Chevalier_1995

Capital Structure and Product-Market Competition: Empirical Evidence from the SupermarketIndustry

Author(s): Judith A. Chevalier

Source: The American Economic Review, Vol. 85, No. 3 (Jun., 1995), pp. 415-435Published by: American Economic AssociationStable URL: http://www.jstor.org/stable/2118181Accessed: 11/02/2010 04:31

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at

http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unlessyou have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and youmay use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/action/showPublisher?publisherCode=aea.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact support@jstor.org.

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to TheAmerican Economic Review.

http://www.jstor.org

Structure Capital and Product-Market Competition: Empirical Evidence from the Supermarket Industry By JUDITH A. CHEVALIER* This paper establishes an empirical link between firm capital structure and product-market competition using data from local supermarket competition. First, an event-study analysis of supermarket leveraged buyouts (LBO's) sug- gests that an LBO announcement the market value of the LBO increases chain's local rivals. Second, I show that supermarket chains were more to likely enter and expand in a local market if a large share of the incumbent in the firms local market undertook LBO's. The study suggests that leverage increases in the late 1980's led to softer product-market in this competition industry. (JEL D43, G14, G32, G34, L13, L81) During the late 1980's corporate debt rose dramatically, due in large part to an un- precedented wave of leveraged buyouts (LBO's). This large-scale experiment with firm capital structure refocused both popu- lar and academic attention on the issue of how a firm's financing choices might affect its performance and behavior. Numerous recent theoretical works have examined one component of this issue-the question of how a firm's capital structure affects compe- tition in the market for the firm's products. However, very little work has been done to determine empirically whether a real link- age exists between capital markets and product markets.' Graduate School of Business, University of Chicago, 1101 East 58th Street, Chicago, IL 60637, and National Bureau of Economic Research. This paper is a revised version of Chapters 1 and 2 of my M.I.T. dissertation. My advisors, Paul Joskow and David Scharfstein, provided invaluable guidance and advice. Christopher Avery, Tasneem Chipty, Glenn Ellison, Franklin Fisher, Ruth Judson, Steven Kaplan, Brigitte Madrian, Christopher Mayer, Wallace Mullin, James Poterba, Nancy Rose, Julio Rotemberg, Fiona Scott Morton, Jeremy Stein, and two anonymous referees provided helpful comments. Charles Hadlock provided invaluable data guidance. Financial assistance from the National Science Foundation at the beginning of this project is gratefully acknowledged. All errors and omis- sions remain my own. 1The only empirical work of which I am aware is A. Michael Spence (1985), Jose C. Guedes and Tim C. Opler (1992), and Gordon M. Phillips (1992). 415 * In this paper, I test between two classes of theoretical models by examining the share price response of supermarket chains to the announcement of a rival chain's leveraged buyout and by examining the entry, exit, and expansion behavior of supermarket chains. The first class of models predicts that increases in firm leverage tend to \"soften\" product-market competition. The second class of models predicts that in- creases in firm leverage tend to \"toughen\" product-market A finding competition.2 that either of these two hypotheses is true would be important in that it would suggest that financing decisions can have real product- market effects. I examine the effect of debt on product- market competition by studying super- market chains in local markets. The super- market industry is a natural laboratory for testing these theoretical models for two rea- sons. First, many large supermarket chains undertook leveraged buyout transactions in the late 1980's. This allows examination of how product-market competition \"shakes out\" after competitors undertake sudden, 2The terms \"tough\" and \"soft\" price competition are used in the sense of John Sutton (1991). The \"toughness\" of price competition in two mar- differs kets if, holding the concentration in the two markets constant, price-cost margins in the two markets differ. 416 THE AMERICAN ECONOMIC REVIEW JUNE 1995 increases in debt. Second, super- dramatic takes place at the local market competition of super- level. This allows a comparison markets. market across competition in this paper are consistent The findings with the group of theoretical models of and product-market com- capital structure petition that suggest that product-market competition becomes \"softer\" when lever- age increases. These models include work At the national level, the supermarket unconcen- to be relatively industry appears supermarket chains trated. The four largest for only 16 percent of U.S. gro- accounted K. Kaufman (Phillip cery store sales in 1982 no R. Handy, 1989). However, and Charles supermarket chain in the United States is chain For example, the largest truly national. Stores, op- in the United States, American while Thus, erated in only 18 states in 1990. by Drew Fudenberg and Jean Patrick Bolton and David S. Scharfstein Tirole (1986), (1990), and Phillips inconsistent with other models, including (1991). The results are James A. Brander and Tracy R. Lewis (1986), Vojislav Maksimovic (1988), and Julio J. Rotemberg which predict that leverage changes man- and Scharfstein (1990), agerial that makes product-market competition and shareholder incentives in a way \"tougher.\" is as follows. The organization of the rest of the paper ity in the supermarket industry. Section I describes LBO activ- presents an event study examining Section nouncement effects of supermarket the an- II aged buyouts Section III describes on rival supermarket chains. lever- tions of the hypotheses the empirical and product-market competition of capital structure predic- exit, and expansion. Section IV describes for entry, the data for the study of entry, exit, and expansion. Section tion for the empirical VI examines Section V presents the results. an alternative explana- and my conlcusions results in Section V, tion VII. are presented in Sec- I. LBO Activity in the Supermarket Industry A supermarket is defined by the publica- food store that has annual sales of more tion Progressive Grocer (1989) as a retail than $2 million and has greater than 9,000 square feet account sales but only 10 percent of retail food for 70 percent of retail food store of selling space. Supermarkets establishments. According to Progressive Grocer (1989), there were approximately 30,754 supermarkets in the United States in of 11 or more stores. 1988, 55 percent of which belonged to chains the industry is relatively unconcentrated on a national concentrated. The average metropolitan level, local markets can be highly statistical four-firm area in the United States had a of 58 percent supermarket concentration ratio between 1985 and 1988.3 Supermarket in 1982. LBO's occurred primarily actions the $4.1 billion Kroger were the $5.3 billion Safeway The largest trans- ization, leveraged recapital- LBO, eral LBO, and the $1.2 the $1.8 billion Supermarkets Gen- LBO. These four companies alone owned billion Stop & Shop nearly of their LBO's. During this period, 4,000 U.S. supermarkets at the time also common for smaller regional chains it was and divisions LBO's. permarket Altogether, of larger chains to undertake undertaken chains in the United States have 19 of the 50 biggest su- proximately $72 LBO's. supermarket sales in 1991. billion They accounted of the $297 billion for ap- in in any LBO activity has not been concentrated ple used in this study, single geographic region. for 16 percent of the stores LBO firms In the sam- accounted markets, ern markets, 21 percent of the stores 17 percent of the stores in South- in Midwestern the stores in Western markets, and 42 percent of in Northeastern large markets. Part of the to the enormous LBO concentration in the West is due importance of Safeway on 3There is also one instance of a leveraged recapital- ization in this industry, which was undertaken by Kroger. A leveraged recap is a transaction in which a firm borrows in order to pay a large dividend to share- holders of at least 50 percent of the former equity value of the firm. Because this recap resulted in debt levels for Kroger similar to typical LBO debt levels, it is included in this analysis as an LBO. VOL. 85 NO. 3 CHEVALIER: CAPITAL STRUCTURE AND MARKET COMPETITION 417 the West Coast. Safeway's 1985 market share in cities in the sample in the West totaled nearly 25 percent. The vast majority of the leveraged buy- outs were not the result of unconstrained decisions by management and shareholders. one examines those local mar- competitor, kets where competition was not an impor- tant cause of the LBO.5 The local-market nature of supermarket competition helps to \"clean out\" the endogeneity of the LBO in the study of entry, exit, and expansion. The Instead, most of them were undertaken response fact, all four of the biggest to unwanted