Boundaries of the firm: evidence from the banking industry

ARTICLE IN PRESS

Journal of Financial Economics 70 (2003) 351–383

Boundaries of the firm: evidence from the banking industry$ James A. Brickleya, James S. Linckb,*, Clifford W. Smith Jr.a a

William E. Simon Graduate School of Business Administration, University of Rochester, Rochester, NY 14627, USA b Department of Banking and Finance, Terry College of Business, University of Georgia, Athens, GA 30602-6253, USA Received 11 September 2000; received in revised form 7 June 2002

Abstract Agency theory implies that asset ownership and decision authority are complements. Using 1998 data from Texas commercial banks, we test whether the likelihood of local ownership of bank offices increases with the importance of granting local managers greater decision authority (for example, due to location or customer base). Our empirical evidence is consistent with this hypothesis. It suggests that complementarities between strategy and organizational structure can foster differentiation among firms in terms of location, customers, and products. It also supports the growing view that small locally-owned banks have a comparative advantage over large banks within specific environments. r 2003 Elsevier B.V. All rights reserved. JEL classification: G32; L22 Keywords: Boundaries of the firm; Banking; Economics of organizations; Ownership incentives; Agency theory; Decision authority; Locational decisions; Riegle–Neal Act; Community banks; Interstate branching

$ We would like to thank Mark Flannery, Lee Heavner, Francine LaFontaine, Laura Lindsey, Michael Pagano, Joe Sinkey, Jerry Zimmerman, seminar participants at Emory, Georgia, Purdue, Rochester, TCU and the Western Finance Association Meetings as well as two anonymous referees for useful comments. We also thank Doug Freeman for research assistance, Sharon Boston and Dorsey Davis at the Dallas Federal Reserve, and Pat Relich of the FDIC, for providing data. *Corresponding author. E-mail address: [email protected] (J.S. Linck).

0304-405X/$ - see front matter r 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0304-405X(03)00170-3

ARTICLE IN PRESS 352

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

1. Introduction Incentive theory implies that asset ownership and decision authority are complements. Asset ownership reduces agency problems and thus the costs of granting the owner more decision authority. Substantial decision authority, in turn, increases the benefits of giving the agent an ownership interest in the assets. This theory suggests at least two testable implications. First, asset ownership and decision authority will be positively correlated.1 Second, the agent will have greater ownership rights and decision authority in environments in which it is important to grant the agent significant decision-making authority (for instance, due to the agent’s specific knowledge). In this paper, we use 1998 data from the Texas banking industry to provide evidence on these propositions. Following the convention of regulators and past researchers, we classify banks as either large or small depending on whether they have more than one billion in assets (similar results hold if we use a $500 million cutoff). Nakamura (1994), for example, indicates that the managers of the individual branch offices within large banks generally are granted limited decision rights, are required to follow standardized operating procedures, and are monitored by supervising managers from the large bank’s headquarters or a nearby banking center. Specialists from these major offices handle many of the more complex products and services. In contrast, local office managers of small banks tend to have broader decision-making authority. We begin by documenting that the ownership of small bank offices is highly concentrated among local office managers and investors from the local community. In contrast, branch managers in large banks have virtually no ownership interests in either the bank or its branches. These findings, coupled with what we know about the decision authority of the managers in large and small bank offices, are consistent with the incentive-theory implication that asset ownership and decision authority will be positively correlated. Our main empirical tests focus on the joint hypothesis that (1) bank offices will be owned by small banks in environments where it is important to grant significant decision-making authority to local office managers; and, (2) It it is more important to grant significant decision-making authority to local office managers in smaller urban and rural areas than in major cities. We expect that it will be relatively more important to grant decision authority to local managers in smaller urban and rural areas for two reasons. First, in major cities it is less expensive to refer inquiries about more complex products and services to specialists at a nearby regional banking center. In more rural markets it is more likely to be efficient to have local bank managers handle a broader set of products and services; thus, specialization is limited by the size of the market. Second, the major loan customers in small urban and rural areas are often small businesses without audited financial statements. The 1 Complementarity along with some additional assumptions relating to the exogenous variables implies . and Milgrom (1994) for positive cross-sectional correlations among the decision variables. See Holmstrom details.

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

353

local manager’s information about the customer, the community, and purpose of the loan is likely to be more important in deciding whether to fund such small business loans. At the same time, this relatively opaque information is more expensive to transfer to a remote decision maker than the more transparent information associated with large public companies that have audited accounting statements. These factors suggest that it will be efficient to grant increased lending authority to the local manager in more rural markets. Conceptually, a large bank also might grant local managers significant decisionmaking authority if they own offices in smaller urban and rural areas. However, given the multi-dimensional output of these offices, it would be difficult to develop effective incentive–compensation plans and the costs of monitoring these managers from a greater distance would be higher.2 Even if it were feasible to write an effective, customized compensation contract for managers of such offices, it might not be optimal for a large bank to do so given the influence costs that variation in the method of compensating managers within the same organization potentially engenders (see Milgrom (1988) for a discussion of influence costs). In addition, the lower demand for extensive branch systems in less populated areas reduces investment costs and thus, the advantages of financing the offices through publicly traded stock. Consistent with our joint hypothesis, we find that the likelihood that a large bank owns an office decreases significantly as the location shifts from a major city to a rural area. The majority of offices in the major Texas cities (Austin, Dallas, El Paso, Fort Worth, Houston, and San Antonio) are part of branch systems owned by large banks with widely held, publicly traded stock; these banks typically are headquartered out-of-state. Small banks, on the other hand, are more prevalent in smaller cities and dominate in rural areas. Moreover, small banks have few branches and normally concentrate in one local market area. Local managers and investors from the local community own almost all the stock in such banks. Although our analysis focuses on the boundaries of the firm, we also offer insights into two additional questions of more general interest. First, economists since Hotelling (1929) have been interested in how firms make choices in terms of geography, products, strategies, and so forth. Of particular interest is whether firms strive for maximum or minimum differentiation. Past empirical work has focused on how factors such as demand conditions, entry costs, competition, and transportation costs affect the degree of differentiation across firms.3 Our study suggests that organizational considerations are important as well; crafting an organizational strategy and structure to serve one type of customer or location can limit a firm’s effectiveness in serving other customers and locations, thus promoting differentiation across firms. Second, some commentators argue that economies of scale imply that large, vertically integrated banks ultimately will displace small community banks now that . and Milgrom (1991, 1994). See, e.g., Fama and Jensen (1983), Grossman and Hart (1986), Holmstrom See Borenstein and Netz (1999), Mazzeo (2001), Netz and Taylor (2002), Pinske and Slade (1998), Salvanes et al. (1997), and Stavins (1995). 2 3

ARTICLE IN PRESS 354

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

branching restrictions have been reduced (for instance, by the Riegle–Neal Act of 1994).4 Coupled with the extant evidence that small banks engage in relatively more small business lending than large banks, this possibility has raised the specific concern that there will be a reduction in the supply of credit to small businesses. Yet as Coase (1937) first argued, production technology should not drive organizational form. Absent contracting costs, any scale economies that a large vertically integrated bank might achieve also could be achieved by a group of small independently owned banks linked by contracts. Our evidence supports the growing view—presented, for example, by Calem (1994) and Nakamura (1994)—that small banks have a comparative advantage over large banks within certain environments. Moreover, we identify a potential explanation for this advantage. Our findings are consistent with the hypothesis that small banks’ ownership structures provide them with an advantage in small business lending. Thus, in contrast to the predictions of some observers, we argue that the recent liberalization of interstate branching regulation is unlikely to result in the elimination of small banks within the foreseeable future. States’ banking laws and regulations historically have had an important effect on organizational patterns in banking. To help control for regulation, we focus on a large sample of banks from a single state, Texas, rather than a similarly sized sample of banks drawn from a set of states with differing historical regulations. Texas has the largest number of banks and banking offices in the United States; it also covers a large geographic area and has a population that exhibits material variation in demographic characteristics across the state. These factors increase the power of our tests. Finally, Texas has permitted both large branch systems and small independent banks since it authorized countywide branching in 1987 and statewide branching in 1988. To help ensure that our results are not driven by historic branching restrictions in Texas, we replicate the analysis using data from California—a state that has never prohibited statewide branching. Although restricting our analysis to the banking industry imposes limitations (for instance, additional work is required to determine how well our results generalize to other, less-regulated industries), banks provide an interesting source for evidence on the boundaries of the firm for at least four reasons. First, physical assets in banking are not particularly firm specific; moreover, the level of physical asset specificity is unlikely to vary materially across banking offices. Thus, banks are especially well suited for providing evidence on how factors other than asset specificity, such as ownership incentives and risk-bearing considerations, affect integration decisions.5 4

For a discussion of the concerns relating to interstate banking, see, e.g., Berger et al. (1998), Rose (1997) and U.S. House of Representatives (1991, 1993). Berger et al. (1995) forecast consolidation under nationwide banking, but also predict that thousands of small banks will remain. They predict a decline in small business lending, but suggest that many of the eliminated loans may have negative net present values. Petersen and Rajan (2002), in contrast, argue that improvements in information technology are likely to increase the amount of small business lending by large banks. 5 Much of the empirical work on the boundaries of the firm has focused on physical asset specificity— generally motivated by the analyses of Klein et al. (1978) and Williamson (1975, 1985). See, e.g., Crocker and Masten (1991), Crocker and Reynolds (1993), Goldberg and Erickson (1987), Joskow (1988), Masten (1984), and Monteverde and Teece (1982). The analysis in this paper is related to those found in the

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

355

Second, commercial banks vary substantially in their degree of horizontal and vertical integration; thus, our tests of explanations of this organizational variation should have power. Third, regulations mandate that banks must disclose financial and organizational information (such as disaggregated operating information and office location) that is rarely disclosed by firms in other industries. Fourth, having a large sample of firms from the same industry controls for a variety of potentially important omitted variables that might confound the interpretation of inter-industry studies. The paper is organized as follows. In Section 2, we discuss organizational arrangements within the banking industry. We develop our empirical predictions in Section 3 and provide an overview of the Texas banking market in Section 4. In Section 5, we present our empirical results, and we conclude in Section 6 with a brief discussion of our results and their implications.

