The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Vol. 25, No. 4 ECONOMIC REVIEW 19 8 9 Quarter 4 Deposit-lnstitution Failures: A Review 2 of Empirical Literature b y A s li D e m irg u c-K u n t Settlement Delays and Stock Prices 19 b y R am on P. D eG ennaro The Effect of Bank Structure and Profitability on Firm Openings b y Paul W . B auer and Brian A . C rom w e ll FEDERAL RESERVE BANK OF CLEVELAND 29 E C O N O M I C R E V I E W 1989 Quarter 4 Vol. 25 , No. 4 2 Deposit-lnstitution Failures: A Review of Empirical Literature Economic Review b y A s li D e m irg u c-K u n t q ua rterly b y the Research is published D e p a rtm e n t of the Federal R e s e rv e B a n k o f C le ve la n d . Turbulence in the U .S . banking and financial system in the 1980s has led to a major government bailout and impending reform of the financial industry. Current literature on the failure of deposit institutions does not seem ade quate to engender complete understanding of the problem. This paper C opie s o f th e Review are availa ble through our Public A ffa irs an d B an k R elations D e p a rtm e n t, 2 1 6 / 5 7 9 -2 1 5 7 . reviews previous studies, giving particular emphasis to the various definitions of insolvent and failed institutions. The paper concludes with recommenda tions to include the regulatory decision-making process into future research. C oordin ating Ec o n o m is t: Randall W . Eb erts 19 Settlement Delays and Stock Prices Ed ito rs : Paul J . N ick e ls Robin Ratliff D e s ig n : M ich ae l G alk a b y R am on P. D eG ennaro T y p o g ra p h y : L iz H a n n a In stock trades made for regular delivery, the buyer need not make payment Economic until the securities are delivered, typically for five business days. No tests O pin ion s s ta te d in have demonstrated whether investors consider the length of this delay and Review the opportunity cost associated with it. The author studies this issue by au th ors an d n ot necessarily modeling stock prices as a function of the federal funds rate during the set those of the Federal R e s e rve are th o s e o f the tlement delay and by conducting regression tests to determine if this variable B an k of C le ve la n d or o f the helps to explain the observed return. He concludes that investors do incorpo Board o f G ove rn o rs o f the F e d rate the effects of the settlement delay into the stock price. eral R e s e rve S y s te m . The Effect of Bank ........ 29 Structure and Profitability on Firm Openings M aterial m a y be reprinted prov id e d th a t the s ource is credited. P iea se send copies of reprjn ted m aterial to the editor. b y Paul W . Bauer and Brian A . C rom w e ll IS S N 0 0 13 -0 2 8 1 A n often-overlooked determinant of firm openings in empirical studies is the price and availability of credit from commercial banks. This study finds that profitable and competitive banking markets are associated with higher rates of firm births in metropolitan areas. These results support the position that bank structure and profitability influence economic development. D eposit-lnstitution Failures: A Review of Em pirical Literature by Asli DernirgilC-Kunt Asli Demirguc-Kunt is an economist at the World Bank in Washington, D.C . This paper was written while she was a visiting scholar at the Federal Reserve Bank of Cleveland, 1988-1989. The author would like to thank Edward Kane, Huston McCulloch, and Jam es Thomson for helpful comments and discussion. Introduction The decade of the 1 9 8 0 s has been a particularly turbulent one for the U.S. banking and financial system. Since the establishment of the Federal Deposit Insurance Corporation (FDIC) in 1933, more than 1,500 banks have been declared offi cially insolvent and were subsequently closed, acquired, or received assistance to prevent closure (see table 1). More than 800 of these closures took place during the 1 9 8 0 s, with 2 0 0 institu tions being closed in 1988 alone. De facto failures, which are defined more broadly to include any regulator-induced cessa tion of autonomous operations, portray an even gloomier picture. This dramatic increase in the bank failure rate has intensified public criticism of deposit-institution regulators, since bank safety and soundness is a major regulatory re sponsibility. 1 The recent crisis in the savings and loan industry7 helped the already existing problem to surface, and the public has become more eager to assess and assign blame. ■ 1 For a thorough discussion of safe and sound banking, see Benston et al. (1986). Deposit institutions fail primarily because they take risks, and subsequent events do not always turn out favorably. However, as Kane (1985) notes, when a series of failures occurs, or a major crisis is threatened, the general public blames regulators as much as it blames deposit-institution managers. Regulators are criticized for not being able to detect and curb different forms of unsuc cessful risk-taking in time to prevent failures. Potentially adverse consequences of bank fail ures include financial losses to bank stockholders and creditors, disruptions of community banking arrangements, contagious losses of confidence in other institutions, and widespread financial dis tress caused by sharp contractions in the money supply (Benston et al. [1986] and Kaufman [1985]). However, the consequences of an indi vidual bank failure on the local economy are unlikely to be any more severe than those of the failure of any other firm of comparable size (Horvitz [1965], Tussing [1967], Kaufman [1985]). Even the commonly feared financial distress thought to result from multiple bank failures is unlikely to occur. Destruction of the means of payment is an indication that govern ment has not fulfilled its macroeconomic respon sibility. Under such circumstances, sensible monetary policy would call for an expansion of the monetary7 base. It is an established view that bank failures that produce a decline in the T A B L E U .S . Bank Closures For Various Subperiods, 19 3 4 -19 8 8 Average Number of Closings per Year Average Deposits in Closed Banks (Millions) All Banks Insured Banks 1934-40 64.2 51.1 1941-50 7.3 6 .1 1951-60 4.3 2 .8 11.5 10.5 1961-70 6.3 5.0 34.2 33.5 Years All Banks Insured Banks 6 8 .2 62.3 9.9 10.3 1971-80 8.3 7.9 537.2 529.1 1981-85 59.8 59.8 6,023.4 6,023.4 19 8 6 138 138 6,471.1 6,471.1 1987 184 184 6,281.5 6,281.5 1988 20 0 20 0 37,200 37,200 1989a 145 145 21,400 21,400 a. As of August 18, 1989. SOURCE: 1987 FDIC Annual Report and telephone calls to FDIC. T A B L E 2 Equity, Insolvency, and Failure Detinitions Federally Contributed Equity = the capitalized value of the deposit-insurance guarantees. Enterprise-Contributed Equity = the capital of the institution net of the federally con tributed equity, Book-Value Insolvency the book value of assets minus the book value of liabilities (book value of the net worth) is negative, Market-Value Insolvency Economic Insolvency De Facto Insolvency market value of assets minus market value of lia bilities net of the value of insurance guarantees ( enterprise-contributed equity) is negative, Official (D ejure) Insolvency Closure Dejure Failure capital is judged inade quate by the regulators and the institution is closed or merged out of existence, De Facto Failure any regulator-induced ces sation of autonomous operations. money supply are the result of errors and mis conceptions by central bankers (Thornton [1939], Friedman and Schwartz [1963], Brunner and Meltzer [1964], Cagan [1965]). The consequences of contagious bank failures are no longer considered serious concerns because of the Federal Reserve System’s macroeconomic responsibilities. Yet the failure of indi vidual institutions still remains a serious prob lem for the general taxpayer. As Kane (1985, 1 9 8 9 ) notes, in a crisis, taxpayers are called upon to underwrite the cost to the Treasury of bailing out these institutions. The burden eventually falls on them in the form of higher taxes or higher rates of inflation.2 The problem for tax payers is to minimize their own loss exposure. By developing an accurate model for predicting bank failures, and by understanding the behavior of bank regulators, it will be possible to identify and/or verify the changes necessary to reform the deposit insurance system, thus minimizing the future loss exposure of the U.S. taxpayer. 3 The purpose of this article is to review empiri cal literature on deposit-institution failures. Sec tion I introduces and discusses concepts crucial in the analysis. Section II compares and contrasts selected empirical studies. Section III identifies weaknesses in the various approaches to study ing the problem and concludes by suggesting future avenues for research. I. Bank Insolvency, Closure, and Failures: Explaining Regulatory Decision-Making The purpose of this section is twofold.4 First, it seeks to define and distinguish between the dif ferent insolvency and failure categories listed in table 2. Second, based on the distinction between insolvency and failure, it describes how failure should be modeled within the framework of a regulatory decision-making process. ■ 2 This fact is exemplified by the recent savings and loan bailout. ■ 3 The problems in the present deposit-insurance system and regulator behavior have been identified by Meltzer (1967), Scott and Mayer (19 71), Merton (19 77, 1978), Kareken and Wallace (1978), Sharpe (1978), Buser, Chen, and Kane (1981), Kane (1981a, 1981b, 198 5,198 6 , 1988, and 1989), McCulloch (1981, 1987), Kareken (1983), Pyle (1983, 1984), and Benston et al. (1986). SOURCE: Author. ■ 4 The definitions and theoretical analysis presented in this section draw largely on Benston et al. (1986) and Kane (1985, 1989). Insolvency Versus Failure Official insolvency occurs when an institution’s chartering authority judges its capital to be inadequate. The procedures by which this deci sion is made are not clear, however. A firm’s capital may be identified as a particu lar measure of its net worth. Net worth is the dif ference between the value of the firm’s assets and nonownership liabilities. In order to deter mine the level of capital, itemization of assets and liabilities and adoption of an appropriate valuation rule are necessary (Kane [1989]). To be able to define capital, various categories of assets and liabilities need to be itemized. A complete definition requires recognition of implicit assets and liabilities as well as explicit ones. Implicit assets and liabilities are defined as all sources of positive and negative future cash flows that are considered “unbookable” by the accounting profession. Valuation of capital is crucial. Using different valuation rules leads to different asset and liabil ity values. Measuring an institution’s capital on the basis of historical cost at which it acquired its various balance-sheet positions is misleading. But historical-cost principles provide the basis for determining the book values of the balance sheet accounts of U.S. banks. Book values are recorded in terms of acquisition costs. As market prices change, these costs tend to depart from market values. Kane (1989) notes two shortcomings of historical-cost accounting. First, using acquisition cost undervalues an institution’s best portfolio decisions and overvalues its worst ones. Second, historical-cost accounting neglects potentially observable changes in the value of a firm’s investments by not modifying the acquisition costs to reflect market developments. This method exaggerates the economic relevance of the acquisition costs of an institution’s assets and liabilities and fails to appraise its investment suc cesses and failures on an ongoing basis. To determine a depository institution’s level of capital for regulatory purposes, it is helpful to break down its capital into two components: enterprise-contributed equity and federally con tributed equity (Kane [1989] )• Enterprisecontributed equity is the capital of the institution net of the capitalized value of its deposit insur ance guarantees. To the extent that federal guar antees are underpriced, the deposit insurer con tributes de facto capital to the institutions. The present deposit insurance system allows aggres sive deposit institutions to pass off poorly moni tored and unpriced risks onto federal insurance agencies. 5 The federally contributed capital is determined by the amount of risk that insurance agencies stand ready to absorb. These valuable guarantees are actually equity instruments that make the U.S. government a de facto investor in deposit institutions. Unless an appropriate recapitalization rule is imposed on managers and stockholders, the capitalized value of the guarantees increases as the institution’s enterprise-contributed equity decreases or as the riskiness of either its portfolio or environment increases. Clearly, the value of the federally con tributed capital should not be counted as a part of the institution’s capital for regulatory purposes. The traditional supervisory approach to regula tion also neglects the role of subordinated debt as a potential source of market discipline, and views debt capital as less desirable than equity. However, permitting institutions to count subor dinated debt toward capital-adequacy determina tions would provide increased protection for the insurance fund in the form of increased market discipline (Benston et al. [1986]). Holders of subordinated debt are a source of market discipline because, as opposed to depos itor debtholders, they cannot withdraw their funds on demand. Also, as opposed to stock holders, they do not share the increased profits that increased risk-taking may bring. Therefore, they prefer safe and conservatively managed insti tutions. If banks were required to maintain rela tively short-term subordinated debt as a certain proportion of equity, thus forcing them into the market on a frequent basis, subordinated debt could protect the insurance agency from losses. An appropriate insolvency criterion is the market value of enterprise-contributed capital, which can be obtained by subtracting the value of federal guarantees from the institution’s market value of equity. 6 De facto or market-value insolvency exists when an institution can no longer meet its con tractual obligations out of its own resources. This occurs whenever the market value of the institu tion’s nonownership liabilities exceeds the market value of its assets; or, in other words, ■ 5 ■ For a thorough review of this issue, see references in footnote 3. 6 An estimate of the capitalized value of the federal guarantees can be obtained using different approaches. For a review of different techniques, see Merton (19 77), Marcus and Shaked (1984), Ronn and Verma (1986), Kane and Foster (1986), Benston et al. (1986), Schwartz and Van Order (1988), and Demirguc-Kunt (1990, forthcoming). when the market value of its enterprisecontributed equity becomes negative. However, in determining official insolvency, regulators tend to look for book-value insolvency rather than market-value insolvency. Book-value insolvency exists when the differ ence between the book values of an institution’s assets and liabilities is negative. Even when an institution is book-value solvent, its market-value or economic insolvency may be suggested by refinancing difficulties that surface as an ongoing liquidity shortage. A liquidity shortage occurs whenever an institution’s cash, reserve balances, and established lines of credit prove insufficient to accommodate an unanticipated imbalance in the inflow and outflow of customer funds. If a continuing liquidity shortage is not relieved by outside borrowing or government assistance, assets may have to be sold at “fire-sale prices,” that is, for less than their equilibrium value. Such sales erode the institution’s capital, and may cause the uninsured customers of the institution to move their funds to safer locations. The resulting run on the institution's resources causes the insti tution to borrow nondeposit funds or to sell earning assets. Given that these runs are typically motivated by the presence of large unbooked losses in an institution’s balance sheet, asset sales push the book value of the institution’s assets toward their market value, eventually resulting in the institution’s book-value insolvency. Official (de jure) insolvency, or closure (de jure failure), occurs when the market-value insolvency is officially recognized and the firm is closed or involuntarily merged out of existence. De facto failure can be defined more broadly than closure as any regulator-induced cessation of autonomous operations. The definitions in this section clarify the dif ference between economic insolvency and fail ure of financial institutions. Economic insolvency is a market-determined event. In contrast, de jure or de facto failure results from a conscious deci sion by regulatory authorities to acknowledge and to repair the weakened financial condition of the institution. Failure is an administrative option that the authorities may or may not choose to exercise even when strong evidence of market-value insolvency exists. lyzes the working of government by applying and extending economic theory to the realm of political or governmental decision-making.7 Myers and Majluf (1984), Narayanan (1985), and Campbell and Marino (1988) apply public choice theory to explain the managerial decision making of an enterprise. Again, based on the public choice theory, Kane (1988 and 1989) develops a model of regulatory decision-making. The Kane model incorporates the economic, political, and bureaucratic constraints as well as the career-oriented incentives of federal regula tors in explaining the regulatory decision-making process. These constraints and incentives foster the difference between market-value insolvency and failure of financial institutions. Due to con flicts of interest between politicians and regula tors, and between regulators and taxpayers, timely resolution of market-value insolvencies is often not attractive to deposit-institution regulators. Kane (1989) argues that this conflict of inter est between regulators and politicians compli cates the regulatory task of serving the taxpayer. Deposit-institution regulators find it difficult to resist budget constraints imposed by politicians because they are subject to appointment and oversight controls from politicians. As appointed officials, they face political pressures to leave problems unsolved, thus keeping involved con stituencies and political action committees will ing to pay tribute to politicians. Regulators also face oversight controls from their regulatory clientele, that is, from the institu tions in the industry they regulate (Stigler [1977]). Federal officials have career-oriented incentives to keep their constituencies and clien tele