P

IIILQIO

2' Illllllllm

3 1293 016885

                                                                                                   

This is to certify that the

dissertation entitled

The Impact of Firm and Market Specific
Characteristics on Product Market
Competition: An Empirical Analysis of the
Discount Department Store Industry

presented by

Sheri Teresa Tice

has been accepted towards fulfillment
of the requirements for

Ph 0 D 0 degree in Finance

 

 

NWW aw»;

Major professor

Date AUEUSt 17, 1998

 

MSU is an Affirmative Action/Equal Opportunity Institution 0-12771

 

l

 

LIBRARY

Michigan State

Unlverslty

 

 

PLACE IN RETURN BOX
to remove this checkout from your record.
TO AVOID FINE return on or before date due.

 

 

out DUE

MTE DUE

DATE DUE

 

W

 

APR 1 0 2005

néogflfi

 

 

 

 

 

 

 

 

 

 

 

 

 

ma WM“

THE IMPACT OF FIRM AND MARKET SPECIFIC CHARACTERISTICS ON
PRODUCT MARKET COMPETITION: AN EMPIRICAL ANALYSIS OF THE
DISCOUNT DEPARTMENT STORE INDUSTRY

By

Sheri Teresa Tice

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Finance

1998

ABSTRACT

THE IMPACT OF FIRM AND MARKET SPECIFIC CHARACTERISTICS ON
PRODUCT MARKET COMPETITION: AN EMPIRICAL ANALYSIS OF THE
DISCOUNT DEPARTMENT STORE INDUSTRY

By

Sheri Teresa Tice

Recent empirical research has documented that leverage
impacts firm aggressiveness in the product market. I extend this
research by studying how managerial ownership as well as other
firm and market specific characteristics affect firm aggressiveness

in the product market.

The discount department store industry is an excellent
industry to use due to a lot of heterogeneity in firm specific
characteristics, essentially homogeneous merchandise, and a large
number of firms in the industry. Because Wal-Mart entered many
local markets, expanded through connected regions, and retained
their cost advantage over the period, their entry into local markets
can be seen as exogenous. Thus, the discount department store
industry provides a natural experiment to study which firm and
market specific characteristics determine how incumbent firms
respond to a new threat, the entry of Wal-Mart, at the local market

level.

In a study of the expansion responses of incumbent firms in
the discount department store industry to new entry by Wal-Mart
into local markets, I find that private firms are less likely to expand
then public firms. I also find that public firms with high managerial
ownership are less likely to expand then public firms with low
managerial ownership. Other results are that a higher level of debt
decreases the likelihood of expansion, a higher level of focus
increases the likelihood of expansion, and the more dependent a
firm is on the market under attack, the greater the likelihood of
expansion.

I also find that Wal-Mart has a harder time gaining market
share in local markets with a high fraction of stores owned by firms
with high levels of managerial ownership. Wal-Mart also has a
harder time gaining market share in local markets with stores owned
by firms with low debt, high focus, or high profitability. Wal-Mart
also has a harder time gaining market share in low Herfindahl
markets. There is evidence that stores in higher Herfindahl markets
are more profitable which supports co-operation within these

markets and/or high switching costs in lower Herfindahl markets.

Copyright by
SI-IERI TERESA TICE
1998

I dedicate this dissertation and all the time and effort put
forth to my parents, my husband, Barry, and my children, Brock and
Crystal. My parent’s encouragement helped to continue when I felt
discouraged. My husband and children supported me in this
endeavor, even when it required sacrifice on their part. They also

helped me to keep the important things in perspective.

ACKNOWLEDGMENTS

I would like to express extreme gratitude to my dissertation
chairman, Dr. Naveen Khanna. His optimism, guidance, energy,
patience as well as the time he invested helping me are very
appreciated. I also want to thank him for the knowledge I gained
from him in the seminar I took from him as well as during the
dissertation process. I cannot imagine having a better mentor.

I also want to thank the other members of my dissertation
committee, Dr. Assem Safieddine, Dr. Peter Schmidt, and Dr.
Richard Simonds, for their insightful and helpful comments
throughout the dissertation process.

In addition, I want to thank the other faculty members in the
finance department particularly those who instructed the doctoral
seminars, or invested their time in the doctoral program in other
ways. I also want to thank Dr. Richard Simonds for accepting me
into the doctoral program while he was chairman.

I also want to thank my fellow doctoral students for their help

during my stay in the program.

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................... ix
LIST OF FIGURES .......................................................................................... x
CHAPTER 1 Introduction .................................................................... l
CHAPTERZ History of the Discount Department Store Industry ........ 4
2.1 Definition of the Industry .............................................. 4
2.2 Firm Locations and Movements ...................................... 6
2.3 Summary ........................................................................ 8

CHAPTER3 The Effect of High Managerial Ownership, Focus
and Capital Structure On Firm Aggressiveness in

the ProductMarket .......................................................... 9
3.1 Introduction .................................................................... 9
3.2 Research Design ............................................................ 13
3.3 Data ............................................................................... 17
3.4 Empirical Results ........................................................... 21
3.4.1 Explanatory Variables ........................................ 21

3.4.2 Significance of Managerial Ownership:Private
Versus Public Firms ............................................. 24
3.4.3 Determinants of Private Firm Expansion .............. 34

3.4.4 The Significance of Managerial Ownership:
Public Firms Only ................................................ 36

3.5 Robustness Checks ........................................................ 42

3.5.1 Operating Profit Margin Instead of Discount

Sales per Square Foot ........................................... 42
3.5.2 Parent Firm Size ................................................ 42
3.5.3 Unmodeled Market Heterogeneity ........................ 44

3.5.4 The Importance of Managerial Ownership
Controlling for Firm Size and Economic

Growth ............................................................ 49
3.6 Summary ......................................................................... 50
CHAPTER4 Incumbent and Market Specific Determinants of an

Entrant’s Penetration Into New Local Markets .............. 51

4.1 Introduction ................................................................... 51

4.2 Research Design ............................................................ 53

4.3 Data ............................................................................... 57

4.4 Empirical Results ........................................................... 61

4.4.1 New Entrant Penetration: The Importance of
Incumbent Ownership ....................................... .61

4.4.2 An Indirect Test for Managerial
Entrenchment ................................................... .70

4.4.3 New Entrant Penetration: Markets With Only
Public Incumbent Firms ...................................... 73

4.4.4 Incumbent Penetration: Markets With Both

Public and Private Firms ..................................... 87

4.4.5 Modified Herfindahls and Profitability ............... 92

4.5 Summary ........................................................................ 95
CHAPTERS Summary ......................................................................... 97
APPENDDKA Tables ........................................................................... 99
APPENDIXB Figures ......................................................................... 136
LIST OF REFERENCES .................................................................................. 149

Table 1
Table 2

Table 3

Table 4

Table 5

Table 6
Table 7
Table 8

Table 9

Table 10

Table 11

Table 12

Table 13

Table 14

Table 15

LIST OF TABLES

Chains Attacked by Wal-Mart ....................................... 99

Summary Statistics: Public and Private
Incumbent Firms ........................................................... 100

Simple Correlation Matrix: All Observations
Public and Private ........................................................ 101

Univariate Probit: Public and Private
Incumbent Firms ........................................................... 102

Univariate Probit: Public and Private

Incumbent Firms Time Dummies Included ................... 103
Summary Statistics: Private Incumbent Firms .............. 104
Univariate Probit: Private Incumbent Firms ................. 105
Summary Statistics: Public Incumbent Firms ............... 106

Simple Correlation Matrix: Public Firm
Observations ................................................................. 107

Univariate Probit: Public Incumbent Firms .................. 108

Univariate Probit: Public Incumbent Firms
Time Dummies Included ............................................... 109

Univariate Probit: Public Incumbent Firms
Using Operating Profit Margin .................................... 110

Univariate Probit: Public Incumbent Firms
Parent Size Added ........................................................ 111

Univariate Probit: Public and Private Incumbents
Regional Dummies ........................................................ 112

Univariate Probit: Public Incumbents
Regional Dummies ........................................................ 113

ix

Table 16

Table 17

Table 18

Table 19

Table 20

Table 21

Table 22

Table 23

Table 24

Table 25

Table 26

Table 27

Table 28

Univariate Probit: Public and Private Incumbents
State Population Changes ............................................. 114

Univariate Probit: Public Incumbents
State Population Changes ............................................. 115

Univariate Probit: Public Incumbents
State Population Changes Herfindahl Instead of
Market Share ................................................................. 116

Univariate Probit: The Importance of Ownership
Controlling for Firm Size, and Economic Growth ......... 117

OLS Regressions Explaining Wal-Mart’s Market
Share as a Function of Ownership in a Local
Market ........................................................................... 118

Summary Statistics: All Markets ................................... 119

Summary Statistics: Markets With All Public
Incumbents .................................................................... 120

OLS Regressions Explaining Wal-Mart’s Market
Share as a Function of Ownership in a Local Market:
Market and Firm Size Controls ..................................... 121

Constrained Least Squares Spline Regressions

Testing for Differences in Slope Coefficients for
Different Ownership Groups: Market and Firm

Size Controls Added ..................................................... 122

OLS Spline Regression: Explaining Wal-Mart’s
Market Share as a Function of Different Ownership
Groups: Market and Firm Size Controls Added ............ 123

Simple Correlation Matrix for Independent
Variables Using Public Incumbent Markets .................. 124

Summary Statistics: Markets With All Public
Incumbents ................................................................... 125

OLS Regressions Explaining Wal-Mart’s Market
Share at Year +2: Markets With Only Public Firms:
Dependent Variable is Penj ......................................... 126

Table 29

Table 30

Table 31

Table 32

Table 33

Table 34

Table 35

Table 36

Table 37

Summary Statistics: Markets With All
Incumbents .................................................................. 127

OLS Regressions Explaining Wal-Mart’s
Market Share at Year +4: Markets With Only
Public Firms ................................................................. 128

OLS Regressions Explaining Wal-Mart’s
Market Share in Year +2 as a Fraction of Stores
In Market in Year -1: Markets With Only

Public Firms ................................................................. 129
Simple Correlation Matrix for Independent

Variables Using All Local Markets ............................... 130
Summary Statistics: All Local Markets ......................... 131

OLS Regressions Explaining Wal-Mart’s
Market Share: All Markets No Debt Ratio, No

Focus Variable .............................................................. 132
OLS Regression Explaining Wal-Mart’s

Market Share: All Markets. No Debt Ratio ................... 133
Summary Statistics ........................................................ 134

OLS Regressions Explaining State Profitability
And Herfindahls ............................................................ 135

Figure 1

Figure 2

Figure 3

Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11
Figure 12
Figure 13

Figure 14

LIST OF FIGURES

Standard Discount Department Store Salesper

Person (U.S.) .................................................................... 136
Fraction of 1975 Industry Sales By Top Five

Chains ........................................................................... 137
Fraction of 1995 Industry Sales By Top Five

Chains ........................................................................... 137
Top Ten Firms Based on Sales ...................................... 138
Wal-Mart 1975 Store Locations .................................... 139
Wal-Mart 1985 Store Locations .................................... 140
Wal-Mart 1990 Store Locations .................................... 141
Wal-Mart 1996 Store Locations .................................... 142
Wal-Mart and Sky City 1981 Store Locations ............... 143
Wal-Mart and Sky City 1990 Store Locations ............... 144
Wal-Mart and TG&Y Store Locations 1977 .................. 145
Wal-Mart and TG&Y Store Locations 1986 ................... 146
Wal-Mart and Duckwall 1975 Store Locations .............. 147
Wal-Mart and Duckwall 1996 Store ocations ................ 148

xii

CHAPTER]

Introduction

Recently, financial economists have started to examine the
relationships between firm characteristics and product market
competition. In the capital structure area, one group of theoretical
models predicts that increases in financial leverage “softens”
product market competition while another group of theoretical
models predicts that increases in financial leverage “toughens”
product market competition. In the incentive contract area, a group
of theoretical papers hypothesize that incentive contracts can
influence the level of managerial aggressiveness in the product
market and entry decisions.

On the empirical side, a couple of recent papers have
documented a link between changes in a firm’s capital structure and
subsequent product market competition. Nonetheless, there are
many unanswered questions regarding the impact of firm specific
characteristics such as managerial ownership, capital structure and
diversification on product market competition. There are also
unanswered questions regarding the impact of market specific

characteristics on product market competition. For example, it is

2

easier or harder for a new entrant to gain market share in a high
Herfindahl market versus a low Herfindahl market? High Herfindahl
markets may provide profit opportunities, however, the firms in
these markets may have developed effective barriers to entry and
expansion. This dissertation is an empirical attempt to broaden our
understanding in these areas.

I use the discount department store industry as a laboratory to
examine the impact of firm and market specific characteristics on
product market competition. This industry provides a natural
experiment for several reasons. First, there are a large number of
firms with heterogeneous characteristics. Second, these firms
compete locally. Third, there is essentially a homogeneous product.
Fourth, Wal-Mart expanded dramatically throughout the United
States presenting a severe competitive threat to the many regional
chains in the industry. This allows me to examine how firms
respond to a virtually identical threat holding industry constant.

In Chapter 2 the industry that is used to conduct this analysis,
the discount department store industry, is defined and described.
Maps showing store locations are used to demonstrate the regional
nature of many of the firms as well as Wal-Mart’s dramatic
expansion. The maps are also used to show differences in firm
responses to Wal-Mart entering their local markets.

Chapter 3 examines the firm and market specific determinants

3

of incumbent firms’ decisions to expand or not when Wal-Mart
enters their local markets. The fundamental question addressed in
this chapter is: Do different firm and market specific characteristics
lead to different investment decisions in the product market when
incumbents face the same threat?

Chapter 4 examines the firm and market specific determinants
of Wal-Mart’s market share gains after entering local markets. The
fundamental question addressed in this chapter is: Which firm and
market specific characteristics does Wal-Mart find leads to easier
market penetration? An indirect test of agency theory and the
entrenchment hypothesis is also done to examine the relationship
between managerial ownership and Wal-Mart’s ability to gain
market share after entering a market. An examination of the link

between market concentration and profitability is also done.

CHAPTERZ

History of the Discount Department Store Industry

2.1 Definition of the Industry

The discount department store industry was first defined by a
trade journal called Discount Merchandiser in 1961. They defined a
discount department store as “a departmentalized retail
establishment utilizing many self-service techniques to sell hard
goods, health-and-beauty aids, apparel and other soft goods and
other general merchandise. It operates at uniquely low margins. It
has a minimum annual volume of $1,000,000 ($500,000 in the
1970’s) and is at least 10,000 square feet in size.” The discount
department store industry grew rapidly. They took sales away from
department stores as they were a cheaper alternative for many items.
They also took sales away from variety stores as they provided a
larger product assortment. More recently, they are adding food
selections and are attempting to take sales away from grocery
stores. Figure 1 demonstrates this by showing standard discount
department store sales per person in the United States.

A list was compiled of all of the discount store chains that

Wm had on the “Leading Discounters” list in at

5
least one year during the time period of 1975 - 1996. To be on the

“Leading Discounters” list they had to have sales of at least $100
million (850 million in the 1970’s). This procedure eliminated the
very small “mom and pop” type chains which have very little public
information available.l A trade journal called W
Discoooi Department Storos was then used to determine store
locations for all of these chains for each year in which they
operated in the industry during the period of 1975 - 1996.2 Thus, a
discount department store chain had to be on the “Leading

Discounters” list in Qiscouot Mcrchondisor and located in Ihe
Dirootory of Disoount Deoartmeot Stores to be included in the

sample.

At the beginning of 1975, there were 105 firms with at least
one discount department store chain. By the beginning of 1996,
there were 34 firms with at least one discount department store
chain. Not only were there fewer players in 1996, but the stores and
sales were concentrated in the largest firms. Figures 2 & 3 show
the fraction of discount department store sales achieved by the top
five firms in the industry ranked by sales in 1975 and 1995

respectively. It is evident that the larger firms are capturing a

 

‘memnmhmemmmmmwmmfle Onlystandrd
discount depttment stores were used. No manbership/warehmse clubs or hyper-markets were included in
thesample.
2Iftwoormorechainsvrmeovvnedlavyfliesumefirmatthesnmetime,theywerecombinedintooneehain.
'Ihenetresultwasll3diains.

6

much larger fraction of industry sales in the United States in 1995

than in 1975.

In addition to the concentration increase that occurred in this
industry, there was also a change in the largest players. Figure 4
shows the ten largest firms in the sample, based on sales, in 1975,
1985, and 1995. As can be seen, there was a lot of change in the

names of the firms that are ranked in the top ten based on sales.

2.2 Firm Locations and Movements

Many of the discount department store chains in the industry
were regionally located. In 1975, Kmart was the only true national
chain. In 1996, Kmart, Wal-Mart and to a more limited extent
Target and Marshalls could be considered national chains. The
chain that experienced the most dramatic growth during the period
1975 to 1996 was Wal-Mart. Wal-Mart grew from a regional chain
to the largest retailer in the United States during the period.
Figures 5,6,7 & 8 show Wal-Mart store locations at the beginning of

1975, 1985, 1990 a 1996 respectively.’

It is interesting to examine the behavior of incumbent chains
when Wal-Mart entered their local markets. Figure 9 shows store
locations for Sky City and Wal-Mart at the beginning of 1981. As
can be seen, Sky City was a regional discount department store

chain. Wal-Mart entered one of Sky City’s markets between the

7
beginning of 1980 and 1981. By the beginning of 1990 Wal-Mart

had blanketed Sky City’s markets with stores. Store locations at the
beginning of 1990 for Wal-Mart and Sky City are shown in Figure
10. Sky City, unlike Wal-Mart, did not change store locations much
during the period of 1981 — 1990. In 1990, Sky City exited the

industry.

Figure 11 shows store locations for TG&Y and Wal-Mart at the
beginning of 1977. Wal-Mart and TG&Y started competing between
the beginning of 1976 and 1977 when TG&Y entered the industry by
converting variety stores to discount department stores. By the
beginning of 1986, Wal-Mart was competing in many of TG&Y’s
markets. Figure 12 shows store locations at the beginning of 1986.
TG&Y, unlike Sky City did expand against Wal-Mart. However, in

1986, TG&Y also exited the discount department store industry.

Duckwall was already competing with Wal-Mart in a couple of
markets at the beginning of 1975. Figure 13 shows store locations
at the beginning of 1975 for Duckwall and Wal-Mart. Unlike the
other two chains, Duckwall has been able to survive in the industry
and still exists today. Duckwall primarily expanded away from
Wal-Mart by going West in the late 1970’s and 1980’s. Store
locations for both chains at the beginning of 1996 are shown in
Figure 14. When looking at Figure 14, it is evident that Wal-Mart

has not penetrated some of the markets that Duckwall is located in

 

ThemapsshowallofthesdigitzipcodesmwhichWal-Mm'thadatleast1staeindiegivenyear.

8

to the same extent that they penetrated some of the markets that the
other two competitors were located in.

The rest of the dissertation examines both sides of the
strategic interaction. First, which firm and market specific
characteristics determine incumbents’ expansion decisions as a
response to Wal-Mart’s entry into their local markets? Second,
which markets does Wal-Mart perceive are weaker and penetrate

quicker?

2.3 Summary

In this chapter, the discount department store industry was
defined. Over the past twenty years there has been consolidation
both in the form of fewer players, as well as a change in the largest
firms within the industry. Most of the firms in the industry over the
past twenty years have been regionally located. Firms exhibit
different responses when they are attacked by Wal-Mart at the local
market level. It appears that some responses may be better at

curtailing Wal-Mart’s penetration into a firm’s local markets.

CHAPTER3
The Effect of High Managerial Ownership, Focus and Capital

Structure On Firm Aggressiveness in the Product Market

3.1 Introduction

Financial economists have recently started researching the
linkages between firm characteristics and product market
competition. Recent empirical research has investigated how
changes in capital structure affect firm behavior in the product
market.4 This chapter of my dissertation builds on previous
research by examining the impact of ownership structure as well as
capital structure and other firm and market specific factors as
determinants of firm responses when facing a severe new
competitive threat.

Fershtman and Judd (1987) show that incentive contracts
affect the level of managerial aggressiveness in the product market.
They show that in oligopolistic situations, owners can make
managers produce more (than in conventional oligopoly models) by
giving them contracts that put weight not only on profits but also on

sales. This finding is especially meaningful in industries with

 

‘ For ample, see Phillips (1995), Chevalier (1995), and Opler and TM (1994).

9

10
higher strategic substitutability.’ Existing papers that attempt to

empirically document the relationship between incentive contracts
and firm aggressiveness do so indirectly. Kedia (1997) and
Aggarwal and Samwick (1997) test to see whether firms whose
outputs are strategic substitutes provide contracts with low
managerial ownership.6 Both of these papers document that as
suggested by theory, the weight in the contract on firm profits
reduces with the extent of substitutability.

In this paper I test the relationship between managerial
ownership and firm aggressiveness directly. I look at the actual
expansion decisions of competing firms, with different levels of
managerial ownership, in response to a common threat. If
managerial ownership is an important determinant of firm
aggressiveness, then we should be able to observe that firms with
higher ownership are less aggressive then competitors with lower
ownership when facing the same threat.

I also examine the role other firm specific characteristics play
in shaping a firm’s behavior in the product market. A firm’s capital
structure, degree of focus, profitability, size, market share, and
dependence on the market may also affect their response to a new

competitive threat.

 

5Reitman(l993)extendsthisreeeardibyshowingdlatstockopfionscanbeusedtocmtrolthedegreeof
aggression.
‘Theimplicatimismatlowa'weightmownashipmflatesmmhighaweightmodlava‘iables,
possl‘blesalesrewedasmggestedbyFershtmanandJudd(l987).

11

The data set that I use is from the discount department store
industry. From 1975-1996 Wal-Mart grew from a small regional
discount department store chain to being the largest retailer in the
U.S. During their expansion across the U.S. they competed with 69
different discount department store chains. Because discount
department stores carry essentially homogeneous merchandise yet
are surprisingly heterogeneous in firm level characteristics, this
data set provided me the opportunity to study the impact of
managerial ownership and other firm specific characteristics on
incumbent firm decisions keeping the industry, product and threat
relatively constant across all observations. Also, since Wal-Mart
expanded across the U.S. through connected regions, their entry into
a market can be seen as exogenous. This avoids some of the
endogeneity problems faced by other studies in this area that
attempt to see if one firm decision causes another firm decision. For
example, Chevalier (1995) examines whether firms undertaking a
LBO are weaker in the subsequent period than their less levered

rivals.7

First I look at extreme levels of ownership by comparing
expansion decisions of privately held and publicly traded firms as a
response to new entry by Wal-Mart. I find that private firms are

less likely to expand in response to the new threat than public firms.

 

7Also,seePhillips(1995)andKovenockandPhillips(1997). Chevalia'uiestoconu'olforpossible
mdogmeityoftheLBOdecisionbydoingmerohlsmwchecks. KovulockandPhillipsuseanvo-stage

12

However, these two organizational forms differ on other dimensions
like access to capital, and inherent differences on the amount of
information that is available to the market. To control for this, I run
similar tests on the sub-sample of public firms. This allows me to
check whether it is managerial ownership that is driving the results
for the private firms. I find that public firms with higher levels of
managerial ownership are less likely to expand then public firms
with lower levels of managerial ownership. This suggests that high
ownership is making both public and private firms less aggressive in
the product market when facing a new competitive threat. Other
significant results are that firms with lower amounts of debt in their
capital structure are more likely to expand as are more focused
firms. Firms with a high market share in a market, as well as firms
that are highly dependent on a market are also more likely to expand
in that market. The more profitable firms are also more likely to
expand as a response to new entry by Wal-Mart into their markets.
These results are robust to changes in the profitability measure, the

addition of regional dummies, and the addition of time dummies.