2. Banking organization Prior to developing our hypotheses, it is important to provide an overview of the relevant institutional features of the U.S. banking industry. This section begins by summarizing the historic trends in the number and size of U.S. banks and bank offices. Next we discuss the services and products offered by banks and how banks vary in their vertical and horizontal structures. Finally, we discuss differences among bank customers and present an argument for why small banks might have a comparative advantage over large banks in small-business lending. 2.1. Bank structure trends Since 1984 there has been a persistent decline in the number of FDIC-insured commercial banks within the United States, falling 41% from an all-time high of 14,496 in 1984 to 8,581 at year-end 1999 (see Table 1, Panel A).6 Yet over this same period, the number of bank offices across the nation increased 28.4% from 56,295 in 1984 to 72,265 in 1999. While the total number of banks has declined substantially over the past decade, small banks (banks with less than $1 billion in total assets) have continued to play an (footnote continued) literature on contracting in retailing, franchising, and trucking. See Anderson (1985), Anderson and Schmittlein (1984), Baker and Hubbard (2000), Barron and Umbeck (1984), Brickley and Dark (1987), John and Weitz (1988), LaFontaine (1992), Martin (1988), Minkler (1990), Muris et al. (1992), Norton (1988), Scott (1995), Shepard (1993) and Slade (1996, 1998). See LaFontaine and Slade (1997) for a review of the literature on retail contracting. 6 To capture all FDIC-insured commercial banks, in this section and Table 1 we report data from the United States plus other areas (Virgin Islands, American Samoa, etc.), as opposed to just those headquartered in the United States. The number of non-U.S.-based FDIC-insured commercial banks is trivial, however, numbering 18 of the 8,581 banks in 1999. For ease of exposition, we refer to these banks as U.S. banks.

ARTICLE IN PRESS 356

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

Table 1 Banking trends—Federal Deposit Insurance Corporation (FDIC)-insured commercial banks Panel A reports the number of banks, branches, total number of banking offices, and per office average assets and deposits for all FDIC-insured commercial banks in the United States and other areas as of 12/31 each year. Panel B reports the number of banks and assets across groupings of bank size as measured by total assets as of 12/31/1999. All data was drawn from the FDIC’s bank data (http://www.fdic.gov/bank/ index.html). Panel A—Banks, branches, offices and size—FDIC-insured commercial banks Year

Banks

Branches

Offices

1999 1998 1997 1996 1995 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 1980

8,581 8,774 9,143 9,530 9,942 10,452 10,960 11,466 11,927 12,347 12,715 13,137 13,723 14,210 14,417 14,496 14,469 14,451 14,414 14,434

63,684 61,957 60,325 57,789 56,512 55,145 52,868 51,935 51,969 50,406 48,005 46,381 45,357 44,392 43,293 41,799 40,853 39,783 40,786 38,738

72,265 70,731 69,468 67,319 66,454 65,597 63,828 63,401 63,896 62,753 60,720 59,518 59,080 58,602 57,710 56,295 55,322 54,234 55,200 53,172

Average assetsa ($millions) 79.4 76.9 72.2 68.0 64.9 61.1 58.1 55.3 53.7 54.0 54.3 52.6 50.8 50.2 47.3 44.6 42.3 40.4 36.8 34.9

Average depositsa ($millions) 53.0 52.0 49.3 47.5 45.6 43.8 43.2 42.6 42.1 42.2 42.0 40.9 39.5 39.0 36.7 34.9 33.3 31.5 28.8 27.9

Panel B—Frequency and assets of banks across the bank’s total assets at 12/31/1999 Category (bank assets ($m)) Assets o ¼ 100 100oAssetso ¼ 500 500oAssetso ¼ 1; 000 1; 000oAssets Totals a

Banks

Percent of total

Total assets ($billions)

Percent of total

5,158 2,729 300 394 8,581

60.1% 31.8% 3.5% 4.6% 100.0%

242.5 547.9 206.6 4,737.7 5,734.8

4.2% 9.6% 3.6% 82.6% 100.0%

Average assets and deposits are on a per office basis.

important role within the economy. Table 1, Panel B shows that 60% of U.S. banks in 1999 had less than $100 million in assets, and 95% had assets of less than $1 billion. But large banks (those with more than $1 billion in assets) held almost 83%

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

357

of all commercial banking assets. Thus, while large banks clearly are important in terms of the assets and deposits that they control, they continue to coexist with many small banks. Mergers and consolidations still occur frequently within the banking industry. Although these transactions indicate that the industry is still evolving, the existence of small banks does not appear to be simply a remnant of past restrictions on branching. California, for example, has never prohibited branching; yet in 1999, 34% of the banks in the state had less than $100 million in assets. Further, new banks continue to be formed; for example, in 1999, 232 new bank charters were issued nationwide, nine in Texas (http://www.fdic.gov/bank/, report cb092). In our empirical analysis, we perform several robustness checks to provide additional assurance that our results are not simply a consequence of prior regulatory restrictions on banking. 2.2. Banking services Most banks, both large and small, offer a combination of retail and commercial products. Retail products include checking and savings accounts, residential mortgages, auto loans, home equity loans, Internet banking, and access to ATM machines. Businesses purchase commercial products including checking accounts, international banking services, term loans, lines of credit, and cash management services, among others. Below we argue that the organizational characteristics of small banks provide a relative advantage over large banks in certain geographic locations as well as in serving certain types of customers, whereas large banks have advantages in other locations and with other customers. That neither type of bank dominates within all settings helps to explain their coexistence. We now examine how organizational characteristics of large and small banks vary in both vertical and horizontal dimensions. 2.3. Vertical organization Most bank offices rely on larger offices to provide products and services such as asset/liability management, bankcard services, regulatory compliance, international banking, electronic fund transfers, federal funds, Internet banking, loan participations, student loans, and back-office operations. Publicly traded holding companies, such as JPMorgan Chase, normally own these large upstream banks and have billions of dollars in deposits. Vertical integration describes the case in which upstream banks, which correspond to our notion of large banks, own downstream offices that are organized as branches. Branches are part of a larger bank; wholly owned subsidiaries are separately incorporated entities. Multi-state banks now are allowed to place branches directly in other states without first establishing a subsidiary bank within the state. This was not true during our sample period in Texas. Prior to 1999, Texas required multi-state banks to own subsidiary banks that, in turn, could form branches. Vertical separation, on the other hand, is the case in

ARTICLE IN PRESS 358

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

which other downstream offices, which correspond to our notion of small banks, receive upstream products and services from large banks on a contractual basis. For example, smaller banks might place deposits with larger correspondent banks in return for products and services. The fact that large and small banks alike buy and sell offices is consistent with the view that vertical integration and vertical separation are organizational alternatives. Offices of small banks become branches in large banks through mergers and acquisitions. Conversely, small banks acquire offices from large banks through market transactions. For instance, Bank of America agreed to sell 14 rural Texas branches to Pacific Southwest Bank of Corpus Christi in 1996. This was a component of Bank of America’s plan to sell 68 of its rural Texas branches. At the same time, Bank of America was opening branches in Dallas, Fort Worth and Houston: ‘‘The goal here is to take the funds that we are getting from these sales and apply them back into our major markets,’’ according to Randy Hicks, a Bank of America, Texas spokesman (Mensheha, 1996, San Antonio Business Journal). Cline (1996) and O’Hara (1997) provide further discussions of branch sales. Finally, banks also change their correspondent relationships.

2.4. Horizontal organization Some offices are stand-alone unit banks (i.e., those operating from a single physical location), while others are part of extensive branch systems. Banks generally establish offices simply to place retail outlets near their customers. Moreover, customers frequently demand bank services at multiple geographic locations; for example, commuters might want to bank near home in the evenings and weekends, but near work weekdays, or family members might value access to branch offices near schools or local shopping centers. Separate banks can contract among themselves to provide services to customers, as in the case of ATM machines. However, branch systems have at least two operational advantages over separate banks when customers have extensive demands for bank services at multiple locations. First, regulations increase the cost of serving customers through a network of independent banks relative to a branch system. For instance, independent banks (including subsidiaries) are limited in their ability to accept each other’s deposits in ways not imposed on branches. Moreover, separately chartered banks must also comply individually with capital and reporting requirements, while a bank with a network of branches must comply only for the bank as a whole. Second, separate banks with independent ownership can have incentives to free ride on cooperative agreements—a manager interested in profits at a given unit will have limited incentives to expend resources to serve customers from other banks. Also, a bank manager might be understandably reluctant to encourage customers to patronize rival banks. Within a branch system, executives can structure incentives to encourage managers to focus on customer service regardless of the customer’s home office. (Brickley (1999), Brickley and Dark (1987) and Klein and Saft (1985) offer similar arguments in analyzing franchise contracts.)

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

359

2.5. Customers Banking customers vary in at least two important dimensions: scale and informational transparency. By scale, we mean the size of loans and the level of demand for special services, such as international banking and corporate hedging. Informational transparency refers to the type of information that is available for evaluating credit risk in transactions such as loan applications. At one end of this informational transparency spectrum is transparent information, which is publicly available or easily documented through audited financial statements, stock prices, 10-K reports, checking account balances, credit ratings/reports, and so on. At the other end of this spectrum is opaque information, which is developed through the banking relationship with the customer. It includes information gathered through an ongoing series of banking transactions as well as interactions with the borrower’s customers and other members of the local business community (for a more extended discussion see Berger et al., 2001). Local office managers acquire much of this information, which can be expensive to aggregate and transfer to a distant supervising manager on a timely basis.7 Large banks have a comparative advantage over small banks in serving large-scale customers. Small banks lack the capital to extend large loans without additional contracting with correspondent banks. Similarly, they generally use their correspondent banking relationships to outsource many specialist services, such as international banking assistance, that are demanded by large-scale customers. It is likely to be more efficient for large banks to serve customers directly that have high demands for such services. The scale of the transactions makes it profitable for large banks to invest in developing relationships with these customers and thus, to meet their demands better. The typical retail customer, in contrast, engages in small transactions and demands loans that are easily evaluated based on readily available information. In fact, such lending decisions are frequently handled by automated credit scoring systems. It is not obvious that either large or small banks have a comparative advantage in serving such customers; these customers are likely to choose banks based on location and price. Some researches (for example, Nakamura, 1994) argue that small banks do have a comparative advantage over large banks in processing opaque information as it relates to small businesses and the self-employed. Consistent with this argument, the evidence suggests that small banks tend to concentrate on small business lending (see, e.g., Berger et al., 1995, 2001; Cole et al., 1999; Petersen and Rajan, 2002). An important question is why? Agency theory provides a potential answer. Agency theory posits that it is important to provide decentralized decision-makers . and Milgrom (1991) with incentives to promote productive decisions. Holmstrom argue that it is difficult to write effective incentive contracts when the agent is responsible for multiple tasks and some of these tasks are expensive to observe: 7

Petersen and Rajan (2002) argue that improvements in information technology are reducing the costs of transferring local information in banking. We discuss the implications of such changes in Section 6.

