunities. This ﬁnding is robust to a variety of speciﬁcation choices. Since foreign

4o

cash ﬂow should have no marginal information content with regard to domestic
advertising investment opportunities, our evidence supports the hypothesis that there is a
causal link between a ﬁrm’s cash ﬂow and its investment decisions. We ﬁnd little
support for alternative explanations of our ﬁndings. Using some calibration exercises, it
appears that our estimates are consistent with a fairly large effect of cash ﬂow variation
on a ﬁrm’s combined investment activities (advertising plus capital investment), on the
order of $.30 of increased spending for each dollar of additional cash ﬂow.

When we group ﬁrms by characteristics that have previously been associated with
ﬁnancial constraints, in many cases we ﬁnd that ﬁrms that are categorized as relatively
more constrained do in fact display a relatively greater sensitivity of US. advertising
spending to foreign cash ﬂow. In particular, our evidence suggests that younger ﬁrms,
ﬁrms that pay lower dividends, ﬁrms with lower credit ratings, and ﬁrms with higher
leverage exhibit particularly large advertising-cash ﬂow sensitivities. We also ﬁnd that
fums that hedge have a relatively smaller sensitivity of US. advertising to foreign cash
ﬂow compared to non-hedgers. Consistent with the arguments of F root, Scharfstein, and
Stein (1993), this result suggests that hedging is often part of a general strategy to
minimize the impact Of ﬁnancial constraints on investment activities.

The evidence we present adds to the important ﬁndings of Rauh (2006) by
conﬁrming in a different context that cash ﬂow does have a causal effect on investment
in large samples, particularly for ﬁrms with certain characteristics. Our ﬁndings suggests
that the effect of variations in internal funds on investment is not restricted solely to ﬁxed
capital investment, but also extends to intangible investments such as advertising. Thus,

the overall impact of ﬁnancial constraints on investment spending by ﬁrms may be even

41

larger than previously believed. Our ﬁndings add to the body of evidence indicating that
ﬁrms have active internal capital markets in which they move funds across different
regions and activities. This suggests that choices regarding how ﬁrms organize their
internal capital markets and how they hedge risk across activities can have a real effect
on the ﬁrm's growth trajectory.

While our results support the general notion that ﬁrms are at times ﬁnancial
constrained and that they have active internal capital markets, they do leave unanswered
the question of what ﬁnancing frictions are generating the relation between cash ﬂow and
investment. This is a vital issue if we seek to identify ﬁnancial strategies that may
alleviate ﬁnancial constraints or to understand the welfare consequences of the presence

of constraints. Hopefully future research will provide insights on this important issue.

42

Table 1.1: Summary Statistics for Compustat Firms

The ﬁgures in this table are derived from the set of all Compustat ﬁrrn-years from 1984-
2002 with the exception of foreign ﬁrms, ﬁnancial ﬁrms, regulated utilities, ﬁrms with no
reported stock price, and ﬁrms with (inﬂation adjusted to 2002) book assets of under $10
million. All variables that are normalized by assets are divided by beginning of period
book assets. Missing values of advertising and R&D are assigned a value of O. Tobin’s
Q is deﬁned as beginning of period (book assets + market equity - book equity)/(book
assets). Cash ﬂow is deﬁned as net income plus depreciation (Compustat data items #14
+ #18). Ratios where the denominator is a 0 are coded as missing. Lagged sales grth
in year t is deﬁned as (salesH — salest-2)/salest-2. All variables are winsorized at the 1%
and 99% points. Panel A includes all observations and Panel B is restricted to
Observations where a ﬁrm reports positive advertising spending during the year.

 

 

 

 

 

Mean Median Observations
Panel A: All Compustat F irm
Advertising/Assets .0155 0 74811
R&D/ Assets .0456 0 7481 1
Capital Expenditures/Assets .0828 .0520 73428
Tobin’s Q 1.94 1.37 74645
Cash F low/Assets .041 .081 74343
Cash/Assets .173 .068 74759
Advertising/ R&D .532 0 37612
Advertising/ Capital Expenditures .455 0 82046
Lagged Sales Growth .245 .100 67566
Dummy Variable for Advertising > 0 .318 0 84730
Panel B: Compustat Firms with
Advertising > 0
Advertising/ Assets .0485 .0265 23981
R&D/ Assets .0430 0 23981
Capital Expenditures/ Assets .0823 .0532 23612
Tobin’s Q 1.97 1.38 23944
Cash Flow/ Assets .044 .082 23947
Cash/Assets .169 .072 23977
Advertising/ R&D 1.586 .355 12614
Advertising/ Capital Expenditures 1.410 .5520 26468
lagged Sales Growth .242 .101 21754

 

43

Table 1.2: Summary Statistics for Top 1000 US. Advertisers

The sample includes all Compustat listed ﬁrrn-years in which the ﬁrm is listed in the top
1000 advertisers in the annual issues of Ad $ Summary frOm 1984 to 2002 and the ﬁrm
reports non-zero foreign income, domestic income, and sufficient information to
calculate Tobin’s Q. Financial ﬁrms, non US. based ﬁrms, and regulated utilities are
excluded. All variables that are normalized by assets are divided by beginning of period
book assets. Missing values for Total Advertising (i.e., the Compustat advertising ﬁgure)
and R&D are assigned a value of 0. Tracked US. Advertising is the sum of advertising
in the six domestic media outlet categories that have been tracked by Ad $ Summary for
all years during the sample period. Tobin’s Q is deﬁned as beginning of period (book
assets + market equity - book equity)/(book assets). Cash ﬂow is deﬁned as net income
plus depreciation (Compustat data items #14 + #18). Foreign proﬁts are deﬁned as pre-
tax foreign income less foreign taxes paid (Compustat item #273- item #64). US. Proﬁts
are deﬁned as pre-tax domestic income less domestic taxes (Compustat item #272 — (item
#16 — item #64)). Ratios where the denominator is a 0 are coded as missing. The ratio
Tracked U.S. Advertising/Total Advertising is set equal to 1 if this ratio is greater than 1
or if Total Advertising is missing. All variables are winsorized at the 1% and 99%
points. The correlation statistics are simple correlations calculated over the entire
sample. The symbol A indicates that annual changes in the indicated variable are
calculated before calculating the indicated correlation.

 

Mean Median Number

 

 

 

of Obs.

Panel A: Spending

Total Advertising/Assets .062 .029 3028
Tracked U.S. Advertising/Assets .022 .008 3028
R&D/Assets .036 .016 3028
Capital Expenditures/Assets .072 .061 3028
Tobin’s Q 2.05 1.60 3028
Cash Flow/Assets .122 .1 19 3018
Total Advertising/R&D 2.55 .494 2052
Total Advertising/Capital Expenditures 1.32 .441 2986
Tracked U.S. Advertising/Total Advertising .503 .355 3028
Panel B: US. and Foreign Proﬁts

U.S. Proﬁts (USP) / Assets .047 .045 3028
Foreign Proﬁts (FP) / Assets .023 .015 3028
(U .S. Proﬁts) / (U.S. Proﬁts + Foreign Proﬁts) .705 .760 3028
(Foreign proﬁts) / (U.S. proﬁts + Foreign proﬁts) .295 .240 3028
(U SP/Assets) and (FP/ Assets) — correlation 0.161 3028
A(USP/Assets) and A(FP/Assets) — correlation 0.197 2473
A[(USP+FP)/Assets] and A(Firm Cash Flow/Assets) — 0.956 2520

correlation

 

44

Table 1.3: Regression Models of Investment, Total Advertising, and Firm Cash
Flow

The sample includes all active Compustat ﬁrm-years from 1984-2002 with the exception
of foreign ﬁrms, ﬁnancial ﬁrms, regulated utilities, and ﬁrm’s with inﬂation adjusted (to
2002) book assets of under $10 million. All models include ﬁrm-ﬁxed effects and year
effects (estimates not reported). Asymptotic standard errors are reported under the
coefficient estimates and are robust to heteroskedasticity and serial correlation. The
dependent variable in column 1 (columns 2-4) is a ﬁrrn’s annual capital expenditures
(total advertising expenditures) normalized by beginning of period book assets. Missing
values of advertising are set equal to 0. Cashﬂow/Assets is deﬁned as
(contemporaneous) net income plus depreciation normalized by beginning of period
assets. In columns 2-4 we also add ﬁrm advertising back to income in our calculation of
cash ﬂow. Tobin’s Q is deﬁned as (book assets + market equity — book equity)/(book
assets) and is measured as of the start of the observation year (i.e., end of previous ﬁscal
year). All variables are winsorized at the 1% and 99% points. The estimates in columns
1 and 2 are derived over the set of all Compustat Observations that satisfy the sampling
criteria. In column 3 we restrict the sample to ﬁrms that have had positive advertising in
the current year or in some year prior to the observation year. In column 4 we restrict
attention to Observations where the ﬁrm is listed in Ad $ Summary’s list of top 1000 US.
advertisers. For cases where a ﬁrm drops off this list, we also include the year following
the exit from the list.

*Signiﬁcant at the 10% level, "Signiﬁcant at the 5% level, *"Signiﬁcant at the 1%
level.

 

 

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

Dependent Variable Cap. Ex/A Advert/A Advert/A Advert/A
Cash Flow/Assets 0.0690*** 0.0185*** 0.0353*** 0.1711***

(0.0043) (0.0017) (0.0036) (0.0237)
Tobin’s Q 0.0113*** 0.0011**"‘ 0.0022*** -0.0006

(0.0004) (0.0001) (0.0003) (0.0015)
Which Observations All All Advert.>0 Top 1000
Number of 73068 74190 33662 6569
observations
R2 0.5620 0.8082 0.8145 0.8861

 

45

Table 1.4: Regression Models of US. Advertising and Foreign Cash Flow

The dependent variable in all columns is a ﬁrm’s annual tracked U.S. advertising in six
media outlets tracked by Ad $ Summary normalized by beginning of period assets. The
sample includes all ﬁrm-years listed in Ad $ Summary’s list of top 1000 advertisers for
US. based ﬁrms listed on Compustat with the exception of regulated utilities and
ﬁnancial ﬁrms. If a ﬁrm drops out of the list we collect and use advertising information
for the ﬁrst year that they exit the list. Asymptotic standard errors are reported under the
coefficient estimates. U.S. cash ﬂow is deﬁned as domestic pre-tax income less domestic
taxes plus tracked U.S. advertising. Foreign cash ﬂow is foreign pre-tax income less
foreign taxes. Both cash ﬂow measures are normalized by beginning of period book
assets and are measured over the same year as the advertising activity. Q is deﬁned as
(book assets + market equity — book equity)/(book assets) as of the start of the ﬁscal year.
The models in columns 1, 3, and 4 are estimated in ﬁrst-differences calculated only over
consecutive annual Observations with heteroskedasticity robust standard errors. All
variables in these columns are winsorized at the 1% and 99% points after taking
differences. The models in columns 2, 5, and 6 are estimated using ﬁrm ﬁxed effects
with an assumed AR(1) error structure and the Baltagi and Wu (1999) estimator. All
variables in these columns are winsorized at the 1% and 99% points. The estimates in
column 3 (column 4) are estimated over the sub-sample of Observations where the ﬁrm’s
tracked advertising spending in the lagged year used to calculate the difference was
greater than (less than) the sample median. The estimates in column 5 (column 6) are
estimated over the sub-sample of observations where the ﬁrm’s advertising spending in
the year prior to the observation year was greater than (less than) the sample median. In
cases where advertising was missing in the prior year, we assign an Observation to
column 5 or 6 based on the current year’s advertising spending.

*, **, "*Signiﬁcant at the 10%, 5%, and 1% levels respectively

 

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

US. Cash 0055‘" 0.076‘" 0085"” 0.005‘" 0127‘" 0.006‘"
Flow/Assets (0.007) (0.004) (0.01 1) (0.002) (0.008) (0.001)
Foreign Cash 0.033” 0.024" 0.052" 0.008‘ 0.046" 0.009‘"
Flow/Assets (0.014) (0.012) (0.023) (0.004) (0.023) (0.003)
Q 0001"" 0.001" 0.002‘" 0.0001 0.001 0.0002"
(0.0004) (0.0003) (0.001) (0.0001) (0.001) (0.0001)
Which Observations All All Big Ad. Small Big Ad. Small Ad.
Ad.
Estimation Procedure FD FE FD FD FE FE
Number of 2473 2623 1232 1241 1267 1276
observations
R2 0.127 0.157 0.201 0.042 0.259 0.064

 

46

Table 1.5: Regression Models of U.S. Advertising for Sub-samples Grouped by
Firm Characteristics

The dependent variable in all columns is a ﬁrm’s annual tracked U.S. advertising
normalized by beginning of period assets. All explanatory variables are deﬁned as in
Table 4. Models are estimated in ﬁrst-differences with asymptotic standard errors
reported under the coefficient estimates. All variables are winsorized at the 1% and 99%
level after taking ﬁrst differences. Each model is estimated over a sub-sample based on
the sorting criterion indicated in the column. These criteria group ﬁrms by the expected
level of ﬁnancial constraints. Each column denotes the expected constraint level relative
to the counterpart sub-sample for the indicated sorting criterion. Age is number of years
since the ﬁrm’s ﬁrst Compustat record with a non-missing stock price. Size is inﬂation
adjusted book value of assets. Divs is annual dividends over net income. Leverage is
deﬁned as short-terrn debt plus long-term debt divided by book assets. Cash holdings are
cash plus marketable securities divided by book assets. Credit rating is the ﬁrm’s S&P
senior debt rating. For the age, size, dividends, leverage, and cash holdings sorting
criteria, we select sub-samples based on whether the indicated ﬁrm characteristic is in the
top or bottom tercile based on start of year or prior year characteristics where younger
ﬁrms, smaller ﬁrms, lower dividend ﬁrms, higher leverage ﬁrms, and lower cash
holdings ﬁrms are assigned to the “High” constraint category. Firms with credit ratings
at or below (above) BBB+ are assigned to the high (low) constraints group. Firms that
indicate (do not indicate) that they undertake currency swaps, future, or options in 1993
are labeled hedgers (non-hedgers). *, **, ***Signiﬁcant at the 10%, 5%, and 1% levels
respectively.

47

Table 1.5

 

 

(I) (2) (3) (4) (5) (6) (7) (8)
Constraint High Low High Low High Low High Low
Level
U.S. Cash 0074‘” 0.016” 0.083'” 0.010 0.059'” 0.061‘” 0.019” 0.024‘”
Flow/Assets (0.010) (0.017) (0.011) (0.012) (0.010) (0.001) (0.009) (0.006)
FOR Cash 0.059” -0.0002 0.057‘ 0.023‘” 0.052” -0004 0.079‘” 0.014
Flow/Assets (0.024) (0.013) (0.029) (0.008) (0.023) (0.018) (0.021) (0.010)
Tobin’sQ 0.002“ 0.001 0.002“ 0.001 0.002" 0.001 0.003” 0.0003
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.0003)
Sorting Age Age Size Size Divs. Divs. Cred. Cred.
Criterion Rat. Rat.
Obs. 869 812 825 824 823 822 609 I 181
R2 0.177 0.061 0.193 0.063 0.161 0.131 0.150 0.078
(9) (10) (ll) (12) (13) (14)
Constraint High Low High Low High Low
Level
U.S. Cash 0061*" 0053*" 0030*" 0079*" 0082*" 0056*"
Flow/Assets (0.013) (0.017) (0.009) (0.01 1) (0.016) (0.01 1)
FOR Cash 0085*" -0.0003 0.015 0.033 0.187*** 0.039"
Flow/Assets (0.029) (0.022) (0.019) (0.021) (0.058) (0.016)
Tobin’sQ 0004*" 0001*" 0.001 0.001" 0.0003 0001*
(0.002) (0.001) (0.001) (0.005) (0.001) (0.001)
Sorting Leverage Leverage Cash/ Cash/ Non- Hedgers
Criterion Assets Assets hedgers
Obs. 821 821 825 824 387 1099
R2 0.206 0.119 0.083 0.198 0.306 0.134

48

ESSAY 2. WHAT HAPPENS IN ACQUISITIONS? EVIDENCE FROM BRAND
OWNERSHIP CHANGES AND ADVERTISING

2.1 Introduction

What happens when a corporate asset changes hands? A large literature on mergers,
acquisitions, and asset sales suggests that ownership changes often lead to substantial
shifts in the way that assets are managed. These shifts have the potential to alter the
efficiency and value of the affected assets. Given this potential, understanding how new
owners manage acquired assets is clearly an issue of much importance.

While prior evidence is consistent with the hypothesis that assets are often
managed differently following changes in control, with a few notable exceptions
discussed below, much of the evidence is indirect. A major impediment to more direct
investigation arises from the fact that it is difficult to track the behavior of individual
assets that are merged into a larger organization. Moreover, accounting adjustments
associated with changes in control can cloud accounting comparisons across different
ownership regimes.

In this paper we provide direct evidence on the consequences of control changes
by studying a unique database of advertising spending at the brand level. Viewing
advertising as a form of investment in a brand, we are able to examine whether new
owners follow different investment policies than prior owners, holding constant the
underlying asset -- the brand. In addition, we are able to assess whether the cross-

sectional variation in post-acquisition advertising behavior supports different hypotheses

49

regarding corporate control activity. The database we examine includes non-public
information collected directly from media outlets by a marketing-intelligence data
vendor. lmportantly, this data is available not only for domestic public ﬁrms, but also for
private ﬁrms and foreign ﬁrms. This rich feature of our data has the important advantage
that it allows us to track advertising investment behavior as brands move across different
owner types (public, private, foreign).

In our sample of 555 brands that experience an ownership change between 1998
and 2003, we ﬁnd strong evidence that new owners often do sharply shift a brand's
advertising strategy. In particular, new owners have a disproportionate propensity to
either sharply increase or decrease advertising in a purchased brand relative to a set of
size-industry matches. This ﬁnding of large changes in both directions may explain why
prior investigations of average changes in post-acquisition investment have yielded
limited results (e.g., Healey, Palepu, and Ruback (1992)). The fact that new owners often
shift a brand's advertising strategy indicates that the identity of an asset's owner can play
an important role in the management of the asset. While we observe both sharp increases
and decreases in spending after acquisitions, the tendency to cut sharply is substantially
more pronounced in our sample. Thus, while acquisitions may occur for a host of
reasons, our evidence suggests that on average they exert downward pressure on
investment spending, at least in the form of advertising.

When we examine differences across buyer and target/seller types, we ﬁnd
evidence that increased (decreased) private ownership of assets is associated with a
downward (an upward) shift in advertising compared to other deals. A similar but

statistically weaker result holds for foreign ownership. When we restrict attention to

50

deals between public ﬁrms, we are unable to detect any signiﬁcant cross-sectional
relation between ﬁrm characteristics and advertising behavior. Our evidence on private
owners suggests that the investment cuts reported by Kaplan (1989) and Muscarella and
Vetsuypens (1990) for certain specialized samples Of LBO ﬁrms are part of a much more
general phenomenon. The tight purse strings of private owners indicates that these
owners either specialize in curbing unproﬁtable spending or, alternatively, are forced to
cut spending in response to heightened ﬁnancial constraints. Our evidence tends to
support the former explanation over the latter, as buyers with ties to sophisticated
investment ﬁrms (i.e., private equity funds, LBO ﬁrms, ﬁnancial ﬁrms) have a slightly
higher tendency to cut post-acquisition advertising than other private buyers (e.g., closely
held family ﬁrms).