The remainder of this chapter is organized as follows: Section
3.2 discusses the research design. Section 3.3 describes the data.
Section 3.4 presents the methodology and empirical results. Section

3.5 presents robustness checks, and Section 3.6 is the summary.

 

regressimwcmuolfamispoblan.

13
3.2 Research Design

During the period of 1975-1996 Wal-Mart transformed itself
from a regional chain consisting of 125 stores in eight states to a
national chain consisting of 2,234 stores in all fifty states.8 Figures
5 through 8 show Wal-Mart store locations at the beginning of 1975,
1935, 1990 and 1996 respectively.’ When looking at the maps it is
apparent that Wal-Mart expanded throughout the U.S., and that they
expanded through connected regions.

Competition between discount department stores takes place at
the local level. Following Chevalier (1995) the United States is
broken down into local markets. The local markets used in this
study are 862 3-digit zip code areas. R n M N 11 Z'

Eiooor defines the five digits of the zip code. The first digit of the
zip code identifies the geographic region of the country. The
second digit identifies a portion of the geographic region. The third
digit identifies a sectional center or multi-coded city within that
portion of the geographic region. A sectional center is usually the
natural center of local transportation and serves the smaller post
offices surrounding it. A multi-coded city is a main city post office
that serves post office branches within a city. The fourth and fifth

digits in the zip code identify the individual post office branches

 

'Numberofstoredataisfi'omDiscotherdimdisa. Loeationsofstoresbystatesisfi‘om'IheDh‘ectg

of Discount Meat Stores.
Themapsshowallofthes digitzipcodesinwhich Wal-mhadatleastonestoreindlegivalyear.
The store locations came from The Directgy of Discmmt Meat Stores.

14

that are served by the sectional center or multi-coded city office.
Because the sectional centers and multi-coded cities are usually the
natural center of local transportation, the three-digit zip code area

served by them is felt to be a reasonable economic market.

In 1975 Wal-Mart had discount department stores in 45 of the
862 local markets. By the beginning of 1996, Wal-Mart had entered
another 655 local markets for a total of 700 markets.10 During the
period of 1975-1996, 69 different discount department store chains
owned by 76 different firms competed against Wal-Mart in the
United States at the local market level.ll A couple of these were
national chains, while the rest were regional. These chains were
surprisingly heterogeneous in characteristics at the firm level. The
local markets differed not only on which of the chains were
competing but also on total number of stores. While some markets
were being served by only one chain, in other markets competition

between 6-7 different chains was not uncommon.

One of the problems well recognized in the literature is that
firm specific characteristics may be caused by the same exogenous
factors that determine investment decisions. Hence, a major
concern in developing the research design of the paper is to
minimize this endogeneity problem. I adopt the methodology used

in Opler and Titman (1994). Instead of trying to establish whether

 

'° In calendar 1996 and 1997, Wal-Mart continued to expmld in the United States.

15

one decision (say a LBO) causes another decision (say investment),
firm specific characteristics are measured prior to an external
threat. Subsequent investment decisions are then examined to see if
any of these characteristics determined the subsequent investment
decisions.12 In this paper, the threat is new entry by Wal-Mart into
a local market where some other firms in the discount department
store industry are incumbents. The investment decision is expansion
or no expansion by each incumbent firm i in the local market j when
first invaded by Wal-Mart.

A time line illustrating the empirical set-up is shown below.

 

 

Firm and
mkt specific Wal-Mart
variables measured enters mkt j
L i 1 1
-1 0 +1 +2 Year
< D

Incumbent firm i’s expansion decision measured fiom Year -1 to Year +2

Year 0 is defined as the beginning of the year in which Wal-Mart
first has a store in 3-digit zip code market j. The firm and market
specific variables are measured at the beginning of Year —1 to avoid

spurious correlation between the variables and the subsequent

 

“LBO’s,spin-ofl‘s,andgoingpublicwerenot cmmtedasfirmdlmgeshaemlyacquisitionswere
counted.
12AshIOpla'andTitrnlm(1994),!1113implicitlyassmnesthatexantefirmspecificcharacteristicsare
exogenous.

l6

expansion decision. If they are flow variables (from the income
statement), they are measured over the fiscal year ending at the
beginning of Year -1.‘3 If they are non-flow variables (balance
sheet items and market specific variables), the values are measured
at the end of the fiscal year which is at the beginning of Year -1.“
The economic activity variable and incumbent expansion decisions
are examined from Year

-1 to Year +2.

This data set has a number of advantages. First, given that
most of the local markets experienced entry by Wal-Mart during the
period examined I am able to generate a large number of
observations where incumbent firms face an exogenous threat.
Second, I am able to keep the threat constant, since I only study the
response of incumbents to new entry by Wal-Mart into their local
markets.” Third, discount department stores carry essentially non-
differentiated merchandise and are thus homogeneous in that
respect. This reduces problems of differential product quality
across firms. Also, this data set provides me with a reliable measure
of profitability “sales per square foot” for both public and private
chains. This enables me to include privately owned firms in many

of the tests. Since this data set permits us to keep industry, product

 

13Inthisindustry, ahnostalloftheretailershadafiscnlyearendinghnuary3l ofthegivenyearoverthe
timeperiodexamined.

l"Iheexceptionha'eisLBO’s. lfafirmmdamtaLBOintheyearspreceedingYeu—l,dlmitwas
treatedasaLBOfirminYec—l.

l7

and threat constant across all observations, I can develop reasonably
clean tests on how different firm and market characteristics affect
incumbent firms’ responses to Wal-Mart’s entry.

Normally, you might not expect firms to expand in response to
new entry if the local markets are in equilibrium. This industry,
however was growing due to the fact that the industry was taking
sales away from other retailers. The question being addressed is
thus, “Which chains are trying to grab part of the growing market in

the face of Wal-Mart’s entry?”

3.3 Data

A trade journal called Discount Merohondiser was used to
identify major discount chains during the period of 1975 - 1996.
They define a discount department store as “a departmentalized
retail establishment utilizing many self-service techniques to sell
hard goods, health-and-beauty aids, apparel and other soft goods
and other general merchandise. It operates at uniquely low margins,
has a minimum annual volume of $1 million ($500,000 in the
1970’s), and has at least 10,000 square feet of total space.” A list
was compiled of all of the discount store chains that Disoount
Mmhonojm had on the “Leading Discounters” list in at least one

year during the time period of 1975 - 1996. To be on the “Leading

 

‘5hembeugmdthnannaflaWal-Mmmecfiayearsposedadifl’eandlreatdlmahrgaWal-Mtt
lateron. [canolfm'fllisdlrmghtimemdregiondmnmies

18

Discounters” list they had to have sales of at least $100 million
($50 million in the 1970’s). This procedure eliminated the very
small “mom and pop” type chains which have very little public
information available."

A trade journal called MW
Sims was then used to determine store locations for all of these
chains for each year in which they operated in the industry during
the period of 1975 — 1996.'7 Thus, a discount department store

chain had to be on the “Leading Discounters” list in Discount

W and located in [he Disootozy of Discount Deonrtment
mm to be included in the sample. The net result was store

location data for 113 discount chains. The “Alphabetical Index” of
the directory lists all cities within each state in which a discount
chain has at least one store open or any planned openings at the
beginning of the year." Each of the cities was then converted to a

9 The first digit ofa zip code identifies the

five-digit zip code.I
geographic region of the country. The second digit tells you the

portion of a State or States the area is located in. The third digit is

 

"Va-ietysta’esaresmallu'thandiscmmtdepamnmtstoresandarenotinthe‘lheD_n_eg_ory' ofDiscount
Went Stores (r the sample. Also, ally standard discomt deputmmt states were used. No
membership/warehouse clubs or hyper-markets were included in the sample.
"Iftwoormoredlainswereownedbyafimatdlesmnefimedleywaecanbinedintoonedlain. The
netresultwas113 chains.

" For 1984- 1986 no “Alphabetical lndex”exists,sothe extended listings of The gm of Diseamt
MaltStoresforthoseyearswereusedtoemmictashnilarlist.
‘TheRanndaniy1996ZJioCodanlgwasusedfirsttocmvertcitiestozipcodes. Ifacityeouldnot
beformd,d1e'l‘imeZioCodeDirecto_r_y1994editionwasused. Ifthereweremorethentwocitieswidlthe
same nameinthe same state the extended listings of The Directgy of Discmmt Merit Storeswas
medtodetexminethecorrecteounty.Oncetheem'reeteamtywasknownmecmeetcitycouldbe
idaltified. Aficdfisprocessathyfiacdonofzipcodescaddnotbefmmdfleasdml%)

19

the sectional center of a multi-coded city, and the last two digits
tell you the individual post office branch.20 The zip codes were
then plotted on a map using a mapping program. A visual
inspection of Wal-Mart’s movements and other firms’ expansion and

retrenchment decisions was then possible.

To generate local markets, the last two digits of the zip codes
were dropped generating 3-digit zip code area markets. Within the
time period of 1975-1996, 862 of these local 3-digit zip code
markets had at least one discount store during at least one year and
69 of the 113 chains competed with Wal-Mart in at least one of the
local markets. Between 1976 and 1994, Wal-Mart entered 540 of
these local 3-digit zip code markets and attacked 59 different chains

owned by 62 different firms.

The trade journals give the chain name and the immediate
company that owns the discount store chain. Oftentimes the
immediate company that owns a chain is a 100% owned subsidiary
of a parent company. In order to determine if there was a parent

firm, and if a company was privately held or publicly traded, Dun &

Bradstreet’s Wm WM
Affiiiojions Who Owns Whom, and Wards Business Directory were

 

used in that order. A firm was classified as “public” if their stock,
or the stock of their parent was traded on the NYSE, ASE or in the

OTC market. Otherwise they were classified as “private”. There

 

ThisdefinitimofthemenningofzipeodesisfimndleRanndNLlly1996ZioCodeFinder.

20

was sometimes a one or two year lag between when ownership
actually changed and when it showed up in one of the above
sources. The NYSE, ASE, and OTC Daily Stock Price Records
books were checked to see if a firm was still trading on January 1 of
a given year. This way it could be identified in which year the
ownership changes occurred. When the name of the firm owning a
chain changed, Discount Merchandiser was used to identify the
underlying reason for the name change. Each name change was
identified as being due to a LBO, being acquired by another firm,
going public, a spin-off from a parent or to simply to change the
firm’s image.“ The 69 chains that competed with Wal-Mart between
1975-1996 were owned by 76 different firms during the years of
competition with Wal-Mart.22 The 76 firms consisted of 40 firms
that were always public, 21 that were always private, 13 that were
both (not at the same time) , and 2 that were foreign held during the
1975 - 1996 period. The foreign held firms were dropped from the
sample, as detailed information regarding ownership was not easily
available.

Discount sales per square foot came from The Directory of

W. Occasionally total square footage was

not shown for a given firm for a given year. The average square

 

"wedidmsdmmsfir 1981- 1989wereusedtoverifythatdlisjrocessdid
notomitanymergasaLBO’s.

22LBO’s, spin-offsmdgoingpublicwerenotco‘mtedasfirmdimgeshaeonlyacquisitimswa'e
counted.

21

footage of a store for that firm in a surrounding year was then used
to determine sales per square foot.”3 The number of stores in the
discount department store industry and sales per square foot (based

on selling space) for the discount department store industry came

from W. Total parent firm sales for both
public and private firms were found in the WW,
The Dirooiory of Coroorato Affiliaiions Who Owns Whom, and
flasds Business Direotosy in that order. Firm level discount sales

came from The Direotory of Discount Deoartmeni Stores except for
1975 through 1978 where they came from W.

Insider ownership was taken from Proxy Statements or Volne Lino
Investmsnt Survey.24 A firm’s total debt to total assets ratio and

operating profit margin came from Compustat.

3.4 Empirical Results
3.4.1 Explanatory Variables

Several factors other than managerial ownership can be
identified that might be expected to contribute to the expansion
decision by public firms in a local market. Several theoretical
models hypothesize that capital structure will impact a firm’s

behavior in the product market. One group of models predicts that

 

23Inaflrweusos,(pa'ticulal'lyobservationsfi'omthe l970’s),sales<rd1emnnberofstaesinadiainwere

notlistedinTheDireetgyofDiscothMmtStores. Discorthadlandiserwasusedindleseenses
tofindthisinformation.

2‘ Ownership labeled as “inside”, “directory”, “family-owned” (I “an individual” were included in the
insider ownership values

22

higher financial leverage leads to more aggressive product market
behavior.” Another group of models predicts that higher financial
leverage leads to less aggressive product market behavior due to
decreased access to capital."5 Recent empirical papers support the
premise that higher debt leads to less aggressive behavior on the
part of the high debt firms.”

Access to internal capital may also affect managers’ decisions.
Meyer, Milgrom and Roberts (1992) and Scharfstein and Stein
(1997) propose that divisions of diversified firms may be subsidized
sub-optimally by more profitable divisions. This could result in
less investment in divisions with good prospects and more
investment in divisions with poor prospects relative to focused
firms. However, Stein (1997) argues that managers of diversified
firms are better able to distinguish between better and worse
divisions than outside capital markets and are thus more likely to
restrict funding to poorly performing units than outside lenders if
these divisions were stand-alone firms. This implies that when
divisions face a tough competitor, and hence face lower future
profits for that division, that division’s funding from headquarters
may be restricted. This could lead to less aggressive investment

and output behavior when facing a new competitive threat.

 

1’ See Brande andLewis (1986), Maksimovic (1988), and Rctemba'g and Sdlarfstein (1990)
" See Fudenbergand T'n'ole (1986) and Bolton and Scharfstein (1990)
1' See Phillips (1995), Chevalia' (Sept. 1995), Cole and Titman (1994), and Safieddine and rm (1997)

23

Existing empirical studies have documented that there is a
value loss associated with diversification. Comment and Jarrell
(1995) document that firms that increase focus have positive
abnormal stock returns. Berger and Ofek (1995) show that the value
of diversified firms is less than the value of a similar set of stand-
alone firms. Recent empirical studies have examined the question
of whether it is sub-optimal cross-subsidization which is driving the
value loss. Berger and Ofek (1995) find evidence that diversified
firms over-invest and provide subsidies to failing segments. More
recently, Scharfstein (1997) finds that diversified firms invest more
than stand-alone firms in low Q industries, but invest less than
stand-alone firms in high Q industries implying there is “socialism”
in allocating funds across divisions.

Additional firm specific factors as well as market specific
factors and economic conditions may also contribute to the decision
of a firm to expand against Wal-Mart after being attacked. Firm
size may play a significant role in determining a firm’s strategic
response to a new competitive threat. Consistent with theoretical
models in the strategic management literature, Chen and Hambrick
(1995) find that in the airline industry, large firms are more likely
to respond when attacked than small firms, and small firms are more
likely to initiate competitive actions than are large firms. A firm’s

response to new competition may also be determined by a firm’s

24
dependence on the market under attack. Chen and MacMillan (1992)

build and test a model which shows that the greater the dependence
of a firm on the market under attack, the more likely they will
respond to an attack. Rivalry variables like the market share of a
firm in a local market as well as the local market’s Herfindahl index
may also be important determinants of firm expansion decisions at
the local level (see Chevalier (June 1995)) and need to be included
in the model specification. Economic conditions over the window
when an incumbent’s expansion decisions are being examined may
also determine whether or not a firm expands and needs to be
controlled for.

Chevalier (June 1995) also includes demand variables like the
number of households, and median income at the local market level
(MSA’s). Unfortunately, reliable population statistics only exist at

the zip code level for 1990.28

3.4.2 Significance of Managerial Ownership: Private Versus
Public Firms
The discount retail store industry contains a large number of
privately held firms. The trade journals have store locations,
discount sales, and sales per square foot for both public and private

firms. This provides the unique opportunity to examine the

 

"Pcpulationnunbersalsocatistfor 1980nmezipcodeleve1, buttheU.S.CensusBlreaudiscourages
dieiruseastheyclaimmel980mnnbersarelmreliable.

25

decisions of firms with extreme levels of insider ownership. If
insider ownership is a determinant of how firms respond to the new
competitive threat, any differences should be the largest by
contrasting private firms with public firms.

Using a univariate probit methodology similar to that of
Chevalier (June 1995), I test to see whether public versus private
ownership determines whether or not incumbent firms will expand
when Wal-Mart enters one of their 3-digit zip code area markets.
Year 0 is the year in which Wal-Mart is first seen competing in a
particular 3-digit zip code area j. The dependent variable is one if
incumbent firm i ever increases the number of cities in which it has
stores in zip code market j during three year period from Year —1 to
Year +2. Otherwise the dependent variable is zero. Year 0 can be
any year in the range of 1976-1994.29

Firm focus was not included here as an explanatory variable as
the private firms were highly focused, and it could act as a proxy

for public/private ownership. The model specification is as follows:

Probability of Expansionij = a + 131 Private/Public Dummy, + B; LBO/No LBO Dummy,
+ [33 Chain Size, + B4 Sales/Sq. FL; + [35 Mkt Shrijaierfindahlj)
+ 86 Dependence Mktij- + [37 Strength ofWal-Mart Dummy
+ 133 Avg. Industry Growth + 8i,-

 

2’Ihavestorelocatitnldutali'om1975-1996. HoweverJneeddataone-yeerpiaandtwoyemsfollowhg
meeventtoa‘eatetheeventwindow.

26

The independent variables used in this specification are defined as

follows:

Firm Soecific Variables:
Public/Private Dummy. : 1 = privately held; 0 = public. This is

recorded at the beginning of Year -1 for firm i who is first seen
competing against Wal-Mart in zip code j at the beginning of Year

0.30

LBO vs No LBO Dummyi: l = LBO done by firm i in the three
years prior to Year —1; otherwise 0. This is recorded for the three
years prior to the beginning of Year -1 for firm i who is first seen

competing against Wal-Mart in zip code j at the beginning of Year

0.31

Chain Size, : The number of five-digit zip codes in which firm i
has at least one store. This is recorded at the beginning of Year -1
for firm i who is first seen competing against Wal-Mart in zip codej

at the beginning of Year 0.

 

3°lfthecompmlytlultovlrrnsthedninisasubsidiarycradivisialofaparultcompanythalthispertainsto
theparult.
3‘Actualdebtratiosarelnlavailableforprivatefirms.Chevalieralsodividedhersample'nlto“1ow”lld
“high”debtfirmsbasedcnwheth¢ornottheyhadmdegcneaLBOasshealsohadprivatefirms'mhc
sample.

27

Inflation Adjusted Sales per Square Foot. : Firm i’s inflation
adjusted discount “sales per square foot” stated in 1974 dollars for
the fiscal year ending at the beginning of Year —1 for firm i who is
first seen competing against Wal-Mart in zip codej at the beginning

of Year 0.32

Rivalry Variables:

Market Share” : Market share of firm i in zip code j. This is
recorded at the beginning of Year -1 for firm i who is first seen

competing with Wal-Mart in zip code j at the beginning of Year 0.33

Modified Herfindahl Index, , This is the sum of the squares of the
market shares (defined above) in zip code j. This is recorded at the
beginning of Year —1 for zip code j when the first reported entry by

Wal-Mart into zip codej is at the beginning of Year 0.

Dependence on Mktu : The fraction of firm i’s stores competing in

zip code j at the beginning of Year -1 for firm i who is first seen

 

3’ This is ameasu'e of operating profitability. Operating income is not available for private firms Due to
diefactthatprofitmarginsaresmallintl'lisindusn'y,thisshouldbeafairlygoodproxyfor<perating
tability. lfsalespasquarefootwasmissingforoneyear,theva1ueforthepreviwsyearwasused.
’I‘hisisdoneusingthenumberofdifl‘erentS-digitzipcodesinwhichfirmihasstoresin3—digitzipcode
marketjdividedbythesameealculationfordleothercompaniesinzipcodemm’ketj. Also,theI-Ia'findahl
cwldbeusedhereinsteadofmarketshm’easdleyarehighlycmrelated.

28

competing against Wal-Mart in zip codej at the beginning of Year

0.34

Strength of Wal-Mart Dummy: 1 = Year 0 is in the range of 1986-
1994; 0= Year 0 is in the range of 1976-1985. Year 0 is the year in

which Wal-Mart is first seen competing in zip code j.”

Economy Variable:

Avg. Industry Growth: Arithmetic average of the annual growth in
the total square footage of discount stores in the discount store
industry during Years —1 through +2 for firm i who is first seen
competing against Wal-Mart in zip code j at the beginning of Year
0.

The first test includes all firm i market j pairs where firm i is
an incumbent when Wal-Mart enters zip code market j for the first
time between 1976 -— 1994. All observations where firm i switches
from public to private or from private to public during Year -1 to
Year +2 are dropped. Table 1 shows the fifty-nine different firms
that are represented in the sample for this specification.36 Table 2
provides the summary statistics for all of the variables used in this

specification. There are 1506 firm i, marketj observation pairs.

 

”Thisisdmemhrgdlemnnbaofdifl‘ms-digitzipcodesinwhidlfirmihasstcresinthe3-digitzip
codemarketjdividedbythetotalnmnberofdifi‘erentS-digitzipcodesinwhichfirmihasstores.
35AdeeadedlnnmyisusedheretorepresentthefactthatWal-Martmayhavebeenastrongercompetitorin
19861994whmmeywerellgerfllmdleywa'efi'an 1976-1985.

29

198 of these observations are for private firms. Table 3 shows the
simple correlations between the explanatory variables.37 The
variance inflation factors are all under 1.6 and the condition index
is 15.5.”

Results for the expansion univariate probit are shown in Model
1 of Table 4. Public firms are more likely to expand than private
firms. The public/private dummy is negative and significant at less
than 1%. The marginal effect using the infinitesimal change based
adjustment at the means of the independent variables is -9.7%. The
infinitesimal change based adjustment can be improved when the
explanatory variable is a dummy. To do this the probability of
expansion for public firms versus private firms is calculated
directly using the means of the other variables. Using that
methodology being private decreases the probability of expansion by
8.1%. This result is consistent with an ownership explanation. It is
also consistent with the hypothesis that private firms have restricted
access to external capital markets. The result could also be due to
differences in the amount of information that is available to the
market between the groups. Later I will test between these

alternative explanations by looking only at public firms.

 

36SomefirmshadtobedroppedastheyfirststartedcmnpetingwithWal-Martafier1994,:n'theirmlesper

somebotdatawasmavailable. Thislefi59firms.
IteanbeseenthatthemarketshareoffirmimmarketjandthemodifiedHerfindahlindexfilrmarketj

haveahighsimplecarelatimwiflieachodler. Thusmespecificationisnmwithjustthemarketshare

variablefirst.

3'Scrnet'ulesofthumbrethatVIF’sover lOoraconditionindexover30indieatepossibleharmfill

multicollinetity.

30

Firms that undergo LBO’s in the period prior to Year 0 are
more likely to expand when attacked. The LBO dummy is positive
and significant at 4.2%. The marginal effect using the infinitesimal
change based adjustment is 15.9% at the means of the independent
variables. Once again, the infinitesimal change based adjustment
can be improved on in the case of a dummy variable. This is done
by calculating the probability of expansion for firms that have
recently undergone LBO’s versus those who have not using the
means of the other variables. Using that methodology recently
undertaking a LBO increases the probability of expansion by 21.6%.
This is consistent with the models that hypothesize that higher
financial leverage leads to more aggressive product market
behavior. It is also possible that the firms who were already more
aggressive did LBO’s. These findings are the opposite of what was
reported in Chevalier (1995). She reported that LBO’s made the
product market softer. However, the only firm specific factor that
Chevalier controlled for was firm size.” Also, the LBO results need
to be interpreted with caution, as there are only 20 observations
where firms undertook LBO’s representing 3 firms. In addition the
reason for the LBO’s needs to be considered.

