Entry into Banking Markets and the Early Mover Advantage Allen

Entry into Banking Markets and the Early-Mover Advantage Allen N. Berger∗ and Astrid A. Dick† This version: January 2006 First version: November 2003 Abstract Using a sample for 1972-2002 with over 10,000 bank entries into local markets, we find a market share advantage for early entrants. In particular, the earlier a bank enters, the larger is its market share relative to other banks, controlling for firm, market and time effects, with a market share advantage for early movers between 1 and 15 percentage points, depending on the order of entry. The strongest early-mover advantage is for banks that were in our sample in 1972 and survive into the 1990s. Moreover, early entrants appear to have such hold in the market by strategically investing in larger branch networks. Even controlling for the potential survivorship bias, we find that a bank’s share decreases by 0.1 percentage points for a change in its order of entry from nth to (n+1)th. High growth markets show a smaller difference between late and early movers, consistent with a larger fraction of consumers yet to be locked in with a bank in these markets. JEL Classification: G2; L1 Keywords: Banks; market entry; market structure; firm strategy; first-mover advantage aberger@frb.gov Board of Governors of the Federal Reserve System, Washington, DC 20551 U.S.A., and Wharton Financial Institutions Center, Philadelphia, PA 19104 U.S.A. † astrid.dick@ny.frb.org Federal Reserve Bank of New York, New York, NY 10045 U.S.A. The opinions expressed do not necessarily reflect those of the Federal Reserve System or the Federal Reserve Bank of New York. The authors would like to thank the Editor, two anonymous referees, Adam Ashcraft, Nicola Cetorelli, Manfred Dix, Bob Hunt, George Kaufman, Matthew Shum, and James Vickery for insighful comments, as well as participants at the Federal Reserve Bank of Chicago Bank Structure and Competition Conference, the International Industrial Organization Conference, the Financial Management Association Conference, and a Bank of Canada seminar. Nathan Miller and Philip Ostromogolsky provided excellent research assistance. All remaining errors are the responsibility of the authors. ∗ 1 Introduction The advantages of early entry, such as a first-mover advantage, are frequently mentioned in both the economics and the business literature, yet the empirical research accompanying the theoretical developments since Stackelberg (1934) has been limited. An early-mover advantage might arise under certain elements that create obstacles to subsequent entry and allow incumbents to earn rents even when entry occurs. These elements include certain capital investments, such as building a clientele when switching costs are present, learning by doing, or economies of scale. The extant empirical literature exploring the relationship between order of entry and a firm’s market presence and performance has focused on differentiated products where innovation is central to product development. In contrast, this paper focuses on a service industry. In particular, within the context of the banking industry, this paper investigates whether early movers have an advantage over later movers by exploring whether the order of entry is related to the degree of market dominance. Various factors could give rise to such an advantage in banking. Both the anecdotal and empirical evidence suggest that consumer switching costs are significant in banking, with some evidence on supply-side factors such as economies of scale as well. The banking industry provides a unique opportunity to study such phenomenon, with the availability of decades of data that allows us to determine the date of entry and exit of thousands of firms into hundreds of local markets, with information on their market shares, as well as other firm characteristics. Moreover, the data provide substantial variation in dates of entry into a given market, as it is common in banking markets to find banks that entered several decades ago coexist with banks that entered only recently. While there is no direct research evidence indicating an early-mover advantage in banking, an abundance of anecdotal evidence suggests that, at least in the short-run and for large banks, a dynamic disadvantage exists for later entrants. In particular, established banks that expand into new markets appear to perform worse than their competitors, losing deposits, loans and profits.1 Some of the reasons put forth for this poor performance include lack of personal contact and loss of personal relationship for the client, “cookie cutter” products that are not tailored to the individual customer’s needs, and the banker’s lack of knowledge about the local community. Conversely, incumbent banks tend to accumulate proprietary information about their customers and their local Bank mergers are associated with runoffs of deposits and loans. DiSalvo (2002) finds that mergers often result in negative deposit growth for the consolidated institutions or smaller growth than that of non-merging competitors. Large banks also tend to have reductions in their small business lending in the aftermath of mergers and acquisitions that is picked up by their local market competitors (e.g., Berger et al. 1998). 1 1 communities. Indeed, an early-mover advantage may be more likely in banking than in other industries due to the importance of relationships. In order to illustrate the advantage of an incumbent, we use a variation of the von Stackelberg two-stage, two firm model based on Spence (1977) and Dixit (1979), which reinterpret the basic model as a reduced-form derived from short-run product-market competition under capacity constraints. While our analytical framework is quite simple, it embodies the main aspects of sequential entry under sunk costs. In particular, it illustrates how early movers can make their investment decisions strategically and fare better than later movers, by “preempting” the market through a higher capital investment, thereby limiting the erosion of rent due to entry. In this model, the incumbent or early-mover, who decides how much capital to invest by taking into account how the entrant will react to its choice of capital, ends up with higher capital investment and subsequent profits than the entrant. This captures the incumbent’s early-mover advantage. In our setup, capital investment is interpreted, without loss of generality, as building a clientele, with incumbents developing larger clienteles than entrants. In terms of the banking data, the model implies that incumbents should enjoy larger deposit market shares than later entrants. Our unique data set, which covers the period 1972-2002 and over 10,000 bank entries, allows us to identify the entry and exit date (if any) of all banking firms in urban markets — defined at the level of the Metropolitan Statistical Area — that occurred throughout the three decades, as well as the incumbent firms as of our starting date in 1972. At each point in time, we divide banks into four groups, based on how long ago they have entered the market, ranging from the last five years to over twenty years ago. The regression analysis focuses on the period 1992-2002, in order to have banks fall into each category. The data include 318 markets and more than 8,000 bank entries throughout this eleven-year period alone (of the 10,655 entries within 1972-2002, 5,381 occur within 1992-2002; of the other 5,274 occurring within 1972-1991, 3,115 of them survive into the 1990s). A test of an early-mover advantage is useful for many reasons. First, it may help in understanding the type of strategic competition that takes place in the industry. Second, measuring the magnitude of any early-mover advantage is a way to gauge how significant barriers to entry are in the industry, indirectly measuring the factors giving rise to them, such as switching costs and other consumer transaction costs on the demand side, and cost advantages from learning by doing, decreasing average costs and capital accumulation on the supply side. Moreover, banking markets make a great laboratory for the study of an early-mover advantage. Unlike some previous work, which relies on surveys, polls and various assumptions in order to distinguish pioneer firms, iden2 tifying incumbents and later entrants is straightforward in the case of banking markets given the nature of bank entry and the available banking data. In addition, the relevant geographic market in the banking industry, which tends to be local and defined at the metropolitan level, provides a wealth of cross-market variation. Our results show that banks that enter markets early enjoy larger deposit market shares, after controlling for firm, market and time effects. The later a bank enters, the smaller is its market share relative to early entrants. While all early entrants have an advantage over later entrants, it is particularly the 1972 incumbents (those firms present at the beginning of our sample that survive into the 1990s) that have the largest market share advantage. While entrants face a disadvantage, they can ameliorate such effect if they enter by merger as opposed to opening a branch or through a de novo charter. Similarly, geographically diversified entrants face a lower disadvantage relative to more locally-limited entrants. Moreover, we find that early entrants appear to have such hold in the market by strategically investing in larger branch networks. In addition, we are able to distinguish the early-mover advantage from alternative stories. In particular, a learning model where firms face a self-reinforcing productivity shock every period and discover their type over time, implies an equilibrium where the size distribution of firms is increasing in the age of firm cohorts. Similarly, under a scenario of imperfect capital markets, such that most of the firm’s ability to invest and therefore grow are derived from internally generated funds, the firm’s assets will be correlated with the firm’s age. In order to rule out these alternative stories, we explore whether earlier entrants have larger shares of the market compared to later entrants, after surviving in the market for the same number of years. That is, we explicitly account for the order in which entry has occurred by holding the number of years since entry constant across all banks, and therefore address the potential survivorship bias in our sample. Even when we control for the number of years in the market, we find that a bank’s share decreases by 0.1 percentage points for a change in its order of entry from nth to (n+1)th. Moreover, each additional year in the market increases a bank’s deposit market share by 0.01 percentage points, regardless of the order in which the bank entered the market. Thus, this test is particularly powerful as it allows us to separate the effects of market tenure from those of order of entry. The test indicates that while market tenure increases a bank’s market share, the later a bank enters a market, the lower its market share relative to early entrants. We also offer another test for whether our earlier results are related to a strategic advantage of early entry. If access to capital markets and/or firm learning are the main factors determining a 3 bank’s entry and growth in a market, we should find that multi-market banks that are incumbents in some markets but entrants in others do not depict large differences in market shares across these markets. If the bank enters a market and wants to grow to the extent of the incumbent, it should be able to do so. However, this is not what we find. On the contrary: larger, multi-market banks do not achieve in new markets the large market shares that they have in markets where they are incumbents. That is, even these bank entrants do worse relative to incumbents. A bank with “deep pockets” can enter a market and build a large branch network to drive out smaller banks if it wants to — that is, if it believes it would be profitable to do so as it would be able to attract other banks’ customers. The evidence in this paper shows that there is a fraction of consumers that will stay with the incumbent regardless, and that is why even a large bank with access to capital and accumulated knowledge might not become as large in markets where it enters later, and branch out in these as much. In fact, compared to the overall sample, the results indicate that the difference between these banks’ market shares when they are entrants and those when they reach maturity in the market are larger. This suggests that larger, multi-market entrants grow faster than smaller, single-market entrants. Moreover, we find that high growth markets show a smaller difference between late and early movers. These results support the prediction from consumer switching costs models that higher population growth markets should exhibit less of an early-mover advantage. Indeed, they suggest that consumer switching costs are likely an important factor behind the documented market share difference. In low growth markets, the number of new consumers is low, and therefore entrants should face greater barriers to their market growth. There are at least two important policy implications of the paper’s main finding that entrants cannot compete on an equal footing with incumbents. The first is that a large entrant might not signify as much of a threat to an established community bank. This may help explain why so many thousands of community banks survive when many had predicted their demise — the large banks merging around the nation cannot enter their markets and take their local market shares away from them. Second, for antitrust analysis, potential entry might not be as effective as a competitive force in a market, especially one with low population turnover. Thus, regulators, who define potential entry as a “mitigating factor” for the possible anticompetitive effects of a merger, should adjust for the fact that entrants may not be able to fully compete head-to-head with incumbents. We discuss both of these implications in the paper. The paper is organized as follows. Section 2 briefly reviews the literature on the advantage of 4 early movers while Section 3 provides a discussion of how such an advantage might arise in banking as well as the prior empirical evidence on this issue. Section 4 describes the empirical framework. Section 5 presents our empirical results and Section 6 concludes. 