The abnormal propensity of buyers to cut advertising in purchased brands
suggests that acquisitions may present opportunities for buyers to realize cost savings by
economizing on their advertising investment. Alternatively, buyers may simply shift
advertising from the purchased brands towards their other brands. We investigate these
possibilities by studying the advertising behavior of the buyer's other brands around the
time of the acquisition. Interestingly, we ﬁnd that buyers tend to cut advertising in the
brands that closely overlap with the purchased brands, but do not signiﬁcantly change
advertising in their non-overlapping brands. The observed advertising cuts in
overlapping brands suggest that acquisitions present opportunities to realize cost savings
in overlapping activities, a notion with some prior support in the literature (e.g., Healey,

Palepu, and Ruback (1992), Houston and Ryngaert (1994)).

51

The efﬁciency of any cost cutting activity depends on whether the anticipated
beneﬁts associated with the foregone spending are less than the savings realized. If they
are, the expectation Of cuts in spending will be associated with deal value creation. In our
sample, we ﬁnd that the combined acquisition announcement returns of the buyers and
targets/sellers are in fact positively related to some measures of post-acquisition
downward shifts in advertising spending. Since the cuts we observe are likely to be an
indicator of more general cost cutting efforts, this ﬁnding suggests that efﬁcient costs
savings are expected and realized in some acquisitions. This evidence on announcement
returns and actual post-acquisition cost cutting behavior complements the recent ﬁndings
of Houston, James, and Ryngaert (2001) and Bemile (2004) who report that the market's
reaction to merger armouncements is closely related to anticipated cost savings.

To further investigate the consequences of post-acquisition changes in advertising
behavior, we study the market share dynamics of our sample brands. If the beneﬁts to
marginal advertising spending are high, post-acquisition downward shifts in advertising
may be associated with losses in brand market share. Interestingly, we ﬁnd that
purchased brands do not experience signiﬁcant losses in market share around the time of
the acquisition, even in cases when the brand experiences a downward shift in
advertising. In addition, the buyers' overlapping brands appear not to experience
abnormal changes in market share subsequent to the acquisition. This evidence is
consistent with the notion that any lost beneﬁts from post-acquisition cuts in advertising
are small, suggesting that the downward revisions in spending we observe are, in fact,

efﬁcient.

52

Taken as a whole, our evidence supports the hypothesis that corporate control
activity has a real effect on how assets are managed. Private entities appear to play a
special role in this activity. Our ﬁndings indicate that purchasers of assets often adopt
strategies to exploit cost savings opportunities brought about by an acquisition, and on
balance the evidence supports the hypothesis that these cost savings strategies are
efﬁciency enhancing. Thus, a generally healthy picture of efficient revisions in spending
strategies following an acquisition emerges. Our ﬁndings complement prior evidence on
changes in plant efﬁciency following ownership changes (e. g., Lichtenber and Siegel
(1990), McGuckin and Nguyen (1995), Maksimovic and Phillips (2001), Schoar (2002)).
These prior studies share a similar feature to ours in that they also track the behavior of
an individual asset across different owners. The efﬁciency changes identiﬁed by these
authors can be thought of as the net effect of various operational changes brought about
by an asset's new owner. Our data lends itself more naturally to identifying some speciﬁc
operational changes implemented by new owners.

The rest of the paper is organized as follows. In section 2.2 we review the
relevant literature and outline our empirical strategy. In section 2.3 we detail our sample
construction and present summary statistics on sample transactions. In section 2.4 we
present our results on advertising and market share changes associated with brand

acquisitions. Section 2.5 concludes.

2.2 Motivation and empirical strategy

2.2.1 Value creation and the market for corporate assets

53

Corporate assets can fall under new ownership in either ﬁrll-ﬁrm or partial-ﬁrm
acquisitions. Maksimovic and Phillips (2001) show that both of these paths towards new
ownership are common. In the case of acquisitions of entire ﬁrms, a large number of
announcement return studies show that targets tend to experience large increases in value
while bidder returns are small and often centered around zero.26 The combined return of
targets and bidders, an indicator of total value creation or synergies from a deal, are on
average positive (e.g., Bradley, Desai, and Kim (1988)). This is an important fact, as it
suggests that mergers often result in a more efﬁcient use of the target ﬁrm's assets.27 In
the case of partial acquisitions, there is ample evidence that asset sellers and buyers both
experience positive average announcement returns (e.g., Alexander, Benson, and
Karnpmeyer (1984), Jain (1985), and Hite, Owers, and Rogers (1987)). Thus, similar to
the evidence on full-ﬁrm acquisitions, this evidence is consistent with the idea that
ownership changes tend to reallocate assets to more efficient uses.28

Since announcement return evidence often admits multiple interpretations, several
studies also consider operating performance information. Some studies ﬁnd that
accounting performance improves on some dimensions after a merger (e. g. Healy,

Palepu, and Ruback (1992), and Andrade, Mitchell, and Stafford (2001)), but the ﬁndings

are far from conclusive. The picture is further clouded by the fact that mergers and

 

2" See the surveys of Jensen and Ruback (1983), Jarrell, Brickley, and Netter (1988) and Andrade, Mitchell
and Stafford (2001).

27 There are other interpretations of the event study evidence based solely on information revelation arising
from transaction announcements. Recent evidence by Moeller, Schlingemann, and Stulz (2005) indicates
that some very large deals in the 1998-2001 period exhibited highly negative combined bidder plus target
announcement returns.

23 In the case of asset sales, the seller's announcement return may reﬂect the market's assessment of what

the selling ﬁrm will do with the sale proceeds rather than pure efficiency gains on the transferred assets.
For a discussion and evidence, see Lang, Poulsen, and Stulz (1995) and Bates (2005).

54

acquisitions typically result in large special adjustments to a ﬁrm's accounting statements,
making it difficult to make intertemporal comparisons of operating performance. Given
these issues and the generally mixed evidence, the post-acquisition operating
performance literature can be viewed as being only suggestive of efﬁciency
improvements following acquisitions.

A richer picture of efficiency consequences following ownership changes
emerges from studies that use micro-level census data on individual manufacturing
plants. McGuckin and Nguyen (1995) and Maksimovic and Phillips (2001) report that
acquired plants often experience increases in productivity under new ownership.
However, Schoar (2002) reports that the purchasers of these plants often display a post-
acquisition decline in the productivity of their existing plants. Consequently, in some
types of acquisitions, notably diversifying acquisitions, the net effect of the ownership
change on overall asset efficiency may in fact be negative on average. These studies
provide important evidence suggesting that (a) new owners often operate purchased
assets differently than former owners, and (b) the presence of newly acquired assets may
affect the operation of a ﬁrm's existing assets. We hope to shed further light on these

issues in our study.

2. 2.2 Sources of value changes in acquisitions

One goal in the literature on acquisitions is to determine whether ownership
changes create value on average. A second goal is to understand the cross-sectional
variation in deal success in light of the many possible costs, beneﬁts, and motivations

underlying observed transactions. For example, some assets may ﬁt particularly well

55

together, in which case the value/efﬁciency gains from merging them may be particularly
large. At the same time, much has been written about the possibility of managerial
incentive problems leading to value destruction in some corporate control activity (e. g.,
Merck, Shleifer, and Vishny (1990) and Lang, Stulz, and walkling (1989, 1991)).

To investigate these issues, a large literature has emerged exploring both the
determinants of acquisition performance and the likelihood of a transaction occurring.
Much of this literature can be viewed as attempts to understand the causes and
consequences of acquisition activity. Factors that have been emphasized as potential
determinants of deals occurring and/or succeeding include the technological ﬁt of an
asset with other assets, agency problems at the divisional and CEO levels, ﬁnancial
constraints, and behavioral factors. The scope of the resulting literature is enormous (see
the survey by Andrade, Mitchell, and Stafford (2001)). While a few consistent ﬁndings
have been reported, no simple picture emerges regarding the determinants of acquisition
success. This likely is a reﬂection of both the richness of corporate control behavior
along with the indirect nature of many of these analyses.

For the purposes of our study, a couple of reported ﬁndings are particularly
relevant. It is frequently suggested that acquisition synergy gains may be larger when
there is substantial asset overlap between the buyer's assets and the purchased assets (e.g.,
Healey, Palepu, and Ruback (1992)). Using detailed data on the locations of bank
branches, Houston and Ryngaert (1994) show that combined bidder plus target
announcement returns are closely related to the degree to which merging banks'

operations overlap. This ﬁnding suggests that cost cutting opportunities arising from

56

overlapping activities are an important part of acquisition synergy gains. We will explore
this possibility in our analysis by examining advertising behavior and brand overlap.

A second relevant ﬁnding is the evidence reported by Houston, James, and
Ryngaert (2001) and replicated on a larger and more general sample by Bernile (2004).
These authors ﬁnd that combined target plus bidder announcement returns are closely
related to management's estimates Of projected costs savings. This evidence suggests
again that cost savings are an important source of the value gain associated with many
ownership changes. We can complement this evidence by using information on post-
acquisition spending changes to assess whether actual, rather than projected, spending

cuts are associated with the market's assessment of deal value creation.

2. 2. 3 Investment policy and changes in control

One important way in which new owners may manage assets differently from
prior owners is in the investment policy they adopt for the acquired assets. Any such
changes in investment policy have the potential to substantially change asset values.
Some theories of corporate control activity suggest that acquirers may tend to cut
investment in acquired assets. For example, we would expect this to occur if acquisitions
are frequently motivated by a desire to curb the overinvesting behavior of empire
building managers (e. g., Jensen (1986)).29 Additionally, investment cuts may be
observed if acquisitions allow ﬁrms to cut duplicate investment in overlapping activities

(e.g., Houston and Ryngaert (1994)).

 

29 Interestingly, Servaes (1994) fails to ﬁnd evidence supporting the hypothesis that takeovers targets are
systematic overinvestors.

57

On the ﬂip side, in some circumstances new owners may choose to increase
investment following an acquisition. This could occur if the old owner of an asset was
ﬁnancially constrained and thus unable to optimally fund the asset's growth.
Additionally, increases in investment may arise if the combination of new and old assets
generates valuable growth opportunities to exploit or if the acquisition is part of a general
growth strategy of the acquirer.

This discussion suggests that there may be high variability in investment policy
changes following acquisitions with an abnormal propensity for both sharp increases and
decreases in investment spending. Consequently, much of the interesting variation in the
data may effectively cancel out in tests that focus solely on average changes in
investment following acquisitions. This may explain why prior research on capital
expenditures following mergers fails to ﬁnd evidence of a signiﬁcant average change in
investment (e.g., Healey, Palepu, and Ruback (1992)). Similarly, researchers have
uncovered little evidence of signiﬁcant average changes in R&D spending following
mergers (e.g., Hall (1988)).30 A recent study by Pesendorfer (2003) examines mergers in
the paper industry and ﬁnds some evidence consistent with abnormal variation in post-
merger investment policies. In particular, he ﬁnds that merged entities are signiﬁcantly
more likely to expand or contract plant capacity than their non-merged counterparts. In
our study we can assess whether this excess variability in post-acquisition spending is a

more general phenomenon.

 

3° Interpreting these prior results is clouded by the difficulty in measuring investment intensity for an asset
base that is often changing because of post-acquisition restructuring and special merger accounting
adjustments. Bhagat, Shleifer, and Vishny (1990) ﬁnd that hostile takeover targets do not display an
abnormal propensity to make large investment cuts. However, overtly hostile deals are a very rare
occurrence during our more recent sample period.

58

In the special case of leveraged buyouts (LBOs), prior authors present evidence of
substantial capital spending cuts (e.g., Kaplan (1989), Smith (1990), Muscarella and
Vetsuypens (1990)). These cuts do not appear to extend to R&D investment (e.g.,
Lichtenberg and Siegel (1990), Smith (1990)), perhaps because LBO ﬁrms typically
operate in low R&D intensity sectors (e.g., Hall (1990)). The ﬁnding that LBOs are
associated with signiﬁcant changes in capital spending suggests that certain types of
ownership structures, for example private ownership, may be associated with substantial
revisions in investment spending. However, since much of the LBO evidence is based on
special samples Of private ﬁrms that eventually re-enter the public equity markets, the

generality of the LBO ﬁndings is an open question.

2. 2.4 Owner type and acquisition activity

Prior event study evidence on acquisitions suggests that there may be something
special about transactions that change the private or public ownership status of an asset.
In particular, Chang (1998) and Fuller, Netter, and Stegemoller (2002) show that bidder
returns are higher when a public ﬁrm acquires a private target rather than a public target.
This suggests that the synergy gains associated with these deals is different from deals
between public ﬁrms, or, alternatively, that the division of the synergy gains is different.
With regard to shifts from public to private ownership, DeAngelo, DeAngelO, and Rice
(1984), and Lehn and Poulsen (1989) show that these transactions are often associated
with particularly large event returns, while Kaplan (1989) shows that LBO deals are often

followed by improved operating performance.

59

While these ﬁndings are suggestive, direct comparisons of how private and public
owners differ in their management of an acquired asset are difﬁcult to make given the
paucity of available data on the management of assets under private ownership. Even
most studies that use information on plant productivity are unable to examine differences
between public and private ownership. Since we can track a certain type of investment
spending, advertising, across different types of owners, our sample has the potential to
further our understanding of differences between public and private ownership.

With regard to investment spending, there are reasons to suspect that there may be
substantial differences between public and private ﬁrms. In particular, since private ﬁrms
are closely held, decision makers at these ﬁrms may have a relatively stronger incentive
to make value maximizing decisions. Thus, if managers have a general tendency to
overspend, private ownership may be associated with lower investment spending as
overinvestrnent at these ﬁrms is more successﬁilly curbed. At the same time, some
private ﬁrms may have less access to capital than public ﬁrms given their inability to tap
public capital markets. Consequently, enhanced ﬁnancial constraints may increase the
propensity of these ﬁrms to underfund proﬁtable investment opportunities. Both of these
possibilities suggest that private owners may spend less freely than their public
counterparts. We will explore for this possibility in our sample and attempt to distinguish
between the two scenarios.

Transactions involving foreign ﬁrms may also be special in certain ways. While
many cross-border deals likely have similar motivations to domestic deals, several
authors have noted that there can be special factors involved in these transactions. In

particular, cross-border deals may be motivated by investor portfolio diversiﬁcation

60

concerns (Caves (1971)), exchange rate factors (F root and Stein (1991)), tax shifting
opportunities (Scholes and Wolfson (1990)), and corporate governance considerations
(Rossi and Volpin (2004)). These special issues, along with factors such as culture or
familiarity, could conceivably lead to differences in the investment policies Of foreign
owners compared to domestic owners. Given the potential differences, in some of our
modeling below we control for whether the buyer or target/seller is a foreign ﬁrm. This
may allow us to provide additional evidence on the nature of cross-border deals and the

investment policies of foreign ﬁrms.3 1

2. 2.5 Advertising as an investment

The idea that advertising can be viewed as an investment has a long history in the
ﬁelds of economics and marketing (see the survey by Bagwell (2005)). In particular,
following Schmalensee (1976), it is reasonable to expect that proﬁt maximizing ﬁrms
will invest in advertising to the point where the marginal cost of this spending are equal
to the marginal beneﬁts. These beneﬁts may include increased future or sales or higher
output prices. Several ﬁnance authors treat advertising as a component of a ﬁrrn's
investment decision in parts of their analyses (e. g., Smith (1990), Holthausen and Larcker
(1996), and Andrade and Stafford (2004)).

In a recent study, Fee, Hadlock, and Pierce (2006) examine advertising
investment for a large sample of publicly traded ﬁrms. They report that advertising

spending is a signiﬁcant component of total investment for ﬁrms that report non-trivial

 

3 ' Event study evidence on foreign transactions is mixed. Harris and Ravenscraft (1991) ﬁnd that target
returns are higher in cross-border deals, but Dewenter(1995) ﬁnds no difference. McCabe and Yook
(1996) ﬁnd higher bidder returns in cross-border transactions, but Moeller and Schligemann (2005) report
the opposite result.

61

advertising spending. In addition, they uncover some patterns regarding the determinants
of advertising spending that suggest important similarities with other types of investment
decisions (e.g., capital expenditures, R&D). The evidence of Fee, Hadlock, and Pierce
(2006) suggests that advertising spending is an important decision to look at in its own
right, at least for the types of advertising-intensive brands that are represented in our
sample. In addition, their evidence indicates that advertising decisions may provide us

with more general insights on investment spending around ownership changes.

2. 2. 6 Empirical Strategy

Our strategy is to use the availability of information at the brand level to track
advertising investment spending as assets experience changes in ownership. An
advantage of this approach arises from the fact that brand advertising is a speciﬁc
investment in the brand. Thus, we can study changes in investment intensity holding the
relevant asset base -- the brand -- fairly ﬁxed. This distinguishes our study from
investigations using accounting data, as the asset base in these prior studies is often
changing as a result of post-acquisition restructuring actions and accounting adjustments.

In our analysis we benchmark advertising decisions of acquired brands against the
advertising behavior of similar brands with no ownership change. This allows us to
assess whether ownership changes are associated with abnormal revisions, upwards or
downwards, in investment spending behavior. Not only can we assess how acquired

brands compare to non-acquired brands, we can also make comparisons of advertising

62

behavior across different types of deals. Thus, we can directly compare deals with
private acquirers to deals with public acquirers. We can also explore whether other
potentially relevant ﬁrm or deal characteristics, for example measures of ﬁnancial
constraints or agency conﬂicts, are related to post-acquisition investment spending
behavior. Our ability to directly compare across different types of deals is a
distinguishing feature of our investigation.

An additional component to our empirical strategy is to explore inter-ﬁrm
linkages in investment decisions. In particular, we can examine how advertising
investment in an acquired brand interacts with an acquirer's investment in their other
brands. This can provide rich information on how acquired assets are digested and
integrated into larger organizations, an issue for which little is known. Finally, by
relating advertising behavior to measures of performance such as announcement returns
and changes in market share, our study can provide information relevant for

understanding the overall efﬁciency of the acquisition process
2.3 Data and Sample Selection

2. 3. 1 Identifying the sample

We identify brands and collect advertising information from publications and
databases issued by Competitive Media Reporting/Taylor Nelson Sofres Company
(CMR/TNS) and its predecessors. CMR/TNS gathers advertising data directly from

major U.S. radio and television stations, newspapers, magazines, billboard companies,

63

etc. This data is relied on heavily by ﬁrms in their marketing intelligence and is used in
many of the empirical studies of advertising surveyed by Bagwell (2005).

Information on brand advertising spending and ownership is reported in annual
releases of the Company Brand $ database issued by CMR/TNS. We choose as our
sample period the most recent 5-year period available from this source (1998-2003). The
database reports information at a very micro level. For example, "Trident Sugarless
Gum" is identiﬁed separately from "Trident Sugarless Mints.” We refer to these
individual listings as microbrands. As we explain shortly, in our subsequent analysis we
aggregate these microbrands together to create a more reasonable deﬁnition of a brand
(i.e., "Trident" in this example).