Firms that have more discount stores are also more likely to

expand. This result is consistent with Chen and Hambrick (1995)

 

”Formple,itislfltelythflmanyfirmsdlatmda’goLBO’sm’ecmvafingfiompubficmpfivate
ownership.

31

who find that large firms are more likely to respond when attacked.
However, this result is significant at the 8.7% level. Firms that
generate higher industry adjusted sales per square foot are more
likely to expand. This is consistent with the idea that firms that
have the best prospects and chance of successfully competing are
more likely to expand. The more dependent firms are on a market,
the more likely they are to expand when attacked. This variable is
positive and significant at 1.1%. This is consistent with the model
developed by Chen and MacMillan (1992) which shows that the
greater the dependence of a firm on the market under attack, the
lower the likelihood that they will not respond. Firms with a higher
market share are less likely to expand when attacked. They may
feel comfortable with the stronghold they have on the market or may
feel they already have the best store locations. Firms are also less
likely to attack in the 1985-1994 period when Wal-Mart is larger as
the decade dummy is negative and significant at less than 1%.

Firms also appear less likely to expand the faster the growth in
sales per square foot for the industry. This is consistent with firms
buying stores from other retailers when economic growth is down
and some retailers are exiting. This is when they can expand
quickly and relatively cheaply as this is when the prices on stores or
leaseholds are likely to be attractive.

A Lagrange Multiplier test statistic was calculated to check

32

for multiplicative heteroskedasticity using all of the right hand side
variables except the constant. The null hypothesis of
homoskedasticity cannot be rejected at the 95% level. As an
additional check, the specification was re-run using White’s robust
estimator of variance as it provides an appropriate estimator for the
variance even when the precise nature of the heteroskedasticity is
unknown. The p-values are virtually identical to the p-values
without the heteroskedasticity robust estimator and thus are not
shown.

The specification was re-run using the modified Herfindahl
index for market j instead of the market share of firm i in market j.
The results are shown as Model 2 in Table 4. The sign on the
modified Herfindahl index is negative as was the sign on the market
share of firm i in market j, but it is not significant. The other
results remain similar to those using the market share variable
instead of the Herfindahl variable.”

It is generally a good idea with large number of firms and a
small number of time periods to allow for separate intercepts for
each time period. This allows for aggregate time effects that may
influence the dependent variable. The nineteen-year period

represented in the sample was broken into 8 groups of two-year

 

”Ibemrkdshmvaiaflemdflafindaflmndrmmgahammespedfimfimasmeyuehighly
emeUedwitheachodler.

33

intervals and 1 group consisting of a three-year interval.‘1 The first
specification was re-run adding 8 time dummies. Time dummies
using one-year intervals could not be done as some years had very
few observations and no incumbent expansions. The results are
shown in Table 5. None of the time dummies is significant. The
main differences between this specification and the one without the
time dummies (see Table 4) is that the decade dummy, the industry
growth in square footage, and the LBO/No LBO dummy are no
longer significant. The LBO/No LBO does have a p-value of 10.5%.
It is not surprising that these three variables are no longer
significant when time dummies are included, as they are the ones
that would be expected to change over time. The private/public
dummy, chain size, inflation adjusted sales per square foot, market
share of a firm in a market, and a firm’s dependence on a market are
still of the same sign and have similar significance as in the
specification without the time dummies.

A Lagrange Multiplier test statistic was done to check for
multiplicative heteroskedasticity using all of the right hand side
variables except the constant. The null hypothesis of
homoskedasticity cannot be rejected at the 99% level. The
specification was re-run using White’s robust estimator of variance

as the null hypothesis of homoskedasticity could be rejected at the

 

1' Twodoesnotgo intonineteen evenly. 'l‘hereare8timedmnmiesrepresentingtwo-yearintervalsand
onet'nne dummy representing thethreeyear interval of1976-1978.

34

95% level. The robust p-values are shown in the third column of
Table 5. There were only small differences between the robust and

non-robust p-values.

3.4.3 Determinants of Private Firm Expansion

The previous set of tests showed that public firms were more
likely to expand against Wal-Mart than private firms. It is still not
clear if this is due to differences in the ownership levels,
differences in the organizational forms or differences in access to
capital between the two groups. It is possible that private firms
have less access to capital, both from less access to external capital
as well as the absence of an internal capital market. If this is true,
then we might expect higher operating cash flows to be a more
important determinant of higher investment for private firms.

Thesameunivm’iateprobitspecificationwasrunonfirmi,marketjpairsfor
private firms only with the exception of the public/private dummy for ownership.

The model specification is as follows:

Probability of Emansions = a. + a, LBO/No LBO Dummyi + a, Chain Size,
+ [33 Sales/Sq. FL, + [34 Min 31%. + B5 Dependence Mktij-
+ Bo Strength of Wal-Mart Dummy
1‘ I37 AVS- Industry Growth + Sij

Table 6 provides the summary statistics for all of the variables

used in this specification. There are 198 firm i, market j

35

observations. 24 private firms are in the sample. Any cases where
a private firm went public between Year —1 and Year +2 were
dropped.

Results for the expansion univariate probit are shown in Table
7. All ofthe coefficients have the same sign as the previous
regression that had both the public and private incumbents. Only
two variables are still significant for private firms. The LBO
dummy and inflation adjusted sales per square foot. The LBO
dummy is still positive and is significant at 5.7%. This again is
consistent with the models of Brander and Lewis (1986) and
Rotemberg and Scharfstein (1990) who hypothesize that higher
financial leverage leads to more aggressive product market
behavior. Inflation adjusted sales per square foot is positive and
significant at 3.3%. This supports the idea that private firms may
be using internal capital markets to fund growth. However, the
magnitude and significance level of this coefficient are very similar
to those for the specification using both public and private firms.
This would imply that both public and private firms rely on funding
from operating cash flows in a similar manner.

The dependence of a firm on zip code market j is statistically
significant only at the 10.2% level. This provides weak support for

the Chen and MacMillan (1992) model which shows that the greater

36

the dependence of a firm on the market under attack, the lower the

likelihood that they will not respond.

3.4.4 The Significance of Managerial Ownership: Public Firms
Only

The previous section showed that both public and private firms
rely on internal cash flows in a similar fashion. It is possible that
the differences in aggressiveness between public and private firms
is due to restricted access to external capital for private firms or
inherent differences between the two organizational forms rather
than a difference in insider ownership. It is also possible that the
difference could be due to the lack of an internal capital market for
private firms as they are more focused then the public firms.

To test these things a similar univariate probit specification is
run on public firms only. More information is available for public
firms, so better specified models can be used. First, the
public/private dummy is replaced with the actual percentage of
insider ownership. If managerial ownership is driving the
differences in aggressiveness between public and private firms then
this same effect should be visible within the group of public firms
as well. It could also be that private firms have different levels of
debt than public firms. It is possible that differences in
aggressiveness between the two groups may be caused by

differences in the debt levels between the two groups. The second

37
change to the model is to replace the LBO/No LBO dummy with the

actual total debt ratios of the firms to control for this. The third
change is to include the fraction of sales coming from discounting
as an explanatory variable. It could be that private firms are less
aggressive than public firms due to the lack of an internal capital
market. The inclusion of this variable will control for this
possibility. The new firm specific independent variables are

defined as follows:

% Insider Ownership; : The fraction of shares held by insiders,
directors, family members and individuals that is reported in lain;
Line investment Survey or from a firm’s proxy statement.42 This is
recorded at the beginning of Year -1 for firm i who is first seen
competing against Wal-Mart in zip code j at the beginning of Year

0.43

Total Debt/Total Assets.: This was calculated as total assets
(Compustat Item #A6) minus Stockholder Equity (Compustat Item
#A216) divided by total assets (Compustat Item #A6). It is thus the
book value of total debt divided by the book value of total assets

and was calculated at the beginning of Year —1 for firm i who is

 

“VahleLinemusedfirst. Ifmsidcownashipwasnothstedmaedlmlwenttoflleproxystatanalts.
”Ifinsideowna'shipwasnotavailableumebeginningonear—l lwentasfarbnckasYeer-4mdasfar
forwardast+2togetitasfllisvmiableisqtitestableovatime

38

first seen competing against Wal-Mart in zip codej at the beginning

of Year 0."4

Focus. : Discount sales divided by total firm sales for the fiscal
year ending at the beginning of Year —1 for firm i who is first seen
competing against Wal-Mart in zip codej at the beginning of Year

0.45

The model specification is as follows:

Probability OfEWIOIIij = a. + B. % Insider Ownership. + BzTotal Debt Ratio.
+ [33 Focus. + B4Chain Size. + B, Sales/Sq. Ft..
+136 MktShr..+B7DependenceMkt.j
+ B. Strength of Wal-Mart Dummy
+ 3.. Avg. Industry Growth + 8.,-

As before, all observations were dropped if the firm either
changed from public to private or from private to public during the
interval of Year -1 to Year +2. Table 8 provides the summary
statistics for all of the variables used in this specification. There

are 1209 firm i market j observation pairs.“5 38 public firms are

 

“Twofirms,l-IecksandS.E.NidlolscmldnotbefmmdonCanpustat. Thisratiowascalmlatedfrcm
statements located in Moodoy’s Industrial Manmls instead
”Sometimesthisratiowasyeatermanlastotalfirmsalesmddiscmmtsalescamefiomdifl’erent
sources Whenever this happened 1 was used.

“Saneobservationsfcrpublicfirmswered‘opped. Allobservationspria'to 1978wered'oppedduetoa
lackochmpustatdata. Alsqscmeobservaticnswu‘ech'oppedduetoalackofownershipdata.

39

represented in the sample."7 Table 9 shows the simple correlations
among the explanatory variables. All nine variance inflation factors

are less than 2.3 and the condition index is 22. 4‘

Table 10 shows the results for the expansion univariate probit
using inflation adjusted sales per square foot as a measure of
operating profitability. The percentage of inside ownership is
negative and significant at less than 1%. This implies that the
higher the inside ownership, the less likely the firm is to expand in
market j when they are attacked there. This lends support to an
ownership explanation rather than an access to capital or a
difference in organizational form explanation as the reason why
private firms are not as likely to expand as public firms. If the
difference between public and private firms had been access to
capital, or differences in organizational form, we would expect this

variable to be insignificant for public firms.49

The total debt to total assets ratio is negative and significant
at less than 1%. This implies that the higher the total debt ratio,
the less likely firm i is to expand in market j when they are
attacked. This is consistent with the models of Bolton and

Scharfstein (1990) and Fudenberg and Tirole (1986). They

 

”Afewfirmswerepublic saneofthetimeandprivate someofthetime. Halce, thenumberofpublic
firmsrqreuented(38 firms)phlsthenlunbu' ofprivatefirmsrepresented (24 firms) exceeds thetotal
munber of firms in the sample (59 films).

“Sanerulesoftlnnnba'ethatVIF’sover10maccnditionindexover301ndieutepoe§blehrmfill
muhieollinetity.

4o

hypothesize that financial leverage leads to less aggressive product
market behavior on the part of the firm. It may also be the case that
weaker firms have taken on debt. However, the proxy for industry
adjusted profitability in the regression specification should help
mitigate this possibility.

The focus variable is positive and statistically significant at
less than 1%. This means that more focused a firm, the more likely
it is to expand when they are attacked by Wal-Mart in a local
market. This appears inconsistent with Meyer, Milgrom, and Roberts
(1992) as they hypothesize that for diversified firms, divisions can
easily subsidize other units. These results are perhaps more
consistent with Stein (1997) who argues that diversified firms can
allocate less resources to divisions with poor prospects forcing
poorly performing units to receive funding that is less than they
could obtain as stand alone firms. These findings are also consistent
with the empirical work of Lang and Stulz (1994) who find evidence
that firms diversify to seek better growth opportunities. Perhaps
when tough competition hits, firms invest in divisions where the
competition is less tough. The evidence also supports an argument
similar to the one proposed by Chen and MacMillan (1992)
regarding the dependence of a firm on a market. Firms who are only

in the discount department store industry are more likely to expand

 

”ThisisalsoccnsistmtuddlKaplandehlgales(l995)whofinddmtmonly15%offirmyearsisthere
memmdmastotheabifityofaMmhaveaccessmintanaloreldanalfimding However,their

41

as that is all they have to rely on. Also, the dependence of firm i on
local zip code market j, and industry growth in square footage of

store space both remain significant.

Lagrange Multiplier test statistic was done to check for
multiplicative heteroskedasticity using all of the right hand side
variables except the constant. The null hypothesis of
homoskedasticity cannot be rejected at the 95% level. As an
additional check, the specification was re-run using White’s robust
estimator of variance. The p-values were virtually identical to the

non-robust p-values and are not shown.

It is generally a good idea with a large number of firms and a
small number of time periods to allow for different intercepts for
each time period. This allows for aggregate time effects that may
influence the dependent variable. The nineteen-year period
represented in the sample was broken into eight groups of two-year
intervals and one group consisting of a three-year interval. ’° The
specification was re-run adding eight time dummies. Time
dummies using one-year intervals could not be done as some years
had very few observations and no incumbent expansions. The results
are shown in Table 11. None of the time dummies is significant.
The results are almost identical to those without the time dummies

included.

 

studycnlylooksatpublicfirms.

42
3.5 ROBUSTNESS CHECKS

3.5.1 Operating Profit Margin Instead of Discount Sales per
Square Foot

Due to the fact that profit margins are small in the industry,
inflation adjusted discount sales per square foot should be a fairly
good measure of operating profitability. However, it is possible
that discount sales per square foot is measuring an attempt to
increase market share instead. As a robustness check the public
incumbent firm specification was re-run using total firm operating
profit margin instead of inflation adjusted discount sales per square
foot as the operating profitability measure. The summary statistics
for operating profit margin are in Table 8. The results from the
probit are in Table 12.51 The results are similar to the results using
inflation adjusted discount sales per square foot as the measure of

operating profitability.

3.5.2 Parent Firm Size

Both Demsetz and Lehn (1985) and Mikkelson and Partch
(1989) report an inverse relationship between insider holdings and
firm size. In the earlier specifications chain size was used. It is

possible that chain size is not fully capturing the parent size as

 

”Twodoesnctgointonineteenevenly. Thaea'estimedmnmiesrep‘esentingtwo-yearmtervalsmd
unethnethmyrqxesent'mgflledneeyear'nnavalofl976-l978.

43

insider ownership is measured at the parent level. Thus it is
possible that the significance of the insider ownership coefficient is
due to parent size instead. As a check, inflation adjusted total
assets of the parent firm is added to the specification for the public

firms. The variable is measured as follows:

Inflation Adjusted Total Assets of Parent. : Firm i’s parent’s
inflation adjusted total assets 1974 dollars at the beginning of Year
-1 for firm i who is first seen competing against Wal-Mart in zip
code j at the beginning of Year 0. Total assets came from Compustat
(Item #A6), and were then adjusted for inflation using the consumer

price index.

The results of the probit are in Table 13. The coefficient and
significance of the insider ownership variable remain unchanged. It
is still significant at less than 1%. All other variables have similar
coefficients and significance levels except the focus variable. It is
now significant at 4.3% instead of less than 1%. It appears that the
previous specification had already captured the effects of parent

firm size.

 

"Onefirmwasch'oppedasitsoperatingprofitmarginwasanmltlier. Whalthatfirmisincluded,
opaatingprofitmagmispositivemdsignificantatlessdlan 1%. The other coefficients and significance

44
3.5.3 Unmodeled Market Heterogeneity

Because the specifications use local market level data, the
concern arises that there could be unmodeled local market
heterogeneity that is correlated with the other independent
variables. For example, it is possible that firms with low ownership
happened to be incumbents in markets where expected growth was
higher. Unfortunately, the results could not be re-estimated with
three-digit zip code or even state dummy variables because there are
not enough observations per geographic area. However, regional
dummies can be put in the regressions. The regions are Northeast,
Southeast, Foundry, Midwest, Rocky Mountain, and West Coast.’2
Because Wal-Mart moved around region by region, they primarily
attacked most markets in a given region during the same years.
Hence, these dummies are basically proxies for a region during a

certain time period.

The probit results for all incumbents both public and private
are shown in Table 14. The results are similar to those done for all
incumbents without the regional dummies. The public/private
ownership dummy is now significant at 2.4% instead of at the 1%
level. The LBO/No LBO dummy is now significant at 1.2% instead
of 4.2%. The market share of firm i in marketj is now significant

at 2.4% instead of 6.9%. Also the significance on the Wal-Mart

 

levelsremainvirulallylmdianged.

45

strength dummy and industry growth in square footage are also
slightly less, but this is to be expected as they vary through time. A
Lagrange Multiplier test statistic was done to check for
multiplicative heteroskedasticity using all of the right hand side
variables except the constant. The null hypothesis of
homoskedasticity cannot be rejected at the 99.5% level. The
specification was also re-run using White’s robust estimator of
variance as the null hypothesis of homoskedasticity can be rejected
at the 95% level. The robust p-values are virtually identical to the
non-robust p-values and are shown in Table 14. One regional
dummy is significant at the 1% level. Another regional dummy is
significant at the 10% level.

The regional dummies were also put in the regression using
public firms only. The results are shown in Table 15. Again the
results are similar to the earlier results without the regional
dummies. The fraction of insider ownership, the debt ratio, the
focus measure, inflation adjusted sales per square foot and the
dependence of the firm on the market under attack are all still
significant at the same levels that they were without the regional
dummies. The only noticeable change is that the market share of
firm i in market j is now significant. One regional dummy is

significant at the 1% level, and another is significant at the 5%

 

”Theseregicnswereselectedusing'lheNineNaticns ofNorthAmericaasawaytodividemeUnited
Statesintoeconomicregions.

46

level. A Lagrange Multiplier test statistic was done to check for
multiplicative heteroskedasticity using all of the right hand side
variables except the constant. The null hypothesis of
homoskedasticity cannot be rejected at the 95% level. The
specification was also re-run using White’s robust estimator of
variance. The robust p—values are virtually identical to the non-
robust p-values and are not shown.

Another way to try to control for local market growth is use
changes in population levels at the zip code level. Unfortunately,
reliable population data only exists at the zip code level for 1990.53
However, population numbers do exist at the state level for 1980 -
1996. Changes in the state population in which marketj is located
during the three-year event window can be included as an
independent variable to control for changes in demand in local
market j. Because the state population data exists only from 1980-
1996, all observations where Wal-Mart enters a market prior to 1981

were dropped.’4 The variable is measured as follows:

State Population Change. : The state population change in
thousands from Year -—1 to Year +2 for the state in which zip code

market j resides where Wal-Mart enters zip code market j for the

 

S3Pcpulntionntunbersalsoexistfor1980atthezipcodelevel,buttheU.S.CensusBlueaudiscctn'ages
their use. They claim that the 1980 values are unreliable.
“Recallthetln‘eeyeureventwindowgoesfrcheu-l toYew+2whereYeIrOisd1eyeurwhenWaL
Mutfirstenta'szipcodemarketj.

47

first time at the beginning of Year 0. State population numbers

came from the U.S. Census Bureau.

The probit results for all incumbents both public and private
are shown in Table 16. State population change is positive and
significant at less than 1% (t statistic is 5.577). This implies that
firms are more likely to expand when population is increasing in the
state in which zip code market j resides. Comparing the results to
Table 4 which shows the basic results for the public and private
incumbents without state population changes we see that the
magnitude and significance of the public/private ownership dummy
is virtually unchanged. The LBO dummy, market share, and
dependence on market j are all stronger under the new specification.
Chain size, inflation adjusted sales per square foot, the decade
dummy and industry growth in square footage are weaker under the

new specification.

The null hypothesis of no multiplicative heteroskedasticity
cannot be rejected at the 99.5% level using a Lagrange Multiplier
test statistic. The specification was also re-run using White’s
robust estimator of variance. The robust p-values are virtually

identical to the non-robust p-values and are shown in Table 16.

The specification for public firms alone was also re-run

adding state population change as an independent variable. The

48

results are shown in Table 17. When comparing the results to the
previous results for public firms in Table 10 some changes are
apparent. First, state population change is positive and significant
at less than 1% which implies that firms are more likely to expand
when the local market is in a state that is experiencing population
growth over the event window. Second, fraction of insider
ownership, the total debt ratio, and the focus measure are all still
significant but now at the less than 5% level instead of the less than
1% level. Third, inflation adjusted sales per square foot is now not
significant, but the market share of a firm in a given market is now
significant at the less than 1% level.

The time period of 1981-1994 was divided into seven groups
of two consecutive years. The specification was re-run using six of
the seven time dummies as independent variables. The results are
similar and are not reported.

The specification for public firms alone was also re-run
adding state population change as an independent variable and
replacing the market share variable with the modified Herfindahl
Index for market j. The results are shown in Table 18. The
Herfindahl Index is negative and significant at the less than 5%
level. Otherwise the results are very similar to those shown in

Table 17 using market share instead of the Herfindahl Index.

49

As with the previous test, the specification was re-run using
six of the seven time dummies as independent variables. The results

are similar and are not reported.

The null hypothesis of no multiplicative heteroskedasticity
cannot be rejected at the 99.5% level using a Lagrange Multiplier
test statistic for the specifications in Table17 and Table 18. The
specifications were also re-run using White’s robust estimator of
variance. The robust p-values are virtually identical to the non-

robust p-values and are shown in the respective tables.

3.5.4 The Importance of Managerial Ownership Controlling for

Firm Size and Economic Growth

As a test of agency theory, it would be helpful to see if
ownership is a significant determinant of the probability of an
incumbent firm expanding controlling for firm size and economic
growth. Other firm specific variables are excluded as they could be
decision variables. The public specification is re—run using
managerial ownership, firm size, how dependent a firm is on a local
market (as this is correlated with firm size), and economic growth
along with the time dummies. The estimation results are shown
under Model 1 in Table 19. As can be seen, managerial ownership
is negative and significant at 1.7%. Controlling for firm size and

economic growth, as well as aggregate time effects, high ownership

50

public firms make different investment decisions than low
ownership public firms.

The specification is re-run on public and private firms. The
managerial ownership variable is replaced with a private/public
dummy. The results are shown under Model 2 in Table 19. The
results on the ownership dummy are negative and significant.
Controlling for firm size and economic growth, as well as aggregate
time effects, private firms make different decisions than public
firms. Also, the private firms are making similar expansion

decisions as the high ownership public firms.

3.6 SUMMARY

The main result of this chapter of my dissertation is that
incumbent firms with high managerial ownership are less aggressive
when facing a new competitive threat then firms with low
managerial ownership. The result holds for public versus private
firms as well as within the sub-sample of public firms. I provide
evidence that suggests differences in the expansion decision are due
to managerial ownership and not differences in access to capital or
organizational form. Other results are that firms with higher debt
levels are less aggressive when facing a new competitive threat as

are diversified firms.

CHAPTER4

Incumbent and Market Specific Determinants of an Entrant’s

Penetration Into New Local Markets

4.1 Introduction

In Chapter 3, firm and market specific determinants of
incumbent investment decisions were examined as a response to a
new competitive threat. This chapter looks at the other side of the
strategic interaction. Which firm and market specific factors led to
a lower market penetration by Wal-Mart after they had already
entered a market? This way of measuring which types of firms are
making “good” or “bad” decisions avoids the possible endogeneity
that could be present if a firm’s performance was measured
following a specific firm decision. It also avoids the need to
collect data regarding the decisions that the firms made which may
be difficult or impossible to do in some cases. Also, firms can fight
in many ways. They can increase output, expand against Wal-Mart,
use aggressive pricing, or fight zoning approvals for new stores. It
would be difficult to gather information regarding all of these types

of decisions.