2 Early-mover advantage: Theory and evidence An early-mover advantage might arise under various demand and supply conditions.2 On the demand side, factors such as switching costs, network externalities, and buyer inertia due to quality uncertainty and/or habit formation can result in an early mover having an advantage over later entrants. The supply-related factors include set up and sunk costs, scale economies, supply chain, and learning by doing. All of these factors have been studied under various setups.3 However, the sequential incumbententrant games that explicitly model the first-mover advantage are few in the literature. One insightful paper in this respect is that by Schmalensee (1982) who studies the long-lived advantages to pioneering brands when buyers face imperfect information about product quality. In a two-period model, two brands enter sequentially, and while being identical, the second brand has a disadvantage because customers have already tried the first brand in the first period and found that it is satisfactory. Within the economics literature, empirical work on the advantage of early movers, which has mostly been interpreted as a market share advantage, has focused on the pharmaceutical industry [Grabowski and Vernon, 1992; Hurwitz and Caves, 1988; Gorecki, 1986], with some other applications such as those to financial innovation [Tufano, 1989] and internet search engines [Gandal, 2001]. Grabowski and Vernon (1992) analyze eighteen drugs experiencing generic competition upon patent expiration by exploring price and market share patterns of incumbents and entrants, generally finding, as the earlier literature, that pioneers maintain premium price positions after entry while their market shares erode over time. Tufano (1989) finds a strong relationship between product innovation and investment banks’ market shares using a database of 58 financial innovations. Gandal (2001), based on a period of one year, uses a discrete choice logit model to find that early entrants in the internet search engine market enjoy a consumer utility premium. There is a significant body of evidence from marketing research on pioneer and early-entrant Mueller (1997) provides a study of several elements that allow a first-mover advantage to exist. For example, Klemperer (1987) analyzes switching costs in a two-period duopoly model where two firms are always in the market. 3 2 5 advantages for non-financial firms. This work, however, is based on surveys or data that are restrictive and present problems related to the levels of industry aggregation and distinguishing the order of entry, among others.4 Examples are Robinson and Fornell (1985) on consumer goods and Robinson (1988) for industrial goods. 3 The early-mover advantage and the banking industry In the banking industry, demand-driven factors are likely to be important in giving rise to an earlymover advantage.5 Both the anecdotal and empirical evidence suggest that consumer switching costs are significant. The evidence on supply-side factors such as economies of scale is mixed, showing both economies and diseconomies of scale in bank products. Switching costs appear to be prevalent in the use of banking services. Anecdotal evidence suggests that depositors find it costly to close an account with their current bank to open an account in another bank.6 Time is invested in doing so, funds may be tied in the process, and the new service might require some specific investment in learning to use it. Other switching costs might include uncertainty over the quality of service, such as branch service quality, product availability, and even how long it takes to get through the phone system to a customer representative.7 Customer inertia is likely to be such that in order for a consumer to switch banks, at least one of the following should occur: current service deteriorates relative to expected new service at another bank and this deterioration is enough to cover switching costs; large discount offered by another bank; some other large expected gain from switching (e.g. new technology that allows new products that the current bank is not offering yet, such as internet access). The empirical evidence suggests that customers stay with the same bank for a long time largely due to the cost of switching banks. Based on survey data, Kiser (2002a) finds that the average household stays with the same bank for ten years, while the most frequently cited motivation for 4 The data are usually from the Profit Impact of Market Strategy. See Kerin et al. (1992), Szymanski et al. (1995), Kalyaran et al. (1995), and VanderWerf and Mahon (1997) for a survey and discussion of the problems of this literature. 5 Note that, in principle, we are agnostic about whether there is an early-mover advantage or disadvantage. The latter might arise from conditions such as being a large banking organization in the form described by Williamson (1967); managers pursuing the quiet life by choosing to grow through in-market mergers, a relatively simple strategy [Mueller, 1997]; as well as other diseconomies. We focus on the former since this is suggested by the anecdotal evidence and our results. 6 A search on American Banker evidences how the notion of switching costs is very much at the forefront of banking strategies. 7 Zephirin (1994) builds a model for the deposit market where consumers care about both price and service quality but the latter is uncertain, giving rise to switching costs that grow proportionally to the length of the bank-customer relationship. 6 changing banks is a household relocation. Moreover, she finds that a third of households report that a main reason why they have stayed with their bank is the inconvenience of changing institutions. Using the same data, Kiser (2002b) finds that owning a home, interpreted as geographic stability, is the leading cause for consumers staying with the same bank for a long period. Kiser also finds that switching costs, as reported by the consumer, appear to be most severe for high income, high education, very low income and ethnic-minority households, suggesting that banking markets where these demographic characteristics are prevalent may present greater costs of entry for new banks. Sharpe (1997), using a model based on Klemperer (1987), tests the prediction that under consumer switching costs, markets with higher population turnover and/or growth should be more competitive, and finds that migration has a positive effect on deposit interest rates. Calem and Mester (1995), in a careful analysis of consumer survey data, find that much of the credit card interest rates stickiness is related to consumers facing switching costs across credit card issuers and the resulting exercise of market power by the firms. Stango (2002) also finds a strong relationship between various measures of switching costs and credit card interest rate pricing by commercial banks. Finally, Kim, Kliger and Vale (2003) estimate a firm model where consumers face switching costs, and, applying it to bank loan data, find that switching costs pay a significant role in the lending market. In particular, they find that switching costs represent about a third of the interest paid on average by loan customers, and are responsible for about a third of the market share of the average bank. Under switching costs, firms enjoy market power over their repeat-purchasers. As a result, a firm’s current market share is an important determinant of its future profits [Klemperer, 1995]. Each period, a firm faces a trade-off between decreasing its price to increase market share through new customers, and exploiting its current customer base through higher prices at the cost of a reduced market share. What the firm chooses to do depends on supply factors such as the threat of new entry as well as demand factors such as market growth in population and income. If buyers face imperfect information and/or firm entry occurs sequentially, building a clientele becomes a strategic decision of the firm. Indeed, building a customer base appears to be a major objective of banks. Dick (Forthcoming) finds that banks with the largest market shares are also the ones with the largest branch networks and the heaviest advertisers. Linking profitability and market concentration to advertising intensity, Örs (2003) finds that advertising plays a role in banking competition as advertising appears to increase profitability, especially in less concentrated markets. Moreover, building a clientele can increase the barriers to entry through its relationship to firm 7 access to capital. A bank that acquires a customer base in a given market has access not only to the customers’ needs for banking services, but importantly, to the customer’s funds, which reinforces an incumbent’s clientele as a barrier to entry to outside firms. Furthermore, the importance of relationships, as an entire body of literature suggests,8 makes an early-mover advantage more likely to occur in banking than in other industries, by giving rise to both customer switching costs and asymmetric information. Over time, incumbent banks accumulate private information about the loan customers, depositors, and the local community, and as a result later entrants can be at a disadvantage due to adverse selection problems. In this last respect, Dell’Ariccia et al. (1999) and Marquez (2002) provide banking models where access to information might create difficulties for potential entrants, as banks obtain much information about the creditworthiness of their borrowers after lending to them. On the supply side, scale economies, a potential factor giving rise to an early-mover advantage, appear to be of importance for some bank types and products, but not for others, though the evidence is somewhat mixed. On the one hand, based on extensive empirical work, economies of scale appear to exist for small-sized banks, while larger banks experience scale diseconomies [Berger and Mester, 1997]. Moreover, while some bank products might enjoy increasing returns to scale, such as electronic payments systems, management of a bank might run into decreasing returns as firms grow into larger institutions. This is also suggested in the work of Williamson (1967), as large firms might become more bureaucratic and have difficulties processing their large amounts of information flows.9 On the other hand, Hughes, Lang, Mester and Moon (Forthcoming) argue that standard techniques to estimate scale economies do not appropriately account for the interplay between bank capital, risk and managerial preferences in the production function of the bank, and find larger scale economies following this adjustment. 