ARTICLE IN PRESS 360

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

Tying pay to those tasks that are readily observable will not only increase focus on these tasks, but will also motivate the agent to ignore other important, but uncompensated, outputs that are more difficult to measure. For example, paying a bank manager on near-term profits might encourage the manager to increase current accounting profits at the expense of future profits, and thus the manager may fund overly risky loans or fail to invest in new technologies. Fama and Jensen (1983) as . and Milgrom (1991) argue that one way to address this problem well as Holmstrom is to locate decision authority and asset ownership jointly. Similarly, authors such as . and Milgrom (1994) argue Hart (1995), Hart and Moore (1990), and Holmstrom that asset ownership can have important incentive effects when agents make decentralized investment decisions and complete contracts are expensive to draft or administer. Local office managers who interact with small businesses and the local community often acquire opaque information that is important in evaluating the credit risk of the enterprise. But to exploit this information effectively, the local manager must be granted substantial decision rights. Incentive theory therefore implies that, to control agency problems, the manager should also be provided material ownership claims, given the difficulty in designing effective incentive compensation schemes within this multi-task environment. These arguments suggest that small banks will allocate broader decision-making authority to local managers and will concentrate ownership locally. Large banks will allocate more limited local decision authority to local managers and will have little local ownership. These predicted patterns appear to correspond with current industry practice. As previously discussed, large and small banks currently coexist in the U.S. In the early history of U.S. banking, the industry was dominated by small banks. Consistent with our arguments, these small banks tended to be locally owned and owners/directors took an active role in decision making. See, for example, Lamoreaux’s (1994) study of the banking industry in New England during the first half of the 19th Century. As described by Nakamura (1994), local managers of small banks have greater decision-making authority (for instance, in making loans) than branch managers in large banks. Large banks generally adopt standardized operating procedures that are used across all branches. If a loan does not fall comfortably within the guidelines given to a local office of a large bank, the local loan officer typically must request an exception from the main office. Customers demanding more specialized products (such as commercial lending, corporate risk management, or bankcard services) usually are referred to specialists from either the bank’s headquarters or its nearby regional banking center. A typical large bank, such as JPMorgan Chase, provides an extensive mix of services and products to a large variety of customers. Often these banks matrix their organizational design based on geography, product and customer. For example, branch offices will be set up to provide services to retail customers within a particular geographic area. However, the bank will also have product specialists (who focus, for instance, on commercial loans, credit cards, student loans, or risk-management products)

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

361

and customer specialists (who focus on selling a variety of products to firms within a given industry, such as the oil industry, or to a specific group, like high net worth individuals). Often the branch makes the initial contact with the customer and the specialists provide the services. In other cases, the specialist might contact the customer directly, sometimes taking a branch officer on the call but other times bypassing the branch office. In contrast, local managers of small banks tend to be less specialized and have broader lending authority. Small banks regularly emphasize these differences in their marketing. For instance, the Independent Community Bankers of America posts the following Community Bank Advantages: ‘‘Community banks focus attention on the needs of local families, businesses, and farmers. Conversely, many of the nation’s megabanks are structured to place a priority on serving large corporations. Community bank officers are generally accessible to their customers on site. CEOs at megabanks are often headquartered in office suites, away from daily customer dealings. Many community banks are willing to consider character, family history and discretionary spending in making loans. Megabanks, on the other hand, often apply impersonal qualification criteria, such as credit scoring, to all loan decisions without regard to individual circumstances. Community banks offer nimble decision-making on business loans because decisions are made locally. Megabanks must often convene loan approval committees in another state.’’ (Source: www.icba.org).

3. Empirical predictions: geographic location and office ownership As discussed in more detail below, the structure of the Texas banking industry has changed substantially since deregulation. Although mergers, office sales, and other restructurings continue to occur, we think it is reasonable to interpret our empirical results presuming that the highest-valued user tends to own each bank office. We provide several robustness checks to help ensure that this interpretation is reasonable. We focus on how geographic location affects office ownership. This theory also suggests that small banks will be relatively more important than large banks in making loans to small businesses and the self-employed. Past studies provide support for this prediction. For example, although large banks make loans to small businesses, these loans are a higher proportion of the loan portfolios for small banks. Nakamura (1994) reports that banks with assets of less than $1 billion made three-fourths of all bank loans smaller than $1 million, while banks with assets of more than $1 billion made nine-tenths of bank loans larger than $1 million. (See also Berger et al., 1995; Cole et al., 1999; Petersen and Rajan, 2002). Because our data do not contain information about small business lending or loans by loan size, we are unable to examine small business lending across the different organization types.

ARTICLE IN PRESS 362

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

3.1. Large urban areas Large banks generally place their headquarters and regional banking centers in metropolitan areas to be near large sets of their potential customers—especially large corporate customers. Given the location of these regional banking centers, we expect that large banks own a relatively large proportion of the bank offices in major urban areas. As noted above, many retail customers within such markets value access to a branch system that blankets the metropolitan area. Large banks with access to well-diversified investors through public capital markets have an advantage in financing these large branch systems. Branch managers within large banks typically have virtually no ownership stakes in their banks; rather, they are granted limited decision authority and are monitored readily by executives from a nearby regional banking center. Complex transactions initiated at the branch normally are referred to a specialist elsewhere in the bank. Small banks, to the extent that they exist within metropolitan areas, are likely to focus on small businesses and other customers that are not well served by large banks, for example, upscale middle-income individuals who demand specialized bank services and attention.

3.2. Smaller urban areas We expect that large banks place some offices in smaller urban areas to serve large-scale customers, as well as retail customers with demands for a larger statewide or nation-wide branch system. Small banks, however, are likely to own a larger fraction of offices in these areas than in large cities. First, small businesses are relatively more important in smaller urban areas than in larger cities, leading to a greater relative importance of small banks. For example, in 1999, 57% of the employees in the five largest metropolitan areas in Texas worked for companies with more than 500 employees, compared to 48% of the employees in smaller metropolitan areas. Second, small banks are at less of a disadvantage in financing branch systems in these areas because a smaller number of offices is required to cover the metropolitan area.

3.3. Rural areas We expect that large banks will own few offices in rural areas. These areas rarely contain large-scale customers. While large banks could place offices in rural areas to serve standard retail customers, on-site monitoring of branch managers at locations further removed from the regional banking centers can be expensive. The population in rural areas is less likely to support multiple bank offices. Thus, a small bank office, which has a comparative advantage in serving small businesses in a rural area, is also likely to serve the area’s standard retail customers.

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

363

3.4. Hypotheses These arguments suggest the following testable hypothesis: Office-location hypothesis: Large banks will own a larger proportion of the offices in major urban areas, a smaller proportion of the offices in smaller urban areas, and few offices in rural areas. Note that this is a joint hypothesis: (1) Bank offices will tend to be owned locally, by local office managers and investors from the local community, when it is important to grant specific decision-making authority to local office managers; and, (2) Giving local office managers more decision-making authority is particularly important in rural markets because of the nature of the customer base and the distance from the regional banking centers of large banks. There are other potential explanations for the ownership patterns predicted by this hypothesis that are unrelated to incentive theory. For instance, some researchers have suggested that small banks are less efficient than large banks and that small banks are therefore more likely to survive within the less competitive rural environments (see, e.g., Berger and Hannan, 1998). Another explanation suggests that large banks simply may have developed procedures, information systems, and products that work well for serving large commercial customers, but that are ill-suited for smaller commercial customers—i.e., there may be limited economies of scope or scale in dealing with smaller customers. By these arguments, large banks might avoid rural locations not for incentive-related reasons, but simply because their target customers are not located there. Following this latter argument, some observers have suggested that locally owned banks are likely to consolidate into larger branch systems and holding companies, possibly across state lines (see Nakamura, 1994, p. 8, for a discussion of this organizational possibility). Conceptually, these larger banking organizations could maintain their decentralized decision rights and operating procedures that have worked well for their target customers and at the same time take advantage of certain scale economies (for instance, not having to comply individually with regulatory requirements). KeyBank, for example, has a stated policy of leaving local management alone when they are taken over. Bank One apparently had a similar policy, but then revised their policy in order to reduce operating costs through more centralized production of standardized products.8 Finally, such combinations would allow the bank to satisfy regulatory constraints such as reserve requirements more effectively. They also might want to diversify geographically to reduce the risk from being overly dependent on a local economy. For example, Gilbert and Belongia (1988) and Lawrence and Klugman (1991) present evidence that rural bank subsidiaries of geographically diversified holding companies have greater opportunities to diversify risk than independent banks. Smith and Stulz (1985) show that value-maximizing firms—including banks—have incentives to reduce volatility due 8

We thank Mark Flannery for these specific examples.

ARTICLE IN PRESS 364

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

to: (1) convex tax schedules; (2) costs of financial distress; and, (3) managerial risk aversion. While a geographically dispersed branch system of formerly independent banks might work well for some types of customers and products, incentive theory suggests that it is less likely to work well in situations where it is important to delegate significant decision rights to local managers. Assuming it is difficult to write effective incentive compensation contracts for these managers, perhaps due to the multidimensional nature of their responsibilities, it is important to locate decision authority and ownership jointly. Thus, in contrast to the above alternative explanation, incentive theory suggests that many small independent banks will continue to exist, each concentrating offices within specific market areas where their managers are most knowledgeable. Local ownership provides incentives for mutual monitoring by local office managers, assuming they own stock, and monitoring by local investors who own blocks of stock. This monitoring becomes less effective as the offices become more dispersed geographically. In contrast, offices of large banks will tend to be more geographically dispersed than those of small banks since large banks are likely to locate branches in most major metropolitan areas, as well as smaller urban areas with large-scale customers. These arguments suggest a second hypothesis: Geographic-concentration hypothesis: Small banks will continue to exist in a deregulated environment with each individual bank concentrating its branches within a given market area. The offices of small banks will be more concentrated geographically than large banks, which are likely to locate offices in multiple urban markets.