happy. Their explicit salaries are lower than what they can make in the private sector. Econ omists conceive this gap as being bridged by implicit wages. As Kane (1989) notes, these implicit wages consist of certain nonpecuniary benefits of holding a high government office and of future increases in wages that accrue in post government employment—very often within the regulated industry. The actions and policy decisions of regulators are closely overseen by their clientele. If regula tors can successfully complete their term in government service, they can generally expect higher wages in postgovernment employment. The importance of the perceived quality of their Failure as a Regulatory Decision Economic theory can explain why deferring meaningful action can be the rational choice for federal officials. The theory of public choice ana ■ 7 See Buchanan (1960, 1967), Tulloch (1965), Niskanen (19 71), Stigler (19 77), and Buchanan and Tollison (1984). performance makes federal officials very sensi tive to the opinions of the institutions they regu late, as well as to those of the trade associations connected with these institutions. These career-oriented incentives introduce political and bureaucratic constraints to regulatory decision-making. Therefore, federal regulators tend to be influenced by their constituencies, avoiding solutions unfavorable to them, or promoting solutions that they find particularly desirable. Lobbying activities exaggerate and make the negative early effects of public policies more visible, further slowing the adoption of substantial changes in financial regulation. For regulators, the economic, political, and bureau cratic constraints increase the career costs of serving the taxpayer well. This conflict of interest between the regulators and the taxpayers leads to the adoption of forbearance policies that allow the continued operation of market-value insolvent institutions. In his model, Kane (1988 and 1989) envisions two extreme types of regulators: the unconflicted or faithful agent of the taxpayer, and the con flicted or self-interested agent. A faithful agent is expected to work toward fulfillment of society’s long-term goals. In the Kane model, faithful agents are modeled as max imizing the unobservable market value of the deposit-insurance enterprise. This value is calcu lated as the net present value of the future cash flows generated by its operations. A faithful agent protects the interests of the taxpayer, resisting politically imposed restraints and careeroriented incentives. Self-interested agents do not resist economic constraints to avoid the possibility of conflict with politicians. In addition, they are tempted by career-oriented incentives and serve their own narrow interests rather than those of the tax payer. In the Kane model, conflicted agents max imize their own perceived performance image in an effort to maximize their postgovernment wages. The self-interested agent’s decision making process is subject to economic constraints implicit in the budget procedures, as well as to the political and bureaucratic constraints implicit in career-oriented incentives. The agent, in an effort to serve himself well, gives in to all of these constraints and incentives, and imposes the resulting costs on the unwary taxpayer. The Kane model is a theoretical model of reg ulatory decision-making that underlines the fac tors leading to the distinction between eco nomic insolvency and failure of financial institutions. Clearly, in a realistic analysis, bank failures need to be modeled within the frame work of a regulatory decision-making process. II. Review Of Empirical Literature On FinancialInstitution Failures A summary of selected empirical studies on thrift-institution and commercial-bank failures is given in table 3. The first group of studies (Sin key [1975], Altman [1977], and Martin [1977]) focuses on developing early warning systems. These systems statistically analyze financial ratios constructed from the balance sheets and income statements that institutions file regularly with federal agencies. The goal is to incorporate this information into monitoring systems and to help regulators by flagging financially troubled institu tions as early as possible. To identify these insti tutions, researchers typically fit cross-sectional models for each year into their sample periods. The second group of studies (Avery and Han weck [1984], Barth et al. [1985], Benston [1985], and Gajewski [1988]) attempts to explain statis tically de jure failures, labeled in this article as the closure process. Their models seek to identify financial factors that affect the likelihood of an institution’s closure. Using cross-sectional data over a given sample period or cross-sectional data pooled from different years, researchers try to pinpoint determinants of closure by analyzing the same types of financial ratios used by the first group of studies. To clarify the model specifications of earlier researchers, it is helpful to review briefly the regulatory supervision process. Bank Supervision and Examination Supervision refers to the oversight of banking organizations and their activities to ensure that they are operated in a safe and sound manner. Examination is a means by which supervisors obtain information on the financial condition of an institution (Benston et al. [1986]). Examina tion is an important part of the supervisory proc ess. Through periodic examinations and contin uous supervision, regulators try to prevent deposit institutions from taking excessive risks that could lead them to economic insolvency. The supervision and examination of depository institutions are performed by one or more of the following institutions: The Federal Reserve System, state and federal chartering agencies, and federal deposit-insurance agencies. The Office of the Comptroller of Currency (OCC) and the Federal Home Loan Bank Board (FHLBB, now the Office of Thrift Supervision) charter national banks and savings and loan institutions, respectively. State D T A B L E 3 A Summary of Selected Empirical Studies on Deposit-institution Failures Author Estimation Technique Institutions and Time Period Dependent Variable Ratioa Sinkey (1975) 110 Problem 110 Nonproblem Commercial Banks (1969-1972) Discriminant Analysis Problem/ Nonproblem Over 100 are tested, 1 0 are chosen, 6 are significant. Altman (1977) 56 Serious Problem/49 Temporary Problem/107 No Problem Savings and Loans (1966-1973) Discriminant Analysis Serious Problem/ Temporary Problem/ No Problem 32/7 Martin (1977) 58 Closed/ 5,642 Nonclosed Commercial Banks (1970-1976) Logit Closed/ Nonclosed 25/4 Avery and Hanweck (1984) 100 Closed/ 1,190 Nonclosed Commercial Banks (12/1978-6/1983) Logit Closed/ Nonclosed 9/7b Barth et al. (1985) 318 Closed/ 588 Nonclosed Savings and Loans (12/1981-6/1984) Logit Closed/ Nonclosed 12/5 Benston (1985) 178 Closed/ 712 Nonclosed Savings and Loans (1981-1985) Logit Closed/ Nonclosed 28/4 Gajewski (1988) 134 Closed/ 2,747 Nonclosed Commercial Banks (1984-1986) Two-Step Logit Closed/ Nonclosed 25/10 a. The ratio of the total number of independent variables screened to significant independent variables. b. Two are significant but have unexpected signs. NOTE: Significant independent variable definitions are given in table 4. SOURCE: See text. banking commissions charter institutions with state charters. The deposit insurance agency for banks is the Federal Deposit Insurance Corpora tion (FDIC) and for savings and loan institutions it is the Federal Savings and Loan Insurance Cor poration (FSLIC), now changed to the Savings Association Insurance Fund by the 1989 Finan cial Institutions Reform, Recovery, and Enforce ment (FIRRE) Act. The 1989 FIRRE Act restructures the savings and loan industry. Under the new law, what was formerly the Federal Home Loan Bank Board is divided into three parts: the Office of Thrift Supervision (OTS), the Savings Association Insu rance Fund (SAIF), and the Federal Housing Finance Board. The Office of Thrift Supervision is responsible for the examination and supervi sion of savings and loans, and has the powers formerly vested in the FHLBB. The Savings Asso ciation Insurance Fund takes the place of FSLIC. In addition, a new Bank Insurance Fund is created. Both the Savings Association Insurance Fund and the Bank Insurance Fund are FDIC agencies. The obligations issued by either fund are backed by the full faith and credit of the United States. A five-member Federal Housing Finance Board is established to oversee credit allocation by the 12 district Home Loan Banks to members in the form of advances. The five mem bers are the secretary of the Department of Hous ing and Urban Development and four others appointed by the president with the advice and consent of the U.S. Senate. In addition, a new agency, the Resolution Trust Corporation (RTC), is created to oversee the liquidation of assets from insolvent thrifts.8 The FDIC is the day-today manager of the RTC. The new law restruc tures the financial institution industry, dismantles the independent Federal Home Loan Bank Sys tem, and gives the FDIC expanded powers. Besides expanding the FDIC’s regulatory7 turf and power, the new law does not substantially alter commercial bank supervision. National banks may be supervised by the Federal Reserve Board, the OCC, and the FDIC. However, unless the banks require assistance from the FDIC or the Federal Reserve, only the OCC supervises national banks. State-chartered banks are exam ined and supervised by the Federal Reserve if they are members of the Federal Reserve System, and by the FDIC if they are nonmembers. Statechartered banks can also be examined by their state banking supervisors, with or without the federal examiners. The Federal Reserve is also responsible for regulating, supervising, and inspecting bank holding companies. Additionally, the states can regulate and supervise holding companies. Fed erally chartered savings and loan institutions are examined and supervised by the FHLBB (now by the OTS). State-chartered savings and loan insti tutions are examined and supervised by their state examiners and the FSLIC (now by the SAIF). ■ 8 See Kane (1989) for an analysis of the savings and loan crisis. Federal examining efforts for banks are coor dinated in such a way that an institution is visited by only one examination team from either the Federal Reserve, the OCC, or the FDIC. Federal and state examiners also coordinate their exami nation schedules and make an effort to conduct joint examinations. If the examinations are con ducted separately, federal and state examiners share information by sending each other copies of their examination reports. Regulators use on-site and off-site methods in order to obtain information about the economic condition of the institutions. Traditionally, regulators have focused their monitoring efforts on sending teams of field examiners to conduct on-site examinations of each institution. On-site examinations are still heavily relied upon in regulatory monitoring efforts. States require exams every 12 to 18 months for their state-chartered institutions. In theory, sound national banks with assets of $300 million and above are supposed to be examined every 1 2 months; smaller banks are examined every 18 months. However, in practice, these schedules are often not met, and federal regula tors tend to concentrate on large institutions, those showing problems on their call reports, and those with low ratings on past examinations, in deciding how to allocate the limited time of their examiners. Principles and standards for federal examina tions are coordinated by the Federal Financial Institutions Examination Council (FFIEC). This council was established by the Financial Institu tion and Interest Rate Control Act of 1978. It coordinates the activities of five regulatory agen cies: the Federal Reserve, the OCC, the FDIC, the FHLBB (OTS), and the National Credit Union Administration, which charters and regulates national credit unions. Efforts of the FFIEC are directed toward making the field examinations conducted by different agencies similar in scope. Examiners focus mainly on the adequacy or inadequacy of the firm’s capital account for meeting the particular forms of risk exposure. Traditionally, they have devoted their attention to risks from nonperforming and questionable loans and from problems rooted in incompetent management (Kane [1985] and Benston et al. [1986]). The documentation, collateral, and payment records of most large loans and a sam ple of small loans are examined, and the loans are classified into good, substandard, doubtful, and loss categories. The institution’s internal control system and managerial practices are reviewed and evaluated. The examiners discuss their find ings with management and may recommend changes in management practices to improve the institution’s performance, and increases in capi tal to strengthen the institution’s balance sheet. After the on-site examination, federal examin ers prepare a formal report pointing out strengths and weaknesses in the firm’s operation. This report is further summarized into a five-point CAMEL rating. CAMEL is an acronym for five cate gories of condition and performance on which the institutions are graded: capital adequacy, asset quality, management, earnings, and liquidity. Capital adequacy is a measure of an institu tion’s buffer against future unanticipated losses. As explained in section I, in the case of financial institutions, the market value of enterprisecontributed equity is the appropriate indicator of capital adequacy. However, regulators tend to focus on the book value of an institution’s equity. As previously mentioned, in evaluating an insti tution’s asset portfolio, examiners focus on loan quality. Examiners go through loan documenta tion and check the quality of collateral, if any, backing each loan. Judgments are made as to the quality of each borrower and his ability to repay the loan. In addition, examiners check to see if the institution has a high concentration of loans to a specific industry' or to a single borrower. The determination of an institution’s man agement quality is very subjective. Typically, examiners decide on the competence of man agement based on the institution’s performance in the other four categories. Examiners rate the earnings of an institution on both recent performance and on the histori cal stability of its earnings stream. Performance and stability are determined by looking at the institution’s profit composition. Examiners tty to see if the profits come from a solid operating base or are driven by one-time gains, such as those generated by the sale of assets (Whalen and Thomson [1988] ). Liquidity of the institution is analyzed to deter mine its exposure to liquidity risk. To determine the institution’s ability to meet unanticipated deposit outflows, examiners look at the bank’s funding sources as well as the liquidity of its assets. Since troubled institutions often try to hide their problems from the public and the regula tors, it is difficult for examiners to detect prob lems by looking at the institution’s accounts and financial statements. On-site examinations are the most effective way of detecting fraud. As studies by Sinkey (1975, 1979) indicate, quality of man agement and honesty of employees are the most important factors leading to bank failures. How ever, examiners were not specifically asked to examine for fraud until 1984. The U.S. House of Representatives Subcommittee on Commerce, Consumer, and Monetary Affairs of the Commit tee on Government Operations (1984) conducted a study of 105 bank and savings and loan failures between January7 1980 and June 1983 and found that “...criminal activity by insiders was a major contributing factor in roughly one-half of the bank failures and one-quarter of the savings and loan failures....” The committee subsequently recommended that federal examiners be trained and advised to specifically examine for fraud. The component ratings of CAMEL categories are subjectively weighed by the examiner to arrive at an overall rating for the institution. A bank’s rating depends on the examining regulatory agency and the examination staff, since subjective judgments are made in obtaining the CAMEL rating (Whalen and Thomson [1988]). The CAMEL system grades an institution on a five-point scale. Institu tions with ratings of 4 or 5 are considered “prob lem institutions.” The FDIC publishes a list of problem banks, but the FSLIC does not publicize its parallel list of problem savings and loan insti tutions. Problem institutions are examined more frequently and monitored more closely. The CAMEL rating is used by the federal exam iners. State examiners conduct similar examina tions, but they do not necessarily use the CAMEL system. Federal and state examiners disclose their overall rating to the institution’s board of directors. Regulators also use off-site monitoring to complement on-site examinations. Off-site m oni toring focuses mainly on analyzing quarterly income and balance sheet statements obtained from Reports of Income and Condition (that is, call reports) filed with the regulatory agencies. Statistical early-warning models have been available to supervisory agencies since the mid1970s. These models were developed to evaluate the financial condition of institutions in order to determine the priority or urgency for on-site examinations. To a limited extent, off-site analy sis also looks at market data (such as growth rates, deposit interest rates, and stock prices), public disclosures, and credit ratings assigned by private analysts. Examiners seek to uncover regulatory viola tions and to identify problem institutions before their condition deteriorates to the extent that the deposit insurance fund is endangered. However, in addition to their inadequate emphasis on fraud □ Definition of Independent Variables Found Significant in Summarized Empirical Studies Author Sinkey (1975) Variable LRTR OETR OEOI LCR SLRTR Altman (1977) LA NWTA NOIGOI RETA ESTA TLTS HLBANW SRETA Martin (1977) GCARA NITA CI2LN GCONI Avery and Hanweck (1984) LNTA NLTA KTA CILNNL NIT A HERE PTD Definition Loan Revenue/Total Revenue Other Expenses/Total Revenue Operating Expense/Operating Income Loans/( Capital + Reserves) Revenue from State and Local Obligations/Total Revenue Loans/Assets Net Worth/Total Assets Net Operating Income/Gross Operating Income Real Estate Owned/Total Assets Earned Surplus/Total Assets Total Loans/Total Savings FHLB Advances/Net Worth Real Estate Owned (SI)/ Total Assets Gross Capital/Adjusted Risk Assets Net Income/(Total Assets-Cash Items in Process) (Commercial and Industrial Loans + Loans to REITs and Mortgage Bankers + Construction Loans + Commercial Real Estate Loans )/Total Assets Gross Charge-offs/(Net Operating Income + Loss Provision) Natural Logarithm of Total Bank Assets Less Loan Loss Reserves (TA) Net Loans/Total Assets ( Equity Capital + Loan Loss Reserve Allowances)/TA Commercial and Industrial Loans/Net Loans Net After-Tax Income/TA Herfindahl Index for Bank’s Local Banking Market3 risk, examiners are typically slow in identifying and evaluating new types of risks as they emerge. The exposure of institutions to interest volatility risk, foreign exchange risk, sovereign risk, and technology risk is still not explicitly priced.9 The recent risk-based capital adequacy guide lines established by the Federal Reserve System seek to explicitly price different categories of risk. The guideline is based on a regulatory meas ure of capital. Capital adequacy is determined by different capital requirement weights attached to assets that fall into broad risk categories. By the end of 1992, institutions are expected to meet a minimum ratio of qualifying total capital to weighted-risk assets of 8 percent. The risk-based capital ratio focuses on broad categories of credit risk and limited instances of interest-volatility risk. However, it does not incorporate other risk factors mentioned above. Most important, “qualifying capital” is not defined in objective economic terms, that is, as enterprise-contributed capital. Helping regulators perform the task of uncov ering financially troubled institutions is the orig inal motivation of the literature on depositinstitution failures. The next two subsections discuss different approaches taken by earlier empirical researchers. Choice of Independent Variables The first group of studies tries to develop early warning systems that are capable of mimicking the regulator’s evaluation process. The hypothe sis of these empirical studies is that appropriately selected financial ratios designed to measure CAMEL’s five categories of information should be able statistically to discriminate between prob lem and nonproblem institutions. According to the definition of failure featured in this article, these studies do not deserve to be called failure studies because they analyze only the financial condition of the institutions. Moreover, their eval uation of this financial condition is accurate only to the extent that book values reported by an institution approximate market values. The second group of researchers has a more ambitious goal. Instead of merely analyzing an institution’s financial condition, these researchers Semiannual Percentage Change in Total Deposits within Each Bank’s Local Banking Market ■ 9 For definitions of these risk categories and a discussion of how they should be priced, see Benston et al. (1986) and Kane (1985,1989). Definition of Independent Variables Found Significant in Summarized Empirical Studies Author Barth et al. (1985) Benston (1985) Variable NWTA NITA ISFTF LATA LNTA NWTA RETTA YLDEAC COSTFDC Gajewski (1988) PKTAHAT NALR LPDR NLTA SENSDTD AGTOTTL CILTL NITA HCN OGINR82 Definition Total RAPb Net W'onh/Total Assets Net Income/Total Assets Interest Sensitive Funds/ Total Funds Liquid Assets/Total Assets Natural Logarithm of Total Assets Independent variables used in both groups of studies are intended to proxy different dimen sions of the CAMEL rating system. Authors typi cally start out with either a large number of finan cial ratios that cover all the CAMEL categories, or selected financial ratios that were found to be sig nificant in earlier studies. Independent variables found to be significant in the reviewed studies are summarized in table 4. Interpretations of some financial ratios vary across different studies. When the same ratios are interpreted differently and classified under separate categories by different authors, this is noted and discussed. Authors’ classifications of significant independent variables into CAMEL categories are given in table 5. Net Worth/Total Assets Net Income/Total Assets Change in Interest and Fee Income/Earning Assets Change in Interest and Depositors’ Dividends/Earning Assets Regulator-Recognized Capital/Assets Nonaccrual Loans/Total Assets Loans Past-Due 90 Days or More, Still Accruing Interest/Total Assets Net Loans/Total Assets Sensitive Deposits/Total Deposits Total Agricultural Loans/ Total Loans Commercial and Industrial Loans/ Total Loans Net Income/Total Assets Corporate Structurec County-Level Oil and Gas Sector Earnings/Total County Earnings, 1982 a. Herfindahl index is the sum of squares of market shares for banking organizations. b. RAP stands for regulatory accounting principles. It is a more lenient set of accounting principles than the generally accepted accounting principles (GAAP). Under RAP, institutions have a higher book net-worth than under GAAP. c. Corporate structure variable equals zero if the bank is independent or a one-bank holding company; it equals the number of banks in the multibank holding company if a subsidiary. SOURCE: See text. set out to explain why it fails. However, although they acknowledge the conceptual distinction between economic insolvency and failure (Avery and Hanweck [1984], Barth et al. [1985], Ben ston [1985], implicitly; and Gajewski [1988], explicitly), their models contain the same finan cial ratios used in the first group of studies. Choice of Statistical Methods Statistical techniques used in these studies also differ. Earlier research used multiple discriminant analysis (MDA), while more recent researchers prefer qualitative response models (QRM ) . 10 Although discriminant analysis (DA) and quali tative response (Q R) models can be used inter changeably, the motivations behind the two m od els are quite different. What distinguishes a DA model from the ordinary QR model is that a DA model specifies a joint distribution of dependent ) variables, not just the ( j ; ) and independent conditional distribution of given t . In econ ometric QR models, the determination of (bank characteristics) clearly precedes that of (failure); therefore, it is important to specify 1 | ), while the specification of the dis tribution of may be ignored. On the contrary, in the DA model, the statement 1 (for exam ple, being a problem bank) logically precedes the determination of (problem-bank character istics); therefore, it is more natural to specify the joint distribution of and (Amemiya [1981]). In simple terms, DA is merely a classification technique, while QR models analyze a causal relationship. Because problem and nonproblem banks do not come from different groups, but the banks become problem banks through time, QR models are intuitively more appealing in our case. In other words, it is more natural to think of problem banks being assigned to the problem list because of their characteristics than vice versa. In addition, QR estimators have desirable sta tistical properties. The discriminant analysis (x i xt yf P (y = X yi X x y= X X ■ 10 y See Am em iya (1981) for a discussion of these two techniques. Judge et al. (1985), Chapter 18 contains a thorough discussion of qualitative response models. Ea T A B L E 5 1 Significant Independent Variables Classified into C A M E L Categories Variables Avery and Hanweck (1984) Barth et al. (1985) Benston (1985) Gajewski (1988) GCARA KTA LNTA NWTA NWTA PKTAHAT GCONI CI2LN NLTA CILNNL Sinkey (1975) Altman (1977) Martin (1977) LCR NWTA HLBANW ESTA RETA SRETA TLTS Capital Adequacy Asset Quality LRTR LA Management Competence OEOI OETR Earnings SLRTR NOIGOI NITA NITA HERE PTD NALR LPDR NITA ISFTF NLTA SENSDTD AGTOTTL CILTL RETTA NITA YLDEAC OGINR82 COSTFDC LATA LNTA Liquidity HCN Fraud SOURCE: See text. ML X estimator is the estimator when is multi variate normal. However, DA is not consistent when this assumption is violated. Still, studies analyzing robustness of discriminant analysis to non-normality report good performance by DA. QR models are not affected by the distribution of Properties of the two estimators are further discussed in Amemiya (1981). Keeping in mind the underlying difference be tween the two models, DA might be useful if a dichotomous classification is the goal. On the other hand, QR models should be preferred when the model, the estimation of the coeffi cients of the independent variables, and thus the determination of the probability of the occur rence of the event, is important. X. Review of Prior Empirical Literature Sinkey’s (1975) problem-bank study is one of the earliest on this topic. He uses linear multiple discriminant analysis (MDA) to evaluate data on 2 2 0 problem and nonproblem commercial banks for the period 1969-1972. Half of his sam ple consists of commercial banks that were listed as problem banks by the FDIC in 1972 and early 1973- Each problem bank is matched with a nonproblem bank based on the following char acteristics: ( 1 ) geographic market area, ( 2 ) total deposits, (3) number of banking offices, and (4) Federal Reserve membership status. The sample contains mostly small banks (total deposits less than $ 1 0 0 million). After testing more than 100 ratios designed to cover all CAMEL categories, 10 financial variables are chosen. Among these, six significantly increase the overall discriminatory power of the model in a stepwise analysis. In table 4, these variables are ranked in decreasing contribution to discrimina tory power. The loan revenue variable ( ), which is an indicator of asset quality, proves to be the best discriminator. Sinkey interprets most of the variables in his study as proxies for management quality and honesty, including two operating efficiency vari ables ( ). The loan-to-capital ratio ) is taken as a measure of adequate bank capital. Sinkey concludes that although the dif ferences in the means of these variables are sta tistically significant, the classification accuracy of LRTR (LCR OEOI, OETR the model is low due to group overlap among the problem and nonproblem banks. Altman (1977) also uses multiple discriminant analysis to analyze three groups of troubled sav ings and loan institutions. Improving on Sinkey’s (1975) study, he tests and rejects the equality of group dispersion-matrices, and therefore uses a quadratic structure. He examines data on 212 sav ings and loan associations during the period 1966-1973. O f these institutions, 56 are classified as having serious problems, 49 as having tempo rary problems, and 107 as having no problems. His definition of “serious problem” closely matches the definition of failure in this paper. He defines “temporary’ problem” institutions as those with problems similar to the ones in the serious problem group, but that have avoided regulatory interference. Finally, the “no problem” group serves as the control group. It consists of institutions that did not show any indication of financial problems on the failure date of the serious-problem group, or thereafter. The range of asset size in all three samples is from $ 1 million to $ 1 0 0 million. Altman tests 32 financial ratios that cover all CAMEL categories. His best predictor model includes only seven variables, listed in table 4. Altman concludes that operating income ( ) and its trend are the most important discriminators. He also finds net worth ( ) and real estate owned ) variables to be important. He interprets these variables as reflecting an institution’s profitability, capital adequacy7, and asset quality. Martin (1977) is the first author to use a logit probability model to evaluate commercial-bank failures. He analyzes data covering all commer cial banks that were members of the Federal Reserve System between 1970 and 1976. In addi tion to closures, his failure definition includes banks whose net worth “...declined drastically over a year or less.” Therefore, his analysis focuses on certain kinds of insolvency7 and not just on failure. Martin’s work represents the transition between the first and second group of studies. He ana lyzes an institution’s probability of becoming insolvent in a book-value sense before analyzing the group characteristics. The second group of studies takes this analysis one step further to explain the closure process rather than merely to approximate an early-warning system. Martin obtains his best results using 1974 data on 23 failed and 5,575 nonfailed commercial banks. He analyzes 25 ratios chosen for their usefulness in previous studies. The preferred model includes only four variables. These varia bles measure earnings ( ), loan quality NOIGOI (RETA NIT A NWTA ( CI2LN, GCONI), and capital ( GCARA ). Avery and Hanweck (1984) study commercial bank closures using semiannual data for 1 0 0 closed and 1,190 nonclosed commercial banks during the period December 1978 to June 1983. Their sample includes only institutions with assets of $250 million or less. Although closure is acknowledged to be a regulatory decision, it is analyzed using only nine financial ratios, chosen because previous authors found them significant. They assume that the probability7 of closure de pends on a distributed lag of the financial condi tion of the institution and estimate a logit proba bility model. Five financial-ratio coefficients prove significant and receive signs expected a priori. These ratios incorporate elements of earn ings ), asset quality ) and capital adequacy ). Avery and Hanweck interpret bank size ( ) as an indicator of ability7 to raise new capital. Observing the reluctance of regulators to fail large banks, they state that larger institutions may raise capital more easily since it may be assumed that they are managed better and able to turn around faltering situations quickly7. Local banking market variables ) are also signifi cant, but receive unexpected signs. Their most puzzling result is a counterintuitive sign for lagged financial-condition variables. They con clude that lagged financial ratios are not impor tant in explaining bank closures. Barth et al. (1985) study thrift institution clo sures using a logit probability7 model. They use semiannual data for 3 1 8 closed and 588 non closed savings and loan associations covering the period December 1981 to June 1984. They also mention that closure is a decision made by the regulators. Again, however, only 12 financial ratios similar to the ones used in earlier studies are analyzed. Five of these variables receive their expected signs and prove statistically significant. These measure capital adequacy ), asset quality ), earnings ), and liquidity7 ). They interpret size ) as an indicator of greater liquidity, since they believe larger institutions have a greater ability to borrow in order to alleviate unexpected liquidity prob lems. A possible alternative interpretation is that this variable captures the reluctance of regulators to liquidate large institutions (Conover [1984], Seidman [1986]). Benston (1985) conducts a logit analysis of 178 closed and 712 nonclosed savings and loans for the period 1981-1985. Among the 28 financial ratios he includes, only four prove statistically significant. These are measures of capital ade quacy ) and earnings and ). (NITA (NITA, CILNNL (KTA, LNTA LNTA (HERE, PTD (ISFTF (LNTA, LATA (NWTA COSTFDC (NITA (NWTA (LNTA (RETTA, YLDEAC, Gajewski (1988) studies commercial-bank clo sures by analyzing a 1986 cross-sectional data set of 134 closed and 2,747 nonclosed banks. Empha sizing the need to differentiate between insol vency and failure, Gajewski is the first author to incorporate this distinction into his modeling. His model has two equations. The first mimics the regulatory screening process, in the spirit of an early-warning model. The second studies the closure process. Although Gajewski recognizes the importance of the regulatory decision making process in explaining bank closures, his two equations differ only in their endogenous variables—book-value insolvency and closure. He analyzes both insolvency and closure using only financial ratios and county characteristics. Characteristics of the bank’s local economy are represented by the percentage of county-level oil and agricultural earnings to total county earn ings. A total of 25 financial ratios covering CAMEL categories are chosen to study the finan cial condition and closure of the institutions. The final specification of the logit probability model develops 1 0 significant variables, listed in table 4. These include measures of capital ade quacy ) obtained from the first equa tion, asset quality ( ), management competence ), earnings and fraud ). What Gajewski interprets as managementcompetence variables are interpreted as asset quality variables by earlier authors. (PKTAHAT NALR\ LPDR (NLTA, SENSDTD, CILTL, AGTOTTL (NITA, OGINRH2), (HCN Relative Importance of Different C A M E L Categories Although cited studies analyze the relative dis criminatory power of different CAMEL categories, it is difficult to compare the findings of one study against another, due to differences in data sets, proxies, and interpretations. Nevertheless, all authors find capital adequacy ), generally proxied by the book value of net worth, to be significant. In addition, earnings usually a measure of net income, are a significant indica tor of financial condition. After capital adequacy and earnings, asset quality ), as proxied by various loan ratios, is found to be a significant indicator of financial trouble by most authors. Fraud and management ) prove to be difficult categories competence to proxy. Instead of explicitly representing them by financial ratios, most authors prefer to con sider the set of included variables as incorporat ing implicitly the effects of management and fraud. With the exception of the study by Barth (C (E ), (A (M et al., liquidity (Z ) is not found to be a signifi cantly important category. III. Possibilities for Improving the Empirical Analyis of DepositInstitution Failures The literature on deposit-institution failures still leaves much room for improvement. The first group of studies seeks to discriminate between problem/nonproblem and closed/nonclosed in stitutions using only financial ratios. The choice of candidate regressors in the accounting-ratio models lacks a compelling theoretical founda tion. Financial ratios are simply utilized in var ious statistical procedures until they “work.” The second group of studies seeks to explain failure using only instrumental variables borrowed from accounting-ratio models. These studies fail to distinguish successfully between insolvency and failure in their modeling and have little theoreti cal underpinning. In studying the failure of financial institutions, it is crucial to make a distinction between eco nomic insolvency and failure. As discussed in section I, economic insolvency is a marketdetermined event. In contrast, the decision to fail an institution requires that a state commission or federal agency realize, often under the urging of the deposit-insurance agency