4 4.1 Empirical framework Analytical framework Entry of new firms is a phenomenon widely observed in banking markets. Banks differ greatly in terms of when and how they enter the market and how long they remain in it. An early-mover advantage can arise under the existence of certain conditions that create a barrier to entry, which There is an extensive literature that documents the focus of smaller banks on lending to small firms or so-called “relationship lending.” See Berger et al. (2005) and Cole et al. (2004). 9 At certain levels of size, in fact, this could be the source of a first mover disadvantage [Mueller, 1997]. 8 8 affect the ability of incumbents to prevent the erosion of rent from entry. These elements, which can therefore result in an early mover having an advantage over later entrants, range from switching costs, on the demand side, to setup costs and scale economies, on the supply side. However different, all of these factors have a common nature, as they fall within the general category of fixed-cost investment decisions of the firm. In the case of the banking firm, the process of building a clientele can be seen as a form of capital investment, and as such, a barrier to entry for later movers. To earn these captive consumers the firm must build branches and invest in advertising and other promotional activities. The costs involved in building this market share are, at least in part, fixed and sunk. They are fixed in the sense that the cost of building the branch/advertising does not vary with the number of consumers served/reached by the branch or ad; and they are sunk in that all or a large part of branch building costs/advertising cannot be recouped. These consumers, by becoming a bank’s clients, now face costs of switching to another bank. Thus, when a new bank enters, it does not compete on a level playing field with the incumbent. Such barrier to entry, even when it does not prevent entry, makes it more difficult, and could allow incumbent firms to enjoy some rents. This is reinforced by how the bank’s client base affects the bank’s access to local funds. Whenever strategic interaction is part of the competition among firms, in that a firm’s behavior can affect another firm’s behavior, one would expect incumbents who face an entry threat to take action to either deter entry or accommodate it, whichever is more profitable. In order to capture the advantage of an incumbent, we use as an illustration a variation of the von Stackelberg two-stage, two firm model based on the Spence and Dixit reinterpretations [Spence, 1977; Dixit, 1979]. In particular, the Stackelberg quantity can be redefined as capacity, which is interpreted here, without loss of generality, as the bank’s clientele, while the profit functions are rationalized as reduced-form functions resulting from a price game under capacity constraints. In the second stage, we have a price game under capacity constraints, where both firms basically dump their capacities in the market, in a manner analogous to Cournot — though, instead of an auctioneer, it is the firms that quote the market price at which demand equals aggregate capacity, as in Kreps and Scheinkman (1983). In the first stage, firms choose capacities sequentially. The incumbent or early-mover decides how much capital to invest by taking into account how the entrant will react to its choice of capital. From the Stackelberg model we know that the incumbent’s capital investment and subsequent profits are higher than those of the entrant’s. This captures the incumbent’s earlymover advantage. 9 While this model is quite simple, it actually embodies some of the main aspects of sequential entry under sunk costs. The focus of the model on market share benefits, which is what is usually referred to as the first-mover advantage, is appropriate in light of the theoretical developments related to entry barriers, as well as the empirical evidence on the link between market share and profitability. Under barriers to entry, market share is usually linked to performance in a direct way. For instance, under switching costs, the firm has a degree of monopoly power over its captive consumers and therefore current market share is an important determinant of future profits [Klemperer, 1995]. Schmalensee (1989), in a Handbook of Industrial Organization article, lists the association between market share and firm profitability to be among the main systematic relationships documented in the inter-industry studies. Mueller (1986), in a lengthy study of longrun profitability, finds that profit differences across manufacturing companies are strongly related to market shares. In the banking literature, the link between profits and market share has not received a lot of attention, but when it has been studied, the relationship came in positive. In particular, Berger (1995) provides support for the relative market power hypothesis that larger shares are associated with greater exercise of market power and higher profits (after controlling for concentration and efficiency). At best, however, we recognize that market share, which is the focus of our analysis, is an imperfect proxy for profits. Intuitively, the model captures the chance of an early bank entrant to obtain a significant portion of the population of a market. These captive consumers open an account with the incumbent, learn whatever specificities there are related to the bank services, and become accustomed to their bank. When another firm enters the market, it has to build a clientele from the noncaptive consumer base. More realistically, when entry occurs the incumbent will usually face two types of consumers: those already captive, and the rest of the consumers in the market, which are up-for-grabs and over whom it is fighting with the entrant. If able to price discriminate, the incumbent would like to charge a high price to its captive consumers, over whom it exerts a large degree of market power10 , and a lower price to the other consumers in order to attract their business. Under no price-discrimination, the firm will have to choose an intermediate price, which will be increasing in the size of the captive clientele. If the firm invests enough in building this clientele, or “overinvests” in the terminology of Fudenberg and Tirole (1984), the price will be high enough to make entry In markets with significant population flows, the incumbent’s advantage should be less [see Beggs and Klemperer, 1992]. 10 10 profitable and to “soften” price competition. In particular, the incumbent would be a “fat cat” that accommodates entry, according to the Fudenberg and Tirole taxonomy, such that the investment in clientele makes the incumbent soft under strategic complements (price competition in the market stage), where an increase in incumbent’s capacity investment leads to an increase in the entrant’s profits. The incumbent’s investment, while reducing the size of the market for the entrant, also increases the price that both of them can charge. 4.2 Empirical specification In the Stackelberg model, the early-mover advantage is reflected in the firms’ market shares, which are central to the firms’ future profits. This is what drives our empirical specification, which is designed to investigate whether early movers have an advantage over later movers by exploring whether there is any relationship between a firm’s order of entry into a market and the size of the firm’s market share. To explore whether early movers are distinctly different from later movers in terms of market presence, we define a set of time ranges for the order of entry into a market. A bank is usually believed to have reached maturity after around twenty years in a market.11 Letting T be the year in which a bank enters a given market, a firm at time t is categorized based on its order of entry into the market, that is, according to whether (1 = yes, 0 = no): (i) the bank entered the market within the last four years, such that 0 ≤ T < 5 years; (ii) the bank entered the market between 5 to 9 years ago, such that 5 ≤ T < 10 years; (iii) the bank entered the market between 10 to 19 years ago, such that 10 ≤ T < 20 years. This leaves out those firms that entered the market 20 years ago or earlier (T ≥ 20). We refer to this latter group as the early entrants or “incumbents,” while referring to the rest of banks as entrants. Note that in the specification, we distinguish between “1972 incumbents,” which are banks that were already in the market in 1972 (the beginning of our sample), and the incumbents that came in later. The latter represent the base case for comparison. While we have data since 1972, the regression analysis focuses on the period 1992-2002, in order to have banks fall into each category. Based on these indicators, we define market share of bank i in market m at time t to be a function of the order of entry as follows: For instance, several papers have found that small banks initially have very high ratios of business loans to assets, and this does not fully decline to equilibrium until about age 20 [e.g., Goldberg and White, 1998; DeYoung, Goldberg, and White, 1999]. 11 11 M arket Sharei,m,t = X j β j ∗ Order of Entryi,m,t + α ∗ IN 1972 + γ m + η i + τ t + ν i,m,t j (1) j where j stands for the three order of entry periods, with the variable Order of Entryi,m,t taking on the value of one if the condition applies, and zero otherwise; IN1972 takes on the value of 1 if the bank was already in the market in1972; γ m is a market fixed effect; η i is a bank specific effect; τ t is a year effect and ν i,m,t is a random disturbance. The fixed components for market, bank and year are included in order to control for those factors that might affect market shares other than the order of entry and that could be correlated with the latter. We also investigate whether the entry method into the market has an effect on market share given the order of entry, such that our specification becomes: M arket Sharei,m,t = X j β j ∗ Order of Entryi,m,t j j k (2) +α ∗ IN 1972 + γ m + η i + τ t + ν i,m,t XX j β jk ∗ Order of Entryi,m,t ∗ Entry M ethodk + i,m where k stands for the three methods of entry, as discussed later: (i) merger; (ii) de novo; and (iii) opening a branch. This is the structure that we will used to explore the effects on market share from another important firm characteristic, namely its geographic diversification. In particular, we divide banks into two groups: one with presence in more than 10 metropolitan markets, and another with presence in 10 or fewer metropolitan markets. correlated with bank size as well as brand recognition. This bank characteristic is usually 4.3 Data The data used in the analysis are taken from the Federal Deposit Insurance Corporation Summary of Deposits.12 Given our sample for the period 1972-2002, we use branch deposit data since 1972 to determine the deposit market share, as well as the entry and exit date, if any, for each firm into each market. To obtain a bank’s deposit market share we add deposits in all of the bank’s branches in a given market. Based on this definition, our regression analysis covers the period 12 We use data from the Report on Condition and Income from the Federal Reserve Board for some of the descriptive statistics, and data from the National Information Center to determine method of entry. 12 1992-2002. Note that while one might also be interested in analyzing other bank output, such as loans, data at the branch level are only available for deposits. Similarly, it would also be interesting to explore an early-mover advantage in terms of profits, prices and risk. However, these data are only available for the bank as a whole, and not for each bank-market observation. At best, market share, which is the focus of our analysis, can be interpreted as a proxy for profits.13 Also, our deposit data aggregate all types of deposit accounts. While most of these are local in nature and involve the consumer switching costs discussed earlier, some (such as negotiable CDs) have a national market and are likely to involve lower switching costs. However, the evidence suggests that consumers tend to buy deposit products locally and to cluster them with a single institution [Amel and Starr-McCluer 2002]. While determining the entry date is sometimes controversial for other industries, the banking data allows us to identify entrants in a straightforward fashion. the method of entry chosen by the firm. We are also able to determine In particular, we can distinguish among the following three types of entry into a market: (i) merger with a target-market bank (absorbing the charter); (ii) de novo (new banks); (iii) opening a branch (outside market bank).14 Note that a banking holding company can enter a market by buying a target-market bank through an acquisition. This changes the bank ownership but does not alter the bank charter, such that from the perspective of the consumer and the regulatory agency the banking institution is still the same. Thus, we do not consider these acquisitions as entry.15 While deposit data are available for each bank branch in the US, we define the relevant geographic banking market at the level of the metropolitan statistical area (MSA), which represents a geographic unit with a large population nucleus with its adjacent communities (a city or town).16 In the results section we briefly explore the link between order of entry and profit rates for single-market banks, for whom profit data at the market level are available. 