4. The Texas banking market Our basic sample consists of FDIC-insured Texas commercial banks in 1998. Before we move to our empirical analysis, it is important to provide some background detail about Texas’ geography, banking regulation, and banking industry. 4.1. Texas geography Texas is the second largest state in the United States, both in terms of population (19.8 million in 1998) and square miles (268,601), and has the largest number of insured commercial banks in the country (798 banks with 4,012 offices in 1998; source: http://www.fdic.gov). The distances across Texas from north-to-south and from east-to-west are approximately 400 and 750 miles, respectively. As depicted in Fig. 1, the state is divided into 254 counties, roughly equal in area (although the western counties tend to be somewhat larger). The six most populous counties, highlighted in Fig. 1, contain the cities of Houston, Dallas, Fort Worth, San Antonio, Austin, and El Paso, and, in 1998, they collectively contained nearly

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

Roberts Pop: 891

King Pop: 376

Tarrant Pop: 1.4m (Fort Worth)

Borden Pop: 723 El Paso Pop: 0.7m (El Paso)

365

Dallas Pop: 2.1m (Dallas)

Loving Pop: 95

Harris Pop: 3.2m (Houston)

Bexar Pop: 1.4m (San Antonio) McMullen Pop: 755 Kennedy Pop: 422

Travis Pop: 0.7m (Austin)

Fig. 1. Counties in the state of Texas. The figure reports the size and population for Texas and its largest and smallest counties. Size and population: covers 268,601 square miles; contains 254 counties; total population (1998): 19,759,614; 2nd most populous in the U.S.; population of top six counties (1998): 9,363,967 (47.4%); population of smallest six counties (1998): 3,262 (0.02%).

half of the state’s population. The six least populated counties collectively had a population of 3,262 (less than 0.02% of the state’s population) and a total of two banking offices in 1998. We generally employ the county as the relevant local market area; however, for some tests, we use Metropolitan Statistical Areas (MSAs). There are 27 MSAs in Texas. To tie to our hypotheses more directly, we further divide the six major MSAs (Houston, Dallas, Fort Worth, San Antonio, Austin, and El Paso) into city and suburb where we define city as the zone within the city limits and suburb as the region within the MSA but outside the city limits. In summary, Texas is quite diverse; it varies substantially in geography, population density, and types of banking customers. Texas contains three of the ten largest cities in the United States, and at the same time, ranks number one in the number of farms and area of land devoted to agriculture. This heterogeneity in Texas geography and demography should increase the power of our tests.

ARTICLE IN PRESS 366

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

4.2. Texas banking regulation Prior to late 1986, Texas law prohibited branch banking.9 During the late 1980s, the Texas banking industry underwent a significant financial crisis due to declines in oil and real estate prices. This crisis prompted deregulation. Effective January 1, 1987, the state’s constitution was amended to permit limited bank branching. Texas allowed countywide branching in 1987 and statewide branching in 1988. Beginning in 1987, banks headquartered outside of Texas were allowed to purchase Texas banks. By 1994, many banks had multiple branches, and eight of the largest ten Texas banks were owned by out-of-state holding companies. Table 2 reports the number of banks, branches, and total number of banking offices for all insured commercial banks in Texas from 1980 through 1999. The branches that the FDIC reports for years before 1987 (when the state constitution was amended to allow branches) generally were established in the 19th century prior to the banning of branching and were grandfathered. Although the number of banks has declined after peaking in 1986 at 1,972, the total number of banking offices has increased steadily. The total number of banking offices in Texas has grown from 2,500 in 1988 to over 4,000 in 1999—an increase of more than 63%. The organization of Texas banks has been relatively unconstrained by government regulation since 1988. Our sample data are from 1998. Given the substantial restructuring that has occurred in Texas banking over the 1988–1998 decade, we believe it is reasonable to presume that the prevailing structure in 1998 is more a result of efficiency concerns than simply an artifact of prior regulation. However, to the extent that Texas is still in transition in 1998, our tests are potentially biased if more banks exist than would be expected if adjustment to the regulatory change were complete. Therefore, to provide further evidence that our results are not driven by prior regulation, we perform two robustness checks. First, we examine the large subsample of Texas banking offices opened since 1988, within the relatively deregulated environment. We argue that the prior regulation should have less of an impact on the ownership of these offices. Second, we examine whether our predictions hold using data from another state, California, a state that has never prohibited branch banking.

9

Although branch banking was prohibited, multi-bank holding companies were permitted and were established in the 1970s and 1980s. As a result, Blackwell et al. (1994) report that by 1986, more than 140 multi-bank holding-company systems had been established in Texas—some with over 70 banks within the holding company. A multi-bank holding company can achieve some of the advantages of branch banking, but not all. For instance, it can provide branding, facilitate the provision of centralized back-office functions, and offer more specialized product offerings. But despite common ownership, transactions between banks within the holding company must be on more of an arm’s-length basis. Moreover, each of the banks has to satisfy regulatory requirements. These additional requirements raise the cost of organizing interrelated offices as a holding company, rather than as system of branches within the same bank.

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

367

Table 2 Federal Deposit Insurance Corporation (FDIC)-insured commercial banks in Texas The table reports the number of banks, branches, total number of banking offices, and per office average assets and deposits for all FDIC-insured commercial banks in Texas as of 12/31 each year. All data is drawn from the FDIC’s bank data (http://www.fdic.gov/bank/-index.html). Year

Banks

1999 1998 1997 1996 1995 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 1980

754 798 839 877 935 980 1,011 1,089 1,122 1,184 1,318 1,501 1,772 1,972 1,936 1,854 1,727 1,598 1,523 1,467

Branches

Officesa

Average assetsb ($millions)

3,335 3,214 3,058 2,812 2,570 2,376 2,190 1,957 1,806 1,591 1,324 999 687 421 372 339 318 305 298 260

4,089 4,012 3,897 3,689 3,505 3,356 3,201 3,046 2,928 2,775 2,642 2,500 2,459 2,393 2,308 2,193 2,045 1,903 1,821 1,727

44.3 44.8 60.3 55.6 57.8 56.0 57.5 57.6 57.7 61.6 65.9 68.4 77.1 86.8 90.6 90.2 89.0 85.9 78.7 68.8

Average depositsb ($millions) 35.0 37.2 49.2 45.6 44.3 45.7 47.2 48.7 50.4 52.5 54.5 57.4 61.5 69.0 72.9 72.4 71.3 68.9 63.1 56.4

a Number of offices differ from that used in the final sample (3,548 shown in Table 5) because (i) our sample data is as of 6/30/99, while the above FDIC-reported data is as of 12/31/99; and, (ii) our sample excludes 427 offices that have no deposits as of 6/30/99. b Average assets and deposits are on a per office basis.

4.3. Texas banks Our sample consists of the 3,548 commercial bank offices (headquarters and branches) in Texas with nonzero deposits as of June 30, 1998 in Sheshunoff Information Services Inc.’s BranchSources database. This sample does not include ATMs, bank loan offices, or non-bank depository institutions such as S&Ls or credit unions. We define a bank as the top holding company. We do not distinguish among offices owned by different subsidiaries of the bank—all are treated as offices of the same bank. For example, we treat all the offices of Chase Manhattan’s two Texas subsidiaries, Chase Bank Texas NA (121 offices) and Chase Bank of Texas-San Angelo, NA (1 office), as offices of Chase Manhattan Corp. of New York (now JPMorganChase). In our sample, most of the Texas offices owned by a given bank are not in different subsidiaries. The 3,548 commercial bank offices in our sample are controlled by 703 different banking organizations. Of these, 243 (35%), which own

ARTICLE IN PRESS 368

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

428 Texas offices, are not organized as holding companies; 392 (55%), which own 1,503 Texas offices, have but one Texas bank within the holding company, and only 68 (10%), which own 1,617 Texas offices, have more than one Texas bank within the holding company. Of these 1,617 offices, 1,143 (71%) are held by the Texas subsidiary that has the largest number of Texas offices. (Note, however, that our basic results are not sensitive to this convention; if we use each Texas bank as the unit of analysis, our conclusions are unaffected). Our primary hypotheses distinguish between small and large banks. We define a small bank as one that has less than $1 billion in total assets—including all its offices, whether or not located in Texas—and a large bank as one that has more than $1 billion in total assets. Similar results are found when we use $500 million to define the categories. 4.4. Ownership structure of small and large banks Our analysis suggests that local office managers and investors own small banks, while large banks are widely held. In this section, we provide evidence that is consistent with this proposition. Ownership information is readily available in the proxy statements of publicly traded companies. But only 35 of the 703 banks in our sample are publicly traded. Proxy statements are not available for the remaining 668 closely held banks in our sample. But many of these banks are organized as holding companies and thus are required to file with the Federal Reserve an ‘‘Annual Report of Bank Holding Companies—FR Y-6’’, which includes ownership information.10 Using these annual reports and proxy statements, we compile ownership information for 27 of the 29 large banks in our sample and a random sample of 50 small banks.11 Table 3 compares the ownership structures of the 50 small banks and 27 large banks for which we have ownership information. Consistent with our analysis, the small banks have significantly more concentrated ownership than the large banks. The median (mean) percentage ownership held by blockholders (owners of at least 5% of the stock) for small banks is 77% (70%) compared to 9% (28%) for large banks. Most blockholders in these banks are officers and directors. The median (mean) percentage ownership held by officers and directors for small banks is 68% (62%) compared to 10% (26%) for large banks. All of these differences are 10 Banks also file other ownership reports with the Federal Reserve; however, these reports are confidential, and we are unable to access them. Since many of the banks in our sample do have holdingcompany structures (392 are single-bank holding companies and 68 are multi-bank holding companies), we are able to use the FR Y-6 reports to gather ownership data for a subset of the banks in our sample. We thank Sharon Boston and Dorsey Davis at the Dallas Federal Reserve Bank for providing copies of these reports. 11 We do not have proxy statements or Y-6 reports from which to gather ownership information for two of the large banks because they are not publicly traded and are headquartered outside the Dallas Federal Reserve Bank’s region (the Federal Reserve branch that supplied us with the Y-6 reports). We do not compile ownership information for all of the small banks with available FR Y-6 reports because of the expense (the reports are not computerized and the data are not reported employing a uniform format).

