' I}: J) I. If; WW: 31’ 134““ 1‘ xv“). .. 4"” «IQ-49% ; a‘ I . ‘I' 3 |:S’I‘;:l“ '. I '::. l i ‘. U' ‘I. :{Iffl‘fi‘ . I \v ' Lt?“ mmfi - 1w WI .. $1: ’ ' 'IJ'LKJ“ "I J‘ a” ., ‘3'? “‘o‘fl'bw‘ 109.39.“ {WP'WQ'I (I. l‘vl I Ip_"_ ‘. nay h ‘.' “IN" '.:‘:::;.:.-I a“. ' ‘ Pr ; 1.. \W, {a ".2 'I ‘ I I I ‘ I .’.:bIttl IO‘\'II‘JV$(V " ".VHH%' .""\'ff‘::"» . . ( . term. ‘I' new.“ .. . 5:1,...” I , I .::“". "I yak)“: ' ';’ ”:0. l‘ ’I: {'3 3.11% "‘1?" ‘3'- ‘I ‘2‘”‘1’03 I "5' in u .'. |_ :'\ ‘Lh’ ‘I'd .’ '.‘ ‘i’VN -‘ am \- o ‘0"- cw 5 MM. .,. ~— fv~ -{ 5W! n EH J'Ilr‘ml‘p ’1' III 3W” IHII I‘ I If" II;le ;- _ 1""!‘M'IIIII I' 1“ ~‘-.-.’.'. _ “.‘..q, I Vv‘i‘ ‘ . I , . . flit-.79. Ll & 41.12.}: —..l.:‘/:. u iWI'I‘ In 0-7639 This is to certify that the thesis entitled THE RESPONSE OF COMMERCIAL BANKS TO INITIAL MARKET ENTRY BY MULTI-BANK HOLDING COMPANIES presented by Gregg K. Dimkoff has been accepted towards fulfillment of the requirements for Ph. D. degree in Finance Major professor LIBRARY Michigan State University mg; “ll! “ll“ ”M a THE RESPONSE OF COMMERCIAL BANKS T0 INITIAL MARKET ENTRY BY MULTI-BANK HOLDING COMPANIES By Gregg K. Dimkoff A DISSERTATION Submitted to l': (9.1;: , .2; , Michigan State University " 'in partial fulfillment of the requirements . for the degree of DOCTOR OF PHILOSOPHY 3“ pen; of‘ Accounting and Financial Administration ti 2.13dapumiy ”' 1973 Te C3 H1 ABSTRACT THE RESPONSE OF COMMERCIAL BANKS TO INITIAL MARKET ENTRY BY MULTI—BANK HOLDING COMPANIES By Gregg K. Dimkoff . The rapid growth in bank holding company acquisitions occurring since the passage of the 1966 Amendments to the Bank Holding Company Act of 1956 has generated considerable interest among regulators con- cerning ways in which recently acquired banks change relative to inde— pendent banks. This interest has resulted in several empirical studies that have been generally consistent in finding that new holding company affiliates do not perform much differently than independent banks. The research described in this report differs in three ways from most of these bank holding company studies. First, the research sample was limited to only those multi—bank holding company acquisitions rep- resenting the initial entry into each market. If the reactions of competing banks to multi-bank holding company entry are a function of their reactions to prior holding company entry, failure to limit the sample to only the initial entry into a market could confound the find— ings. Next, an independent bank will be paired with both the new af— filiate and with a non-competing independent bank. The purpose of this pairing technique is to check for differences between the new affiliate and competing independent bank and to check for coincident changes occurring to both of these competing banks. Finally, the usefulness f of multivariate instead of only univariate statistics will be I81 an: '4 rr‘ Gregg K. Dimkoff investigated. Only the most recent research has used multivariate statistical techniques. A multivariate multiple linear regression model was hypothesized which included three dependent variables (return on loans, before-tax return on investments, and operating expenses to operating revenue) and nine independent variables (calendar year, Federal Reserve System membership, state branching law, bank asset size, market concentration, county-wide personal income, deposits at competing financial institu- tions, and two variables describing bank type). The sample was selected from eight states: Florida, Iowa, Michigan, Missouri, New Mexico, Ohio, Texas, and Wisconsin. Counties were used as banking markets. All coun- ties experienceing initial market entry through acquisition by a multi- bank holding company between 1/1/69 and 12/31/76 were included in the research sample if there were no holding company relationships between the affiliate and holding company prior to acquisition. Altogether 83 markets were identified. In each one, two banks were selected: the new affiliate and a same size independent bank which remained independent throughout the period investigated. Further, a same size non-competing independent bank located in a county experiencing no holding company activity was paired with the competing independent bank. Thus for each of the 83 markets, a triplet of three same size banks was selected: affiliate, competing independent bank, and non-competing independent bank. Analysis of the sample began with inspection of graphs of the de- pendent variables over time for each of the three bank groups. The graphs suggested that holding company affiliates achieved slightly higher return on their loan portfolios, shifted into lower yielding Séfil rev: C01 Gregg K. Dimkoff securities, and experienced no change in operating expenses to operating revenue compared with independent banks. Using both univariate and multi— variate statistics, the null hypothesis that the coefficients of the variables designating bank type were equal to zero could not be re- jected in 29 out of 30 multivariate tests or 81 out of 90 univariate tests at the 95 percent confidence level. Both univariate and multi— variate statistics were consistent in finding few significant differences between bank groups. Thus there is no evidence that the univariate model distorts the findings compared with the multivariate model. The conclusion is that only univariate statistics was needed in this study. These results suggest that when no clear-cut factors dictate approv- al or denial of an application for acquisition, the acquisition should be approved. Neither competing bank will significantly increase its riskiness, but there may be benefits from increased services, better management, and capital infusion at the new affiliate. ACKNOWLEDGMENT ' I am deeply grateful for the guidance I received from the members of my dissertation committee who were Professors Jack Brick and Harold Sollenberger, both of the Department of Accounting and Financial Admin- istration, and Professor Robert Rasche of the Department of Economics. ‘ I also wish to thank the staff of the Research Department of the Federal Reserve Bank of Chicago, especially Drs. Chayim Herzig-Marx and .3816 Drum, without whose help and cooperation this dissertation would finot’have been possible. Finally, I wish to acknowledge the patience and assistance of my wife, Evie. Her sacrifices made it all possible. 11 4““ at"? u 3.33.21 .A'. ‘rv‘rsro - “ V ‘7. .~ I.‘ LIST 0 11 III .'-—.‘. TABLE OF CONTENTS LIST OF FIGURES Chapter I. INTRODUCTION Background Methodological Issues Research Model Null Hypothesis Policy Implications Overview Summary II. HISTORY OF LEGISLATION AFFECTING BANK HOLDING COMPANIES Early History Developments During the Last Decade Summary ' III. LITERATURE REVIEW Bank Mergers David Smith Thomas Snider Branching Kuhn and Carlo New York State Branching Department Schweiger and McGee Horvitz and Shull Bank Holding Company Acquisitions Thomas Piper Joe McLeary Robert Lawrence Samuel Talley Rodney Johnson and David Meinster Robert Ware Stuart Hoffman John Mingo Lucille Mayne Edward Daley and Duane Graddy Different Approaches of Present Research Sample Selection Pairing Technique Appropriateness of Multivariate Statistics Summary iii vi ix \OQCDNO‘WD-I Chapter IV. THE RESEARCH MODEL 49 The Research Model 49 Dependent Variables 56 Diversification 57 Management Succession 62 Financial Leverage 63 Inefficient Markets 63 Economics of Affiliation 64 Summary 65 Independent Variables 66 Bank Types X and Y 66 Federal Reserve System Membership, M 66 State Branching Law, L 67 Bank Asset Size, S 68 Herfindahl Index, H 68 County Income, I 69 Deposits at Competing Institutions, D 70 Summary 70 V. THE SAMPLE 73 Market Definition 73 Sample Selection 74 Sample Characteristics 77 Average Bank Size 77 Accuracy of Pairing Technique 82 Summary 84 VI. TESTING AND RESULTS 87 Characteristics of the Research Variables 87 Characteristics of the Continuous Variables 87 Correlations among the Variables 91 Graphical Analysis 93 Dependent Measures 94 Return on Loans, R 94 Before-Tax Return on Investments, R 96 Operating Expenses to Operating Revenue, R3 96 Herfindahl Index 98 Statistical Testing 100 Comparisons 100 Results 101 Statistical Significance 102 Need for Multivariate Analysis 106 Estimation 107 Direction of Effect of Independent Variables Compared with the Hypothesized Effect 108 Analysis of Beta Coefficients 113 R1 115 R2 116 R3 116 Summary 116 iv n... .'.rye VT" ‘1'. SUMMARY AND CONCLUSIONS , up . Summary of Objectives and Methodology -"’ Summary of Findings ' ‘ Policy Implications and Recomendationa Insights Suggestions for Further Research " 'I A): 0 . t , ‘ ~ I-rhpflldix ’ ‘ ' Turf. RAH REGRESSION COEFFICIENTS FOR THE UNIVARIATE TESTS 4.1 - a. DIRECTION OF EFFECT THE INDEPENDENT VARIABLES HAVE , ' ‘ ON THE DEPENDENT VARIABLES '1 ~. I .‘Saleeted Bibliography ..... 120 120 122 123 125 126 129 140 149 ' . '5“ A} \ . -_... ‘g Q '3 7:05. “'6 .'.w - -. u. '5'“ 3r“ Cally-f .. 1‘ “V r—l n) L I" to £‘ (1‘ (11 Table 2.2 3.2 3.3 LIST OF TABLES Page Classification of States According to Types of Branching Prevalent, December 31, 1976. ................. 18 Classification of State Laws Affecting the Acqui- sition of Bank Stocks by Corporations, May 31, 1973 ..... 19 Summary of Selected Research into the Effects of Merger on Bank Performance............ .................. 24 Summary of Selected Research into Branching Effects on Bank Performance............ ..... . ..... . ....... . ..... 27 Summary of Bank Holding Company Research Findings: Relative Changes Occurring in New Affiliates.. .......... 32 Consolidated Report of Condition of Bank and Domestic SUbSidiaries-noon-.....oo...............u......... sssss O 52 Consolidated Report of Income of Bank and Domestic subSidiarieSoo..-.........o.a..o.....oo-..........-..... 54 Number of Triplets by State for Each Year Relative to Acquisition Year............................. ....... . 78 Number of Triplets Available for Testing by State and Calendar Year......................... ...... ........ 79 Average Asset Size of Bank Groups During the Acqui— sitionYearCIOIOIIIIOIOOOOIOIOIOCIOUIOOI.OIIOCIOCOIOIIO. 80 Accuracy of Bank Pairings......... ....... ................. 83 Statistical Characteristics of Continuous Variables....... 88 Correlation Matrix of All Variables in the Study.......... 92 Results of F-Tests for Significant Differences Between GrOWBOIOOCOI.'...'....IOOICOIOCIOOIIOOODDOOI0.. ..... I0. 103 Results of Tests for Significant Differences Between Groups-O...IIICICOOOOCCOC.'....OIOICI.IIIIOIIOOOIOOQIIII 104 vi cl} Table Page 6.5 The Need for Multivariate Statistics Measured by Bartlett Tests of Sphericity ........................... 107 6.6 Beta Coefficients for Dependent Variables ................ 114 A.1 Raw Regression Coefficients for Tests Made on Data from the Third Year Prior to Acquisition ............... 130 A.2 Raw Regression Coefficients for Tests Made on Data from the Second Year Prior to Acquisition .............. 131 A.3 Raw Regression Coefficients for Tests Made on Data from the Year Prior to Acquisition ..................... 132 A.4 Raw Regression Coefficients for Tests Made on Data from the Year of Acquisition ........................... 133 A.5 Raw Regression Coefficients for Tests Made on Data from the First Year after Acquisition ..... . ......... ... 134 A.6 Raw Regression Coefficients for Tests Made on Data from the Second Year after Acquisition.. ............... 135 A.7 Raw Regression Coefficients for Tests Made on Data from the Third Year after Acquisition ........ . ......... 136 A.8 Raw Regression Coefficients for Tests Made on Data from the Fourth Year after Acquisition........... ...... 137 A.9 Raw Regression Coefficients for Tests Made on Data from the Fifth Year after Acquisition.................. 138 A.10 Raw Regression Coefficients for Tests Made on Data ' from the Sixth Year after Acquisition.................. 139 B.1 Direction of Effect on Dependent Variables from State Branching Law..... ...... . ......... ......... ...... 141 8.2 Direction of Effect on Dependent Variables from County Personal Income................................. 142 B.3 Direction of Effect on Dependent Variables from Deposits at Competing Financial Institutions........... 143 3.4 Direction of Effect on Dependent Variables from Calendar Year........................ ....... ........... 144 3.5 Direction of Effect on Dependent Variables from Federal Reserve System Membership...................... 145 i 3.6 Direction of Effect on Dependent Varaibles from hnkAsset sizeCIIOOODI.........OI......OOOIOIUOIOIOOOO 146 vii Page Direction of Effect on Dependent Variables from mrket concentration.IIIOOOCI...OCIOII.I.......OII....' 147 Figure LIST OF FIGURES Egg; Pairing Technique Used in Most MBHC Research ..... . ..... ... 4 Pairing Technique Used in This Research............ ..... .. 5 Initial MBHC Position....... ............. ............ ..... 58 MBHC Position after Acquiring Bank B.... ....... ........... 59 Behavior of R , Return on Bank Loans, for Bank Groups Relative to Acquisition Year..... ......... . ..... ... ..... 95 Behavior of R , Return on Investments, for Bank Groups Relative to Acquisition Year............................ 95 Behavior of R3, Operating Expenses to Operating Revenues, for Bank Groups Relative to Acquisition Year............ 97 Changes in the Herfindahl Index Relative to Acquisition Year for Markets with and without MBHC Affiliates....... 