takeover attempts. in of the smaller ones) were deals (and many In thwart undertaken to the Haft family, which controls the Dart the unwanted takeover attempts of drugstore chain. supermarket industry leaves This description of LBO activity question of what caused the LBO's in unanswered the in the industry.4 Magowan (1989), the CEO of It has been suggested by this stores, which undertook Safeway Peter that the main effect of the an LBO in 1986, was to force Safeway to sell Safeway LBO off divisions which were not or spin Magowan divisions in which it was not as suggests that Safeway eprofitable. xcised the competitor as its rivals (see This suggestion by Magowan is Magowan, 1989). strong a restatement rationale for LBO's suggested by of the basic \"empire-building\" practically a C. Jensen Michael bad divisions, (1989). If a good firm has a few buying the firm and turning then value can be gained divisions to oby LBO asset sales were common higher-valued users. ver the bad chains. If this model of LBO's aTmong hese post- then, in the study sion, by examining the local markets in of entry, exit, and expan- is correct, LBO which Safeway chose to remain an active 4Perhaps, given the importance of the Haft family in initiating takeover attempts in this industry, the better question to ask is: what factors contributed to the Haft family's choices of takeover targets? An examination of newspaper and magazine accounts of these takeovers reveals no statements by the Haft family indicating why it chose particular targets. One might hope to analyze the family's plan by examining the reforms that the family instituted at the supermarket chains that were actually taken over. However, the only supermarket chain that the Haft family actually took over was Shoppers Food Warehouse. It is difficult to try to infer what the Haft family would have done with Safeway or Kroger by examining what changes they effected at a chain with only 30 stores. issue of how the affects the results tion VI. wendogeneity of LBO's ill be taken up in Sec- II. An Event Study of Supermarket LBO's response of supermarket In this section, I examine the stock-return nouncement ing a leveraged that a rival chains to the an- looking was pioneered by B. at the event buyout.6 chain is undertak- responses This approach of rival firms of and Robert literature. If leveraged S. Stillman Espen Eckbo (1983) to make buyouts (1983) in are expected the merger rival itive abnormal returns supermarket chains product-market competition ssofter, of LBO around the time hould exhibit pos- expected to lead to announcements; if LBO's are market chains turns around the time should competition, rival supermarket tougher product- exhibit negative abnormal re- ments. of LBO announce- A. Methodology and Data order to separate the I focus my analysis on a single industry in LBO leads to a change in hypothesis product-market that the 5The explanation suggested here for LBO's-that they serve to force firms to excise bad divisions, and to leave them only with divisions in which they are rela- tively efficient producers-is supported by the evi- dence. LBO firms are not statistically significantly more likely to be closing stores in unclosed divisions than non-LBO firms. At least in this regard, their behavior in the divisions not sold off seems unaffected by the LBO. 61 do not focus on the stock-return response of the leveraging firm itself. A leveraged buyout transaction occurs when the managers of the firm (or others) offer to pay a premium over the prevailing market price of the firm. The stock price rises to reflect this premium. The managers of the firm would not undertake a leveraged buyout if they did not believe that they could improve firm value. 418 THE AMERICAN ECONOMIC REVIEW JUNE 1995 from the alternative hypothesis competition are due to in- event returns that abnormal creased speculation that more LBO's will occur in the industry. I use information to separate about local-market competition com- that are directly firms in the industry firms from firms peting with the leveraging that are not directly competing with the leveraging firms.7 If an LBO greatly im- dates were ob- ments and announcement tained from the Wall Street Journal Index. studied Since all four of the transactions in response to take- here were undertaken window begins prior over attempts, the event to the first public announcement suggesting that a takeover might occur. Because the event window extends until the announce- as ment of the LBO, it can be interpreted proves dertaking the LBO and the LBO announce- the financial outlook of the firm un- ment increases speculation that other supermarket chains will also undertake an LBO, then one would expect all firms the industry return response to the LBO announce- to experience a positive stock- in ment. However, soften product-market if the LBO is expected to one would expect supermarket competition, then ating aging in the same local markets chains as the lever- oper- response chain to exhibit a positive stock-return markets the leveraging tthat do not compete directly o the LBO announcement. Super- response to the LBO announcement. chain should have no return with make Fi- nally, if LBO's are expected to product-market competition tougher, direct then ative share price response, while noncom- competitors should experience a neg- petitors should experience share no significant way, Supermarkets I examine price response. the leveraged bof Safe- Shop Kroger, the largest leveraged and the leveraged General, and Stop & uyouts recapitalization of undertaken dow begins 30 days prior to the in the industry. The event win- transactions first an- nouncement suggestive leveraged that an LBO or ends on the day of the recapitalization might occur and or leveraged recapitalization.8 it was undertaking an LBO firm's final an- nouncement that Announce- separate competing 7Michael D. Whinston and Scott C. Collins (lines in their event study airlines from 1992) press. of the entry of People Ex- noncompeting air- victed insider-trader Ivan Boesky 8This wide event window was chosen because con- the Securities ing in at least one of these Exchange Commission for insider-trad- was investigated by transactions. reflecting the market's expectation change the LBO, as long as no confounding in the value of the rival firm of the dmation was released within the event iue to nfor- win- dow. each leverage transaction The first announcement leading up its date are listed in Table 1. Table under study and to lists the date of the announcement that the 1 also leverage The daily stock returns transaction would definitely rivals are studied. The rivals are listed of 13 supermarket occur. Table derived 2. These rivals represent all firms that in from at least 80 percent of their rcevenues ontinu- ously from January supermarket sales and traded ber 10, 1988, of the Kroger the date of the announcement 1, 1985, through Octo- rdata on the stock-market The event study leveraged was conducted using ecapitalization. returns of the daily supermarket timated, chains. The equation to be es- has the form a variant of the basic following form:9 market model =it,a +,3iRmt + E,SijDjt + eit where Rit = the return on the value-weighted firm i's return at date t, Rmt = AMEX index at date t, NYSE/ events, estimated, ai, f8i, and j indexes the four 3ij are parameters to be dummy event window variable eit is an error for event which equals term, 1 and D - is a during the start on the first trading day of The stock return data j uand 0 otherwise. sed for estimation extend through Kroger's announcement 1985 and of 9For a discussion of the market model, see Eugene F. Fama (1976). VOL. 85 NO. 3 CAPITAL CHEVALIER: STRUCTURE AND MARKET COMPETITION TABLE 1-EVENTS INCLUDED IN THE EVENT ANALYSIS 419 Event Safeway LBO First announcement Date 6/13/86 3/10/87 1/15/88 9/13/88 Date final transaction announced 7/29/86 4/23/87 3/1/88 10/10/88 Dart Group announces that it has acquired a 6-percent stake in Safeway Supermarkets General LBO Dart Group proposes to buy Supermarkets General Stop & Shop LBO Dart Group announces that it seeks a major stake in Stop & Shop Kroger leveraged recap The Haft family (who control Dart Group) reveals that it has a major stake in Kroger Note: The event window is from 30 days prior to the first announcement, through the announcement of the final transaction. TABLE 2-RIVAL SUPERMARKET CHAINS INCLUDED IN THE EVENT STUDY Albertsons American Stores Brunos Delchamps Food Lion Foodarama Giant Food Stores Great Atlantic and Pacific Tea Company Hannaford Brothers Marsh Supermarkets Ruddick Weis Markets Winn-Dixie to have a single event response, and the firms not directly competing with the lever- aging firm are constrained to have a single event response for each event. These con- straints cannot be rejected at conventional significance levels. For each event, volumes of the Supermar- which chains competed with the chain un- dertaking the LBO. This book lists the names of stores operating in each of the Metropolitan Statistical Areas (MSA's) in the United States. The Supermarket News ket News's annual Distribution Study of Grocery Store Sales were used to determine its leveraged recapitalization in October 1988. These data are obtained from the Prices. for Research in Security Center are esti- The event response parameters mated using