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

369

Table 3 Ownership-structure of banking organizations From our full sample, we select a subsample of banks for which we collect detailed ownership data. The subsample consists of 27 of the 29 large banks in our sample for which we could obtain ownership data, and a random sample of 50 small banks. A large bank is one that has more than $1 billion in total assets (including assets held by offices owned by the bank but outside Texas), and a small bank is one with less than $1 billion in total assets. We collect ownership data from proxy statements and from the Federal Reserve’s Y-6 report (Annual Report of Bank Holding Companies), obtained from the Dallas Federal Reserve Bank. Ownership is defined as all shares either directly or beneficially owned and includes options exercisable within 60 days (if reported). An outside owner is one that is not an employee, former employee or family member of an employee. Panel A reports size statistics for the subsamples, and Panel B reports their ownership concentrations. Panel C reports the mean and median distances (in miles) from outside owners’ residence to the bank’s headquarters for small banks. The p-values for difference in means is from an ANOVA test and difference in medians is from a Kruskal-Wallis Chi-square test. Panel A—Sample statistics Large

Small

p-value

N 27 Mean assets ($millions) 42,410.6 Median assets ($millions) 4,447.0

50 103.0 66.8

(o0.01) (o 0.01)

Panel B—Ownership concentration Percent held by 5% ownersa

Median Mean First quartile Third quartile Minimum Maximum Standard deviation

Percent held by officers/directors

Large Small

Diff

p-value

Large Small

Diff p-value

8.5 28.2 0.0 42.0 0.0 100.0 35.9

68.0 41.0 51.9 48.6 0.0 0.0 7.1

ðo0:01Þ ðo0:01Þ

9.7 67.9 25.8 62.4 3.9 36.0 22.0 90.6 1.0 2.2 100.0 100.0 33.8 31.0

58.2 ðo0:01Þ 36.6 ðo0:01Þ 32.1 68.6 1.2 0.0 2.8

76.5 69.2 51.9 90.6 0.0 100.0 28.7

Panel C—Distance (in miles) from outside owners’ residences to the bank’s headquarters for small banks

Outsiders on the board Outsiders not on the board (5% owners) All outsiders

Nb

Meanc

40 26 44

22.9 160.9 86.3

Medianc 8.7 34.9 18.4

a The 5% blockholders include all owners who either directly or beneficially own 5% of the firm’s outstanding shares. b Note, N varies, and is not equal to 50 (the size of the small firm subsample) for several reasons: (i) Insufficient address information for the owners (1 bank); (ii) The bank has no outsiders on the board (9 banks); and/or, (iii) The bank has no non-director outsiders with a 5% ownership stake (24 banks). c We first calculate the mean distance from each owner to its bank’s headquarters. This panel then reports the mean (median) of those means across all banks in the sample.

ARTICLE IN PRESS 370

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

statistically significant and appear economically material. Panel C of Table 3 shows that the small-bank, outside directors and other non-management, non-director 5% stockholders tend to live in reasonable proximity to the bank, normally within the same community.12 4.5. Frequency of banks across size categories Our sample consists of 703 commercial banks with an office located in Texas. Of these banks, 658 (94%) have fewer than ten offices in Texas, and 675 (96%) have less than $500 million in Texas deposits. The eight largest banks (by number of Texas offices) control 55% of the total commercial banking deposits in Texas. These statistics suggest that while large banks are quite important, small banks continue to play a prominent role in Texas. Consistent with our geographic-concentration hypothesis, these offices have not consolidated into large banking organizations. 4.6. Population, banking offices, and deposits within Texas In untabulated analysis, we find that the six largest Texas counties (in terms of population) have 47% of the population and 53% of the total commercial banking deposits within Texas. Using Metropolitan Statistical Areas (MSAs) rather than counties yields similar conclusions; the six largest MSAs have 61% of the state’s population and 62% of its commercial banking deposits. Approximately 16% of Texas’ population and 17% of its commercial banking deposits are outside any MSA.

5. Empirical results 5.1. Market characteristics and office ownership Table 4 presents selected summary statistics for large versus small banks and their corresponding market areas, which we define as the county of operation. Small banks own slightly more offices in the state (1,828) than large banks (1,720). The percentage of offices established after 1988 (in the less regulated environment) is slightly higher for large banks (43.8%) than small banks (38.9%). The median size of the large banks’ offices is larger than for the offices of small banks—approximately $30.9 million in deposits compared to $20.5 million. The market characteristics of these banks reported in Table 4 are quite consistent with our office-location hypothesis: although large banks locate some of their offices in smaller urban areas, they concentrate the majority of their offices in major 12

We have insufficient owner address information to estimate the proximity of large bank owners (proxy statements, from which we collected most of our large bank ownership data, normally does not provide owner address information). However, since many of the large banks are headquartered outside Texas (eight of the top ten)—it is likely that most of the large blockholders in these banks live outside Texas.

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

371

Table 4 Comparison of market characteristics for offices of large and small banks The sample consists of the 3,548 bank office locations in Texas with nonzero deposits as of 6/30/98 in Sheshunoff Information Services Inc.’s BranchSource(c) database. A large bank is a bank where the banking organization has more than $1 billion in total assets (including assets held by offices owned by the bank but outside Texas), and a small bank is a bank with less than or equal to one billion dollars in total assets. County-level data is obtained from the Texas Data Center (http://www.txsdc.tamu.-edu/tpepp/), and is 1998 data unless otherwise noted. Distances are calculated based on the 5-digit zip code of the banking office compared to all five-digit zip codes in the six largest Texas cities (Austin, Dallas, El Paso, Fort Worth, Houston and San Antonio).a The p-values, from a standard ANOVA t-test, test the null hypothesis that the values are the same between the two groups. Panel B reports the proportion (in percent) of each organization type’s offices grouped by county population. Panel A—Market characteristics across organization type Description Full sample

Small banks

Large banks

Number of commercial banking organizations Number of banking offices Percent of offices established 1989 and later Office’s total deposits ($millions, median) Percent of offices in an MSA Percent of offices in a major city County population (000s, median)

674 1,828 38.9 20.5 59.6 11.9 83.5

29 1,720 43.8 30.9 92.5 43.7 1,351.3

703 3,548 41.3 24.5 75.5 27.3 269.2

p-value

(o0.01) (o0.01) (o0.01) (o0.01) (o0.01)

Panel B—Frequency of offices by county population across organization type (in %) Small Large banks banks

Total

Population > 1:5 million 1:5 milliono ¼ Population > 750 thousand 750 thousando ¼ Population > 250 thousand 250 thousando ¼ Population > 100 thousand 100 thousando ¼ Population > 50 thousand 50 thousando ¼ Population > 25 thousand 25 thousando ¼ Population Total

24.8 10.4 15.8 17.0 7.1 10.3 14.7 100.0

a

13.0 5.0 11.1 20.1 8.8 16.7 25.3 100.0

37.2 16.2 20.9 13.7 5.2 3.4 3.4 100.0

Zip code distances are computed using the program ZipFind 2.0 from Bridger Systems, Inc.

metropolitan areas; small banks focus primarily on smaller urban and rural areas. In particular, 93% of the offices owned by large banks are located in MSAs compared to 60% for small banks. Only 12% of the small banks’ offices are in one of the major cities, yet 44% of the offices owned by large banks are located within major cities. The median office owned by a large bank is in a county with a population of 1,351,300, compared to 83,500 for small banks. In contrast to those owned by small banks, untabulated results indicate that the typical office owned by a large bank is in a county with more large employers, higher household income, a higher population growth rate, less land in farms, and a larger number of bank offices. All of the statistics in Panel A are significantly different between large and small banks.

ARTICLE IN PRESS 372

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

Table 4 Panel B reports the frequency of offices for each bank type classified by the population of the county. These numbers also are consistent with the office-location hypothesis. Large banks concentrate their offices within large metropolitan areas. The highest frequencies of small bank offices are in rural areas, and to a lesser extent, within small to mid-sized urban areas. While not reported in Table 4, we also compare the frequencies of large bank ownership in cities and suburbs. Large banks own approximately 78% of the offices within the city limits of the six major cities in Texas, and they own about 55% of the offices in the suburbs of these cities (the remaining part of the MSA). These proportions are significantly different at the 0.01 level. In untabulated analysis, we examine the office locations for the eight largest Texas banks in greater detail. These banks generally follow a similar pattern: The vast majority (85%) of their offices are located within 100 miles of one of the six major MSAs in Texas. For the typical large bank, all the offices established since 1989 are in one of Texas’ 27 MSAs. To test the office-location hypothesis more formally, Table 5 contains estimates of logistic models that predict the likelihood that an office is owned by a large bank given that it is in one of the following mutually exclusive and exhaustive location categories: (i) a major city, (ii) a suburb of a major city, (iii) a smaller MSA, or (iv) a rural location (defined as not within an MSA). Three models are presented, each estimated with an intercept and with dummy variables for the last three location categories; the intercept thus captures the impact of locating an office in a major city. In the second model, we add the natural logarithm of the county’s population, because county population varies within our location categories. Adding population is expected to add explanatory power to the model, since the theory suggests that the likelihood of being owned by a large bank increases with the population of the market area. In the third model, we add additional controls for the size of the office (as measured by the natural logarithm of deposits), the county’s population growth rate over the period from 1990 to 1998, and the Herfindahl-Hirschman Index (HHI) for the county based on total deposits. Our reported HHI is based on total deposits in the county, including deposits from non-bank depository institutions such as Savings & Loans and Credit Unions. For robustness, we also calculate HHI using only bank deposits. Both definitions yield similar results. We add these variables as controls since they are correlated with the key explanatory variables and potentially affect ownership patterns. For example, if small banks are inefficient relative to large bank, as some suggest (see footnote 4), they will be more likely to survive in less competitive markets (i.e., those with higher HHI’s). The results are consistent with our office-location hypothesis, which predicts the highest frequency of large bank ownership within the city, but less ownership in rural areas.13 In all specifications, the coefficients on the suburb dummy, the smaller MSA 13

We also estimated similar models where the key explanatory variable was the distance to a major city (and various definitions thereof) instead of dummy variables representing the type of market area in which the office is located. Results of this analysis yield the same conclusions as those reported above—the likelihood of being owned by a large bank decreases as an office’s distance to a major city increases.