involved, that a natural propensity to forbear is no longer in its bureaucratic interest (Kane [1985]). Failure is a regulatory decision, influenced by conflicts of interest that exist between regulators, politicians, and taxpayers. These conflicts of interest allow political, bureaucratic, and eco nomic pressures, and career-oriented incentives of the regulators, to shape failure decisions. Therefore, economic insolvency and failure of financial institutions should be distinguished but studied simultaneously. Furthermore, failure should be modeled for mally as the outcome of a regulatory decision making process, explicitly taking into considera tion regulators’ constraints and conflicts of interest. In studying economic insolvency of financial institutions, the appropriate measure is the market value of enterprise-contributed capital. Assuming an efficient stock market, the market value of enterprise-contributed capital summa rizes the institution’s financial condition, freeing the researcher of the dilemma of picking and choosing the “right” financial ratios among many possibilities. Also, if one uses financial ratios cal culated from balance sheets and income state- ments, the implicit assumption is that book values adequately proxy market values. Adopting the market value of enterprisecontributed equity as the measure of economic solvency and analyzing failure within a theoretical model of regulatory decision-making brings a much-needed structure to the choice of inde pendent variables, establishing a theoretical basis for the empirical research on deposit-institution failures. Most studies of problem and failed banks con centrate on small-bank failures. They include few, if any, large banks in their samples. How ever, recent increases in large-bank insolvencies indicate the importance of developing a model of large-bank failures. Developing a large-bank failure model has the further advantage of allowing us to use stockmarket data. In addition, as Kaufman (1985) states, consequences of insolvency and failure of large banks are blown out of proportion by the regulators. Regulators publicly show a fear of large-bank failures, ostensibly because of the possible repercussions on the banking system and on economic policy. At the time of the Conti nental Illinois National Bank crisis, Comptroller of the Currency C. T. Conover (1984), in defense of his rescue of the bank, argued: In our collective judgement (directors of the FDIC, the chairman of the Federal Reserve Board, and the Secretary' of the Treasury'), had Continental failed and been treated in a way in which depositors and creditors were not made whole, we could very well have seen a national, if not an international, financial crisis the dimensions of which were difficult to imagine. None of us wanted to find out. What leads to forbearance policies and ineffi cient insolvency resolution methods, however, is not necessarily these vague and poorly docu mented consequences, but the hidden fears of what particularly visible large-bank failures can do to the perceptions of the quality of regulators’ performance in office (Kane [1989] )." Thus, one would expect the political and bureaucratic constraints of the regulators to be especially binding when their decision to fail concerns a large bank. Demirgiic-Kunt (1990, forthcoming) addresses the above issues. 12 It is a study of large commercial-bank failures for the period 19731989. Annual panel data are used in estimation. The failure model developed distinguishes between economic insolvency and failure, study ing them simultaneously. An estimate of the market value of enterprise-contributed equity is taken as the measure of economic insolvency. Failure determination is based on a theoretical model of failure decision-making in the spirit of the Kane model. The theoretical model identifies and explicitly incorporates important regulator constraints and incentives. In the empirical model, the FDIC’s number of examiners and size of the insurance fund are proxies for economic constraints, whereas failure rate (for banks and businesses), number of problem banks, variance of interest rates, and bank size are included to proxy political and bureaucratic constraints implicit in the career-oriented incentives of regulators. As expected, results indicate that regulator constraint and incentives play a significant role in failure determination. The empirical model of bank failures developed in Demirgiic-Kunt (forthcoming) is more complete because it takes into consideration a previously ignored determi nant of the decision-making process and brings theoretical structure to the empirical depositinstitution failure literature. One possibility for future research in this area of deposit-institution failures is to investigate changes in regulatory decision-making through the years. Periodic restructuring of the financial system (most recently by the 1989 FIRRE Act) leads to shifts of power among different regula tory bodies and may affect failure decisions. It is also important to take into consideration differ ences among various insolvency resolution methods, that is, different categories of de facto failure (Maddala [1986] ) . 13 Development of a failure model that distinguishes between differ ent methods of insolvency resolution is the next challenging task facing economists. ■ 12 See Demirgiic-Kunt (1989) tor a preliminary version of the study and empirical results. The theoretical model is fully developed in Demirguc-Kunt (1990, forthcoming). ■ 11 A discussion of these policies can be found in Kane (1985, 1989), Benston et al. (1986), and Caliguire and Thomson (1987). ■ 13 For a discussion, see references in footnote 11. References Altman, Edward I., "Predicting Performance in the Savings and Loan Association Industry7,” October 1977, 443-66. Journal of Monetary' Economics, 3, Amemiya, T., “Qualitative Response Models: A Survey,” De cember 1981, 1483-1536. Journal of Economic Literature, 19, Avery, Robert B. and Hanweck, Gerald A., “A Dynamic Analysis of Bank Failures,” Conference Pro ceedings, Federal Reserve Bank of Chicago, 1984, 380-95. Bank Structure and Competition, Barth, James R., Brumbaugh, R. Dan, Jr., Sauerhaft, Daniel, and Wang, George H. K., “Thrift Institution Failures: Causes and Policy Issues,” Conference Proceedings, Federal Reserve Bank of Chi cago, 1985. Bank Structure and Competition, An Analysis of the Causes of Savings and Loan Association Failures, Mono Benston, George J., graph Series in Finance and Economics, New York University, 1985. ________, Eisenbeis, Robert A., Horvitz, Paul H., Kane, Edward J., and Kaufman, George J., Perspectives on Safe and Sound Banking, Cambridge, MA: MIT Press, 1986. The Federal Reserve’s Attachment to the Free Reserve Con cept, Washington, D.C.: House Committee on Brunner, Karl and Meltzer, Allan H., Banking and Currency, 1964. Fiscal Theory and Political Buchanan, James M., Chapel Hill, N.C.: University of North Carolina Press, I960. Economy, Public Finance in Democratic Process, ------ , Chapel Hill, N.C.: University of North Carolina Press, 1967. The Theory of ------ and Tollison, Robert D., Ann Arbor, MI: University of Michigan Press, 1984. Public Choice, Buser, Stephen A., Chen, Andrew H., and Kane, Edward J., “Federal Deposit Insurance, Regu latory Policy, and Optimal Bank Capital,” March 1981, 51-60. Journal of Finance, 35, Cagan, Phillip, Determinants and Effects of Changes in the Stock of Money, New York: Columbia University Press, 1965. Caliguire, Daria B. and Thomson, James B., “FDIC Policies for Dealing with Failed and Troubled Institutions,” Federal Reserve Bank of Cleveland, October 1, 1987. Economic Commen tary, Campbell, T. S. and Marino, A. M., “On the Incentives for Managers to Make Myopic Investment Decisions,” Los Angeles: Univer sity of Southern California, Working Paper, 1988. Conover, C. T., Testimony in U.S. Congress, Sub committee on Financial Institutions Suspen sion, Regulation and Insurance of the Com mittee on Banking, Finance and Urban Affairs, Inquiry into Continental Illinois Corp. and Continental Illinois National Bank: Hearings, September 18 and 19 and October 4, 1984, 91-111, 98th Congress, 2nd Session: 287-88. Demirgiic-Kunt, Asli, “Modeling Large Commercial-Bank Failures: A SimultaneousEquations Analysis,” 8905, Federal Reserve Bank of Cleveland, May 1989. Working Paper ________, “Modeling Large Commercial-Bank Failures: A Simultaneous-Equations Analysis,” Ph.D. Dissertation, The Ohio State University, 1990 (forthcoming). A Monetary^ History of the United States, 18671960, Princeton, N.J.: Princeton University Friedman, Milton and Schwartz, Anna J., Press, 1963Gajewski, Gregory R., “Bank Risk, Regulator Behavior, and Bank Closure in the Mid-1980s: A Two-Step Logit Model,” Ph.D. Dissertation, The George Washington University, 1988. Horvitz, Paul M., “Stimulating Bank Competition Through Regulatory Action,” March 1965, 1-13- Finance, 20, Journal of Judge, George G„ Griffiths, W.E., Hill, C. Carter, Lutkepohl, Helmut, and Lee, Tsoung-Chao, The Theory and Practice of Econometrics, & Sons, Inc., 1985. New York: John Wiley Kane, Edward J., (1981a) “Deregulation, Savings and Loan Diversification, and the Flow of Housing Finance,” National Bureau of Eco nomic Research Working Paper No. 640, March 1981. ------ , (1981b) “Impact of Regulation on Economic Behavior: Accelerating Inflation, Technological Innovation, and the Decreasing Effectiveness of Banking Regulation,” May 1981, 355-67. of Finance, 36, Journal The Gathering Crisis in Federal Deposit Insurance, Cambridge, MA: MIT Press, 1985. ________, ________, “Appearance and Reality in Deposit Insurance,” 1986, 175-88. 10, Journal of Banking and Finance, ________, “Changing Incentives Facing FinancialServices Regulators,” paper presented at the Federal Reserve Bank of Cleveland Confer ence on Bank Structure, 1988. The S&L Insurance Mess: How Did It Happen? Cambridge, MA: MIT Press, 1989. ________, ________ and Foster, C., “Valuing and Eliminat ing the Subsidies Associated with Conjectural Government Guarantees of FNMA Liabilities,” Working Paper, Ohio State University, 1986. Karaken, John H., “Deposit Insurance Reform or Deregulation Is the Cart, Not the Horse,” Federal Reserve Bank of Minneapolis, Spring 1983, 1-9. Quarterly Review, ------ and Wallace, Neil, “Deposit Insurance and Bank Regulation: A Partial Equilibrium Exposition,” July 1978, 413-38. 51, Journal of Business, Kaufman, George J., “Implications of Large Bank Problems and Insolvencies for the Banking System and Economic Policy,” Occasional Papers, SM-85-3, Federal Reserve Bank of Chi cago, 1985. Maddala, G. S., “Econometric Issues in Empirical Analysis of Thrift Institutions’ Insolvency and Failure,” Invited Research Working Paper No. 56, Federal Home Loan Bank Board, 1986. Marcus, Alan J. and Shaked, Israel, “The Valua tion of FDIC Deposit Insurance Using OptionPricing Estimates,” November 1984, 446-60. and Banking Journal of Money, Credit, 20, Martin, Daniel, “Early Warning of Bank Failure,” November 1977, 249-76 . Journal of Banking atid Finance, 1, McCulloch, J. Huston, “Interest Rate Risk and Capital Adequacy for Traditional Banks and Financial Intermediaries,” in Sherman Maisel, ed., Chicago: University of Chicago Press for the National Bureau of Economic Research, 1981, 223-48. Risk and Capital Adequacy in Commer cial Banks, ________, “The Ohio S&L Crisis in Retrospect: Implications for the Current Federal Deposit Insurance Crisis,” paper presented at the Fed eral Reserve Bank of Chicago Conference on Bank Structure and Competition, 1987. Meltzer, Allan H., “Major Issues in the Regula tion of Financial Institutions,” August 1967, Supplement, 482-501. Journal of Polit 75, ical Economy, Merton, Robert C., “An Analytical Derivation of the Cost of Deposit Insurance and Loan Guar antees,” June 1977, 3-11. Journal of Banking and Finance, 1, ________, “O n the Cost of Deposit Insurance When There Are Surveillance Costs,” July 1978, 439-52. of Business, 51, Journal Myers, Stewart C. and Majluf, Nicholas, “Corpo rate Financing and Investment Decisions When Firms Have Information that Investors Do Not Have,” June 1984, 187-221. ics, Journal of Financial Econom 13, Narayanan, M.P., “Managerial Incentives for Short-Term Results,” De cember 1985, 1469-84. 40, Journal of Finance, Bureaucracy and Repre sentative Government, Chicago: Aldine Ather Niskanen, William A., ton, 1971. Pyle, David H., “Pricing Deposit Insurance: The Effects of Mismeasurement,” Federal Reserve Bank of San Francisco and University of Cali fornia, Berkeley, 1983________, “Deregulation and Deposit Insurance Reform,” Federal Reserve Bank of San Francisco, 1984. Economic Review, Ronn, Ehud I. and Verma, Avinash K., “Pricing Risk-Adjusted Deposit Insurance: An OptionBased Model,” Sep tember 1986, 871-95. Journal of Finance, 41, Schwartz, Eduardo and Van Order, Robert, “Valuing the Implicit Guarantee of the Federal National Mortgage Association,” April 1988, 23-34. Journal of Real Estate Finance and Economics, 1, Scott, Kenneth W. and Mayer, Thomas, “Risk and Regulation in Banking: Some Proposals for Deposit Insurance,” Fall 1971, 857-902. Stanford Law Review, Seidman, W., presentation by FDIC chairman to the National Press Club, Washington, D.C., October 1986. ia Sharpe, William F., “Bank Capital Adequacy, Deposit Insurance and Security Values,” Jour nal of Financial and Qualitative Analysis, Proceedings Issue, November 1978, 701-18. 13, Sinkey, Joseph, “A Multivariate Statistical Analysis of the Characteristics of Problem Banks,” March 1975, 21-36. Journal of Finance, 30, ------ , Problem and Failed Institutions in the Commercial Banking Industry, Greenwich, CT: JAI Press, 1979. The Citizen and the State: Essays on Regulation, Chicago: University of Stigler, George J., Chicago Press, 1977. An Enquiry>into the Nature and Effects of the Paper Credit of Great Britain, New York: Farrar and Rinehart, 1939. Tulloch, Gordon, The Politics of Bureaucracy, Thornton, Henry, Washington, D.C.: Public Affairs Press, 1965. Tussing, Dale, “The Case of Bank Failures,” October 1967, 140-41. Journal of Law and Economics, 10, U.S. Congress, House Committee on Govern ment Operations, Federal Response to Crimi nal Misconduct and Insider Abuse in the Nation’s Financial Institutions, 1984, Fiftyseventh Report, 98th Congress, 2nd Session, House Report 98-1137. Whalen, Gary and Thomson, James, “Using Financial Data to Identify Changes in Bank Condition,” Federal Reserve Bank of Cleveland, Quarter 2, 1988, 17-26. Economic Review, Settlem ent Delays and S to c k Prices by Ramon P. DeGennaro Ramon P. DeGennaro is a visiting scholar at the Federal Reserve Bank of Cleveland. The author would like to thank Randall W . Eberts, Jam es T. Moser, and Jam es B. Thomson for helpful comments. Introduction The typical stockbroker requires only about two minutes to execute and confirm a market order. During that time, the order is routed electroni cally either to the specialist or to the Intermarket Trading System, which connects eight regional markets including the New York Stock Exchange and the National Association of Securities Deal ers. These agents then pair the order with another buy or sell order. 1 Thanks to modern technol ogy, the process of executing a trade and pro ducing a confirmed order is quick and efficient. Although this confirmed order represents a binding contract between the buyer and seller, neither the security nor payment for the security changes hands at the time the trade is con firmed. Instead, payment for the stock occurs five business days later, when the buyer delivers a bank check to the seller and the seller delivers the promised securities. 2 Until final payment is made, the stock trade remains conditional, and ■ 1 ■ 2 For a further discussion of trading details, see Jakus and Chandy (1989). In practice, these transactions usually are executed by brokers acting as agents. official title remains with the seller, who cannot use the proceeds of the sale. The equity markets have no provision to com pensate the seller for the opportunity cost he bears while waiting for the trade to clear. In con trast, bond-market procedures call for explicit adjustment of the cost of the bond for interest accrued since the most recent coupon date. Interest is calculated using the number of days from the last coupon payment until the date of delivery, not the date of the trade. If the terms of the trade call for delivery tomorrow instead of today, the buyer must pay an extra day’s worth of interest. Another important market, residential real estate, while not explicitly adjusting the pur chase price for the date of closing, does prorate taxes and rents for the date of occupancy. Although the stock markets make no explicit adjustment for the opportunity cost of settlement delays, rational investors do not ignore the fact that they lose several days’ worth of interest. Indeed, much empirical work has assumed that investors consider delivery procedures in pricing assets, although few studies have tested this theory. This paper studies whether investors do, in fact, consider settlement delays in determining stock prices. We construct two models of stock returns. The first expresses returns as a function of changes in the settlement delay. The second models returns as a function of changes in the length of the delay and in the federal funds rates during the delay. The first model controls for variation in the length of the delay, while the second controls for both the opportunity cost and the length of the delay. We then conduct regres sion tests of the significance of these variables. Both models show that in the full sample and all subperiods, investors apparently do consider the settlement delay; the variables controlling for it are statistically significant and correctly signed. Section I reviews previous research regarding payment delays, and section II develops our model of the return-generating process. In sec tion III we describe the data, conduct prelimi nary tests, and report the results. A summary concludes the paper. I. Previous Research and the Impact of Delivery Procedures Lakonishok and Levi (1982) speculate that set tlement and check-clearing delays might explain the “weekend effect” in stock prices. The week end effect refers to the well-documented ten dency of stock prices to decline on Monday. 