99 ix 5'11 at Ft? th- hr Ci. to: f0: {10' CHAPTER I INTRODUCTION The objective of this chapter is to briefly describe why the response of commercial banks to initial market entry by multi-bank holding companies is of research interest and to explain how the re- sponse will be measured. Discussion begins with a background section describing the rapid increase in the number of holding company acqui- sitions and summarizing research results of several studies directed at finding how new affiliates change relative to other banks. The methodology that sets this research apart form previous studies is then explained. After the research model and null hypotheses are presented, policy implications are described. Finally, following an overview of the remainder of this report, the chapter concludes with a brief summary. BACKGROUND A bank holding company is a corporation which owns, controls, or has power to vote at least 25 percent of the common stock of a commer- cial bank or controls the election of a majority of the board of direc- tors of a bank. While bank holding companies have been in existence for decades, their importance in the commercial banking industry was not great until passage of the 1966 Amendments to the Bank Holding Company Act of 1956. An examination of bank holding company growth san Sea 86‘! since the passage of these amendments makes evident the explosive growth that has occurred. For example, at the start of 1966, forty- eight multi-bank holding companies (MBHC's) controlled 8 percent of total deposits.1 By the end of 1970 the number of MBHC's had grown to 121. They controlled 16 percent of total bank deposits while 1,352 one-bank holding companies controlled an additional 33 percent of the nation's bank deposits.2 As of December 31, 1974, 276 MBHC's controlled 38 percent of total bank deposits while 1340 one-bank holding companies controlled an additional 30 percent.3 This explosive growth has occurred with the Federal Reserve's sanction since holding company acquisitions must be approved by the Board of Governors. Approval or rejection is based on an analysis of several banking factors and on the impact the acquisition has on poten- tial competition.4 Even though any one of these numerous factors can be the basis on which a bank holding company application is denied, one single factor—-potential competition-—has been either a major or minor factor in 50 percent of the denials since 1957.5 This rapid growth and its impact on competition has provided the impetus for several research studies aimed at finding how both the ac- quired bank and the remaining independent banks change after acquisition. Empirical results have been fairly consistent in finding that compared to unaffiliated banks, acquired banks shift funds out of U.S. government securities and into both state and local government securities and loans (6,7,8,9,10,11,12,l3). Studies have also shown that acquired banks lend a greater proportion of their loanable funds to consumers (7,8,10,12) and experience no change in net income (6,10,14). METHODOLOGICAL ISSUES In most of these studies acquired banks were paired with banks un- affiliated with MBHC's so that differences in their performance could be compared before and after acquisition. However, in studies like these, reactions of unaffiliated banks to the initial MBHC entry must be con— sidered. This is necessary because when an MBHC first enters a market comprised of only unaffiliated banks, these independent banks may react in ways changing their operating and financial characteristics. Reac- tions might be in response to their actual or alleged inability to com- pete with the larger MBHC system. It is possible that the reactions of the unaffiliated banks to the second MBHC entry would be different than their reactions to the initial entry, that their reactions to the third MBHC entry would be different than their reactions to the first or sec— ond, and so on. Thus to measure the initial response of independent banks to entry by an MBHC, it is necessary to limit the study to only those banking mar— kets which are experiencing MBHC entry for the first time. In this way effects of holding company entry will not be distorted by confounding variables introduced from inclusion in the research sample of MBHC ac- quisitions occurring after the initial acquisition within a market. Previous studies have not included this limitation on sample selection. Nor have previous studies adequately considered coincident changes that might have occurred to both the acquired and unaffiliated paired banks. Figure 1.1 below illustrates the pairing technique used by most researchers. Pai I'~ FIGURE 1.1 PAIRING TECHNIQUE USED IN MOST MBHC RESEARCH Banking Market I Affiliate Independent Bank New MBHC affiliates have been paired with independent banks of similar asset or deposit size. Differences in performance between the pairs at some point in time prior to acquisition were compared with dif— ferences at a point in time after acquisition. If statistically signif- icant differences between these time points were found, they were at- tributed to the MBHC entry. If no differences were found, it was rea- soned that MBHC's have little impact on competing independent banks. However, if managers of the paired independent banks felt threat- ened by acquisition of a competitor, they might take actions deemed nec- essary to remain competitive. Such actions might include changes in asset and liability structure, changes in pricing, etc. If the MBHC also made proportionate changes in its newly acquired bank, no differ- ences in performance would be found between the acquired and independent banks even though both banks did experience a significant change. In an attempt to overcome this possible shortcoming, several re- searchers have included other comparisons with this before and after pairdwise comparison (8,15). These other comparisons usually consist of ~r-. m ‘L’s' r U A 1 i no 14:» 9‘. 5.“, 1 I sun: ‘5‘. a before and after analysis of all new affiliates or all paired inde- pendent banks considered as a group. Thus if the pair-wise comparison indicated that no statistically significant differences existed, the be- fore and after analysis on either group would indicate whether both groups did change proportionately after MBHC entry. If a change were found, it could be only weakly inferred that the holding company acqui- sition caused it and not other exogenous changes. This research is designed to overcome the two possible shortcomings just described regarding market selection and comparison techniques in the measurement of bank responses to MBHC entry. The sample will be limited to only those banking markets which have experienced MBHC entry by acquisition for the first time. Further, to check for coincident changes occurring to both banks in the pairs, comparisons illustrated in Figure 1.2 will be made. FIGURE 1.2 PAIRING TECHNIQUE USED IN THIS RESEARCH Banking Market I @x \ Independent MBHC ‘ \ Bank Affiliate ‘ x O _ - - — - Banking Market II Independent Bank So: competi non-co: banks ( acre Vie] Not only will the new affiliates be compared with independent banks competing in the same banking market, but independent banks located in non—competing markets will be compared with the competing independent banks (Independent Bank in Market II will be compared with Independent Bank in Market I). This last comparison will indicate whether coinci- dent changes occur to both banks in Market I in Figure 1.2. If coinci- dent changes do occur, this will give more conclusive evidence that the change was in reaction to MBHC entry and not just a response to exoge- nous changes in the economy affecting all commercial banks. Additionally, most prior research has relied upon univariate sta— tistical techniques instead of using the more powerful multivariate techniques for analysis. This research will investigate whether the use of multivariate statistics on the measures used in this study would be more appropriate. RESEARCH MODEL The research model may be expressed in a regression framework as R1 = bio + bilx + b12Y + bi3C + bi4M + biSL + bies + b17H + b18I + bigD + e1 where R1 represents the dependent measures of bank performance as follows: R1 8 interest on loans. total loans R2 = before—tax interest on investments, total investments ’ R3 = operating egpenses, operating income ’ bij represent net regression coefficients; _____..._I- __ .__..__—______ N and Y have values of 0 or 1 and represent bank type. By manipulating the values of these two variables, a given bank type can be contrasted with either of the two others; represents represents represents represents represents represents represents tutions; e represents U H m m r': 0 Since there are calendar year; Federal Reserve System membership; state branching law; bank asset size; the Herfindahl Index; county personal income; county-wide deposits in competing financial insti- and the error term. three different dependent measures, multivariate statistics may be appropriate. The model is based on findings by other researchers and on an assessment of the usefulness of these measures. It is explained in detail in Chapter IV. W The null hypothesis, H0, is Ho‘ bil = biz = 0 i = 3 dependent variables. That is, bank type does not result in statistically significant per- formance differences on the three dependent measures. However, it is expected that new MBHC affiliates will experience higher return on both their loan and investment portfolios and higher ratios of opera- ting expenses to operating revenue compared with paired competing inde- pendent banks. An explanation of the reasoning for these expectations is-presented in Chapter IV. IESE. beu: r6136: this 'a'“ POLICY IMPLICATIONS A11 acquisitions of banks by corporations require approval by the Board of Governors of the Federal Reserve System. The Board considers specific criteria contained in the Bank Holding Company Act of 1956 relating to present and potential competition, management competency, and the convenience and needs of the community served. Evaluation of these factors requires timely data and measurements of the effects acquisitions will have on the new affiliates and competing banks. This research will provide regulators with additional information which can be used to evaluate holding company applications. Results will show whether holding company affiliates change their loan and investment portfolios after acquisition and whether their operating expenses rise compared with operating revenues. Additionally, whether competing independent banks change in response to the initial MBHC entry on these same performance measures will be determined. If it is found that affiliates and/or competing independent banks switch to riskier loans and investments, regulators might use these findings to deny acquisition applications in markets where one or more of the banks has a relatively risky asset structure already. Approval of the application might result in even riskier banks and possibly damage their financial strength. The Federal Reserve Bank of New York has for several years con- ducted research aimed at finding measures of bank performance that forewarn of financial deterioration.16 One of the measures that has proven most useful is the ratio of operating expenses to operating revenue, R3 in the research model described earlier. The findings of this research may indicate that R3 falls for competing banks as a result that th from hc of ban] Cbapte cor-p a: ted t model the s istic Finaj sion SQ! Sit ate 30: result of MBHC entry. If that is the case, it would be recommended that the Board carefully determine whether the decline in R3 resulting from holding company entry will contribute to a financial deterioration of banking firms in a given market. mm The remainder of this dissertation is subdivided into six chapters. Chapter II describes the history of regulation affecting bank holding companies. Then Chapter III presents a review of the literature rela- ted to holding company acquisitions. A discussion of the research model used to test for significant differences between bank groups is the subject of Chapter IV. Next, Chapters V and VI describe character- istics of the research sample and results of the testing, respectively. Finally, Chapter VII concludes this dissertation with a summary, discus- sion of implications, and suggestions for further research. SUMMARY This chapter provided an introduction to the six chapters that follow. Discussion began with a description of the rapid growth in bank holding company acquisitions occurring since passage of the 1966 Amendments to the Bank Holding Company Act of 1956. Consistency in the findings of several empirical studies was described. The methodology that sets this research apart from these past studies was then de- scribed. In brief, this study will investigate only those MBHC acqui- sitions occurring in markets previously void of holding company affili- ates, will compare affiliates with both competing independent banks and non-competing independent banks, and will investigate the need for :ultive was the tax If will b H- r) (1‘) (D 10 O . 3“; W ~variate instead of only univariate statistics. The research model I . ‘ '4‘ ‘ ovaigfi then described. Three dependent measures--return on loans, before- '. ‘ C(‘Il “tflflpneturn on investments, and Operating expenses to operating revenue—- "I111 be used to test for significant differences among groups of paired . - ... banks beginning with the third year prior to acquisition and continuing .i A turbugh the sixth year after acquisition. Ways bank regulators could .' l— .‘ .use findings of this research to help reach decisions approving or deny- 4 in; Acquisition applications was then described. 1 \ I l I l CO? 11 FOOTNOTES T0 CHAPTER I 1One—bank holding companies were excluded from the regulations contained in the Bank Holding Company Act of 1956 and were not subject to them until the 1970 Amendments. 2Harvey Rosenblum, "Bank Holding Companies: An Overview," Business Conditions, Federal Reserve Bank of Chicago (August, 1973): 4. 3Annual Statistical Digest 1971-1975, (Washington, D. C.: Board of Governors of the Federal Reserve System, October 1976), 280. 4Banking factors include such items as capital adequacy, quality of management, earnings prospects, etc. SHarvey Rosenblum, "Bank Holding Companies--Part II," Business Conditions, Federal Reserve Bank of Chicago (April, 1975): 14. 6Robert J. Lawrence, The Performance of Bank Holding Companies, (Washington, D. C.: Board of Governors of the Federal Reserve System, June, 1967). 