14 This last category includes the case of a bank entering a market by purchasing branches of another bank. 15 Note that when a bank is acquired without a change in its bank charter, it becomes affiliated to a Multi-Bank Holding Company (MBHC). MBHCs might serve as internal capital markets for their member institutions [Houston, James and Marcus, 1997; Houston and James, 1998; Campello, 2002]. Thus, these acquisitions that provide MBHC membership without changing the bank charter might alter the target’s behavior if they provide the target access to an internal capital market. To deal with this issue, as a robustness exercise we control in our regressions for whether the bank is subjected to an acquisition that preserves its charter. We find that our results are robust to the inclusion of such control, as will be discussed later. We focus on metropolitan markets rather than rural markets primarily because these are the markets that are generally considered to be more competitive with easier entry/exit and acquisition of local market shares. Finding an early-mover advantage in these markets would therefore be more surprising to us. Since this is the first analysis done of this kind, we also think it is better to focus on metropolitan markets in order to learn about whether there exists an early-mover advantage in banking, given that these banks have over 80% of industry deposits. However, the study of rural markets is an interesting direction for future research, given that the interaction among banks in rural markets is likely to be systematically different. 16 13 13 This definition is supported by surveys of consumers and businesses as well as the bulk of the empirical banking literature.17 There are 318 MSA markets in our data, which include almost all metropolitan markets in the U.S. (over 96 percent, based on 1999 Census definition). The appendix contains the summary statistics for the variables used in the analysis. Based on our regression sample of surviving banks for the period 1992-2002, there are 8496 bank entries into local markets since 1972.18 Table 1 presents the number of banks by year and number of years in the market (under “Time”), based on bank-market observations across time.19 Date of entry into a market can be established up to 1972, given that our sample starts that year. We use these data to categorize firms by order of entry. The numbers in italics represent banks that entered in 1972 or earlier and have survived since then (and for whom date of entry into the market cannot be determined), while the rest of the numbers of the table are bank entries. The first line on the table, where Time=0, corresponds to the new entrants each year of our analysis. Thus, there was an average of 489 entries into local markets per year, with a larger number of entries towards the second part of the period (275 entries in 1992 vs. 517 in 2002). The total number of entries throughout 1992-2002 is 5,381, as indicated in the last column. The second line, where Time=1, corresponds to the surviving banks that entered the previous year —one-year old banks—, and so forth for the following lines of the table. For instance, under the 2002 column, the 105 banks in the row where Time=10 correspond to those firms that entered ten years ago, or, in other words, have been in the market for ten years. Therefore, of the 275 banks that entered in 1992, only 105 of them remain in 2002. Table 2 shows the distribution of banks across markets in terms of their order of entry in each year of the analysis. While in the earlier part of the sample almost half of the firms in the market are old (with market presence of over 20 years), towards the second half most market banks are new entrants. Table 3 depicts the average deposit market shares by time in the market. One fact that comes out of the table is the stark difference in market shares between entrants and banks that have been in the market for twenty or more years. While entrants average a market share of 3.3 percent, 17 For a detailed discussion on the relevant geographic market definition in banking, see Dick (2002) and the references therein. 18 Note that our sample includes 10,655 entries over 1972-2002. Of those, 5,381 entries occur during 1992-2002, while the other 5,274 entries occur throughout 1972-1991. Out of these 5,274 entries, however, only 3,115 entrants survive into the 1990s. That is how we come up with the number of 8,496 entries since 1972 for the sample 1992-2002. 19 Note that our original data set included some bank observations that revealed multiple entries into the same market throughout our sample period. This happens in the uncommon situation where a bank enters a market, stays for a few years, then leaves the market, and returns again after a few years. These observations constituted only around one percent of the observations, and we have dropped them from our sample. 14 incumbents have more than double this market share, with an average of 7.1. Entry methods are distributed more or less equally, though the most popular method of entry is de novo, with 37 percent of the total entries in our sample, followed by 33 percent of the entries through branch opening and the rest using a merger. Table 4 provides information on the entrants’ average market shares based on method of entry used: not surprisingly, banks that enter through mergers have on average the largest market shares, relative to entering via de novo or opening a branch. Another feature from the table is that market shares usually rise on average as a bank accumulates years in a market (note that the figures for the oldest entrants via merger is based on a small number of observations). 5 Results The regressions include market and year effects This section introduces the estimation results. (318 and 11, respectively). We also introduce bank effects (over 7000) in most regressions. Table 5 shows the results from estimating the deposit market share of a bank in a market (in a given year) as a function of the order of entry of the bank into that market, as described in expression 1. Column (i) is our base regression, while column (ii) introduces bank fixed effects and column (iii) adds further time-varying bank and market controls. Based on the results from all the specifications, the coefficients on all three of the defined order of entry periods are negative and statistically significant at the one percent l.o.c., indicating that entrants have, on average, a lower market share than incumbents.20 Moreover, this market share disadvantage increases the more recent is the entrant. That is, later entrants do worse than earlier entrants, so that a bank that has only been in the market for four years or less has, on average, the lowest market share. Also, note that 1972 incumbents have the highest market shares in the market. Indeed, the biggest difference is between these incumbents and all other banks. Lastly, note that our dependent variable is bounded between zero and one. However, when we estimate the model with a logit transformation of market shares to allow the dependent variable to take on any real value, we find that our results are robust. We keep the bounded dependent variable, however, for ease of interpretation of the regression coefficients.21 Note that the results are robust to using a continuous order of entry measure (with market share increasing a quarter of a percentage point for every year in the market), as well as various other order of entry periods. 21 Clearly, analyzing the effect of time of entry on profits would be interesting, but as mentioned earlier, there are data limitations to doing so. When we explore the link for single-market banks, however, for whom profit data are available at the market level, we find that our results are similar to those obtained using market share, at least for the newest entrants. Based on various specifications (controlling for bank, year and market effects), in particular, we find that the newest entrants have over 1 percentage point lower profit rates (the mean/median of the distribution is 20 15 While the results are similar across the specifications on the column, we focus on the results in column (ii), which incorporate individual specific effects to control for differences in market share across banks other than those reflected in our explanatory variables. We do this since taking into account bank variation appears to be important. In particular, when we break up the total sample variation into bank, market and year effects, we find that most of the variation in market share comes from variation between banks (55 percent), followed by variation between markets (17 percent) and variation between years (less than 1 percent). percent to 69 percent when bank fixed effects are included. Thus, based on the results from column (ii), the most recent entrants have on average market shares that are 4.1 percentage points lower than those of our base case, after-1972 incumbents; entrants with 5 to 9 years in the market have lower shares by 3.0 percentage points; and entrants with 10-19 years of market presence have market shares that are 1.4 percentage points lower. Note that 1972 incumbents have, on average, 10.5 percentage points higher market share than incumbents that entered after 1972. This implies, for instance, that relative to 1972 incumbents, the most The R-squared increases from 23 recent entrants have almost 15 percentage points fewer in deposit market share. With a regression fit close to 70 percent, order of entry as well as bank, market and year effects appear to explain most of the variation in firms’ market shares. Column (iii) incorporates to the specification in (ii) time-varying bank and market controls. While bank fixed effects take into account unobserved individual characteristics, these are assumed to remain constant over time. However, we can think of certain bank characteristics that are likely to change over time and that can have a potential effect on market share and on the decision of a bank to enter a market. In particular, we introduce a proxy for a bank’s cost and profit X-efficiency, constructed as in Berger and Mester (2003), as these could be important controls for bank quality.22 Berger and Mester (2003) find that these vary considerably during the 1990s, 11 percent, based on net income over equity, and after removing outliers). This result is significant at the 1 percent l.o.c. The other two coefficients of interest (entrants from 5-9 years ago and entrants from 10-19 years ago) are also negative and decrease in magnitude, but when we include enough controls (bank fixed effects, for instance), they become insignificant (results not reported). We also find similar results using loan shares. These results do provide evidence of a link between profits and market shares. However, given that they are based on single-market banks only, it is difficult to make inferences more generally. 22 The main difference between standard financial ratios of cost and profit rates and the cost and profit X-efficiency ranks we use is that the latter remove some of the differences in conditions facing the individual banks. In particular, these ranks are based on the residuals from regressions of costs and profits, respectively, on bank outputs (asset categories, off-balance sheet activities), fixed inputs (capital, premises), and market prices for variable inputs (labor, deposits, purchased funds). The residuals are put in rank order for each year and converted to a uniform scale over the [0,1] interval, where 0 is the least efficient (highest cost or lowest profit) and 1 is the most efficient (lowest cost, highest profit), to make the ranks comparable across years. Thus, a bank’s rank in a year is the proportion of sample banks with lower efficiency (e.g., rank = 0.70 for a bank that is more efficient than 70% of the sample). 