3 Lakonishok and Levi note that, in addition to the settlement delay, the check presented at settle ment requires another business day to clear. They claim this makes the total payment delay six bus iness days. For their empirical work, they add and subtract interest based on the prime rate, but, more important for our purposes, they con duct no tests to determine if buyers actually do compensate sellers in the manner they suggest. DeGennaro (1990, forthcoming) tests the con jecture that the combined settlement and checkclearing delays explain the weekend effect. He concludes that, while the combined delay fails to explain the weekly return pattern, it does appear to influence measured stock returns. However, he also reports that the estimated rate of com pensation for the combined delay varies substan tially, suggesting that further work is necessary. ■ 3 The weekend effect was first identified by Cross (1973). An important paper by French (1980) reexamined this apparent anomaly, demonstrating that returns on M onday are so persistently negative that rational investors must expect to suffer losses on M ondays. Lakonishok and Smidt (1988) extend the evidence of negative Monday returns to a 90-year sample. Gibbons and Hess (1981) show that Treasury bills also earn below-average returns on Mondays, although returns are not negative for bills. Another example is Choi and Strong (1983), who study “when-issued” common stock. Firms announce stock issues well in advance of the time the new securities are issued; investors trade these securities on a “when-issued” basis. Choi and Strong attempt to determine why this when-issued stock commands a premium over the corresponding stock that is currently out standing. They speculate that when-issued stock represents the existing share plus a zero-interest loan. They find that adjusting prices for the interest savings is insufficient to explain the dis crepancy, but again, they do not test to see if investors price the zero-interest loan. More recently, Flannery and Protopapadakis (1988) assume that settlement and clearing delays are priced in their test of the generality of the weekend effect. They study three stock indexes and seven Treasury bond maturities to learn if intraweek seasonality is the same across these assets. Following the suggestion of Lakoni shok and Levi, they adjust the returns on the 10 assets to control for the financing costs incurred during the payment delays. They find that the returns on these assets do not vary in a similar manner during the week, but again, the authors do not test if the delay is actually priced. DeGennaro (1988) shows such payment delays can have important implications for inter est rates. If delays exist in the Treasury bill market, but are not explicitly incorporated into pricing equations, certain common estimators of term premiums are biased in favor of finding positive premiums. He shows that this bias is sufficiently large to explain the results of McCulloch (1975). However, he does not test if inves tors do, indeed, consider these delays. The results of the present paper are important for several reasons. First, if the delays have no impact on observed prices, then the aforemen tioned studies must be flawed: theoretical work begins with inappropriate assumptions, and empirical studies are misspecified. Second, if investors do consider settlement delays in deter mining equity prices, then observed prices diverge from true prices. This has implications for the event-study methodology commonly used in empirical tests (see, for example, Hite and Owers [1983]). To conduct an event study, the researcher first estimates the parameters of a model using time-series data prior to the event in question. He then calculates abnormal returns, defined as realized returns less the returns pre dicted by the model. Significance tests can be conducted using the cumulative sum of these residuals. To date, all event studies known to the author have ignored the possibility that payment delays may influence the measured stock price and return. If these delays do affect stock prices, events that may seem to be economically signifi cant may in fact be negligible once proper accounting for the delays is made. Conversely, events judged to be insignificant may be important. Consider, for example, an event that the researcher expects to generate positive returns, but which in fact does not. The total compensa tion for the settlement delay, capitalized in the observed price, may be higher than usual on the event date (due to a holiday that lengthens the delay, or perhaps simply to an increase in interest rates). This would make the observed price higher than usual, biasing the significance of test statistics. The reverse might also be true. The economic effect of an event may be positive and significant, but if the number of calendar days in the delay is lower than usual, or if the opportunity cost on a daily basis is less, then the impact of a true economic event might be negated and appear insignificant. Other important results might also be affected. For example, French and Roll (1986) document a large decrease in volatility when markets are closed. The variance of stock returns from Friday’s close to Monday’s close is only about 10 or 15 percent higher than during a one-day holding period. If the opportunity cost of the settlement delay varies systematically— for example, if inter est rates or the delay varies according to the day of the week— French and Roll’s variance ratio measures both the true volatility and the variance in the opportunity cost. While this is unlikely to be sufficient to overturn their results, divergences from true prices are especially important in stud ies of variance, which is a particularly sensitive measure due to the squaring of deviations from the mean. Perhaps the most important reason for studying whether delivery procedures are important and whether settlement delays are priced is their implication for market efficiency. If the settle ment delay does not affect prices, then research ers must not only reinterpret research that pre sumes it does, but they must also explain why rational investors ignore the fact that the present value of the purchase price is reduced because of these delays. The choice of delivery procedures may become an increasingly important policy issue for the securities industry. For example, the present fiveday delivery terms trace to the inability of tech nology to handle heavy' trading volume during the late 1960s. Prior to February 9, 1968, the set tlement period was only four business days; extending it to five ensures that brokers have a weekend between the trade date and the delivery date to complete the necessary paperwork. Con ceivably, further increases in volume could force another extension, while technological advances might permit a reduction. A reduction in the time between the trade date and the delivery7 date may be important in preventing defaults on trades. For example, although the buyer and seller commit to trade at the confirmation of their order, large price changes create incentives for one side of the transaction to renege. For example, equity pur chasers during the week of October 12, 1987, expected to receive stock worth a given amount; instead, they received stock worth about 2 0 per cent less. Although the safeguards against such defaults proved adequate in this case, the increasing volatility of financial markets observed in recent years means larger losses can be sus tained between the time of the trade and the date that the trade becomes final, increasing the likelihood that the buyer will default. II. The Model If investors consider delivery procedures in pric ing stocks, then observed prices contain the true value of the underlying asset plus an adjustment for the settlement delay. Observed prices mis state true values. Since empirical work must use observed prices, we must devise a model that removes any adjustment the market incorporates for the delay. To do this, we first define the true as the price observed in the stock price, absence of delays. The expected true price at time / asa function of the true price at the beginning of the holding period (time 1 ) is P*, t- (1 ) £ ,.,( ? ;) = P *_,exp [E (R *) - Et _ ,(d t)], Et _ x where is the expectations operator condi tioned on information available at time 1 , is the unobservable true price at time , is the unobservable (constant) expected continuously compounded daily rate of return on the stock in the absence of delays, is the dividend yield, and is the base for natural logarithms. Equation (1) states that if no divi dends are expected to be paid, the expected price at is the price at 1 adjusted for the expected continuously compounded rate of price appreciation. If dividends are expected to be paid, the expected price is adjusted down ward accordingly. t- P* E (R *) dt exp t t- t bQ To incorporate the settlement delay, the observed price is written as P st P - P *exp(% c/t), j= i (2) bx 7 st where is the number of days in the settlement delay and is the continuously compounded rate of compensation on day for trades made at If investors ignore delivery procedures, equals zero and the true price equals the observed price. If sellers demand and receive compensation for the settlement delay, is posi- cJt j t. c c stcjt tive. In equation (2), ]£ represents the total j= i compensation to the seller for financing the position until he receives the proceeds of the sale at settlement. Equation (2) is also true at 1, so t- (3) P t _ ! = />*_ st - i l exp( X C/o-i))j =1 P*, Solving equations (2) and (3) for substitut ing into equation (1), and assuming that the and are uncorrelated, we can rearrange equa tion ( 1 ) to obtain P c (4) st ~st-Xcj(t Xcjt J = j= log[£, _ x( P t) / P t . J + Et „ x(dt) = E (R *) + 1 1 i c Et _x(Rt ) i) • t, = E (R *) s, as the (c (7) y/j, = cj„ jJt t, st 'XS, as X fj, , so that J= 7 X R ,. R R, = bQ+ b xAs, + et . st ^ S t = X cj f j= Substi- 1 tuting 2)5 into equation (4) and combining terms yields Et _ x(Rt ) = £ (/? * ) + XS, + c X As,, j where is the federal funds rate on day of the settlement delay for trades at and 7 is a constant. For notational convenience, we define (8 ) where the total expected return on the stock— capital gains plus dividends— is written as Intuitively, equation (5) says the observed expected return equals the expected return in the absence of delays plus changes in the impact of delays. To proxy for the dependent variable t , we use the return on the value-weighted portfolio, including dividends, provided by the Center for Research in Security Prices (CRSP). Although we have derived our model in terms of an individ ual stock, if the settlement delay affects any stock, it must affect all stocks. Further, this effect is not diversifiable: any settlement effects must appear in the observed return on a portfolio. Substituting ex post values, we obtain our test equation: (6 ) c, 1 Letting be constant and defining A change in 5 at time we obtain (5) In this model, estimates the unobservable expected continuously compounded daily rate of return on the stock in the absence of delays, and estimates the rate sellers receive as compensation for the settlement delay. Theory suggests that both coefficients should be posi tive. This is because risk-averse investors require a premium to compensate for the nondiversifiable risk contained in stocks, and increases in the financing costs during the settlement delay require buyers to raise their bids to compensate the sellers. Therefore, one-tailed tests are appropriate. One potential problem with this specification is that As varies relatively little. To circumvent this, we also estimate a second specification. Rather than letting the settlement cost per day ) be constant in equation (4), we use the federal funds rate as a proxy for c. The federal funds rate is both readily available and respon sive to changes in the economic environment. Formally, we write 7 XS X AXS t , t, where A is the change in at and the total expected return on the stock is again writ ten as Substituting ex post values, we obtain Rt . (9) Rt = b'0 + b\AXSt + e't . b'0 As in equation ( 6 ), estimates the unobserv able expected continuously compounded daily rate of return on the stock in the absence of estimates 7 , the delays, but in this model, proportion of the federal funds rate sellers receive as compensation for the settlement delay. One-tailed tests are again appropriate. Equation (9) offers both advantages and dis advantages relative to equation ( 6 ). In equation ( 9 ), the independent variable is a function of the federal funds rate, and therefore may be simul taneously determined with the stock return. However, it controls for both the length of the settlement delay and the opportunity cost during that delay rather than for simply the number of b\ T A B L E Regression Results Estimates obtained by regressing the rate of return on the CRSP value-weighted index, including dividends on the change in the settlement delay (A corrected for heteroscedasticity: (Rt), st), Rt= b Q + b jAs, + u t , Full sample period: January 1, 1970-December 31, 1986 (4,296 observations) Parameter estimate (f-statistic) 4.68 x IO ' 4 (3.03)a 7.27 x 10“ 4 (4.07)a e III. Data and Results Data ut = et - 8et _ j . Variable st AXSt Equation (1 0 ), Full Sample (10) This asymmetric treatment permits the brokerage firm to use the funds between the two dates. The firm generates revenue by imposing an added cost of trading on its customers. If these inves tors are the marginal traders, neither A nor measures the true cost of the delays these investors face. Again, the estimated slope coeffi cients could be insignificant. 0.23 (15.5)a a. Significant at the 1 percent level. NOTE: Significance levels are for one-tailed tests on SOURCE: Author’s computations. b() and bY The stock-return measure is the return on the CRSP value-weighted index, with dividends. We use 4,296 observations from January 1, 1970, through December 31, 1986. Federal funds rates used to compute the opportunity cost of the set tlement delay are from the Federal Reserve Board. We estimate equation ( 6 ) in the full sam ple and in three subperiods partitioned at October 6 , 1979 and October 9, 1982, the dates of important changes in the Federal Reserve’s operating procedures. O n the former date, the central bank began focusing on the level of nonborrowed reserves rather than on the level of the federal funds rate. On the latter date, it began attempting to stabilize interest rates. Preliminary Tests days in the delay. It is also much more variable than in equation ( 6 ). Which economic or institutional forces could cause the slope coefficients in equations ( 6 ) and ( 9 ) to be not significantly different from zero? First, is the settlement delay Although the exchanges alter only rarely, brokerage firms may not credit and debit accounts as accurately as the exchanges. For example, they may err and credit a customer’s account later than promised. Such mistakes may not always be discovered. Even if the customer does detect the error, he must take the time to complain. Investors may, therefore, base com pensation on the expected value of the delay rather than on the promised delay. If so, the independent variables in equations ( 6 ) and ( 9 ) are incorrect proxies for the true values, and the estimated coefficients could be insignificant. In addition, some investors face different values of because of the procedures of their agents. For example, some brokers debit accounts for purchases on the trade date, but credit accounts for purchases only on delivery. As st promised st st The ordinary least squares residuals from equa tion ( 6 ) exhibit positive first-order serial correla tion, while higher-order autocorrelations are small. This is consistent with the use of an index as the dependent variable and with the results of Scholes and Williams (1977). To see this intui tively, note that some securities composing the index do not trade at the closing bell. The most recent prices for these securities are “stale.” If the market moves up or down since the last trade, these stale prices tend to move in the same direction when the securities subsequently do trade, inducing serial correlation at lag one. Therefore, we fit a first-order moving average to equation ( 6 ) and estimate (10) Rt = b 0 + b jAs, + u t , ut = et - 6et _ j . To formally investigate the possibility that the parameters in equation ( 1 0 ) may not be stable across subperiods, we conduct the test according to Chow (I960) for each subperiod partition. These tests show that both break points are T A B L E 2 F Regression Results Equation (1 0 ), Subperiods Estimates obtained by regressing the rate of return on the CRSP value-weighted index, including dividends (/?,), on the change in the settlement delay (As,), corrected for heteroscedasticity: (10) Rt= b 0 b ,As, + u t , F + ut - e t - 6el -1 • t First sample period: January 1, 1970 - October 6 , 1979 (2,467 observations) Variable Parameter estimate (^-statistic) x 10"4 (1.47)a 3 .1 0 b o b\ 7.20 x 10“ 4 Q 0.31 (I6.3)b Results Using the Change in the Length of the Delay (3.43)b Second sample period: October 7, 1979 - October 8 , 1982 (760 observations) Variable Parameter estimate (^-statistic) bo 5.60 x IO ' 4 ( 1.41 ) a b\ 8.73 x 1 0 - 4 (1.76)c e 0.17 (4.86)b Third sample period: October 9, 1982 - December 31, 1986 ( 1 , 0 6 9 observations) Variable F necessary. For the first partition, the -value is 7.64, which exceeds the 1 percent critical value of 3-78. For the second, the -value is 5.59, which again is significant at the 1 percent level. In addition, the test rejects the conjecture that the first and third subsamples can be combined. Because of weekends and holidays, holding periods range from one to four days. Given the results of French and Roll (1986), we would expect heteroscedasticity to be present, depend ing on the holding period for the observations. This proves to be the case. In the full sample, for example, the -ratio using the variance of the three-day holding period and the one-day hold ing period is 1 .3 1 , which exceeds the critical 1 percent value of 1.15. Similar results are found for both subperiods. Therefore, we weight observations by the inverse standard deviation of the residuals for the holding period in all reported results. Parameter estimate (^-statistic) b o b\ 6 7.61 x IO " 4 ( 2 .9 2 ) b x 10 -4 ( 1 .6 0 )a 5 .9 8 0 .1 0 (3.42)b a. Significant at the 10 percent level. b. Significant at the 1 percent level. c. Significant at the 5 percent level. NOTE: Significance levels are for one-tailed tests on b(j and bv SOURCE: Author’s computations. Table 1 contains the results obtained by estimat ing equation (10) using the full sample. Given the results of the Chow tests reported above, these estimates must be interpreted with caution, but we report them for completeness. All parameters have their expected signs and are sta tistically significant. The intercept, which esti mates the expected daily stock return in the absence of delays, implies an annual rate of about 11.80 percent. This is quite close to the actual realized value of 10.97 percent. The parameter estimates the rate of compensa tion for the settlement delay. This parameter is also significant, with a -statistic of 4.07. Table 2 contains the results from the subperi ods, which are broadly consistent with the full sample. For the first subperiod, the intercept is positive and significant at the 1 0 percent level, and is almost exactly the correct magnitude. The estimated value of . 