seemingly unrelated regres- sions (SUR). This methodology is employed because the error terms from the market- chain model equation for a supermarket correlated should be contemporaneously with the error terms for other supermarket the For each debt event, I calculate chains. that com- event coefficient of firms average and I with the leveraging firm, pete directly calculate the average event coefficient of with the that do not compete directly firms compet- leveraging firm. The firms directly are constrained ing with the leveraging firm guide lists store names, not parent firms. Information from annual 10K filings with the Securities and Exchange Commission used to link store names to parent firms. Two supermarket chains to were considered be in direct competition with one another if there was any MSA in which both owned stores listed in the Supermarket News guide. B. Results and the 1988 Retail Tenants Directory were Table 3 shows the results of an SUR estimation of return responses to the four events. The return responses for the com- peting firms are shown in column 1; the return responses for noncompeting firms are shown in column 2. The table shows that of the competing the return firms responses 420 THE AMERICAN ECONOMIC REVIEW JUNE 1995 TABLE 3-EVENT COEFFICIENTS Event coefficient for other supermarket chains Event (1) (2) Competing Not competing Safeway LBO 0.003168** 0.001614 (0.000966) (0.001044) Supermarkets 0.001782 - 0.000175 General LBO (0.001141) (0.000817) Stop & Shop LBO 0.001573 -0.000082 (0.001381) (0.000793) Kroger leveraged 0.001857* 0.000991 recapitalization (0.000930) (0.001229) Notes: Column 1 reports event coefficients for firms competing in some of the same MSA's as the \"event\" firm; column 2 reports event coefficients for firms com- peting in none of the same MSA's as the \"event\" firm. The coefficients were estimated using seemingly unre- lated regressions. Standard errors are given in paren- theses. *Statistically different from zero at the 5-percent level. **Statistically different from zero at the 1-percent level. are positive for all with the pected to become softer hypothesis four events, consistent that competition is ex- The return following the LBO. icant at the 5-percent responses are statistically signif- and Kroger events but level for the Safeway cant at only the statistically sSupermarkets General 12-percent level for the ignifi- percent level for the event and at the 25- The joint hypothesis tStop & Shop event. hat the event coeffi- jected at the 1-percent cients for all four events equal zero is re- significance level. ing firms are positive for The return responses for the noncompet- events, zero at standard but are not statistically different from two of the four Safeway zero at the event is significance levels. The prising 12-percent statistically different level. It is not sur- from there is some that this event is the one in which firms since the response of the noncompeting large LBO in this Safeway LBO was the first speculation tant for this LBO. effect might have been impor- industry, and thus, the pothesis that the event coefficients However, the joint for all hy- four events even the 55-percent equal zero cannot significance level. be rejected at competing firms, while The measured return responses for the cant as a group, The Safeway event, appear to be quite statistically ssmall. ignifi- mated to lead to an abnormal for example, is esti- value for the competing percent, firms oif only 0.32 ncrease in even smaller. and the other event coefficients are sponse However, this small event re- even the \"direct may be due to the small the firm undertaking competitors\" compete amount that ple, at the time of the the LBO. For exam- with recapitalization, American Kroger leveraged approximately Kroger 1Stores owned 25 states. However, owned approximately ,500 stores in 29 1,400 stores states; in the two firms in only seven competing I could find in the same MSA records of of the coefficients states. Thus, the magnitudes the changes estimated do not reflect those operations in the expected profitability of that actually competes with the of each supermarket rival leveraging chain.10 present discounted value of The event-study results suggest that the future profits when a rival supermarket chain of a supermarket chain the expected rises that it is recapitalization. undertaking an LBO or leveraged announces tent competition with the hypothesis that These results are consis- product-market to become softer. In the following the LBO is expected hypothesis the entry and exit of supermarket is tested further next section, this using data on chains. III. Empirical Predictions for Entry and Expansion The theories of capital structure and product-market competition changes in capital structure posit that toughness of product-market change the an LBO competition. If market changes the toughness of product- competition, then, following the 10R. Preston McAfee and Michael A. Williams (1988) make a similar point about the estimated magni- tudes of the effects of mergers on rival firms. VOL. 85 NO. 3 CHEVALIER: CAPITAL STRUCTURE AND MARKET COMPETITION 421 LBO of a supermarket market chain in a local market, one should observe a change in the structure of the local market.\" Specifically, if an LBO leads to an increase in the \"toughness\" of product-market com- petition, then rival firms would want to exit the local market or close stores in the local market. On the other hand, if a leverage increase leads to a decrease in the tough- established supply channels and, therefore, can begin to respond quickly to a change in the competitive situation in the local mar- ket. Supermarket trade sources confirm that de novo entry into a city requires more time and planning. Obviously, this empirical strategy assumes that the LBO changes capital structure and that these changes in capital structure im- ness of product-market competing to add in the local market stores in the local market competition, there, awould want firms wnd ould firms wnot competing ine whether, controlling In the analysis which follows, ant to enter. I will exam- conditions, adding or subtracting stores in supermarket chains tend to be for local market dominated by LBO firms. I markets percentage change in total supermarkets measure the across ential importance. cities in which LBO's were of differ- cated by the fact that the LBO The analysis is compli- very recent. The empirical taken strategy episode is under- gradual here assumes ture with more firms movement ttoward a market struc- hat one should observe tion is \"softened\" by and stores if competi- observe structure with gradual movement LBO's; one should petition fewer firms atnd stores oward a market if com- adjustments may not have occurred To address is \"toughened\" by LBO's. the problem that all post-LBO time the data were by the tional tests are undertaken. constructed, decision or subtract stores in cities in which by large supermarket chains I examine two addi- to add the were incumbents. I they local markets nonincumbent eseparately study whether done because operating osxperience ne would upermarket chains. This is de novo entry by expect tlocal real estate and in a local market conditions and has is familiar hat a chain with competition 1\"Sutton (1991) refers to the \"toughness\" of will refer to the \"toughness\" of as determining market structure. Here, I price petition. place on both the price and I use this term because product-market com- methodology in this paper price competition from changes cannot quality the competition takes in quality competition. separate dimensions. The changes in pact subsequent tion. If, on the other product-market competi- some way endogenously hand, the LBO product-market determined by were in ence about competition, then any infer- market issue will be taken up in competition would the effect of LBO's bon product- Section VI. Section IV and e spurious. This IV. Data market The data consist Areas (MSA's) chains in 85 oin 1985 and 1991. Metropolitan Statistical f information on super- The data are drawn from Progressive lication 1992) publication Market Grocer's (1986, Scope.12 This pub- number lthe most-populated 100 MSA's in oists the supermarket chains f stores operated by each chain and the in United States.'3 The book also lists the total number the dependent studied consist of those MSA's firms of supermarkets owned in the MSA. The 85 MSA's by in- among the largest 100 in both 1985 and which were 1991 and for which nition of the MSA borders the official remained un- Census defi- 12This to determine supermarket is a different data source The Progressive of the number Grocer data locations for than the one used the event study. chain counts, but they are obtained in an MSA. of stores owned provide high-quality counts The Supermarket by each supermarket News data include newspapers study, and are of lesser quality. by surveying the local mine whether firms competed the Supermarket News data were used to deter- For the event MSA's. unlike MSA's. the Progressive For this reason, a source in any of the same Grocer data wsource, as needed included which, all 13Some against microfilm of the Progressive Grocer data were checked confirm the quality of the data copies of old telephone books to source. 