7Samuel H. Talley, The Effect of Holding Company Activity on Bank Performance, (Washington, D. C.: Staff Economic Studies, Number 69, Board of Governors of the Federal Reserve System, 1972). 8Robert F. Ware, "Performance of Banks Acquired by Multibank Holding Companies in Ohio," Economic Review, Federal Reserve Bank of Cleveland (March-April, 1973): 19-28. 9Joe W. McLeary, "Bank Holding Companies: Their Growth and Per- formance," Monthly Review, Federal Reserve Bank of Atlanta (October, 1968): 131—138. 10Stuart G. Hoffman, "The Impact of Holding Company Affiliation on Bank Performance: A Case Study of Two Florida Multibank Holding Companies," Working Paper Series, Federal Reserve Bank of Atlanta (January, 1976). 11Thomas R. Piper, The Economics of Bank Acquisitions by Regis- tered Bank Holding Companies, (Boston: Research Report to the Federal Reserve Bank of Boston, No. 48, 1968). 12Edward V. Daley and Duane B. Graddy, "Optimal Firm Size, Market Penetration, and Holding Company Performance: A Test of the Jessup- Piper Hypothesis," Conference Papers Series: #22, Business and Economic Research Center, Middle Tennessee State University, (June, 1977). 13Gerald C. Fischer, Bank Holding Companies, (New York: Columbia University Press, 1961). 14John J. Mingo, "Capital Management and Profitability of Pro— spective Holding Company Banks," Journal of Financial and Quantitative Apglygig, (June, 1975): 191-201. 12 & 15Rodney D. Johnson and David R. Meinster, "The Performance of Qéflflnh Holding Company Acquisitions: A Multivariate Analysis," Journal .139; Business, (April, 1975): 204—212. ' 1‘6For example, see "A Nationwide Test of Early Warning Research '. in Banking," Quarterly Review, Federal Reserve Bank of New York (Autumn, ’ ‘ 1.51977): 37-52. .‘.)L". .’.” . ‘9‘ ...-w r 7: l ( str CHAPTER II HISTORY OF LEGISLATION AFFECTING BANK HOLDING COMPANIES In order to understand the factors that have led to present re- strictions on bank holding company formation and expansion, it is neces— sary to review the history of regulation affecting holding companies. That is the major objective of this chapter. After a brief description of the early history of attempts to control the movement toward holding company formation, a more detailed discussion of recent regulation is presented. EARLY HISTORY Interest in determining the impact bank holding companies have on bank structure has been a topic of great interest over the last twenty years. Hovever, the bank holding company concept has been around since the start of the nineteenth century. During the first third of that century, banks exercised their common law right to establish branches. Because many of these branches had their own presidents and boards of directors, they closely resembled today's holding companies. Few branch banks remained in existence after the 1860's because of detrimental legislation included in the National Bank Acts of 1863 and 1864 and because of adverse conditions during the Civil War. Thus by the start of the twentieth century unit banking was dominant in the United States. For example, Klein points out that: 13 (,5 14 The great distances between population centers, the difficulties in travel and communication, and the primarily agrarian nature of the economy all fostered the development of unit banks—- locally owned, operated, and controlled. By 1921 for example, there were 31,076 banks operating 1,455 branches.i However, the thousands of banks that failed in the 1920's and 1930's convinced bank regulators of the need to prevent such "overbanking" whereby many small and inefficient banks were eking out a marginal existence.2 As a result, both state and national authorities made it much more difficult to establish new banks or branches. One of the first such attempts to restrict any form of banking ex- cept unit banking was contained in the Banking Act of 1933. Included in this act was the first federal attempt at regulating bank holding companies: those holding companies which included a member of the Federal Reserve as a constituent were required to register with the Federal Reserve System. Further, among its many provisions to finan- cially strengthen the banking industry, this act included a provision which introduced the "need" test in chartering new banks and branches. It requires that banks must demonstrate to the appropriate regulators that a clear need exists for a new bank or branch in the community to be served. The feelings against competitive banking that led to such restric- tions were more deeply engrained during the 1930's than this act might suggest. Indeed, the popular sentiment against all forms of multiple banking was so pervasive in the U.S. that even the bankers themselves favored only unit banking. For example, during the American Bankers Association Boston Convention of 1937, a resolution was adopted which described unit banking as "peculiarly adapted to the highly diversified community life of the United States."3 ‘Pd .6 'i: l 0! I \. ha' ha ha“ F._____—_ 15 DEVELOPMENTS DURING THE LAST TWO DECADES Although the sentiment against multiple banking has abated some- what since the 1930's, it still exists and still influences bank legis- lation. In fact, the most recent legislation affecting bank structure and competition has been a series of acts regulating bank holding com— pany operations and expansion. The initial legislation was the 1956 Bank Holding Company Act giving the Federal Reserve the responsibility for regulating bank hold— ing companies. Additionally, bank holding companies were defined by this act as any corporation or trust association controlling 25 percent or more of the voting stock of two or more banks. Finally, registered bank holding companies were required to divest themselves of most of their non-bank businesses. At that time 53 corporations owned 428 banks and controlled 7.5 percent of total bank deposits.4 During the latter 1950's the banking industry entered a period of strong merger activity. Resultant concern amOhg regulators, bankers, and Congress over the monopoly power of a few large banks led to the Bank Merger Act of 1960. This act specified to the appropriate regu- lators that in deciding whether a merger of FDIC insured banks should be approved, the following list of factors were to be considered:5 1. The financial history and condition of each of the banks involved; . Capital structure; . Future earnings prospects; The general character of management; . The convenience and needs of the community to be served; and . The effect of the merger on competition. O‘MJ-‘UJN C Thus, in addition to the competitive effects of a proposed merger, regulators were instructed to also weigh other factors such as earnings and management competency. Mergers could be approved under the Act \ o conf 1655 defi thES 16 even if competition were reduced as long as the benefits from the other factors outweighed the reduction in competition. However, in a landmark decision, the United States Supreme Court ruled on June 17, 1963 that Section 7 of the Clayton Act, as amended in 1950 by the Celler—Kefauver Act, applied to the banking industry.6 Under Section 7, a merger is not allowed whenever the courts find a reasonable probability that competition will be substantially lessened in the relevent market. Most notably, the Court refused to accept any of the other factors listed above in determining the legality of a pro- poSed bank merger: only the effects of competition mattered. Thus, a proposed merger could be found to be in the public interest under the Bank Merger Act but would be struck down by the courts based on this landmark decision. Prior to this time, bankers and regulators had be- lieved themselvesxto be outside the scope of the Clayton Act.7 The Supreme Court decision threw the banking industry into a sea of confusion. Bankers were unsure of reactions by the courts to proposed bank acquisitions. Would the courts disallow the merger because of a lessening of competition? How restrictive would the courts be in their definition of competition? As a result of unanswered questions like these, there were few bank acquisitions in the ensuing three years. Congressional concern regarding this confusion led to the Bank Merger Act of 1966. The Act provides that the appropriate federal bank- ing agencies shall not approve a merger that would result in a monopoly. However, if public benefits outweigh the harm caused by lessened compe- tition, the merger can be approved. Hence this act is not as restric- tive as the Court's Philadelphia decision regarding the lessening of competition, but is more restrictive than the Bank Merger Act of 1960. vvvv wv 'TFIIl‘F-""""""""""""""""""""""""""""""""""""""-"""""""'- 17 Also that year the Bank Holding Company Act of 1966 was enacted to extend the provisions of the Bank Merger Act of 1966 to holding company acquisitions. The last significant legislation affecting bank holding companies was the 1970 Amendments to the Bank Holding Company Act of 1966. The original act had excluded one-bank holding companies from its provi— sions. By 1970 however, these unregulated one—bank holding companies accounted for 38 percent of total bank deposits and operated in every major sector of the economy.8 Concern over the financial power of these companies led to the 1970 Amendments which brought one—bank holding companies under the provisions of the Bank Holding Company Act of 1956 as amended in 1966. This just-described stream of legislation restricting bank acquisi— tions is strong evidence that many still feel the preservation, if not the expansion, of unit banking is the best defense against the monopoly powers of big banks and the trend toward consolidation in the banking industry. Today the extent to which a bank or holding company can ex— pand through branching or acquisition is determined by the laws of the state in which it is located. Twenty states permit state-wide branch— ing (provided, of course, that the state or federal regulators approve the branCh), seventeen states permit limited area banking, and thirteen states, the unit banking states, permit no branching whatsoever. Twenty states also restrict or prohibit the formation of bank holding companies. Tables 2.1 and 2.2 on the two following pages present state laws regula- ting branching and holding company acquisitions, respectively. TABLE 2.1 CLASSIFICATION OF STATES ACCORDING TO TYPES OF BRANCHING PREVALENT, DECEMBER 31, 1976 Statewide branch banking prevalent Alaska Arizona California Connecticut Delaware Hawaii Idaho Maine Maryland Nevada New Jersey New York North Carolina Oregon Rhode Island South Carolina South Dakota Utah Vermont Washington Limited branch banking prevalent Alabama Arkansas Georgia Indiana Iowa Kentucky Louisiana Massachusetts Michigan Mississippi New Hampshire New Mexico Ohio Pennsylvania Tennessee Virginia Wisconsin Unit banking prevalent Colorado Florida Illinois Kansas Minnesota Missouri Montana Nebraska North Dakota Oklahoma Texas West Virginia Wyoming SOURCE: Rand McNally International Bankers Directory, First 1977 Edition, New York, NY, Laws Section, L7-L154. SC TABLE 2.2 CLASSIFICATION OF STATE LAWS AFFECTING THE ACQUISITION OF BANK STOCKS BY CORPORATIONS, MAY 31, 1973 State approval Restricted No limitations required or prohibited Alabama Nevada California Alaska Arizona New Mexico Connecticut Arkansas Colorado North Carolina Florida Georgia Delaware North Dakota Iowa Illinois Dist. of Ohio Maine Indiana Columbia South Dakota Massachusetts Iowa Hawaii Tennessee Missouri Kansas Idaho Texas New York Kentucky Maryland Utah Oregon Louisiana Michigan Virginia South Carolina Mississippi Minnesota Wisconsin Missouri Montana Wyoming Nebraska New Hampshire New Jersey Oklahoma Pennsylvania Rhode Island Vermont Washington West Virginia SOURCE: "Bank Holding Company Facts," Association of Registered Bank Holding Companies, Washington, D.C., Spring, 1973 Edition. Reg tomerci with a c to the I tir pet stricti: destruc during 9098!. 20 SUMMARY Regulations affecting the ability of holding companies to acquire commercial banks has been presented in this chapter. Discussion began with a description of the legislation affecting holding companies prior to the 1956 Bank Holding Company Act. Regulatory philosophy during this time period can be described as favoring unit banking and severely re- stricting multiple banking. This philosophy reflected concerns about destructive competition, about the great number of banks that failed during the 1920's and 1930's, and that bigness would lead to abuses of power. The Bank Holding Company Act of 1956 defined bank holding companies"1 and placed restrictions on the types of businesses that holding companies could acquire. Strong merger activity within the banking industry that began during the latter 1950's was severely restricted by a Supreme 9/ Court ruling in 1963 that held mergers or acquisitions could not be aPProved if a lessening of competition resulted. This was applicable even if there were offsetting public benefits of a proposed combination. To remedy this undesireable effect, the Bank Merger Act of 1966 and the Bank Holding Company Act of 1966 were enacted. The purpose was to per- mit combinations of banks if benefits outweighed the lessening of compe- tition. Finally, the Bank Holding Company Act of 1966 was amended in 1970 to bring one-bank holding companies under provisions of the 1966 Act. They had been omitted from the original act. The description of regulations affecting bank holding companies describes the regulatory environment that currently exists. With this background, a better understanding of research into the effects of holding company acquisitions can be gained. A review of the research literature is presented in the next chapter. 1: vaancri cussion found in 1101. 3 One-Bank 21 FOOTNOTES TO CHAPTER II 1John J. Klein, Money and the Economy, (New York: Harcourt Brace Jovanovich, 1974): 45. 