51

52
Agency theory predicts that managers with high ownership

will make value- maximizing decisions.” This suggests that they
will be more efficient. This efficiency could lead to higher
switching costs for customers and quicker responses on their part.
It may be difficult to gain market share from such firms.“ It is also
possible that managers with low ownership are over-investing in
market share as an empire building activity. If this is the case, they
may be setting prices too low and maximizing market share instead
of profits. This could make the markets they are in tougher to
compete in. It is thus an empirical question to determine whether
high or low managerial ownership deters market share gains by the
new entrant, Will-Mart. A finding that supports either of these two
hypotheses being true is important at it would suggest that
managerial ownership has real product market effects.

It is also of interest to examine whether or not Wal-Mart has
an easier or a harder time gaining market share in highly
concentrated markets. Highly concentrated markets may be difficult
to penetrate if the incumbent firms have barriers to entry and
expansion. They may also be easier to penetrate if competition is

not as fierce and there are profit opportunities.

 

”InsularldMecklingU976)
”Mmhawexmhedwhefiammgchlownashiphnmfirmdedsimswhmfilmsficeam
FaexampleseeBerger,0fekdeermack(l997),KhamamdPoulsen(1995),andSongandWalking
(1993). Thesepnpauhowevernrenotinthecontextofaproductmrket setting.

53
4.2 Research Design

Chapter 3 presented evidence that certain firm and market
specific factors do determine expansion decisions by firms when
facing a new competitive threat (entry by Wal-Mart into an
incumbent firm’s local market). Firms can compete in other ways
too. For example, incumbent firms can compete on price, invest in
advertising, or try to influence zoning decisions regarding
expansion by the new entrant. If firm and market specific factors
determine different responses by incumbents, a logical extension is
to try to determine which of these characteristics imply potential
profits to the new competitor.

One of the problems well recognized in the literature is that
firm specific characteristics may be caused by the same exogenous
factors that determine their decisions. Hence, a major concern in
developing the research design of this chapter is to minimize this
endogeneity problem. The methodology of Opler and Titman
(1994) is employed. Instead of trying to establish whether one
decision (say undertaking a LBO) causes weakness in the product
market, firm specific characteristics are measured prior to an
exogenous threat. Because this minimizes the reverse causality
problem, one is better able to decipher if any of these

characteristics determine subsequent product market performance.

54

The threat is new entry by Wal-Mart into a local market where
some other firms in the discount department store industry are
incumbents. Entry by Wal-Mart can be thought of as exogenous
because of the way Wal-Mart expanded and the fact that by the end
of the sample period they had entered virtually every market.
During the period of 1975-1996 Wal-Mart transformed itself from a
regional chain consisting of 125 stores in eight states to a national
chain consisting of 2,234 stores in all fifty states. Figures 5
through 8 show Wal-Mart store locations at the beginning of 1975,
1985, 1990 and 1996 respectively.” As can be seen, Wal-Mart
expanded throughout the United States through connected regions.

Competition between discount department stores takes place at
the local level. Following Chevalier (1995) the United States is
broken down into local markets. The local markets used in this
study are 862 3-digit zip code areas. WW
Finoor defines the five digits of the zip code. The first digit of the
zip code identifies the geographic region of the country. The
second digit identifies a portion of the geographic region. The third
digit identifies a sectional center or multi-coded city within that
portion of the geographic region. A sectional center is usually the
natural center of local transportation and serves the smaller post

offices surrounding it. A multi-coded city is a main city post office

 

”Themapsdiowallofthe5digitzipcodesinwhichWal-Mmthadatleastonestm'emdlegivalyem'.
Thestm‘elocafimscamefi‘anTheDh‘ectoryofDiscmthentStores.

55

that serves post office branches within a city. The fourth and fifth
digits in the zip code identify the individual post office branches
that are served by the sectional center or multi-coded city office.
Because the sectional centers and multi-coded cities are usually the
natural center of local transportation, the three-digit zip code area
served by them is felt to be a reasonable economic market. In 1975
Wal-Mart had discount department stores in 45 of the 862 local
markets. By the beginning of 1996, Wal—Mart had entered another
655 local markets for a total of 700 markets.

To employ this methodology, firm and market specific
characteristics for a local market are measured prior to entry by the
new competitor, Wal-Mart. The penetration rate of Wal-Mart into a
local market at the end of a subsequent period is then measured to
determine how well firms successfully prevented losses in market
share to Wal-Mart once Wal-Mart entered the local market. In other
words, which firm and market specific characteristics determined
“toughness” in product market competition as measured by
penetration rate of the new entrant.

A time line illustrating the empirical set-up is shown below.

Firm and Mkt Specific Wal-Mart Wal-Mart’s
Variables Measured iantheginning Mkt Slmein
For Market j of Year 0 Mid j Measured

L i 5 1

'— V

-1 0 +1 +2 Year

56

The trade journal used gives store location data at the
beginning of each year. Year 0 is defined as the first year in which
Wal-Mart has a store in 3-digit zip code market j. The firm and
market specific variables are measured at the beginning of Year —1
to avoid Spurious correlation between the variables and the
subsequent penetration of Wal-Mart. If they are flow variables
(from the income statement), they are measured over the fiscal year
ending at the beginning of Year —1.58 If they are non-flow variables
(balance sheet items and market specific variables), the values are
measured at the end of the fiscal year which is at the beginning of
Year -1.59 The economic activity variable is examined from Year -1
to Year +2.

This data set has a number of advantages. First, given that
most of the local markets experienced entry by Wal-Mart during the
period examined I am able to generate a large number of
observations. Second, I am able to keep the threat constant, since I
only study the entry and penetration of Wale-Mart into local markets.
Third, discount department stores carry essentially non-
differentiated merchandise and are thus homogeneous in that
respect. This reduces problems of differential product quality or

product differentiation across markets. Also, this data set provides

 

”mmishuhrsuy,ahnostallofther¢ailcshadafiscnlyeumding1anmy3lofthegivuryeurovcthe
timepn'iudexamined.

57

store location data and a reliable measure of profitability “sales per
square foot” for both public and private chains. This enables me to
include privately owned firms in many of the tests. Since this data
set permits us to keep industry, product and threat constant across
all observations, I can develop reasonably clean tests on how
different firm and market characteristics affect Wal-Mart’s

penetration into local markets.

4.3 Data

A trade journal called W was used to
identify major discount chains during the period of 1975 - 1996.
They define a discount department store as “a departmentalized
retail establishment utilizing many self-service techniques to sell
hard goods, health-and-beauty aids, apparel and other soft goods
and other general merchandise. It operates at uniquely low margins,
has a minimum annual volume of $100 million ($50 million in the
1970’s), and has at least 10,000 square feet of total space.” A list
was compiled of all of the discount store chains that Rim
Merchandiser had on the “Leading Discounters” list in at least one

year during the time period of 1975 - 1996. This procedure

 

”TheemaptionhceisLBO’s. IfafirmmdtmtaLBOinmeyuarspncaudingYu—meitwas
tnatadasaLBOfirminYear-l.

58

eliminated the very small “mom and pop” type chains which have

very little public information available.“

 

A trade journal called _ .
$919; was then used to determine store locations for all of these
chains for each year in which they operated in the industry during I
the period of 1975 -— 1996.61 Thus, a discount department store
chain had to be on the “Leading Discounters” list in Diecgggt
Merchandiser and located in lhehirectcrxemiscscnnflcuarment
Mes to be included in the sample. The net result was store
location data for 113 discount chains. The “Alphabetical Index” of
the directory lists all cities within each state in which a discount
chain has at least one store open or any planned openings at the
beginning of the year."2 For 1980 -l996 I was able to get exact
counts of stores for each firm in each city.63 The city and firm
location data was updated to reflect this information. Each of the

cities was then converted to a five-digit zip code.“

 

“VfiamemflhmMMthmmmmmmm-mm
Wammple. Also, mlymdudchsasmdepcmnaamudblo
membaship/wwdiwse clubs were included an the sample.
“IfMGmuediahswaeownedbyafimnmemefimemeywaeecmbinedhmmem The
netresultwaslliichains.
“Farm-ImnowphaheuaiwmmmmdedfimorW
forthoseyearswu'eusedtoccnsu'uctasimiltlist.

For 1980- 1993 theenmdedccmpanylistingswereused. Fa- l994-l996noextendedccmpmy
listingsezdstedsostateextendedhstingswaeused.
“MWI lwegmggermdawasusedfimmmnvamnatozipcoda Ifacitycouldnot
hemmeWedfimmused lffliereweremcreflientwocitieswiflime
samenameinmesamestatemeextendedlistingsof .~ .- . ~ .. .. ~
medtodetermmetheeorrectcmmty. Oncemecarectcountywashownfliecarectcitycmldbe
identified. Aflaflfismafinyfiacfimofzipeodescmfldnotbefomdflessflml%)

 

59

To generate local markets, the last two digits of the zip codes
were dropped generating 3-digit zip code local markets. The
methodology used requires data for one year prior to and two years
following Year 0. Because exact the number of stores for each firm
in each city is known for 1980 —1996, only local markets who first
had Wal-Mart in them at the beginning of 1981 - 1994 were used in
the sample. Between 1981 and 1994 Wal-Mart entered 488 of these
3-digit zip code local markets.

Data was needed on the incumbent firms in the markets
entered by Wal-Mart. The trade journals give the chain name and
the immediate company that owns the discount store chain.
Oftentimes the immediate company that owns a chain is a 100%
owned subsidiary of a parent company. In order to determine if

there was a parent firm, and if a company was privately held or

publicly traded, Dun & Bradstreet’s W Ihc

Directgry gt Qgrporetc Affiliations Whg ang Whom, and Ward;
Business Directgry were used in that order. A firm was classified

as “public” if their stock, or the stock of their parent was traded on
the NYSE, ASE or in the OTC market. Otherwise they were
classified as “private”. There was sometimes a one or two year lag
between when ownership actually changed and when it showed up in
one of the above sources. The NYSE, ASE, and OTC Daily Stock

Price Records books were checked to see if a firm was still trading

60

on January 1 of a given year. This way it could be identified in
which year the ownership changes occurred. When the name of the
firm owning a chain changed, Dirceunt Merchandiser was used to
identify the underlying reason for the name change. Each name
change was identified as being due to a LBO, being acquired by
another firm, going public, a spin-off from a parent or to simply to
change the firm’s image.“

Discount sales per square foot came from Ihe Directory of
Qiecount Department Stores. Occasionally total square footage was
not shown for a given firm for a given year. The average square
footage of a store for that firm in a surrounding year was then used
to determine sales per square foot.“5 The number of stores in the
discount department store industry and sales per square foot (based

on selling space) for the discount department store industry came

from W. Total parent firm sales for both
public and private firms were found in the Million Qeilar Qireerery,
The Directory of Cernorate Affilintions; Who Owns Whem, and
Wnrde aninese Directory in that order. Firm level discount sales

came from The Directory of Discount Department Stores. Insider
ownership was taken from Proxy Statements or W

 

 

“Qanta'lyeditionsofMggersandAggn' 'tionsfor l981-1989wereusedtoverifythatthisprocessdid
notomit mymergersorLBO’s.
“Inafewcases,salesorthenunbaofstoresirraMwuenotlistedirr'l’heMy’ ofDiscount
Mag. Discathadnmdisawasusedmmesemsesmfindmismfumafim.

61
Investment Survey.” A firm’s total debt to total assets ratio came

from Compustat.

4.4 Empirical Results
4.4.1 New Entrant Penetration: The Importance of Incnmbent
Ownership
Agency theory predicts that there are value losses associated

with having less than 100% ownership. The prediction is that the

 

higher the managerial ownership, the lower the value losses. It
seems reasonable to predict that firms with low managerial
ownership might behave differently in the product market.
However, managers who have low ownership may be given incentive
contracts and be monitored to try to minimize the agency problem.
If agency problems cannot be mitigated with incentive contracts, we
might expect that the new entrant, Wal-Mart, would view firms with
lower managerial ownership as being weaker competitors and may
penetrate their markets quicker. It is also possible that managers
with low ownership may over-invest in market share at the expense
of profits. If this is the case, there may be higher switching costs,
and it may be tougher to gain market share from low ownership

firms.

 

”MhbeMmW,“W,“fimWa“mmfiWwaemchmmmemsida
ownadiipvahies. IfhsidcownashipwasnotinValrreLinepoxystatanmtswereused.

62

Indirect tests of Agency Theory can be done. I first test to
see if managerial ownership unconditionally determines the market
share of the new entrant (Wal-Mart) after a three-year period.

The methodology here is to measure the market share that

Wal-Mart has in a local market j in Year +2 where Wal-Mart has

 

entered market j sometime between the beginning of Year —I and the E
beginning of Year 0. Wal-Mart’s market share is measured by i
determining the fraction of total stores in local market j at the

beginning of Year +2 that are Wal-Marts. The dependent variable E

will be denoted as “penj”. This is felt to be a reasonable proxy for
the unobservable sales based market share.68 The underlying
assumption is that the local market was in equilibrium before Wal-
Mart entered, then a new equilibrium is established after entry.
This new equilibrium may have more stores if for example the
population is expanding over the period. Pen,- is capturing the
fraction of total stores that exist at the beginning of Year +2 owned
by Wal-Mart. Thus it is a measure of market share at that point in
time. The independent managerial ownership variable is defined as

follows:

fracprij; The fraction of stores in local market j at the beginning

of Year --1 owned by firms that are privately held.

 

“Manychahishavedifi‘eremsizestamfordifl'aentsizemarkets. Hmeeitisnotmreasonabletoassmne
thatdiainshavesimilssizedsta‘esinsimilm'sizedmm'kets.

63

The Model 1 specification to be estimated using OLS is as follows:

Pen“; = a + B1fi'acprij,-l+ B. annual time dummies + a,-

Wal-Mart entered 488 local markets between 1981 and 1994.

 

1:-
All local markets were dropped from the sample that had any
foreign held firms in them, or that had no stores in them at the
beginning of Year —1. This left 471 local markets. The first r

estimation ofthe model uses all 471 ofthese local market
observations. Annual time dummies are included in the model to
allow for aggregate time effects that may influence the dependent

variable, but their coefficients are not shown.

The results are shown in Table 20 for Model 1. Standard errors
are White’s heteroskedasticity robust standard errors. The higher
the fraction of stores owned by privately held firms in local market
j, at the beginning of Year --1, the lower Wal-Mart’s market share at
the beginning of Year +2. This result is significant at 3.6% and
implies that Wal-Mart finds product market competition to be
tougher in markets where there are a large fraction of privately held
firms.

An alternative assumption is that the market is in equilibrium
prior to Wal-Mart entering, and Wal-Mart is trying to capture some

of the existing stores. If one assumes that the equilibrium number

64

of stores is static, then another possible dependent variable would
be to determine what fraction of the stores in a market at the
beginning of Year —1 are owned by Wal-Mart at the beginning of
Year +2. Therefore, as a robustness check, the results were re-
estimated using the number of stores that Wal-Mart has in local
market j at the beginning of Year +2 divided by the total number of
stores in local market j at the beginning of Year —1. This dependent
variable will be denoted as “penbasej”. The Model 2 specification

to be estimated using OLS is as follows:

Penbasej=a+01fiacprij,.1+fltannua1tirnedummies+ej

All 471 market observations are used again. Annual time
dummies are again included, but their coefficients are not shown.
The estimation results are shown in Table 20. The higher the
fraction of stores owned by privately held firms in local market j, at
the beginning of Year -1, the lower Wal-Mart’s market share at the
beginning of Year +2 as a fraction of the number of stores that
existed immediately prior to Wal-Mart’s entry. This result is
significant at 2.5% and implies that Wal-Mart finds product market
competition to be tougher in markets where there are a large

fraction of privately held firms.

 

65

Will managerial ownership be a determinant of Wal-Mart’s
penetration into markets with only public firms in them? As
mentioned earlier, Wal-Mart entered 488 local markets between
1981 and 1994. All local markets were dropped from the sample
that had any foreign held firms in them, or that had no stores in
them at the beginning of Year —1. This left 471 local markets. F
About half of these markets, 233 to be exact, had only public firms

in them at the beginning of Year —1 and had ownership data

 

available for all firms in the local market at the beginning of Year
-1. The two specifications are re-run using these only these 233
local market observations and changing the fraction stores in the
market owned by privately held firms to the weighed average of the
managerial ownership levels of the firms in local market j. The new

managerial ownership variable is defined as:

w opr: The weighted average of insider ownership of the firms
with at least one store in local marketj at the beginning of Year —1.
Insider ownership is measured as the fraction of shares held by

insiders, directors, family members and individuals that is reported

in Ynlne Line Inveermeni Snrvey or from a firm’s proxy statement.69

 

”ValueLinewasusedfirst. Ifmsidaawnashipwasnothstedma'etlicnlwenttotheproxy statements.

66

This is recorded at the beginning of Year -1 for firm i.70 This

variable is reported in decimal form.

The Model 3 & Model 4 specifications to be estimated using OLS

are as follows:

Pen,-,+2 = a + [31w oij,.1+ [5. annual time dummies + s,-

Penbasej=a+l31woij,.1+B.annualtimedummies+ej

 

The results are shown in Table 20. In both of the
specifications, using only markets that have all public firms,
managerial ownership is negative and significant at less than 1%.
Wal-Mart has less penetration into local markets that have higher
managerial ownership irrespective of whether these are publicly
traded or privately held firms. Annual year dummies were included
in all four of the model estimations, but their coefficients are not
shown. The standard errors used in all three models are White’s
heteroskedasticity robust standard errors.

There is the possibility that there are some market size and
firm size factors that can jointly affect the penetration of Wal-Mart

into a local market and managerial ownership and thus may induce a

 

”Ifinsideownashipwasnotavailableatthebeginn'mgonem-lIwentasfirbackasYear-4andasfar
fawnrdasYenr+2togetitasthisvariableisquitestableovertime.

67

spurious correlation between them. To control for firm size and

market size three additional variables were added as controls:

lnstrsj: This is the natural log of the number of stores in local
market j at the beginning of Year -1. It is done to scale for market
size. One would expect that it is physically easier to get a higher

penetration rate in smaller markets.

proplgj: This is the proportion of stores in local marketj that
belong to firms in the upper quartile based on chain size at the
beginning of Year -1. This variable is reported in decimal form.
There may be economies of scale to being large. It is difficult for
the managers to own a large part of a bigger firm, raising the

possibility that managerial ownership is a proxy for firm size.

w dep,: The weighted average of how dependent a firm is on a local
market j for the firms with at least one store in local market j at the
beginning of Year -1. ”. The weights were calculated by
determining what fraction of the total stores in local market j firm i
had at the beginning of Year —1. How dependent firm i is on local
market j is defined as the fraction of firm i’s total stores competing
in zip code j at the beginning of Year —1. This variable is reported

in percent form. Some in the management strategy literature have

68

argued that the more dependent a firm is on a market, the tougher
they may fight. However, firms that are more dependent on a given
local market tend to be smaller firms. Hence, this may be another
measure of firm size.

The four model specifications were re-run using these three
additional variables. All four of the model specifications are shown
below and were estimated using OLS:

Model 1:

Palm = a + B: fiWilma-1+ I32 “Susi-1 + Bapmplai.-I + 54‘" den-n
+ B. annual time dummies + e,-

Mode12:
Penlmei = a + B: Wj.-I+ [52 Ma + Bapmpln-n + 34‘" depL-I

+ [3: annual time dummies + a,-

Model 3:

Pens+2 = a + 131W OWE-1 + I32 muss-I + fistPQ-l + 54‘” dept-1
+ B.annualtimedummies+ej

Model 4:
PM = a + 131W owl’s-1+ 92 1118mm + B3P1'OP18J'A + 34W dept-1

+B.annualtimedummies+s,~
The summary statistics for all markets and for the public markets
only are shown in Tables 21 & 22 respectively. The estimation
results are shown in Table 23. All models have t-statistics using
White’s heteroskedasticity consistent standard errors. All models
include annual time dummies, but their coefficients are not shown.

In all four models, the measure of managerial ownership is negative

 

In

an

an

m:

01

“1‘

69

In all four models, the measure of managerial ownership is negative
and significant at less than 5%. This implies that holding firm size
and market size constant, Wal-Mart has a harder time gaining
market share in markets with higher managerial ownership.

In Models 1 & 2, the fraction of stores in a local market
owned by private firms is negative and significant at 1.8% and 1.3%
respectively. The results also show that Wal-Mart has a harder time
gaining market share in markets that have more stores. This is to be
expected due to the physical effort of adding stores. In Models 3 &
4, the fraction of shares held by insiders is negative and significant
at 2.9% and less than 1% respectively. Again, the results also show
that Wal-Mart has a harder time gaining market share in markets
that have more stores. In Model 1, the marginal effect is that a 10%
increase in the fraction of private firms in a local market leads to
about a 1% decrease in Wal-Mart’s penetration rate. In Model 3,
the marginal effect is that a 10% increase in weighted managerial
ownership in a local market leads to about a 2% decrease in Wal—
Mart’s penetration rate.

The results suggest that firms with high managerial ownership
do make different decisions in the product market than firms with
low managerial ownership, and that Wal-Mart has a harder time

gaining market share from the firms with low managerial ownership.

I.I| ll. I'll!

70

4.4.2 An Indirect Test for Managerial Entrenchment

Morck, Shleifer and Vishny (1988) document a changing slope
when they regress managerial ownership on Tobin’s Q. They argue
agency problems decrease as ownership increases but managerial
entrenchment may set in for some range of ownership values." They
use two thresholds, 5% ownership and 25% ownership to test for the
entrenchment hypothesis. They do find some evidence of
entrenchment and agency problems using a piecewise linear
regression when they control for size, industry, and intangible
assets. They concede, however, that Tobin’s Q is undoubtedly a

noisy signal of managerial performance.

It is possible that entrenchment could affect the decisions that
managers make when fighting Wal-Mart, and thus affect Wal-Mart’s
penetration in local markets. Given this industry is very
competitive, and that there is low barriers to entry, it would be
expected that entrenchment would be unlikely in this industry. As
an indirect test of managerial entrenchment, a spline regression,
using the same thresholds as Morck, Shleifer and Vishny(1988) was
run to examine the relationship between managerial ownership and
Wal-Mart’s penetration. The test was run using the markets that had
only public firms, controlling for local market and firm size. The
model specifications are shown below and were estimated using

constrained least squares:

71
Model]:

Penm = a + 51W OWPL-I + 51di(0WPj,-1 - 95) + 52d2 (OWPL-I - .25) + 52 Md
+ ngroplgH + [34 w depjfl + [3, annual time dummies + e,-

whered1=1ifoij,-lgreaterthanorequalto 5%; otherwise=0
d2=lifoij,-1greaterthanorequalto 25%; otherwise = 0
Model 2:

Penbasej = a + [in w 0WPj.-1 + 51d1(0WPj.-1- .05) + 52d2(0WPj.-1- .25) + B: Irma-1
+ ngroplgi,‘ + [54 w dcpjfl + B, annual time dummies + c,-

whered1= l ifoij,—l greatcrthanorequalto 5%; otherwise=0
d2=1ifoij,-l greaterthanorequalto 25%;otherwise=0

The results are shown in Table 24. In both models have t-
statistics were calculated using White’s heteroskedasticity
consistent standard errors. Both models include annual time
dummies, but their coefficients are not shown. In Model 1, neither
of the threshold variables is significant individually. This implies
that there is not a statistically significant change in the slopes
between group one and group two or group two and group three. A
test of the hypothesis that the slope of the function is constant
across all three groups is tested with a joint test of 81= 82 = 0.
The null hypothesis that the slope is constant cannot be rejected at
the 99% level. In Model 2, the threshold variable at 5% ownership
is statistically significant. This means that the slope between group

one and group two is significantly different, but the slope between

 

7‘ See Fame and Jam (1933) regarding the mtrmdlmmt hypothesis.