If a microbrand's reported owner in 2003 differs from its 1998 owner, we conduct
news searches in the Factiva database to investigate whether there was a substantive
change in ownership.3'2 If we can conﬁrm that there was a change in the entity
controlling the brand, we collect details on the ownership change. Control is deﬁned to
be any ownership stake exceeding 50%. We supplement news article information with
any deal details reported in the SDC mergers and acquisitions database. In some cases a
brand changes hands more than once during the sample period, in which case each
ownership change is treated as a separate Observation.

For each microbrand with a conﬁrmed ownership change, we use visual
inspection to group together for each year all other microbrands listed in the CMR/TNS

data ﬁles that share a common name with the microbrand. Annual information on these

 

32 Spellings and abbreviations of microbrands sometimes change slightly with each annual data release.
Our electronic screening for potential ownership changes can only detect cases in which at least one
microbrand with no change in spelling has a change in its listed owner.

64

microbrands is derived from the Ad $ Spender database, as this data is available in an
electronic format on an annual basis.33 The identiﬁed set of microbrands is what we refer
to as a brand in what follows. In all but a handful of cases, all Of a brand's component
microbrands were transferred in the ownership change event. In the few cases in which
only a subset of a brand's microbrands were transferred, the subset is treated as the
relevant brand for further study.

The preceding procedures yield a sample of 555 distinct brand ownership
changes. For each of these changes, we record the identity of the buyer and target/seller
of the brand and match these owners to the Compustat and CRSP ﬁles when possible. If
we determine that the owner of a brand is a subsidiary of another entity, we attribute
ownership to the ultimate parent. If an identiﬁed owner is not a CRSP or Compustat
listed ﬁrm, we conduct news searches to identify whether the owner is a private or
foreign ﬁrm.

In table 2.1 we report some initial summary statistics describing the sample.
Consistent with economy-wide patterns in acquisition activity, the number of brand
ownership changes falls sharply between 2000 and 2001. Approximately 2/3 of the
ownership changes entail a full acquisition of one ﬁrm by another (377 of 555), with the
remaining 1/3 representing partial acquisitions of the seller's assets, typically the
purchase of a division. Since the motivations for full and partial acquisitions may differ,

in our later analysis we explore for possible differences. In what follows we will use the

 

33 To identify ownership changes, we hand coded data from the Company Brand 8 volumes for the years
1998 and 2003. The Ad $ Spender CDs are not useful for identifying brand owners, as the CDs start in
2001 and each CD only lists a brand's current owner along with S-years of advertising information. The
advertising information in the two data sources matches exactly as they are issued by the same firm.

65

word seller to interchangeably refer to the seller of an asset in a partial acquisition and the
entire target ﬁrm in the case of a full acquisition.

As we report in the table, the sample is composed of 348 distinct control change
events. These events lead to 555 brand ownership changes, as 84 events are cases in
which multiple brands are purchased in a single acquisition. Since we view the brand as
the unit of interest, most of our analysis below is conducted at the brand level. However,
since there could be some within-deal correlation when multiple brands are acquired, we
do explore the robustness of some of our ﬁndings to the treatment of each brand as an
independent observation.

For each brand, we calculate total tracked advertising as the sum of all of the
calendar-year advertising of the component microbrands in the ten media categories
tracked by CMR/TN S. Focusing on the year prior to the deal announcement, we report in
table 1 that the mean level of tracked advertising is $11.46 million. This ﬁgure suggests
that the brands we study are advertised fairly intensely. Normalizing a brand's tracked
advertising by the owner's total reported advertising in Compustat, the sample mean is
21.33%. The fact that this ﬁgure is substantially below 100% is not surprising, as the
CMR/TNS numbers pertain to advertising in an important but limited set of major
domestic media outlets while Compustat ﬁgures pertain to all advertising. In addition,
the CMR/TN S data pertains to a speciﬁc brand while Compustat total advertising ﬁgure
pertains to all of a ﬁrm's brands. As long as the CMR/TNS data are a reliable indicator of
an owner's promotion of a given brand, a very reasonable assumption in our view, the

data should be informative for the purposes of our study.

66

The ﬁgures in the bottom rows of table 2.1 indicate that there is substantial
heterogeneity in the types of buyers and sellers represented in the sample.
Approximately half of the buyers and sellers are publicly traded U.S. ﬁrms. The
remaining buyers and sellers are either private domestic ﬁrms or foreign ﬁrms (both
public and private). Within the private domestic category, there are a fair number of
traditional private Operating ﬁrms, but also a substantial number of specialized
investment companies (e.g., private buyout ﬁrms such as KKR). It is possible that
private operating companies and private investment companies have somewhat different
motivations for engaging in assets sales or purchases, a possibility we examine in some
of our analysis below. While we have limited information on the foreign ﬁrms in our

sample, it appears that most of these ﬁrms are large multinational entities.

2. 3.2 Characteristics of buyers, sellers, and transactions

Since ﬁrm characteristics may affect acquisitions, we present in table 2.2
summary statistics on the buyers and sellers in our sample. Most of these statistics are
based on Compustat data, so the ﬁgures are typically missing for private and foreign
ﬁrms. In panel A (panel B) we present statistics on sellers (buyers) as a group, while in
panel C we present statistics on the characteristics of sellers relative to buyers. Given the
typical skewness of ratio measures, in panel C we report medians only.

The ﬁgures in the table indicate, not surprisingly, that buyers are generally larger
than sellers. However, sample acquisitions are certainly substantial from the buyer's
perspective, as the amount paid in the median transaction is equal to 40% of the buyer's

book assets. Sellers tend to have higher leverage than buyers, suggesting that alleviation

67

of ﬁnancial constraints may motivate some ownership changes. Sellers also tend to have
lower levels of Tobin's Q and sales growth than buyers. This is consistent with the idea
that buyers have better managerial skills or growth prospects than sellers, enabling them
to more effectively manage an acquired brand. Finally, sellers and buyer's are in the
same 2-digit industry in 72.79% of all cases, indicating that the majority of the deals we
study are somewhat horizontal in nature.

In interpreting our later evidence, it is useful to gauge the market's overall
reaction sample deals. As we report in table 2.3, the market reaction to the control
changes in our sample is Similar what others report for full ﬁrm and partial ﬁrm
acquisitions. Using a 3-day event window, the ﬁgures in the table reveal that the typical
market reaction for sellers is positive and highly signiﬁcant while for buyers it is negative
and insigniﬁcant. Using the combined return of the buyer and the seller as a measure of
synergy gains, the mean and median gains are both positive (2.13% and 1.51%
respectively) and highly signiﬁcant. This indicates that sample deals on average create
shareholder value. Our subsequent analysis may shed some light on the sources of this
value creation.

In panel B of table 2.3 we report method of payment information. Since
approximately one third of these transactions entail purchases of only part of the seller's
assets, our sample has a fairly higher proportion of pure cash deals (51.72%). Given the
well-known importance of payment method in announcement returns, we include it as a
control variable in some of what follows.

In panel C of table 2.3 we report information on stated reasons behind sample

deals. These reasons are not mutually exclusive and are based on our reading of the

68

longest news article in F activa released around the transaction announcement. Deals with
no indication of an economic motivation are treated as having a missing stated reason.

As we report, reported reasons for sample deals are fairly typical of what one would
expect. The most common reasons are that the assets are a good/better ﬁt with the buyer
(43%) and that the buyer can better ﬁnance/nurture/ grow the acquired asset (39%). Many
deals are also attributed to a buyer's desire to diversify into new regions/product lines. In
partial acquisitions (i.e., asset sales) the seller's desire to focus is often given. Only a
small number of deals (5%) have cost savings or efficiency improvements as an explicitly
stated reason. However, the better ﬁt category may include many cases in which there

are anticipated cost savings or efﬁciency gains from a deal.

2.4 Advertising and Market Share Changes Associated with Brand Transfers

2. 4. I Variability in advertising policy around ownership changes

As we discuss earlier, one way to detect whether ownership changes have an
effect on how assets are managed is to examine whether there are sharp changes in the
intensity with which new owners advertise purchased brands. To investigate, for each
purchased brand we calculate the percentage change in advertising expenditures between
the year prior to deal announcement (hereafter year -1) and the year following deal
completion (hereafter year +1). We seek to determine whether these changes in
advertising are abnormally large on average relative to the benchmark of what one would

expect had there been no ownership change.34

 

3" lrnplicitly we are conditioning on a brand surviving from year -1 to year +1. Our benchmarks are also
chosen conditional on surviving from year -1 to year +1. Thus, we can make inferences on the role of

69

To calculate this benchmark, we use a size-industry matching approach.
Speciﬁcally, for each transferred brand we select brands that are in the same industry and
have advertising falling between 50% and 200% of the sample brand as of year -1. Exact
details on the procedure used for selecting these matches are reported in the appendix.
Since below we use decile-based measures of advertising changes, we create matching
portfolios only for sample brands for which 9 or more matches can be identiﬁed. The set
of matches for each sample brand is referred to hereafter as its matching portfolio.

For each brand in a matching portfolio, we calculate its percentage change in
advertising over the year -1 to year +1 period. We then calculate decile breakpoints for
the matching portfolio based on the derived empirical distribution of advertising changes.
The purchased brand is assigned to a decile based on where its percent advertising change
falls relative to the decile breakpoints derived from its matching portfolio. Observations
in decile 1 (decile 10) are cases in which the sample brand decreases (increases)
advertising most sharply relative to its size-industry group.

If control changes have no effect on advertising strategies, the decile assignments
for purchased brands should be unifonnly distributed from 1 through 10. However, an
inspection of the data reveals an abnormally large portion of observations in the top and
bottom deciles. The ﬁgures in table 2.4 illustrate the basic story. The percentage of
observations in the bottom decile is 19.86%, a ﬁgure that is almost double the expected
rate of 10%. The difference is highly signiﬁcant (p<.01). We also ﬁnd that 13.16% of all
purchased brands fall in the top decile, a ﬁgure that is substantially larger than the

expected rate of 10%. The difference here is signiﬁcant at the 5% level using a simple

 

acquisitions in the management of surviving brands. 1f acquirers tend to eliminate brands with an elevated
frequency, the overall effect of acquisitions on how assets are managed will be understated by our analysis.

70

binomial test, and at the 1% level using a binomial test that conditions on the fact that an
observation is not in the bottom decile. The conditional test is more appropriate if we
seek to understand whether there is an abnormally large upper tail in the distribution of
advertising changes after accounting for the fact that there is certainly an abnormally
large lower tail.

Grouping together the lower and upper deciles, fully 33.03% of sample brands fall
in the top or bottom decile of advertising changes measured relative to their size-industry
group. This substantially exceeds the theoretical expectation of 20% and is highly
signiﬁcant (p<.01). Thus, putting it all together, the results offer quite strong evidence
that ownership changes are associated with abnormally large variability in changes in
advertising spending. The presence of large advertising cuts appears more pronounced
than large increases, but we do ﬁnd evidence for both types Of behavior.

If we create a measure Of a brand's overall change in advertising by subtracting
from each brand's change the median of its matching portfolio, we ﬁnd that the sample
median for these benchmarked changes is -6.03%, a ﬁgure that is not quite signiﬁcant at
the 10% level (p=.105). Thus, the presence of a very large lower tail and a moderately
large upper tail combine to hint at an overall decrease in advertising, but the evidence is
not overwhelming. This may explain why prior tests of investment changes following
control changes have failed to yield strong results; the two abnormally sized tails partially
cancel out and mask much of the interesting variation.

The purchased brands we study may have had abnormally large variation in their
advertising/investment policies even before the change in control. If this is the case, the

event we focus on may not be the driving factor behind the elevated variability uncovered

71

above. To investigate, we identify the brand within each matching portfolio that was
closest to the purchased brand in its percent advertising change between year -2 and year
-1. Thus we essentially match on size, industry, a_nd recent advertising changes. For
each selected match (one for each sample brand), we calculate its percentage change in
advertising from year -1 to year +1 and the decile to which this change corresponds.

As we report in column 2.4 of table 4, 9.77% of these matches fall in the bottom
decile and 8.62% are in the top decile. Neither of these ﬁgures is signiﬁcantly different
from the theoretical expectation of 10%. Thus, it does not appear that brands with similar
recent advertising changes to our purchased brands have an abnormal propensity to fall in
the extreme deciles in their subsequent advertising behavior. Formal tests, reported in
table 4, conﬁrm that the propensity for sample purchased brands to fall into the bottom
and/or top deciles differs signiﬁcant from the selected matches. This evidence lends
further support to the hypothesis that the abnormal variability we report above is in fact

caused by the change in control of the brand.

2. 4.2 Owner type and post-acquisition advertising

Our preceding results demonstrate that changes in ownership are associated with
abnormally large changes in advertising spending in both directions. This is highly
consistent with the notion that the identity of an asset's owner affects how the assets are
managed, presumably because of differences in underlying ﬁrm characteristics. In this
subsection and the next, we attempt to identify some of the underlying characteristics that

inﬂuence a new owner's investment strategy.

72

As we discuss in the introductory sections, there are reasons to expect that private
owners may invest less heavily than their public counterparts either because of lessened
agency problems or heightened ﬁnancial constraints. If this is the case, we would expect
advertising to decrease (increase) when assets go from public to private ownership. In
addition, since foreign owners may have unique motivations for buying and selling U.S.
based assets, these owners may also invest differently than their domestic counterparts.

To investigate, we estimate cross sectional regressions in which the dependent
variable measures advertising changes and the explanatory variables indicate owner type.
We use two different dependent variables to measure post-acquisition advertising
changes over the year -1 to year +1 window. The ﬁrst measure, which we refer to as the
log measure, is deﬁned as the log Of (l + a brand's percentage change in advertising) less
the median of this quantity for the brand's matching portfolio. The second measure,
which we refer to as the decile measure, is the decile assignment for a brand's percentage
advertising change relative to its matching portfolio using the procedure described earlier.
The decile measure will attribute less weight to large changes in advertising since it is
constrained to lie between a ﬁxed lower bound of 1 and an upper bound of 10.

In columns 1 and 2 of table 2.5 we present results for parallel models using the
two alternative dependent variables. These speciﬁcations include four dummy variables
indicating the presence of private buyers, private sellers, foreign buyers, and foreign
sellers. Consequently, the constant in these models will represent an estimate of
advertising changes for deals with a domestic public buyer and seller.

As we report in the table, the coefﬁcient on the private buyer variable is negative

and signiﬁcant in both column 1 and column 2. This indicates that private buyers tend to

73

sharply cut advertising after acquiring a brand. The magnitude of the estimated effect is
surprisingly large. The column 1 estimate indicates that deals with private buyers are
associated with a 60% decrease in industry-adjusted advertising relative to other deals.
The column 2 estimate indicates that the presence of a private buyer is associated with a
movement downwards by 1.6 deciles in a brand's advertising. The coefficient on the
private seller variable in these models is positive, but in both cases insigniﬁcant (t=1.1
and 1.5 respectively). While little can be deﬁnitively concluded from insigniﬁcant
coefficients, the negative sign here is at least consistent with the ﬁndings on the private
buyer variable in suggesting that advertising intensity under private ownership is
relatively low compared to under public ownership

Turning to the foreign ownership variables, the estimated coefficient for foreign
sellers is positive and signiﬁcant in both models 1 and 2 of table 2.5. The estimated
effect is of reasonably large magnitude (36% increase, 1.04 decile increase). This
suggests that foreign owners advertise brands relatively lightly and that a transfer of
ownership to domestic ﬁrms presents an opporttmity for a substantial revision in
advertising strategy. The coefﬁcient on the foreign buyer variable is negative but
insigniﬁcant in both estimated models (t=1.0 and t=1.4 respectively). Again, without
overstating the case, the negative sign for foreign buyers is consistent with the positive
sign for foreign sellers in suggesting that foreign owners advertise brands relatively
lightly compared to domestic owners.

An alternative modeling choice is to assume that ownership type has a symmetric
effect on advertising regardless of whether a certain type of ﬁrm is on the buying or

selling side of a transaction. While this choice imposes additional restrictions on the

74

assumed behavior, the restrictions may enhance our ability to precisely estimate the
underlying effect of ownership type on advertising intensity. To estimate models of this
type, we create a “more private” variable which assumes a value of +1 (-1) if the buyer
(seller) is a private ﬁrm and the seller (buyer) is not. This variable is assigned a value of
0 if the buyer and seller are both private and also if neither are private. We deﬁne a
“more foreign” variable in an analogous way.

Estimates for models using the two dependent variables and these alternative
explanatory variables are reported in columns 3 and 4 of table 2.5. The estimated
coefficient on the more private variable is negative and highly signiﬁcant in both models
(t=2.4 and t=3.0 respectively). This provides additional evidence indicating that control
transactions that increase the private ownership of an asset are associated with signiﬁcant
decreases in advertising spending. The estimated coefﬁcient on the more foreign variable
is negative in both models, but insigniﬁcant in column 3 (Fl .3 8) and signiﬁcant at the
10% level in column 4 (t=1.85). These coefﬁcients are consistent with foreign owners
advertising brands at relatively low levels compared to alternative owners of these assets,
but the evidence appears substantially less conclusive than it is for private owners.

The fact that private ownership is associated with lower levels of investment
spending in the form of decreased advertising is a particularly interesting ﬁnding. This
behavior appears to be particularly prominent for private buyers who frequently cut
advertising sharply. As we discuss earlier, lower spending by private owners is
consistent with these owners either curbing unnecessary spending because of superior
incentives/govemance or cutting proﬁtable spending because of heightened ﬁnancial

constraints. To distinguish between these explanations, we conduct Factiva news

75

searches for each private owner and use this information to characterize each private
buyer and seller as either an Operating company or an investment company. We place in
the investment company group all private ﬁrms that appear to be in the business of
buying, selling, and restructuring individual assets. This category primarily includes well
known private equity, LBO, and ﬁnancial ﬁrms. F inns in the operating company
category are more traditional ﬁrms that appear to be primarily focused on operating their
basic business. Many of these ﬁrms are family owned ﬁrms with a history of stable
ownership.

Given their differences, it is reasonable to expect that the private operating
companies in our sample would have less access to capital than the private investment
companies. Consequently, if our preceding results are driven by private ﬁrm ﬁnancial
constraints, we would expect advertising cuts to be relatively larger for deals involving
the operating company subgroup. However, if, as some have argued, investment
companies have a special skill in creating value by improving operating decisions, then
more spending cuts by these types of acquirers would suggest reduction in overspending.

To investigate, we add to the models of columns 1 and 2 of table 2.5 additional
terms indicating whether the private buyer or seller in a transaction was an operating
company rather than an investment company. These coefficients will indicate the
difference in typical post-acquisition advertising spending revisions between the two
types of private ﬁrms. When we add to columns 1 and 2 of table 2.5 coefﬁcients on the
term indicating that the private buyer is an operating company, both are positive, but only
signiﬁcant at the 10% and 15% levels respectively. This weakly suggests that private

operating companies that acquire new brands tend to cut advertising investment less than

76

their private investment company counterparts. This is what we would expect under the
improved incentives scenario outlined above, and it is opposite of what we would expect
under the ﬁnancial constraints explanation. In these same two models, the coefﬁcients on
the term indicating that the private seller is an Operating company are small, insigniﬁcant,
and change in sign across the two speciﬁcations. This suggests that any difference in
advertising between types of private ﬁrms appears to be driven by the behavior of private
buyers rather than sellers.35

Taken as whole, the results in this section are highly consistent with the notion
that private ownership of a brand is associated with a lower level of advertising spending.
This behavior is most prominent on the buying side of control transactions, as private
buyers appear particularly likely to shift advertising spending on purchased brands
downwards. The tighter purse strings of private owners could reﬂect an ability to avoid
overspending because of better incentives or a propensity to underspend because of
heightened ﬁnancial constraints. The private buyers in our sample who would appear to
have less access to capital do not display a tendency to cut advertising more than other
private buyers. In fact, the evidence points weakly in the other direction. Thus, on

balance, our ﬁndings support the better incentives interpretation of the evidence.