2Between 1921 and 1933, over 15,000 banks failed. Further dis- cussion of the hardships encountered by banks during this period can be found in "Banking in the United States," The Banker, (September, 1974): 1101. 3George S. Eccles, "Registered Bank Holding Companies," in The One-Bank Holding Company, (Chicago: Rand McNally and Company, 1969): 90. 4Ibid., "Banking in the United States," p. 1101. 5Act of May 13, 1960, Public Law 86—463, 74 Stat. 129, 12 U.S.C. 1828(c). 6U.S. vs. The Philadelphia National Bank and Girard Trust Corn Exchange Bank, 374 U.S. 321 (1963). 7Ibid., "Banking in the United States," p. 1102. 8"One-Bank Holding Companies Before the 1970 Amendments," Bulletin, Board of Governors of the Federal Reserve System, (December, 1972): 1101. their of a l focuse search compare Re three c sition 80?? th formanc or with acquire ing Com Ethodo Each ta The rem hnfr with a CHAPTER III LITERATURE REVIEW A question raised by a number of researchers is: Do banks change their operating and financial characteristics in response to acquisition of a competing bank by another bank? However, most of the research has focused on changes occurring in the acquired bank. Although this re- search is concerned with changes in both the acquired bank and the re- maining competing banks, a review of the literature will be fruitful. The methodology used and results obtained in this research can then be compared with the past research. Research related to changes in acquired banks can be divided into three categories. The first compares performance pre— and post—acqui- sition for banks Which have merged into a larger bank. In a second cate— gory the Performance of bank branches has been compared with the per- formance characteristics of either their home offices, with unit banks, or with their own performance prior to acquisition. And finally, banks acquired by MBHC's have been examined to determine the effects the hold- ing company had on the acquired bank. With only a few exceptions, the “Ethodology and results are similar in all three of these categories. ESCh category is examined in more detail in the discussion that follows. The remainder of the chapter then describes the ways this research dif- fers from that reviewed in the first three categories. The chapter ends with a summary, 22 lath 3.1 meant into 1 the table ate baits-me by '. scribed below. 1m SHIT. ”he in huh 1 ed Stile} 03 the = :atia: ten 23 BANK MERGERS Table 3.1 on the following page summarizes key results of selected research into the effects of merger on bank performance. Included in the table are two notable investigations into the performance of merged banks——one by David Smith and the other by Thomas Snider. Each is de— scribed below. DAVID SMITH The first investigation is a study of 81 merging banks in the Fourth Federal Reserve Districtl during the 1960-65 period by David Smith.2 These 81 merged banks were compared with 81 non—merging banks of the same approximate asset size located in the same geographical lo- cation. Twenty—two banking ratios were then selected to measure bank performance. Predmerger data were collected at year-end for the year preceding merger and post-merger data were collected for year-end, three years later. A univariate t—test for statistical significance was performed on the mean difference between the two groups for each ratio. Smith found that the merged banks significantly decreased their cash assets compared with~n°n'mer8ed banks, increased their relative holdings of time and sav- ings deposits, and changed the composition of their loan portfolio less than non-merging banks. THOMAS SNIDER The second notable investigation of merged banks shown in Table 3.1 was conducted by Thomas Snider.3 He investigated whether loan port- f°11°3 0f formerly independent rural banks changed significantly after the rural banks were acquired by urban banks. His sample consisted of 24 TABLE 3.1 SUMMARY OF SELECTED RESEARCH INTO THE EFFECTS OF MERGER ON BANK PERFORMANCE PERFORMANCE RESEARCHER MEASURE Smith Snider Cash Assets ' Time and Savings + Deposits as a percent of Total Deposits Changes in Loan - Portfolio Composition Five Loan Portfolio NSD Ratios Where - represents a lower level after merger compared with before. + represents a higher level after merger compared with before. NSD represents no significant differ— ence before and after merger. 35 urban-m 3m 3%! X was me asset 3". :17 ESE: 25 36 urban-rural mergers which occurred in Virginia during the period June 30, 1960 to June 30, 1968. Each of the formerly independent banks was paired with a non-merging independent bank of the same approximate asset size and located in the same banking market. Data were collected for three years prior and for three years after the merger (but not including the year of the merger) and were used to calculate five loan portfolio ratios based on averaged data pre— and post—merger for each bank. Like Smith, Snider used univariate t-tests to check for statistical significance. Tests were applied to the mean difference for each ratio (average of the differences between each bank pair) and to the difference in each ratio before and after merger. Results indicated that there were no statistically significant dif— ferences between the merged and non—merged banks either prior to or fol- lowing merger. Snider therefore concluded that "...recent changes in the Virginia banking structure have not materially affected the amount or type of bank credit available in rural areas."4 In summary, there have been two notable investigations into the effects of merger on bank performance, one by David Smith and one by Thomas Snider. If new MBHC affiliates behave as merged banks, then based on the Smith and Snider findings, the affiliates will experience decreases in cash assets, will increase their time and savings account deposits relative to total deposits, and will make few changes in their loan portfolios, BRANCHING More evidence exists describing the performance of bank branches than exists for merged banks. As in the case of the merged banks, the major interest of most of the research has been concerned with how Mrs W 31 0n the idiom '1: research into diseussion 0' an: ‘ M is an h Y V hm mum“ i ‘gftkKM A h in: 26 branches (or acquired banks) change compared to independent banks. On the following page Table 3.2 summarizes key findings of selected research into the effects of branching on bank performance. A brief discussion of each study follows. KOHN and CARLO Kohn and Carlo studied the competitive impact of new bank branches opened in New York State between 1950 and 1961.5 Their objective was to determine whether recently created branches of commercial banks, sav- ings banks, and savings and loan associations had an adverse effect on competing institutions. They used three profitability measures and the growth rate of total deposits to measure the impact of new branches. To eliminate erratic fluctuations, three year averages of these measures were employed. Each measure was then calculated for every bank in a nine county sample (79 banks total) for the three year period prior to opening of the first branch in that county and for the three year period following the Year in which the first branch appeared. Additionally, ratios were calculated for various three year periods terminating with 1967 to de— termine whether adverse influences required more than three years to appear. The ratios were contrasted with identical measures for various c°ntr01 groups over the same time periods. The control groups were se- lected 80 as to be similar in size, structure, and environment. Sign tests and t-tests were used to check for statistical significance. There was general agreement among the tests that no statistically sig- nificant adverse effects in any of the time periods existed for the three measures of bank profitability. However, significant adverse 27 TABLE 3.2 SUMMARY OF SELECTED RESEARCH INTO BRANCHING EFFECTS ON BANK PERFORMANCE PERFORMANCE Kohn* MEASURE Carlo Profitability NSD Deposit Growth - Rate Lending Activity Operating Expenses Loans to Assets Interest Rates Charged on Loans Service Charges on Checking Accounts Interest Paid on Savings Accounts *Changes occurring to a bank or branch after RESEARCHER New York State Schweiger Branching Department* McGee+ acquisition compared with performance before acquisition. fBranch bank performance compared with unit bank performance. Horvitz Shu11* Where - represents a lower value for the branch bank. + represents a higher value for the branch bank. NSD represents no significant difference. 28 effects were found for rates of deposit growth in two of the nine counties studied. Deposit growth slowed by 4-6 percent for these two counties depending upon the control group used for comparison. These same results, viz., branching had no significant impact on bank profitability but did decrease the rate of deposit growth of com- peting banks, were found when Kohn and Carlo made.similar tests on a subset of banks whose home offices were close enough to recently estab- lished branches to be subject to maximum competitive impact of branch- ing. Kohn and Carlo concluded with a suggestion that if geographic limits to branching were extended, no adverse effects would be notice- able in the performance measures they studied. NEW YORK STATE BANKING DEPARTMENT Research reported by the New York State Banking Department analyzed 39 banks which converted to branch offices between 1946 and June, 1957.6 Various loan ratios pre- and post-branching were compared. They showed that the lending activities of these banks increased substantially after merger. However, field interviews indicated that remaining independent banks in areas where a unit bank was absorbed through merger were not adversely affected by this increased lending activity. In fact, the in- dependent banks "sharpened up" to meet this new competition. This find- ing suggests that pairing of independent banks with similar recently acQuired banks located in the same banking market may hide some of the Changes occurring to the acquired banks because the control group (in- dependent banks) may change. SCHWEIGER and McGEE In a much quoted study by Schweiger and McGee, the adequacy of the financial structure in Chicago and in Illinois in 1960 was investigated.7 29 { Although this study was not directed at comparing the effect of bank acquisitions on independent banks, some light was shed on the matter. They found: 1. Branch banks grant a larger volume of loans than unit banks of the same capital and deposit size; 2. Unit banks competing with branch banks in the same areas had higher loan to asset ratios than did unit banks in unit bank- ing areas; and 9) 3. Branch banks have higher operating expenses than unit banks of the same size. Hence this research suggests that if independent competing banks do in fact change their operations in reaction to the entry of a larger bank through merger of a competitor, it would be expected that the in— dependent banks might raise their loan volume and thereby experience higher operating expenses than similar non-competing banks. Based on these findings and on the results of research into several other areas related to financial institutions in Chicago and Illinois, Schweiger and McGee recommended relaxed entry laws for banking. HORVITZ and SHULL Additional support for the Schweiger and McGee findings was es- tablished through research conducted by Paul Horvitz and Bernard Shull.8 Questionnaires requesting information about prices and services pre— and [nastdmerger were sent to all national banks that acquired other banks tlirough merger in 1962. Postemerger data represented bank operating ex- Perience at a point in time at least two years after merger. Sixty- three responses were studied. Results indicated that the pricing and :I‘DEIn.policies of most acquired branches were identical to those of the atlquiring bank. However, there was a tendency for the acquired banks ‘3‘) ‘pay a higher rate on savings accounts and charge lower rates with 30 more liberal terms on loans after the merger than before. Additionally, service charges on checking accounts and the number of services offered by the acquired branch both increased. These results are similar to those found in the New York State Banking Department study described earlier. Horvitz and Shull then investigated 3,000 Federal Reserve member banks to determine the effect of branch banking on bank performance. Results showed that bank profitability did not appear to be related to branch banking laws. There was no consistent pattern to the degree of profitability of banks in branch banking states versus banks in unit branching states. However, they did find that unit banks located in states permitting branch banking have lower loan to asset ratios and higher ratios of time deposits to total deposits than.branch banks. That branch banks generally have the highest loan to asset ratios was also found. Finally, Horvitz and Shull found that, in isolated towns, the rate of interest paid on time deposits was higher when unit banks established branches. In summary, much more research into the effects of bank branching has occurred compared with research into bank mergers. If new MBHC affiliates behave in a manner similar to new branches, the results of this research provide insights into how the affiliates might change £1fter acquisition. The studies suggest that new affiliates will ex- Perience no significant change in profitability, will increase lending £i<:tivity, and will incur higher levels of operating expenses. Results aalso indicate that state branching law must be considered when investi- EEerting the performance of banks. Although changes caused by branching law were not great, they did exist. 