72

group two and group three is not significantly different. This
implies that ownership greater than 5% does increase penetration by
Wal-Mart versus what it is when ownership is less than 5%. Given
that the slope coefficients are statistically different between groups
one and two, another estimation is done to estimate the slope
coefficients for each of the three groups using penbasej as the
dependent variable. The model specification is shown below and was
estimated using constrained least squares (piecewise linear

regression):

Penbasej = a + B] woijfl 0 to .05 + B; woij..1.05 to .25
+ 83 w owp,-,.1 over .25 + B. lnstrsj,-1+ [is proplgm + Ba w depj_-1
+ B. annual time dummies + e,-

where w owp5,.1 0 to .05 = w oijn if w OWpM < .05
= .05 fiwom.1h> or= .05

w own-,4 .05 to .25 = 0 if w opr is < .05
= w oij,.1 minus .05 if .05 < or = w own-,1 < .25
= .20 if w oijn > or = .25

woijfl over .25 =0ifw oijfl <.25
= wopr minus .25 if w opr > or = .25
The results are shown in Table 25. The slope coefficient is negative
on the ownership measures in all three groups in the piecewise
regression. However, it is only statistically significant in the less

than 5% ownership group.

 

73
4.4.3 New Entrant Penetration: Markets With Only Public

Incumbent Firms
This section is an attempt to determine why Wal-Mart finds
firms with lower managerial ownership to be easier prey in the
product market. Only markets with all public incumbent firms at
the beginning of Year —1 are used, as debt ratios are only available f

for public firms.

 

Firms with lower managerial ownership may have more debt in
their capital structures as debt can be used as an incentive device to
lower the agency problem. The use of debt, however, may increase
predatory behavior on the part of the other competitors.72 Wal-Mart
may find markets with high debt firms to have “softer” product
market competition. However, there also is a strand of literature
that argues that firms may take on more debt to commit to more
aggressive output behavior in the product market.73 Hence, Wal-
Mart could find competition to be “tougher” when there are a large
fraction of high debt firms in a local market. All of these papers
suggest that debt is a choice variable. This may not be the case.
There is some evidence in the sample that debt may be used as an
incentive device. Table 26 shows the simple correlation matrix for
the variables in the sample with the 233 public incumbent firm

observations. As can be seen in Table 26, markets whose stores are

 

”SeeforexampleBoltonmdeiai-fitein0990)
”Seefor-t-pleBranderandLewis

74

owned by firms with low managerial ownership are markets whose
stores are owned by firms that use more debt.

Firms with lower managerial ownership may be more
diversified. They may be “building empires” to satisfy their own
perquisites. There is some evidence in the sample that this is the
case. Markets whose stores are owned by firms with low managerial
ownership are markets whose stores are owned by firms that have a
lower fraction of their sales coming from discounting. Theoretical
work differs on its predictions here as to how this will impact the
“toughness” of product market competition. One group of models
predicts that failing divisions may be subsidized sub-optimally by
more profitable divisions.74 This would argue that diversified firms
may fight longer and harder than focused stand-alone firms. Stein
(1997) argues that firms that are diversified into related areas can
more easily judge relative profitability and may take funds away
from divisions where future prospects are relatively poor. This
would imply that diversified firms could retrench quicker when
threatened.

It is also possible that either high or low ownership firms tend
to be located in high Herfindahl markets. In this sample, high
Herfindahl markets are negatively correlated with the level of
managerial ownership of firms in the markets. It is not clear if

Wal-Mart will gain a higher or lower market share in high

 

"for exmnple see Meyer, Milgrom and Roberts (1992) and Scharfstein and Stein (1997)

75

Herfindahl markets. If the incumbents cooperate we would expect
higher penetration. Wal-Mart may expect collusion in these
markets, and the incumbents do collude. If the incumbents have
mechanisms to block entry and expansion we would expect lower
penetration in high Herfindahl markets. With respect to low
Herfindahl markets, it may be easier for Wal-Mart to expand in

these markets, as it may be easier to eliminate marginal performers

A‘—

 

due to Wal-Mart’s cost advantage. On the other hand, it may be
harder for Wal-Mart to penetrate low Herfindahl markets as
switching costs would be expected to be higher. Competitors may
be more efficient in these markets, and the marginal benefits would
be low to customers to drive farther to get better prices. This effect
may be particularly pronounced in local markets with higher income

populations.

It is also possible that other ownership incentive contracts
which can affect output and entry decisions may also impact the
penetration of Wal-Mart into local markets that consist of firms that
have such contracts. Hence, managerial ownership may still
determine new entrant penetration even after controlling for the

other variables discussed.”

Other variables that will be used as control variables as they
may impact the ability of Wal-Mart to gain market share are

incumbent firm profitability, firm size, firm dependency on the

76

local market under attack, and the size of the market. The new

variables are defined as follows:

w feel: The weighted average focus of the firms with at least one
store in local market j at the beginning of Year —1. The weights
were calculated by determining what fraction of the total stores in
local market j firm i had at the beginning of Year —1. Focus is
calculated as discount sales divided by total firm sales for the fiscal
year ending at the beginning of Year —1 for firm i.76 This variable

is reported in decimal form.

w debt,: The weighted average total debt ratio of the firms with at
least one store in local marketj at the beginning of Year —1. The
weights were calculated by determining what fraction of the total
stores in local market j firm i had at the beginning of Year —1. The
total debt ratio is calculated as total assets (Compustat Item #A6)
minus Stockholder Equity (Compustat Item #A216) divided by total
assets (Compustat Item #A6). It is thus the book value of total debt

divided by the book value of total assets and was calculated at the

 

7’ For exanpleseeKhm(l998)andFershtman andJudd(1987)
”Somethnesthisratiowasgreaterthm l foragivmfirmiastotalfirmsalesmddiscmmtsalmcamefran
difl‘erentsomces. Whmevcrthishappmedlwasused.

—_'r-—_ '- I ~11

77

beginning of Year —1 for firm i." This variable is reported in

decimal form.

w ssqj: This is a measure of profitability. The weighted average
inflation adjusted sales per square for the firms with at least one
store in local market j at the beginning of Year -1. Inflation

adjusted sales per square foot is measured in 1974 dollars for the

fiscal year ending at the beginning of Year —1 for firm i."a

mod herfj: This is a measure of market concentration. It is
calculated as the sum of the squared market shares of the firms in
each local market j. The market share of each firm in a local market
j is defined as the fraction of stores in the local market owned by

firm i.”

The first set of tests in this section will use “penj” as the
dependent variable. The model specification to be estimated using

OLS is as follows:

Pen-,2 = a + B: W f°°i.-l + 132 W debts-1 + B: W OWPi-l + 54‘” dept-1 + 135 W “Qt-I
+ Bo ploplgm + [37 mod herfi,.1 + B; lustre-,1 + B. annual time dummies + a,-

 

"MfirmnflecksandSE.NidmlscmldnotbefmmdonCompuflat. Thisratiowascalaflatedfian

statements located in Moody’s Industrial Manuals instead.

7' Due to the low margins and homogenous goods in this indmIry, higher turnover lends tohigher

grofitability. This measure of profitability is the only one available for both public and private firms.
Recallthattheverysmalldiainsmenotmthesample. Onlychainsthatmetthesizea'itciastatedinthe

dataswtiona'einthesample.

78
The level of debt is not available for privately held firms. As

mentioned earlier, Wal-Mart entered 488 local markets between
1981 and 1994. All local markets were dropped from the sample
that had any foreign held firms in them, or that had no stores in
them at the beginning of Year —1. This left 471 local markets.
About half of these markets, 233 to be exact, had only public firms
in them at the beginning of Year —1 and had debt and ownership
data available for all firms in the local market at the beginning of
Year -1. The first few specifications are run using these 233 local
markets.

Table 27 shows the summary statistics for all of the variables
used in this first specification. Table 28 shows the estimation
results for three models. OLS estimation was used in all of the
models.” Annual year dummies were included in all three of the
model estimations, but their coefficients are not shown. In all three
models the Cook-Weisberg (1983) test rejects the null hypothesis of
no homoskedasticity at 5%. Hence, the standard errors used in all

three models are White’s heteroskedasticity robust standard errors.

Model I is the full model where the estimation is run using all
of the variables. The results show that Wal-Mart has an easier time
penetrating local markets if the incumbent firms have high debt

levels prior to new entry by Wal-Mart into local market j. This

 

”SincemedepcndmtvariableisbmmdedbetweenwomdoneaTobitestimationprocedmemightbe
appuprime. I-lowever,observationsateitherextremearerare,soOLS estimationisusedasitiseonsidaed

79

result is significant at less than 1% and is consistent with predatory
behavior on the part of the new entrant. A ten percent increase in
the weighted average debt ratio of the market will lead to about a
7% increase in Wal-Mart’s penetration rate. Wal-Mart also has an
easier time gaining market share in local markets that have higher
Herfindahls prior to Wal-Mart entering, controlling for market size.
This result is also significant at less than 1%. Ex-ante it is unclear
whether Wal-Mart would have an easier or a harder time penetrating
high Herfindahl markets. The results suggest that either there is co-
operative behavior in the higher Herfindahl markets or there are

higher switching costs in the low Herfindahl markets.

Wal-Mart has a harder time penetrating local markets if the
incumbent firms are highly focused prior to entry by Wal-Mart into
the local market. This result is significant at 6.6%. A ten percent
increase in the weighted average focus of the firms in a market
leads to a 1.2% decrease in Wal-Mart’s penetration rate. Meyer,
Milgrom and Roberts(1992), and Scharfstein and Stein (1997) both
argue that firms may sub-optimally finance poorly performing
divisions. The implication is that there is more investment in
poorly performing divisions and less investment in good performing
divisions relative to stand-alone firms. From Chapter 3 it is known
that diversified firms are less likely to expand than focused firms.

It may also be that they are less likely to initiate price wars or

 

tobeconsistent.

80

influence zoning decisions. The findings here would be more
consistent with the predictions for good performing divisions, as it
appears that the diversified firms do not fight as hard as the focused
firms. However, this does not seem plausible given that many of
these firms entered Chapter 11 and/or exited the industry after
competing against Wal-Mart. This would imply that future

prospects would have been poor for many firms.

The results are more consistent with Stein who argues that
funding may be restricted to those divisions that are expected to
perform poorly relative to other divisions. He argues that this is
most relevant for firms that are diversified in related industries.
Most of the diversified firms in this industry are diversified
retailers and thus are diversified in related areas. The results are
also consistent with diversified firms lacking a credible threat to

fight, as they have lower opportunity costs.

Once firm debt levels, firm focus levels, firm profitability,
market Herfindahls have been added to the regression specification,
the significance of the managerial ownership variable disappears.
From the correlation matrix in Table 26 it can be seen that markets
whose stores are owned by firms with higher managerial ownership
have stores with lower debt levels, and higher focus than markets
whose stores are owned by firms with lower managerial ownership.

Since lower debt and higher focus each marginally make Wal-Mart’s

81

penetration more difficult, this appears to be what was driving the

ownership results before the additional variables were entered.

With respect to Model 1, the Ramsey RESET test cannot reject
the null hypothesis that the model has no omitted variables at the
5% level. Multicollinearity does not appear to be too large a
problem. The VIF’s for all of the variables other than the annual
year dummies are less than three. The highest VIF is 16.0 for one
of the annual year dummies. Regional dummies were also put in the
model to see if there was any un-modeled market heterogeneity.

The United States was divided into six regions, the Northeast,
Southeast, West, Foundry, Mid-West, and the Empty region. Five of
them were put in the regression, and none of them was significant.
An F test was done to test the joint significance of the regional
dummies. The null hypothesis that the coefficients are all zero
cannot be rejected at the five percent level. Thus, the results with

the regional dummies included in the specification are not shown.

Local markets that have a high fraction of stores owned by
focused firms also tend to have a high fraction of stores owned by
profitable firms as these two variables are positively correlated.
The model was re-estimated first dropping the profitability measure
w ssqj (Model 2), then dropping the focus variable, w fOCj (Model
3). When one of these variables is dropped, the significance of the

other is increased. When the profitability measure is dropped, the

82

focus variable becomes significant at less than 1%. When the focus
variable is dropped, the profitability variable remains negative but
becomes significant at 2%. This implies that Wal-Mart has a harder
time penetrating markets with focused firms in them, and these
markets tend to have more profitable firms in them. Note that none
of the other variables’ coefficients or significance levels change

much when either of these variables is dropped.

A graph was done plotting the residuals versus the fitted
values. There was no curvature or pattern in the residuals. The
assumption that the penetration of Wal-Mart is linear in the

independent variables appears to be reasonable.

The results are basically consistent with the results in Chapter
3 and imply that expanding against Wal-Mart in a local market
reduces Wal-Mart’s penetration in that market. In Chapter 3, the
findings show that incumbents with higher debt, and higher market
shares (higher market Herfindahl’s) are less likely to expand against
new entrant, Wal-Mart, in local market j. Wal-Mart gains higher
market share in markets that consist of high amounts of firms with
these traits. The findings in Chapter 3 show that incumbents with
higher focus and higher profits are more likely to expand against
new entrant, Wal-Mart, in local market j. Wal-Mart gains lower
market share in markets that consist of high amounts of firms with

these traits less. However, another result from Chapter 3 was that

83

incumbent firms with high insider ownership or low dependence on
a market were less likely to expand against new entrant, Wal-Mart
in local market j controlling for the other variables.

It is assumed that when Wal-Mart enters a new market that
they disturb the equilibrium and a new equilibrium will eventually
emerge. It is not clear that the new equilibrium is in place in a
local market by the beginning of Year +2. As a robustness check,
the results were re-estimated using Wal-Mart’s market share in local
market j at the beginning of Year +4. However, a potential problem
exists with a longer time window. With a longer window, various
factors may change and other variables may determine Wal-Mart’s
market share other than the firm and market specific variables
measured at the beginning of Year —1. Wal-Mart’s market share is
measured by determining the fraction of total stores in local market
j at the beginning of Year +4 that are Wal-Marts. The dependent
variable will be denoted as “penfivej”. All other variables remain
unchanged. All observations where Wal-Mart is first seen
competing in a local market at the beginning of 1993 or 1994 now
have to be dropped as I only have store location data until the
beginning of 1996. This lowers the number of observations from
233 to 176. The model specification to be estimated using OLS is

as follows:

penfivej,4=a+ B. wfoc,;.1+ B;wdebt,-,.1+ Bawowp,“ + B4Wdcpj..| + Bswssqid
+ 05 proplgm + [Emodherfjn + B; lustre-,1 + B.annualtime dummies+ s,-

34

Summary statistics are located in Table 29. Table 30 shows the
estimation results for three models. OLS estimation was used in all
of the models." Annual year dummies were included in all three of
the model estimations, but their coefficients are not shown. In all
three models the Cook-Weisberg (1983) test rejects the null
hypothesis of no homoskedasticity at 5%. Hence, the standard
errors used in all three models are White’s heteroskedasticity robust

standard errors.

Model 1 is the full model where the estimation is run using all
of the variables. All of the coefficients have the same sign as they
did using the shorter event window except the coefficient on the
weighted ownership variable, which is still not significant. The
results are stronger for the coefficients on the weighted firm debt
variable, the Herfindahl variable, and the weighted firm
profitability variable (wssq).

The coefficient on the firm focus variable is no longer
significant, but on inspection, the correlation of this variable with
weighted firm profitability is higher in this smaller sample. Due to
the correlation between these two variables, the model was re-
estimated first dropping the profitability measure w 83% (Model 2),
then dropping the focus variable, w fOCj (Model 3). When one of

these variables is dropped, the significance of the other is

 

"SincethedepmdmtvariableisbormdedbetwemwomdmeaTobitestimationprocedmemightbe
approprime. I-Ioweva,observationsateithcrexu‘emearerare,soOLSestimationisusedasitisbelievedto

85

increased. When the profitability measure is dropped, the focus
variable becomes significant at less than 1%. When the focus
variable is dropped, the profitability variable remains negative but
becomes significant at less than 1%. This implies that Wal-Mart has
a harder time penetrating markets with focused firms in them, and
these markets tend to have more profitable firms in them. Note that
as before, none of the other variables’ coefficients or significance

levels change much when either of these variables is dropped.

It would be reasonable to conclude that widening the event
window from three to five years made the results slightly stronger.
It appears that a new equilibrium was relatively established after

the three-year period.

If one assumes that the equilibrium number of stores is static,
then a third possible dependent variable would be to determine what
fraction of the stores in a market at the beginning of Year —1 are
owned by Wal-Mart at the beginning of Year +2. As a robustness
check, the results were re-estimatcd using the number of stores that
Wal-Mart has in local market j at the beginning of Year +2 divided
by the total number of stores in local market j at the beginning of
Year —1. This dependent variable will be denoted as “penbasej”.

All other variables remain unchanged. The model specification to be

estimated using OLS is as follows:

 

be consistent.

86

penbase5=a+fll Wfocikl + B:wdebt,-,1+ B3W0WPj.-i +54Wd€pj,-i + BsWSSqi-i
+ [36 P1013183}: + B7modhert},. + B; lnstrs,-,.1+ Biannualtime dummies+aj

The summary statistics are the same as those shown in Table
27, as these are the same observations used in the first
specification. Only the dependent variable has changed. Table 31
shows the estimation results for three models. OLS estimation was
used in all ofthe models. Annual year dummies were included in
all three of the model estimations, but their coefficients are not
shown. The standard errors used in all three models are White’s

heteroskedasticity robust standard errors.

Model 1 shows the results using all ofthe variables. The
results are similar to earlier results. Wal-Mart gains a higher
fraction of the number of stores that existed at the beginning of
Year —1 by the beginning of Year +2 in local markets that have
higher Herfindahls and that have firms with higher debt. As
before, having focused firms in a market increases the “toughness”
of a market as Wal-Mart gains a lower fraction of the number of
stores that existed in the beginning of Year -1 in local markets that
have highly focused firms in them. In this case the dependent
variable does not adjust for increases in the number of stores by any
other firms in the event window. It assumes that a local market is
in equilibrium and will stay the same size. It makes sense that Wal-

Mart would want to add more stores to markets that have firms that

 

87

are more profitable with this dependent variable. It may be that
these firms are more profitable because they are in markets that are
under-stored. The competitors may also be expanding in these
markets. The other dependent variables, pen,- and penfivej
automatically adjust for this possibility. In Models 2 & 3, where
weighted sales per square foot and weighted firm focus are each
dropped, only market herfindahls and market size remain

significant.

4.4.4 Incumbent Penetration: Markets With Both Public and
Private Firms

In Section 4.4.1, tests showed that Wal-Mart had a harder time
gaining market share in local markets the higher the proportion of
stores in the market owned by private firms. When looking at local
markets consisting of only public firms, Wal-Mart had a harder time
gaining market share in local markets that had firms with higher
weighted managerial ownership. These results held even after
controlling for local market size and firm size. In Section 4.4.3, the
results using local markets with only public firms were examined
more closely to determine why Wal-Mart had a difficult time
gaining market share in the markets consisting of stores with high

levels of firm managerial ownership. In this section all local

 

88

markets are examined, including those with private firms, to see if
private firms act similarly to high ownership public firms.

The specifications run on markets with stores owned by only
public firms, are re-estimatcd using all markets and making some
variable changes. As in the earlier tests done in Sections 4.4.1, the
fraction of private firms in the local market is used to see if having
extreme measures of managerial ownership is important. The
information needed to create the focus variable was not available
for all of the private firms, and thus this variable was not used in
the first tests to be shown below. Debt ratios are not available for
private firms, but a LBO dummy variable is used if a market has at
least one store in it at the beginning of Year -1 that is owned by a
firm that has recently done a LBO. This variable is defined as

follows:

LBO dummy]: Equals 1 if local marketj has at least one stores
owned by a firm that has done a LBO in the three years prior to the

beginning of Year —1; Otherwise = 0.

All other variables are the same as were used in earlier
specifications. The models to be estimated are as follows and differ

only with respect to the dependent variable:

89

Model 1:
Penj.+2 = a. + 131130 dummyi + 52 W 134 + 153 W delta-1 + 34 W 53%
+ 35 1910131ng + 56 find heft} + 57 “Susi-I
+B.annualtimedummies+sj

Model 2:
Penfivej,“ = a + BILBO dummyj + Bzfi‘acprim + B;wdep,-,.1+ [34 w ssq;
+ 135 Pmplgj.-1 + 136 End herfi + 137 lnstrsiq

+ 8. annual time dummies + e,-

Model3:
Penbasej=a+ BlLBOdUIDij +Bzfi’acprij,.1+B3wdep,-,.i +B4wssqi

+BstP18i-1 + 56 1110411611} + B71nStTSj.-l
+Btannualtimedummies+ej

A potential problem exists when including markets with
privately held firms in the sample. The debt ratios for the privately
held firms are not available. Also, the focus variable is not
available for all private firms. This could least to biased estimates
of the remaining coefficients.

Table 32 shows the simple correlation matrix using the
observations for all local markets. These relationships between
variables can be compared to those using local markets consisting of
stores owned by only public firms in Table 26. The correlations
between markets with a high fraction of stores owned by private
firms in them and other variables are very similar to the correlations
between markets with stores owned by firms with high managerial
ownership and those variables. Local markets with a high fraction
of stores owned by private firms tended to be low Herfindahl

markets, have less profitable firms in them, and have smaller firms

90

in them. However, the correlations with profitability and
Herfindahl are close to zero. The markets with stores owned by
public firms only, with high managerial ownership, tended to have
smaller firms in them, have less profitable firms in them, be in low

Herfindahl markets, have more focused firms in them, and have

firms with lower debt in them. E”
The summary statistics are shown in Table 33. Table 34

shows the estimation results using the three different dependent

variables. OLS estimation was used in all of the models. Annual 5

 

year dummies were included in all three of the model estimations,
but their coefficients are not shown. The standard errors used in all
three models are White’s heteroskedasticity robust standard errors.

In all three models, the fraction of private firms in the market
is negative and significant at 1.4%, less than 1% and at 6.6%. This
implies that Wal-Mart has a harder time penetrating markets with a
high fraction of private firms in them. Modified Herfindahl is still
positive and significant at less than 1%, and the size of the market,
lnstrsj, i is still negative and significant at less than 1%. The LBO
dummy is not significant in any of the specifications.

The specifications can be run adding the focus variable. Six
local markets have to be dropped due to a lack of information
regarding the focus variable for at least one firm in the local

market.

91

Modell:
Peni+2=a+ Bifi'acprii-I+Bzwf0¢i+fisWd¢Pi4+l54WSSqi
+BsPr°P18m + Bomodhert} + B7m-I
+ Biannual time dummies + 8]-
Mode12:

Pem+2=a+Biwowm.i + Bzwfoci4+ B3Wdepi-1+B4WSSQj,-l

+135 Proplai-l +Bsm0dherfi.-1+Bilnstrsi-i

+ B, annualtirne dummies+ c,-
The results are shown in Table 35. In both specifications, the
managerial ownership variable is negative and~significant at less
than 1%. This implies that Wal-Mart has a harder time penetrating
markets with a high fraction of stores owned by private firms, and a
harder time penetrating markets with only public firm owned stores
in them when the firms owning the stores have high managerial
ownership. This again supports that private firms are apparently
taking similar actions, and Wal-Mart finds these actions to be
“tougher”. In both specifications Wal-Mart has a tougher time
gaining market share in markets that have more profitable firms in
them. When the debt variable is omitted, the focus variable
becomes positive and significant. In the earlier specification when
debt was included for the public only markets, focus was negative
and significant. The switch in the sign on the focus variable is
probably due to an omitted variable bias. Given that the public and
private market coefficients are very similar to those for public firms
only, it again appears that private firms are taking actions similar to

high managerial ownership public firms.