422 THE AMERICAN ECONOMIC RE;VIEW JUNE 1995 changed between the two years.'4 The use since be of concern, data may of MSA-level that an MSA is to assume there is no reason market of the relevant measure the correct when considering supermarket competition. my examination of the Progressive However, Grocer data shows that the MSA's corre- supermar- to divisions of large closely spond ket chains. In general, all of the supermar- From these sources, a defini- publications. A lever- tive list of LBO's was assembled. aged or LBO firm is defined as a firm that recapital- an LBO (or leveraged underwent ization) any time between 1981 and 1990. local mar- typically exit several LBO firms kets following the LBO, usually by selling to another chain or spin- the local division soon managers ning it off to the division's kets that one chain has in overseen a division are are served by a single division by a single divisional manager and Furthermore, one important way in which warehouse. supermarkets weekly describe circulars in the local compete is by distributing newspapers that week. A single flyer is generally the sales in the supermarkets that all of the supermarkets approximately, in a division or, issued for MSA. all of the supermarkets in an classified All of the firms in the 85 MSA's are dertaken by whether or not they have un- divide firms into low-debt and high-debt an LBO. I use this mechanism to firms available for privately owned firms. The because actual leverage ratios are un- power of that many the test is weakened by the fact have omay they did not undertake reasonably high f the \"low-leverage\" firms levels of debt although in two ways. First, quarterly editions of The information on LBO's an LBO.'5 was obtained Mergers and Acquisitions contain all owner- ship transactions (including greater than $1 million. Second, LBO's) of ences to transactions market parent companies involving in the sample the all refer- super- were searched using Predicasts Funk and Scott Index, United States, which indexes Super- market News, Supermarket Business, and Progressive Grocer, the major industry trade 14Unfortunately, because the MSA's were redefined for most of New England, the Bridgeport, Connecticut, MSA is the only New England MSA appearing in the sample. This removes from consideration most of one LBO chain which was very successful (Stop & Shop) and most of another which was very unsuccessful (Supermarkets General). I have confirmed that the debt ratios of non-LBO firms with publicly traded debt or equity are in fact, much lower than the debt levels of LBO firms. after the LBO. In total, 633 of the supermarkets in the study were sold in a 13,512 post-LBO asset sale. Of markets, 187 were sold these 633 super- management sion. in a second LBO of the divi- to the division's they were always owned by My approach is to treat the assets as if the eventual purchaser.'6 proach because, otherwise, I take this conservative ap- increased entry into LBO markets simply one would see because of these asset cause of a change in post-LBO product- transfers, not be- market California division For example, Safeway competition. Safeway LBO tsold its Southern after the in my sample that were part of Safeway's in 1986. o Vons shortly Here, I add the stores Southern total for 1985. California division to Vons's nia city, Thus, Cstore the net total of Safeway the change in stores for a Southern and Vons stores for Vons equals alifor- opened and 1991. The change in the or closed in that city between number of 1985 Safeway city equals zero. stores in any Southern Cdent variables ialifornia ndepen- market, the same convention such as the LBO share of In constructing is used. a dled in a similar way. The stores of Mergers among non-LBO firms were han- firms two were always which merged one exception to this rule owned by the same firm. were treated as if they The tion about the acquisitions is that informa- independent able. Purchases of small chains was not generally of very small avail- by in the independent chains entry chains or expansion. sample are thus counted as 16This the dependent treatment is undertaken in constructing and the independent variables that both will be described later. VOL. 85 NO. 3 CHEVALIER: CAPITAL STRUCTURE AND MARKET COMPETITION 423 Information about asset sales was ob- tained by checking the Wall Street Journal Index, Mergers and Acquisitions, Supermar- ket News, Supermarket Business, and Pro- gressive Grocer. Demographic obtained from Donnelly Marketing data are mation Services, a market research firm Infor- which provided the demographic data for the Progressive Grocer volume. V. Methodology and Results A. Full-Market Regressions changed I first test the hypothesis that LBO's termining product-market fewer stores whether LBO's lead to more or competition by de- a market \"fitting\" in the local market. If an LBO than it could before (adjusting for can support more stores following other changes the hypothesis in the market), tuct-market competition. that LBO's \"soften\" prod- his supports this supports the hypothesis that LBO's If fewer stores fit, \"toughen\" product-market competition. the percentage change in the number of The strategy employed here is to measure stores and to check whether this measure is re- in each MSA between 1985 and 1991 lated to the share of stores in each market in 1985 owned by chains that eventually undertook LBO's. of how market period structure This allows measurement changed for several The specifications in this section control in which the LBO's took place. over the contribute factors that might be expected to supermarkets to the growth in a local market of the number period: over this of holds in the MSA and household growth the growth in the number of house- adjusted dian income for MSA area, the growth in me- in the share of households that have an and its square, and the change income of less than $10,000 These variables are described wiere included. n Table 4. tics implicitly assumes The use of these five market characteris- in an equilibrium state in 1985: that each market the market changes was should be due to changes in the market structure between 1985 and 1991 in characteristics between 1985 and adjust for the possibility that an MSA was 1991. To librium in an \"over-stored\" or \"under-stored\" equi- teristics in 1985, I also include ment of a city's of the market in 1985: tawo charac- measure- number of stores per household deviation from the expected in 1985, and a measure oin Table 4. The variable 1985. These variables f market aconcentration of LBO firms in the MSA, ore also described f interest, the share in stores in the market in 1985 owned by a is the share of supermarket chain LBO by 1990. that would undertake an of the variables used in this and subsequent Table 4 provides summary statistics for all specifications. Results for least-squares an ordinary centage city between 1985 and 1991 on the LBO change (OLS) regression in the number of stores of the per- in a share of the market and the controls for market conditions described above are shown market has a positive coefficient, Table 5 shows in column t1hat the LBO share of Table 5. of the coefficient the 22-percent is only statistically significant but the coefficient level. The magnitude of the at owning an LBO, 10 percent implies tohat, if a firm f the stores undertakes in an MSA is expected the number it would otherwise. to grow bof stores in the market more than may of LBO's on the toughness of product- be due to there being, Ty 1-percent his insignificant effect in fact, no effect market fact that there simply may not have been competition or may be due to the enough time spond fully tfor market conditions to re- preliminary attempt The second column of Table 5 makes a o the LBO's. effect might have been observed to ascertain whether atime elapsed since the LBO's. It repeats had more n tbut separates specification f the first column he regression ointo two groups: the share of stores that the LBO share of total stores undertook LBO's prior share of stores that undertook to 1988 and the ing or after 1988. This is done because, LBO's dur- if it takes time for the market to adjust changes in one might not ex- to pect competition, LBO's. Indeed, Table 5 shows that the co- to see much response to the later efficient for which took place prior the store share to 1988 is of LBO firms positive 424 THE AMERICAN ECONOMIC REVIEW TABLE 4-SUMMARY STATISTICS Variable Description A. Variables: Change in households (10,000's) Change in the number of households in [percentage change households] the MSA between 1985 and 1991 Change in households per square mile Change in households per square mile in [percentage change households the MSA between 1985 and 1991. This per square mile] is included because the change in households may have a different im- pact if spread over a very large or very small area. Change in median income ($10,000's) Change in median income in the MSA [percentage change in median income] between 1985 and 1991 Change in median income squared Change in squared median income in the (in $1 x 108) [percentage change MSA between 1985 and 1991 in squared income] Change in the share households with Change in the share of households with income less than $10,000 annual incomes of less than $10,000 in the MSA between 1985 and 1991 Deviation in mean stores per household MSA's deviation