31 BANK HOLDING COMPANY ACQUISITIONS Most of the recent research concerned with the impact of acquisi- tions and branching on bank performance has dealt with bank holding company affiliation. In general, all of the holding company studies have found that independent banks acquired by holding companies do not operate or perform much differently than independent banks. For ex- ample, Peter Rose and Donald Fraser state that: "Bank holding companies seem to have much less effect on the performance of their affiliated institutions than both the supporters and the critics of the holding company movement would have us believe."9 A brief description of the pertinent research in this area will make clear the extent of influence holding companies have had on their affiliates. Table 3.3 on the fol- lowing page summarizes selected findings of this research. THOMAS PIPER In a research paper for the Federal Reserve Bank of Boston, Thomas Piper investigated the acquisition activities of registered bank hold- ing companies10 during the years 1947 through 1967.11 One area investi- gated was the post-acquisition performance of the acquired banks. The study was based on the 146 acquisitions for which complete data were available during this twenty year period. Omitted from this sample Ivere holding company acquisitions of lead banks. These are typically Jearge banks which are acquired not as investments but represent merely a reorganization to the corporate form of business. They are not repre- 8entative of most banks acquired by holding companies. In fact, Piper f<>L1nd that at the time of acquisition the holding companies were twenty t131-mes larger on average that the acquired banks in his sample. \ \ h a.“ v 55 last: its ( L396: Hinge «rd 32 TABLE 3.3 SUMMARY OF BANK HOLDING COMPANY RESEARCH FINDINGS: RELATIVE CHANGES OCCURING IN NEW AFFILIATES RESEARCHER 9-339) ”MHMZHMK Zomzmog ~c1 cram PE RFORMAN CE mASURE mozmwz>r mwtersL-‘oti NMEH>H was»: Z>§ZHdHiO£fl WNWH’U OQZHK MUU>WQ lwmmanm Cash to Assets - - - Loans + + + + + + + S 1: ate and + + + + + + Municipals U-S - Gov't. - - - - - — Securities Demand Deposit + + Fe es Operating Revenue + + + Operating Expenses + + Othe 1' Operating + + + + Expenses Fringe Benefits + Re t Income NSD NSD NSD beige-11d Deposits to + + otal Deposits C ap 1tial to Assets - Ca pital to Deposits + + + + Co nannier Loans + + \ EEITEB NSD represents no significant difference between bank pairs, + represents a higher value for the acquired bank relative to the paired independent bank. and - represents a lower value for the acquired bank relative to the paired independent bank. size the 1 nizel repo' Edie the ' cper; fora; He f: Peri.- inde; Eigh- arEa Vere tiSt: fete: 33 He paired acquired banks with unacquired banks of the same deposit size, his only pairing criterion. The independent and acquired banks in the pairs were similar at the time of acquisition. However, he recog- nized that "...affiliation might result in significant operating changes "12 ‘both by the acquired bank and by the (paired) competing institution. Thus the problem of a changing control group described earlier in this report was recognized in writing for the first time. Piper, therefore, made a second test comparing asset structure of the acquired banks at the time of acquisition and at year-end 1967 with all commercial banks operating in the host state. His intent was to check for actual per- formance differences that may not have appeared in the paired comparison because of proportionate changes occurring to each of the paired banks. He found evidence that affiliates tended to change their asset compo- sition after acquisition, but these tendencies were not significant at the 5 percent level. ;T(3E McLEARY In somewhat related research, Joe McLeary investigated the 1966 Performance of a group of holding conmany banks versus a paired group of independent banks of approximately the same total deposit size.13 Eighty-two pairs of banks located in the same county or metropolitan a‘4"5'ea in the Sixth Federal Reserve District comprised the sample.14 Data wet‘e collected for eleven measures of performance. To check for sta- tistical significance, univariate t-tests were applied to the pair dif- fele’ences for each measure. Results of this cross-section study indicated that the holding Q(Jllapany banks were very similar to the independent banks with only the fOllowing exceptions. Significant differences were found on loan 34 interest rates, U.S. Government securities held as a fraction of total assets, municipal securities to total assets, and time deposits to total deposits. While the ratio of municipal securities to total assets in- <2reased for the acquired banks relative to the independent banks, the three other measures were all lower for the affiliated banks. McLeary concluded that each proposed acquisition must be evaluated on its own merits; results of this research were not convincing enough to use for making policy decisions. One possible defficiency in McLeary's study arises from the rather large difference in average bank size of the acquired banks compared saith the independent paired banks. Deposits of the affiliates were on average fourteen million dollars greater than deposits of the indepen- dent banks. Because it is possible that economies of affiliation exist for small banks but decrease as bank size increases, the relatively large difference in average bank size may have affected McLeary's find- 11138.15 In a second area of research, McLeary investigated whether absen- tee ownership of banks affiliated with holding company groups affected tl‘teir responsiveness to the banking needs of the local areas served. Using ten performance measures and a 1967 sample of 23 locally con- t"~‘<>1led holding companies versus 59 holding company banks located out- sZlcle the county or metropolitan area of the lead bank (the absentee group), he found that the operating performance of the local groups was fa:Lrly similar to the absentee group and also similar to that of inde- pendent banks in the same area. 35 ROBERT LAWRENCE In another study of the effect holding companies have on acquired banks, Robert Lawrence compared 43 acquired banks with 55 banks of simi- lar size in the same town.17 The acquired banks were selected from 123 holding company acquisitions occurring from 1954 through 1963. Lawrence found that the acquired banks were typical banks one year prior to ac- quisition. Differences in means of selected performance measures for the acquired group versus the independent group indicated that at the 95 percent confidence interval, the only statistically significant dif- ference prior to acquisition was "due to banks as a percent of total deposits." After acquisition Lawrence found that, compared with indepen- dent banks, the affiliates held greater amounts of state and local obli- ,gations and loans to total assets, less cash, and more capital. He .also found that service charges on checking accounts were higher for the affiliates. :SAMUEL TALLEY Samuel Talley updated Lawrence's study by investigating 82 pairs of flanks based on acquisitions taking place between 1966 and 1969.18 He found that the affiliates held more municipal securities and made more consumer loans than independent banks, but held less U. S. governments. 3E!) contrast to Lawrence's study, checking account service charges were found to be lower at the affiliate banks. JOHNSON and MEINSTER Rodney Johnson and David Meinster completed one of the first Studies using multivariate statistics. They stated: 36 We hypothesize that some performance measures used in the earlier studies are highly correlated with each other and that some measures might act differently in combination than they would if tested separately. Therefore multivariate analysis should yield different results. They used data acquired from questionnaires on 20 different mea- sures for 36 pairs of banks. Performance before acquisition was com— pared with bank performance 1, 2, and 4 years after acquisition. Using univariate statistics, the greatest changes were found to be in the asset structure of the acquired banks. However, using multi— variate discriminate analysis, largest changes were found in bank pric- ing, not asset composition. This demonstrates that when several mea- :sures are used, multivariate estimates can be differenct from uni- wrariate'estimates. Johnson and Meinster found that after acquisition, service charges on deposits, interest rates received on U.S. governments, and loans as :3. percent of total assets all increased for affiliates, but interest rates received on loans decreased compared with independent paired banks . ROBERT WARE Robert Ware studied acquisitions occurring in Ohio during a three year period.20 Acquired banks were paired with independent banks lo- Qated in the same market. Pair differences one year prior to acquisi- tion were compared with pair differences 1, 2, and 3 years later using ‘Jtlivariate t-tests. Like other researchers, he found that MBHC affili- atea held less cash and U.S. government securities and more loans and Inlinicipal securities compared with the paired independent banks. He also fOund that other operating expenses were higher for affiliates. 37 STUART HOFFMAN Stuart Hoffman studied the acquisitions of two Florida MBHC's.21 Thirteen affiliates of each holding company were paired with independent banks. Using univariate statistics on 28 measures, Hoffman's findings were consistent with those of other researchers, viz., the affiliates held less cash and U.S. governments but more loans and municipal secu- rities as a percent of total assets than independent banks. He also found that increases in revenues offset increases in operating expenses for the affiliates compared with independent banks. JOHN MINGO In another study that used multivariate statistics, John Mingo in- wrestigated whether holding company affiliates take greater risks com- }Jared with independent banks.22 He theorized that this might be the (:ase because: 1. There are activities available to affiliates that are not available to independent banks, such as leasing, factoring, and insurance selling; 2. To justify the relatively high prices paid for new affiliates, return and therefore risk must be high- er; and 3. The relatively great separation between owners and managers of holding company banks may induce mana- gers to operate affiliates in a riskier manner. To test this proposition, 384 banks in 9 states were studied - 134 were holding company affiliates and 250 were independent banks. All banks had assets less than $70 million. Mingo found that the affili- a~tes operated with more financial leverage (capital to assets was low- er) compared with independent banks. This was evidence that affiliated banks in Mingo's sample were operated in a riskier manner than indepen- clent banks. 38 IUCILLE MAYNE A study whose objective most closely resembles the objective of this research was conducted by Lucille Mayne.23 Her research goal was to determine whether there were significant differences in the opera- tions of bank holding company affiliates and independent banks during the 1969 through 1972 period. A sample of 656 banks, evenly divided be- tween affiliates and independent banks, was studied. No attempt was ‘nade to measure the affects of holding company acquisitions. Rather a cross-section analysis was performed on each of the four years inves- tigated. In this manner operating differences between affiliates and :competing independent banks that continually persist for years after ac- quisition could be measured. The sample was drawn from 28 states. Differences on eighteen per— :ftarmance measures were investigated using multivariate linear regres- sion. Results of the study indicated that compared to affiliates, in- dependent banks 1. Held more cash and liquid assets to total assets but less municipal securities to assets, 2. Had less capital to non-liquid assets, 3. Had lower service charges on deposits, 4. Had higher wage expense as a percent of assets, 5. Experienced higher loan losses, and 6. Had lower profitability (net profit to total assets, net profit to total capital, and after-tax income before total gains and losses on securities to total assets were all higher for affiliates). Sh" (:oncluded that the higher profitability of affiliates appeared to be cans ed by three contributing factors--assulllP111°n 0f greater risk, high- e r 8ervice charges on deposits, and economies of operation. Because the 39 other research described in this section compared new affiliates with independent banks, but Mayne's research was not directed at new affil— iates, these findings are omitted from Table 3.3. Her study differs from this research in several respects. First, the average length of time of affiliation in the markets she studied was 10 years. In this research only those markets experiencing initial MBHC entry were selected for study. Another difference is the inclu- sion in this research of several independent variables not included in the Mayne study. Except for holding company affiliation, only bank asset size, bank charter type, and bank location were included by Mayne as independent variables. Further, the sample selected in this study is limited to affiliates of only MBHC's. Mayne's study included af- filiates of one-bank holding companies as well. A final difference is that the error in matching bank pairs by asset size is considerably less in this research than the approximate 50 percent error accepted by Mayne. The degree of success in matching banks is explained in more detail in Chapter V. DALEY and GRADDY Finally, the most recent study of MBHC acquisitions was completed by Daley and Graddy.24 Using 150 bank pairs from the four states that have experienced the greatest amount of MBHC acquisition activity since ‘19 70, a cross section multiple discriminate analysis using 1975 data was IDeirformed. Excluded from the sample were de novo acquisitions and one batik holding company acquisitions. Counties were defined as banking Iairkets. Their most important findings were that affiliates earned a l..‘zlgher return on net worth and experienced higher operating expenses E Q operating revenues than paired independent banks. 