 

92

4.4.5 Herfindahls and Profitability

Earlier results showed that Wal-Mart gained higher
penetration in high Herfindahl markets then in low Herfindahl
markets. It was unclear whether this was due to more co-operative
behavior on the part of the incumbents in the high Herfindahl
markets or if it was due to higher switching costs in the low
Herfindahl markets. It would be beneficial to empirically test to
see if high Herfindahl markets had high profits. If they do, this
would support co-operative behavior in these markets. It would also
be instructive to see what happens to market Herfindahls and market

profits when Wal-Mart enters a local market.

The correlation matrices show a positive simple correlation
coefficient between market Herfindahl’s and the profitability of the
firms that are in the high Herfindahl markets (see Tables 26 & 32).
The problem with this is that the Herfindahl is based on only the
local market j, while the firm profitability is based on the firm—wide
profits ofthe firms who have stores in local market j and is not the
firm profits in market j alone. Unfortunately, profits at the 3-digit
zip code level are unavailable, The empirical results shown earlier
demonstrate that Wal-Mart ends up with a higher market share in the
markets that have high Herfindahls. This provides some evidence of
profit potential in these markets and cooperation. If Wal-Mart had

a lower market share in high Herfindahl markets after a few years of

93

competing in those markets, this would provide some evidence that
the incumbents in these markets had effective mechanisms to block
entry and expansion.

Discount Merchandiser, a trade journal for the discount
department store industry, reports sales per square footage of
selling space for standard discount department stores for every
State, for every year during the sample period. I calculated
Herfindahl’s for every local market for the earlier tests that were
reported. The arithmetic average of the Herfindahls for all of the
local markets within each State was then calculated. The ending
result was the creation of the average local market Herfindahl for
each State for each year. With a measure of profitability per State
and a measure of market concentration per State for each year, tests
can be done to see if higher Herfindahl markets have higher profits.

The variables of interest are defined as follows:

herfnz The arithmetic average of the local market (3-digit zip code)
modified Herfindahls as defined in section 4.4.2 for every 3-digit

zip code in State i for year t.

infssqm The sales per square footage of selling space for standard

discount department stores for every State i, for every year t as

reported in Dieeeunr Merchandiser.

 

94

propwalm The number of Wal-Marts as a fraction of the total
number of stores in State i during year t. (Stated in percentage

form.)

propwal" sqd : This is propwali, raised to the second power.

Table 36 shows summary statistics for these variables. The data
consists of observations on all fifty states for 1980 — 1995. The first
specification tests to see if States with higher average local market
Herfindahls have higher profits. The model to be estimated by

pooled OLS is:

infssqg=a+filherfh +B,annualyeardummies+eh

The results are shown in Table 37. Annual year dummies were put in
the specification to control for any aggregate time effects. The
results show that States with higher average local market
Herfindahls have higher profits (inflation adjusted sales per square

foot).

The second specification tests to see if the proportion of Wal-
Marts in a State determines the average herfindahl for the State.

The proportion of Wal-Mart’s in the state squared is also included

95

to control for any non-linearities. The model to be estimated by

pooled OLS is:

herfa=a+Bipropwalh +szropwala. sqd+Bgannualyeardummies+au

The results are shown in column two of Table 37 and imply a u
shaped effect on Herfindahl. The Herfindahl measure falls when the
proportion of Wal-Marts in a State increases but then rises.

The third specification tests to see if the proportion of Wal-
Marts in a State determines the profitability levels in the State. The
proportion of Wal-Mart’s in the state squared is also included to
control for any non-linearities. The results imply a u shaped effect
on profitability within the State. The profitability in the State first
falls when the proportion of Wal-Marts in a State increases but then

rises.

4.5 Summary

This chapter explored the relationship between firm and
market specific characteristics and Wal-Mart’s ability to gain
market share. The findings suggest Wal-Mart had a difficult time
gaining market share in markets with a high proportion of high
ownership firms. This result holds even after controlling for firm

and market size and whether the high ownership firms are publicly

96

or privately held. There does not appear to be any support for the
entrenchment hypothesis, but this is not surprising given that this is

a competitive industry with low barriers to entry.

The results also support predatory behavior on Wal-Mart’s
part when entering markets whose stores are owned by firms with
high amounts of debt. Interestingly, Wal-Mart also has an easier
time penetrating markets whose stores are owned by diversified
firms. There is also some evidence that Wal-Mart has a tougher
time gaining market share in local markets whose stores belong to
firms that are more profitable.

Another finding is that Wal-Mart gains higher market share in
more highly concentrated markets. Evidence is presented that these

markets are more profitable as well.

 

CHAPTERS

Summary

In this dissertation I extend existing research in the product

 

:-
market literature area by studying how managerial ownership as
well as other firm and market specific characteristics affect firm
aggressiveness in the product market. Very little empirical work *-

has been done to test varying firm specific determinants of product
market interactions.

I find that several firm and market specific characteristics do
determine investment decisions on the part of incumbent firms when
they face a new competitive threat. I find that private firms are less
likely to expand then public firms. I also find that public firms
with high managerial ownership are less likely to expand then
public firms with low managerial ownership. I present evidence
suggesting the results are due to differences in managerial
ownership and not differences in access to capital, or organizational
form. Other results that emerge are that a higher level of debt
decreases the likelihood of expansion, a higher level of focus
increases the likelihood of expansion, and the more dependent a
firm is on the market under attack, the greater the likelihood of

expansion.

97

 

98

I also find that several firm and market specific
characteristics determine Wal-Mart’s ability to gain market share.
Wal-Mart has a more difficult time gaining market share in local
markets whose stores are owned by firms with high managerial
ownership. This result holds even after controlling for firm and
market size. This provides support for agency theory. The
implication here is that managerial incentives and/or monitoring
cannot totally mitigate agency problems. A spline function was also
estimated to test for the entrenchment hypothesis, but no evidence
supporting it was found.

Other results are that the new entrant, Wal-Mart, does prey on
firms with high debt levels. Firms that are focused as well as the
more profitable firms also tend to deter Wal-Mart’s penetration into
their markets.

Higher Herfidahl markets are more profitable, and Wal-Mart
gains higher market share in these markets. This provides support

for cooperative behavior occurring in these markets.

APPENDICES

APPENDIX A

 

APPENDIX A- TABLES

Table l

Chains Attacked by Wal-Mart

Thechainnamesareshownbelow. Ifmorethanonefirmownedachaindm-ingthc
periodflnnthcfirmnameisshowninparentheses.

Ames

Anderson’s

Bargain Town USA (Bargain Town)
Bargain Town USA (Kinder-Care)
Bradlees

Caldor

Clover

Cook United

Danners

Duckwall-Alco

Gold Circle & Richway
Fishers Big Wheel

Fred Meyer

Gambles (Gamble-Skogmo)
Gambles (Wickes)
Gaylords

Gee Bee

GI Joe

Giantway

Grandpa’s

Harts (Big Bear Inc)

Harts (Penn Trafiic)

Hecks

Hills

Howard Brothers

Jacks

Jamesway

Jefl‘erson Ward

Kings

Kmart

Kuhns
Lechmere
Magic Mart
Maloneys
Marshalls
Maxway
Meijer
Murphy G.C.
Pamida
Prangeway
Quality
Rich’s
Roses
Schottenstein
SE Nichols
Shopko
Sky City
Stein Mart
Stuarts
Swallens
Target
TG&Y
Treasury
Van Leunens
Variety W.
Venture
Whitney
Woolco
Zayre

 

100
Table 2
Summary Statistics: Public and Private Incumbent F inns.
Summarystatisticsfor15066rmimarketjpairswhereWa1-Martwasfirstseencompetingina

3-digitzipcodennrketjbetween1976-1994,andinwhichfirmiwasalreadyanincumbem. 59
firmsarerepresentedinthesample.

 

 

 

 

 

 

 

 

 

 

Englanatgry Variables Mean Minimum Maximum Std Dev
Private/Public Dummy i .1315 0 1.0 .3380
1 = private 0 = public
LBO/No LBO Dummy: .0126 0 1.0 .1116
1 = LBO 0 = No LBO
Chain Size 3 636.4655 4 1811 693.8854
Inflation Adjusted Sales 60.5824 15.0816 204.7293 20.3759
per Square Foot i
Market Share of FirmI in .2925 .0256 1.0 .1960
Marketj
Modified Herfindahl Index .3399 .1224 1.0 .1592
1
Dependence of Firm i on .0164 .0006 .3333 .0323
Market 1
Strength of Wal-Mart .71 18 0 1 .0 .4531
Dummy:
1=1986-1994 0 =1976-
1985
Avg. Indusu'y Growth in .0227 .0062 .0469 .0098

 

 

 

 

 

 

Square Feet

 

101

Table 3
Simple Correlation Matrix: All Observations Public and Private

Using 1506firmi,marketjpairswhereWal-Martwasfirstseencompetingina3-digitzipcode
marketjbetween1976-1994,andinwhichfirmiwasalrendyanincumbent. 59firmsare

 

 

 

 

 

 

 

 

 

 

 

representedinthesample.
Pri/LBO/ChainS/SqutMod.Depi,-Decadelnd.
Pllb NO 82" F15 Shrg Herf Dum. GTOWth
Dum LBO Index,-

Dums

Pri/

Pub 1.00

Duma

LBO/

No .220 1.00

LBO

Dumi

Chain -.322 -.086 1.00

82,-

S/Sq -.021 -.137 .356 1.00

Fig

Mk1 -.131 -.037 .351 .058 1.00

Shrfi

Mod. 1.00

Herf. -.116 -.005 .225 .060 .765

Indeg '

Depfi .508 .006 -.373 -.046 -.095 -.162 1.00

Decade .096 .072 .094 .048 -.017 -.038 .004 1.00

Dum.

Ind. -.114 -.092 .024 .007 .110 .124 -.054 .082 1.00

Growth

 

 

 

 

 

 

 

 

 

 

 

 

102
Table 4

Univariate Probit: Public and Private Incumbent Firms

Maxinmmlikelihoodestimationresults. Thedependentvariableyij=liffirmieverexpands
duringYears—lto+2,otherwiseyg=0. Thereare1506firmimarketjobservatiom. Expansion
occurredin219 observations. 59firmsarereprescntedinthe sample. Thevaluesinparentheses
amp-values. Marginalcffectsarecalculatedatthemeamofthex’s. Inthecaseofdummy
var'mbles,themarginalefi'ectsforthediscretechangeofthedummyfiom0to1isshownbelow
theclassicmarginalefi‘ectvalue.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Model 1 Marg. Effects Model 2 Marg. Effects
Variables Coefficients dprob[y=1]/dx Coefficients Dprob[y=1]ld
x
Constant -1.4148 W” - 0.3131 -1.4315 ""” -0.3176
(0.000) (0.000)
Private/Public Dummy: - 0.4385 ‘” -0.0970 -0.4297 “W -0.0953
1: private 0 = public (0.009) -0.0807 (0.0097) -0.0800
LBO/No LBO Dummy; 0.7169 ” 0.1586 0.7198 ” 0.1597
1 = LBO o = No LBO (0.042) 0.2160 (0.041) 0.2176
Chain Size; 0.00012“ 0.00003 0.00008 0.00002
(0.087) (0.2137)
Inflation Adjusted Sales 0.0048 ” 0.0011 0.0050 ” 0.0011
Per Square F cot. (0.024) (0.017)
Market Share of Firm i in -0.4230 ‘ - 0.0936
Market j (0.069)
Modified Herfindahl -0.2182 -0.0484
Index,- (.4182)
Dependence of Firm i on 3.6565 ” 0.8091 3.3618 ” 0.7460
Market j 40.01 1) (0.0195)
Wal-Mart Strength -0.2440 ‘” - 0.0540 -0.2407 ”’ -0.0534
Dummy (0.005) - 0.0570 (0.006) -0.0563
1 8 ‘86-’94 0 = ‘76-‘85
Industry Growth in Sq. Ft -8.4801 ‘ -1.8765 -8.9678 ” -1.9899
(0.0505) (0.038)
Significance Level 0.00008 0.00025
(HO: B =01
Pseudo R-squared .317 .315
N 1506 1506

 

 

 

 

 

 

 

Statistically significant at 10% ‘, at 5% "', at 1% ”*

A LM test statistic was done to check for multiplicative heteroskedasticity using all ofthe right
hand side variables except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 95% level for the specification using market share. The null hypothesis of
homoskedasticity cannot be rejected at the 97. 5% level for the specification using modified
herfindahl. The specifications were re-run using White’s robust estimator of variance. The p-
values are virtually identical to those shown above for both specifications. Pseudo R-Squared
was calculated in Limdep using a formula given by Zavoina and McKelvey (1975).

103
Table 5

Univariate Probit: Public and Private Incumbent Firms Time Dummies Inchided

Maximum likelihood estimtion results. The dependent variable Yij=1 iffirm i ever expands
duringYears-l to+2; MCYfij = 0. Thereare1506 firmimarketj observations. Expansion
occurred in 219 observations. 59 firms are represented in the sample. The values in parentheses
are p-values. Marginal effects are calculated at the mean of the x’s. In the case of dummy
variables, the nmrginal effects for the discrete change of the dummy fiom 0 to 1 is shown below
the classic marginal efl‘ects value. The sample was broken into 8 two-year periods and 1 three-

 

 

 

 

 

 

 

 

 

 

 

 

 

year period. 8 time dummies were included in the specification, but the coefficients are not F
shown below.

Explanatory Variables Coefficients Marginal Coefficients
Effects Robust p-values
dprob1y=llldx (White’s)

Constant - 1.8029 ”" - 0.3871 - 1.8029 ”" L-

(0.006) (0.005)

Private/Public Dummy; - 0.4332 ” -0.0930 - 0.4332 ”"

1: private 0 = public (0.011) ~0.0772 (0.009)
LBO/No LBO Dummy; 0.5702 0.1224 0.5702
1 = LBO o = No LBO (0.113) 0.1598 (0.105)

Chain Size; 0.00013 " 0.00003 0.00013 "
(0.082) (0.073)

Inflation Adjusted Sales 0.0060 ”" 0.0013 0.0060 ”*

Per Square F cot; (0.007) (0.004)

Market Share of Firm i in - 0.5089 " - 0.1093 - 0.5089 "'

Market j (0.034) (0.040)

Dependence of Firm i on 3.7834 ”* 0.8123 3.7834 ‘"“

Market j (0.009) (0.005)

Wal-Mart Strength Dummy - 0.2311 - 0.0496 - 0.2311

1 = 1986-1994 (0.317) - 0.0524 (0.317)

0 = 1976-1985

Industry Growth in Sq. Ft - 1.7699 -0.3800 - 1.7699

(0.900L (0.897)

Significance Level 0.00000 0.00000

(Ho: 5 =01

Pseudo R-squared .336

N 1506

 

 

 

 

 

 

Statistically significant at the 10% *, at 5% “, at 1% ”W

A LM test statistic was done to check for multiplicative heteroskedasticity using all of the right
hand side variables except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 99% level. None of the 8 time dummies is significant. The specification was re-
nm using White’s robust estimator of variance. The p-values are shown above. Pseudo R-
Squared was calculated in Limdep using a formula given by Zavoina and McKelvey (1975).

104
Table 6
Summary StatisticszPrivateIncumbentFirm.
Summarystatisticsfor l98firminnrketjpairswhereWal—Martwasfirstseencompetingina3-

digitzipcodemarketjbetween1976-l994,andinwhichfirmiwasalreadyanincumbent. 24
firmsarerepresentedinthesample.

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Mean Min. Max. Sthev.
LBO/No LBO Dummy .0758 0 1.0 .2653
1 = LBO 0 = No LBO

Chain Size 62.19 4 162 49.79
Inflation Adjusted Discount 59.50 19.98 204.73 34.08
Sales per Square Foot

Market Slme of Firm i in .2265 .0476 1.0 .1800
Market j

Dependence of Firm i on .0585 .0062 .3333 .0642
Market j

Strength of Wal—Mart Dummy .8232 0 1.0 .3824
l =1986-1994 0= 1976-1985

Avg. Industry Growth in .0198 .0062 .0376 .0085
Square Feet

 

 

 

105
Table 7

Univariate Probit: Private Incumbent Firms

Maximum likelihood estinntion results. The dependent variable Yij = 1 if firm i ever expands
duringYears-l to+2; otherwiseyg=0. Thereare 198 firmimarketj observations. Expamion
occurred in 21 observations. 24 firm are represented in the sample. The values in parentheses
are p-values. Marginal effects are calculated at the means of the x’s. In the case of dummy
variables,themarginalefl'ects forthediscretechange ofthedummy fiomOto 1 isshownbelow
the classic marginal efl‘ects value.

 

 

 

 

 

 

 

 

 

 

 

 

r-
Explamtory Variables Coemcients Margiml Efl‘ects
dProb[y=1]/dx
Constant -l.8529 "" -0.3034
(0.001)
LBO/No LBO Dummyi 0.8290 " 0.1358
1 = LBO 0 = No LBO (0.057) 0.2038 7
Chain Size; 0.0031 0.0005
(0.390)
Inflation Adjusted Sales 0.0076 ” 0.0012
Per Square Footi (0.03 3)
Market Share of Firm i in Market -0.3192 -0.0523
j (0.725)
Dependence of Firm i on Market j 3.5922 0.5883
(0.1024)
Decade Dummy -0.3921 -0.0642
1 = 1986-1994 0 = 1976-1985 (0.231) -0.0757
Industry Growth in Sq. Ft. -0.0589 -0.0096
(0.997)
Significance Level (Ho: [3 =0) 0.1607
Pseudo R-squared .3176
N 198

 

 

 

 

 

Statistically significant at 10% ‘, at 5% ”, at 1% "W

A LM test statistic was done to check for multiplicative heteroskedasticity using all of the right
hand side variables except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 95% level. The specification was re-run using White’s robust estimator of
variance. The p-values are virtually identical to those shown above except for inflation adjusted
sales per square foot which has a p-value of.018 and the dependence offirm i on marketj which
has a p-value of .053. Pseudo R-Squared was calculated in Limdep using a formula given by
Zavoina and McKelvey (197 5).

106
Table 8
Summary Stat'mtics: Public Incumbent Firms
Summary statistics for 1209firmi, marketjpairswhere Wal-Martwasfirstseencompetingina

3-digitzipcodemarketj between 1976- l994,andinwhichfirmiwasalreadyanincumbem. 38
public firms are represented in the sample.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Mean Min. Max. Std. Dev.
% Insider Ownershij i 0.1056 0.0009 .7548 0.1763
Total Debt to Assets ; 0.6445 .2894 1.1530 0.1537
Focus ; 0.7360 .0077 1.0 0.3090
Chain Size i 765.65 19 1811 715.81
Inflation Adjusted Sales per Sq. 62.18 15.08 110.55 17.06
Ft. i

Operating Profit Margin 5 (11111929 .0649 -.0598 .5247 .0753
‘Operating Profit Margim am a!) .0535 -.0598 .1740 .0266
Market Slmre of Firm i in Mkt j 0.2988 .0256 1.0 0.1934
Dependence on Market j 0.0099 .0006 .1739 0.0167
Strength of Wal-Mart Dummy 0.7378 0 1.0 0.4400
1 = 1986—1994 0 = 1976—1985

Avg. Industry Growth in Square 0.0228 .0062 .0376 0.0097
Feet

 

 

 

 

 

’NOTE: Onefirmhadanoperatingprofitmarginthatwasmuchlargerthanthatoftheother
firmsinthesample. Itwasdroppedinthespecificationthatusesoperatingprofitmargin.

 

 

107
Table 9

Simple Correlation Matrix: Public Firm Observations
Using 1209 firmi, marketjpairswhereWal-Martwasfirstseencompeting ina3-digit

zipcodemarketjbetween1976-1994,andinwhichfirmiwasah'eadyanincumbent.
38firmsarerepresentedinthesample.

 

% Debt ChainS/Sq Mkt.Dep DecadeInd Focus
fl.

 

 

 

 

 

 

Mgr to Size Shr J Dum Growth
Assets
%Mgr 1.00
pr
Debtto -.048 1.00
Assets
Chain -.412 -.083 1.00
Size
S/Sq. -.297 -.159 .426 1.00
Ft.
Mkt. -.102 .038 .390 .123 1.00
Shr.

 

DQJ .328 -.032 -.454 -.264 -.070 1.00
Decade -.146 .214 .081 .080 .020 -.054 1.00

 

 

Ind. .006 .231 .005 -.006 .045 -.021 .126 1.00

 

 

 

 

 

 

 

 

 

 

 

 

Focus .247 .234 .358 .077 .212 -.237 -.049 .061 1.00

 

108
Table 10

Univariate Probit: Public Incumbent Firms

Maximmlikelihoodestimationresults.Thedependentvariableyfi=liffirmieverexpands
duringYears-lto+2;otherwiseyg=0. Thespecification ha31209firmimktjobservatiom.
Expansion occurred in 173 obs. 38 firms are represented in the sample. The values in
parenthesesarep-values. Marginaleffectsarecalculatedatfliemeamofthex’slnthecaseof
dummy variables, the marginal effects for the discrete clunge of the dummy fiom 0 to 1 is shown
belowtheclassicmargimlefi‘ectsvalue.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Coefiicients Marg. Efl‘ects
dProb[y=l]/dx
Constant -0.7896 ” -0.1684
(0.015)
% Insider Ownership, -1.2025 ”" -0.2565
(0.002)
Total Debt to Assets; -1.1979 "W -0.2555
(0.002)
Focusi 0.5493 “‘ 0.1172
(0.0099)
Clnin Size; 0.000005 0.000001
(0.962)
Inflation Adjusted Sales Per 0.0065 ” 0.0014
Square Ft, (0.031)
Market Share of Firm i in -0.4319 -0.0921
Market 1 (0.104)
Dependence of Firm i on 10.135‘" 2.1620
Market j (0.0004)
Wal-Mart Strength Dummy —0.0529 -0.0113
1 = ‘86-’94 0 = ‘76-‘85 (0.614) -0.1145
Industry Growth in Sq. Ft - 7.6560 -1.6331
(0.121)
Significance Level 0.000002
(HO: 13 =0)
Pseudo R-squared .334
N 1209

 

 

 

 

 

Statistically significant at 10% a, at 5% n, at 1% m

LM test statisties were done to check for multiplicative heteroskedasticity using all of the right
hand side variables except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 95% level. The specification was re-run using White’s robust estimator of
variance. The p-values are virtually identical to those shown above and are not shown. Pseudo
R-Squared was calculated in Limdep using a formula given by Zavoina and McKelvey (1975).

109
Table 11

Univariate Probit: Public Incumbent Firms The Dummies Included

Maximum likelihood estimation results. The dependent variable y;,- = 1 if firm i ever expands
dmingYears-lto+2;otherwiseyg=0. Thereare1209firmimktj observations. Expansion
occturedin173obs.and38firmsarerepresentedinthesample. Thevaluesinparenthesesarep-
values. Marginal effects are calculated at the means of the x’s. In the case of dummy variables,
themarginaleffectsforthediscretechangeofthedummyfiomOtolisshownbelowtheclassic
marginal effects value. The sample was broken into 8 two-year pciods and 1 three-year period 8

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

time rhunmies were inchtded in the specification, but the coeflicients are not shown below. H
Explanatory Variables Coefficients Marg. Efl‘ects Coefl’icients
dProb[y=1]/dx Robust p-values
(White’s)
Constant -0.6778 -0.1405 -0.6778
(0.384) (0.363) t
% Insider Ownership, -l.3l76 “‘ -0.2731 -1.3176 "M
(0.001) (0.001)
Total Debt to Assets, -1.1548 "” -0.2394 -1.1548 ’”
(9.004) (0.007)
Focus. 0.5522 ” 0.1145 0.5522 ”‘
(0.013) (0.007)
Chain Size. 0.00003 0.000006 0.00003
(0.785) (0.762)
Inflation Adjusted 0.0058 " 0.0012 0.0058 ”
Sales Per Square Ft, (0.069) (0.039)
Market Share of Firm i -0.5270 " -0.1092 -0.5270 "
in Market j (0.053) (0.063)
Dependence of Firm i 9.820 ”“ 2.0355 9.820 ‘”
on Market j (0.0008) (0.000)
Wal-Mart Strength -0.2785 -0.0577 -0.2785
Dummy (0.260) (0.269)
1 = ‘86-’94 0 = ‘76-‘85
Industry Growth in Sq. -2.8889 -0.5988 -2.8889
Ft (0.867) (0.857)
Significance Level 0.00000
(Ho: B =01
Pseudo R-squared .3512
1209

 

 

N
Statistically significant at 10% ’, at 5% ”, at 1% ""

LM test statistics were done to check for multiplicative heteroskedasticity using all of the right
hand side variables except the constant. The null hypothesis of homoskedasticity can be rejected
at the 99.5% level. The specifications were re-nm using White’s robust estimator of variance.
The p-values are shown above. None of the 8 time dummies is significant in either of the
regressions. Pseudo R-Squared was calculated in Limdep using a formula given by Zavoina and
McKelvey (1975).