16 while Hasan and DeYoung (1998) find that bank profit efficiency varies during the early years of de novo banks’ lives. The results are robust to these additional controls. We also introduce the ratio of small business loans to total loans as a way to control for the bank’s business focus (data available for 1993 onwards). Since market share and these controls are likely to be jointly endogenous, we lag these control variables and only use this specification for sensitivity analysis. Lastly, we incorporate personal income growth to control for a market characteristic that is likely to change over time. After we introduce these controls, we find that the results are similar to those presented in column (ii). Note that the specification is now based on the period 1994-2002 due to the limitations on small business loan data, which implies a loss of close to 18,000 observations — yet the results are very similar to those in column (ii). Note that the only control that appears to matter is the cost efficiency rank, which indicates that a relatively more efficient bank has a higher deposit market share, which is consistent with the Efficient Structure Hypothesis.23 Does the method of entry affect the early-mover advantage? We now explore whether the method of entry used by the entrant has an effect on the latermover disadvantage we found earlier. The order of entry is interacted with two types of entry: by merger and by branch opening, leaving as base case the de novo entry method. As can be seen in column (i) of Table 6, entering a market through a merger gives the bank a higher market share, since the bank is buying up another bank’s existing branch network in the market. Entering by opening a branch, however, provides a lower share relative to de novo banks. Note that while a specification with bank fixed effects would be desirable given the large variation across banks, it would not be appropriate in this exercise. This is due to the selection of banks into a particular entry method and therefore the sources of variation, which changes the interpretation of the within-bank regression coefficients. Banks that enter by merger are large and geographically diversified compared to other banks, and they tend to choose the same method of entry every time they enter a new market. Thus, the bank fixed effect cannot be distinguished from the entry method to provide the interpretation that interests us here. Given the significance of the variation across banks, the reported findings should be viewed with caution.24 In particular, the median asset size of a merger entrant is $1.5 billion compared to $295 million 23 Under the Efficient Structure Hypothesis, the largest players in a market will be those that are most efficient, as efficiency is the mechanism thereby they grow and become large. 24 We also tried distinguishing “toeholds” from larger mergers. Indeed, we find that large mergers do increase the market share of the entrant, whereas toehold acquisitions do not have such an effect. Again, we caution about drawing conclusions from these findings because we have to drop the bank fixed effects. 17 of entrants via branch and $30 million of de novo entrants. Also, 94 percent of entrants by merger are multi-market banks. De novo entrants, however, are virtually single-market banks by definition (with the exception of only 54 entrants who entered more than one market the same year). Moreover, banks tend to stick to the same entry method whenever they enter a new market.25 Thus, the sources of variation to identify the coefficients are different depending on the entry method (for merger entrants, which are mostly multi-market, it is both the across-time and across-market variation, while for de novo banks, it is only across-time). Thus, the bank fixed effect would already control for bank quality and size, and with that implicitly capturing the effect of entry method. Has “geographic diversification” an effect on the early-mover advantage? We now explore whether large, geographically diversified entrants that offer extended branch networks outside the new market have a different experience from smaller, more locally-limited entrants. In particular, we define geographically diversified entrants to be those that have presence in at least ten other metropolitan markets.26 Such measure is correlated with the size of the branch network, the size of the bank in terms of assets, and is also likely to be correlated to how much of a brand name consumers ascribe to this bank — especially as branches themselves are believed to be a form of bank advertising.27 Column (ii) of Table 6 introduces interactions between the order of entry and the indicator variable for geographic diversification, which takes on the value of 1 if the bank operates in at least 10 other metropolitan markets. Indeed, geographically diversified entrants have larger market shares than entrants that are not. Note that here as well, the within regression is not the appropriate specification, since there is virtually no variation within a bank across time in terms of whether it is geographically diversified according to our definition (a bank is always one or the other). As a result, the geographic diversification indicator and the bank fixed effect cannot be separated from each other in a way as to provide us with the interpretation of interest. Are investments in branch networks larger for early movers? The entry method Herfindahl within a bank (the sum of the squared proportions of entry methods within a bank.), for instance, is 0.7 for banks that enter more than one market, suggesting that most banks use the same entry method each time they enter a market. 26 We chose this cutoff based on the distribution of the number of markets served across banks. In the appendix, we provide some statistics for the number of outside markets in which banks in our sample operate. The mean is 5 markets, while the median is zero. Presence in at least ten other markets happens to be around the 90th percentile of the distribution, which is why we chose it for our definition. Trying various cutoffs above 5 markets actually provide similar results, though the sharpest appear to be around the current cutoff of 10 markets. 27 See Dick (Forthcoming) for further evidence on bank advertising. 25 18 We now test whether “within-market” coverage, as measured by the size of the branch network inside the market, plays a role. Indeed, an early entrant’s advantage could be due to strategic investments in the branching network, which, ceteris paribus, might make it harder for later entrants to penetrate the market to the extent of the incumbent. Table 7 shows the results from estimating a bank’s branch density in a given market, measured as branches per square mile, as a function of our three order of entry variables and a series of controls. The results are similar to our earlier ones using deposit market share. In terms of magnitudes and from the bank fixed effects regression, they indicate that the newest entrants have about half of the after-1972 incumbents’ branch network (our base case), while less recent entrants (5 to 9 years) have about two thirds, and the least recent entrants (10 to 19 years) have close to 90 percent of the incumbents’ branch network. Interestingly, the 1972 incumbents, who have the largest shares in the market, also have the largest rates of branch penetration. In particular, they have almost double the size of the network of incumbents that came in after 1972. Also note that this specification explains 77 percent of the variation in branch networks among banks. Is the early-mover advantage stable over time? While the removal of geographic barriers to a bank’s expansion started in some US states as early as the 1970’s, the final phase of deregulation took place during 1994-97 with the passage of the Riegle-Neal Interstate Banking and Branching Efficiency Act in 1994, which allowed for nationwide branching. The regulatory framework in place before these changes could have temporarily benefited incumbents relative to entrants, and as a result, our results could derive from such asymmetry. If this were the case, we would expect that, following deregulation, any advantages originally bestowed on incumbents should vanish over time. To explore this possibility we split the sample into two time periods: 1992-1997 and 1998-2002, before and during deregulation, and after, respectively. Table 8 shows that the early-mover advantage, if anything, increases towards the later period, with sometimes a doubling in the market share disadvantage for entrants.28 An F-test rejects the null hypothesis of coefficient stability over time (one percent l.o.c.). One possible explanation for this result is that in the latter period, during which banking markets are more deregulated and more entry is observed (see 1), new entrants have to deal with more competitors also entering those markets at the same time, and therefore have lower shares. Note that this result, however, does 28 Our results also hold for each year of the analysis separetely, with the market share gap between entrants and incumbents increasing slightly over the years. 19 not rule out the possibility that the earlier regulatory framework could have permanently helped incumbents in developing their customer base more than under a different regime. 5.1 Ruling out alternative stories to the early-mover advantage While our tests up to this point establish that early entrants capture a larger share of the deposits in a market, there remains the question of whether this correlation is really the result of the strategic advantage of early movers. That this empirical regularity exists is certainly a novel and striking finding. are ignored. However, our sample is one of survivors, and as such, it is biased, as failures For instance, one alternative story could be that of a learning model where firms face a self-reinforcing productivity shock every period and discover their type over time, as in Hopenhayn (1992). In such a world, as a higher productivity shock today makes it more likely that the productivity shock will be higher in the future, the model implies a stationary equilibrium with a size distribution of firms that is stochastically increasing in the age of firm cohorts. Similarly, under a scenario of imperfect capital markets, such that most of the firm’s ability to invest and therefore grow are derived from internally generated funds, the firm’s assets will be correlated with the firm’s age. In order to address these issues, we carry the following test. In particular, we explore whether earlier entrants have larger shares of the market compared to later entrants, after surviving in the market for the same number of years. That is, we explicitly account for the order in which entry has occurred by holding the number of years since entry constant across all banks. To do this, we estimate market share as a function of two main variables: (i) “order of entry,” which ranges from 1 to 30, based on whether the bank entered the market first, second, third, and so on, over our 30-year data; and (ii) the “number of years in the market,” which holds the time since entry constant. Thus, we can make use in estimation of almost our entire data, by pooling observations for the period 1973-2002. Note that we drop the year 1972, that is, all the “1972 incumbents,” since for these banks we cannot determine the year of entry (whether it was in 1972 or before that). This makes it a stringent test, since it is based only in the sample of banks that entered after 1972, and not on the 1972 incumbents which appear to be a main force behind our earlier results. Moreover, this sample includes many banks that failed that are no longer in the regression sample used in the previous analysis. Also note that unlike our previous exercise, the group of first entrants, for instance, is made up of banks that have been in the market from 1 to 30 years. Table 9 shows results for this test. The first column in Panel A of the table shows that the 20 higher the order of entry in which the bank entered the market, the lower its market share is relative to banks with a lower order of entry. Based on the coefficient on the“order of entry,” we find that a bank loses 0.1 percentage points for each change in its order of entry into the market from nth to nth + 1 (or for each year that the bank delays entry), which is a 1 percentage point per decade of delayed entry. Moreover, each additional year in the market increases a bank’s deposit market share by 0.01 percentage points, regardless of the order in which the bank entered the market — according to the coefficient on “number of years in the market.” Thus, this test is particularly powerful as it allows us to separate the