in 1985 from the num- [percentage deviation in mean stores ber of stores that it would be pre- per household] dicted to have given the number of households in the MSA in 1985. I estimate that MSA's in 1985 have 40 stores plus 2.3 x 10-4 stores per household. Share of LBO firms Share of stores in the MSA in 1985 owned by firms that would undertake LBO's by 1991 Share of early-LBO firms Share of stores in the MSA in 1985 owned by firms that would undertake LBO's prior to 1988 Share of late-LBO firms Share of stores in the MSA in 1985 owned by firms that would undertake LBO's in 1988 or later Herfindahl index Sum of the squared market shares of the five firms with the largest market shares in the MSA, where a firm's market share is defined as its share of total stores in a market B. Incumbent Firm Variables: Store share, non-LBO incumbents Incumbent firm's share of the total stores in the MSA in 1985 Store share, LBO incumbents Incumbent firm's share of the total stores in the MSA in 1985 Total stores in chain, non-LBO incumbents Total stores in the sample of 100 MSA's owned by the incumbent in 1985 Total stores in chain, LBO incumbents Total stores in the sample of 100 MSA's owned by the incumbent in 1985 JUNE 1995 Standard Mean deviation 5.50 5.93 [12.0] [10.1] 49.7 157 [0.01] [0.03] 1.53 4.73 [64.2] [16.5] 9.95 4.66 [172] [53.7] - 0.0534 - 0.0227 0.00 30.0 [-7.11] [24.8] 0.220 0.191 0.108 0.156 0.111 0.128 0.120 0.020 0.116 0.098 0.143 0.108 307 284 283 227 VOL. 85 NO. 3 CHEVALIER: CAPITAL STRUCTURE AND MARKET COMPETITION TABLE 425 4- Continued. B. Incumbent Firm Variables: (Continued) Number of firm-MSA observations in which non-LBO incumbents add stores: Number of firm-MSA observations in which non-LBO incumbents neither add nor subtract stores: Number of firm-MSA observations in which non-LBO incumbents subtract stores: Number of firm-MSA observations in which LBO incumbents add stores: Number of firm-MSA observations in which LBO incumbents neither add nor subtract stores: Number of firm-MSA observations in which LBO incumbents subtract stores: Number of MSA's in which de novo entry occurs: 79 20 85 47 19 47 39 TABLE 5-OLS SPECIFICATIONS Coefficients Variable Constant Percentage change in households Percentage change in income Percentage change in income squared Change in share with income less than $10,000 Percentage change in households per square mile Percentage deviation from mean stores per household Herfindahl index Share LBO Share early LBO Share late LBO Regional dummies included? R2 (1) - 0.0070 (0.1386) 0.6347** (0.1569) - 0.6731 (1.1708) 0.1974 (0.3558) - 0.0071 (0.0084) 58.7209 (54.0745) -0.0866 (0.0575) - 0.1079 (0.2818) 0.0966 (0.0775) (2) 0.0075 (0.1380) 0.5952** (0.1581) - 0.7661 (1.6374) 0.2282 (0.3537) - 0.0076 (0.0084) 60.0863 (53.6788) -0.0831 (0.0571) -0.1171 (0.2798) 0.1736* (0.0931) -0.0175 (0.1094) (3) - 0.0006 (0.1572) 0.5377* (0.1651) - 0.5778 (1.1935) 0.1918 (0.3630) - 0.0044 (0.0088) 52.8402 (54.2677) -0.1194 (0.0620) -0.2920 (0.3036) 0.1438 (0.1369) 0.0280 (0.1141) yes 0.34 85 N no 0.30 85 no 0.32 85 Notes: The dependent variable is the percentage change in the total number of stores in the MSA between 1985 and 1991. Standard errors are in parentheses. *Significantly different from zero at the 5-percent level. **Significantly different from zero at the 1-percent level. 426 THE AMERICAN ECONOMIC REVIEW JUNE 1995 significance and significant at the 7-percent for the share of later level. The coefficient LBO's is insignificant at standard levels. It has been suggested that regional in Section III, 48 acquisitions as described It is the expansion chains are left for study. decisions of these firms in each of the 85 in which they are incumbents that markets dummy variables regression to control for unmodeled city should be included in this heterogeneity. The regression was reesti- mated using dummy variables for the Northeast, Midwest, and South (with the West as a base case). None of the dummy variables the 30-percent was statistically significant at even dummies level. The inclusion of the the share of early LBO's slightly, shrank the estimated coefficient of from 0.174, and decreased the statistical to 0.144 significance level of the coefficient to 30 percent. The coefficient became ftion that the presence These results offer a preliminary sugges- positive, but remained or late LBO's insignificant. the market firms structure. However, does lead to a change of leveraged in market in than a change in the \"toughness\" of many hypotheses other product-market competition could be put forth to explain if undertaking an LBO greatly these results. Ffirm's total costs, then LBO dor example, ecreased it profitable by these firms to expand. Markets firms might find a store growth than other markets. In the might experience fpaster total opulated following section, will avoid these alternative I introduce tests which which attempt hypotheses and changes expected to occur relatively in market to measure structure which separately those ing the LBO quickly would follow- be expected to take more time to occur. from those which might be B. Expansion by Incumbent Firms owhy large supermarket In this section, I examine tively competing to add or subtract in a market chains that are ac- the question f might choose net. To do this, I identify stores in that market on the largest sample.17 After number adjusting of stores in 1985 in the the 50 chains with for mergers and firms. 170ne set of firms is left out of the sample of top These firms are those involved in the only major will be studied. These firms account for 6,068 of the 13,512 supermarkets MSA's firm-city pairs in the study. in the bent in the city in 1985. in which There are a total of 297 the firm is an incum- stores added or subtracted Because of the small, integer number of local market, one should not ignore the by a chain in a discreteness of the data when analyzing these decisions; vations for incumbent in 36 percent adds or subtracts Thus, I adopt an ordered-probit methodol- no more than one store. firms the incumbent of the obser- ogy, estimating market whether each large super- ther adds nor subtracts chain adds stores in a market, or subtracts stores in a nei- the determinants of these decisions stores in a market.18 Because market, very different for LBO and non-LBO in- may be cumbent firms, the specifications mated separately for the two sets of firms. are esti- firm's I measure the relationship between a the share of rival stores in the market in decision to add or subtract stores and 1985 owned by firms undertake LBO's. I control that would eventually changes in the market for demographic possibility over-stored in 1985. that the market was under- or and control for the are described in Table 4. These control variables antitrust challenge period. After the federal antitrust to a supermarket merger during the decided by American Stores, not to challenge Office decided to pursue a challenge the California Attorney the purchase of Lucky supervisory bodies General's Stores under the California of the merger tied up in the courts antitrust statutes. The case was American tions and of the two firms Stores was not allowed for over a year, during which time and wto merge the opera- merger closing new stores in California. The parties as restricted from opening results similar. iancluding them re left out of the specifications here, though to this were checked and are extremely 18The context of measuring ordered-probit firms the determinants of how methodology was used in the many Peter cCompete . Reiss i(n a city in Timothy 1987, 1990). F. Bresnahan and VOL. 85 NO. 3 CHEVALIER: CAPITAL AND MARKET STRUCTURE COMPETITION TABLE 6-MAXIMUM-LIKELIHOOD ESTIMATION RESULTS FOR INCUMBENT FIRMS 427 A. Non-LBO incumbents Marginal effects dPr[y=-1] Variables Change in households Change in income Change in income squared Change in share with income less than $10,000 Change in households per square mile Deviation from mean stores per household Total stores Market share Herfindahl index Share LBO Exit threshold Entry threshold Number of observations Coefficient 0.0339a (0.0198) -2.7210 (2.0230) 0.2590 (0.1850) - 0.2128 (0.1435) -0.0008 (0.0006) 0.0010 (0.0038) -0.0012** (0.0004) 2.1970* (1.0566) 2.1396 (1.7995) 1.7016* (0.7557) 0.0996 (0.6693) 0.4173 (0.6697) 184 dPr[y=1] B. LBO incumbents Marginal effects dPr[y=-1] Coefficient - 0.0262 (0.0262) 3.9310 (2.6440) - 0.3690 (0.2400) 0.2154 (0.1764) 0.0005 (0.0063) 0.0032 (0.0046) 0.0007 (0.0006) -2.7866a (1.5580) 0.1315 (3.6680) 1.7620 (1.1292) 0.8998 (0.9534) 1.3547 (0.9561) 113 dPr[y=1] dx -0.0134 1.0788 - 0.1027 0.0844 0.0003 -0.0004 0.0005 -0.8711 -0.8483 -0.6746 dx 0.0129 -1.0358 0.0986 - 0.0810 - 0.0003 0.0004 -0.0004 0.8363 0.8145 0.6477 dx 0.0102 -1.5337 0.1440 - 0.0840 - 0.0002 - 0.0012 -0.0003 1.0872 -0.0513 -0.6875 dx - 0.0054 0.8153 - 0.0765 0.0447 0.0001 0.0007 0.0001 -0.5780 0.0273 0.3655 Notes: The dependent variable has the following values: Yij = + 1 if firm i withdraws at least one store from market j, Yij = 0 if firm i neither adds nor withdraws stores from market j, and Yij = + 1 if firm i adds stores in market j. Standard errors are reported in parentheses. aSignificantly different from zero at the 10-percent level. *Significantly different from zero at the 5-percent level. **Significantly different from zero at the 