40 In summary, past empirical research has been directed at examing changes occurring to acquired banks after affiliation. Using a variety of ratios and growth rates as performance measures, most studies have been fairly consistent in finding that compared with paired independent banks, affiliates held less cash and U.S. governments but more loans and municipal securities as a percent of total assets. No significant changes have been found in net income as a result of affiliation. Only the most recent studies have used multivariate statistics. DIFFERENCT APPROACHES OF PRESENT RESEARCH The research described in the remainder of this report differs in tihree ways from most of the bank holding company studies just described. These different approaches are: l. The sample will be limited to only those MBHC acquisi- tions representing the initial MBHC entry into each market; 2. An independent bank competing with a new MBHC affiliate will be paired with the affiliate and with a non-compet- ing independent bank located in a different market to check for changes in bank behavior; and 3. The appropriateness of multivariate statistics in meas- uring performance differences among banks for the variables used in this study will be investigated. Discussion of each of these different approaches is presented below. SAMPLE SELECTION Past researchers have in most cases selected their samples by in- QJ~uding all MBHC acquisitions occurring during a specified time inter- ."‘Elnl or in a certain geographical or political area. Thus included in their samples were MBHC acquisitions which represented the first, sec- 011d, third, etc. MBHC acquisition in that market. However, to measure tlle full undistorted impact of MBHC entry on the new affiliates and 41 remaining independent banks, reactions of independent banks to the in- itial MBHC entry must be considered. This is necessary because if af- filiates or competing banks do change their operations or policies in response to MBHC entry through acquisition, it is possible that their response to the first MBHC acquisition in the market would be different than their response to the second MBHC acquisition. Likewise their re- sponse to the third MBHC entry might be different than their response to either the first or second, and so on. If reactions of competing banks to MBHC acquisitions within their market are a function of their reactions to prior holding company en- tries, failure to consider the extent of MBHC influence already present :in the market may introduce confounding variables into the research and Iaossibly distort the findings. In order to measure the full undis- t:orted impact of MBHC acquisitions on the new affiliates or competing independent banks then, the research sample must be limited to only tzhose banking markets experiencing initial MBHC entry. In this way confounding variables caused by existing MBHC influence in the market can be minimized . PA IRING TECHNIQUE A second difference in approach concerns the pairing technique ‘13 ed to measure changes. Most researchers have paired the new affili- ate with an independent bank located in the same banking market. Us- :1413Lg this technique, if no statistically significant differences between ‘31E1e pairs were found before and after acquisition, the researchers rea- :E3‘:>ned that MBHC acquisitions did not significantly affect Operating or :1573Lnancial characteristics of either bank in the pair. Yet a possible 1:”Gamson that relatively few significant differences were found may be that th larger ence be their 0 Tc in race banks a nificar fore ar mtaff this re threats deemed to redr CESh hc aSSEts banks ( Sc States the aCc tiOns, researc d0ne SC a bEfoI Pairs a WhethEr thanged 42 that the control group of independent banks reacted to the entry of a larger competitor by themselves changing. Hence a significant differ- ence between the banks may not appear even though both did change their operations significantly. To see this deficiency, suppose the ratio of cash to total assets in recently acquired banks is studied and that independent competing ‘banks are paired with these banks. It is quite possible that no sig— riificant differences would be found between the two groups using a be- fore and after comparison. It might then be concluded that MBHC's do riot affect the liquidity position of their affiliates as measured by t:his ratio. However, if managers of the paired independent banks felt t:hreatened by acquisition of a competitor, they might take actions Cieemed necessary to remain competitive. One of these actions might be t:o reduce cash balances to a minimum. If the MBHC also reduces the cash holdings of its new affiliate, no difference in the cash to total Elaisets ratio would be found between the two banks even though both banks did experience a significant change. Some researchers have recognized this problem. For example Piper £31:zates that "...affiliation might result in significant changes both by t311.1.eacquired bank and by the competing institution. Under such condi- tions, use of paired bank comparisons could be misleading."25 Those researchers who have attempted to overcome this potential problem have ‘1<>Ine so in one of two ways. The most common approach has been to make ‘1 lbefore and after acquisition test for all independent banks in the I>£iilrs and then for all affiliates in the pairs. This will indicate whether the independent banks as a group, or the affiliates as a group, (:}1£Mnged after MBHC entry. If no statistically significant differences as m gt e: D: 1 | O 43 between the pairs are found, and then it is found that both groups did change after acquisition compared to before, researchers concluded that the changes were attributable to holding company entry. Evidently the independent banks "sharpened up" in response to the acquisition as the New York State Branching Department study suggested. A problem with this technique is that the differences may represent secular changes in the banking industry and may not necessarily be a result of MBHC entry. A second approach used to avoid the problem of not finding re- :sponses occurring equally to both paired banks was used by Piper. He (rampared a group of acquired banks with a group of all other commercial t>anks in the host state. The intention was that if only MBHC entry czaused the acquired banks to change, differences would be found between these two groups, unless secular changes in the banking industry sys- tematically affected one group more than the other. Unfortunately the Eleverage bank size of Piper's two groups was substantially different. Ufllis is a major limitation of that technique. Most researchers have found that banks acquired by holding companies are not typical banks. They are smaller than average size banks. A more appropriate technique, the one used in this research, is to unailse a three-way comparison. After checking for statistically signifi- <2£111t differences between the paired independent banks and affiliates, <>Il€2 of the groups could be compared with a group of similar size banks :Ltl other markets void of MBHC affiliates. Secular changes in banking $11(Duld affect both groups equally. If no difference between the inde- r>‘311dent-affiliate pairs are found, it may be that MBHC entry does not 81Stlificantly affect competing banks or it may be that banks in both EgjbcDUps changed equally. But if differences are then found between the 44 independent banks and a group of similar size independent banks in mar- kets void of MBHC affiliates, it may be more strongly deduced that MBHC entry was responsible for some changes occurring to competing banks in its market. Changes resulted from MBHC entry, not secular changes. APPROPRIATENESS OF MULTIVARIATE STATISTICS The final difference in approach concerns the reliance of research- ers on univariate statistics to test for statistical significance. Uni- variate statistics may be inadquate when used alone to test several highly correlated measures of performance. For example, Johnson and Meinster explain that if 40 ratios are being used and the confidence in- terval is 95 percent, then on average two of the ratios (5 percent x 40) will turn out significant when, in fact, the true differences for all 40 measures is zero.26 Evidence that findings may be different if multi- variate statistics are used is furnished by the Johnson and Meinster study cited earlier. Performing a series of univariate tests to check for statictical significance may be inappropriate for a second reason. When several de- Pendent variables are used as measures of whether an independent varia~ ble makes a significant difference in performance, a series of univari- ate tests might detect a small but insignificant difference on each of the dependent variables. The researcher might be undecided as to wheth- er the independent variable had no affect on performance or whether all of the slight differences together indicate a significant difference. Multivariate statistics provides a method of working with these types of problems. The differences among the groups stratified by the independent variables are evaluated in terms of all dependent variables Qonsidered simultaneously. In general terms univariate scalar values C3: >(‘J 45 are replaced with matrices of dependent and independent variable char- acteristics. It is the use of these matrices that allows all of the dependent variables to be considered simultaneously. Multivariate a- nalysis will generate a T- or F-ratio which can be used to accept or re- ject the null hypotheses based on analysis of all dependent measures considered together. This is quite unlike a series of univariate tests which would give a t- or F-ratio for the null hypothesis on each of the dependent measures. In this research both univariate and multivariate tests will be performed. A comparison of the estimated coefficients obtained will in- dicate whether multivariate tests are more appropriate than univariate tests. SUMMARY To gain insights into how researchers have tested for performance differences between banks and into what responses might be expected in this present study, the investigations of several researchers have been described. These investigations were categorized by the type of bank- ing activity studied and included bank mergers, new branches, and hold- ing company acquisitions. Two studies into the effects of merger on bank performance and four studies describing the responses of branches compared with independent banks were described. Considerably more re- search into the impact of bank holding company acquisitions on new af- filiates exists than for both of these two other categories combined. The research is also more recent. A description of the different hold- ing company studies concluded that findings were fairly consistent. Compared with independent banks, new affiliates held less cash and U.S. governments but more loans and municipal securities as a percent of 46 total assets. Affiliation did not appear to affect profitability of affiliates versus independent banks. After these three categories of past research were reviewed, two unique approaches of the present study and one approach seldom employed in most research were described. In short, this research is unique be- cause the sample is limited to only those MBHC acquisitions representing the initial holding company entry into each market and because competing independent banks will be paired with both the new affiliates and non- competing independent banks to check for coincident changes occurring to both affiliates and competing banks. The appropriateness of using mul- tivariate instead of only univariate statistics, something only the most recent MBHC research has considered, will be examined. In the chapter that follows the research model is developed and explained. Ke at: 52 O. r a a Cult. B~B rL. 3.x MU. 47 FOOTNOTES TO CHAPTER III 1The Fourth District includes Ohio and parts of Pennsylvania and Kentucky. 2David L. Smith, "The Performance of Merging Banks," Journal of Business, (April, 1971): 184-192. 3Thomas E. Snider, "The Effect of Merger on the Lending Behavior of Rural Banks in Virginia," Journal of Bank Research, (Spring, 1973): 52-57. 41bid., p. 52. 5Ernest Kohn and Carmen J. Carlo, The Competitive Impact of New Branches, (Albany, New York: New York State Banking Department, December, 1969). 6Postwar Banking Developments in New York State--A Summary Report, (Albany, New York: New York State Banking Department, 1958). 7Irving Schweiger and John S. MbGee, "Chicago Banking," Journal of Business, (July, 1961): 203-366. 8Paul Horvitz and Bernard Shull, "The Impact of Branch Banking on Bank Performance," The National Banking Review, (December, 1964): 143-188. 9Peter 8. Rose and Donald R. Fraser, "The Impact of Holding Company Acquisitions on Bank Performance," Bankers Magazine, (Spring, 1973): 91. 10A registered bank holding company was a bank holding company required to register with the Federal Reserve and required to gain its approval for any acquisition involving more than five percent of the acquired bank's stock. A holding company was required to register if it consisted of two or more banks. Excluded from registration until the 1970 Amendments were one-bank holding companies. 11Thomas R. Piper, The Economics of Bank Acquisitions by Registered Bank Holdinngompanies, (Boston: Research Report to the Federal Reserve Bank of Boston, No. 48, 1968). lzIbid., p. 165. 13Joe W. McLeary, "Bank Holding Companies: Their Growth and Per- formance," Monthly Review, Federal Reserve Bank of Atlanta (October, 1968): 131-138. 14Included were banks in Tennessee, Georgia, and Florida. 15For a summary of research that has been directed at measuring the economies of affiliation in banking, see Midwest Banking in the Sixties, "Bank Costs and Output--A Commentary on the Evidence," pp. 186-193, published by the Federal Reserve Bank of Chicago, March, 1970. 