2. 4.3 Firm characteristics and advertising changes

 

3’ We also experimented with altering the models in columns 3 and 4 of table 5 by adding dummy variables
interacting the more private variable with an indicator for whether the private party in the transaction was
an operating company. This added dummy variable was in both cases positive, but far from signiﬁcant. By
treating the buying side and selling side symmetrically, this modeling approach may have low power to
detect differences that exist only on the buying side.

77

The differences we uncover above between private and public owners raise the
possibility that other ﬁrm characteristics also play a role in post-acquisition revisions to
advertising investment strategies. In particular, the existing literature suggests that
factors related to growth prospects, ﬁnancial constraints, and managerial agency
problems may be related to acquirers' investment decisions regarding newly acquired
assets. To investigate, we identify variables related to these factors and examine whether
differences between buyers and sellers along these dimensions are related to post-
acquisition revisions in advertising. Since the selected variables rely on Compustat data,
the sample sizes in this analysis are substantially reduced.36 Thus, our power to detect
any underlying effects may be limited.

One variable that may be particularly relevant is a ﬁrm's Tobin's Q. We expect
that ﬁrms with higher levels of Tobin's Q generally have better investment prospects
which may lead them to advertise more heavily. Alternatively, Tobin's Q may be a
reﬂection of management quality or management's discipline in avoiding overinvestrnent.
Under this alternative scenario, the transfer of a brand from a low Q seller to a high Q
buyer may result in a decrease in post-acquisition advertising as overinvesting behavior is
curbed.

With regard to ﬁnancial constraints, a variety of different measures have been
suggested in the literature. Following recent work in this area (e.g., Almeida, Campello,
and Weisbach (2004), Rauh (2006)), we select dividends, leverage, cash, ﬁrm size, ﬁrm

age, and credit ratings as factors that may be related to ﬁnancial constraints. We expect

 

3" The necessary Compustat data is unavailable (usually unavailable) when either the buyer or seller is a
private (foreign) ﬁrm. Since our prior analysis suggests that foreign acquirers differ from domestic
acquirers, we further restrict the analysis in this subsection to purely domestic deals.

78

that when the buyer is less constrained than the seller, it is more likely that a brand will
experience an increase in post-acquisition advertising as ﬁnancial constraints are relaxed.

For each of the selected ﬁrm characteristics, we create a dummy variable
indicating whether the buyer's value surpasses the seller's. We then investigate in a
regression context whether these dummy variables are related to post-acquisition changes
in industry-adjusted advertising. The dependent variables in this analysis are the same
log and decile based measures discussed and utilized earlier.

As we report in table 2.6, none of the coefﬁcients on the ﬁrm-characteristic
indicator variables are even close to signiﬁcant in these regressions. Since many of these
variables may be collinear, in unreported results we repeat the table 2.6 regressions using
only a single indicator variable at a time. Even in these models, the coefficients on the
indicator variables are in all cases insigniﬁcant. Thus, a simple relation between relative
firm characteristics and post-acquisition advertising changes is not readily apparent.

To further search for a relation, in unreported results we also experimented with
using the log of [1 + (buyer's characteristic / seller's characteristic)] in place of the
indicator variables. With a couple of minor exceptions, the explanatory variables remain
insigniﬁcant. In a few limited cases some of the ﬁnancial constraints variables are
signiﬁcant at the 10% level, but in a subset of these cases the Sign is opposite of our
expectations. We also experiment with using a retum-on-assets-based variable in place
Of the Tobin's Q-based variable. The coefﬁcient is insigniﬁcant in columns 1-3 of table
2.6 and negative and signiﬁcant at the 10% level in the column 4 model. This at best
hints at the possibility that more proﬁtable acquirers are tighter with their advertising

spending as they digest acquired brands. Taken as a whole, this investigation reveals

79

little convincing evidence that ﬁrm characteristics, beyond an owner's
public/private/foreign status, play an important role in post-acquisition advertising

strategies.

2. 4. 4 Buyer Advertising and Brand Overlap

In our initial analysis, we uncovered a particularly high rate of sharp cuts in
advertising following acquisitions. Consistent with some prior discussions, this behavior
suggests that buyers may often be able to exploit valuable cost savings strategies after
completing an acquisition. To further investigate, it is useful to also consider advertising
spending for non-purchased brands held by the buyer. If advertising cost savings are a
beneﬁt underlying some sample deals, we would expect to observe cuts in advertising in
both the purchased brands an_d_ some of the buyer's previously held brands. Alternatively,
if buyers reconﬁgure the allocation of advertising investment dollars over their brand
portfolio after an acquisition, it is possible that buyers shift advertising dollars away from
purchased brands and towards their existing brands, in which case the sign on total
spending changes could go either way.

Prior discussions of cost savings associated with acquisitions suggest that these
savings may vary by the degree of overlap between the buyer and seller's assets. In the
case of advertising, it would appear likely that cost savings are easier to realize for brands
that are closely related to one another. For example, when closely related brands fall
under common ownership, it may become feasible to economize on spending by
advertising the brands together or eliminating advertising intended to shift demand from

one brand to the other. These possibilities suggest that we would observe a decrease in

80

advertising both in the purchased brand and in the buyer's brands that closely overlap
with the purchased brands.

Under the alternative scenario in which acquisitions present opportunities to
optimally shift advertising from one brand to another, it is reasonable to expect this
Shifting to take place more frequently for closely related brands. In particular, after the
purchase the common owner can choose to reallocate spending between the purchased
brand and the closely overlapping brands based on which brand has relatively more
promising future prospects. This possibility suggests that we may actually observe an
increase in advertising for the buyer's bands that most closely overlap with the purchased
brands.

To investigate these issues, we assign each of the buyer's microbrands to either an
overlapping or a non-overlapping group based on whether the brand shares a 4-digit PIB
industry code (assigned by CMR/TN S, additional details in appendix) with one of the
component microbrands of the purchased brand. We then aggregate all of the
overlapping (non-overlapping) microbrands together into a single group that we refer to
as the buyer's overlapping (non-overlapping) brands. Finally, we calculate the percent
change in advertising of the buyer's overlapping and non-overlapping brands over the
same -1 to +1 window used for the purchased brands. We industry adjust these ﬁgures
by subtracting off the median percent change in advertising of a set of size-industry

matches using the procedure described in the appendix.

81

In table 2.7 we report median changes in industry-adjusted advertising for the
different groups of brands.37 As we report in the ﬁrst row, for purchased brands the
median brand experiences a 6.03% decrease in industry-adjusted advertising. This
statistic is signiﬁcant at the 11% level (p=.102) and is a restatement our earlier ﬁnding
that large advertising cuts are more common than large increases, even though both occur
at abnormally large rates.

The second row of table 2.7 reveals that the buyers' overlapping brand experience
a median 11.46% decrease in industry-adjusted advertising, a ﬁgure that is highly
signiﬁcant (p<.01). Thus, it appears that purchased brands and the buyer's overlapping
brands both display a tendency towards decreases in advertising after an acquisition is
completed. This ﬁnding is consistent with what we would expect under the cost savings
scenario outlined above, and appears inconsistent with a scenario in which advertising
dollars are simply shifted from one brand to another.38 Expanding on this ﬁnding, in the
third row of the table we group the purchased brand and the buyer's overlapping brands
together into a single portfolio. Treating this portfolio of brands as a single unit, the
median change in industry adjusted advertising is -7.53%, a ﬁgure that is highly
signiﬁcant (p<.01). This helps conﬁrm that there is an aggregate decrease in advertising
investment for the set of closely overlapping assets involved in the acquisition.

Turning to the non-overlapping brands, the statistics in the ﬁnal row of the table

indicate that these brands experience a median 1.34% increase in industry-adjusted

 

37 Since advertising changes (in percent terms) are extremely skewed and industry benchmarks are based on
medians, we use medians rather than means to gauge the general direction of industry-adjusted advertising
changes.

38 We do not detect any signiﬁcant positive or negative correlation between advertising changes for the

purchased brand and the buyer's overlapping brands. This also suggests that funds are not actively shifted
ﬁom one brand to the other after an acquisition.

82

advertising, a ﬁgure which is insigniﬁcantly different from 0. This insigniﬁcant increase
contrasts sharply with the signiﬁcant decrease for overlapping brands and suggests that
brand overlap is an important determinant Of post-acquisition advertising strategies.39
The fact that advertising is not cut in non-overlapping brands also helps conﬁrm that we
are not simply picking up across the board downward revisions in spending by ﬁrms that
undertake acquisitions. Taken as a whole, the evidence here is highly consistent with the

hypothesis that acquisitions allow buyers to realize cost savings brought about by the

presence of overlapping activities.

2. 4.5 Acquisition returns and advertising behavior

The evidence reported above suggests that some acquisitions allow acquirers to
exploit cost savings opportunities brought about by a transaction. If this cost cutting
activity is efﬁcient, it may explain some of the value gains associated with a change in
ownership. Prior research suggests that anticipated cost savings are, in fact, positively
related to the market's overall assessment of an acquisition (e. g., Houston, James, and
Ryngaert (2001), Bernile (2004)). To investigate this possibility in our context, we
examine whether total buyer plus seller abnormal returns around acquisition
announcements are related to post-acquisition decreases in advertising spending. If post-
acquisition downward revisions in spending are efﬁcient and at least partially anticipated
by the market, we would expect to observe a positive relation between advertising cuts

and market measures of deal synergy gains.

 

39 A formal test for a difference in medians between the buyers' overlapping brands and non-overlapping
brands is signiﬁcant at the 5% level using a non-matched Wilcoxon rank sum test.

83

Before conducting this analysis, we note some potential issues. Since we need
data on both buyer and seller returns to measure anticipated synergy gains, this analysis is
restricted to deals with both a public buyer and seller. In addition, we must conduct this
analysis at the deal level rather than at the brand level, since when multiple brands are
acquired in a deal the same buyer and seller announcement returns will pertain to all of
the transferred brands. These requirements substantially reduce the size of the sample
that we can analyze, potentially leading to limited power.

A ﬁnal issue to note is that actual post-acquisition advertising spending may be a
noisy indicator of anticipated cost savings at the time of deal announcement for at least
two reasons. First, there will likely be some errors in the market's ability to predict
exactly how the acquired assets will be integrated into the buyer's Operations. Second,
many of the cost savings arising from an acquisition are likely related to factors other
than advertising such as savings on material, labor, and other types of investment (e. g.,
capital expenditures, R&D). As these other types of cost savings activities are surely less
than perfectly correlated with advertising cuts, variables based solely on advertising
spending will serve as a noisy proxy for the overall intensity of cost savings efforts. This
noise should bias downwards our estimates of the relation between cost savings activity
and the market's assessment of acquisition synergy gains.

With these considerations noted, we following the prior literature and estimate
models in which the dependent variable is the combined abnormal return of the buyer and
seller upon the announcement of the transaction. The key explanatory variables of
interest are various measures of post-acquisition changes in advertising Spending. To

create these variables, we initially calculate the percent change in advertising between

84

year -1 and +1 for a portfolio composed of all purchased brands plus all of the buyer's
other brands. In some models we adjust this ﬁgure by subtracting off a size/industry
benchmark using the procedure described in the appendix.

Our initial measure of anticipated advertising cost cutting is a simple dummy
variable indicating whether the percent change in advertising is less than the sample
median. Our alternative measure is a logged version of the actual percent change in
advertising spending. If the market is better at predicting the general direction of
spending after an acquisition rather than the precise level of spending changes, the
dummy variable approach may be more appropriate.40 However, it is difficult to
determine whether this is the case on purely a priori grounds. Thus, we experiment with
both possibilities.

In column 1 of table 2.8 we present a simple model in which the sole explanatory
variable is the dummy variable indicating downward pressure on post-acquisition
advertising. The coefficient on this variable is positive and highly signiﬁcant (t=2.78).
This provides some initial support for the hypothesis that the market is optimistic about
cost cutting possibilities created in some acquisitions. The coefﬁcient of .050 is
reasonably large and suggests that deals with above median cost cutting experience a
combined return that is approximately 5% greater than other sample deals. In column 2
we use as an alternative measure of cost cutting a simple log transformation Of the
percent change in advertising [i.e., log (1 + % change in advertising)]. The coefﬁcient on
this variable has a negative Sign, which is consistent with advertising cuts being viewed

positively by the market, but the coefficient is not even close to signiﬁcant (t=0.88). The

 

4° It is also possible that variations in the size of substantial positive changes in advertising are not related
to announcement returns because these deals have primarily non-cost cutting motivations.

85

difference in signiﬁcance of the advertising variables moving from the column 1
speciﬁcation to the column 2 speciﬁcation suggests that the market may have a better
sense of the general direction of future spending changes rather than the precise level.

In columns 3 and 4 of table 2.8 we present analogous regressions but use
size/industry benchmarked (1 changes in advertising rather than raw changes to create the
explanatory variables. The character of the results here are similar to the earlier columns.
In particular, the model using an advertising dummy variable indicates a signiﬁcant
positive relation between cost cutting and announcement returns (t=2.68), while the
model using the level of the advertising change is consistent in sign but not even close to
signiﬁcant (t=-0.70).

To check whether the results on the advertising cut dummy variables are robust to
inclusion of other variables thought to affect synergy gains in control transactions, we
report results in columns 5 and 6 for models that include variables related to buyer and
seller Size, the horizontal nature of the deal, the relative Q levels of the buyer and seller,
the method of payment, and the presence of a foreign buyer or seller. As we report, the
coefficients on the advertising cut variables remain positive and signiﬁcant in these
models (t=1.77 and t=2.59 respectively). None of the added control variables is
signiﬁcant at conventional levels.4|

Taken as a whole, the results are generally supportive of the notion that the

market is more optimistic about deals with more Observed cost cutting. However, these

 

4' The only variable that is close to signiﬁcant is the all stock method of payment variable. If this is
included as the only additional control variable, its coefﬁcient is signiﬁcantly negative (5% level for
column 5 and 10% level for column 6). We have experimented with adding to the column 5 and 6 models
both year dummy variables and a partial sale variable indicating whether the deal was for only part of the
selling ﬁrm. The coefﬁcients on the advertising cut variables are not substantively altered with these
modiﬁcations.

86

results depend on how cost cutting behavior is measured. The data suggest that the
market may more accurately anticipate the general direction of cost cutting activities
rather than the exact level, at least in the case of advertising.42 The evidence we report,
which is based on actual post-acquisition behavior, complements evidence reported by
others on announcement returns and cost savings projections at the time of deal
announcement. Thus, our results help add to the evidence that cost savings are an

important source of value gains in some acquisitions.

2. 4. 6 Acquisitions, advertising, and market share

The abnormal return evidence provides some support for the hypothesis that
downwards shifts in post-acquisition advertising are associated with efficient cost savings
strategies that are enabled by an ownership change. If observed spending cuts are
efficient, the costs of marginal advertising investment should exceed the beneﬁts, with
the principal anticipated beneﬁt being increased future sales. To provide some evidence
on the magnitude of these foregone beneﬁts, we examine changes in market share
subsequent to an ownership change. If advertising or other spending cuts result in
significantly lower sales, we would expect sample brands to exhibit market share losses
in the period following acquisition completion. However, if the spending cuts have very

little effect on sales, market share losses following acquisitions may be negligible.

 

‘2 An alternative way to model the market's ability to predict spending cuts is to create dependent variables
based on advertising changes and use the market's reaction to the deal as an explanatory variable. This is
essentially the reverse regression of the table 8 models. Logit models of this type corresponding to the
speciﬁcations in columns 1, 3, 5 and 6 of table 8 indicate a positive and signiﬁcant relation between the
market's reaction to a deal and the likelihood of a downward shift in advertising.

87

We obtain market share information for a large set of consumer products brands
from Mediamark Research for the period 1995-2003.43 For each acquired brand we
identify the market in this database in which the brand has the largest number of
consumers of one of the brand's products as of year -1. Since the Mediamark data reports
sales in physical units (e. g., cases of beer sold), our measure of brand market share is the
brand's total units sold during the year divided by total units sold for all listed brands in
the market. For later reference, we note that the median (mean) market share of acquired
brands in year -1 is 5.12% (11.14%).

We compute each brand's change in market share in the identiﬁed market between
year -1 and year +1 (the one year window) and between year -1 and year +3 (the three
year window). Since changes in market share may evolve slowly over time, we suspect
that the three year window may be a better indicator of the effects of the new owner's
strategy for managing the brand. However, given the 2003 cutoff for the market share
data, we have many fewer observations when we rely on a longer window. As a
compromise, we also consider a modiﬁed 3-year window for measuring a brand's change
in market share, deﬁned as the change in market share between year -1 and +3 when
available and between -1 and +1 otherwise.

Statistics on the median change in market share for purchased brands over the
different time windows are reported in the ﬁrst row of table 2.9. The median change in

market share is small and insigniﬁcant over all three of the selected windows, with a

 

‘3 Mediamark data has been used in several prior studies (e. g., Agres and Dubitsky (1996), Harless and
Hoffer (2002), and Depken and Wilson (2004)). The entire Mediamark database is prohibitively expensive.
We purchased an extract pertaining to markets for products sold in mass merchandise, general
merchandise, and grocery outlets. Many sample brands compete in these markets and products in these
markets are often heavily advertised.

88

median change of -.02% over the one year window and +09% over the three year
window. These ﬁgures provide no support for the presence of large changes in market
share around the time of control changes.44

In light of our earlier ﬁnding that buyers decrease advertising in brands that
overlap with the purchased brands, it is possible that market share changes are more
apparent in the buyer's overlapping brands. To investigate, for each observation we
identify the buyer's brand that appears to most closely overlap with the purchased brand
using the market deﬁnitions in the Mediamark database (for details see appendix). The
median (mean) market share for these brands as of year -1 is 5.04% (11.21%). As we
report in the second row of table 2.9, market share changes for these buyer brands are
also small and insigniﬁcant. The sample median change is an increase of .01% over the
1-year window and -.20% for the 3-year window.

TO check the robustness of the table 2.9 results, we have conducted the analogous
analysis using means rather than medians. The basic conclusions Of no signiﬁcant
decreases in market share for purchased brands and buyer overlapping brands remain
unaltered. We have also conducted the analysis using alternative deﬁnitions for the
purchased brand's market share including (i) an equally weighted average of market
shares in all markets the brand is listed in and (ii) a value weighted average of market
shares in all markets the brand is present in using weights based on the number of
domestic consumers using the product in year -1. The results under these modiﬁcations

are very similar to what we report in table 2.9.