0 0 0 3 1 0 implies an annual rate of about 7.81 percent; the actual value was 7.13 percent. The estimate of is reliably dif ferent from zero, with a -ratio of 3 43. After the first change in Federal Reserve oper ating policy, the results are somewhat different. The intercept is still marginally significant and again about the correct size (it implies a daily rate of 14.10 percent versus the actual 12.08 per cent). Despite being larger in magnitude, how ever, the significance of the slope coefficient is smaller. The -ratio is 1.76. The larger standard error is consistent with the smaller sample size and with the increased volatility during this bx c, t t t bx Regression Results Equation ( 1 1 ) , Full Sample Estimates obtained by regressing the rate of return on the CRSP value-weighted index, including dividends on the change in total return from an investment in federal funds during the settlement delay (ASS,), corrected for heteroscedasticity:___________________________________ _ (Rt), (11) Rt= b'0 + b{&XSt + u't, b y, t u \ - e't - d'e't _ j . t Full sample period: January 1, 1970-December 31, 1986 (4,296 observations) Variable Parameter estimate (^-statistic) bo For completeness, table 3 contains the results obtained by estimating equation ( 1 1 ) using the full sample. Again, all parameters have their expected signs and are statistically significant. The intercept, which estimates the expected daily stock return in the absence of delays, is very close to the value in table 1. The parameter j estimates the proportion of the federal funds rate that buyers receive as compensation for the settlement delay. This parameter is also significant, with a -statistic of 4.02. The coeffi cient of 2.73 is also reliably different from unity. A -ratio testing the hypothesis that the esti mated value equals one is 2 .5 5 , which rejects the null hypothesis at the 1 percent level. Thus, we reject the conjecture that the rate of compensa tion is the federal funds rate. The federal funds rate is too low or too stable to serve as the rate of compensation. Table 4 contains the estimates from the sub periods, which are again similar to those from equation (10). For the first subperiod, the inter cept is the same size and is equally significant as in table 2. The estimate of the slope coefficient, is 3-80. As is the case for the full sample, this is reliably different both from zero and from unity. The -ratios are 3.64 and 2.69, respectively After the first change in Federal Reserve oper ating policy, the intercept is still significant and again about the correct size, but the slope coeffi cient is much smaller. The estimated value is 1.84. This differs from zero at the 10 percent level, but unlike the case in the first subsample, it does not differ from unity. The -statistic is only 0.71. We cannot reject the hypothesis that the rate of compensation for settlement delays equals the average realized federal funds rate during the sample. The third sample begins on October 9, 1982. The results are similar to the second subsample and comparable to equation (10). The estimate of o is significant and implies a stock return of 19.18 percent, compared to the actual value of 19.07 percent. The estimated slope coefficient, \, is 2.59, which differs from zero at the 5 per cent level, but does not differ from unity. The -statistic is only 1.03. The results suggest that during the first sample, the Federal Reserve’s intervention in the federal funds market prevented the federal funds rate from tracking market conditions as well as it did during periods when the Federal Reserve concen trated on other policy vehicles. When the federal 4.69 x 1 0 - 4 (3.03)a 2.73 (4.02)a d' 0.23 (15.6)a b \, a. Significant at the 1 percent level. NOTE: Significance levels are for one-tailed tests on b’} and b\. SOURCE: Author’s computations. period, when the Federal Reserve did not attempt to stabilize interest rates. The third sample begins on October 9, 1982. The results of this subsample are similar to those of the second subsample. The estimate of implies a stock return of 19.17 percent; the actual value was 19.07 percent. The -value of 2.92 is significant at the 1 percent level. The estimated slope coefficient is 0.000598, which differs from zero at the 1 0 percent level. b0 t Results Using the Change in the Opportunity Cost During the Delay The preliminary tests using equation (9) yield results similar to those of equation ( 6 ). Chow tests confirm that the subperiods are best esti mated separately. Heteroscedasticity is again present, and a first-order moving average is required. We estimate t t b b t Regression Results Equation ( 1 1 ) , Subperiods Estimates obtained by regressing the rate of return on the CRSP value-weighted index, including dividends on the change in total return from an investment in federal funds during the settlement delay (A corrected for heteroscedasticity: (R,), XSt), Rt= b'0 + b{AXSt + u't , (11) u \ - e \ - d'e't _ j . First sample period: January 1, 1970 - October 6, 1979 (2,467 observations) Variable Parameter estimate (^-statistic) bo b\ Q' x IO ' 4 (1.46)a 3 .1 0 3.80 (3.64)b 0.31 ( l 6 .2 ) b Second sample period: October 7, 1979 - October 8, 1982 (760 observations) Variable Parameter estimate (^-statistic) K 5.62 x 1 0 - 4 (1.41 )a b\ 1.84 (1.55)a Q’ 0.17 (4.85)b Third sample period: October 9, 1982 - December 31, 1986 (1,069 observations) Variable Parameter estimate (f-statistic) b'o 7.61 x IO ' 4 (2.92)b b\ 2.59 ( 1 .6 8 )c 6' b\ not b t 0 .1 0 (3.44)b a. Significant at the 10 percent level. b. Significant at the 1 percent level. c. Significant at the 5 percent level. NOTE: Significance levels are for one-tailed tests on SOURCE: Author’s computations. funds rate is permitted to float freely, we cannot reject the notion that stock purchasers compen sate sellers for the settlement delay at the federal funds rate. However, when the central bank intervenes, the federal funds rate appears to be too stable to serve as the rate of compensation. Since the estimates of exceed unity, they are higher than predicted by Lakonishok and Levi (1982), who argue that delays should be compensated at the riskless rate. To the extent that the overnight federal funds rate is riskless, the coefficient should be one if Lakonishok and Levi are correct. The results in table 4 are, how ever, consistent with their empirical results. Lakonishok and Levi assume that settlement and check-clearing delays are priced at the prime rate and test to see if the prime rate is large enough to explain the weekend effect. Although a strict interpretation of their story requires that sellers be compensated at the riskless rate, they report that the prime rate is too low to eliminate these effects completely. This suggests that if the set tlement and check-clearing delays were in fact the sole reason for the weekly pattern, rates of compensation during these delays must be larger than the riskless rate. Since our results apply only to the settlement delay and not to the check-clearing delay, they do not directly relate to those of Lakonishok and Levi. However, they do suggest the possibility that rates of compensa tion are larger than the riskless rate. Conceivably, though, the rate of compensation should be the riskless rate: errors in posting to brokerage or bank accounts do occur. While restitution is always made if the error is caught, the seller may not notice it. Even if he does, complaining is time-consuming. The seller may therefore require a premium over the riskless rate. In addition, the buyer may very well be wil ling to pay this premium. If he monitors his account, it cannot be debited early, but through bank or brokerage error, it may be debited late. Since the buyer can only win, he is willing to pay extra for this possibility. Using the brokers’ call money rate as the interest rate proxy7 would probably produce smaller values of This rate tends to be higher than the federal funds rate, so smaller propor tions of the call money rate imply the same lev els of compensation. If the call money rate is as variable as the federal funds rate, -tests would be less likely to reject the notion that the rate of compensation is the call money rate. b'() and b\. IV. Conclusion This paper shows that investors consider delivery procedures in pricing stocks. We model stock returns in two ways. The first uses a function of the length of the settlement delay, while the second uses a function of both the length of the delay and interest rates during the delay. We find that the coefficient on this variable is always cor rectly signed and statistically significant. This means that observed prices diverge from the prices that would be observed in the absence of this trading mechanism. This, in turn, means that measured returns diverge from true returns. While this result is comforting to researchers who have assumed that settlement delays are priced, it does have implications for empirical studies using daily stock-return data. Since the observed price equals the true price plus a pre mium to compensate for financing costs, meas ured returns diverge from true returns if the pre mium changes during the holding period. This could, for example, affect event studies either by masking the impact of a true economic event or by lending statistical significance to “events” which result only from changes in the premium and not from any underlying economic force. References Choi, Dosoung and Strong, Robert A., “The Pric ing of When-Issued Common Stock: a Note,” September 1983, 1293-98. Journal of Finance, 38, Chow, Gregory C., “Tests of Equality' Between Sets of Coefficients in Two Linear Regres July I960, 591-605. sions,” Econometrica, 28, Cross, Frank, “The Behavior of Stock Prices on Financial Analysts Fridays and Mondays,” November/December 1973, 67-9. Journal, DeGennaro, Ramon P., “Payment Delays: Bias in Journal of Money', Credit 20, 684-90. the Yield Curve,” November 1988, and Banking, ------ , “The Effect of Payment Delays on Stock Prices,” 1 9 9 0 (forthcoming). Journal of Financial Research, Flannery, Mark J. and Protopapadakis, Aris A., “From T-Bills to Common Stocks: Investigat ing the Generality7 of Intra-Week Return Sea sonality,” June 1988, 431-50. Journal of Finance, 43, French, Kenneth R., “Stock Returns and the Journal of Financial Eco 8, 55-69. Weekend Effect,” March 1980, nomics, and Roll, Richard, “Stock Return Var iances: the Arrival of Information and the Reaction of Traders,” September 1986, 5-26. ----------- Economics, Journal of Financial 17, Gibbons, Michael R. and Hess, Patrick, “Day of Journal the Week Effects and Asset Returns,” October 1981, 579-96 . of Business, 54, Hite, Gailen L. and Owers, James E., “Security Price Reactions Around Corporate Spin-off Announcements,” December 1983, 409-36. nomics, Journal of Financial Eco 12, Jakus, Judy and Chandy, P.R., “After Your Order: Financial 4, the Saga of a Stock Transaction,” Summer 1989, 13-16. Management Collection, Lakonishok, Josef and Levi, Maurice, “Weekend e," Journal of Effects on Stock Returns: a Not 883-89. Finance, June 1982, 37, Lakonishok, Josef and Smidt, Seymour, “Are Sea sonal Anomalies Real? A Ninety-Year Perspec tive,” Winter 19 8 8 , 403-25. Revieu' of Financial Studies, 1, McCulloch, J. Huston, “An Estimate of the Li Journal of Political Econ quidity7 Premium,” February 1975, S3, 95-119. omy, Scholes, Myron and Williams, Joseph, “Estimat Jour ing Betas from Nonsynchronous Data,” December 1977, 5, 309-27. nal of Financial Economics, The E ffe c t of Bank Structure and Profitability on Firm Openings by Paul W. Bauer and Brian A. Cromwell Paul W . Bauer and Brian A . Crom well are economists at the Federal Reserve Bank of Cleveland. The authors thank Randall Eberts, Kathe rine Sam olyk, Jam es Thomson, Gary Whalen, and David Whitehead for useful discussion and suggestions. Ralph Day and Lynn Downey pro vided valuable assistance with the data and systems. Fadi Alameddine and Kristin Smalley provided excel lent research assistance. Introduction The banking industry has undergone significant changes in recent years. Much attention has been given to the effect of financial deregulation and interstate banking on the structure of the bank ing industry7. Attention has also been directed at the systematic effects of financial structure on the national economy. However, bank structure can also affect local economic development. 1 The availability and the cost of financing potentially varies across regions due to differences in bank structure and in the health of the local banking sector. Since bank credit is an important source of financing for new firms, differences in bank structure can affect regional growth. This paper examines the effects of bank struc ture and profitability on the birth of new firms, an important component of economic develop ment. Specifically, we enter measures of profita- ■ 1 W e use the term “bank structure" to refer to both the organization of banks themselves (number of branches, employees per bank, etc.) and the market structure of the banking sector (concentration, ease of entry, etc.). bility, concentration, size, and entry of a region’s banking sector (as well as an overall measure of lending activity) into a standard model of firm location. This enables us to test for independent effects of bank structure and profitability on re gional growth, as measured by business openings. Our results suggest that bank structure and profitability have significant effects on firm open ings. A profitable and competitive banking mar ket is associated with a higher rate of firm births. In particular, firm births are found to be asso ciated with higher bank profits, higher numbers of bank employees, lower levels of concentration, higher proportions of small banks, and freer entry of new banks into the region. These results support the position that bank structure and profitability influence economic development. Section I briefly reviews previous work relat ing banking and economic activity and discusses the implications of bank structure for regional growth. Section II presents a standard model of firm location and extends it to include measures of bank structure and profitability. Section III describes the data, and section IV provides results on the impact of banking on firm loca tion. Finally, section V presents conclusions. I. Bank Structure and Regional Growth With the advent of deregulation and interstate banking, the banking industry has changed sig nificantly in recent years. Much attention has been given to the effects of these developments on the structure of the banking industry itself. 2 Attention has also been directed at the systematic effects of bank failures and financial structure on aggregate economic activity. 3 The effect of changes in bank structure on regional econo mies, however, remains an open question.4 For example, Eisenbeis (1985), in a recent article on interstate banking, comments that: The most controversial issues surrounding considera tion of modifying interstate banking laws deal with the implications of proposed changes for competi tion and concentration of resources. There is little doubt that restrictions on geographic expansion have, in the past, insulated many local markets from competition and have restricted economic growth. While casual inspection of the data suggest that states with more liberalized policies toward intrastate bank ing have generally had higher economic growth rates than unit banking states, empirical studies show no convincing relationship between banking structure and economic development. More detailed study would have to be done to determine whether this is just a matter of correlation or causation, (p. 231-32) ■ 2 For example, Lee and Schweitzer (1989) use event-study analysis to determine the effect on stock prices of decisions by bank holding companies (B HCs) to establish subsidiaries within Delaware and find no evidence of long term stock price changes during the postannouncement period. Trifts and Scan lon (1987) use a sample of interstate mergers to provide early evidence of the effects of interstate bank mergers on shareholder wealth. Born, Eisenbeis, and Harris (1988) provide evidence on the market evaluation of financial firms entering into interstate banking when restrictions are relaxed and find no signif icant effect of an announced geographic interstate expansion on shareholder values. ■ 3 Gertler (1988) provides an overall review. Bernanke (1983) argues that extensive bank runs and defaults in the 1930-1933 financial crisis reduced the efficiency of the financial sector in performing its intermediation function, caus We approach this issue by studying the effect of bank structure on business openings. If bank structure and the health of the local banking sec tor affect the cost and availability of credit for new firms, changes in bank structure will poten tially affect regional growth. Financial institutions, especially banks, are the primary supplier of external funds to new busi nesses, which are typically small, independent enterprises. Unlike medium-sized (100 to 500 employees) or large corporations, small busi nesses have limited access to organized open markets for stocks, bonds, and commercial paper. Approximately three of every four existing small businesses have borrowed from banks. 5 The availability of credit at affordable rates for the start-up and the continued operation of new firms is not necessarily a given. 