0 mm H u... 3% in .w b. a m Q C s: ...: um N. R F, “a H ......i 48 16Joe W. Mtleary, "Absentee Ownership-Its Impact on Bank Holding Company Performance," Monthly Review, Federal Reserve Bank of Atlanta (August, 1969): 99-101. 17Robert J. Lawrence, The Performance of Bank Holding Companies, (washington, D. C.: Board of Governors of the Federal Reserve System, June, 1967). 18Samuel H. Talley, The Effect of Holding Company Activity on Bank Performance, (washington, D. C.: Staff Economic Studies, Number 69, Board of Governors of the Federal Reserve System, 1972). 19Rodney D. Johnson and David R. Meinster, "The Performance of Bank Holding Company Acquisitions: A Multivariate Analysis," Journal of Business, (April, 1975): 204-212. 20Robert F. were, "Performance of Banks Acquired by Multibank Holding Companies in Ohio," Economic Review, Federal Reserve Bank of Cleveland (March-April, 1973): 19-28. 21Stuart G. Hoffman, "The Impact of Holding Company Affiliation on Bank Performance: A Case Study of Two Florida Multibank Holding Companies," WOrking Paper Series, Federal Reserve Bank of Atlanta (January, 1976). 22John J. Mingo, "Capital Management and Profitability of Pro- spective Holding Company Banks," Journal of Financial and Quantitative Analysis, (June, 1975): 191-201. 23Lucille S. Mayne, "A Comparative Study of Bank Holding Company Affiliates and Independent Banks, 1969-1972," Journal of Finance, (March, 1977): 147-158. 24Edward V. Daley and Duane B. Graddy, "Optimal Firm Size, Market Penetration, and Holding Company Performance: A Test of the Jessup- Piper Hypothesis," Conference Papers Series: #22, Business and Economic Research Center, Middle Tennessee State University, (June, 1977). 25Ibid., Piper, p. 58. . 26Rodney D. Johnson and David R. Meinster, "An Analysis of Bank Holding Company Acquisitions: Some Methodological Issues," Journal of Bank Research, (Spring, 1973): 59. CHAPTER IV THE RESEARCH MODEL The last two chapters have explained the factors that led to present restrictions on bank holding company formation and the research that has been directed at finding how banks have changed in response to mergers, branching, and holding company acquisitions. Furthermore, the ways in which this research differs from the described research was explained. This chapter now describes in detail the research model that will be used to test for statistical differences between bank groups. Discussion begins with a description of the research model. The rationale for in- cluding each dependent performance measure is then described. That sec- tion is followed by a similar section describing the rationale for in- cluding each independent variable in the model. Finally, the chapter ends with a summary. THE RESEARCH MODEL The effect of MBHC entry will be examined using the following multi- variate multiple linear regression model: 13165 + bi7H + b (l) 181 + b19D + e1 where R1 represents the dependent measures of bank performance as follows: R1 = interest on loans, total loans 49 50 R2 = before-tax interest on investments ; and total investments R3 = pperating expenses operating income ° The denominator and numerator in each of these measures are in dollars. R1 is the ratio of line Ala to line 9a on the Consolidated Report of Income. R2 is calculated as follows: R2 = Ald + Ale + Alg + Alf/(1 - A6b/A3). 2 + 3 + 4 + 5 In the numerator all terms are found on the Consolicated Report of Income while all terms in the denominator are found on the Report of Condition. Each term is defined below. Ald = dollars of interest received on Treasury securities. Ale 8 dollars of interest received on federal agency securities. Alg 8 dollars of interest received on other securities. Alf = dollars of interest received on state and municipal securities. The divisor of Alf is necessary to convert the yield on state and municipal securities to a before-tax value comparable with the three other sources of interest income. This step is required because interest income on state and municipal securities is not taxed while all other sources of interest income are. A6b divided by A3 is income tax paid divided by taxable income, i.e., the effective average tax rate. Lines 2, 3, 4, and 5 represent dollars of Treasury, federal agency, state and municipal, and other securities held, respectively. R is the ratio of line A21 to line Aln found on the Consolidated 3 Report of Income. It is the ratio of interest income, trust department income, demand deposit service charge income, and other income to operating expenses. These expenses include items such as salaries and wages, fringe benefits, interest paid on deposits, interest paid on borrowed money, occupancy expenses, depreciation, and loan losses. Tables 4.1 and 4.2 on the following pages are a Consolidated Report of Condition and a Consolidated Report of Income, respectively. They help explain the calculation of these three dependent measures. b represent net regression coefficients. 11 atqu: thEr. Year 51 C represents calendar years from 1969 through 1976. M represents Federal Reserve System membership. If M I l, the bank is a member. If M = O, the bank is not a member. L represents state branching law. If L = 1, limited area branching is permitted. If L 8 0, no branching is permitted. represents bank asset size in dollars. represents the Herfindahl Index for each market. represents dollars of personal income in each market. represents dollars of deposits in competing financial institutions in each market. CHEM X and Y have values of O or 1 and represent bank type. By manipulating the values of these two variables, a given bank type (affiliate, competing independent, or non-competing independent) can be contrasted with either of the two others. Although only X and Y are necessary to differentiate among the three types of banks, a third variable, 2, was created to simplify the analysis. Each type of bank can then be represented by a different variable and then the three comparisons to be tested can be represented as follows: 2 - X Independent banks (different market than the affiliate) versus MBHC affiliates. X - Y Affiliates versus competing independent banks (same market as the affiliate). Y - Z Competing independent banks versus affiliates. For example, consider the first comparison shown above (2 - X). Variable X in equation 1 would take on a value of O or 1 indicating either independent banks (2) or affiliates (X), respectively. Thus variable Y in equation 1 will hold constant effects of the remaining bank group--the competing independ- ent banks. The two other comparisons will be made in a similar manner. In every comparison the Y varia- variable represents the bank groups whose effect is held constant while the X variable represents the two groups being compared. The model will be tested for each of the ten years relative to the acquisition year (-3 through +6). The objective is to determine whether there are statistically significant differences between groups in any year or from year to year, The rationale for including each dependent Cm Mm” CO [Inc 6’5"? 52 TABLE 4.1 CONSOLIDATED REPORT OF CONDITION OF BANK AND DOMESTIC SUBSIDIARIES O Comptroller oi the Currency Administrator of National Banks CONSOLIDATED REPORT OF CONDITION 75E i'rrr‘l (Including Dornsstic Subsidiaries) CHARTER NUMBER: NAME OF BANK. CITY: STATE: ZIP CODE. cm; ARCH 31. 1978 ma 0.. m: I. Nuns NOTfiszmMNWbyanm.» wwwmmmmmmmnmm mummmmmmmmwd mumam.mmdnm mumwummmmmm momspsdfisd. TRIO oi the shove-hm bonkdo hsrshydscisrs that this Fisport oiOonditionistmshd comettothsbsstoimyknowisdgs and belief. mini! Vb. ummmunmaumam. mmdmmummdmm miss. WWNWW. Mdsdsrslhstithssbssnumnsdbym. Ioihshsstoiunumshdhslsiis unwound. mutnmmamupummmmwaymumsc m. mmanmumuum inrflsponolm. mummmedmm Hauntithsndphomnunbsroipsrsontowhomin- °"°°‘°" gums; may be new; LIABILITIES .8388: 83888858 8 BOW CAPITAL ”EMA up» 53 TABLE 4.1, CONT'D. NAME OF BANK: CHARTER NUMBER: BALANCE SHEET at the close 01 bum on m is ’- Ststement oi Resomoesend Lusbrlmes Sch. Item Col 1. Ceshendduetrombsnks .................... C 7 ........................... t 2. U.S.Treesury sewrities ...................... B 1 - E ........................... 2 3. ObligationsolomerU.S.Gov't.eoenoicssndoorps.. a 2 E .............................. 3 4. ObigetionsoISIetessndpoiiticsisubdivisions .. B 3 E ........................... 4 5. Other bonds. notes. and debentures ........... B 4 E ........................... 5 6. Federal Reserve stock and corporate stock ................................................. 6 7. Treding account securities ................................................................ 7 s. Federsl lundssoldsnd securities purohesed underegreementstoresetl .................. D 4 ...... r IXXXIXXIB 9. s. Loens.ToteI(exoluding unesmedinoome) A 10 XXX 0 ntummmmnmmmunbmwuu ....................... IIIIIIIIEZME c. Loens.Net ........................................................................... 10. Direct ioeselinsnoing ................................................................... 11. Benkprernises.iumnumendfixtum.endoMerssseunprmnungheMpmnim ............. 12. Reelesteteosmedotherthenbenkpremises ................................................ 13. Imestmentsin unconsolidated subsidieries sndessocietedcompsnies ......................... 14. Customers'IiebiIitytoIhisbenkonmptenoesoutstending .................................. 15. Other essets ................................ G 7 ........................... 16. TOTAL ASSETS (sumoi Iternst thru 15) ................................................... 17. Demand deposits otindividusls,prtnshps..endcorps. F 11 A ........................... 18. Time and sevings depositsoilndividuels. prtnshps..endoorps. ...................... F. It B+C ........................... 19. DepositsoIUritedStstes Government ......... F 2 A+B+C ........................... 20. OepositsoiStstessnd poiiticsiswdivisions F 3 A+B+C ........................... 21. OepositsoiioreignoovtandoiflcIel institutions... F 4 A+B+C ........................... 22. Depositsoioomnercielbm ................. F 5+6A+B+C ........................... 23. Certifiedlnd officers'oheoks ................. F 7 A ............................ 24. TOTALDEPOSITS(sumoIIIems17Ihru23) ....... s. Toteldemenddeposits .................... F 8 A ‘ XXX XX b.ToteItIrneendsevlnosm ............ F 3 8+0 XXXXX . Federeihmdspwohesedendsewritiessold under edreementstoroourohsse ........... E 4 ........................... Liehldeeiorborrowedrnoney ............................................................. mmwwmwum ..................................... mmwaumammmm .......................... Otheriebilities .............................. H 9 ........................... TOTAL LIABILITIES (sum or Items 24 m 29 em Items 24s A o) ......................... Subordineted notes end deoemees ....................................................... per veiue) ...................... Preluredstook No.sheresoutstendng Conmionstooke. Noshueseuthorized Reesrveioroontingenoieesndotheromitei reserves ......................................... TOTAL EQUITY CAPITAL (sum at Items 32 thru 36) ......................................... . TOTAL LIABILITtEs AND EQUITY CAPITAL (sum oi Items 30. 31. end 37) ..................... AverspeiorSOoeIenderdeysendingthrepofl dete: e Ceshendduetromoenks(oorresoondstoltem1ebove) .................................. Fw.mmmmwmmwem(mnmsm1.. .Totsiloens(oorrespondstosubitem9eebove) ........................................... Tunedepoeiuoifloo.000ormore(oomepondstome1m1dsmmapm4 below) .Toteldepoeltstoorrespondstoltemuebove) ............................................ Fed. Iundspurohssedsndsewntieseoidundersomtstoreourdisse(oon. amass“). .IJebIlItiesiorhorrowedmoneywoneepondstottemawove) .............................. h.TOTALASSETS(oorreepondsbitem16ebove) ....................................... . ShnmyiettersoioremINtetendingesoireportdete) ....................................... Tmoertticebsddeooutmdummmd81oo.0000m(mnoimondfl) ..... -Ommdepodtsinmmbd81oo.0000rmore(Mssotmndde) ................ F 09.9.0 moon MW 8881’ d 3°...QOU‘. ‘0” 2 IN) 54 TABLE 4.2 CONSOLIDATED REPORT OF INCOME OF BANK AND DOMESTIC SUBSIDIARIES (J Comptroller oi the Currency Administrator 01 National Banks CONSOLIDATED REPORT OF INCOME (motoring Donnell: end Foreign SubeIdIeries) CHARTER NUMBER; NAME OF BANK: CITY; STATE: ZIP CODE; "' ' m - rwm-amu:(mmuwm)mdm) mu Meme. title end phone wmher ot person to whom in- quiriesrney bedorected. l H Fen poems-u 55 TABLE 4 .2 , CONT 'D. NAME OF BANK; CHARTER NUMBER. __ SECTION A—SOUFICES AND DISPOSITION OF INCOME YEAR TO DATE (Indicate losses in parentheses) 1. Operating income: . lowest and tees on loans ............................................................. interest on balances with banks ....................................................... Income on Federal tunds sold and securities purchased under agreements to resell in domestic oltlces . . Interest on U.S. Treasury securities ..................................................... Interest on obligations 01 other U.S. Government agencies and corporations .................. Interest on obligations 01 States and political subdivisions oi the US. ........................ . Interest on other bonds. not”. and debentures ........................................... Dividends on stock .................................................................... Income lrom direct lease financing ...................................................... income lrorn fiduciary activities ......................................................... .Servicechergesondepositaccountsindomesticottices ................................... Other service charges. commissions. and tees ........................................... . Other income (Section D. Item 4) ....................................................... . Total operating income (sum ol Items 1a1hru 1m) ......................................... 