 

4‘ Additionally, the correlation between benchmarked advertising changes (year -1 to +1) and market share
changes is small and insigniﬁcantly negative (positive) using the 1-year (3-year) window to measure
market share changes. This further suggests that downward revisions in advertising for purchased brands
are not associated with losses in market share.

89

Our analysis of market share changes assumes implicitly that, absent a control
change, the typical change in a brand's market share will be centered around 0. This
seems reasonable, since any loss in market share by one brand will show up as an
increase in market share for a competing brand. To check this assertion, we calculate
changes in market share for a set of control brands. These control brands are selected
from markets immediately following each purchased brand's market in alphabetical order
and are matched to the purchased brands based on year -1 market share. As we report in
the third row of table 2.9, market share changes for this control set are in fact centered
around 0 for all three windows. Formal tests for differences between the brand groupings
in the other rows of the table and the control set are insigniﬁcant for all windows. This
further strengthens our conclusion that there is nothing abnormal about the median
market share change subsequent to an acquisition.

Our market share analysis indicates that, despite the generally downward pressure
on advertising after acquisitions, there is little systematic movement in the market share
of acquired brands and buyers' overlapping brands. The presence of cost cutting behavior
with no signiﬁcant offsetting loss in market share would appear broadly suggestive of
improvements in efficiency. This evidence nicely complements our earlier ﬁnding that
the market attributes larger synergy gains to deals with more post-acquisition cost cutting
in advertising activities. Taken together, a generally healthy picture of post-acquisition

restructuring emerges.

2.5 Conclusion

90

We examine 555 brands that experience an ownership change between 1998 and
2003 and ﬁnd that new owners often alter a brand's advertising strategy. In particular,
buyers display an abnormal propensity to sharply increase or decrease a brand's
advertising spending relative to benchmarks. This ﬁnding is consistent with the
hypothesis that there is a real effect of corporate control activity on the management of
assets. Post-acquisition cuts in advertising spending are particularly common in our
sample. Thus, in an average sense, our ﬁndings suggest that acquisitions exert downward
pressure on investment spending, at least in the form of advertising.

Comparing different types of owners, we ﬁnd that increased (decreased) private
ownership of assets is associated with a downward (an upward) shift in advertising
compared to other deals. A similar but statistically weaker result holds for foreign
ownership. The propensity of private owners to spend less on advertising may reﬂect
their greater ability to avoid unproﬁtable spending. Alternatively, private ﬁrms may be
more ﬁnancially constrained. When we consider acquirers who we suspect have
particularly good access to capital (i.e., private equity ﬁrms, ﬁnancial ﬁrms), we ﬁnd that
they have a somewhat elevated tendency to cut spending in an acquired brand relative to
other private ﬁrms. Thus, on balance, the evidence appears to support the efficient cost
cutting explanation for the private ownership ﬁndings.

Our sample reveals that acquisitions affect no only purchased assets, but also a
buyer's existing assets. In particular, we ﬁnd that buyers tend to cut advertising in their
brands that closely overlap with purchased brands. However, buyers do not signiﬁcantly

change advertising in their other brands. These ﬁndings suggest that acquisitions give

91

rise to cost costing in overlapping activities rather than a shift in spending across the
owner's portfolio of assets.

To help ascertain the efficiency consequences of changes in spending strategies
around acquisitions, we examine the relation between the market's estimate of synergy
gains and changes in advertising spending. Here we ﬁnd that the combined buyer-seller
announcement returns are positively related to some measures of post-acquisition
downward shifts in advertising spending. Since we suspect that observed advertising cuts
are often an indicator of more general cost cutting efforts, this ﬁnding suggests that
efficient costs savings are expected and realized in some acquisitions. Adding further
support to the presence of efﬁcient cost cutting, we ﬁnd that purchased brands do not
tend to experience signiﬁcant losses in market share, even when advertising spending is
revised downwards. This suggests that acquisitions allow cost savings to be realized
without any substantial offset in terms of decreased sales.

Taken as whole, our study provides important evidence that ownership changes
often result in substantial shifts in resource allocation decisions, with private ﬁrms
playing an especially active role in this process. New owners frequently take advantage
Of cost cutting opportunities, particularly in overlapping activities, and on balance the
evidence points to increases in efficiency as a result. Thus, a generally healthy picture Of
ownership changes emerges. While we believe these ﬁndings provide some valuable
insights, they also raise some additional questions. In particular, it would be useful to
assess whether these results apply to other forms of investment, most notably capital
expenditures and R&D. In addition, it would be informative to understand the efficiency

implications of cases in which an acquirer sharply increases investment in an acquired

92

asset, as many acquisitions appear to be motivated by reasons that have little to do with

cost savings. These and related questions must await future research.

93

Creation of matching portfolios for purchased brands

After choosing the initial sample, the advertising information we use is derived
from the Ad $ Spender CDs. We have access to two CDs, one for the 1997-2001 period
and one for the 1999-2003 period. Each CD lists microbrand advertising for tens of
thousands of microbrands for each year plus the owner associated with the microbrand at
the end of the CD coverage period. For each ownership change, we select the CD that
includes both year -1 and year +1 and calculate advertising changes using data from the
selected CD. If both CD3 include years -1 and +1, we select the more recent CD. Since
it is not practical to aggregate microbrands into brand groupings by visual inspection for
an entire CD, for matching portfolio purposes we deﬁne a brand to be the aggregation of
all of the microbrands associated with the same owner in any given year. This should
represent a reasonable measure of aggregate advertising spending on a group of
connected microbrands. By grouping microbrands together by a common eventual
owner, we eliminate any changes in advertising associated with acquisitions or
divestitures. Thus, advertising changes for matching portfolio brands will represent
organic changes in advertising for a related set of microbrands -- an appropriate
benchmark if we wish to assess typical levels of advertising changes for a brand.

For each sample brand with an ownership change, we select for inclusion in its
matching portfolio all brands (as deﬁned above) that satisfy size and industry matching
criteria based on year -1 characteristics. First we require that the advertising of the
potential matching brand falls between 50% and 200% of the sample brand and above a
minimum of $50,000. Second, we require that the matching brand operates in the same

2-digit PIB industry code as the sample brand. PIB industry codes are assigned by

94

CMR/TN S at the microbrand level and we assign a code to a brand as a whole based on
which code is responsible for the largest portion of a brand's total advertising as of year -

1.

Creation of matchigportfolios for buyers' brands

When we select matches for the buyer's brands (overlapping brands as a group,
non-overlapping brands as a group, and all brands as a group), we select matches in a
method analogous to our procedure for purchased brands. In this matching, size is
deﬁned as the sum of all spending on all component microbrands for the group being
considered and industry is deﬁned as the 2-digit PIB code representing the largest portion
of advertising in year -1 within the set of component microbrands. If sufficient matches
cannot be found, we again repeat the procedure at the 1-digit level. Since we do not use a
decile analysis for buyer's brands, we insist on only a minimum of three size/industry
matches to create an industry benchmark for buyers. The size/industry benchmark for
any brand group is set equal to the median of all other brands that satisfy the matching

criteria.

Creation of matching portfolios forﬁvertisingﬁ purchased and buvers brands

To benchmark the combined advertising of purchased brands plus buyers' existing
brands we identify all possible permutations that can be created by selecting one brand
from the matching portfolio of the relevant purchased brand and one brand from the
matching portfolio of the relevant buyers' brands. In creating these portfolios we weigh

the two component parts by the relative level of year -1 advertising in the purchased

95

brand and the buyers' existing brands. After identifying all associated permutations we
deﬁne the relevant benchmark to be the median percent change in advertising within the

set of permutations.

Selection Of buyer's overlapping brands for marlgt share analysis

If a buyer has a brand in the exact same market as the purchased brand's largest
market, we select this brand as the matching brand. If there are multiple qualifying buyer
brands, we select the one with market share closest to the purchased brand's market share.
If there are no buyer brand matches in the purchaser's largest market, we repeat the
procedure on the purchaser's second largest market (based on number of consumers using
the product in year -1) and if this fails the third largest market, fourth largest, etc. If this
procedure yields no buyer brand matches, we select the buyer brand in the market that is
most similar to the markets in which the purchaser's brands are listed (e. g., the hair
shampoo market could be used as a match for the hair conditioner market). If the buyer
does not have any brands listed in markets that are, in our judgment, at least somewhat
related to the purchased brands markets, we exclude the buyer ﬁ'om the sample of buyer

market share changes.

96

 

Table 2.1: Sample Characteristics

The sample includes all brands with an ownership change identiﬁed by screening the
1998 and 2003 lists of brand owners in Company Brand $ and investigating news
accounts of the potential ownership change. A microbrand is an individual listing in the
advertising data sources while a brand is the aggregate set of microbrands that share a
common brand name. A partial acquisition is a case in which the buyer purchases some,
but not all, of the assets of the seller. Firms are assigned to the public/private/foreign
categories based on information in Compustat/CRSP and Factiva news searches. Figures
on the identity of buyer and seller and pre-deal advertising levels are calculated using the
brand as the level of observation. All dollar ﬁgures are inﬂation adjusted to 2003 dollars
and reﬂects advertising in the year prior to the ownership change. Pre-deal advertising
ﬁgures are the brand's total advertising in the 10 tracked media categories reported in the
CMR/TNS data ﬁles as reported in the Ad $ Spender database. Compustat reported
advertising is the total advertising ﬁgure reported in Compustat. For all private ﬁrm and
many foreign firms, the seller's book assets and Compustat total advertising are missing.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Statistic
Number of brands with an ownership change -- all years 555
Number of brands transferred in 1998 99
Number of brands transferred in 1999 126
Number of brands transferred in 2000 164
Number of brands transferred in 2001 63
Number of brands transferred in 2002 61
Number of brands transferred in 2003 42
Brands transferred in full acquisition/merger 377
Brands transferred in a partial acquisition 178
Mean number of microbrands composing the brand 5.10
Number of distinct control events 348
Deals with single brand transferred 264
Deals with multiple brands transferred 84
Mean level of pre-deal advertising ($ millions) 11.46
Mean level of (pre-deal brand advertising/Compustat reported 21.33%
advertising)
Mean level Of(pre-deal brand advertising/seller's book assets) 1.12%
Seller is a public U.S. ﬁrm (%) 55.14%
Seller is a private U.S. ﬁrm (%) 21.08%
Seller is foreign ﬁrm (%) 23.78%
Buyer is apublic U.S. ﬁrm (%) 49.91%
Buyer is a private U.S. ﬁrm (%) 14.23%
Buyer is foreign ﬁrm (%) 35.86%

 

97

 

Table 2.2: Characteristics of Buyers and Sellers

All variables except for those related to advertising and amount paid are derived from
Compustat data as of the ﬁscal year end irrrrnediately preceding a deal announcement.
Consequently, these statistics are limited almost exclusively to domestic public ﬁrms. In
full ﬁrm acquisitions the entire acquired ﬁrm is considered to be the seller. All dollar
ﬁgures are adjusted to 2003 dollars. Tracked advertising is the total reported advertising
in the year prior to deal announcement (in $ millions) in the 10 media outlets tracked by
the CMR/T'NS data sources. Amount paid in a deal is the stated deal value as revealed by
news articles or the SDC entry on the transaction. Age is deﬁned as the number of years
since a ﬁrm's ﬁrst listing on Compustat. Tobin’s Q is deﬁned as beginning of period
(book assets + market equity - book equity)/(book assets). Sales grth is deﬁned as the
percentage change in ﬁrm sales between two years prior to deal announcement and the
year prior to deal announcement. Dividend paying status is a dummy variable assuming
a value of 1 if the ﬁrm reports non-zero dividends.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Mean Median Observations
Panel A: Seller characteristics
Book assets 15,240.49 6,824.47 366
Age 27.73 30 366
Tobin's Q 2.44 1.68 364
Past year sales growth 13.91% 8.08% 365
Book long-term debt / book assets 28.10% 24.15% 366
Dividend paying status 73.15% 1 365
Panel B: Buyer characteristics
Book assets 25,372.97 9,364.40 375
Age 27.24 30 375
Tobin's Q 3.48 2.11 364
Past year sales growth 17.90% 6.71% 373
Book long-term debt / book assets 21.65% 16.91% 375
Dividend paying status 70.96% 1 365
Buyer's tracked advertising 212.57 55.309 365
Panel C: Buyers vs. sellers
Seller assets / Buyer assets 0.61 272
Amount paid / Buyer assets 0.40 328
Seller age / Buyer age 1.00 272
Seller Q / Buyer Q 0.88 268
Seller sales growth / Buyer sales growth 0.49 270
Seller leverage / Buyer leverage 1.21 269
Dummy for buyer and seller in same 2- 72.79% 1 272
digit SIC industry

 

 

Table 2.3: Details of Ownership Change

In calculating returns, each control transaction is treated as a single observation, even
when multiple brands are transferred in a single deal. The ﬁgures in Panels B and C are
calculated treating each transferred brand as a separate Observation. Abnormal Return is
the abnormal return for a three-day window centered on the announcement date and
calculated from a market model estimated over the period from 240 to 41 days before the
announcement. We require at least 100 trading days over the estimation window to
calculate abnormal returns. Combined abnormal returns are the value-weighted average
of the bidder and target returns. Signiﬁcance levels for means are based on tests that
standardized prediction errors are equal to zero (t-statistics in parentheses below reported
means). In the case of combined returns, t-statistics are calculated using the value
weighted average of the standardized prediction errors of the buyer and seller/target for
each event and then aggregating across events. Signiﬁcance levels for medians are based
on sign tests on the number of positive vs. negative observations (number positive and
number negative reported in this order in parentheses below medians). Method of
payment information is revealed by news articles or the SDC entry on the transaction.
Debt assumption is ignored in determining the method of payment. Reasons for the deal
are non-mutually exclusively categories assigned based on reading the longest news
articles released at the announcement of the deal.

*** (**,*) denotes signiﬁcance at the 0.01 (0.05, 0.10) level using the respective tests
described above.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Mean Median Obs.
Panel A: Announcement returns
Seller/target abnormal return 12.34%*** 6.23%*** 246
(9.94) (191, 55)
Buyer abnormal return 0.004% 0.001% 255
(0.51) (130 ,125)
Combined abnormal return 2.13%*** 1.51%*** 169
(3.69) (109, 60)
Panel B: Method Of payment
Dummy variable for all stock deal 35.13% 464
Dummy variable for all cash deal 51.72% 464
Dummy variable for mixed cash and stock 13.15% 464
Panel C: Stated reasons for deal
Good/better ﬁt with buyer 43.00% 379
Buyer can better ﬁnance/nurture/ grow the asset 38.79% 379
Buyer desire to diversify (geographic and product 25.59% 379
line)
Seller desire to focus 23.48% 379
Efficiency/costs savings 5.01% 379
Other 2.37% 379

 

 

 

99

Table 2.4: Variability in Advertising and Brand Ownership Changes

Each brand that experiences a change in ownership is treated as a single sample
observation. For each observation we select a portfolio of size/industry matching brands
(minimum of 9) using the criteria outlined in the appendix. The percent change in
advertising is calculated for sample brands and matching brands over the year -1 to year
+1 window. For each matching portfolio decile breakpoints are chosen based on the
derived empirical distribution of advertising changes. Each purchased brand is then
assigned to a decile based on the breakpoints for its matching portfolio. Breakpoints are
chosen based on the "altdef" interpolation rules implemented by Stata. The ﬁgures in
column 1 indicate the fraction of sample brands that fall into the different decile groups.
The ﬁgures in column 2 are p-values for a binomial test of whether the ﬁaction of
observations in column 1 differs from the theoretical expectation of 10% for each decile.
The conditional binomial test in column 3 for decile l (decile 10) is a test for whether the
number of brand transfers in the indicated decile differs from theoretical predictions
under the null hypothesis of a uniform distribution across deciles, conditioning on the fact
that the observation does not lie in decile 10 (decile 1). In column 4 we select a single
match for each sample brand based on the matching observation that is closest to the
sample observation in its percentage change in advertising between year -2 and year -1.
The reported ﬁgures in the column are the fraction of these matches that fall into each
decile category using the same cutoff points that are used for the sample observations.
The p-values in column 5 are for a simple t-test of whether the fraction of observations in
column 1 differs from column 4. There are 433 sample brands with sufficient matches to
assign to a decile in column 1. There are 348 matches for these sample brands using the
matching criteria in column 4. The difference arises because of a lack of machine
readable advertising information for the matching brands prior to 1997.

 

 

 

 

 

 

 

 

 

 

 

 

 

Column 1 Column 2 Column 3 Column 4 Column 5
% of deals Binomial test Binomial % selected Sample vs.
unconditional test matches matches
conditional
Advert. change in decile 1 19.86% p=0.00 p=0.00 9.77% p=0.00
Advert. change in decile 10 13.16% p=0.04 p=0.00 8.62% p=0.03
Advert. change in decile l or 33.03% p=0.00 18.39% p=0.00
10
Advert. change in decile 2-9 66.97%

 

100

 

Table 2.5: Advertising Changes and Type of Owner

All estimates are OLS estimates with heteroskedasticity consistent standard errors
reported in parentheses under the coefficient estimates. Each brand with a change in
owner is treated as a separate observation. Advertising changes are calculated as the
percent change in advertising in year +1 relative to year -1. The dependent variable in
columns 1 and 3 is deﬁned as log (1+ brand's percent change in advertising) minus the
median of this same quantity calculated over the brand's size-industry matching portfolio
created using the procedure outlined in the text and appendix. The dependent variable in
columns 2 and 4 is the decile ranking (from 1 to 10) of a brand's percentage change in
advertising between year -1 and year +1 calculated relative to the empirical distribution
of advertising changes for its matching portfolio. The private buyer (seller) variable is a
dummy variable assuming a value of 1 if the buyer (seller) is a non-foreign private ﬁrm.
The More Private variable assumes a value of -1 (+1) if the seller (buyer) is a private
non-foreign ﬁrm and the buyer is not. Otherwise this variable assumes a value of 0. The
foreign buyer (seller) variable is a dummy variable assuming a value of 1 if the buyer
(seller) is a foreign ﬁrm. The More Foreign variable assumes a value of -1 (+1) if the
seller (buyer) is a foreign ﬁrm and the buyer is not. Otherwise this variable assumes a
value of 0. Pre-deal advertising is the advertising of the brand in the year prior to the
deal announcement (inﬂation adjusted to 2003 dollars).