effects of market tenure from those of order of entry. The test indicates that while market tenure increases a bank’s market share, the later a bank enters a market, the lower its market share is relative to early entrants. The results are similar if we use branch density instead of market share (results not shown). Column (ii) in Panel A is similar, though instead of allowing the order of entry to be continuous, it divides order of entry into six different categories, as follows: (i) first to fifth entrants; (ii) sixth to tenth; (iii) eleventh to fifteenth; (iv) sixteenth to twentieth; (v) twenty-first to twenty-fifth; (vi) twenty-sixth to thirtieth. Thus, if an early-mover advantage exists, we would expect the first group (of first, second, third, fourth and fifth entrants) to have higher market shares than the second group, and this group, in turn, should have higher market shares than the next, and so on. Since we let our base case be the last group (vi) in the regression, we expect all the coefficients to be positive and in decreasing magnitude. Indeed, this is what we find. In particular, the first group of entrants has 3.0 percentage points higher market share than the base group, followed by a 2.8 percentage point advantage for the second group, 2.4 percentage points for the third group, 2.2 percentage points for the fourth group, and 1.0 percentage points for the fifth group. Each additional year in the market, as before, increases market share by 0.1 percentage points. We also offer another test for whether our earlier results are related to a strategic advantage of early entry. If imperfect capital markets or learning models are the main explanation for our earlier results, we should find that multi-market banks that are incumbents in some markets but entrants in others, do not depict large differences in market shares across these markets. Why? Because if the bank wants to enter and grow in a new market, and the alternative stories above apply, the bank will have access to its internal capital markets and firm-level knowledge in order to do so — thus, if it does not grow to be like the incumbent, there has to be another reason besides availability of capital and/or accumulated firm learning. However, we do not find this. On the contrary: larger, multi-market banks do not achieve in new markets the large market shares that 21 they have in markets where they are incumbents. That is, even these bank entrants do worse relative to incumbents. To do this, we reestimate our earlier model on a subsample of banks that are both incumbent in at least one market and entrant in at least one other market. This subsample is made up of larger, more geographically diversified banks than the overall sample. The results are shown in Panel B of Table 9. The specifications are exactly like the ones presented on Table 5, columns (i) and (ii), with the difference that now the estimation is based on a subsample that is one fifth of the original sample. Yet the results are strikingly similar, and if anything, the later-mover disadvantage appears to be actually stronger for these banks. If the above alternative stories were behind our earlier results, we should find no difference between incumbents and these larger entrants, since both are likely to have similar access to funds when they want to grow.29 In fact, compared to the overall sample, the results indicate that the difference between these banks’ market shares when they are entrants and those when they reach maturity in the market are larger. This suggests that these banks grow faster than smaller, less geographically diversified banks. Note that similar results are obtained here if we use branch density instead of market share (results not shown). These results are particularly enlightening because while there could be some market level learning that is relevant to a bank considering entry and growth in a market, its access to capital is likely to play a larger role. For example, a bank such as Bank of America already knows a lot about banking. When this bank is deciding whether to enter new markets, the bank researches which markets offer the most profitable opportunities for entry. Thus, there is little learning left to do about these new markets. However, with its “deep pockets” Bank of America can come in and build a large branch network to drive out smaller banks if it wanted to — that is, if it believes it would be profitable to do so as it would be able to attract other banks’ customers. The evidence in this paper shows that there is a fraction of consumers that will stay with the incumbent regardless, and that is why a bank like Bank of America might not become as large in markets where it enters later, and branch out in these as much. Does the market turnover in population have an effect on the early-mover advantage? 29 As mentioned earlier, when a bank is acquired, it becomes affiliated to a MBHC, which might serve as an internal capital market for its member institutions [Houston, James and Marcus, 1997; Houston and James, 1998; Campello, 2002]. When we control for whether the bank is subjected to an acquisition (Table 9, Panel B, col. (i), without bank fixed effects to allow for the proper identification of the coefficient of interest), we find that our results are unchanged, and that the coefficient on the indicator variable for whether the bank is acquired (which takes on the value of 1 if it is acquired and 0 otherwise) is positive and significant, as we would expect, though contributing only 0.6 percentage points to market share. While our reported results do not separately show this effect, our inclusion of bank fixed effects should already account for this. 22 While there are several factors giving rise to an early-mover advantage, consumer switching costs are likely to be important in banking markets, where repeated transactions and relationship building is central to banking services. The theory on consumer switching costs suggests a relationship between the degree of market power over a firm’s consumers and a market’s growth and turnover in its customer base. In an infinite-period two-firm market with consumer switching costs and customer turnover, Beggs and Klemperer (1992) provide a proposition that states that prices (and profits) are decreasing in the rate of growth in the market, due to the reduction in the proportion of locked-in consumers relative to new market consumers, as well as the greater importance of the future stream of profits relative to current profits. Similarly, in a two-firm, two-period model, Klemperer (1987) finds that the greater the share of unattached consumers in the second period, the more sensitive consumers are to price differences among the firms and the more competitive the market. Intuitively, the incumbent is likely to have a large degree of market power over its current customer base. In the face of entry, the incumbent confronts two types of consumers, those already captive and the rest of the consumers in the market, over which both the incumbent and the entrant fight. When population growth is high, the latter consumer type grows, thus providing the entrant with a more similar playing field. As mentioned earlier, Sharpe (1997) tests these predictions with banking data and finds that greater customer flows into banking markets are associated with higher, more competitive deposit interest rates. In lending markets, the model in Marquez (2002) predicts that entry will be easier in markets with a high degree of borrower turnover, given that incumbents enjoy an information advantage over entrants as they obtain proprietary information through the lending process. Similarly in our current setup, if consumer switching costs are a significant element behind our findings, we should expect the early-mover advantage to be stronger in markets with low population growth and weaker in markets with high population growth. As new, potential consumers move into the market, a smaller proportion of old consumers are locked-in with a given bank, and therefore the advantage of having entered that market earlier should diminish, given that both new and old market banks should be perceived more or less equally by new consumers (controlling for other firm characteristics). To test for this possibility, we construct two groups of markets, with high and low population growth. Taking the average of the annual market growth rates over our sample period (1972- 2002) for each market, a high (low) growth market is defined as the top (bottom) quartile of the distribution of markets. Reestimating our specifications for each group, we find that, indeed, 23 markets with high population growth present lower barriers to entry relative to low growth markets (see Table 10). Particularly interesting is the fact that there is no difference between recent entrants and those that entered after 1972, but rather, the market share difference is between 1972 incumbents and all other market banks — and this difference is large. In particular, we find that while in high growth markets the difference between 1972 incumbents and those incumbents that entered later is of around 5 percentage points, in low growth markets this difference is 18 percentage points. Moreover, this later-mover disadvantage is the same across all groups of banks other than the 1972 incumbents. In high growth markets, conversely, even the difference between the newest entrants and 1972 incumbents is no more than 11 percentage points. An F-test rejects the null hypothesis that the order of entry coefficients are equal across the two samples (one percent l.o.c.). These results support the prediction from consumer switching costs models that higher population growth markets should exhibit less of an early-mover advantage. Indeed, they suggest that consumer switching costs are likely an important factor behind the documented market share difference.30 In low growth markets, the number of new consumers is low, and therefore entrants should face greater barriers to their market growth. 5.2 Discussion Our results show a robust empirical relationship between a bank’s order of entry and its predominance in the market. If banks were all identical, we would unambiguously expect later entrants to have lower market shares, on average, than those of earlier entrants. However, banks are not identical. It is not uncommon in banking markets to find some of the largest US banks with branch networks across the country coexisting with small, single branch local banks. Moreover, there is large variation in terms of order of entry within each bank category, with national brand-name banks entering markets along side with small de novo firms every year. When we introduce the fact that banking firms are differentiated, we get a more nuanced story of bank entry. Measuring bank differentiation in terms of geographic diversification, for instance, we learn that even entrants can attain a significant market presence if they offer large branch networks, thus reducing the disadvantage of entering later. Moreover, our results suggest a similar situation for incumbents, where their market position This test is particularly useful to disentangle the factors behind the early-mover advantage that we find. For instance, early-movers could simply be higher quality providers, that when entered the market captured the prime location in town, for instance. The fact that population turnover appears to significantly matter for our results is indication that consumer switching costs are relatively important. 