110
Table 12

Univariate Probit: Public Incumbent Firms Using Operating Profit Margin

Maximum likelihood estinmtion results. The dependent variable ya = 1 if firm i ever expands
during Years -1 to +2; otherwise y“ = 0. The specification has 1178 firm i mkt j observatiom.
Expansion occlu'red in 168 obs. and 37 firms are represented in the sample. (One firm was
dropped as its opm was a large outlier.) The values in parentheses are p-values. Marginal effects
are calculated at the means of the x’s. In the case of dummy variables, the marginal effects for the
discrete change of the dummy from 0 to 1 is shown below the classic margiml efl'ects value.

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Coefficients Marg. Efl‘ects
Variables dprobjy=1]/dx

Constant -0.5841 ’ -0.1228
(0.0556)

% Insider -1.3937 ‘” -0.2930

Ownership; (0.0005)

Total Debt to Assets, -1 .71 12 "" -0.3598
(0.00008)

Focus; 0.8572 *" 0.1802
(0.0003)

Chain Size; 0.00002 0.000005
(0.8736)

Operating Profit 3.444 * 0.7241

Margim (0.0949)

Market Share of -0.5210 " -0.1095

Firm i in Market j (0.0577)

Dependence of Firm 10.5099 ”" 2.2096

i on Market 1 (0.0003

Wal-Mart Strength 0.0898 0.0189

Dummy (0.419) 0.1842

1 = ‘86-’94 0 = ‘76-

‘85

Industry Growth in - 7.0078 -1.4733

Sq. Ft (9.156)

Significance Level 0.00000

(Ho: B =0)

Pseudo R-squared .3422

N 1178

 

 

 

 

 

Statistically significant at 10% *, at 5% n, at 1% m

A LM test statistic was done to check for multiplicative heteroskedasticity using all of the right
hand side variables except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 95% level. The specification was re-run using White’s robust estimator of
variance. The p-values are virtually identical to those shown above and are not shown. Pseudo
R-Squared was calculated in Limdep using a formula given by Zavoina and McKelvey (1975).

 

111
Table 13

Univariate Probit: Public Incumbent Firms Parent Size Added

Maximum likelihood estimation results for expansion at the 3-digit zip code level for public
incumbents. The dependent variable Yij = l iffirm i ever expands during Years —1 to +2;
otherwise yg = 0. There are 1209 firm i mktj observations. Expansion occurred in 173 obs. The
values in parentheses are p-values. 38 public firms are represented in the sample. Marginal
effects are calculated at the means of the X’s.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Coeflicients Marginal Effects
dprob[y=1]/dx @t mean of X)

Constant -0.7799 ‘"" -0.1664
(0.016)

% Insider Ownership -l.1989 *" -0.2558
(0.002)

Total Debt to Assets -1.l684 ”" -0.2493
(0.003)

Focus 0.4819 " 0.1028
(0.043)

Clnin Size -0.00009 -0.00002
(0.600)

Inflation Adjusted Discount 0.0066 ” 0.0014

Sales Per Square Foot (0.028)

Market Share of Firm i in -0.4344 -0.0927

Market 1 (0.102)

Dependence of Firm i on Market 10.085 ‘” 2.1517

j (0.0004)

Wal-Mart Strength Dummy «0.0486 -0.0104

1 = 1986-1994 0 = 1976-1985 (0.645)

Industry Growth in Sq. Ft -6.8498 -1.4614
(0.180)

Inflation Adjusted Parent Total -0.00003 -0.000006

Assets (.552)

Significance Level .000005

Gio=£=0)
Pseudo R-squared .3335
N 1209

 

 

 

 

 

Statistically significant at 10% t, at 5% u, .t 1% m

A LM test statistic was done to check for multiplicative heteroskedasticity using all of the right
hand Side variables except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 95% level.

112
Table 14

Univariate Probit: Public and Private Incumbents Regional Dummies

Maximum likelihood estimation results for expansion at the 3-digit zip code level for private and
public incumbents. Thedependentvariableyg= l iffirmiever expands dlu'ingYears—l to+2,
otherwise yij = 0. There are 1506 firm i mktj observations. Expamion occurred in 219
observations. The values in parentheses are p-values. Observations were divided into 6
geographic regions and 5 regional dummies were used. The coefficients on the regional dummies
are not shown. Marginal effects are calculated at the mean ofthe x’s. In the case ofdummy
variables, the marginal effects for the discrete change of the dummy fiom 0 to 1 is shown below
the classic marginal efl’ects value.

 

 

 

 

 

 

 

 

 

 

 

 

ExplanatoryVariables Coefficients Marg. Effects Coeflicients
dprob[y=1]/dx Robust p-values
(White’s)
Constant -1.5743 "* -0.3388 -1.5743 "””
(0.000) (0.000)
Public/Private Dummy - 0.3927 ” - 0.0845 - 0.3927 "
1= private 0 = public (0.024) - 0.0715 (0.025)
LBO/No LBO Dummy 1.0223 ” 0.2200 1.0223 "
1 = LBO 0 = No LBO (0.012) 0.3297 (0.011)
Chain Size 0.0001 " 0.00003 0.0001 "'
(0.085) (0.069
Inflation Adjusted Discount 0.0051 “ 0.0011 0.0051 ”
Sales Per Square Foot (0.023) (0.025)
Market Share of Firm i in -0.5729 ” -0. 1233 -0.5729 "
Market j (0.024) (0.022)
Dependence of Firm i on 3.3741 ” 0.7260 3.3741 "
Marketj (0.021) (0.012)
Wal-Mart Strength Dummy -0.1912 " -0.0412 -0.l912 *
1 = 1986-1994 0 - 1976-1985 (0,059) -0.0430 (0.057)
Industry Growth in Sq. Ft -8.2697 "' -1.7794 -8.2697 "
(0.070) (0.061)
Significance Level 0.0000
THO: [5 =0)
Pseudo R-squared .3373
N 1506

 

 

 

 

 

Statistically Significant at 10% ’, at 5% ", at 1% *"

A LM test statistic was done to check for multiplicative heteroskedasticity using all of the right
hand side variables except the comtant. The null hypothesis of homoskedasticity cannot be
rejected at the 99.5% level. The specification was re-run using White’s robust estimator of
variance. The p-values are shown above. One regional dummy is significant at the 1% level.
Another regional dummy is significant at the 10% level.

 

 

113
Table 15

Univariate Probit: Public Incumbents Regional Dummies

Maximum likelihood estimation results. The dependent variable yij = 1 if firm i ever expands
dining Years—l to+2; otherwiseyg =0. Thereare1209 firmimktj observations. Expansion
occurred in 173 obs. The values in parentheses are p-values. 38 public firms are represented in
the sample. Marginal effects are calculated at the means of the x’s. In the case of dummy
variables, the mginal effects for the discrete change of the dummy from 0 to 1 is shown below
the classic marginal effects value. Observations were divided into 6 geographic regions and 5
regional dummies were used. The coefficients on the regional dummies are not shown.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Coeflicients Marginal Effects
dprob[_y=1]/dx (at mean of X)

Constant - 1.2094 ’” -0.2496
(0.002)

% Imider Ownership. - 1.4680 ‘” -0.3029
(0.0003)

Total Debt to Assets, - 1.0684 “‘ -0.2205
(0.007)

Focusi 0.6764 ”" 0.1396
(0.002)

Chain Size. -0.000008 -0.000002
(0.936)

Inflation Adjusted Discount 0.0067 "“" 0.0014

Sales Per Square Foot, (0.040)

Market Share of Firm i in - 0.5800 ” -0.1197

Market j (0.048)

Dependence of Firm i on 11.3364 ”W 2.3392

Market j (0.0001)

Wal-Mart Strength Dummy 0.0170 0.0035

1 = 1986-1994 0 = 1976- (0.886) 0.0035

1985

Industry Growth in Sq. Ft - 7.0386 -l.4524
(0.180)

Significance Level .000000

(Ho: [3 =0)

Pseudo R-Squared .3541

N 1209

 

 

 

 

 

Statistically significant at 10% *, at 5% ", at 1% "W

A LM test statistic was done to check for multiplicative heteroskedasticity using all of the right
hand side variables except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 95% level. The specification was re-run using White’s robust estimator of
variance. The p-values are virtually identical to those Shown above and are not shown. Pseudo
R-Squared was calculated in Limdep using a formula given by Zavoina and McKelvey (1975).
One regional dummy and the constant are significant at the 1% level. Another regional dummy is
Significant at the 5% level.

114
Table 16

Univariate Probit: Public and Private Incumbents State Population Changes

Maximum likelihood estimation results for expansion at the 3—digit zip code level for private and
public incumbents. Thedependentvariableyij= l iffirmiever expandsdm'ing Years-1 to+2,
otherwise yg = 0. There are 1422 firm i mkt j observations. Expamion occurred in 205

observations. The values in parentheses are p-values. Marginal efl’ects are calculated at the means
of the x’s. In the case of dummy variables, the rmrginal effects for the discrete change of the
dummy from 0 to 1 is Shown below the classic marginal effects value.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Coefficients Marginal Effects Coefi‘tcients
dprob[y=1]/dx Robust p-values
(White’s)
Constmt -1.4771 “* -0.3172 -1.4771 *”
(0.000) (0.000)
Public/Private Dummy, -0.4517 *” -0.0970 -0.4517 ”
1= private 0 = public @0097) -0.0802 (0.012)
LBO/No LBO Dummyi 1.0729 *” 0.2304 1.0729 ”*
1 = LBO 0 = No LBO (0.007) 0.3493 (0.008)
Chain Sizet 0.0001 0.00002 0.0001
(0.145) (0.140)
Inflation Adjusted Discount 0.0027 0.0006 0.0027
Sales Per Square Foot; (0.227) (0.230)
Market Share of Firm i in -0.6242 ” -0.l340 -0.6242 ”
Marketj (0.015) (0.019)
Dependence of Firm i on 4.5467 "* 0.9764 4.5467 *"
Market j (0.002) (0.001)
Will-Mart Strength Dummy —.2077 " -0.0446 -.2077 "
1 = 1986-1994 0 - 1976-1985 (0,032) -o.o473 (0,035)
Industry Growth in Sq. Ft -6.3177 -1.3567 -6.3177
(0.175) (0.172)
State Population Changej 0.0006 *” 0.0001 0.0006 *"
(0.000) (0.000)
Significance Level 0.0000
(Ho: [3 =0)
Pswdo R-squared .3313
N 1422

 

Statistically significant at 10% ‘, at 5% “, at 1% ""

A LM test statistic was done to check for multiplicative heteroskedasticity using all of the right
hand side variables except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 99.5% level. The specification was re-nm using White’s robust estimator of
variance. The p-values are shown above.

 

 

115
Table 17

Univariate Probit: Public Incumbents State Population Changes

Maximum likelihood estimation results. The dependent variable yij = 1 if firm i eva' expands
dtu'ingYearS-l to+2; otherwiseyg=0. Thereare 1155 firmimktj observations. Expamion
occmredin 160 obs. Thevalues inparenthesesarep—values. 38publicfirmsarerepresentedin
the sample. Marginal effects are calculated at the means ofthe x’s.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Coefficients Marginal Coefficients
Effects Robust p-values
dprobly=ll/dx W68)
Comtant -0.9417 ‘" -0.1930 -0.9417 ”"
(0.005) @002)
% Insider Ownership, -0.8159 ” -0.1672 -0.8159 ”
(0.043) (0.045)
Total Debt to Assets; -0.9433 ” -0.l933 -0.9433 ”
(0.016) (0.025)
Focus, 0.4663 " 0.0956 0.4663 ”
(0.036) (0.035)
Chain Sizet 0.0001 0.0002 0.0001
(0.240) (0.205)
Inflation Adjusted Discount 0.0026 0.0005 0.0026
Sales Per Square Foot (0.435) (0.392)
Market Share of Firm i in -0.8441 "" 01730 -0.8441 ”"
Market j (0.004) (0.008)
Dependence of Firm i on 12.9963 "* 2.6636 12.9963 "‘
Market j (0.000) (0.000)
Wal-Mart Strength Dummy 0.0445 0.0091 0.0445
1 = ‘86-‘94 0 = ‘76-‘85 (0.700) (0.718)
Industry Growth in Sq. Ft -7.4920 -l.5355 -7.4920
(0.152) (0.145)
State Population Change; 0.0007 ”" 0.0001 0.0007 ”‘
(0.000) (0.000)
Significance Level 0.0000
0: B =0)
Pseudo R-Squared .3381
N 1155

 

 

 

 

 

Statistically significant at 10% ’, at 5% u, at 1% see

A LM test statistic was done to check for multiplicative heteroskedasticity using all of the right
hand side variables except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 99.5% level. The specification was re-run using White’s robust estimator of
variance. The robust p-values are shown above. Pseudo R-Squared was calculated in Limdep
using a formula given by Zavoina and McKelvey (1975).

116
Table 18

Univariate Probit: Public Incumbents State Popuhtion Clunges Herfindahl Imtead of
Market Share

Maxirmmlikelihoodestinntionresults. Thedependentvariableyg=1iffirmieverexpands

duringYears—lto+2;otherwisey,,=0. Thereare1155firmimktj observatiom. Expansion
occurredinl600bs.Thevaluesinparenthesesarep-values. 38publicfirmsarerepresemedin
thesample. Marginal efi'ectsarecalculated atthemeansof the x’s.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Coefi'lcients Marginal Efl‘ects Coefiicients
dprob[y=l]ldx Robust p-values
Mites)
Comtant -0.8400 ” -0.1731 -0.8400 ”
(0.013) (0.007)
% Imider Ownership, -0.8574 ” -0.1767 -0.8574 ”
(0.033) (0.034)
Total Debt to Assets, -0.9823 ” -0.2024 -0.9823 ”
(0.012) (0.021)
Focus, 0.4351 “ 0.0896 0.4351 “
(0.0499) (0.046)
Chain Size, 0.00004 0.000009 0.00004
(0.644) (0.621)
Inflation Adjusted Discount 0.0030 0.0006 0.0030
Sales Per Square Foot, (0.355) (0.312)
Modified Herfindahl j -0.7419 "" -0.1529 -0.7419 “
(0.033) @037)
Dependence of Firm i on 11.3573 "" 2.3400 11.3573 *"
Market j (0.000) (0.000)
Wal-Mart Strength Dummy 0.0376 0.0077 0.0376
1 = ‘86-‘94 0 = ‘76-‘85 (0.744) (0.761)
Industry Growth in Sq. Ft -7.1080 -1.4645 -7.1080
(0.174) (0.169)
State Population ChangeJ 0.0007 ”"' 0.0001 0.0007 ‘"”
(0.000) (0.000)
Significance Level 0.0000
(1'10: 13 =0)
Pseudo R-squared
N 1155

 

 

 

 

 

Statistically significant at 10% ’, at 5% ”, at 1% "*

A LM test statistic was done to check for multiplicative heteroskedasticity using all of the right
hand side variabla except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 99.5% level. The specification was re-nm using White’s robust estimator of
variance. The robust p-values are Shown above. Pseudo R-Squared was calculated in Limdep
using a formula given by Zavoina and McKelvey (1975).

117

Table 19

Univariate Probit: The Importance of Ownership Controlling for Firm Size, and
Economic Growth

Maximum likelihood estimation results. Model 1 using public firns only. Model 2 uses both

publicandprivatefirms. Thedependentvariableyg=1 iffirmieverexpandsduringYears-lto

+2; otherwise y,,- = 0. There are 1209 firm i mkt j observations. Expansion occurred in 173 obs.
The values in parentheses are p-values. 38 public firms are represented in the sample. Margiml

effects are calculated at the means of the x’s. In the case of dummy variables, the marginal efl’ects

forthediscretechangeofthedummyfiomOto 1 isshownThesamplewasbrokeninto8two—

year periods and 1 three-year period 8 time dummies were included in the Specification, but the

 

 

 

 

 

 

 

 

 

coefficients are not shown below.
Explanatory Variables Model 1 Model 1 Model 2 Model 2
Coeflicients Marginal Coefficients Marginal
Robust Efl‘ects Robust Efl‘ects
p-values dprobly=1]/dx p-values dprob[y=1]/dx
(White’s) (White’s)
Comtant -0.5177 -1.3669 ”
(0.408) (0.026)
% Insider Ownership, -0.8122 " -0.1741
(0.017)
Public/Private Dummy, -0.3082 “ -.0592
l= private 0 = public (0047)
Chain Sizel 0.0001” 0.00003 0.0001 ” 0.00003
(0.047) (0.035)
Dependence of Firm i on 7.2150 ”" 1.5468 3.4149 ” 0.7459
Market 1 (0.007) (0.013)
Industry Growth in Sq. Ft -4.2646 -0.9143 -3.3955 -0.7417
(-0.792) (0.801)
Significance Level 0.0000 0.0000
(HO: 13 =0)
N 1209 1506

 

 

 

 

 

 

Statistically Significant at 10% ‘, at 5% ”, at 1% ”"'

A LM test statistic was done to check for multiplicative heteroskedasticity using all of the right
hand side variables except the constant. The null hypothesis of homoskedasticity cannot be
rejected at the 99.5% level. The Specification was re-run using White’s robust estimator of
variance. The robust p-values are Shown above.

 

118
Table 20

OLS Regressions Explaining Wal-Mart’s Market Share as a Frmction of Ownership in a
Local Market

The dependent variable for Model 1 & Model 3 is the fiaction of total stores owned by Wal-Mart
in local marketj at the beginning onear +2 (M). The dependent variable for Models 2 & 4 is
the number of stores owned by Wal-Mart in local market j at the beginning of Year +2 as a
fiaction of the total number of stores in local market j at the beginning of Year -1 (penbasq). All
models are estimated using OLS with t-statistics in parentheses using White’s consistent standard
errors. Annual time dummies wee included in the specifications, but their coefficients are not
shown. Tobeinthesample, Wal-Marthastobefirstseen inalocal marketbetween 1981 — 1994.

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Model 1 Model 2 Model 3 Model 4
Variables Coefficients Coefficients Coefl’lciems Coeflicients
PM Yi‘Penbfisei Yl=P°ni Yl=Penbasq
(All Markets) (All Markets) (Public Markets) (Public Markets)
constant 0.414 an 1.026 "* 0.559 ”"' 1.059 ”*
(7.407) (3.103) (8.377) (7.981)
fracpri, - 0.072 I" - 0.165 in
(4.105) (-2241)
w oij - 0.320 ""' - 0.984 ”‘
(-3.398) (6.733)
Numbe Of 471 471 233 233
Obs.
R-Squared .149 .158 .166 .225

 

Statistically significant at 10% r, at 5% n, at 1% m

119
Table 21

Summary Statistics: All Markets

Tobeinthesample, Wal-Marthastobefirstseeninalocalnnrketbetween 1981—1994.

 

 

 

 

 

471 obs.

Explanatory Mean Standard Minimum Maximum

Variables Deviation

fracprii .1434 .2233 0.0 1.0

w depl- 1.565 % 2.031% .0437°/o 22.577%
W .0271 .0951 0.0 1.0

lnstrs;L 1.831 .7939 0 3.784

 

 

 

 

 

 

 

120
Table 22

Summary Statistics: Markets With All Public Incumbents

To be inthe sample, Wal-Martlms to be first seen ina local market between 1981 - 1994.

 

 

 

 

 

233 obs.

Explanatory Mean Standard Minimum Maximum

Variables Deviation

w oij .0917 .1033 .0038 .5859

w clep-L .8047 % .8124 % .0437% 6.1587%
m .9315 .1519 0 1.0

lnstrsj 1.6003 .8055 0 3.5264

 

 

 

 

 

 

 

 

121
Table 23

OLS Regressions Explaining Wal-Mart’s Market Stare as a Function of Ownership in a
LocalMarket: Market andFirmSize ControlsAdded

ThedependentvariableforModel l &Model3isthefi'actionoftotalstoresownedbyWal-Mart
inlocalmarketj atthebeginningonear+2 (penj).ThedependentvariableforMode182&4is
the number of stores owned by Wal-Mart in local market j at the beginning of Year +2 as a
fi'actionofthetotalnumberofstoresinlocalmarketj atthebeginningonear-l (penbasej). All
models are estimted using OLS with t-statistics in parentheses using White’s consistent standard
errors. Annual time dummies were included inthe specificatiom, buttheir coefficients arenot
shownTobeinthesample,Wal-Marthastobefirstseeninalocalmarketbetween1981—1994.

 

 

 

 

 

 

 

 

 

Explanatory Model 1 Model 2 Model 3 Model 4
Variables Coefficients Coeflicients Coeflicients Coeflicients
1=pens ”=9ch yin-pen: y1=lmbtsq
(All Markets) (All Markets) (Public Markets) (Public Markets)
comtant 0.630 m 1.387 m 0.642 m 1.113 m
(9.277) (4.683) (4.230) (6.179)
impri- - 0.079 n - 0.118 n
(.2374) (.2498)
wow; -0.198 n - .0517 m
(.2195) (-2.963)
locus,- - 0.122 m - 0.273 m - 0.145 m - 0.282 m
(44.594) (12.074) (-9.362) (42.450)
proplg - 0.0008 0.130 n 0.054 0.235 '-
(-0.014) (2.382) (0.494) (1.923)
Wdepj 0.001 0.001 0.047 9* 0.026
(0.439) (0.233) (2.284) (0.244)
Number of 471 471 233 233
Obs.
R-squared 0.501 .538 .4817 .607

 

 

 

 

 

 

 

Statistically significant at 10% ‘, at 5% ”, at 1% ”"

 

122
Table 24

Constrained Least Squares Spline Regressions Testing For Differences in Slope
Coefficients for Difl‘erent Ownership Groups: Market and Firm Size Controls Added

The dependent variable for Model 1 is the fi'action of total stores owned by Wal-Mart in local
market j at the beginning of Year +2 (penj). The dependent variable for Model 2 is the number of
stores owned by Wal-Mart in local market j at the beginning of Year +2 as a fraction of the total
number of stores in local market j at the beginning of Year —1 (penbasq). All models are
estimated using OLS with t-statistics in parentheses using White’s consistent standard errors.
Annual time dummies were included in the specifications, but their coefficients are not shown. To
be inthe sample, Wal-Marthasto be first seen in a local market between 1981 —1994.