*Signiﬁcant at the 10% level, “Signiﬁcant at the 5% level, ***Signiﬁcant at the 1%
level

101

Table 2.5

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Dependent Variable
Log Change Decile Log Change Decile
Ad. Change Ad. Ad. Change Ad
(1) (2) (3) (4)

Private Buyer -0.60** -l .60***

(0.24) (0.49)
Private Seller 0.18 0.65

(0.16) (0.40)
More Private -0.31*"‘ -0.93***

(0.13) (0.31)

Foreign Buyer -0.15 -0.53

(0.15) (0.37)
Foreign Seller 0.36** 1.04**

(0.16) (0.43)
More Foreign -0.18 -0.61*

(0.13) (0.33)

Log(pre-deal -0.06 -0.15 -0.03 -0. 10
advertising) (0.04) (0. l 0) (0.04) (0.09)
Constant 0.22 6.37* * * 0.02 5.91 ***

(0.39) (0.85) (0.33) (0.78)
R2 .0293 .036 0.016 .025
Number of 433 433 433 433
Observations

 

102

 

Table 2.6: Advertising Changes and Relative Characteristics of Buyer and Seller

All coefficients are OLS estimates with heteroskedasticity-consistent standard errors
reported under each coefficient estimate in parentheses. Each brand with a change in
owner is treated as a separate observation. All models are restricted to brands with U.S.-
based Compustat-listed buyers and sellers. Advertising changes are calculated as the
percent change in advertising in the year following deal completion relative to the year
prior to deal announcement. The dependent variable in columns 1 and 2 is deﬁned as log
(1+percent change in advertising) minus log (1 + median advertising change of
size/industry matching portfolio) where the matching portfolio is chosen using the
procedure outlined in the text and appendix. The dependent variable in columns 3 and 4
is an integer (1 through 10) corresponding to the decile a brand's change in advertising
falls relative to its size/industry matching portfolio. All explanatory variables that begin
with the word "Buyer" are dummy variables indicating whether the buyer is different
from the seller/target on the indicated dimension in the year prior to the deal
announcement. Dividend comparisons are based on the ratio of dividends to net income.
Leverage comparisons are based on the ratio of short term plus long-term debt divided by
book assets. Cash comparisons are based on the ratio of cash to book assets. Size
comparisons are based on book assets. Age comparisons are based on years since ﬁrst
listing in Compustat. The credit rating variable is set equal to 1 if both ﬁrms have S&P
senior credit ratings listed in Compustat and the buyer's rating is strictly higher than the
seller. If one or both ﬁrms do not have a reported rating the associated explanatory
variable is set equal to 0. Tobin’s Q is deﬁned as beginning of period (book assets +
market equity - book equity)/(book assets). Pre-deal advertising is the advertising of the
brand in the year prior to the deal announcement (inﬂation adjusted to 2003 dollars).
ll'Signiﬁcant at the 10% level, “Signiﬁcant at the 5% level, ***Signiﬁcant at the 1%
level

103

Table 2.6

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Dependent Variable
Log Log Decile Decile
Change Change Change Change Ad
Ad. Ad. Ad. (4)
(1) (2) (3)
Buyer div. payout > Seller -0.027 -0.102
(0.267) (0.622)
Buyer leverage > Seller -0.063 0.172
(0.266) (0.619)
Buyer cash/assets > Seller 0.213 0.509
(0.255) (0.744)
Buyer book assets > Seller 0.059 -0.l31
(0.306) (0.726)
Buyer age > Seller -0.146 -0.114
(0.253) (0.669)
Buyer credit rating > Seller -0.251 -0.3 53
(0.283) (0.628)
Buyer Tobin's Q > Seller -0.213 -0.177 -0.188 -0.256
(0.185) (0.218) (0.534) (0.655)
Log(pre-deal advertising) -0.012 0.005 -0.098 -0.086
(0.067) (0.089) (0.143) (0.169)
Constant 0.062 -0.052 6.235*** 6.102***
(0.587) (0.706) (1.291) (1.404)
R2 0.006 0.018 0.004 0.009
Number of Observations 153 153 153 153

 

104

 

Table 2.7: Advertising Changes in Buyer's Brands and Purchased Brands

Each purchased brand with a change in owner is treated as a separate observation.
Advertising changes are calculated as the percent change in advertising in the year
following deal completion relative to the year prior to deal announcement. A purchased
brand's industry adjusted change in advertising is the percent change in advertising less
the median percent change in advertising for a size/industry matching portfolio chosen
using the procedure outlined in the text and appendix. For each purchased brand we
identify all microbrands of the buyer that are in the same 4-digit industry group as any of
the purchased microbrands using the PIB industry assignment codes in the CMR/TNS
data ﬁles. These brands are deﬁned to be overlapping brands and all other buyer brands
are considered non-overlapping brands. Advertising spending for buyer overlapping and
non-overlapping brands are aggregated up to a single number and changes for these
numbers are calculated over the same window as the corresponding purchased brands.
Industry-adjusted ﬁgures for the buyer’s brands are created in the same manner as the
purchased brands except that we require only a minimum of three size/industry matching
brands in the calculation of the industry median benchmark. In the row pertaining to the
purchased brand plus the overlapping buyer's brands we aggregate the ﬁgures for these
brands together and calculate the percent change. This ﬁgure is industry-adjusted using
the median return from the set Of all possible permutations Of portfolios created from one
of the purchased brand's matches and one of the buyer's matches weighted by the relative
year -1 advertising of the purchased brand and the buyer's overlapping brands. P-values
for whether the median industry adjusted change in advertising is different from zero are
calculated using a two-sided sign test.

 

 

 

 

 

Median Percent P-value Obs.

Change in Negative

Ind. Adj.

Adv.
Purchased brand -6.03% 54.04% 0.10 433
Buyer's overlapping brands -11.46% 61.76% 0.00 238
Purchased brand plus buyer's -7.53% 57.21% 0.00 208
overlapping brands
Buyer's non-overlapping 1.34% 47.86% 0.51 280
brands

 

 

 

 

 

 

 

105

Table 2.8: Acquisition Announcement Returns and Advertising Behavior

Reported coefﬁcients are OLS estimates with heteroskedasticity-consistent standard
errors reported in parentheses under the coefficient estimates. The dependent variable in
each regression is the total percent change in the combined value of the buyer and the
seller upon announcement of the acquisition. Combined buyer and seller returns are
calculated over a three-day window centered on the announcement date in the manner
detailed in the note to table 3. Deals in which multiple brands are transferred are treated
as a single observation and all associated advertising variables of the component brands
are aggregated together. For each transaction we calculate the total percent change in
advertising for all purchased brands and all of the buyer's existing brands between the
year prior to deal announcement (year -1) and the year following deal completion (year
+1). To create the advertising benchmark we form portfolios from all possible
permutations of the matching portfolio of the most heavily advertised purchased brand
and the matching portfolio of the buyers' brands, weighted by the relative year -1
advertising of the purchased brands and the buyer's brands (additional details in
appendix). The benchmark is deﬁned to be the median percent change in advertising
within this set of permutations. The dummy variables for advertising changes assume a
value of 1 if the indicated advertising change in a transaction is less than the sample
median and 0 otherwise. In construction of the explanatory variables the bidder and
seller's market equity are calculated as of the last trading day prior to the event window
using CRSP data. The size of the bidder relative to the target is measured as [Log(buyer
market equity/seller market equity)]/1000. An acquisition is treated as horizontal if the
buyer and seller have the same 2-digit SIC code. Tobin's Q is deﬁned as in the earlier
tables and is measured as of the ﬁscal year end preceding the deal announcement.
Method of payment is determined from Thompson SDC information and news articles
detailing the transaction. The increased foreign ownership variable assumes a value Of l
(-1) if the asset is transferred from a U.S. (non-U.S.) owner to a non-U.S. (U .8.) owner
and 0 if the deal involves no foreign buyer or seller.

*Signiﬁcant at the 10% level, ”Signiﬁcant at the 5% level, ***Signiﬁcant at the 1%
level

106

Table 2.8

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

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

Durruny for advertising 0050*" 0039*

change < median (0.018) (0.022)

Log (1 + percent -0.013

change advertising) (0.015)

Dummy for (advert. 0.051 *** 0.057"

change - benchmark) < (0.019) (0.022)

median

Log (1 + percent -0.010

change advertising) - (0.014)

Log (1 + benchmark)

Log (bidder market 0.002 0.005

equity) (0.005) (0.005)

Size of bidder relative -0.001 -0.004

to target (0.003) (0.003

Horizontal deal dummy 0.013 0.014
(0.025) (0.026)

(Bidder Q > Target Q) -0.012 -0.025

dummy (0.022) (0.024)

All stock dummy -0.042 -0.043
(0.029) (0.028)

Mixed cash and stock 0018 -0.004

dummy (0.029) (0.029)

Increased foreign 0.019 0.005

ownership (0.029) (0.030)

Constant -0.020 0.006 -0.023* 0.001 -0.032 -0.079

(0.013) (0.010) (0.014) (0.010) (0.085) (0.087)
Observations 84 84 74 74 70 64
R2 0.082 0.009 0.091 0.007 0.127 0.164

 

107

 

Table 2.9: Market Share Changes following Brand Ownership Changes

The reported ﬁgure in the top of each cell is the median change in market share for the
indicated set of brands over the indicated time window. The ﬁgure in square brackets
refers to the number of observations over which this median was calculated and the ﬁgure
in parentheses is the p-value for a Sign test indicating whether the median is different
from zero. Market share information is collected from a Mediamark database extract of
all consumer products markets. For each purchased brand we identify the market in
which the brand has the most reported consumers in the year prior to the deal
announcement. Using this as the relevant market, we calculate the change in the brand's
market share from the year prior to deal announcement (year -1) to the year immediately
following deal completion (year +1) and the annual period 3 years following deal
completion (year +3). The modiﬁed -1 to +3 column indicates brand market share
changes between years -1 to +3 when available and years -1 to +1 otherwise. The ﬁrst
row includes all sample purchased brands with the required data. The buyer's brand row
indicates changes in market share for the buyer's brand that appears to be most closely
related to the purchased brand using the algorithm outlined in the appendix. The control
set of brands are matches chosen for each purchased brand based on having the closest
market share (subject to a 10% maximum difference) to the purchased brand within the
set of brands in the market alphabetically following the purchased brand.

 

 

 

 

 

 

 

 

 

 

 

Period for measuring market share
changes
-1 to +1 -1 to +3 Modiﬁed
-1 to +3
Purchased brand: all -0.0002 0.0009 -0.0001
[139] [68] [141]
(0.4976) (0.2750) (1 .000)
Buyer's brand 0.0001 -0.0020 -0.0020
[105] [56] [111]
(0.8454) (0.5703) (0.2546)
Control set of brands -0.0005 -0.0002 -0.0003
[139] [65] [141]
(0.3964) (0.8043) (0.5006)

 

 

ESSAY 3. QUALITATIVE MEASURES OF FINANCIAL CONSTRAINTS: A
CLOSER EXAMINATION

3.1 Introduction

A large body of literature has explored the role of ﬁnancial constraints on
investment behavior45 . A common way to detect the presence of these constraints is to
examine investment-cash ﬂow sensitivities. A seminal paper in this area is by Fazzari,
Hubbard, and Petersen (1988), (hereafter FHP), who examine the sensitivity of
investment to cash ﬂow across sub-samples of constrained ﬁrms identiﬁed by dividend
payout. They document that ﬁrms in the smallest dividend payout group display the
largest investment-cash ﬂow sensitivity and thus conclude that investment-cash ﬂow
sensitivities capture the ﬁnancial constraint status of a ﬁrm. In an inﬂuential follow-up
paper to FHP, Kaplan and Zingales (1997) (hereafter KZ) use a detailed qualitative
analysis of the ﬁnancial statements of ﬁrms that FHP classify as ﬁnancially constrained
in order to classify ﬁrms into ﬁnancial constraint groups based on qualitative measures.
KZ ﬁnd that these qualitative measures of ﬁnancial constraints are not positively
correlated with investment-cash ﬂow sensitivities. In fact, they argue that the relation is
actually negative and present evidence that ﬁrms that appear to be less constrained based
on qualitative measures display higher investment-cash ﬂow sensitivities. They proceed
to argue that investment-cash ﬂow sensitivities are not valid measures of ﬁnancial

constraint status. In the ensuing literature, many authors have borrowed from the K2

 

'5 See Stein (2003), Fazzari, Hubbard, and Petersen (1988), Kaplan and Zingales (1997), Lamont (1997),
Rauh (2006), and Shin and Stultz (1998)

109

ﬁndings to construct an index of ﬁnancial constraints based on the qualitative information
in the KZ sample“.

In this paper we extend the work of KZ by replicating their methodology on a
much larger sample (400 ﬁrms vs. 49 ﬁrms) over a more recent time period (1995-2004
vs. 1970-1984). My sample also displays much more heterogeneity, as we do not restrict
attention to ﬁrms that are initially suspected of being ﬁnancially constrained and do not
restrict ﬁrms to be in the manufacturing industries (4-digit SIC between 2000 and 3900
inclusive).

We show that the distribution of ﬁnancial constraint status has drastically changed
over time. This could either be directly due to the difference in time period studied
between this study and KZ or due to the lack of ﬁrm heterogeneity in previous work.

For instance, the percent of ﬁrms classiﬁed as Not Financially Constrained has decreased
from 54.50% to 11.50% between the 19703 and 19803 and mid 19903 and mid 20003 and
the percent of ﬁrms classiﬁed as Possibly Financially Constrained has increased from
7.30% to 46.96% over the same time periods. These results hold even after controlling
for industry across the two time periods.

Additionally, the correlation between qualitative measures of ﬁnancial constraints
and quantitative accounting variables has also changed. More importantly, the
correlations have not only changed but have reversed sign and become insigniﬁcant in
the vast majority of cases. This may not come as a surprise, as many of the quantitative
variables used to proxy for ﬁnancial constraint status of a ﬁrm are endogenous (such as

leverage and cash holdings).

 

4" See Malmendier and Tate (2005), Almedia, Campello, and Weisbach (2004), Baker, Stein, and Wurgler
(2003), and Lamont, Polk, and Saa-Requejo (2001).

110

Finally, as KZ document, we also show that using investment-cash ﬂow
sensitivities as a gauge Of constraint status may lead researchers astray. However, our
updated estimates indicate that at the very least the relation between investment and cash
ﬂow has drastically reduced over time.

The rest of the paper is organized as follows. In section 3.2 we ﬁrst discuss the
relevant literature on ﬁnancial constraints. In section 3.3 we discuss our empirical
strategy, data, and sample selection procedures. In section 3.4 we present our main (4
analysis of the correlation between qualitative and quantitative measures of ﬁnancial

constraints and investment-cash ﬂow sensitivities. Section 3.5 concludes.

 

(ﬁ—

3.2 Related Literature
3. 2.1 Financial constraint status and investment

As we indicate in the introduction, a large body of literature has investigated both
the role of cash ﬂow on investment as well as the classiﬁcation of ﬁnancially constrained
ﬁrms. The two seminal papers in the area of the classiﬁcation of ﬁnancially constrained
ﬁrms are F HP and K2. FHP put forth evidence that sample ﬁrms with the lowest
dividend payout ratio, and therefore presumably the most ﬁnancially constrained, display
a higher sensitivity of investment to cash ﬂow. They interpret this result as evidence that
investment-cash ﬂow sensitivities provide meaningful measures of ﬁnancial constraint
status. KZ present results driven by the novel idea of classifying ﬁrms as ﬁnancially
constrained based not only on quantitative measures (debt ratio, cash holdings, dividend
payout ratios, etc.), but also on qualitative statements made in annual reports to

shareholders and 10-Ks. Further, they investigate the correlation between qualitative and

111

quantitative measures and the impact of cash ﬂow on investment by constraint sub-
sarnples when the sub-samples are created using qualitative indicators of equity
dependence or ﬁnancial constraint status. KZ examine the 49 low dividend paying
(ﬁnancially constrained) ﬁrms of F HP and ﬁnd opposite results when utilizing qualitative
measures of ﬁnancial constraint status. In other words, unconstrained ﬁrms have higher
sensitivity of investment to cash ﬂow than constrained ﬁrms. However due to the sample
originally investigated by FHP, the work by KZ is restricted to 49 Value Line if
manufacturing ﬁrms between 1970-1984, all of which are required to be present for the

entire sample period. '

 

3.3 Empirical Strategy, Data, and Sample Selection
3. 3.1 Empirical strategy

In the spirit of KZ, we reproduce their methodology and provide large sample
evidence of the correlations between qualitative and quantitative measures of ﬁnancial
constraint status and the relation between investment and cash ﬂow by qualitative
ﬁnancial constraint status.

We ﬁrst present sample-wide ﬁnancial constraint classiﬁcation by ﬁrm-year in
order to address the stability of ﬁnancial constraint classiﬁcation over time. After
examining the stability of ﬁnancial constraints between 1970-1984 and 1995-2004 we
present the time series properties of ﬁnancial constraint classiﬁcation within the 1995-
2004 sample period. Following KZ, we then obtain the correlations between qualitative

and quantitative measures of ﬁnancial constraint status. After partitioning the sample

112

ﬁrms into qualitative measures of ﬁnancial constraint status, we estimate investment-cash

ﬂow sensitivities by ﬁnancial constraint sub-samples.

3. 3.2 Data and sample selection

Following KZ, data for quantitative variables are taken from COMPUSTAT and
we obtain data used to create qualitative measures of ﬁnancial constraint status from
statements made by ﬁrms in 10-Ks and annual reports. Following standard practice we
eliminate ﬁnancial and regulated ﬁrms (ﬁrms in 4-digit SIC codes between 6000 & 7000
and 4900 & 4949 inclusive), ﬁrms without a stock price, and ﬁrms incorporated outside
the U.S.

In order to examine the robustness of the results of K2 to a random sample of 400
COMPUSTAT ﬁrms we ﬁrst obtain a sample of 9775 unique ﬁrms that were on the
COMPUSTAT tapes for at least one year during the time period 1995-2004. We then
sort by GVKEY and select every 24th entry to obtain the random sample.47 After arriving
at the ﬁnal sample we utilize Lexis-Nexis to obtain qualitative statements made by ﬁrms
in annual 10-Ks and/or annual reports in order to create the ﬁnancial constraint variables
and utilize COMPUSTAT in order to create the quantitative variables.

Table 3.1 presents the median of the quantitative variables used in the study.
Column 1 of table 3.1 reports the medians of the full sample, for comparison pmposes
column 2 restricts the updated sample to contain only manufacturing ﬁrms as in KZ, and

column 3 reports the medians of the quantitative variables reported in table 3 of KZ. As

 

‘7 This procedure gives us 407 unique ﬁrms. We sort by GVKEY and take the ﬁrst 400 to create the
sample.

113

table 1 indicates, sample ﬁrms have lower investment as a percent of plant property and
equipment (0.210 vs. 0.348), lower amounts Of debt (0.225 vs. 0.349), larger cash
holdings (0.455 vs. 0.168), and smaller cash ﬂow as a percent of plant property and
equipment (0.227 vs. 0.421) compared to the sample-wide medians reported by K2. The
ﬂavor of these differences holds whether we examine sample-wide medians or restrict

attention to manufacturing ﬁrms.

3. 3.3 Creation of financial constraint variables

As in KZ, we create ﬁve mutually exclusive measures of ﬁnancial constraint
status by ﬁrm-year. Not Financially Constrained (NFC) ﬁrm-years are those in which a
ﬁrm either (1) initiated or increased dividends (2) repurchased stock (3) made explicit
statements in annual reports of having excess liquidity (4) had large amounts of cash
relative to investment or (5) were not restricted from paying dividends. Likely Not
Financially Constrained (LNFC) ﬁrm-years were those in which a ﬁrm either (1) made
no clear statements of excess liquidity in annual reports or (2) had lower cash to
investment ratios than ﬁrms classiﬁed as NFC. Possibly Financially Constrained (PFC)
ﬁrm-years are those in which a ﬁrm made contradictory statements regarding liquidity in
annual reports. Likely Financially Constrained (LFC) ﬁrm-years are those in which ﬁrms
(I) mention difficulties in raising money in annual reports (2) postpone equity offerings
due to market conditions (3) are prevented from paying dividends or (4) have little cash
relative to investment. Financially Constrained (FC) ﬁrm-years are those in which ﬁrms

(1) are in violation of debt covenants (2) have been cut out of credit agreements (3)

114

mention in annual reports that they are forced to cut investment due to liquidity issues or

(4) report renegotiating debt payments.