1-percent level. I also include variables to describe the rivalry faced by firm i in market j. These are a measure of concentration in the local market and firm i's share of total stores in market j. To control for chain size, I in- clude a variable measuring the total number of stores that firm i has in the entire sample in 1985. These are described in Table 4 as well. The variable of most interest, Share LBOij, is the share of LBO firms among firm i's rivals in market j. Firm i's own stores are not counted when constructing either the numerator or the denominator of this share. Thus, the variable characterizes the rivalry facing firm i in market j. Part A of Table 6 shows the results of this specification for non-LBO incumbents (a to- tal of 184 firm-market pairs). The coeffi- cients for the demographic variables, with the exception of the change in households per square mile, have the same sign as the corresponding variables in Table 5. Only the coefficient for the change in households is statistically significant. The coefficient for the firm's market share is positive and sig- nificant. This suggests that firms with large market shares in a market are the most likely to expand. The results also suggest that large chains are less likely to expand than smaller chains. This result should be interpreted with caution because the coef- ficients are only estimated for a sample of fairly large chains. The coefficient for the share of firm i's rivals in market j which have undertaken LBO's is positive and significant at the 3- 428 THE AMERICAN ECONOMIC REVIEW JUNE 1995 percent level. The \"marginal effects\" show that, in a city in which all market character- istics are held at their mean, adding the LBO of a firm with a 10-percent market share would increase the probability that a given non-LBO firm will add stores in the market by approximately 6.5 percent. This result supports the results in Section I which suggest that LBO's lead to a decrease in the toughness of competition in the market; when a firm undertakes an LBO, rival non- LBO firms in the market find expansion attractive. Part B of Table 6 repeats the specifica- tion of Part A, except the expansion deci- sions of LBO firms are used as the depen- dent variables. The coefficients for all of the demographic variables are statistically in- significant, and many have the opposite signs from the previous specifications. In contrast to the specification for non-LBO stores, the coefficient for total stores is positive, and the coefficient for the firm's own market share is The coefficient for the share of firm i's negative. rivals in market j which are LBO firms is positive, as in the previous specification, but significant only at the 12-percent level. This result provides some evidence for the hy- pothesis that LBO's decrease the \"tough- ness\" of product-market competition, al- though the results are clearly weaker than for non-LBO firms. As mentioned before, one reason for ex- amining expansion by incumbent firms sepa- rately from new entry into local markets is that one would expect that incumbent firms would be able to begin to respond relatively quickly to local market conditions. Thus, one would expect that a firm in response to an LBO would have added at adding stores least one store by 1991, since the last LBO took place in early 1990. In Table 7, I divide LBO's into those that took place prior to 1988 (early LBO's) and those that took place during or after 1988 (late LBO's) and re- peat the specifications of Tables 6. In both columns, the coefficient for the early-LBO share is only slightly larger than the coeffi- cient for the late-LBO share. The sis that the coefficient for early LBO's is hypothe- larger than the coefficient for late LBO's is not rejected at standard significance levels. This is consistent with the view that expan- sion by rival incumbents in response to LBO's should begin quickly. Several tests were undertaken to test the robustness of the results. First, because these specifications use firm-level data, the concern arises that unmodeled firm hetero- geneity may affect the basic results. These results were reestimated including firm dummy variables. Because of the number of firm dummy variables relative to the num- ber of observations, it was necessary to pare down the specification in order to estimate this relationship. The basic specification was thus reestimated pooling data from LBO and non-LBO firms and including firm dummy variables. The coefficient for the LBO share of the market remains posi- tive and is statistically significant at the 6-percent level. Unfortunately, the results could not be reestimated with city or even state dummy variables because there are not enough ob- servations per geographic area. However, the results for both LBO and non-LBO firms are robust to the inclusion of dummy variables for the Midwest, West, South, and East Coast. The coefficient for share LBO in the non-LBO firm regressions remains positive and significant at the 3-percent level; the coefficient for share LBO in the LBO firm regressions remains positive and significant at the 6-percent level. Cross-firm within-city correlation of the error term might lead to the estimation of inflated significance levels. The regression results were checked for robustness to this possibility. Table 7 was reestimated using one observation per MSA. The dependent variable took the value of 1 if the supermar- ket chains in the sample in the MSA added stores on net; it took the value of 0 if they neither added nor subtracted stores, and it took the value of - 1 if they subtracted stores on net. The right-hand-side variables took their mean value for the MSA. The specification was done separately for LBO firms and non-LBO firms, as before. In the specification for non-LBO firms, there were VOL. 85 NO. 3 COMPETITION STRUCTURE AND MARKET CHEVALIER: CAPITAL TABLE 7-MAXIMUM-LIKELIHOOD ESTIMATION RESULTS FOR INCUMBENT FIRMS 429 A. Non-LBO incumbents Marginal effects dPr[y=-1] Variables Change in households Change in income Change in income squared Change in share with income less than $10,000 Change in households per square mile Deviation from mean stores per household Total stores Market share Herfindahl index Share early LBO Share late LBO Exit threshold Entry threshold Number of observations Coefficient 0.0330a (0.0202) -2.6630 (2.0190) 0.2550 (0.1850) - 0.2125 (0.1438) -0.0008 (0.0006) 0.0012 (0.0038) -0.0012** (0.0004) 2.1043* (1.0503) 2.0584 (1.8184) 1.7032a (0.9073) 1.4087 (0.8865) 0.0996 (0.6693) 0.4173 (0.6697) 184 dx -0.0131 1.0553 -0.1010 0.0842 0.0003 -0.0005 B. LBO incumbents Marginal effects dPr[y=-1] Coefficient -0.0283 (0.0265) 4.2400 (2.6840) -0.3950 (0.2430) 0.2358 (0.1793) 0.0008 (0.0064) 0.0032 (0.0046) 0.0006 (0.0006) -2.7507a (1.5434) - 0.2282 (3.6914) 1.9126 (1.1192) 1.13798 (1.3221) 0.9035 (0.9542) 1.3866 (0.9569) 113 dx 0.0110 -1.6511 0.1538 - 0.0918 - 0.0003 - 0.0013 - 0.0002 1.0712 0.0889 - 0.7448 dPr[y=1] dx 0.0126 -1.0152 0.0972 -0.0810 - 0.0003 0.0004 - 0.0004 0.8022 0.7847 0.6493 0.5370 dPr[y=1] dx -0.0057 0.8565 -0.0798 0.0476 0.0002 0.0007 0.0001 -0.5557 - 0.0461 0.0005 -0.8339 -0.8157 -0.6749 0.3864 0.2787 -0.5582 -0.5373 if firm i withdraws at least one store from market j, Notes: The dependent variable has the following values: Yij =-1 Yij = 0 if firm i neither adds nor withdraws stores from market j, and Yij = + 1 if firm i adds stores in market j. Standard errors are reported in parentheses. aSignificantly different from zero at the 10-percent level. *Significantly different from zero at the 5-percent level. **Significantly different from zero at the 1-percent level. 79 observations, and the coefficient for the average share of LBO rivals remained positive and was statistically significant at the 5-percent level. In the specification for LBO firms, the coefficient for the average share of LBO rivals remained positive but was statistically significant at only the 25- percent confidence level. The results for this LBO incumbent specification should be treated with extreme caution, however, as only 28 observations were available. Finally, the importance of the use of the ordered-probit specification was investi- gated. Table 8 reestimates Table 7 using the actual change in the number of stores for each incumbent firm as the dependent vari- able. Because the left-hand-side variable varies so much in scale, heteroscedasticity- robust standard errors are used, following the method of Halbert White (1980). The coefficient for the share of early-LBO firms remains positive and statistically significant at the 6-percent level in the regression for non-LBO incumbents. The coefficients for the share of late-LBO firms is ap- proximately zero. This is not surprising. The ordered-probit methodology examines whether any response to the LBO has oc- curred, while this specification measures the magnitude of the response. It is not surpris- ing that, by 1991, very few stores have been built \"responding\" to the later LBO's. The 430 THE AMERICAN ECONOMIC REVIEW JUNE 1995 TABLE 8-OLS ESTIMATION RESULTS FOR INCUMBENT FIRMS Coefficients (1) (2) Non-LBO LBO Variables incumbents incumbents Change in households - 0.0660 - 0.2690 (0.1170) (0.2350) Change in income 0.2860 14.0750 (6.2420) (14.1650) Change in income squared 0.0900 - 1.3100 (0.5750) (1.2500) Change in share with income - 0.4167 0.5441 less than $10,000 (0.4084) (0.9351) Change in households per - 0.0033 0.0003 square mile (0.0039) (0.0352) Deviation from mean stores 0.0575* 0.0268 per household (0.0250) (0.0435) Total stores - 0.0055* - 0.0005 (0.0022) (0.0042) Market share 7.0556 - 28.1176* * (5.4637) (10.2838) Herfindahl index 29.6984 30.6566 (11.9950) (27.0131) Share, early LBO 7.9665a 0.7648 (4.1466) (5.7235) Share, late LBO - 1.0105 - 5.8049 (3.9945) (9.0537) Constant - 6.4550* - 3.2238 (3.1960) (6.7199) R 2 0.12 0.14 Number of observations 184 113 Notes: The dependent variable is the number of stores that firm i has added to market j. It is negative if the firm has subtracted stores from market j. White (1980) robust standard errors are in parentheses. aSignificantly different from zero at the 10-percent level. *Significantly different from zero at the 5-percent level. **Significantly different from zero at the 1-percent level. results in column 2 of Table 8 offer no support respond ftor the hypothesis that LBO firms o the in this subsection LBO's of their rivals. sthat non-LBO The results markets dominated firms find expansion auggest ttractive in by LBO firms. As ex- pected, the results in this subsection are stronger additions than those which fpool the store trants. of incumbent irms and new en- tion that This is consistent expansion or contraction with the expecta- by in- cumbent response firms would market. to a change ibn conditions e the first dietectable n a local C. Entry entry by a large supermarket chain into a In this subsection, I examine de novo local market.19 I extended the data set to 1993 by searching announcements of new entry. This is done the Nexis data base for for the entry specifications and not for the expansion sons. First, schanges in market conditions more than bpecifications above for two rea- ecause one expects entry to lag expansion, recent data as possible for new entry. Sec- it is more important to have as ond, the supermarket trade press does not, in general, announce that a supermarket chain is opening new stores in a city in which it already report that a large supermarket has stores, though entering chain is it does column The specification a new local market. dependent variable 1 of Table 9 is a for the entry model in simple probit. no entry occurs in the city between 1985 takes the value of 0 if The and 1993. value of 1 if entry occurs in that period. The dependent variable takes the Entry is defined to occur when a large supermarket chain of more than 25 stores opens it was not an incumbent at least one store in an MSA in which cluded as market by KMart entry is the opening of in 1985. Also in- a hyper- gressions The variables included in the entry re- or WalMart.20 are a subset of those included in into local 19This is not the first paper de novo entry W. Cotterill markets in the supermarket to study novo entry not consider into cities by supermarket chains. and Lawrence E. Haller (1992) study industry. Ronald de firms. cof incumbent They do using Tthe leverage entry oharacteristics ver a different time period with 20A hypermarket is a full supermarket combined a different data hey consider set and methodology. nor WalMart owned a general merchandise store. While they opened hypermarkets, they are obviously 25 supermarkets at the time that neither KMart retailing chains and were thus included. large VOL. 85 NO. 3 CHEVALIER: CAPITAL STRUCTURE AND MVARKET COMPETITION TABLE 9-RESULTS FOR NEW ENTRY 431 (1) Variable Change in households Change in income Change in income squared Change in share with income less than $10,000 Change in households per square mile Deviation mean stores per household Herfindahl index Share, early LBO's Share, late LBO's Constant 2 R Coefficient 0.0841* (0.0395) 4.6880 (2.9970) - 0.4950a (0.2910) 0.2859 (0.1806) - 0.0025 (0.0032) 0.0012 (0.0064) 0.0887 (2.9218) 2.4183a (1.1330) 0.6756 (1.2177) -1.518 (1.0364) (2) Marginal effects dPr[y =1] dx 0.0126 0.7038 -0.0743 0.0429 - 0.0004 0.0002 0.0133 0.3630 (3) Coefficient 2.23 x 10-6* (1.13 x 10-6) 1.15 x 10-4 (7.0x 10-5) 1.22 x 10-9a (6.75 x 10- 10) 0.0702 (0.0388) (3.97x 10-4) 6.95 x i04 (0.0021) 0.0147 (1.0290) 0.7904a (0.3835) 0.2960 (0.4317) 0.0758 (0.3248) 0.2010 4.19 X 10-4 Notes: The first column shows probit results. The dependent variable has the following = 0 if no entry occurs in market j; Yj = 1 if entry occurs. The second column values: Yj shows the marginal effects implied by the coefficients in column 1. The third column shows the results of the linear probability specification. Entry occurs in 39 of the 85 markets. Standard errors are reported in parentheses. aSignificantly different from zero at the 10-percent level. *Significantly different from zero at the 5-percent level. the expansion regressions above. Since these regressions examine entry at the market level rather than the firm level, firm-specific characteristics must be excluded. Table 9 shows that the coefficient for the early-LBO share is of a larger magnitude than the coefficient for the share of late- LBO firms. However, the hypothesis that the coefficient for the share of early LBO's is different from the coefficient for the share of late LBO's is rejected at only the 24- percent significance level. As expected, the share of early-LBO firms in the local mar- ket has a positive coefficient, statistically significant at the 3-percent level. The coef- ficient for the share of late-LBO firms is positive, but statistically significant at only the 58-percent level. This table thus sug- gests that large supermarket chains find en- try into local markets dominated by firms that undertook LBO's prior to 1988 attrac- tive. There is only very limited evidence, however, that entry has responded to LBO's that took place during or after 1988.21 The results of a linear probability model, shown in column 3 of Table 9 are substantially the same as the probit results. 21Once again, the results are robust to the inclusion of regional dummy variables. The coefficient for the share of early LBO's is positive and statistically signif- icant at the 10-percent level; the coefficient for the share of late LBO's is positive and significant at the 42-percent level. 432 THE AMERICAN ECONOMIC REVIEW TABLE 10-COMPARISON OF 1985 ACCOUNTING VALUES FOR FIRMS THAT WOULD EVENTUALLY UNDERTAKE AN LBO AND FIRMS THAT WOULD NOT EVENTUALLY UNDERTAKE AN LBO JUNE 1995 Mean Non-LBO t statistic of Accounting ratio Operating income/sales Net income/sales Market value/book value of assets Capital expenditures/assets Retained earnings/net income Dividends/net income LBO firm 0.0363 0.0040 0.8316 0.1461 0.3703 0.2375 firm 0.0395 0.0043 0.8194 0.1300 0.3266 0.1896 difference 0.48 0.14 0.10 0.80 0.76 0.39 VI. An Alternative Hypothesis If the LBO event selects for firms that were underperformers, then one might ex- pect accounting data to reflect that LBO firms were underperformers on average prior to their LBO's. Table 10 contains ac- counting data for all of the supermarket chains that were publicly traded in 1985.22 Eleven of these firms undertook an LBO after 1985, and 20 did not. The table shows that, on average, LBO and non-LBO firms do not differ signifi- In particular, LBO firms cantly. do not gen- A. Asset Sales erate significantly less operating income as a share of sales or less net income as a share were underperformers of sales. Furthermore, the ratio of market Even if LBO firms this would to book value of assets, a proxy for the to their LBO's, on average prior not necessarily affect the results of Sec- market's estimation of a firm's future is slightly higher for LBO firms. tion V if LBO firms sold off all underper- prospects, to assets firms in post- The ratio of capital expenditures to non-LBO divisions forming in Section IV, and the ratio of retained earnings to net LBO asset sales. As discussed higher for LBO firms, used in Section V assigns income are somewhat the methodology assets sold after LBO's to their eventual although the difference is not statistically owners. The sale of the division is not significant.23 for the LBO firm, as a loss of stores counted for nor is it counted as entry or expansion 22These data are from Compustat. More importantly, the pres- the purchaser. 