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

373

Table 5 Logistic regression models of being owned by a large bank The sample consists of the 3,548 bank office locations in Texas with nonzero deposits as of 6/30/98 in Sheshunoff Information Services Inc.’s BranchSource(c) database. The dependent variable is equal to one if the banking office is owned by a large bank (defined as a bank that has more than one billion in total assets, including all assets held by offices owned by the bank but not located in Texas). The logistic regressions are estimated with three dummy variables, indicating whether an office is located in (i) a suburb of a major city, (ii) a smaller MSA, or (iii) a rural location, defined as not within an MSA. The intercept captures when an office is located in a major city (within the city limits of Austin, Dallas, El Paso, Fort Worth, Houston, or San Antonio). Control variables include the natural log of the county’s 1998 population, the county’s percentage population growth from 1990 to 1998, the natural log of the bank office’s total deposits, and a Herfindahl-Hirschman index (HHI) measure of market concentration for each county. HHI is defined as the county’s sum of squared deposit market shares for all depository institutions in Texas (banks, savings and loans, and credit unions). Asymptotic t-statistics are reported in brackets. Pseudo-R2 is ð1  Lð0Þ=LðbhatÞ2=nÞ where Lð0Þ is the likelihood of the intercept-only model, LðbhatÞ is the likelihood of the specified model, and n is the sample size (Cox and Snell, 1989, pp. 208–209). Difference tests comparing estimated coefficients are based on a Wald Chi-square statistic. Logistic regressions

Dependent variable ¼ 1 if office owned by a large bank

Intercept dummy (omitted category—major city) Major city suburb dummy Smaller MSA dummy Rural (non-MSA) dummy

1.24 [16.13] 1:05 ½10:40 1:58 ½14:78 2:99 ½24:36

In(Population)

4:84 ½8:19 0:54 ½4:76 0:61 ½4:35 1:09 ½5:03 0.42 [10.35]

County population growth 1990–1998 In(Deposits) Herfindahl-Hirschman index Pseudo-R2 Model pðw2 Þ

0.205 (0.00)

0.230 (0.00)

6:66 ½8:20 0:58 ½4:94 0:53 ½3:55 0:89 ½3:90 0.43 [8.64] 0.01 [2.97] 0.17 [5.91] 0:70 ½1:17 0.240 (0.00)

Difference tests of coefficients on suburb vs. smaller MSA and smaller MSA vs. rural dummies: Smaller MSA  major city suburb 0:53 0:07 0.05 Probðw2 Þ (0.00) (0.52) (0.70) Rural  smaller MSA 1:41 0:48 0:36 Probðw2 Þ (0.00) (0.00) (0.02)

dummy and the non-MSA (rural) dummy are negative and significant, as predicted. Difference tests comparing a smaller MSA to a rural location are significant at the 1% level in the first two specifications, and at the 5% level in the third specification. The difference between a smaller MSA and a major city suburb is significant only in the model with no control variables, suggesting that, once we control for population,

ARTICLE IN PRESS 374

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

moving from a major city to a suburb is similar to that of moving from a major city to a smaller MSA. Taken together, these results suggest that the likelihood that a large bank owns an office decreases as one moves from the city to smaller urban areas to rural areas, consistent with our predictions. The coefficient on the natural log of population also is positive and significant, consistent with the hypothesis that the likelihood of being owned by a large bank increases with the population of the county (market area). The models explain approximately 21–24% of the crosssectional variation in the dependent variable. To provide more intuition on the economic significance of the results in Table 5, we use model 1 to calculate the implied change in probability of being owned by a large bank as we move across the different market areas. The probability that a large bank owns an office decreases by 36 percentage points (from 78% to 42%) with a move from a major city to a smaller MSA, and by 27 percentage points with a move from a smaller MSA to a rural location. We also calculate the change in probability given a change in the population of the market area. Using Model 3, starting in the smaller MSA category and setting all other variables constant at their mean values, the probability that a large bank owns the office increases by eight percentage points (from 37% to 45%) as we move from the first to third population quartile within smaller MSAs. One concern about the estimates in Table 5 is the potential lack of independence of the error terms. In particular, the error terms for offices within a given county arguably are correlated. As a check, we use as the dependent variable the percentage of offices in each county that are owned by large banks ðN ¼ 247Þ: These estimated coefficients also are significant and have the predicted signs. 5.2. On the impact of prior regulation Texas banking data are extremely useful for testing hypotheses about the boundaries of the firm because of its large number of counties, its wide dispersion in demographic characteristics across counties, its large number of banking offices, and its substantial variation in horizontal and vertical integration across banks. However, because Texas regulation required unit banking prior to 1988, these data impose the potential disadvantage that aspects of our results might be driven by prior regulation. Note, however, that the existence of multi-bank holding companies should speed the process of adjustment to a change in regulation. Blackwell et al. (1994) report that by 1986, multi-bank holding companies controlled almost half of the more than 1,800 separately incorporated Texas banks and nearly three-quarters of the bank deposits in Texas. To examine whether aspects of our results are driven by prior regulation, we perform two additional tests. First, we replicate the analysis in Table 5 using the subsample of Texas banking offices that were established since 1989 - i.e., within the deregulated environment. The results (available from the authors) are similar to those for the full sample. Second, we replicate the analysis using June 30, 2000 data from another state—California—a state that never prohibited bank branching. Although California originally had no laws specifically permitting or prohibiting

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

375

branching, California permitted branch banking under the discretion of the Attorney General’s office in Sacramento beginning in the 1860s. The Banking Act of 1909 included an explicit provision to permit branching, subject to approval of the Superintendent of Banking, an office established within the Act. California has significantly less urban/rural diversity than Texas, which reduces the power of the tests. For example, approximately 5% of California’s population lives outside an MSA, compared to 16% for Texas. Nonetheless, the results for California, which are reported in Table 6, are significant and similar to those for Texas, although the significance levels are somewhat lower. Unlike Texas, our measure of market competition, the Herfindahl Index, is significantly positive when we restrict deposits just to bank deposits. Note that this significant positive sign is the opposite of that implied by the argument that small banks are inefficient and more likely to survive in a less competitive (more concentrated) environment. 5.3. Concentration of offices within small banks The geographic-concentration hypothesis implies that many small banks will remain independent, concentrating their individual offices within a given market area. We already have documented the existence of many small banks in Texas. Table 7 provides descriptive information on the geographic concentration of bank offices. Approximately 41% of the commercial banks in Texas are unit banks (those with only one physical office), another 51% are small banks with fewer than ten offices; and only 24 (3%) small banks have more than nine offices. Excluding unit banks, the median distance between a small bank’s headquarters and its branch offices is 21 miles and the median distance between these banks’ most distant office and their headquarters is 35 miles. Large banks’ offices are substantially more distant from their headquarters under both of these measures. The median small bank has 67% (100%) of its offices within 25 (50) miles of its headquarters and 67% within the same county. In contrast, the median large bank has 34% (40%) of its Texas offices within 25 (50) miles of its Texas headquarters, and has 50% of its Texas banking offices within one county. These descriptive facts suggest that the offices within both large and small banks are geographically concentrated. However, consistent with our geographic-concentration hypothesis, large banks appear to be less concentrated and more likely to locate in multiple market areas than smaller banks. To provide more formal evidence on the geographic concentration of these banks’ offices, we estimate a measure of spatial agglomeration for each independent branch bank similar to that proposed by Ellison and Glaeser (1997). This measure allows us to compare the relative degree of concentration of large and small banks in a manner that accounts for the fact that large banks tend to have significantly more offices than small banks.14 14

Ellison and Glaeser (1997) originally designed this measure to capture the combined effect of natural advantages and spillovers causing firms within an industry to locate plants within the same geographic region. For example, shipbuilding companies might concentrate plants along the coast because of the natural advantage of being near a waterway. There also may be spillover effects; for example, the positive

ARTICLE IN PRESS 376

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

Table 6 Logistic regression models of being owned by a large bank—California Bank Offices The sample consists of the 4,469 bank office locations in California with nonzero deposits as of 6/30/2000 in Sheshunoff Information Services Inc.’s BranchSource(c) database. The dependent variable is equal to one if the banking office is owned by a large bank (defined as a bank that has more than one billion in total assets, including all assets held by offices owned by the bank but not located in California). The logistic regressions are estimated with three dummy variables, indicating whether an office is located in (i) a suburb of a major city; (ii) a smaller MSA; or, (iii) a rural location, defined as not within an MSA. The intercept captures when an office is located in a major city (within the city limits of one of the 12 largest cities in California). Control variables include the natural log of the county’s 1999 population, the county’s percentage population growth from 1990 to 1999, the natural log of the bank office’s total deposits, and a Herfindahl-Hirschman index (HHI) measure of market concentration for each county. HHI is defined as the county’s sum of squared deposit market shares for all depository institutions in California (banks, savings and loans, and credit unions). Asymptotic t-statistics are reported in brackets. Pseudo-R2 is ð1  Lð0Þ=LðbhatÞ2=nÞ where Lð0Þ is the likelihood of the intercept-only model, LðbhatÞ is the likelihood of the specified model, and n is the sample size (Cox and Snell, l989, pp. 208–209). Difference tests comparing estimated coefficients are based on a Wald Chi-square statistic. Logistic regressions

Dependent variable ¼ 1 if office owned by a large bank

Intercept dummy (omitted category—major city) 1.54 [19.30] Major city suburb dummy 0:18 ½1:90 Smaller MSA dummy 0:66 ½6:40 Rural (non-MSA) dummy 1:43 ½9:37 In(Population)

1:23 ½2:48 0:18 ½1:87 0:43 ½3:83 0:74 ½3:79 0.19 [5.64]

County population growth 1990–1999 In(Deposits) Herfindahl-Hirschman index Pseudo-R2 model pðw2 Þ

0.026 (0.00)

0.033 (0.00)

6:67 ½8:26 0:16 ½1:55 0:31 ½2:59 0:59 ½2:92 0.22 [5.02] 0.02 [4.20] 0.43 [13.80] 1.76 [1.53] 0.078 (0.00)

Difference tests of coefficients on suburb vs. smaller MSA and smaller MSA vs. rural dummies: Smaller MSAmajor city suburb 0:48 0:25 0:15 Probðw2 Þ (0.00) (0.01) (0.13) Ruralsmaller MSA 0:77 0:31 0:28 Probðw2 Þ (0.00) (0.06) (0.11)

(footnote continued) spillover effect of concentrating multiple plants might induce suppliers to locate near the plants. As Ellison and Glaeser note, their measure also can be used to test whether there is spatial concentration among the plants or offices within a given firm, as opposed to an industry. See Ellickson (1999) for a specific application in the supermarket industry.