30 24 is also related with what they have to offer to their clients. In particular, incumbents are not guaranteed market dominance by the mere fact of having entered early. While that certainly is part of the story behind the distribution of firms’ market shares in banking markets, incumbents that achieve and maintain a large market presence appear to do so by offering a higher quality service as suggested by the different results for branded and unbranded incumbents. As shown on Table 11, early-movers also tend to have a greater branch density relative to entrants (44 percent higher), and are almost twice the age of newer entrants — where age could be interpreted as a proxy for bank experience. In terms of geographic diversification, an interesting pattern arises over the period. While in 1992 incumbents were more geographically diversified than entrants, towards the latter part of our sample, many new entrants tend to be larger and more geographically diversified than incumbents, with new entrants having on average presence on 4 states, versus incumbents with presence on 2 states (though the medians of the distribution are the same). This result is driven by some national banks in the tail of the distribution, that have operations in dozens of states and continue to move into new markets. Following nationwide branching deregulation, banks had the opportunity to expand to new markets across the country, and in doing so becoming higher quality banks offering the convenience of large branch and ATM networks. Many market incumbents have taken advantage of this opportunity and some have by now become regional or even national banks, sometimes joining the group of geographically diversified banks that are buying up market leaders and incumbents in other markets. Other incumbents, however, have chosen to stay put solely with operations in the home market, and have sometimes been recent targets of mergers or face increasing competition from new banks. However, their early-mover advantage aids them in retaining some of their market leadership. The market dominance of early-movers can be appreciated on the last column on Table 11. When we rank banks in each market by deposit shares, we find that about three quarters of the banks with the largest share in the market are banks that entered over 20 years ago, reflecting once again the strong relationship between market leadership and order of entry. Incumbents’s Following market shares are also ranked much higher, on average, than those of later entrants. deregulation, however, banks have had the opportunity to enter new markets, and the evidence indicates that large, geographically-diversified banks that enter new markets through a merger can also attain market leadership relatively quickly. suffer as much from the early-mover advantage. 25 As a result, this kind of later entrants do not For instance, while in 1992, 87 percent of top market banks are incumbents, in 2002 their share in market leadership decreases to 57 percent. An interesting question that arises from this is how consumers are affected by these changes. On the one hand, the anecdotal evidence suggests that consumers tend to be disrupted whenever their current bank is bought by a new one, and moreover, it is not clear what the competitive effects from mergers are. On the other hand, many of the buyers in the large mergers are big national brands and geographically diversified banks that might provide a higher quality service to consumers, at least over time. Table 12 shows the preferred method of entry by order of entry and bank size. Traditionally, large banks have preferred mergers while small banks de novo (results on the table are stable over the period). The specific case of Bank of America, for example, which is one of the largest retail banks, is useful. Bank of America entered over a hundred markets in the 1990s, mostly by merger, such that it was usually able to obtain a big stake in the market right after entering. However, compared to markets where it is an incumbent (usually with presence since 1972 or before), the deposit market shares of Bank of America, as well as its branch networks, are lower in market where it is an entrant — especially a recent one. 6 Concluding remarks The entry literature provides an abundance of work, some of which has emphasized the advantages of entering early into a market. Empirical research documenting the effects of early entry, however, is not as abundant, and has focused on differentiated goods where innovation is central to the industry. In contrast, this paper explores the existence of an early-mover advantage in the service industry of banking. Using a unique data set for 1972-2002 on all banks in all urban markets in the US, this paper analyzes whether there are differences in deposit market shares among banking firms in a given market based on how early they entered. The results indicate that later entrants have lower market shares, on average, controlling for bank, market and year effects. While identifying the factors driving the early mover advantage documented here is outside the scope of this paper, our measure of the market share advantage of early entry may be interpreted as an indirect measure of consumer switching costs and thus of barriers to entry in the industry, which can be useful for assessing the effect of competition policy. 26 References [1] Amel, D.F. and M. Starr-McCluer (2002). “Market Definition in Banking: Recent Evidence,” Antitrust Bulletin, 47: 63-89. [2] Beggs, A. and P. Klemperer (1992). “Multi-Period Competition with Switching Costs,” Econometrica, 60: 651-666. [3] Berger, A.N. (1995). “The Profit-Structure Relationship in Banking — Tests of Market-Power and Efficient-Structure Hypotheses,” Journal of Money, Credit, and Banking, 27: 404-31. [4] Berger, A.N and L.J. Mester (1997). “Inside the Black Box: What Explains Differences in the Efficiencies of Financial Institutions,” Journal of Banking and Finance, 21: 895-947. [5] Berger, A.N. and L.J. Mester (2003). “Explaining the Dramatic Changes in Performance of US Banks: Technological Change, Deregulation, and Dynamic Changes in Competition,” Journal of Financial Intermediation 12: 57-95. [6] Berger, A.N., N.H. Miller, M.A. Petersen, R.G. Rajan, and J.C. Stein (2005).“Does function follow organizational form? Evidence from the lending practices of large and small banks,” Journal of Financial Economics, 76: 237-269. [7] Berger, A.N., A. Saunders, J.M. Scalise, and G.F. Udell (1998). “The Effects of Bank Mergers and Acquisitions on Small Business Lending,” Journal of Financial Economics, 50: 187-229. [8] Calem, P.S. and L.J. Mester (1995). “Consumer Behavior and the Stickiness of Credit-Card Interest Rates,” American Economic Review, 85: 1327-1336. [9] Campello, M. (2002). “Internal Capital Markets in Financial Conglomerates: Evidence from Small Bank Responses to Monetary Policy,” Journal of Finance, 57: 2773-2805. [10] Cole, R. A., L.G. Goldberg, and L.J. White (2004).“Cookie-cutter versus character: The Micro Structure of Small Business Lending by Large and Small Banks,” Journal of Financial and Quantitative Analysis, 39: 227-251. [11] Dell’Ariccia, G., E. Friedman, and R. Marquez (1999). “Adverse selection as a barrier to entry in the banking industry,” RAND Journal of Economics, 30: 515-534. [12] DeYoung, R., L.G. Goldberg and L.J. White (1999). “Youth, Adolescence, and 27 [13] of Banks: Credit Availability to Small Business in an Era of Banking Consolidation,” Journal of Banking and Finance, 23: 463-492. [14] DeYoung, R. and I. Hasan (1998). “The Performance of De Novo Commercial Banks: A Profit Efficiency Approach,” Journal of Banking and Finance, 22: 565-587. [15] DiSalvo, J. (2002). “Deposit Runoff from Bank Mergers,” Federal Reserve Bank of Philadelphia working paper. [16] Dick, A.A. (2002). “Demand Estimation and Consumer Welfare in the Banking Industry,” Finance and Economics Discussion Series, Federal Reserve Board, 2002-58. [17] Dick, A.A. (Forthcoming). “Market Size, Service Quality and Competition in Banking,” Journal of Money, Credit and Banking. [18] Dixit, A. (1979). “A Model of Duopoly Suggesting a Theory of Entry Barriers,” Bell Journal of Economics, 10: 20-32. [19] Fudenberg, D. and J. Tirole (1984). “The Fat-Cat Effect, The Puppy-Dog Ploy, and the Lean and Hungry Look,” American Economic Review Papers and Proceedings, 74: 361-366. [20] Gandal, N. (2001). “The dynamics of competition in the internet search engine market,” International Journal of Industrial Organization, 19: 1103-1117. [21] Goldberg, L.G. and L.J. White (1998). “De Novo Banks and Lending to Small Businesses,” Journal of Banking and Finance, 22: 851-867. [22] Gorecki, P.K. (1986). “The Importance of Being First: The Case of Prescription Drugs in Canada,” International Journal of Industrial Organization, 4: 371-395. [23] Grabowski, H.G. and J.M. Vernon (1992). “Brand Loyalty, Entry, and Price Competition in Pharmaceuticals After the 1984 Drug Act,” Journal of Law and Economics, 35: 331-350. [24] Hopenhayn, H.A. (1992). “Entry, Exit, and Firm Dynamics in Long Run Equilibrium,” Econometrica, 60: 1127-1150. [25] Houston, J. and C. James (1998). “Do Bank Internal Capital Markets Promote Lending?,” Journal of Banking and Finance, 22: 899-918. 28 [26] Houston, J., C. James and D. Marcus (1997). “Capital Markets Frictions and the Role of Internal Capital Market in Banking,” Journal of Financial Economics, 46: 135-164. [27] Hughes, J.P., W. Lang, L.J. Mester and C. Moon (Forthcoming). “Recovering Risky Technologies Using the Almost Ideal Demand System: An Application to U.S. Banks,” Journal of Financial Services Research. [28] Hurwitz, M. A., and R. E. Caves (1988). “Persuasion or Information? Promotion and the Shares of Brand Name and Generic Pharmaceuticals,” Journal of Law & Economics, 31: 299320. [29] Kalyanaram, G., W.T. Robinson, and G.L. Urban (1995). “Order of Market Entry: Established Empirical Generalizations, and Future Research,” Marketing Science, 14: G1212-G1221. [30] Kerin, R.A., P. R. Varadarajan, and R.A. Peterson (1992). “First-Mover Advantage: A Synthesis, Conceptual Framework, and Research Propositions,” Journal of Marketing, 56: 33-52. [31] Kim, M., D. Kliger and B. Vale (2003). “Estimating Switching Costs: The Case of Banking,” Journal of Financial Intermediation, 12: 25-56. [32] Kiser, E. (2002a). “Household Switching Behavior at Depository Institutions: Evidence from Survey Data,” Antitrust Bulletin, 47: 619-622 . [33] Kiser, E. (2002b). “Predicting Household Switching Behavior and Switching Costs at Depository Institutions,” Review of Industrial Organization, 20: 349-365. [34] Klemperer, P. (1987). “The Competitiveness of Markets with Switching Costs,” RAND Journal of Economics, 18: 138-151. [35] Klemperer, P. (1995). “Competition when Consumers have Switching Costs: An Overview with Applications to Industrial Organization, Macroeconomics, and International Trade,” Review of Economic Studies, 62: 515-539. [36] Kreps, D.M. and J.A. Scheinkman (1983). “Quantity precommitment and Bertrand competition yield Cournot outcomes,” Bell Journal of Economics, 14: 326-337. [37] Marquez, R. (2002). “Competition, Adverse Selection, and Information Dispersion in the Banking Industry,” Review of Financial Studies, 15: 901-926. 29 [38] Muller, D.C. (1986). Profits in the Long Run. Cambridge: Cambridge University Press. [39] Muller, D.C. (1997). “First-mover Advantages and Path Dependence,” International Journal of Industrial Organization, 15: 827-850. [40] Örs, E. (2003). “The Role of Advertising in Commercial Banking,” 39th Annual Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, May. [41] Robinson, W.T. (1988). “Sources of Market Pioneer Advantages: The Case of Industrial Goods Industries,” Journal of Marketing Research, 25: 87-94. [42] Robinson, W.T. and C. Fornell (1985). “Sources of Market Pioneer Advantages in Consumer Goods Industries,” Journal of Marketing Research, 22: 305-317. [43] Schmalensee, R. (1982). “Product Differentiation Advantages of Pioneering Brands,” American Economic Review, 72: 349-365. [44] Schmalensee, R. (1989). “Inter-Industrty Studies of Structure and Performance,” in R. Schmalensee and R.D. Willig (eds.), Handbook of Industrial Organization, Vol. 2, Amsterdam: North-Holland. [45] Sharpe, S.A. (1997). “The Effect of Consumer Switching Costs on Prices: A Theory and its Application to the Bank Deposit Market,” Review of Industrial Organization, 12: 79-94. [46] Spence, A.M. (1977). “Entry, Capacity, Investment, and Oligopolistic Pricing,” Bell Journal of Economics, 8: 534-44. [47] Stackelberg, H. (1934). Marktform und Gleichgewicht, Vienna: Julius Springer. English translation,entitled The Theory of the Market Economy (1952), by Oxford University Press. [48] Stango, V. (2002). “Pricing with Consumer Switching Costs: Evidence from the Credit Card Market,” Journal of Industrial Economics, 50: 475-492. [49] Sutton, J. (1991). Sunk Costs and Market Structure: Price Competition, Advertising, and the Evolution of Concentration. Cambridge: MIT Press. [50] Szymanski, D.M., L.C. Troy, and S.G. Bharadwaj (1995). “Order of Entry and Business Performance: An Empirical Synthesis and Reexamination,” Journal of Marketing, 59: 17-33. 30 [51] Tufano, P. (1989). “Financial Innovation and First-Mover Advantages,” Journal of Financial Economics, 25: 213-240. [52] VanderWerf, P.A. and J.F. Mahon (1997). “Meta Analysis of the Impact of Research Methods on Findings of First-Mover Advantage,” Management Science, 43: 1510-1519. [53] Williamson, O.E. (1967). “Hierarchical Control and Optimum Firm Size,” Journal of Political Economy, 75: 123-138. [54] Zephirin, M.G. (1995). “Switching Costs in the Deposit Market,” The Economic Journal, 104: 455-461. 