3. 3.4 Summary statistics of ﬁnancial constraints

Our ﬁnal sample includes 394 unique ﬁrms and 2217 ﬁrm-years“. As table 3.2
indicates, by using the exact methodology outlined in KZ, the percent of ﬁrm-years that
are classiﬁed as potentially ﬁnancially constrained (PFC, LFC, and FC) increases from
14.70% to 52.28%. Not surprising then, the percent of ﬁrms classiﬁed as NFC decreases
from over half of the sample-years 54.50% to 11.50%. The lack of stability of ﬁnancial
constraint classiﬁcations between this study and KZ could be due to the different time
period examined or due to the more heterogeneous sample ﬁrms in the current study.
However, if we restrict the sample ﬁrms to be in manufacturing industries (4-digit SIC
between 2000 and 3900 inclusive) as in KZ, the distribution of constraint status is
relatively the same. As such, it is more likely that the time period studied is responsible
for the divergence in binning as opposed to the difference in industry.

Table 3.3 documents the time series properties of qualitative measures of ﬁnancial
constraint status among the sample years 1995-2004. It can been seen from table 3.3 that
as the U.S. experienced a recessionary period between 2000 and 2004 the percent of
ﬁrms classiﬁed as NFC decreased from 9.43% in 1999 to a low of 5.24% in 2003 while
the percent of ﬁrms classiﬁed as FC increased from 3.69% to 7.41% over the same time

period.

 

48 We lose 6 of the 394 ﬁrms because of missing lO-Ks or annual report to shareholders for each ﬁrm year.

115

3.4 Models of Constraints and Investment-Cash Flow Sensitivities
3. 4. 1 Correlation of qualitative measures of ﬁnancial constraints and quantitative

variables

After establishing the time series properties of ﬁnancial constraint status we now
turn our attention to the correlation between qualitative measures of ﬁnancial constraints
and quantitative accounting variables. This is a crucial area in the literature as it is
potentially impossible to read the 10-Ks and annual reports to shareholders of every ﬁrm-
year on COMPUSTAT. Thus, the point estimates taken from the correlation between
these two measures of ﬁnancial constraints from a random sample Of ﬁrms is needed in
order to apply the results to out of sample studies.

Table 4 replicates the ordered logit performed in table 9 of Lamont, Polk, and
Saa-Requejo (2001) using the data from KZ. This ordered logit model is the basis for the
construction of the “K2 Index” which is now widely used by researchers as a measure of
ﬁnancial constraints.49 In Column 1 of table 3.4 we report regression estimates from an
ordered logit framework based on the current sample and in column 2 of the table we
report the results from the current sample but restrict observations to ﬁrms in the
manufacturing industries (4-digit SIC between 2000 and 3900 inclusive). Finally, in
column 3 of the table we report the results from the “KZ Index” reported in table 9 of
Lamont, Polk, and Saa-Requejo (2001). Whether we compare the results from

speciﬁcation 1 or 2 to speciﬁcation 3 we notice stark differences between the correlations

 

‘9 An indication of the general acceptance of the “KZ Index” is found in Baker, Stein, and Wurgler (2003)
in which they indicate, “The KZ index has some very attractive features from our perspective. It is an
objective, off-the-shelf index that has already gained substantial currency as an indicator of ﬁnancial
constraints.”

116

of qualitative measures of ﬁnancial constraints and quantitative accounting variables. As
many of the accounting variables are endogenous (for example, leverage and cash
holdings) this should not come as a surprise. For instance, one could speculate either than
ﬁrms with small cash holdings are more likely to be constrained, or that ﬁrms that are
more likely face binding ﬁnancial constraints are more likely to hoard cash. For all but
one variable, (Tobin’s Q) coefficient estimates in the updated sample are opposite in sign
and insigniﬁcant. Additionally, the magnitude of the coefﬁcients has drastically reduced
over time. For example, the magnitude on the coefficient of Tobin’s Q is approximately
ten times as large (0.283 vs. 0.023) in KZ as it is in the updated sample. These results
suggest that either that (l) the correlations of ﬁnancial constraints based on qualitative
information and quantitative variables are not stable over time or (2) this approach does
not provide useful information in determining ﬁnancial constraint status. If one believes
the former, at a minimum future studies focusing on partitioning ﬁrms into ﬁnancial
constraint status should modify the factor loads on the quantitative variables used in the
partitioning process.
3. 4.2 Investment-cashﬂow sensitivities by financial constraint status

We now examine the relation between investment and cash-ﬂow as investigated
in FHP and K2. Following KZ we use the ﬁve mutually exclusive ﬁrm-year
classiﬁcations of ﬁnancial constraint status to form two broader measures of constraints
in order to investigate investment-cash ﬂow sensitivities by constraint. We replicate the
classiﬁcation used in table 5 of KZ and classify ﬁrms as Not Financially Constrained
(NFC) over the entire sample if they were classiﬁed as either NFC or Likely Not

Constrained (LNFC) in every year and classify ﬁrms as Likely Financially Constrained

117

 

(LF C) over the entire sample period if they were classiﬁed as LFC or Financially
Constrained (PC) at some point during the sample period.

In columns 1 and 2 of panel A of table 3.5 we replicate Table 9 of Lamont, Polk,
and Saa-Requejo (2001) which uses data from K2 and report the coefﬁcient estimates
from a ﬁxed effects estimation of the relation between investment, cash ﬂow, and
Tobin’s Q using the updated sample. The coefﬁcient on cash ﬂow in speciﬁcation 1 of
panel A (NFC group) is 0.208 and highly signiﬁcant (t=5.06) and the coefﬁcient on cash
ﬂow in speciﬁcation 2 (LFC group) is much smaller in magnitude, -0.0004, and
signiﬁcant (t=6.11). The nature of this result is similar in ﬂavor to the relation between
investment and cash ﬂow by constraint status as reported by K2 and reproduced in
speciﬁcation 1 and 2 Of panel B in that ﬁnancially unconstrained ﬁrms display a much
higher sensitivity of investment to cash ﬂow than constrained ﬁrms. However, there are
stark differences between the two results. First, the coefficient on the cash ﬂow variable
in the updated sample for the unconstrained group is less than 1/3 (0.208 vs. 0.680) of the
coefficient reported by KZ. Additionally, while the coefﬁcient on the constrained group
is smaller than the unconstrained groups in both samples, the coefficient in the updated
sample is virtually meaningless in economic terms though signiﬁcant. Finally, while the
coefficient on Tobin’s Q is highly signiﬁcant across both constraint groups in KZ it is
insigniﬁcant across both sub-samples using the updated sample.

In order to ascertain whether the reduction in investment-cash ﬂow sensitivities is
due to the (1) time period studied or (2) greater ﬁrm heterogeneity in the updated sample
we re-run the models on the updated sample with the restriction that sample ﬁrms be in

the same industries (4-digit SIC between 2000 and 3900 inclusive) as KZ. We report the

118

results of this restricted sample in columns 3 and 4 in panel A of table 3.5. Interestingly,
the difference in the sensitivity of investment to cash ﬂow is just as stark or starker after
restricting the models to include only manufacturing ﬁrms. The coefficient on cash ﬂow
for the restricted sample of manufacturing ﬁrms in the NFC sub-sample is 0.182 and
highly signiﬁcant (t=4.04). This is only a slight reduction (0.208 vs. 0.182) from the
sample-wide estimate reported in column 1 of panel A. The coefficient on cash ﬂow in
the constrained sample remains very small and negative (-0.010). These results indicate
that it is more likely the difference in time period studied that accounts for the drastic

reduction in investment-cash ﬂow sensitivity”.

3.5 Conclusion

By replicating the methodology of KZ on a much larger, more recent, and more
heterogeneous sample we are able to revisit the classiﬁcation of ﬁnancial constraints
based on qualitative measures, correlation between constraint status and accounting
variables, and implications of investment-cash ﬂow sensitivities. We show that the
stability of ﬁnancial constraint status is far from stable over time. Using the same
methodology, approximately 37.58% more ﬁrm-years are classiﬁed as having some
affirmative statement pertaining tO ﬁnancial constraint status (PFC, LFC, and F C ﬁrrn-
years) over the period 1995-2004 compared to 1970-1984. What naturally follows is a
reduction in classiﬁcation of ﬁrms as NFC from 54.50% to 11.50% between 1970-1984
and 1995-2004. This lack of stability of ﬁnancial constraint status over time holds even

after controlling for industry. Additionally, we show that the correlations between

 

5° Another possibility that may lead to the reduction in investment-cash ﬂow sensitivity between the two
samples is that FHP and thus KZ require ﬁrms to survive for all 14 sample years. Since surviving ﬁrms
may be more likely than others to be unconstrained, this may contribute to the higher coefﬁcient in KZ.

119

 

ﬁnancial constraint classiﬁcation and accounting variables have starkly changed over
time. Not only have the signs of the coefﬁcients of the quantitative variables ﬂipped in
predicting ﬁnancial constraint status, but four of the ﬁve are no longer signiﬁcant at
conventional levels. This either casts doubt on the use of this methodology to classify
ﬁrms as ﬁnancially constrained or, at a rrrinimum, suggests that the factor loadings used
in future studies should be updated. Finally, we report that consistent with the results of
KZ, the sensitivity of investment to cash ﬂow cannot be used as valid measure of
ﬁnancial constraint status. As in KZ, the investment-cash ﬂow sensitivity of sample
ﬁrms classiﬁed as unconstrained is much larger (0.208 vs. -0.0004) than those classiﬁed
as likely constrained. However, within the unconstrained sub-sample the sensitivity
drastically reduced over time (0.680 vs. 0.208) and within the likely constrained group
has reduced drastically and ﬂipped signs (0.340 vs. -0.0004). However, the economic
magnitude of the coefﬁcient is negligible.

These results at a minimum suggest that future work that classiﬁes ﬁrms as
ﬁnancially constrained based on factor loads from the correlations between qualitative
and quantitative measures should update the factor loads used. In the extreme, these
results cast doubt on the ability to bin ﬁrms into ﬁnancial constraint status based on
insigniﬁcant correlations between qualitative and quantitative measures. Areas of future
research investigating these correlations include examining the relation between these
qualitative classiﬁcations and other quantitative variables such as ﬁrm size or ﬁrm age.
Additionally, the creation of liberal and conservative deﬁnitions of ﬁnancial constraint

status could provide robustness checks to the original speciﬁcations.

120

Table 3.1: Sample Summary Statistics

Following Kaplan and Zingales (1997), Cap. Ex/K, is capital expenditures (data128) /
plant property and equipment (data8), Cash Flow/K is computed as earnings before
depreciation and extraordinary items (data18) + depreciation (data14) / plant property and
equipment (data8). Tobin’s Q is calculated as book assets (data6) + market equity
(data24*data25) — book equity (data60) — deferred taxes (data74) / book assets (data6).
Debt/Total Capital is calculated as long term debt (data9) + debt in current liabilities
(data34) / long term debt (data9) + debt in current liabilities (data34) + total stockholders
equity (data216). Dividends/K is calculated as preferred dividends (data19) + common
dividends (data21) / plant property and equipment (data8). Cash/K is calculated as cash
and short term investments (datal) / plant property and equipment (data8). In all
variables plant property and equipment is start of period. As is customary in the literature
and in performed in Kaplan and Zingales (1997), Tobin’s Q is calculated as start of
period. The ﬁrst number in each cell represents the median of the quantitative variable
and the second number reports the number of observations over which the median is
calculated. Column 1 reports sample-wide medians, column 2 reports medians of
manufacturing ﬁrms, and column 3 reports the medians and number of Observations
given in table 3 of KZ.

 

 

_— “in

 

(1) (2) (3)
Sample Pierce (2007) Pierce (2007) K2 (1997)
Cap. Ex./K 0.210 0.181 0.348
1786 902 719
Tobin’s Q 1.600 1.617 1.231
1961 999 719
Debt/Total Capital 0.225 0.194 0.349
1961 999 719
Dividends/Total Capital 0.000 0.000 0.000
1961 999 719
Cash/K 0.455 0.507 0.168
1961 999 719
Cash Flow/K 0.227 0.242 0.421
1961 999 719
Which Observations All Mfg All

 

121

F— _
a H-
-.

 

Table 3.2: Summary Statistics for Annual Financial Constraint Classiﬁcation

Following Kaplan and Zingales (1997) ﬁrms are assigned one of ﬁve mutually exclusive
categories (NFC, LNFC, PFC, LFC, F C). As in Kaplan and Zingales (1997) NFC ﬁrm-
years are those in which a ﬁrm either (1) initiated or increased dividends (2) repurchased
stock (3) made explicit statements in annual reports of having excess liquidity (4) had
large amounts of cash relative to investment (5) were not restricted from paying
dividends. LNF C ﬁrm-years were those in which a ﬁrm either (1) made no clear
statements of excess liquidity in annual reports (2) and lower cash to investment ratios
than ﬁrms classiﬁed as NFC. PFC ﬁrm-years are those in which a ﬁrm either (1) made
contradictory statements regarding liquidity in annual reports. LFC ﬁrm-years are those
in which ﬁrms (I) mention difficulties in raising money in annual reports (2) postpone
equity offerings due to market conditions (3) are prevented from paying dividends (4)
have little cash relative to investment. FC ﬁrm-years are those in which ﬁrms (1) are in
violation of debt covenants (2) have been cut out of credit agreements (3) mention in
annual reports that they are forced to cut investment due to liquidity issues (4)
renegotiating debt payments. The “Pet” column indicates the percentage of the total
number of Observations that fall into each mutually exclusive category and the “Number
of Obs.” coltunn indicates the raw number of sample observations in each category.

 

 

 

 

 

Number

Pet of Obs.
Panel A: Sam Wide Financial Constraint Classification
(Pierce 2007)
Total Number of Firm Years 2217
Unique F inns 394
Not Financially Constrained (NFC) 11.50% 255
Likely Not Financially Constrained (LNFC) 36.22% 803
Possibly Financially Constrained (PFC) 46.96% 1041
Likely Financially Constrained (LFC) 0.81% 18
Deﬁnitely Financially Constrained (FC) 4.51% 100
Panel B: Sample Wide Financial Constraint Classiﬁcation
(KZ 1997]
Total Number of F irm Years 719
Unique Firms 49
Not Financially Constrained (NFC) 54.50% 391
Likely Not Financially Constrained (LNFC) 30.90% 222
Possibly Financially Constrained (PFC) 7.30% 52
Likely Financially Constrained (LFC) 4.80% 35
Deﬁnitely Financially Constrained (FC) 2.60% 19

 

122

 

Table 3.3: Annual Financial Constraint Status

Financial constraint status is determined as deﬁned in Table 3.1. The top number in each
cell is the percentage of ﬁrms in each category for the respective year and the number in
parentheses in each cell is the raw number of sample observations for each given
constraint status per year.

 

Not Likely Not Possibly Likely Deﬁnitely
Financially Financially Financially Financially Financially
Constrained Constrained Constrained Constrained Constrained

 

Constraint NFC LNFC PFC LF C F C
Level
1995 27.27% 39.39% 30.30% 0.00% 3.03%
(45) (65) (50) (0) (5)
1996 13.22% 47.93% 35.95% 0.41% 2.48%
(32) (1 16) (87) (1) (6)
1997 15.87% 39.29% 42.86% 0.40% 1.59%
(40) (99) (103) (1) (4)
1998 12.20% 45.12% 37.40% 1.22% 4.07%
(30) (111) (92) (3) (10)
1999 9.43% 34.02% 50.41% 2.46% 3.69%
(23) (83) (123) (6) (9)
2000 8.03% 29.32% 58.63% 0.00% 4.02%
(20) (73) (146) (0) (10)
2001 7.69% 29.91% 55.98% 0.85% 5.56%
(18) (70) (131) (2) (13)
2002 12.68% 24.39% 54.63% 0.98% 7.32%
(26) (50) (1 12) (2) (15)
2003 5.24% 36.13% 50.26% 1.05% 7.33%
(10) (69) (96) (2) (14)
2004 5.82% 35.45% 50.79% 0.53% 7.41%
(11) (67) (96) (1) (14)

 

123

Table 3.4: Ordered Logit for Predictability of Financial Constraint Status

The dependent variable is ﬁnancial constraint status as described in Table 3.1. Standard
errors are reported in parentheses below coefficient estimates. All independent variables
are created from data items available in COMPUSTAT. Following Kaplan and Zingales
(1997), Cash Flow/K is computed as earnings before depreciation and extraordinary
items (data18) + depreciation (data14) / plant property and equipment (data8). Tobin’s Q
is calculated as book assets (data6) + market equity (data24*data25) — book equity
(data60) — deferred taxes (data74) / book assets (data6). Debt/Total Capital is calculated
as long term debt (data9) + debt in current liabilities (data34) / long term debt (data9) +
debt in current liabilities (data34) + total stockholders equity (data216). Dividends/K is
calculated as preferred dividends (data19) + common dividends (data21) / plant property
and equipment (data8). Cash/K is calculated as cash and short term investments (data1)/
plant property and equipment (data8). In all variables plant property and equipment is
start of period. As is customary in the literature and in performed in Kaplan and Zingales
(1997), Tobin’s Q is calculated as start of period.

*Signiﬁcant at the 10% level, "Signiﬁcant at the 5% level, ***Signiﬁcant at the 1%
level.

 

(1) (2) (3)
Sample Pierce (2007) Pierce (2007) KZ (1997)
Dependent Variable Fin. Const. Fin. Const. Fin. Const.
Cash F low/K 0.0001 0.0002 -1.002***
(0.0001) (0.0002) (0.234)
Tobin’s Q 0023*" 0025*" 0283*"
(0.005) (0.007) (0.078)
Debt/Total Capital 0010 -0.013 3139*"
(0.008) (0.016) (0.449)
Dividends/Total Capital 0.051 0.132* -39.368***
(0.036) (0.080) (6.097)
Cash/K 0.0003” 0.001 -1.315***
(0.0001) (0.0003) (0.289)
Which Observations All Mfg All
Number of observations 1961 999 719
Pseudo-R2 0.033 0.040 0.134

 

124

 

Table 3.5: Investment-Cash Flow Sensitivity by Financial Constraint Status

All independent variables are created from data items available in COMPUSTAT.
Standard errors are reported in parentheses below coefficient estimates. All models are
estimated using ﬁrm ﬁxed effects. Following Kaplan and Zingales (1997) the dependent
variable in each speciﬁcation, Cap. Ex/K, is Capital Expenditures (data128) / plant
property and equipment (data8). Cash Flow/K is computed as earnings before
depreciation and extraordinary items (data18) + depreciation (data14) / plant property and
equipment (data8). Tobin’s Q is calculated as book assets (data6) + market equity
(data24*data25) — book equity (data60) — deferred taxes (data74) / book assets (data6).
In all variables plant property and equipment is start of period. As is customary in the
literature and in performed in Kaplan and Zingales (1997), Tobin’s Q is calculated as
start of period. Following Kaplan and Zingales (1997) ﬁrms are classiﬁed as Not
Financially Constrained (NFC) if they are never classiﬁed as constrained (NFC or
LNFC) during the entire sample period and classiﬁed as Likely Financially Constrained
(LFC) if at some point in time during the sample period they were classiﬁed as LFC or
F C.