6 For small start up firms (typically “mom and pop” operations), financing comes mostly from private sources, such as personal savings, home equity loans, and loans from friends or relatives. For larger small businesses, capital for start-ups comes from financial institutions and organized venture capi tal firms, as well as from friends, relatives, and informal investors. Even after being established, firms may require financing when cash inflow lags behind cash outflow due to a rise in receiv ables or an inventory buildup. When external financing is used, it is received primarily from commercial banks. The rates charged for small start-up firms are typically 2 to 3 percentage points above that charged for larger firms. This is due in part to the high-risk nature of new small businesses, which lack collateral and a credit history and suffer high rates of failure. Some researchers and many policymakers argue that banks do not meet the needs of various types of businesses, particularly small businesses. They contend that due to high monitoring costs and a lack of adequate information about risk, a market failure exists— popularly referred to as the “credit gap.” It has been argued that the price of credit, especially working capital, pro vided to small and middle-sized firms is too high after controlling for appropriate risk factors. The ing adverse effects on real output, other than through monetary channels. Samolyk (1988) conducts a similar test on British data, using corporate and noncorporate insolvencies as proxies for the health of the financial sector, and also finds that credit factors matter empirically on output. Gilbert and Kochin (1990, forthcoming) provide additional tests of the hypothesis that bank fail between savers and investors. They argue that the role intermediaries play in ures have adverse effects on economic activity using rural county-level data improving the efficiency of intertemporal trade is an important factor governing and find that closing banks has adverse effects on local sales and nonagricul- general economic activity. The correlation between economic development and tural employment. financial sophistication across time and across countries has often been noted. See Goldsmith (1969) and Cameron (1972) for examples of such studies. ■ 4 A s discussed in Gertler (1988), the literature on financial structure and economic development has principally focused on variations across countries. ■ 5 Small Business Administration (1985), p. 206. supply process. They note that in developed countries there typically exists a ■ 6 Current information is not available on the sources of internal financing highly organized system of financial intermediation facilitating the flow of funds to small firms. For historical data, see Small Business Administration (1984). Gurley and Shaw (1955) emphasize the role of intermediaries in the credit credit gap is aggravated in times of tight credit, during which banks ration funds, with larger firms receiving a disproportionately large share. This perception of market failure is reflected in how public-sector development agencies lower the cost of credit by providing access to sheltered pools of money (such as public pen sion funds), by passing on the favorable tax treatment of funds (through tax abatement and public bonds), or by accepting risks greater than private institutions are willing to bear (such as the loan guarantee program of the Small Busi ness Administration) . 7 While there are no direct measures of the price and availability of credit for small busi nesses across regions, they are likely to vary with bank structure. 8 Concentrated banking markets with large banks and high barriers to entry may be unresponsive to the credit needs of small businesses and new firms. Lending to new firms entails higher risks than lending to established firms, since a large proportion of new firms fail in the first few years. Heggestad (1979), Rhoades and Rutz (1982), Clark (1986), and Liang (1987) argue that banks in highly concentrated markets trade potential monopoly profits for lower risk. Alternatively, a highly competitive bank market, characterized by large numbers of smaller banks and easy entry, may result in a greater availability of credit at lower prices for small businesses. Finally, a prof itable banking sector is expected to result in less credit rationing and a greater supply of credit for small firms. Even if most start-ups do not rely directly upon commercial banks for their initial financing, the expectation of ample credit for future expansion at low cost potentially affects the decisions of entrepreneurs to start a firm. 9 An understanding of the impact of bank struc ture on firm kx:ation and regional growth is important because of the significant changes occurring due to deregulation and interstate banking. By the end of 1988, all but three states ■ 7 See Hill and Shelley (1990, forthcoming). ■ 8 This would not be true if banks were perfectly contestable; the actual permitted some form of interstate acquisition of their banks, 14,600 offices of banking organiza tions existed outside the organizations’ home state, and more than half of these were permit ted to offer all banking services. 10 To the extent that this results in freer entry and increased competition among banks, the availability of cap ital for small businesses and new firms could increase. In the Southeast and New England, however, these developments have increased the number of extremely large banks, called “superregionals,” at the expense of regional banks. Increased concentration could reduce the supply of credit for small businesses. A recent survey of state bank regulators by Hill and Thompson (1988) found that advancing eco nomic development is an important goal of state bank regulators. 11 If changes in bank structure do indeed affect regional growth, however, policy makers may be misjudging the costs and benefits of deregulation and interstate banking. We now turn to an empirical analysis of this issue. II. A Model of Firm Location To study the effect of bank structure and profita bility on local economic activity, we concentrate on firm openings because they are driven by current and expected economic conditions, as opposed to expansions, contractions, and deaths, which will be greatly affected by the large fixed costs associated with changing locations. The model estimated here was originally developed by Carlton (1979), though we more closely fol low Eberts and Stone (1987).12 The number of new establishments in a city is assumed to depend on the number of potential entrepreneurs in the city and on the probability that a given entrepreneur will start a new firm. The higher the level of economic activity in a city, the greater the number of potential entre preneurs. Also, the higher the expected profita bility of new firms, the larger the probability that they will actually emerge. number and size distribution of competitors would not affect the price or the ■ availability of credit. Whalen (1988) found that there is evidence that bank per banking by King et al. (1989). Earlier surveys include Whitehead (1983a, formance is systematically related to proxies designed to measure the inten 1983b, and 1985), and Am el and Keane (1986). 10 These figures come from a recent comprehensive review of interstate sity of actual and potential competition in rural banking markets in Ohio and concludes that these n o n -S M S A banking markets are contestable, since ■ potential competition matters, but are not perfectly contestable. Our results positors’ funds and providing banking (depository) services throughout their suggest this m ay be true for S M S A s as well. states. ■ 9 Unfortunately, we do not have measures of sources of funds from non bank entities, which potentially compete with commercial banks. ■ 11 12 It ranked third, just behind ensuring the safety and soundness of de For reviews of the firm-location literature, see Bartik (19 8 5 ,19 8 8 ), Wasylenko (1988), and Wolkoff (1989). Carlton (1979) modeled this birth process as a Poisson probabilistic model, since the birth of new establishments is a discrete event. Let be the probability that a potential entrepreneur will start an establishment in a given city; then let ln p t= x (b + ejy i - ,..., M , where x i is a vector of independent variables affecting firm profitability, b is a vector of fixed (1 ) 1 coefficients, c; is an error term composed of the variance of the Poisson process and a random error, and is the number of cities in the sam ple. Consistent estimates of the mean and var iance of are given by M pl (2) E (p t) = (.Nt/BP ,), (3) Var(P') = (Nj/BPf), where iV; is the observed number of births and is the birth potential as proxied by employ ment in the standard metropolitan statistical area (SMSA) . 13 Carlton shows that a consistent and asymptotically efficient estimate of can be obtained by weighted least squares, with weights equal to the standard error of the Poisson process. The independent variables typically used to measure expected profitability include wage rates, tax rates, unionization rates, and energy7 prices. We extend this list by including measures of bank structure and profitability. As discussed in the previous section, these measures deter mine, at least in part, the price and availability of credit and thus expected profitability and firm openings. Measures of bank structure and profit ability are employed because direct measures of the price and the availability of credit are unavailable. To control for the effects of bank structure and the availability of credit on firm births, we include measures of the number and size distribution of banks as well as a measure of the financial health of banks. BP\ b 1980 to 1982 in the USELM data) to existing employment in the SMSA14 A birth is defined as an establishment that did not exist in 1980 but did exist in 1982. Births within this two-year period are treated as comparable. We divide the independent variables into two types. The first are measures of local economic conditions, and the second are measures of bank structure and profitability. All data are measured at the SMSA level unless otherwise noted. The measures of local economic activity are the natural logs of the wage rate ( the num ber of establishments the gross state product the personal income and the population Also included is the effective state corporate tax rate ( We control for population by entering it directly into our equation rather than using per capita varia bles that would impose additional structure. Bank data are obtained from the Federal Financial Institutions Examination Council’s Reports on Condition and Income, known as call reports, for 1980. (We assume that the lagged 1980 variables on banking are exogenous to firm births occurring between 1980 and 1982.) Meas ures of bank structure and profitability are created by aggregating data from individual banks up to the SMSA level. The total amount of loans and leases is a measure of the level of bank intermediation. The average rate of net income divided by assets, return measures the amount of resources available for future lending and the health of the banking sec tor. 16 This variable may also be measuring the effects of bank structure and the general eco nomic health of the region. The empirical analy sis will thus explicitly control for these effects. We employ standard measures of market struc ture such as the total number of banks and branches the number of bank employees per bank and a Herfin dahl index of the concentration of deposits ) . 1 7 We also include a measure of bank (GSP), (LOANS) (RETURN), (BRANCH), (BANKEMP), (HQS) (HERE ■ III. Data WAGE), (FIRMS), (PINC), (POP). TAX).15 14 U S E L M stands for the U .S . Establishment and Longitudinal Microdata file constructed for the Small Business Administration by Dun and Bradstreet. Data from 259 SMSAs across the country7 are employed to estimate the model. The dependent variable ( ) is the natural log of the ratio of new firm births (as reported for the years BIRTHRATE ■ 15 WAGE and M X GSP, PINC, and POP are are 19 77 variables from the Census of Manufactures. 1980 variables from the Census Bureau and the Department of Commerce. The number of establishments is a 1980 variable from the U S E L M data. ■ 16 Specifications using income divided by equity capital yield similar results. ■ ■ 13 17 The Herfindahl index is defined as the sum of the square of each Although policymakers concerned with economic development value bank’s share of deposits for a given S M S A . While we are interested in the the employment resulting from new firms, the firm location literature explicitly effect of concentration in the lending market, we assume that deposits are models the birth of the firm itself. Using job creation (instead of firm births) as subject to less geographic dispersion than loans, and thus provide a more the dependent variable, however, yielded similar results. accurate indicator of concentration in the local banking sector. T A B L (ENTRY E Descriptive Statistics Variable Mean Standard Deviation BIRTHRATE (firm 0.008 0.003 5.986 1.183 0.403 0.039 13,150 24,713 635.4 1 0 6 0 .2 2,656.4 9,411.5 0.009 0.003 birth/ employment) WAGE (manufacturing) TAX (effective tax rate) FIRMS (number of establishments) POP (population, , thousands) LOANS (total loans and leases, millions) RETURN (net income/assets) HQS (number of banks) BRANCHES 23 39 132 252 1 9 6 .8 324.6 2,499 1,849 0.456 0.224 0.180 0.129 0.084 0.092 0.058 0 .1 0 0 0.042 0.073 0.028 0.081 -0.014 0.156 6,740.4 12,413.0 (number of branches) BANKEMP (employees/bank) HERF (Herfindahl concentration index) SIZE 1 (percent of banks with $0-$25 million assets) SIZE 2 (percent of banks with $25-$50 million assets) SIZE 3 (percent of banks with $50-$75 million assets) SIZE A (percent of banks with $75-$ 100 million assets) SIZE 5 (percent of banks with $100-$250 million assets) SIZE 6 (percent of banks with $250-$400 million assets) ENTRY (percentage change in the number of banks) PINC (personal (SIZE I-SIZE SIZE SIZE SIZE SIZE SIZE 5 SIZE LOANS, IV. Estimation and Results income, millions) GSP (gross state entry ), the percentage net change in the number of banks from 1978 to 1980.18 Our last measures of bank structure are a set of variables 6 ) that control for the size of banks. 1 is the proportion of banks in an SMSA with assets less than $25 million, 2 is the proportion of banks with assets between $25 and $50 million, 3 is the pro portion of banks with assets between $50 and 4 is the proportion of banks $75 million, with assets between $75 and $100 million, is the proportion of banks with assets between $100 and $250 million, and 6 is the propor tion of banks with assets of $250 to $400 million. The proportion of banks with assets greater than $400 million is the omitted category in our esti mations. 19 Summary statistics for these variables are presented in table 1 . A pervasive problem with this data set for the purpose of looking at how banking activity affects the regional economy is that regions for which data are collected (SMSAs and states) and economic regions do not necessarily match. In addition, for some variables, such as though the total dollar value of loans is known, it is not possible to determine where the loans were made. For example, loans made by an Ohio bank to firms in Florida and Ohio are counted in the same way. With the banking data, there is an additional measurement problem in that a call report for a consolidated banking unit may include data for branches not located in the SMSA In states that allow branch banking, activity at the branches may be reported solely in the SMSA headquarters. Thus, our measures of competition and concen tration are potentially subject to errors. The sen sitivity of our full sample results to this potential errors-in-variables problem is tested by running the model without SMSAs in states that have state wide branching, and then again without SMSAs in states that have limited branching (that is, only SMSAs in unit banking states). Full Sample Results 100,680 product, millions) 84,277 Estimates of variations of the above model for the full sample are presented in table 2. Equa- NOTE: Changes are measured as log differences. SOURCE: Authors’ calculations. ■ 18 Note that this measure treats entry and exit symmetrically. ■ 19 Alternative measures of size were also tested. In general, only the measures of the smaller banks were statistically significant. ■ 1 A B L E 2 Estimation Results Coefficient W AGE (1 ) (3 ) (2 ) -0.6823a (0.1131) -0.44263 (0.1023) -0.50763 (0.1140) TAX -1 .8 3 6 8 a (0.5694) -1.70323 (0.5442) -1.51933 (0.5490) F IR M S 0.28253 (0.0940) 0.3453a (0.0939) 0.30463 (0.1090) POP -0.24123 (0.1015) -0.l694b ( 0 .1 0 0 2 ) -0.35323 (0.1692) -0.0393 (0.0870) -0 . 0 6 0 2 (0.0872) 31.7890a ( 6 .8 2 3 8 ) 31.29403 (6.8055) -0.0693 (0.1294) -0.0451 (0.1293) -0.227 l a (0.0555) -0.19453 (0.0574) 0.3192a (0.0942) 0.31913 (0.0938) -0.19873 (0.0687) -0.19113 (0.0684) 0.86503 (0.2463) 0.85503 (0.2450) 0.3396 (0.2537) (0.2525) 0.4889b (0.2746) 0.4486 (0.2742) 0.4387 (0.2688) 0.4101 (0.2677) -0.0085 (0.3159) -0.0432 (0.3146) -0.0803 (0.2784) -0 . 0 8 1 6 (0.2770) 0.4314a (0.1319) 0.42393 ( 0 .1 3 1 2 ) LOANS — RETURN — HQS — BRANCHES — BANKEM P — HERF — S IZ E 1 — S IZ E 2 — S IZ E 3 — S IZ E A — S IZ E 5 — S IZ E — 6 ENTRY — P IN C — — 0 .3 1 6 8 0 .1 8 3 8 (0.1785) GSP CO N STANT Log likelihood function — — -4.05023 (0.4267) -4.6572a (0.7856) 0.0427b (0.0239) -6.3725a (1.5336) -95.4467 -46.6358 -44.1093 0.2109 0.4579 0.4683 Mean of the dependent variable -4.9267 -4.9267 -4.9267 259 259 R-square No. of obs. 259 a. Significant at the 95 percent confidence level. b. Significant at the 90 percent confidence level. NOTE: Standard errors of the coefficients appear in parentheses. SOURCE: Authors’ calculations. tion ( 1 ) is a basic, static model of firm location, where the probability that a birth will occur depends on the wages, taxes, number of estab lishments, and population. This set of variables differs somewhat from that employed by Carlton (1979), who also used the unionization rate and energy prices in his estimates for selected indus tries. Eberts and Stone (1987) found that energy prices do not matter when the model is esti mated with aggregate manufacturing data, and it is even less likely that energy prices would mat ter since we are looking at all industries. Because we are not concerned about differ ences across industries and are interested only in whether there are statistically significant effects on aggregate regional economic activity as a result of bank structure and profitability, energy prices can safely be omitted. The unionization rate was omitted due to lack of available data. We assume that unionization is not systemati cally related to the banking variables. All the coefficients in equation (1) are statisti cally significant at the 95 percent confidence level. As expected, we find that higher wages and higher effective corporate tax rates reduce the probability of firm births in an SMSA. Also, the probability of firm births increases with a greater number of establishments and a lower population. Though the coefficient on population is somewhat unexpected, this result suggests that given the similar magnitude and opposite signs of these two coefficients, perhaps the number of firms per capita is the appropriate regressor. We continue entering population as a separate regressor because this is the most gen eral way of including population in the model. 