31 Table 1: NUMBER OF BANKS BY YEAR AND NUMBER OF YEARS IN THE MARKET Year Time 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Total 0 275 222 241 356 438 706 696 683 614 633 517 5, 381 1 273 262 200 226 332 378 631 635 652 596 575 4, 760 2 275 254 250 193 208 299 351 579 598 597 556 4, 160 3 298 248 237 226 180 180 263 296 526 565 549 3, 568 4 253 276 231 226 204 154 160 228 274 497 504 3, 007 5 209 237 260 219 210 166 135 137 207 240 444 2, 464 6 194 186 227 241 205 186 152 118 129 192 226 2, 056 7 231 172 171 204 212 177 172 134 112 118 179 1, 882 8 204 210 158 158 189 194 164 151 125 104 110 1, 767 9 174 187 196 149 147 154 161 147 140 118 100 1, 673 10 129 155 159 182 130 129 139 140 133 123 105 1, 524 11 111 121 143 143 156 111 117 126 124 121 110 1, 383 12 102 102 117 128 125 146 100 102 113 96 112 1, 243 13 79 96 91 107 116 118 124 84 94 103 90 1, 102 14 58 72 89 87 98 102 103 110 81 87 96 983 15 74 56 65 86 79 82 88 86 103 71 73 863 16 86 65 55 58 78 67 76 78 77 97 65 802 17 108 80 58 50 48 72 65 67 65 71 90 774 18 107 103 75 53 48 45 64 56 65 58 63 737 19 93 100 97 71 46 44 36 59 51 62 55 714 20 3 , 292 87 93 88 64 39 38 31 50 46 56 3, 884 21 . 3 , 105 80 85 81 56 35 32 28 44 42 3, 588 22 . . 2 , 915 74 79 78 49 32 30 28 43 3, 328 23 . . . 2 , 707 68 71 72 47 31 27 28 3, 051 24 . . . . 2 , 452 58 68 64 44 31 24 2, 741 25 . . . . . 2 , 241 56 56 55 43 30 2, 481 . . . . . . 2 , 078 52 55 53 41 2, 279 26 27 . . . . . . . 1 , 926 49 49 50 2, 074 28 . . . . . . . . 1 , 792 47 40 1, 879 . . . . . . . . . 1 , 675 42 1, 717 29 30 . . . . . . . . . . 1 , 596 1, 596 Total 6, 625 6, 396 6, 208 6, 117 5, 993 6, 053 6, 193 6, 256 6, 417 6, 592 6, 611 69, 461 NOTE.– Based on bank-market observations. Dates of entry into a market are based on a sample for 1972-2002. Numbers in italics represent banks that entered in 1972 or earlier and have survived since then (and for whom date of entry into the market cannot be determined), while the rest of the numbers of the table are bank entries. The first line on the table, where Time=0, corresponds to the new entrants each year of our analysis. The second line, where Time=1, corresponds to the surviving banks that entered the previous year —one-year old banks—, and so forth for the following lines of the table. 32 Table 2: AVERAGE MARKET DISTRIBUTION OF FIRMS BASED ON ORDER OF ENTRY Year 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Entrant 0≤t<5 0.23 0.22 0.22 0.24 0.26 0.34 0.39 0.43 0.45 0.47 0.43 Entrant 5 ≤ t < 10 0.14 0.15 0.16 0.15 0.16 0.14 0.12 0.11 0.12 0.13 0.18 Entrant 10 ≤ t < 20 0.14 0.14 0.15 0.14 0.14 0.13 0.13 0.13 0.13 0.12 0.12 Entrant 20 ≤ t 0.50 0.49 0.48 0.46 0.44 0.39 0.36 0.33 0.31 0.28 0.27 In1972 0.50 0.48 0.46 0.43 0.39 0.35 0.31 0.29 0.26 0.23 0.22 NOTE.– Dates of entry into a market are based on a sample for 1972-2002. In1972 indicates banks that were already in the market in 1972, for whom entry date cannot be determined. 33 Table 3: AVERAGE DEPOSIT MARKET SHARES BY TIME IN THE MARKET Time 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20+ Market Share % 3.2 3.3 3.5 3.5 3.6 3.4 3.0 3.1 3.2 3.0 3.2 3.5 3.2 3.3 3.4 3.5 3.8 3.8 3.5 3.5 7.1 NOTE.– Dates of entry into a market are based on a sample for 19722002, while the figures are based on the regression sample for 19922002. 34 Table 4: ENTRANTS’ AVERAGE DEPOSIT MARKET SHARES BY ORDER AND METHOD OF ENTRY Method of Entry Order of Entry Merger De Novo Open Branch 0≤t<5 6.8 1.4 1.6 5 ≤ t < 10 7.2 2.0 2.0 10 ≤ t < 20 8.2 2.7 2.2 20 ≤ t† 6.5 4.1 4.4 NOTE.– In percentages. Dates of entry into a market are based on a sample for 1972-2002. †Entrants 20 ≤ t only include those that entered after 1972. 35 Table 5: OLS REGRESSIONS OF MARKET SHARE ON ORDER OF ENTRY Dependent Variable: Deposit Market Share (ii) (iii) −0.041 −0.050 (0.002)∗∗ (0.007)∗∗ −0.030 −0.037 (0.002)∗∗ (0.007)∗∗ −0.014 −0.019 (0.002)∗∗ (0.005)∗∗ 0.105 (0.002)∗∗ 0.103 (0.002)∗∗ −0.002 (0.002) 0.003 (0.001)∗ −0.008 (0.006) 0.002 (0.012) NO 68946 0.23 — YES 68946 0.69 12.6** YES 51309 0.68 10.0** Explanatory Variable Entrant 0 ≤ t < 5 Entrant 5 ≤ t < 10 Entrant 10 ≤ t < 20 In 1972 (i) −0.018 (0.002)∗∗ −0.013 (0.002)∗∗ −0.006 (0.002)∗∗ 0.040 (0.002)∗∗ Profit efficiency rank (t − 1) Cost efficiency rank (t − 1) Small-business loan ratio (t − 1) Personal income growth Bank fixed effects Observations R-squared F-stat NOTE.– ALL REGRESSIONS INCLUDE MSA AND YEAR FIXED EFFECTS. Regression sample is for 1992-2002, with time of entry determined throughout 19722002. The dependent variable is the deposit market share for bank i in MSA market m in a given year. Standard errors are in parenthesis. *significant at 5%; **significant at 1%. F-statistic for the test that the bank fixed effects are jointly zero. 36 Table 6: DOES THE METHOD OF ENTRY AND GEOGRAPHIC DIVERSIFICATION MATTER? Dependent Variable: Deposit Market Share Explanatory Variable Entrant 0 ≤ t < 5 Entrant 5 ≤ t < 10 Entrant 10 ≤ t < 20 Entry by merger Ent 0 ≤ t < 5 * Merger Ent 5 ≤ t < 10 * Merger Ent 10 ≤ t < 20 * Merger Entry by branch Ent 0 ≤ t < 5 * Open Branch Ent 5 ≤ t < 10 * Open Branch Ent 10 ≤ t < 20 * Open Branch Multimarket (10+) Ent 0 ≤ t < 5 * Multimkt (10+) Ent 5 ≤ t < 10 * Multimkt (10+) Ent 10 ≤ t < 20 * Multimkt (10+) In 1972 0.032 (0.002)∗∗ (i) −0.032 (0.002)∗∗ −0.021 (0.002)∗∗ −0.013 (0.002)∗∗ −0.004 (0.005) 0.036 (0.006)∗∗ 0.027 (0.006)∗∗ 0.027 (0.006)∗∗ −0.038 (0.004)∗∗ 0.025 (0.004)∗∗ 0.020 (0.004)∗∗ 0.013 (0.004)∗∗ 0.084 (0.002)∗∗ −0.044 (0.003)∗∗ −0.043 (0.004)∗∗ −0.049 (0.004)∗∗ 0.040 (0.002)∗∗ (ii) −0.022 (0.002)∗∗ −0.012 (0.002)∗∗ −0.002 (0.002)∗∗ Bank fixed effects Observations R-squared NO 68946 0.26 NO 68946 0.26 NOTE.– ALL REGRESSIONS INCLUDE MSA AND YEAR FIXED EFFECTS. Regression sample is for 1992-2002, with time of entry determined throughout 1972-2002. The dependent variable is the deposit market share for bank i in MSA market m in a given year. The entry method base case is de novo. Standard errors are in parenthesis. *significant at 5%; **significant at 1%. 37 Table 7: INVESTMENTS IN BRANCH NETWORK Dependent Variable: Branch density Explanatory Variable Entrant 0 ≤ t < 5 Entrant 5 ≤ t < 10 Entrant 10 ≤ t < 20 In1972 (i) −0.0068 (0.0011)∗∗ −0.0067 (0.0012)∗∗ −0.0030 (0.0012)∗∗ 0.0241 (0.0011)∗∗ (ii) −0.0236 (0.0011)∗∗ −0.0152 (0.0011)∗∗ −0.0063 (0.0010)∗∗ 0.0466 (0.0012)∗∗ Bank fixed effects Observations R-squared F-stat NO 68946 0.37 - YES 68946 0.77 15.4** NOTE.– ALL REGRESSIONS INCLUDE MSA AND YEAR FIXED EFFECTS. Regression sample is for 1992-2002, with time of entry determined throughout 1972-2002. The dependent variable is for bank i in MSA market m in a given year. Standard errors are in parenthesis. *significant at 5%; **significant at 1%. F-statistic for the test that the bank fixed effects are jointly zero. 38 Table 8: IS THE EARLY-MOVER ADVANTAGE STABLE OVER TIME? Dependent Variable: Deposit Market Share 1992-1997 1998-2002 Explanatory Variable Entrant 0 ≤ t < 5 Entrant 5 ≤ t < 10 Entrant 10 ≤ t < 20 In1972 (i) −0.034 (0.003)∗∗ −0.022 (0.003)∗∗ −0.006 (0.002)∗∗ 0.105 (0.003)∗∗ (ii) −0.064 (0.003)∗∗ −0.051 (0.003)∗∗ −0.027 (0.003)∗∗ 0.105 (0.003)∗∗ Bank fixed effects Observations R-squared F-stat YES 37100 0.75 12.8** YES 31846 0.67 13.3** NOTE.– ALL REGRESSIONS INCLUDE MSA AND YEAR FIXED EFFECTS. Regression sample is for 19922002, with time of entry determined throughout 19722002. The first (second) column is based on the 19921997 (1998-2002). The dependent variable is the deposit market share for bank i in MSA market m in a given year. Standard errors are in parenthesis. *significant at 5%; **significant at 1%. F-statistic for the test that the bank fixed effects are jointly zero. Moreover, another F-test rejects the equality of order of entry coefficients across the two sample periods at the one percent level of confidence (F=95.9. 39 Table 9: FURTHER TESTS OF THE EARLY-MOVER ADVANTAGE PANEL A Dependent Variable: Deposit Market Share Sample: 1973-2002 (i) (ii) −0.001 (0.000)∗∗ 0.030 (0.005)∗∗ 0.028 (0.004)∗∗ 0.024 (0.003)∗∗ 0.022 (0.002)∗∗ 0.010 (0.001)∗∗ 0.001 (0.000)∗∗ YES 81649 0.60 10.1** YES 81649 0.60 10.1** 0.001 (0.000)∗∗ Explanatory Variable Order of entry 1st to 5th entrant 6th to 10th entrant 11th to 15th entrant 16th to 20th entrant 21st to 25th entrant No. of yrs. in the market Bank fixed effects Observations R-squared F-stat Entrant 0 ≤ t < 5 Entrant 5 ≤ t < 10 Entrant 10 ≤ t < 20 In1972 PANEL B Dependent Variable: Deposit Market Share Sample: 1992-2002 Inc.+entrant multi-mkt. banks (i) (ii) −0.045 −0.069 (0.004)∗∗ (0.004)∗∗ −0.034 −0.062 (0.005)∗∗ (0.005)∗∗ −0.021 −0.048 (0.005)∗∗ (0.005)∗∗ 0.098 0.098 (0.005)∗∗ (0.005)∗∗ NO 14626 0.35 YES 14626 0.54 9.48** Bank fixed effects Observations R-squared F-stat N O T E .– A L L R E G R E S S IO N S IN C L U D E M S A A N D Y E A R F IX E D E F F E C T S . T h e d e p e n d e nt va ria b le is th e d ep o sit m a rket sh a re fo r b a n k i in M S A m a rke t m in a g iven yea r. S ta n d a rd erro rs a re in p a re nth e sis. * sig n ifi ca nt a t 5 % ; * * sig n ifi c a nt a t 1 % . F -sta tistic fo r th e test th a t th e b a n k fi x e d eff ects a re jo intly z ero . In Pa n el A , th e b a se c a se is th e 2 6 th to 3 0 th e ntra nt g ro u p , a n d th e sam p le ex c lu d e s 1 9 7 2 in cu m b e nts. In P a n el B , th e reg ressio n sa m p le is b a se d o n b a n k s th a t a re b o th in c u m b ent in a t lea st o n e m a rket a n d e ntra nt in a t le ast o n e o th e r m a rke t. W h ile th e sa m p le covers 1 9 9 2 -2 0 0 2 , tim e o f entry is d eterm in e d th ro u g h o u t 1 9 7 2 -2 0 0 2 . 40 Table 10: ARE HIGH VS LOW POPULATION GROWTH MARKETS DIFFERENT? Dependent Variable: Deposit Market Share High Growth Low Growth Explanatory Variable Entrant 0 ≤ t < 5 Entrant 5 ≤ t < 10 Entrant 10 ≤ t < 20 In1972 (i) −0.067 (0.003)∗∗ −0.047 (0.003)∗∗ −0.028 (0.002)∗∗ 0.047 (0.003)∗∗ (ii) 0.001 (0.004) 0.001 (0.004) 0.003 (0.003) 0.180 (0.004)∗∗ Bank fixed effects Observations R-squared F-stat YES 18864 0.72 12.73** YES 16394 0.75 11.54** NOTE.– ALL REGRESSIONS INCLUDE MSA AND YEAR FIXED EFFECTS. Regression sample is for 19922002, with time of entry determined throughout 19722002. The first (second) column is based on the top (bottom) quartile of markets from the distribution of average population growth over 1972-2002. The dependent variable is the deposit market share for bank i in MSA market m in a given year. Standard errors are in parenthesis. *significant at 5%; **significant at 1%. F-statistic for the test that the bank fixed effects are jointly zero. Moreover, another F-test rejects the equality of order of entry coefficients across the two samples at the one percent level of confidence (F(3,31624)=86.97). 41 Table 11: AVERAGE BANK CHARACTERISTICS BY ORDER OF ENTRY INTO THE MARKET Bank Characteristics Gross No. of No. of Branch Age Top Total States Outside Density Bank Assets Markets % 0≤t<5 27.2B 2.85 11.27 0.00453 58 14.3 5 ≤ t < 10 9.1B 1.54 4.38 0.00364 43 6.3 10 ≤ t < 20 4.1B 1.22 2.31 0.00403 36 7.1 20 ≤ t 5.8B 1.20 1.92 0.00605 76 72.3 NOTE.– Dates of entry into a market are based on a sample for 19722002. The last column shows the breakup of banks with the largest market share in the market among the four order of entry categories. 42 Table 12: PREFERRED METHOD OF ENTRY INTO A MARKET BY BANK SIZE AND ORDER OF ENTRY 20 ≤ t† Assets 10B+ Merger Merger Merger Merger 1B-10B Merger Merger Merger Open branch 300M-1B Open branch Open branch De novo De novo 100M-300M Open branch De novo De novo De novo 100M and less De novo De novo De novo De novo NOTE.– Dates of entry into a market are based on a sample for 1972-2002. †Entrants 20 ≤ t only include those that entered after 1972. Bank size is measured by Gross Total Assets, adjusted by 1994 dollars. 0≤t<5 Order of entry 5 ≤ t < 10 10 ≤ t < 20 43 APPENDIX: SUMMARY STATISTICS 1992-2002 Variable Gross total assets ($000) Deposit market share Entrant 0 ≤ t < 5 Entrant 5 ≤ t < 10 Entrant 10 ≤ t < 20 Entrant 20 ≤ t In1972 Entry by merger Entry by branch opening Entry by de novo Number of outside markets Branded Market per capita income Market annual population change Market population Mean 12,400,000 0.049 0.301 0.142 0.146 0.412 0.371 0.165 0.215 0.255 5 0.114 25,763 0.0129 1,559,255 St. Dev. 55,100,000 0.093 0.458 0.349 0.353 0.492 0.483 0.371 0.411 0.436 16 0.318 6,011 0.011 2,030,660 Min 107 0.000 0 0 0 0 0 0 0 0 0 0 10,253 -0.050 56,780 Max 543,000,000 0.931 1 1 1 1 1 1 1 1 140 1 60,839 0.085 9,677,220 Number of observations (bank-market-year) 69,337 Constructed on the basis of the FDIC Summary of Deposits; Federal Reserve Report on Condition and Income; U.S. Census; Bureau of Economic Analysis. 44