 

 

 

 

 

 

 

 

 

 

 

 

Explamtory Model 1 Model 2
Variables Coefficients Coeflicients
Y1: P601 Y1: MM
(Public Markets) (Public Markets)

Constant 0.648 ”’ 1.166 ”"
(4.245) (6.841)

w oij - 0.354 - 3.254 ""
(-0.515) (-2.775)

(11(0ij - .05) 0.135 3.025 “
(0.166) (2.220)

(12 (oij - .25) 0.138 0.177
(0.352) (0.267)

lnstrsj - 0.144 ”" -0.271 ”t
(-9.467) (-12.l62)

proplg 0.050 0.221 ‘
(0.439) (1.770)

w depj 0.047 " 0.023
(2.278) (0.329)

Number of Obs. 233 233

R-squared .186 .617

 

 

 

Statistically significant at 10% t, at 5% n, at 1% m

123
Table 25

OLS Spline Regression: Explaining Wal-Mart’s Market Share as a Function of Different
Ownership Groups: Market and Firm Size Controls Added

The dependent variable is the number of stores owned by Wal-Mart in local mrket j at the
beginning onear +2asafi-actionofthetotal numberofstores in local marketj atthebeginning
of Year —1 (penbasej). The model is estimated using constrained least squares with t-statistics in
parentheses using White’s consistent standard errors. Annual time dummies were included in the
specifications, but their coefficients are not shown. To be in the sample, Wal-Mart has to be first
seen in a local rmrket between 1981 - 1994.

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Model
Coefficients
10 = penbasei
(Public Markets)
Constant 1.166 ”"
(6.841)
w own-,1 0 to .05 -3.254 "’
(-2.775)
w opr .05 to .25 -0.229
(-0.765)
w opr over .25 -.052
{-0.111)
lnstrsj -0.271 ”"
(-12.l62)
proplg, 0.221 "
(1.770)
w depj 0.023
(0.979)
Number of Obs. 233
R-squared .617

 

 

 

Statistically significant at 10% ‘, at 5% ”, at 1% ”"'

 

 

124
Table 26

Simple Correlation Matrix for Independent Variables Using Public Incumbent Markets

There are 233 local market observations inthe sample. All 233 markets have only public
incumbentfirmsinthematthebeginning onear—l. Tobeinthesample, Wal-Marthasto

befirstseeninalocalmarketbetween 1981 -1994.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

W focj W debt, W oij W depj W ssqi Proplg; Mod Lnstrsj

(1974 H311}
$29.1

W foo) 1.000

W debt; .264 1.000

W 0W1}; .268 -.060 1.000

W dep, -.086 .072 .336 1.000

Wssqj .178 -.118 -.357 -.319 1.000

p (1974 M)

PrOPlgL .030 .036 -.151 -.508 .057 1.000

Mod .186 -.105 -.372 -.369 .338 .125 1.000

Haj

Lmtrsj -.163 .007 .166 .495 -.103 -.010 -.654 1.000

 

 

125
Table 27

Summary Statistics: Markets With All Public Incumbents

Tobeinthesample,Wal-Marthastobefirstseeninalocalmarketbetween1981-1994.
233obs.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Mean Standard Minimum Maximum

Variables Deviation

w focj .7728 .1743 .0390 1.0

w delji .6660 .1145 .2935 1.0235

w owpi .0917 .1033 .0038 .5859

w depj .8047 % .8124 % .0437% 6.1587%

w ssqi (1974M 59.4619 11.6496 15.5549 90.9609
Mg, .9315 .1519 0 1.0

mod Herfi .5079 .2342 .2099 1.0

lnstrsj 1.6003 .8055 0 3.5264

 

 

126
Table 28

OLS Regressions Explaining Wal-Mart’s Market Share at Year +2: Markets With Only
Public Firms: Dependent Variable is Pen,-

Dependent variable is the fiaction of total stores owned by Wal-Mart in local market j at the
beginning of Year +2 for all models (penj). All models are estimated using OLS with t-statistics
in parentheses using White’s consistent standard errors. Annual time dummies were included in
the specifications, but their coefficients are not shown. To be in the sample, Wal-Mart has to be
firstseeninalocalmarketbetween 1981 -1994.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Model 1 Model 2 Model 3
Coefficients Coefficients Coeflicients

comtant 0.324 ” 0.239 ” 0.373 m"
(2.426) (2.105) (2.882)

WfOCj -0.116" - 0.145 ”*
(-1.848) (-2.693)

w debt,- 0.678 *” 0.701 ‘” 0.603 ‘”
(7.269) (7.897) (6.880)

w 0ij 0.026 0.0757 - 0.081
(0.327) (1.035) (-1.106)

w depj 0.014 0.020 0.013
(0.923) (1 .467) (0.860)

w ssqi - 0.001 - 0.002 ”
(-1.055) (-2.087)

proplg, - 0.036 - 0.007 - 0.055
(-0.455) (-0.100) (-0.706)

mod hert} 0.240 ‘” 0.231 "* 0.216 t”
(4.421) (4.283) (3.920)

lnsn'sj - 0.091 W" - 0.096 ""' - 0.090 “W
(6050) (-7.103) (-5.956)

Number of Obs. 233 233 233

R-scLuared 0.64 0.63 0.63

 

Statistically significant at 10% *, at 5% ‘t, at 1% us

 

127
Table 29

Summary Statistics: Markets With All Public Incumbents

Tobeinthesample,Wal-Marthastobefirstseeninalocalmarketbetween1981—1992.

 

 

 

 

 

 

 

 

 

176 obs.
Explanatory Mean Standard Minimum Maximum
Variables Deviation
W foq .7522 .1829 .0390 1.0
W debt,- .6488 .1103 .2935 1.0235
W oij .0899 .1026 .0038 .5859
W depj .8147% .8619% .0437% 6.1587%
W - 1914 60.6942 12.1748 15.5549 90.9609
Proplgi .9256 .1606 0 1.0
Mod Hert} .5191 .2335 .2222 1.0
Lnstrsj 1.5953 .7865 0 3.5264

 

 

 

 

 

 

 

"”7

128

Table 30

OLS Regressions Explaining Wal-Mart’s Market Share at Year +4: Markets With Only
Public Firms

Dependentvariableisthefiactionoftotalstoresownedby Wal-Martinlocal marketjatthe
beginning of Year +4 for all models (penfivej). All models are estimated using OLS with t-
statistics in parentheses using White’s consistent standard errors. Annual time dummies were
included in the specifications, but their coefficients are not shown. To be in the sample, Wal-Mart
hastobefirstseeninalocal marketbetween 1981 — 1992.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Model 1 Model 2 Model 3
Coefficients Coefficients Coefficients
Constant 0.197 0.005 0.221 t
(1.401) (0.043) (1.700)
w foo,- - 0.064 - 0.180 m
(-0.763) (-2920)
w debt,- 1.022 m 1.033 m 0.999 m
(10.831) (10.386) (10.013)
wowpj -0.069 0.119 -0.l39
(0.443) (0.889) (4.146)
w depj 0.017 0.032 r 0.017
(1.029) (1.899) (1.015)
w ssqj - 0.004 8* - 0.004 m
(-2.477) (4.212)
Proplg, - 0.053 0.008 - 0.060
(.0574) (0.094) (-0.689)
Mod hat} 0.310 m 0.284 m 0.297 m
(5.846) (5.198) (5.396)
lnstrsj - 0.065 m - 0.081 m - 0.064 m
(-3.649) (.4619) (-3.630)
Number ofObs. 176 176 176
R-squared 0.64 0.63 0.64

 

Statistically significant at 10% ‘, at 5% “, at 1% ""

 

 

129
Table 31

OLS Regressions Explaining Wal-Mart’s Market Share in Year +2 as a Fraction of Stores
in Market in Year —1: Markets With Only Public Firms

DependeMvmiableisthefi’actionoftotalstoresatthebeginningonear-l inlocalmarketj
owned by Wal-Mart in local market j at the beginning of Year +2 (palbasq). The models is

estimated using OLS with t-statistics in parentheses using White’s consistent standard errors.
Annual time dummies were included in the spwification, but their coefficients are not shown.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Model 1 Model 2 Model 3
Variables Coefficients Coefficients Coefficients
Constant 0.538 ” 0.805 "‘ 0.620 ‘”
(2.316) (4.192) (2.731)
WfOCj -0.195 t -0.105
(-1.773) (-1.044)
w debtj 0.350 "' 0.275 0.224
(1.935) (1.620) (1.559)
w oij 0.005 -0.146 -0.l76
(0.031) (-0.801) (-1.069)
w depj 0.017 -0.002 0.016
(0.681) (-0.098) (0.627)
w ssqi 0.003 ” 0.002
(2.022) (1.384)
Proplgi 0.197 0.109 0.165
(1.541) (0.916) (1.289)
Mod hert} 0.456 ‘" 0.484 "* 0.415 ”"
(4.971) (5.261) (4.283)
Lnstrsj - 0.205 W" -0.188 *” -0.203 ”t
(-6.763) (6789) (-6.720)
Number of Obs. 233 233 233
R-squared 0.66 0.65 .65

 

Statistically significant at 10% *, at 5% "a at 1% m

 

 

130
Table 32

Simple Correlation Matrix for Independent Variables Using All Local Markets
There are 471 localmarket observations inthe sample. Marketswithstoresownedby

privatefirmsareincluded. Tobeinthesample, Wal-Marthastobefilstseeninalocalmarket
between 1981 -1994.

 

LBO,- Fracprij W depj W ssq; Proplgi Mod Hert} Lmtrsj
(1974 gall-3)
LBO 1.000

fi'acpri-l .2623 1.000
W dep; .0803 .4048 1000

W ssq; -.0725 -.0548 -.0548 1.000

 

 

"'7

 

 

 

 

 

 

1974(th
Propl§_ -.0977 -.4972 -.5643 .0809 1.000 7
Mod -.1635 -.0708 -.2483 .1460 .2004 1.000

Heft}

 

 

 

 

 

 

 

 

 

 

Lnstrsj .0765 .0206 .3577 .0174 -.l305 -.6843 1.000

 

131
Table 33

Summry Statistics: All local markets

Tobeinthesample, Wal-Marthastobefilstseeninalocalmarketbetween 1981-1994.
471 obs.

 

 

 

 

 

 

 

 

 

Explanatory Mean Standard Minimum Maximum

Variables Deviation

LBO dummy) .1125 .3163 0.0 1.0

flagprij .1434 .2233 0.0 1.0

w deL 1.5650 % 2.0305 % .0437 % 22.5768 %

w . 1974”,, 59.2232 11.1508 15.5549 105.7653
M81 .8545 .1955 0.0 1.0

mod Hertj .4288 .2174 .1358 1.0

lnstrs) 1.8313 .7939 0.0 3.7842

 

 

 

 

 

 

132
Table 34

OLS Regressions Explaining Wal-Mart’s Market Share: All Markets No Debt Ratio, No
Focus Variable

The dependent variable for Model 1 is the fraction oftotal stores owned by Wal-Mart in local
market j at the beginning of Year +2 (pent). The dependent variable for Model 2 is the fraction of
total stores owned by Wal-Mart in local unrket j at the beginning of Year +4 (penfivej). The
dependent variable for Model 3 is the number of stores owned by Wal-Mart in local market j at
the beginning of Year +2 as a fiaction of the total number of stores in local market j at the
beginning of Year -1 (penbase,). All models are estimated using OLS with t-statistics in
parentheses using White’s consistent standard errors. Annual time dummies were included in the E
specifications, but their coefficients are not shown. To be in the sample, Wal-Mart has to be first '

seen in a local market between 1981 - 1994.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Model 1 Model 2 Model 3
Coefiicients Coefficients Coefficients
Y1: pent Yj= penfivei y1= penbusei
Constant 0.6000 *" 0.681 ”"' 1.004 *”
(9.708) (7.246) (3.462)
LBO dummy; -0.00006 0.003 0.020
(-0.004) (0.126) (076$
fi-acprij -0.076 ” -0.121 t” -0.090 "
(-2.469) (-2.860) (-1.842)
w depj -0.001 -0.001 -0.003
(-0.474) (-0.l67) (-0.478)
w ssqi -0.001 -0.001 0.003 ”
(-1.491) (-0.856) (2.289)
proplgj -0.027 -0.147 " 0.051
(—0.476) (4.852) (0.867)
mod herf} 0.140 "" 0.212 ‘““ 0.401 t”
(3.319) (3.617) (3.958)
lnstrsj -0.095 ”“ -0.082 ’“ «0.196 “‘
(-10.017) (-5.767) (-8.892)
Number of Obs. 471 370 471
R-squared .5195 .4121 .5789

 

Statistically significant at 10% ‘, at 5% ”, at 1% ""

 

 

OLS Regressions Explaining Wal-Mart’s Market Share: All Markets. No Debt Ratio

Thedependentvariable forbothModel l & Model 2 isthefiactionoftotalstores ownedby Wal-
Mart in local market j at the beginning of Year +2 (M). Both models are estimated using OLS
with t—statistics in parentheses using White’s consistent standard errors. Annual time dummies
were included in the specifications, but their coefficients are not shown. To be in the sample,

133

Table 35

Wal-Mart has to be first seen in a local market between 1981-1994.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory Variables Model 1 Model 2
Coefficients Coefficients
Y1: pen, Y1: pen,-
(All Markets) (Public Markets)
Constant 0.569 ”" 0.776 ”"
(8.908) (4.774)
fi’acprij -0.090 ”"
(-2.734))
w ()pr- -0.285 ”‘
(-3.081)
w focj 0.066 0.124 "
(1.591) (1.877)
w depj -0.002 0.023
(-0.552) (1.102)
w ssqi -0.001 * -0.003 ”'
(-1.821) (-2.753)
proplg, -0.041 -0.063
(-0.698) (-0.547)
mod hert} 0.144 ”‘ 0.141 "‘
(3.483) (2.421)
lnstrs, -0.092 t" -0.105 ‘”
(-9.231) (-6.008)
Number of Obs. 465 233
R-s_quared .5236 .5154

 

 

Statistically significant at 10% *, at 5% ", at 1% "‘"

_ 'IT I,“ uh"-"v a."

I34

 

 

 

 

 

 

Table 36
Summary Statistics

Observations for eachofthe50 States for eachyearwithin 1980- 1985. 8000bservations.
Explanatory Mean Standard Minimum Maximum
Variables Deviation
Herf. .453 .141 .188 1.0
Infssga- (1974de 67.63 23.98 36.57 232.46
Propwal. 13.26% 18.61% 0.0% 75.25%
Propwali. Sqd 5.21 10.35 0.0 56.62

 

 

 

 

 

 

 

 

135
Table 37

OLS Regressiom Explaining State Profitability and Herfindahls

The dependent variable is shown at the top of the column below. The models are estimated using
OLS with t-statistics in parmthsses using White’s consistent standard errors. Annual time
dummies were included in the specifications, but their coefficients are not shown. There are
observatiom for each ofthe 50 states for each year in the range of 1980 -1995.

 

 

 

 

 

 

 

 

 

 

 

 

Explanatory DepVariable = Dep Variable = Dep Variable
Variables infssqi. herfa Infssqg
Comtant 33.029 ”" .464 "" 64.152 ”"
(7.179) (23.236) (17.534)
Herf. 65.211 ‘”
(7.038)
Propwali - 0.652 ”‘ - 0.622 ‘"
(-10.315) (5.792)
Propwal. 1.378 "“ 0.835 “"
sqd (15.086) (5.563)
Number of 800 800 800
Obs.
R- Squared .2394 .132 .1242

 

Statistically significant at 10% r, at 5% n, at 1% m

.
B.‘ .— OI‘IIq‘FIL ”tire? .5.—

APPENDIXB

APPENDIX B — FIGURES

 

Standard Disc. Dept. Store Sq.Ft per

    

 

 

 

Person (U.S.)
3
2.8
2.6 .
2.2
2 .
Q? 9‘? a? a? 61' 99
Calendar Year
Figure]

Standard Discount Department Stores Sales per Person (U.S.)

Sources: U.S. Population from U.S. Census Bureau
Industry Sales Data From Discount Merchandiser

136

137

 

Discount Department Store Sales
Calendar 1975

 

lTop 5 Firms
l Other Firms

 

 

 

 

 

 

 

Figure 2

Fraction of 1975 Industry Sales By Top Five Chains

 

Discount Department Store Sales
Calendar 1 995

 

lTop 5 Firms
lOther Firms

 

 

 

 

 

 

 

Figure 3

Fraction of 1995 Industry Sales By Top Five Chains

Source: Sales Data From Discount Merchandiser

138

1975 1985 1995
l. Kmart 1. Kmart 1. Wal-Mart
2. Woolco 2. Wal—Mart 2. Knnrt
3. Zayre 3. Target 3. Target
4. Vormdo 4. Zayre 4. Meijer
5. Korvette 5. TG&Y 5. Fred Meyer
6. Target 6. Bradlees 6. Marshalls
7. Fred Meyer 7. Fred Meyer 7. Caldor
8. TheTreasury 8. Ames 8. Ames
9. Skaggs 9. Caldor 9. Shopko
10. Fed-Mart 10. Marslmlls 10. Venture

Figure 4
Top Ten Firms Based on Sales

Source: Sales Data fi'om Discount Merclmndiser

a V "" u—u '1 'n
s

I, -
H

 

Sauce:StoreLocationDataFromDirectoryofDisco\mtDepamnemStorss

140

 

Figure 6

Wal-Mart 1985 Store Locations

Source: Store Location Data From Directory of Discount Department Stores

141

 

Figure 7

Wal-Mart 1990 Store Locations

Source: Store LocationDataFrom Directory ofDiscountDepartment Stores

I42

 

Figure 8

Wat-Mart 1996 Store Locaions

Source: Store Location Data From Directory of Discomt Department Stores

 

 

F iiiiii

Source: Store Location Data From Directory of Discount Department Stores

144

' oath-ammonia.
an

:2”
‘ cult—um .

 

Figure 10

Wal-Mart and Sky City 1990 Store Locations

Source: Store Location Data From Directory of Discount Department Stores

145

 

Figure 11

Wal-Mim and TG&Y Store Locations1977

Source: Store Location Data From Directory of Discount Department Stores

harm
' ammonia“. w

 

Figure 12

Wal-Mart & TG&Y Store Locatiom 1986

Source: Store Location Data From Directory of Discount Department Stores

147

 

 

Figure 13

Wal-Mart and Duckwall 1975 Store Locations

Source: StoreLocationDataFromDirectoryofD’ncomtDepartmentStores

148

I
I
mm
* ammonium

. mm
' nun-antennas.“

 

 

Figure 14
Wal-Mart and Duckwall 1996 Store Locations

Source: Store Location Data From Directory of Discount Department Stores

LIST OF REFERENCES

LIST 0]" REFERENCES

Aggarwal, Rajesh, and Andrew Samwick, 1997, Executive
compensation, and relative performance evaluation: Theory
and evidence, working paper, Dartmouth College.

Berger, Philip and Eli Ofek, 1995, Diversification’s effect on firm
value, Journal of Financial Economics 37, 39-65.

Bolton, Patrick and David Scharfstein, 1990, A theory of predation
based on agency problems in financial contracting, The
American Economic Review 80, 93-106.

Brander, James, and Tracy Lewis, 1986, Oligopoly and financial
structure: The limited liability effect, The American Economic
Review 76, 956-970.

Chen, Ming-Jet, and Donald Hambrick, 1995, Speed, stealth, and
selective attack: How small firms differ from large firms in
competitive behavior, Academy of Management Journal 38,
453-481.

Chen, Ming-Jcr, and Ian MacMillan, 1992, Nonresponse and delayed
response to competitive moves: The roles of competitor
dependence and action irreversibility, Academy of Management
Journal 35, 539-570.

Chevalier, Judith, 1995, Capital structure and product market
competition: Empirical evidence from the supermarket
industry, The American Economic Review, 415-435.

Chevalier, Judith, 1995, Do LBO supermarkets charge more? An
empirical analysis of the effects of LBO’s on supermarket
pricing, The Journal of Finance 50, 1095-1112.

Comment, Robert, and Gregg Jarrell, 1995, Corporate focus and
stock returns, Journal of Financial Economics 37, 67-87.

149

150

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

Directory of Corporate Affiliations Who Owns Whom, New Jersey:
National Register Publishing., 1975-1990.

Directory of Discount Department Stores, New York: Business
Guides Inc.1975-1982, 1987-1994.

Directory of Discount Stores, New York: Business Guides Inc.,
1983-1986.

Directory of Discount and General Merchandise Stores, New York:
Business Guides Inc., 1995-1996.

Discount Merchandiser, New York: Schwartz Publications, 1975-
1996.

Fama, Eugene, and Michael Jensen, 1983, Separation of ownership
and control, Journal of Law and Economics 26, 301-325.

Fershtman, Chaim, and Kenneth Judd, 1987, Equilibrium incentives
in oligopoly, The American Economic Review, 77, 927-940.

Fudenberg, Drew and Jean Tirole, 1986, A “Signal Jamming” Theory
of Predation, Rand Journal of Economics, 17 (3), 366-376.

Jensen, Michael, and William Meckling, 1976, Theory ofthe firm:
Managerial Behavior, agency costs and ownership structure,
Journal of Financial Economics 3, 82-137.

Kaplan, Steven, and Luigi Zingales, 1995, Do financing constraints
explain why investment is correlated with cash flow? NBER
Working Paper Series, 5267.

Khanna, Naveen, 1998, Optimal contracting with moral hazard and
cascading, forthcoming Review of Financial Studies.

Khanna, Naveen and Annette Poulsen, 1995, Managers of financially
distressed firms: villains or scapegoats? The Journal of
Finance 50, 919-940.

Kedia, Simi, 1997, Product market competition and top management
compensation, working paper, New York University.

151

Kovenock, Dan, and Gordon Phillips, 1997, Capital structure and
product market behavior: An examination of plant exit and
investment decisions, The Review of Financial Studies 10,
767-803.

Lang, Larry, and Rene Stulz, 1994, Tobin’s q, corporate
diversification, and firm performance, Journal of Political
Economy 102, 1248-1280.

Maksimovic, Vojislav, 1988, Capital structure in repeated
oligopolies, Rand Journal of Economics 19 (3), 389-407.

Meyer, Michael, Paul Milgrom, and John Roberts, 1992,
Organizational prospects, influence costs, and ownership
changes, Journal of Economic & Management Strategy 1, 9-35.

Mikkelson, W., and M. Partch, 1989, Managers’ voting rights and

corporate control, Journal of Financial Economics 25, 263-
290.

Million Dollar Directory, Pennsylvania: Dun & Bradstreet, 1975-
1996.

Opler, Tim, and Sheridan Titman, 1994, Financial distress and
corporate performance, The Journal of Finance 49, 1015-1040.

Phillips, Gordon, 1995, Increased debt and industry product
markets: An empirical analysis, Journal of Financial
Economics 37, 189-238.

Reitman, David, 1993, Stock options and the strategic use of

managerial incentives, TheAmerican Economic Review 83,
513-524.

Rotemberg, Julio, and David Scharfstein, 1990, Shareholder-value
maximization and product market competition, The Review of
Financial Studies 3, 367-391.

Safieddine, Assem, and Sheridan Titman, 1997, Debt and corporate

performance: Evidence from unsuccessful takovers, NBER
Working Paper Series, No. 6068.

Scharfstein, D., 1997, "The dark side of internal capital markets 11:

Evidence from diversified conglomerates,” Working paper,
MIT and NBER.

152

Scharfstein, D. and J. Stein, 1997, ”The dark side of internal capital
markets: divisional rent-seeking and inefficient investment,"
Working paper, MIT and NBER.

Song, Moon, and Ralph Walking, 1993, The impact of managerial
ownership on acquisition attempts and target shareholder
wealth, Journal of Financial and Quantitative Analysis, 28,
439-457.

Stein, Jeremy, 1997, Internal capital markets and the competition
for corporate resources, The Journal of Finance, 52, 111-133.

Value Line Investment Survey, New York: Value Line Publishing
Inc., 1975-1996.

Wards Business Directory, New York: Gale Research, 1987-1996.