20 Equation (2) estimates the same model, only now the measures of bank structure and profita bility are included. The results strongly support the view that bank structure and profitability have a statistically significant effect on firm births. The addition of the bank structure varia bles did not affect the estimates of the basic firm location variables. The basic firm location coeffi cients have roughly the same magnitude and remain statistically significant at the 90 percent confidence level or higher. The measure of the total amount of financial intermediation is negative but not sta tistically significant. The variable has a positive and statistically significant coefficient, (FIRMS) (LOANS) ■ 20 RETURN More restrictive specifications using per capita variables yielded sim ilar results. ■T A B L E 3 Unit and Limited Branching States Coefficient WAGE TAX FIRMS POP LOANS (1 ) -0.45593 (0.1075) -0.46103 (0.1340) -3.04843 (0.6175) -1.50433 (0.6943) -0.7901 (0.8031) 0.44373 ( 0 .1 1 3 2 ) 0.40133 (0.1392) 0.4063a (0.1654) -0.43373 (0.1224) -0.30013 (0.1367) -0.3458b (0.2088) -0 . 1 1 6 2 (0.1352) -0 . 1 6 1 2 (0.1371) 44.34303 (9.9812) 43.40403 (9.9638) 0.1324 ( 0 .2 0 0 0 ) 0.2018 (0.2031) -0.27783 (0.0735) -0.26473 (0.0736) 0.54933 (0.1412) 0.58173 (0.1419) -0.21633 (0.0863) -0.21043 ( 0 .0 8 6 1 ) 1.24283 (0.3579) 1.22873 (0.3569) 0.70643 (0.3454) 0.6672b (0.3449) 0.86703 (0.3380) 0.86773 (0.3370) 0.94563 ( 0 .3 2 8 1 ) 0.94593 (0.3270) 0.7980b (0.4074) 0.7962b (0.4068) — — RETURN — HQS — BRANCHES — — BANKEMP — — HERF — — SIZE 1 — — SIZE 2 — — SIZE 3 SIZE A SIZE 5 SIZE — — — — — — 6 ENTRY — — — — PINC GSP CONSTANT Log likelihood function (3 ) (2 ) -0.75583 (0.1137) (0.4510) 0.1004 (0.4527) 0.1757 (0.2295) 0.1948 (0.2311) 0 .0 3 6 0 — — — — — — — — -37568a (0.4690) -5.16423 (1.0234) 0.0108 (0.2472) -5.92763 (1.9894) -19.2143 -17.4198 0.3675 0.5569 0.5652 Mean of the dependent variable -4.9699 -4.9699 -4.9699 No. of obs. 190 HERF HQS BRANCHES, BANKEMP, HERF), SIZE ENTRY In equation (3), two more measures of regional activity and are added to the model to see whether the bank structure and profitability effects are merely reflecting regional economic conditions. O f the added regressors, only is statistically significant and only at the 90 percent confidence level. The bankrelated coefficient estimates do not change appreciably with the addition of these regressors. In particular, retains its positive and sta tistically significant value even when we control as much as possible for local economic condi tions, suggesting that this variable is doing more than just reflecting a robust local economy. 21 (PINC GSP) GSP RETURN 0 .0 6 6 l b (0.0372) -53.0456 R-square suggesting that (controlling for structure) a prof itable banking sector is associated with a higher probability of firm births. Profitable banks could have more opportunities for providing interme diation services and engage in less credit ration ing, suggesting a positive relationship with firm births. Alternatively, high profits in the banking sector could merely be indicating profitable market conditions for other industries as well. (We will therefore control for regional economic activity in equation [3 ].) The number of banks ( ) is not statistically significant, but and are, suggesting that the greater the number of branches and the more concentrated the banking market (at least as measured by the lower the probability of firm births. More branches could reflect more of a retail orientation of the banks. Also, the more employees per bank, the higher the probability of firm births. The statistical significance and the magnitude of 1 suggest that smaller banks (those with less than $ 5 million in assets) are more involved in firm births than larger banks: the higher the proportion of small banks, the higher the proba bility of firm births. Finally, the coefficient on is positive and statistically significant, implying that the more contestable the banking market (as indicated by a larger value for entry), the higher the probability of firm births. 190 a. Significant at the 95 percent confidence level. b. Significant at the 90 percent confidence level. NOTE: Standard errors of the coefficients appear in parentheses. SOURCE: Authors’ calculations. 190 Partial Sample Results As previously discussed, the banking data are potentially subject to significant measurement ■ 21 Specifications that included the complete set of economic variables but entered the various bank structure variables separately (instead of the full set) yielded similar results. A n exception was our measure of concentration, HERF, which was statistically significant only when the included as well. SIZE variables were Unit Banking States Coefficient WAGE TAX FIRMS POP LOANS (1) -0.88473 (0.1994) (2 ) -0.54943 (0.1951) -0.34663 (0.2724) -1.6874 (1.0677) -0 . 2 8 1 6 (0.9922) -0.9859 (1.7693) 0.51933 (0.1778) 0.3525 (0.2747) 0.5890b (0.3543) 0.50293 (0.1885) 0.0184 (0.2915) 0.2364 (0.3563) 0.2934 (0.3359) 0.1598 ( 0 .3 6 0 6 ) — — RETURN — HQS — BRANCHES — — BANKEMP — — HERF — — SIZE 1 — — SIZE2 — — SIZE 3 — — SIZE A — — SIZE 5 — — SIZE 6 — — ENTRY — — PINC GSP CONSTANT Log likelihood function R-Square 3 6 .6 8 0 0 b (22.1410) 43.88105 (23.4160) -0.4136 (0.6288) -0.1035 (0.6956) -0.3807b ( 0 .2 1 3 6 ) -0.4427b (0.2367) 0.0810b (0.4796) 0.1937 (0.5147) -0.1543 (0.2107) -0.0565 (0.2396) 2.71953 ( 1 .3 6 6 2 ) 2.5134b (1.4066) 1.9879 (1.2694) 1.7754 (1.3086) 2.34523 (0.9367) 2 .2 6 0 1 3 (0.9560) 0.7998 (1.1518) 0.7543 (1.1646) 2 .0 3 0 0 b (1.0934) 1.7276 (1.1633) (1.0377) 1.1365 (1.0511) 1.58433 ( 0 .6 2 3 8 ) 1.36823 (0.6601) 1 .1 3 8 6 — — — — — — — — -4.2875a (0.6673) -13.6582 (3) -0.4996 (0.4562) -10.08503 (2.8175) -5.8005 (4.9151) 1 2 .8 3 2 6 13.7363 0.4021 0.7603 0.7677 -4.7994 -4.7993 58 In table 3, we reestimate the model omitting SMSAs in states with statewide branching. 22 Although the magnitude of the coefficients tends to be larger, there is no qualitative change in the results in equation (1). In equation (2), the results are again quite similar to those in table 1 , except that more of the size variables are statisti cally significant, but ENTRY is no longer statisti cally significant. These differences carry over to the results for equation (3). Thus, omitting the SMSAs in the statewide branching states has little effect on our results. Though we remove most of the measurement problems in the banking variables by omitting the SMSAs in the statewide branching states, the same problems hold to a much lesser degree for the SMSAs in the states with limited branching, which generally allow branches to operate only in contiguous counties. In table 4, the model is reestimated with only the SMSAs in the unit banking states. 23 These sta tistical results are not as strong, but our sample has fallen from 259 in table 2 , to 190 in table 3, to only 58 in table 4. O f the bank structure and profitability variables (reported in equation [2 ]), 1, 3, and 5 all remain statistically significant. and lose their statistical significance, but once again becomes statistically signifi cant. When we add and in equation (3), is no longer statistically significant, but the number of establishments is. O f the banking variables, RETURN, BRANCHES, SIZE SIZE SIZE BANKEMP HERE ENTRY PINC GSP WAGE (FIRMS) RETURNS, BRANCHES, -0.0231 (0.0741) Mean of the dependent variable -4.7987 No. of obs. error. In states that permit statewide branching, a call report for a consolidated banking unit may include data for branches not located in the SMSA. While the standard errors-in-variables problem in econometrics results in a bias toward zero in the estimated coefficients, we wanted to test whether our results were due to measure ment error. We therefore estimate the model without SMSAs in states that have statewide branch banking, and then again without SMSAs in states that allow statewide or limited branch ing. These results are reported in tables 3 and 4. ■ 22 Thus, we omit S M S A s in the following states: Alaska, Arizona, Cali fornia, Connecticut, Delaware, Florida, Hawaii, Idaho, Maine, Maryland, N e v ada, New Hampshire, New Jersey, New York, North Carolina, Oregon, Rhode Island, South Carolina, South Dakota, Utah, Vermont, Virginia, and 58 a. Significant at the 95 percent confidence level. b. Significant at the 90 percent confidence level. NOTE: Standard errors of the coefficients appear in parentheses. 58 Washington. ■ 23 Thus, only S M S A s in the following states are included in this sample: Colorado, Illinois, Kansas, Missouri, Montana, Nebraska, North Dakota, Okla homa, Texas, West Virginia, and Wyoming. SIZE SIZE ENTRY 1, 3, and all remain statistically significant. In the basic firm-location model (equation [1 ]), the coefficients retain the same signs and magnitudes, though the state corporate tax rate ( ) is no longer statistically signifi cant. When we add the bank variables, only retains its statistical significance. Clearly, the model does not perform as well with this sample. Even the coefficients in the basic firm location model lose their statistical significance (except for Whether this is due to the small sample size or to possibly pecu liar characteristics of the included SMSAs is unclear. 24 Yet even with this sample, bank struc ture (as measured by 1, 3, and retains a statistically significant effect on firm births. In summary, the error-in-variables problem discussed in the previous section does not appear to severely bias our results. Estimates of the model using the full sample are very similar to the estimates obtained using only SMSAs in states with unit or limited branching. When the model is estimated with just the SMSAs in unit branch banking states, the estimates change much more, but the profitability of the banking sector, the number of branches, the proportion of small banks, and entry all have a statistically significant effect on the probability of firm births. Our measure of concentration ( retains the same sign and magnitude but is not statisti cally significant. Banking structure and the avail ability of credit appear to have measurable effects on firm births. TAX WAGE FIRMS). SIZE SIZE RETURN, BRANCHES, ENTRY) HERE) V. Conclusion This study presents evidence on the effects of bank structure and profitability on the births of new firms. The attraction of new firms is an important goal of local economic development policies, which often provide public-sector financial incentives. Private-sector financial struc ture, however, potentially influences firm loca tion through the price and availability of credit from commercial banks. The empirical analysis examines the relation ship between banking activity and regional development from 1980 through 1982. Using bank-level data, we construct measures of lend- ■ 24 The remaining S M S A s in the sample tend to be in states with large energy and agricultural sectors. ing, profitability, concentration, size, and entry in the banking sectors of 259 SMSAs. Measures of bank structure are included in a standard model of firm location in order to test for independent effects of banking on regional growth as meas ured by firm births. As with other firm location studies, we find firm births to be positively associated with low wages, low taxes, and a large number of existing firms. Our analysis, however, also shows that the private banking sector appears to be systemati cally related to the probability of firm births. Higher rates of firm openings are associated with a healthy and competitive banking sector. Specif ically, firm births are associated with higher rates of bank profits, higher numbers of bank employ ees, lower levels of concentration, higher pro portions of small banks, and higher rates of entry of new banks into the SMSA These results are robust across several specifications and samples and support the position that bank structure and profitability are significant factors in facilitating economic development. m References Amel, D. and Keane, D., “State Laws Affecting Commercial Bank Branching, Multibank Hold ing Company Expansion, and Interstate Bank ing,” Autumn 198 30-40. Issues in Bank Regulation, 6,9, Bartik, Timothy J., “Business Location Decisions in the United States: Estimates of the Effects of Unionization, Taxes, and Other Characteristics of States January 1985, 14-22. Statistics, "Journal of Business and Economic 3, ________, “The Effect of Policy on the Intramet ropolitan Pattern of Industry Location and Employment Growth,” paper prepared for Uni versity of Tennessee Symposium on Industry7 Location and Public Policy, April 1988. Bemanke, Ben S., “Nonmonetary7 Effects of the Financial Crisis in the Propagation of the Great Depression,” June 1983, 73, 257-76. American Economic Review, Bom, J., Eisenbeis, Robert A., and Harris, R., “The Benefits of Geographical and Product Expansion in the Financial Services Industry7,” Janu ary 1 9 8 8 , 1 6 1 -8 2 . Journal of Financial Services Research, Banking and Economic Development: Some Lessons of History\ New Cameron, Rondo, ed., York: Oxford University Press, 1972. Gertler, Mark, “Financial Structure and Aggregate Economic Activity: An Overview August 1988, Part 2, 559-88. ," foum al of Money, Credit, and Banking 20, Gilbert, Alton R. and Kochin, Levis A., “Local Economic Effects of Bank Failures 1990 (forthcoming). of Financial Services Research, "Journal Financial Structure and Goldsmith, Raymond, New Haven: Yale University Press, 1969. Development, Gurley, J. and Shaw, Edward, “Financial Aspects of Economic Development,” September 1955, 515-38. American Eco 45, nomic Review, Heggestad, Arnold A., “Market Structure, Compe tition, and Performance in Financial Indus tries, in Franklin R. Edwards, ed., New York: McGraw Hill Book Company, 1979. Issues in Financial Regulation, Hill, Edward W. and Shelley, N., “An Overv iew of Economic Development Finance,” in R. Bingham, E. Hill, and S. White, eds., Financ ing Economic Development: Institutional Responses, 1990 (forthcoming). Hill, Edward W. and Thompson, John Clair, “State Bank Regulators: Their Role in Eco nomic Development Finance,” Winter 1988, Development Commeyitary, Economic 24-29. Carlton, Dennis W., “Why New Firms Locate Where They Do: An Econometric Model,” in William C. Wheaton, ed., Washing ton, D.C.: The Urban Institute, 1979, 13-50. King, B. Frank, Tschinkel, Sheila L., and W hite head, David D., “F.Y.I.: Interstate Banking De velopments in the 1980s,” Federal Reserve Bank of Atlanta, May/June 1989, 32-51. Clark, Jeffrey A., “Single-Equation, MultipleRegression Methodology: Is It an Appropriate Methodology7 for the Estimation of the Structure-Performance Relationship in Bank ing?” No vember 1986, 2 9 5 -3 1 2 . Lee, Insup and Schweitzer, Robert, “Shareholder Wealth Effects of Interstate Banking: The Case of Delaware’s Financial Center Development Act,” June 1989, International Movements and Regional Growth, Journal of Monetary>Economics, 18, Eberts, Randall W. and Stone, Joe A., “Determi nants of Business Openings in Manufacturing: Wages, Unionization, and Firm Size,” paper presented at the Southern Economic Associa tion meetings in Washington, D.C., November 1987. Eisenbeis, Robert A., “Economic and Policy7 Issues Surrounding Regional and National Approaches to Interstate Banking,” in T. Havrilesky, R. Schweitzer, and J. Boorman, eds., Arlington Heights, 111.: Harlan Davidson Inc., 1985. Dynamics of Banking Economic Review, 71, Journal of Financial Services Research, 2, 65-72. Liang, Nellie, “Bank Profitability and Risk,” Unpublished paper, Board of Governors of the Federal Reserve System, November 1987. Rhoades, Stephen A. and Rutz, Roger D., “Market Power and Firm Risk: A Test of the ‘Quiet Life’ Hypothesis,” January7 1982, 73-86. Economics, Journal of Monetary’ 9, Samolyk, Katherine A., “An Empirical Test of the Credit View: Insolvent Borrowers and Aggre gate Output in the British Economy,” mimeo, Federal Reserve Bank of Cleveland, August 1988. The State of Small Business: A Report of the President, Small Business Administration, Washington, D.C.: Government Printing Office, March 1984. The State of Small Business: A Report of the President, Washington, D.C.: Govern ________, ment Printing Office, March 1985. Trifts, Jack W. and Scanlon, Kevin R., ‘ Interstate Bank Mergers: The Early Evidence,” Winter 1987, 305-11. of Financial Research, Journal 10, Wasylenko, Michael, “Empirical Evidence of Inter-Regional Business Location Decisions and the Role of Fiscal Incentives in Economic Development,” presented at University of Tennessee Symposium on Industry Location and Public Policy, April 1988. Whalen, Gary, “Actual Competition, Potential Competition, and Bank Profitability in Rural Markets,” Federal Reserve Bank of Cleveland, Quarter 3, 1988, 14-23. Economic Review, 24, Whitehead, D., (1983a) “Interstate Banking: Tak ing Inventory,” Federal Reserve Bank of Atlanta, May 1983. Economic Review, 68, ________, ( 1983b) A Guide to Interstate Banking 1983, Federal Reserve Bank of Atlanta, July 1983. ________, “Interstate Banking: Probability or Reality?” Federal Reserve Bank of Atlanta, March 1985, 6-19. Economic Review, 70, Wolkoff, Michael J., “Economic Development Financing Policy: A State and Local Perspec tive,” chapter for Sage Publications Inc., August Finance Tools, 1989. Economic Development m Fourth Q uarter Working Papers Working Paper Notice Current Working Papers of the Cleve land Federal Reserve Bank are listed in each quarterly issue of the Review. Economic Copies of specific papers may be requested by completing and mail Single copies of individual papers will Institutional subscribers, such as librar be sent free of charge to those who ies and other organizations, will be request them. A mailing list service for placed on a mailing list upon request personal subscribers, however, is not and will automatically receive available. Papers ■ 8 9 15 ■ 8 9 17 T h e T im in g o f Working as they are published. ing the attached form below. ■ 8 9 13 P o rtfo lio R is k s and B a n k A s s e t C h o ice Re g im e C h a n g e s in S to c k R e tu rn s by Katherine A. Samolyk by Nan-Ting Chou In te rg e n e ra tio n a l T r a n s fe r s , T a x P o lic y , and Ramon P. DeGennaro and A g g re g a te S a v in g s by David Altig and Steve J. Davis ■ 8 9 14 ■ 8 9 16 T h e R ole o f B a n k s in In flu e n c in g R e g io n a l A n A n a ly s is o f B a n k Fa ilu re s : 1 9 8 4 - 1 9 8 9 F lo w s o f Fu n d s by James B. Thomson by Katherine A. Samolyk ■ 8 9 18 A ltr u is m , B o rro w in g C o n s tra in ts , and S o c ia l S e c u rity by David Altig and Steve J. Davis Please com plete and detach the fo rm below and mail to: Federal Reserve B ank of Cleveland Research D epa rtm en t RO . Box 6387 C leve lan d, O h io 44101 Check item(s) Please send the fo llo w in g W o rkin g Paper(s): requested □ 8913 □ 8914 □ 8915 □ 8916 □ 8917 □ 8918 Send to: ________________________________________ Please print Name A ddress C it y S ta te Z ip