*Signiﬁcant at the 10% level, "Signiﬁcant at the 5% level, ***Signiﬁcant at the 1%
level.

125

Table 3.5

 

 

 

 

Panel A: Pierce {2007) (1)
Financial Constraint Status NFC
Dependent Variable Cap. Ex/K
Cash Flow/K 0208*"
(0.041)
Tobin’s Q 0.054
(0.043)
Which Observations All
Year Dummies Yes
Firm Fixed Effects Yes
Number of observations 366
R2 0.131
Panel B: Kaplan and Zingales (1)
(1997)
Financial Constraint Status NFC
Dependent Variable Cap. Ex/K
Cash Flow/K 0680*"
(0.041)
Tobin’s Q 0010*"
(0.006)
Which Observations All
Year Dummies Yes
F irm Fixed Effects Yes
Number of Observations 279
R2 0.721

(2)
LFC

Cap. Ex/K Cap. Ex/K

-0.0004***
(0.00007)

0.002
(0.001)

All
Yes
Yes
263
0.199

(2)

LFC
Cap. Ex/K

0340*"
(0.042)

0070*"
(0.018)

All
Yes
Yes
327
0.456

(3)
NFC

0182*"
_ (0.045)

0.046
(0.031)

Mfg
Yes
Yes
1 91

0.246

(4)
LFC
Cap. Ex/K

-0.010**
(0.004)

-0.00008
(0.001)

Mfg

Yes

Yes
1 00

0.203

 

126

References

Acharya, Viral, Heitor Almeida, Murillo Campello, 2005, “13 Cash Negative Debt? A
Hedging Perspective on Corporate Financial Policies,” NBER Working Paper 11391.

Agres, Stuart J. and Tony M. Dubitsky, 1996, “Changing Needs for Brands,” Journal of
Advertising Research 36, 21-30.

Alexander, Gordon J ., George Benson, and Joan M. Kampmeyer, 1984, “Investigating
the Valuation Effects of Announcements of Voluntary Corporate Selloffs,” Journal of
Finance 39-2, 503-517.

Allayannis, George and Abon Mozumdar, 2000, “Cash Flow, Investment, and Hedging,”
Working Paper, University of Virginia.

Almedia, Heitor and Murillo Campello, 2005, “Financial Constraints, Asset Tangibility,
and Corporate Investment,” Working Paper, NYU and University of Illinois.

Almeida, Heitor, Murillo Campello, Michael S. Weisbach, 2004, “The Cash Flow
Sensitivity of Cash,” Journal of Finance 59-4, 1777-1804.

Andrade, Gregor and Erik Stafford, 2004, “Investigating the Economic Role of Mergers,”
Journal of Corporate Finance 10-1, 1-36.

Andrade, Gregor, Mark Mitchell, and Erik Stafford, 2001, “New Evidence and
Perspectives on Mergers,” Journal of Economic Perspectives 15-2, 103-120.

Alti, Aydogan, 2003, “How Sensitive is Investment to Cash Flow When Financing is
Frictionless?” Journal of Finance 58-2, 707-722.

Bagwell, Kyle, 2005, “The Economic Analysis of Advertising,” to appear in Handbook
of Induatrial Organization Volume 3. Elsevier Publishers, Edited by M. Armstrong and
RH. Porter.

Baker, Malcolm, Jeremy C. Stein, and Jeffrey Wurgler, 2003, “When Does the Market
Matter? Stock Prices and the Investment of Equity-Dependent Firms,” Quarterly Journal
of Economics 118, 909-968.

Baltagi, Badi H. and Ping X. Wu, 1999, “Unequally Spaced Panel Data Regressions With
AR(1) Disturbances,” Econometric Theory 15, 814-823.

Bass, Frank M., 1969, “A Simultaneous Equation Regression Study of Advertising and
Sales of Cigarettes,” Journal of Marketing Research 6-3, 291-300.

127

Bates, Thomas W., 2005, “Asset Sales, Investment Opportunities, and the Use Of
Proceeds,” Journal of Finance 60-1, 105-135.

Bemdt, Ernst R., 1991, “Causality and Simultaneity Between Advertising and Sales,”
appearing as Chapter 8 in ER. Bemdt (1991), The Practice of Econometrics: Classic and
Contemporary. Addison-Wesley, Reading, MA.

Bernile, Gennaro, 2004, “The Information Content of Insiders' Forecasts: Analysis of the
Gains from Mergers in the 903,” University of Rochester, working paper.

Bhagat, Sanjai, Andrei Shleifer, and Robert W. Vishny, 1990, "Hostile Takeovers in the
19803: The Return to Corporate Specialization," Brookings Papers on Economic Activity:
Microeconomics, 1-72.

Blanchard, Olivier J ., Florencio Lopez-de-Silanes, and Andrei Shleifer, 1994, “What do
Firms do with Cash Windfalls?” Journal of Financial Economics 36, 337-360.

Bodnar, Gordon M. and Joseph Weintrop, 1997, “The Valuation of the Foreign Income
of U.S. Multinational F irms: A Growth Opportunities Perspective,” Journal of
Accounting and Economics 24, 69-97.

Boyer, Kenneth D. and Kent M. Lancaster, 1986, “Are There Scale Economics in
Advertising?” Journal of Business 59-3, 509-526.

Bradley, Michael, Anand Desai, and E. Han Kim, 1988, “Synergistic Gains from
Corporate Acquisitions and Their Division Between Target and Acquiring Firms,”
Journal of Financial Economics 21, 3-40.

Brown, Lawrence D., 2001, “A Temporal Analysis of Earnings Surprises: Proﬁts versus
Losses,” Journal of Accounting Research 39-2, 221-241.

Campello, Murillo, 2002, “Internal Capital Markets in Financial Conglomerates:
Evidence from Small Bank Responses to Monetary Policy,” Journal of Finance 57-6,
2773-2805.

Caves, Richard E., 1971, "International Corporations: The Industrial Economics of
Foreign Investment," Economica 38-149, 1-27.

Chang, Saeyoung, 1998, “Takeovers of Privately Held Targets, Methods of Payment, and
Bidder Returns,” Journal of Finance 53-2, 773-784.

Chevalier, Judith, 2004, “What Do We Know About Cross-subsidization? Evidence from
Merging Firms,” Advances in Economic Analysis and Policy 4-1, Article 3.

Cleary, Sean, 1999, “The Relationship between Firm Investment and Financial Status,”
Journal of Finance 54, 673-692.

128

DeAngelo, Harry, Linda DeAngelo, and Edward M. Rice, 1984, “Going Private:
Minority Freezeouts and Stockholder Wealth,” Journal of Law and Economics, 367-401.

Degeorge, Francois, J ayendu Patel, and Richard Zeckhauser, 1999, “Earnings
Management to Exceed Thresholds,” Journal of Business 72-1, 1-33.

Depken, Craig A., and Dennis Wilson, 2004, “13 Advertising a Good or a Bad? Evidence
from U.S. Magazine Subscriptions,” Journal of Business 77-82, S61-S80.

Desai, Mihir, C. Fritz Foley, and James R. Hines Jr., 2004, “A Multinational Perspective
on Capital Structure Choice and lntemal Capital Markets,” Journal of Finance 59, 2451-
2488.

Desai, Mihir A., C. Fritz Foley, and Kristin J. Forbes, 2005, “Financial Constraints and
Growth: Multinational and Local Firm Responses to Currency Depreciations,” Working
Paper, Harvard University.

Deshmukh, Sanjay and Stephen C. Vogt, 2005, “Investment, cash ﬂow, and corporate
hedging,” Journal of Corporate Finance 11, 628-644.

Dewenter, Katheryn, L., 1995, “Does the Market React Differently to Domestic and
Foreign Takeover Announcements? Evidence from the U.S. Chemical and Retail
Industries,” Journal of Financial Economics 37-3, 421-441 .

Erickson, Timothy, and Toni M. Whited, 2000, Measurement Error and the Relation
Between Investment and Q,” Journal of Political Economy 108, 1027-1057.

Fazzari, Steven M., R. Glenn Hubbard, and Bruce P. Petersen, 1988, “Financing
Constraints and Investment,” Brookings Papers on Economic Activity 1, 141-195.

F azzari, Steven M., R. Glenn Hubbard, and Bruce P. Petersen, 2000, “Investment-Cash

Flow Sensitivities are Useful: A Comment on Kaplan and Zingales,” Quarterly Journal
of Economics 115, 195-205.

Fee, C. Edward, Charles J. Hadlock, and Joshua R. Pierce, 2007, “Investment, Financing
Constraints, and lntemal Capital Markets: Evidence from the Advertising Expenditures
Of Multinational Firms.” Working paper, Michigan State University.

Froot, Kenneth, David Scharfstein, and Jeremy C. Stein, 1993, “Risk Management:
Coordinating Corporate Investment and Financing Policies,” Journal of Finance 48,
1629-1658.

Froot, Kenneth A., and Jeremy C. Stein, 1991, "Exchange Rates and Foreign Direct
Investment: An Imperfect Capital Markets Approach," Quarterly Journal of Economics
106-4,1191-1217.

129

 

1%

 

Fuller, Kathleen, Jeffry Netter, and Mike Stegemoller, 2002, “What DO Returns to
Acquiring Firms Tell Us? Evidence from Firms That Make Many Acquisitions,” Journal
of Finance 57-4, 1763-1793.

Hadlock, Charles J ., 1998, “Ownership, Liquidity, and Investment,” RAND Journal of
Economics 29, 487-508.

Hall, Bronwyn, 1988, "The Effect of Takeover Activity on Corporate Research and
Development." In Corgorate Takeovers: Causes and Congeqmences. edited by Alan J.
Auerbach, University of Chicago Press.

Harless, David W. and George E. Hoffer, 2002, “Do Women Pay More for New
Vehicles?” American Economic Review 92, 270-279.

Harris, Robert S. and David Ravenscraft, 1991, “The Role Of Acquisitions in Foregin
Direct Investment: Evidence from the U.S. Stock Market,” Journal of Finance 46-3, 825-
844.

Healy, Paul M., Krishna G. Palepu, and Richard S. Ruback, 1992, “Do Mergers Improve
Corporate Performance?,” Journal Financial Economics 31-2,135-175.

Hines, James R. Jr., 1996, “Dividends and Proﬁts: Some Unsubtle Foreign Inﬂuences,”
Journal of Finance 51-2, 661-689.

Hirschey, Mark and Jerry J. Weygandt, 1985, “Amortization Policy for Advertising and

Research and Development Expenditures,” Journal of Accounting Research 23-1, 326-
335.

Hite, Gailen L., James E. Owers, and Ronald C. Rogers, 1987, "The market for interﬁrrn

asset sales: Partial sell-offs and total liquidations," Journal of Financial Economics 18-2,
229-252.

Houston, Joel, Christopher James, and David Marcus, 1997, “Capital Market Frictions

and the Role of lntemal Capital Markets in Banking,” Journal of Financial Economics
46, 135-164.

Holthausen, Robert W., and David F. Larcker, 1996, "The Financial Performance of
Reverse Leveraged Buyouts," Journal of Financial Economics 42-3, 293-332.

Houston, J ., James, C., Ryngaert, M., 2001, “Where Do Mergers Gains Come From?
Bank Mergers From the Perspective of Insiders and Outsiders,” Journal of Financial
Economics 60-2-3, 285-331.

Houston, Joel F., and Michael D. Ryngaert, 1994, “The overall gains from large bank
mergers,” Journal of Banking and Finance 18-6, 1 155-1 176.

130

Hubbard, R. Glenn, 1998, “Capital Market Imperfections and Investment,” Journal of
Economic Literature 36, 193-227.

Iizuka, Toshiaki, 2004, “What Explains the Use of Direct-To-Consumer Advertising of
Prescription Drugs,” Journal of Industrial Economics 52-3, 349-3 79.

Jain, Prem C., 1985, “The Effect of Voluntary Sell-Off Announcements on Shareholder
Wealth,” Journal of Finance 40-1, 209-224.

Jarrell, Gregg A., James A. Brickley, and Jeffry M. Netter, 1988, “The Market for

Corporate Control: The Empirical Evidence Since 1980,” Journal of Economic
Perspectives 2-1, 49-68.

Jensen, Michael C., and Richard S. Ruback, 1983, “The Market for Corporate Control:
The Scientiﬁc Evidence,” Journal of Financial Economics 11-1-4, 5-50.

Jones, Jennifer, 1991, “Earnings Management During Import Relief Investigation,”
Journal of Accounting Research 29, 193 -228.

 

Kaplan, Steven, 1989, “The Effects of Management Buyouts on Operating Performance I3
and Value,” Journal of Financial Economics 24-2, 217-254. I

Kaplan, Steven N., and Luigi Zingales, 1997, “Do Investment-Cash Flow Sensitivities
PrOvide Useful Measures of Financial Constraints,” Quarterly Journal of Economics 112,
159-216.

Kaplan, Steven N., and Luigi Zingales, 2000, “Investment-Cash Flow Sensitivities Are
Not Valid Measures of Financial Constraints,” Quarterly Journal of Economics 115, 707-
712.

 

Lamont, Owen, 1997, “Cash Flow and Investment: Evidence from lntemal Capital
Markets,” Journal of Finance 52, 83-109.

Lamont, Owen, Christopher Polk and Jesus Sad-Requejo, 2001, "Financial Constraints
and Stock Returns," Review of Financial Studies, 14, 529—554.

Landes, Elisabeth M. and Andrew M. Rosenﬁeld, 1994, “The Durability of Advertising
Revisited,” Journal of Industrial Economics 42-3, 263-276.

Lang, Larry, Annette Poulsen, and Rene Stulz, 1995, “Asset Sales, Firm Performance,
and the Agency Costs of Managerial Discretion,” Journal of Financial Economics 37-1,
3-3 7.

Lang, Larry H.P., Rene M. Stulz, and Ralph A. Walking, 1991, “A Test of the Free Cash

Flow Hypothesis: The Case of Bidder Returns,” Journal of Financial Economics 32-1,
45-60.

131

Lang, Larry H.P., Rene M. Stulz, and Ralph A. Walking, 1989, “Managerial
Performance, Tobin’s Q, and the Gains from Successful Tender Offers,” Journal of
Financial Economics 24-1, 137-154.

Lee, Bong 800 and Jungwon Suh, 2004, “Exchange Rate Inelasticity of Foreign Income
of U.S. Multinationals: An Exploration of Why Their Exchange Rate Exposure is not
Signiﬁcant,” Working paper, University Of Houston.

Lehn, Kenneth and Annette Poulsen, 1989, “Free Cash Flow and Stockholder Gains in
Going Private Transactions,” Journal of Finance 44-3, 771-787.

Maksimovic, Vojislav and Gordon Phillips, 2001, “The Market for Corporate Assets:
Who Engages in Mergers and Asset Sales and Are There Efficiency Gains?” Journal of
Finance 56-6, 2019-2065.

Malmendier, Ulrike and Geoffrey Tate, 2005 , “CEO Overconﬁdence and Corporate
Investment,” Journal of Finance 60-6, 2661-2700.

McCabe, GM. and KC. Yook, 1996, "The Effect of International Acquisitions on
Shareholders’ wealth,” Mid-Atlantic Journal of Business 32, 5—17.

McGuckin, Robert H. and Sang V. Nguyen, 1995, “On Productivity and Plant Ownership
Change: New Evidence form the Longitudinal Research Database,” RAND Journal of
Economics 26-2, 257-276.

Moeller, Sara B. and Frederik P. Schlingemann, 2005, “Global Diversiﬁcation and
Bidder Gains: A Comparison between Cross-border and Domestic acquisitions,” Journal
of Banking and Finance 29-3, 533-564.

Moeller, Sara, B., Frederik P. Schlingemann, and Rene M. Stulz, 2005, “Wealth
Destruction on a Massive Scale? A Study of Acquiring-Firm Returns in the Recent
Merger Wave,” Journal of Finance 60-2, 757-782.

Morck, Randall, Andrei Shleifer, and Robert W. Vishny, 1990, “Do Managerial
Objectives Drive Bad Acquisitions?” Journal of Finance 45-1, 31-48.

Moyen, Nathalie, 2004, “Investment-Cash Flow Sensitivities: Constrained versus
Unconstrained Firms,” Journal of Finance 59-4, 2061-2092.

Peek, Joe and Eric S. Rosengren, 1997, “The International Transmission of Financial
Shocks: The Case of Japan,” American Economic Review 87-4, 495-505.

Peles, Yoram, 1971, “Rates of Amortization of Advertising Expenditures,” Journal of
Political Economy 79-5, 1032-1058.

132

 

Pesendorfer, Martin, 2003, "Horizontal Mergers in the Paper Industry," RAND Journal of
Economics 34-3, 495-515.

Rauh, Joshua D., 2006, Investment and Financing Constraints: Evidence from the
Funding of Corporate Pension Plans, Journal of Finance 61-1, 33-72.

Ross, Stephen A., 1983, “Accounting and Economics,” The Accounting Review 58-2,
375-380.

Rossi, Stefano and Paolo F. Volpin, 2004, “Cross-country determinants of mergers and
acquisitions,” Journal of Financial Economics 74-2, 277-304.

Schmalensee, Richard, 1976, “Advertising and Proﬁtability: Further Implications of the
Null Hypothesis,” Journal of Industrial Economics 25, 45-54.

Schoar, Antoinette, 2002, “Effects of Corporate Diversiﬁcation on Productivity,” Journal
of Finance 57-6, 2379-2403.

Scholes, Myron S., and Mark A. Wolfson, 1990, "The Effects of Changes in Tax Laws on
Corporate Reorganization Activity," Journal of Business 63-1, Sl41-Sl64.

Schmalensee, Richard, 1976, “Advertising and Proﬁtability” Further Implications of the
Null Hypothesis,” Journal of Industrial Economics 25, 45-54.

Shin, Hyun-Han and René Stulz, 1998, “Are lntemal Capital Markets Efﬁcient,”
Quarterly Journal of Economics 113, 531-552.

Stein, Jeremy C., 2003, “Corporate Investment Policy,” appearing as Chapter 3 in the
Handbook of the Economics of Finance, Volume 1A, Elsevier North Holland:
Amsterdam. Edited by GM. Constantinides, M. Harris, and RM. Stulz.

Teoh, Siew Hong, Ivo Welch, and T. J. Wong, 1998, “Earnings management and the
long-run market performance of Initial Public Offerings,” Journal of Finance 53-6, 1935-
1974.

Whited, Toni, 1992, “Debt, Liquidity Constraints, and Corporate Investment: Evidence
from Panel Data,” Journal of Finance 47, 1425-1460.

Wooldridge, Jeffrey M., 2002, Econometric Analysis of Cross Section and Panel Data
(MIT Press, Cambridge, MA).

133

 

   

l11111111111111l