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

377

Table 7 Geographic concentration of Texas banks—descriptive statistics Panel A reports the number of banks and banking offices for Texas banks, and Panel B reports geographic concentration information for the banks in our sample that have more than one office. A small bank is defined as a bank with less than one billion dollars in total assets (including assets held by offices owned by the bank but outside Texas). In Panel B, the mean (median) distance represents the mean (median) distance from each bank’s Texas headquarters to all of its non-headquarters Texas banking offices. Distances are calculated based on the five-digit zip code of the banking office compared to the five-digit zip code of the bank’s Texas headquarters.a Panel A—Number of banks and offices Texas banks Description

Number of banks

%

Number of offices

%

Small unit banks Small non-unit banks 1o Number of offices o10 10p Number of offices Large banks

290

41.3

290

8.2

360

51.2

1,206

34.0

24

3.4

332

9.4

29

4.1

1,720

48.5

Totals

703

100.0

3,548

100.0

Panel B—Geographic concentration of Texas banking offices (small banks with more than 1 office) Small banks Description Distance to headquarters (HQ) office (miles) Distance to bank’s furthest office from HQ (miles) Pct of offices within 25 miles of HQ office (%) Pct of offices within 50 miles of HQ office (%) Pct of offices in county with largest num of offices (%)

Large banks

Mean

Median

Mean

Median

40.7 63.2 59.9 80.0 70.6

20.5 35.2 66.7 100.0 66.7

181.5b 433.8b 31.8b 38.5b 47.1b

180.6b 336.0b 33.5b 40.4b 50.0b

a

Zip code distances are computed using the program ZipFind 2.0 from Bridger Systems, Inc. The difference between large and small banks is significant at the 1% level. The difference in means is from an ANOVA F -test and the difference in medians is from a Kruskal-Wallis Chi-square test. b

As applied to Texas banks, the null hypothesis is that each bank chooses the locations of its offices as though it were making independent throws at a map weighted by total deposits in each county. Rejecting the null suggests that individual banks concentrate their offices either because there are natural advantages (for example, a bank focusing on oil loans will want to locate near its customers) or positive spillovers (for example, concentrating offices lowers the cost of on-site monitoring). Although we cannot separate the two effects, we expect both are important - incentive theory suggests that these banks normally would not want to

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

378

locate their offices in geographically dispersed areas, even if the customer bases were similar. We compute an Ellison and Glaeser ‘‘Gamma’’ for each bank using the following formula: PM g

i¼1

P PN 2 P 2 ðsi  xi Þ2  ð1  M G  ð1  i x2i ÞH ; i¼1 xi Þ j¼1 zj P  P PN 2 2 ð1  i x2i Þð1  HÞ ð1  M i¼1 xi Þð1  j¼1 zj Þ

ð1Þ

where si is the bank’s share of deposits in each county: i ¼ 1 to M counties; zj the office j’s share of the bank’s total deposits: j ¼ 1 to N offices; xi the county i’s share of total banking deposits in Texas; G the sum of the squared differences between the bank’s deposit concentration in county i and county i’s proportion of all Texas deposits; H the Herfindahl Index of the firm’s office-size distribution. The expected value of Gamma is zero under the null hypothesis. It will be positive if the bank’s offices are more highly concentrated than under the null, i.e., the higher the Gamma, the higher the concentration. Table 8 reports the average Gamma for small banks, large banks and the two subsamples combined. Gamma is not defined for banks with only one office. Our basic analysis focuses on banks with at least 10 branches. These multi-office banks are the ones that have the greatest potential to be dispersed geographically. However, our results are unaffected if we use 2, 5, or 20 branches. While a positive Gamma suggests high concentration, its magnitude is difficult to interpret. Ellison and Glaeser argue that 0.05 is economically material in their application at the industry level. Based on this benchmark, the Gammas in Table 8 appear quite large. The average Gamma for the 24 small banks is 0.42, and for the 21 large banks is 0.26; both are significantly different from zero at the 1% level. Moreover, the difference in these estimates is significant at the 5% level (one-tailed p-value of 0.03). The results of these tests reinforce the impression derived from the descriptive statistics—small banks concentrate their offices within selected market areas and generally are more concentrated than large banks. While the results suggest significant concentration within small banks, they simply might reflect the observation that small banks as a whole tend to concentrate in smaller urban areas, i.e., the observed concentration potentially reflects a small bank industry effect rather than a firm-specific effect. As an additional check, we reestimate Gamma, weighting by total deposits of all small banks with at least ten offices and all large banks with at least ten offices, instead of total deposits in Texas. Although the Gammas for each bank type continue to be significant and large, the difference between large and small is no longer significant. As a whole, our results appear to be consistent with the hypothesis that small banks concentrate their offices within a given market area. While the results are somewhat mixed, our data also suggest that small banks concentrate more than large banks.

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

379

Table 8 Spatial agglomeration of office locations for Texas banks The table reports summary information on the 45 banks in our sample that have at least ten Texas offices. A large bank is one that has more than $1 billion in total assets (including assets held by offices owned by the bank but outside Texas), and a small bank is one with less than $1 billion in total assets. The first Gamma column reports average Gammas when the industry, for purposes of calculating spatial agglomeration, is defined as all commercial banking deposits in Texas. The second Gamma column reports average Gammas when the industry is defined as only those deposits held by the respective organization type—for small banks, the industry is all deposits held by small banks; for large banks, the industry is all deposits held by large banks. Gamma, based on Ellison and Glaeser (1997), measures the geographic concentration of the firm’s offices. The p-values for comparing gamma between small and large banks are from a Kruskal-Wallis Chi-square test. Description

Full sample Small banks Large banks Difference (SmallLarge) p-value for small > large (one-tailed) a

Number of banks

45 24 21

Average number of offices

Industry¼Total Commercial Banking Deposits Average Gammaa

Industry ¼ Deposits Held by Respective Organization Type Average Gammaa

0.348b 0.422b 0.263b 0.159

0.321 0.336 0.304 0.032

45 14 80

(0.03)

(0.37)

Gamma is calculated as follows: PM 2 PN 2 PM P 2 G  ð1  i x2i ÞH i¼1 ðsi  xi Þ  ð1  i¼1 xi Þ j¼1 zj P 2  g PM 2 PN 2 ð1  i xi Þð1  HÞ ð1  i¼1 xi Þð1  j¼1 zj Þ

where xi is the county i’s share of total banking deposits in Texas; si the bank’s share of deposits in each county: i ¼ 1 to M counties; zj the branch j’s share of the bank’s total deposits, j ¼ 1 to N offices; G the sum of the squared differences between the bank’s deposit concentration in county i and county i’s proportion of all Texas deposits; H the Herfindahl index of the firm’s branch-size distribution. b Average Gamma significantly different from zero at a significance level less than 0.01 level, using a standard t-test and a Wilcoxon sign rank test.

6. Conclusions and discussion There are over 3,000 commercial bank offices in Texas. Large banks with dispersed ownership own 49% of these offices, while small banks with relatively concentrated ownership own the remaining 51%. Most of the offices are affiliated with other offices as part of a branch system, though 8% are organized as standalone unit banks. In this paper, we use incentive-related arguments to analyze ownership patterns within this industry. Our results are consistent with the hypotheses that bank offices are more likely to be owned locally (vertically separate) when it is more important to grant significant decision rights to the local office managers (for example, due to the location of the office, the customer base, and the

ARTICLE IN PRESS 380

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

distribution of knowledge). In contrast to much of the prior empirical work on the boundaries of the firm, our focus is on ownership incentives, rather than firm-specific physical assets. Although our study examines the boundaries of the firm, we also provide insights into two other questions of broad interest. First, while theorists since Hotelling (1929) have questioned how firms make locational choices, limited empirical evidence exists on this issue. Of particular interest is whether firms strive for maximum or minimum differentiation. Our evidence suggests that certain types of banks strive to differentiate themselves from other banks both through locational choice and customer focus. Large banks tend to concentrate primarily on urban areas, while smaller locally owned banks focus more on smaller urban and rural areas. Individual banks tend to concentrate geographically. Designing an organization to serve one type of customer or location limits the ability of that organization to serve other types of customers and locations. This promotes differentiation across organizations. Second, the Riegle–Neal Act of 1994 materially reduces constraints on interstate bank branching. Opponents of this legislation have argued that large interstate banks would eliminate small community banks. Our evidence supports the growing view that small banks have a comparative advantage over large banks within identifiable environments. Our evidence suggests that one advantage of small banks is the incentives provided to local managers through concentrated share ownership. Petersen and Rajan (2002) argue that improvements in computer and information technology have reduced the importance of local information in evaluating loans to small businesses and thus have reduced the advantage that small banks traditionally have had in small business lending. Their empirical evidence supports the hypothesis that geographic distance between banks and their loan customers has become less important over time. An interesting question is whether or not the patterns we have found in our data will continue to hold as computer and information technology continues to improve. Our conjecture is that, while large banks are likely to continue to increase the amount of business they do with distant customers, small banks will continue to own many of the bank offices in less urban areas. We base this forecast on three arguments: First, local ownership provides important incentives to provide quality service to customers. An alternative is to have an employee manage the office and be monitored by central bank personnel. This alternative, however, is likely to be more expensive in rural locations (Brickley and Dark (1987) use related arguments to explain why isolated units in franchised systems tend to be franchised rather than company-owned). Second, while improvements in technology reduce the costs of processing and transferring information, they are unlikely to eliminate the importance of local manager information for all banking transactions (see Berger et al. (2001) for a related discussion). Third, some customers appear to prefer to deal with a community bank rather than a branch of a large bank with a distant headquarters. Indeed, this point is often the focus of marketing by small banks. Such product differentiation has the potential to help offset any cost disadvantages from operating at a smaller scale. These arguments imply that small