23Firms which undertook LBO's early do not differ that were ence in a city of supermarkets significantly from firms which undertook them later, asset sale is not counted except that firms which undertook sold in a post-LBO LBO's early have in calculating the store share of LBO firms higher ratios of capital expenditures to assets than do in the city. The asset-assignment procedure firms that undertook LBO's later. The results of Section V suggest that su- firms find entry and expansion permarket attractive in markets dominated by firms that I have suggested that undertook LBO's. because are attractive entry and expansion less \"tough\" following competition becomes the explore I briefly an LBO. In this section, hypothesis that LBO's did not alternative change the toughness of product-market that un- that firms but rather, competition, dertook LBO's were weak firms. If weak LBO's, then one might ex- firms undertook pect to see entry and expansion occur in by LBO firms. dominated those markets helps to diminish the effect of these under- performing divisions on the results. B. Accounting Evidence VOL. 85 NO. 3 CHEVALIER: CAPITAL AND MARKET COMPETITION STRUCTURE 433 C. Event-Study Evidence The event-study evidence presented in Section II does not support the alternative hypothesis. If it was common knowledge that LBO firms were underperformers and their behavior was unchanged by the LBO, then the LBO announcements would not contain positive information for rival firms, and there would be no share price response. On the other hand, if LBO's were expected to increase the toughness of product-market competition, then rivals would be expected to experience a negative share price re- sponse to the LBO announcement. Thus, the event-study finding of a positive share price response of firms to rivals' LBO's is consistent with the findings here that LBO's decrease the toughness of product-market competition. The event-study evidence is in- consistent with the hypothesis that LBO's were undertaken by firms that were known underperformers and that product-market competition did not change much following the LBO's. D. Evidence from Early versus Late LBO's If LBO's select for firms that were chronic underperformers, and the LBO did not change the LBO firm's behavior, then one would expect to find that the LBO firm attracted entry and expansion even prior to its LBO. The pattern of coefficients for early and late LBO's in Tables 7 and 9 do not lend support to this suggestion. In particular, if LBO's select for firms that were chronic weak then one would expect that new underperformers, entry into a city (in Table 9) would have responded to late LBO's. It seems unlikely that these firms were chronically weak and yet did not attract entry. It might be suggested that this is because early-LBO firms were very poor firms, while late-LBO firms were good firms. The results in Table to this 7, however, run counter suggestion. These results show that non-LBO incumbents are almost as likely to have expanded in the presence of late LBO's as in the presence of early LBO's. The grocery trade press and firm lOK's and prospectuses indicate that the time elapsed between the decision to build a new store in a city and the actual building of the store is much longer for firms that are not incumbents in the city. Thus, if the LBO's that took place during or after 1988 actually changed product-market competition, one would expect to see that expansion by in- cumbent firms had responded by 1992, but one would not necessarily expect to see much evidence that new entry had re- sponded by 1993. However, if nothing changed at the time of the LBO, then one would expect to see either that neither entry nor expansion is correlated with these LBO's or that both entry and expansion are corre- lated with these LBO's. The findings in Tables 7 and 9 are consistent with the hy- pothesis that competition did change at the time of the LBO's. VII. Summary and Conclusion The principal results of this paper are that the announcement of an LBO in- creases the expected future profits of a firm's product-market rivals and that the presence of LBO firms encourages local entry and expansion by rivals. Both sets of results are suggestive that market leverage makes product- competition less \"tough.\" The re- sults lend empirical support to the theoreti- cal models of Bolton and Scharfstein Fudenberg and Tirole (1986), (1990), and Phillips (1991). The basic finding, that markets in which LBO's have occurred attract entry and ex- pansion, is consistent with the alternative hypothesis that LBO firms were simply underperformers prior to their LBO's. Evi- dence against this The results of this hypothesis was presented. that paper strongly suggest when firms product-market competition changes radically increase their leverage. However, it would be mine the dimensions on which interesting to deter- in the competition product market changes. For exam- ple, neither the event-study evidence nor the evidence on can entry, exit, and expansion distinguish the hypothesis that price 434 AMERICAN ECONOMIC REVIEW THE JUNE 1995 following less vigorous competition becomes an LBO from the hypothesis that quality following less vigorous competition becomes it has been suggested an LBO. For example, com- that LBO firms in the trade literature in the area of store quality, pete less fiercely placing their stores on slower renovation It has also been sug- and repair schedules. gested that LBO firms refuse to become embroiled in price wars with rivals. This which, if any, of these cannot identify study mechanisms brings about the change in product-market competition. An examina- and by leveraged tion of price competition unleveraged supermarket chains is the sub- ject of future research. that does suggest Finally, while this paper the nature of competition changes when firm leverage changes, the results do not to the necessarily make any contribution or not leveraged debate concerning whether Clearly, buyouts were value-maximizing. did not their rivals would rather LBO firms it and enter their markets. However, expand to know whether the amount is impossible and expansion deterred this entry that firms prior to LBO's was efficient or inefficient. Thus, while these results cannot prescribe an optimal capital structure based on prod- make clear the results uct-market outcomes, that product-market effects of capital- a com- be considered must market decisions ponent of the choice of optimal capital structure. REFERENCES David S. \"A Bolton, Patrick and Scharfstein, Theory of Predation Based on Agency Problems in Financial Contracting.\" American Economic Review, March 1990, 80(1), pp. 93-106. Brander, James A. and Lewis, Tracy R. \"Oligopoly and Financial Structure.\" American Economic Review, December 1986, 76(5), pp. 956-70. F. and Reiss, Peter C. \"Do Bresnahan, Timothy Entry Conditions Vary Across Markets?\" Brookings Papers on Economic Activity, 1987, (3), pp. 833-71. \"Entry in Monopoly Markets.\" Re- view of Economic Studies, October 1990, 57(4), pp. 531-53. E. Cotterill, Ronald W. and Haller, Lawrence \"Barrier and Queue Effects: A Study of Leading U.S. Supermarket Chain Entry Patterns.\" Journal of Industrial Economics, December 1992, 40(4), pp. 427-40. \"Horizontal Mergers, Collu- B. Espen. Eckbo, sion, and Stockholder Wealth.\" Journal of Financial Economics, April 1983, 11(1-4), pp. 241-73. F. Foundations of finance. New Eugene Fama, York: Basic Books, 1976. Jean. \"A 'Signal- and Tirole, Fudenberg, Drew Jamming' Theory of Predation.\" Rand Journal of Economics, Autumn 1986, 17(3), pp. 366-76. Guedes, Jose C. and Opler, Tim C. \"The Strategic Value of Leverage: An Ex- ploratory Study.\" Mimeo, Cox School of Business, Southern Methodist University, 1992. C. \"The Eclipse of the Public Michael Jensen, Corporation.\" Harvard Business Review, September-October 1989, 67(5), pp. 61-74. Kaufman, Phillip R. and Handy, Charles R. Supermarket prices and price differences: City, firm, and store-level determinants, Economic Research Service Publication TB-1776. Washington, DC: U.S. Depart- ment of Agriculture, December 1989. Peter. \"The Case for LBOs: The Magowan, Safeway Experience.\" California Manage- ment Review, Fall 1989, 32(1), pp. 9-18. Vojislav. \"Capital Structure in Maksimovic, Repeated Oligopolies.\" Rand Journal of Economics, Autumn 1988, 19(3), pp. 389-407. McAfee, R. Preston and Williams, Michael A. \"Can Event Studies Detect Anticompeti- tive Mergers?\" Economics Letters, 1988, 28(2), pp. 199-203. Mergers and Aquisitions. Philadelphia, PA: MLR, bimonthly, 1985-1992. Phillips, Gordon M. \"Financial Slack, Refi- nancing Decisions, and Firm Interaction.\" Working paper, Krannert School of Man- agement, Purdue University, 1991. _ . \"Increased Debt and Product Mar- ket Competition: An Empirical Analysis.\" VOL. 85 NO. 3 CHEVALIER: CAPITAL STRUCTURE AND MARKET COMPETITION 435 Working paper, Krannert School of Man- agement, Purdue University, 1992. Predicasts Funk and Scott index, United States. Cleveland, OH: Predicasts, annual, 1985- 1991. Progressive Grocer. \"Annual Report of the Supermarket Industry.\" April 1989, 68(4). . Market scope. White Plains, NY: MacLean Hunter Media, 1986, 1992. Retail tenants directory. Clearwater, FL: Na- tional Mall Monitor, 1988. Rotemberg, Julio J. and Scharfstein, David S. \"Shareholder-Value Maximization and Product Market Competition.\" Review of Financial Studies, 1990, 3(3), pp. 367-91. Spence, A. Michael. \"Capital Structure and the Corporation's Product Market Envi- ronment,\" in Benjamin Friedman, ed., Corporate capital structures in the United States. Chicago: University of Chicago Press, 1985, pp. 353-82. News. Distribution study of gro- Supermarket cery store sales. New York: Fairchild, an- nual, 1987-1990. Stillman, Robert S. \"Antitrust Horizontal Merger Policy.\" Journal of Financial Eco- nomics, April 1983, 11(1-4), pp. 225-40. Sutton, John. Sunk costs and market struc- and the ture: Price competition, advertising, evolution of concentration. Cambridge, MA: MIT Press, 1991. WaUl Street Journal Index. New York: Dow Jones, 1985-1991. Whinston, Michael D. and Collins, Scott C. \"Entry and Competitive Structure in Deregulated Airline Markets: An Event Study Analysis of People Express.\" Rand Journal of Economics, Winter 1992, 23(4), pp. 445-62. White, Halbert. \"A Heteroskedasticity-Con- sistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity.\" Econometrica, May 1980, 48(4), pp. 817-38.