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

381

banks will continue to play an important role in the economy in the foreseeable future.15

References Anderson, E., 1985. The salesperson as outside agent or employee: a transaction cost analysis. Marketing Science 4, 234–254. Anderson, E., Schmittlein, D., 1984. Integration of the sales force: an empirical examination. Rand Journal of Economics 15, 385–395. Baker, G., Hubbard, T., 2000. Contractibility and asset ownership: on-board computers and governance in U.S. trucking. Unpublished working paper, NBER. Barron, J., Umbeck, J., 1984. The effects of different contractual arrangements: the case of retail gasoline markets. Journal of Law and Economics 27, 313–328. Berger, A., Hannan, T., 1998. The efficiency cost of market power in the banking industry: a test of the ‘quiet life’ and related hypotheses. The Review of Economics and Statistics, 80, 454–465. Berger, A., Kashyap, A., Scalise, J., 1995. The transformation of the U.S. banking industry: What a long, strange trip it’s been. Brookings Papers on Economic Activity, Washington 2 (55), 164. Berger, A., Saunders, A., Scalise, J., Udell, G., 1998. The effects of bank mergers and acquisitions on small business lending. Journal of Financial Economics 50, 187–229. Berger, A., Klapper, L., Udell, G., 2001. The ability of banks to lend to informationally opaque small businesses. Journal of Banking and Finance 25, 2127–2167. Blackwell, D., Brickley, J., Weisbach, M., 1994. Accounting information and internal performance evaluation: evidence from Texas banks. Journal of Accounting and Economics 17, 331–358. Borenstein, S., Netz, J., 1999. Why do all flights leave at 8 am?: competition departure-time differentiation in airline markets. International Journal of Industrial Organization 17, 611–640. Brickley, J., 1999. Incentive conflicts and contractual restraints: evidence from franchising. Journal of Law and Economics 42, 745–774. Brickley, J., Dark, F., 1987. The choice of organizational form: the case of franchising. Journal of Financial Economics 18, 401–420. Calem, P., 1994. The impact of geographic deregulation on small banks. Federal Reserve Bank of Philadelphia Business Review Nov–Dec, 17–31. Cline, K., 1996. NationsBank branch sale illustrates trend of big players fleeing from small markets. American Banker. Coase, R., 1937. The nature of the firm. Economica 4, 386–405. Cole, R., Goldberg, L., White, L., 1999. Cookie-cutter versus character: the micro structure of small business lending by large and small banks. In: L. Blanton, A. Williams, S.L.W. Rhine (Eds.), Business Access to Capital and Credit: A Federal Reserve System Research Conference, Federal Reserve Bank of Chicago. pp. 362–389. Crocker, K., Masten, S., 1991. Pretia ex machina? Prices and process in long-term contracts. Journal of Law and Economics 34, 69–100. Crocker, K., Reynolds, K., 1993. The efficiency of incomplete contracts: an empirical analysis of air force engine procurement. The Rand Journal of Economics 24, 126–146. Ellickson, P., 1999. Vertical product differentiation and concentration in the supermarket industry. Unpublished working paper, University of Rochester. Rochester, NY. 15

In the three-year period from our sample year of 1998–2001 restructuring of the Texas banking industry continued to occur with bank openings, closures, and mergers. Nevertheless, there was almost no net change in the number of offices owned by small banks as reported in the Sheshunoff database. In 1998, Shesunoff reported 1,828 offices that were owned by small banks, compared to 1,826 in 2001. The percentage of bank offices owned by small banks declined from 51% to 47% over the period because of an increase in total banking offices in the state.

ARTICLE IN PRESS 382

J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

Ellison, G., Glaeser, E., 1997. Geographic concentration in U.S. manufacturing industries: a dartboard approach. Journal of Political Economy 105, 889–927. Fama, E., Jensen, M., 1983. Separation of ownership and control. Journal of Law and Economics 26, 301–325. Gilbert, R., Belongia, M., 1988. The effects of affiliation with large holding companies on commercial bank lending to agriculture. American Journal of Agricultural Economics 70, 69–78. Goldberg, V., Erickson, J., 1987. Quantity and price adjustment in long-term contracts: a case study of petroleum coke. Journal of Law and Economics 30, 369–398. Grossman, S., Hart, O., 1986. The costs and benefits of ownership: a theory of lateral and vertical integration. Journal of Political Economy 94, 691–719. Hart, O., 1995. Firms, Contracts, and Financial Structure. Clarendon Press, Oxford. Hart, O., Moore, J., 1990. Property rights and the nature of the firm. Journal of Political Economy 98, 1119–1158. . Holmstrom, B., Milgrom, P., 1991. Multitask principal-agent analysis: incentive contracts, asset ownership, and job design. Journal of Law Economics and Organization 7, 24–52. Holmstr.om, B., Milgrom, P., 1994. The firm as an incentive system. American Economic Review 84, 972–991. Hotelling, H., 1929. Stability and competition. The Economic Journal 39, 41–57. John, G., Weitz, B., 1988. Forward integration into distribution: an empirical test of transaction cost analysis. Journal of Law, Economics and Organization 4, 337–355. Joskow, P., 1988. Asset specificity and the structure of vertical relationships: empirical evidence. Journal of Law, Economics and Organization 4, 95–117. Klein, B., Saft, L., 1985. The law and economics of franchise tying contracts. Journal of Law and Economics 28, 345–361. Klein, B., Crawford, R., Alchian, A., 1978. Vertical integration, appropriable rents, and the competitive contracting process. Journal of Law and Economics 21, 297–326. LaFontaine, F., 1992. Agency theory and franchising: some empirical results. Rand Journal of Economics 23, 263–283. LaFontaine, F., Slade, M., 1997. Retail contracting: theory and practice. Journal of Industrial Economics 45, 1–25. Lawrence, D., Klugman, M., 1991. Interstate banking in rural markets: the evidence from the corn belt. Journal of Banking & Finance 15, 1081–1090. Lamoreaux, N., 1994. Insider lending: Banks, personal connections, and economic development in industrial New England. Cambridge University Press, Cambridge. Martin, R., 1988. Franchising and risk management. American Economic Review 78, 954–968. Masten, S., 1984. The organization of production: evidence from the aerospace industry. Journal of Law and Economics 27, 403–417. Mazzeo, M., 2001. Competitive outcomes in product-differentiated oligopoly. Unpublished working paper. Kellogg School of Management, Northwestern University. Mensheha, M., 1996. Bank of America has inked deals to sell 63 rural branches. San Antonio Business Journal, September 23, http://sanantonio.bizjournals.com/sanantonio/stories/1996/09/23/story4.html. Milgrom, P., 1988. Employment contracts, influence activities, and efficient organization. Journal of Political Economy 96, 42–60. Minkler, A., 1990. An empirical analysis of a firm’s decision to franchise. Economics Letters 34, 77–82. Monteverde, K., Teece, D., 1982. Supplier switching costs and vertical integration in the automobile industry. Bell Journal of Economics 13, 206–213. Muris, T., Scheffman, D., Spiller, P., 1992. Strategy and transaction costs: the organization of distribution in the carbonated soft drink industry. Journal of Economics & Management Strategy 1, 83–128. Nakamura, L., 1994. Small borrowers and the survival of the small bank: is mouse bank Mighty or Mickey? Federal Reserve Bank of Philadelphia Business Review Nov–Dec, 3–15. Netz, J., Taylor, B., 2002. Maximum or minimum differentiation? Locational patterns of retail outlets. Review of Economics and Statistics 84 (1), 162–175. Norton, S., 1988. An empirical look at franchising as an organizational form. Journal of Business 61, 197–217. O’Hara, T., 1997. Branch sales creating a growth opportunity in Upstate New York. American Banker. Petersen, M., Rajan, R., 2002. Does distance still matter? The information revolution in small business lending. The Journal of Finance 57 (6), 2533–2570.

ARTICLE IN PRESS J.A. Brickley et al. / Journal of Financial Economics 70 (2003) 351–383

383

Pinske, J., Slade, M., 1998. Contracting in space: an application of spatial statistics to discrete-choice models. Journal of Econometrics 85, 125–154. Rose, P., 1997. Banking across State Lines: Public and Private Consequences. Quorum, Westport. Salvanes, K., Steen, F., Sorgard, L., 1997. Hotelling in the air? flight departures in Norway. Unpublished working paper. Norwegian School of Economics and Business Administration. Scott, F., 1995. Franchising vs. company ownership as a decision variable of the firm. Review of Industrial Organization 10, 69–81. Shepard, A., 1993. Contractual form, retail price, and asset characteristics. Rand Journal of Economics 24, 58–77. Slade, M., 1996. Multitask agency and contract choice: an empirical assessment. International Economic Review 37, 465–486. Slade, M., 1998. Strategic motives for vertical separation: evidence from gasoline. Journal of Law, Economics, and Organization 14 (1), 84–113. Smith, C., Stulz, R., 1985. The determinants of firms’ hedging policies. The Journal of Financial and Quantitative Analysis 28, 391–405. Stavins, J., 1995. Model entry and exit in a differentiated-product industry: the personal computer market. Review of Economics and Statistics 77, 571–584. U.S. House of Representatives, 1991. The effect of interstate branching on national, state and local economies. Hearing before the Subcommittee on Economic Stabilization, Committee on Banking, Finance, and Urban Affairs (May 15). U.S. House of Representatives, 1993. Interstate banking and branching. Hearing before the Subcommittee on Financial Institutions Supervision, Regulation, and Deposit Insurance, Committee on Banking, Finance, and Urban Affairs (July 29). Williamson, O., 1975. Markets and Hierarchies: Analysis and Antitrust Implications. The Free Press, New York. Williamson, O., 1985. The Economic Institutions of Capitalism. The Free Press, New York.