2. Operating expenses: Salem and employee benefits ......................................................... lntemstonmcennmeddeposhdfloopmanmissuedbydonesticofiices ......... . Interest on deposits in toreign othces .................................................... interest on other deposits .............................................................. =a='*'-"-".=r0-*9999- .0969.- 30‘QQOUN N - . oi 33-x..- 000 3 Interest on subordnated notes and debentures ...................................... 885 » B313}. :79. . 1. Occupancyexpenseolbenkpremises.eross ................ IXXXIQI 2. LeeszRenlalincome ....................................... IXX‘! 3. Occupancy expenseoibanltpremises. Net ............................................ I. anitureandequipmentexpense ....................................................... j. vaisionlorpoeslblebenloeeesloractusinetloenbeses)(8ecbonc.hem4) .............. I1. Otl'lerexpenseHSectionEJhma) ..................................................... I. Total operating expenses (sum oi Items 2a thru 2k) ....................................... .mcomebeloreincometaxesanducunhesgainsorloseeultem-tnmzl) ................... . Applicable income taxes (domestic and toreign) (tax credits in paren.) ............................ .Incomebetoresecuritiesgainsorlosses(ltem3minus4) ..................................... . a. Securities gains (losses). Gross ............................... X XX' 1:. W income lanes (W and torsigltlltn credits in paren.) —mm $0 GUI c Sealing-“nonessenfim ........................................................... I 7. Netir1comelltem5plusorminus6c) ....................................................... F— OR .Incomebeloreeirtraordinaryitems ......................................................... .Extraordineryitenis.NetoltaxelIect(8el:tlon F.1tern2c) ..................................... 00‘] .NetlncomeIIteni7plusorrninus8) ........................................................ Fern WM S6 and independent variable in the regression model is presented in the two sections that follow. DEPENDENT VARIABLES Underlying the discussion that follows is the hypothesis that busi- ness managers seek to maximize shareholder wealth as measured by com- mon stock price. This implies that when a manager must make a decision which has more than one alternative, the one which maximizes the firm's common stock price will be chosen. Market price can be expressed as n P0 = ZEt/(l+ke)t (2) t=0 where Po is the current common stock price, Et represents the earnings expected at the end of time period t, and ke is the rate of return investors require on common stock of a given risk class. In other words, market price equals the present value of all future earnings dis- counted at the investor required rate of return. It follows then that there are two general methods of increasing price-~increasing the ex- pected earnings or decreasing the required rate of return. Whatever reasons MBHC managers have in mind when the decision is made to acquire another bank, the ultimate purpose must be an increase in the holding company's market price. Favorable effects must appear in earnings, the investor required rate of return, or both if the acquisi- tion is to be a successful investment. Some of the more important rea- sons for acquiring a bank that would have the effect of increasing return or decreasing risk, and the implications these reasons have for bank performance, are presented below. For each reason discussed, one or two 57 measures of bank performance are described that will indicate whether the expected benefits of acquisition.were achieved. Because the same measures can be used for several of the reasons presented, only three different independent measures are needed to determine whether the hy- pothesized benefits were realized. DIVERSIFICATION Officers of an MBHC can reduce the overall risk of their firm by acquiring a bank whose portfolio of assets and liabilities does not have perfect positive correlation with its own portfolio. For instance, con- sider the case of an MBHC whose affiliates are located in large metro- politan areas and a rural bank located in an agricultural region. The loan portfolios of the urban banks may consist mostly of business and consumer loans while for the rural bank the loan portfolio may be domi- nated by agricultural loans. Less than perfect positive correlation between the portfolios of the rural and urban banks would be expected. Acquisition of the rural bank by the MBHC would reduce the riskiness of the combined loan portfolio because variability of returns would be reduced. Even when this "product line" diversification is not present, geo- graphical diversification will be. It results from the Federal Reserve Board's reluctance to approve those acquisitions where the proposed af— filiate and the holding company banks are located in the same banking markets. Asset and liability portfolios of these banks will not have perfect positive correlation because the banks will have neither the same borrowing customers nor the same depositors. Riskiness of the com- bined portfolio will be less than the riskiness of either prior to ac- quisition. 58 A conclusion from the foregoing discussion is that MBHC's may ac- quire banks to reduce risk. This is consistent with the hypothesis that managers seek to maximize share value: when risk decreases, the required rate of return will fall and thereby tend to drive stock prices upward, ceteris paribus. Prior to the proposed affiliation an.MBHC might be located at point A on the risk-return space shown in Figure 4.1 below. Expected Return U1 A Risk FIGURE 4.1: Initial MBHC Position Point A lies on line fAfA which represents the holding company's effi- cient frontier, i.e., the locus of all points which represent maximum return for any level of risk. Lines U1Ul and U are utility curves 2112 for risk averse bankers. Management of holding company A selects point A because its utility is maximized at that point, i.e., the highest utility curve that can be achieved, U1U1, is tangent to the efficient frontier at that point. If it could be achieved, management would pre- fer a risk-return point along curve U202 because U2U2 represents a greater level of utility. If this holding company acquires a bank of lower risk than repre- sented by point A, holding company A can reduce its over-all risk re- gardless of the new affiliate's return. Figure 4.2 shows this point. 59 Expected Return Risk FIGURE 4.2: MBHC Position After Acquiring Bank B. Suppose holding company A acquires bank B located at a lower risk, lower return position compared with point A. Management of bank B has selected point B along its efficient frontier, foB in Figure 4.2. Af- ter acquisition the holding company can achieve any risk-return position within or along envelope ABC. How will the holding company decide upon a final risk-return point? It will choose that point which maximizes its utility. Initially the holding company, considering its combined portfolio of A assets and B assets, may find itself at point D. Point D repre- sents the combined risk-return position immediately after acquisition. Return at this point is simply a weighted average of the returns of the affiliate and holding company together. Combined risk is a function of relative bank size, risk, and correlation between the two banks. The MBHC may choose to remain at this risk-return position con- tented with the same return at lower risk. If this were the case, meas- surements of risk should show that it has decreased. Unfortunately the time span covered by this research, 8 years, is not sufficiently long to allow an accurate calculation of risk. Because returns can be 6O calculated only annually,1 the number of observations for each acquisi- tion will range from 3 to 8 depending on the number of years for which data on each bank is available. A risk measure such as the standard deviation calculated on these few observations would be accurate only by coincidence. And even if the data for a sufficient number of years were available, such a measure would include risk for conditions that existed several years prior to the time of interest. However, instead of being contented with the same return but lower risk, management of the holding company may maximize its utility if it can achieve a point such as point E representing higher return and risk. This result can be accomplished by acquiring a bank and then changing the riskiness of its loan and investment portfolio and by increasing the degree of financial leverage.2 Immediately after acquisition the MBHC has reduced its risk through diversification and will be located at a point such as D in the risk-return space. The shift from D to E can be made by lengthening the maturity and/or increasing the average risk of both loans and investment securities. For instance, short-term securities could be replaced with longer term securities while cash and Treasury securities could be replaced with municipal securities. Both of these actions would increase risk and return without causing the banks to be illiquid because MBHC's could shift liquidity to affiliates if the need arose. The same approach can be applied to the loan portfolio to increase return and risk. This effect can be measured by comparing the rate of return gen- erated by the investment and loan portfolios of acquired and unaffilia- ted banks. Returns of MBHC affiliates should increase after acquisi- tion compared with unaffiliated or independent banks. The two measures 61 below would capture this effect: Interest on loans total loans Before-tax interest on investments total investments The contention that owners increase the riskiness of their affiliates to generate greater returns will be supported if differences between affiliates and independent banks on these two performance measures turn out to be statistically significant. Figure 4.2 shows affiliate B located at a position of both lower risk and lower return than holding company A. If risk is to be lower, then it follows that expected return probably is lower. After all, if an investor wishes to avoid risk, expected return will be lower com- pared to that available if riskier investments are acquired. However, the preceding discussion is not affected if a bank of lower risk but equal or greater return than the holding company is acquired. Combined risk will still be lower. But if the affiliate has the same return as the holding company, combined return will not change as a result of the acquisition. And if the affiliate has a greater return than the hold- ing company, combined return will be greater after acquisition. What- ever the case, risk will be reduced, and the effect of this risk reduc- tion will show up in the two measures cited earlier. If banks riskier than the holding company are acquired, it is likely that these new affiliates also have greater return compared with the holding company--higher risks are necessary to earn greater returns. Thus combined return after acquisition will be higher. And if the acquired banks are riskier than the holding company, a reason might be 62 that the new affiliates' loans are concentrated in one or a few local industries such as agriculture or tourism. These loans would exhibit high positive correlation among themselves. Yet the loan portfolio of the potential affiliate might be negatively correlated with the holding company's portfolio for the reasons described at the beginning of this section. Thus the holding company may still shift assets in the new af- filiate's portfolio to a higher return, riskier composition by substitu- ting municipal securities for Treasury securities and by channeling more consumer loans to it. It can safely do this because of the diversifica- tion achieved. Both performance measures cited earlier will capture this effect by increasing for the acquired bank relative to an independ- ent bank located in the same banking market. MANAGEMENT SUCCESSION Federal Reserve Board Orders indicate that many small banks, prior to acquisition by an.MBHC, are controlled by managers who are at or near the customary retirement age.3 Management succession in these banks is a potential major problem. If management retires in the absence of trained and competent successors, the performance of the bank is likely to suffer. Knowledge of this possibility leads investors to require a higher rate of return than otherwise. But an.MBHC would be a source of competent managers. The effect of acquisition would be a reduction in risk and the subsequent reduction in return required by investors. The relationship in Equation (2) shows how this will lead to a higher market price. It follows that MBHC's may acquire small banks because the po- tential exists to lower the investor required rate of return by supply- ing competent management to the bank. The lower risk return position thereby achieved can be altered by holding company owners by rest: manna this grapl sure: cent FINAL beca ance long acqu nanc fina inpa men: meas May the bani bani tive Uflde 63 restructuring the affiliate's loan and investment portfolios in the manner described earlier. The effect would be an increase in risk, but this is offset by the risk reduction achieved from product line and geo- graphical diversification. This effect will be captured by the two mea- sures cited earlier measuring return on the affiliate's loan and invest- ment portfolios. FINANCIAL LEVERAGE The ability to issue more debt than otherwise may be another reason MBHC's acquire banks. Debt capacity will be greater after affiliation because risk will have been reduced. Affiliation will reduce the vari- ance of the combined cash stream of the MBHC and its new affiliate as long as the streams do not have perfect positive correlation and the acquired bank is not riskier than the holding company. Favorable fi- nancial leverage achieved through the use of greater amounts of debt financing will increase bank returns. This should have a beneficial impact on the bank's stock. The operating expense to revenue ratio mentioned in the discussion of economies of affiliation can be used to measure this higher return. INEFFICIEnT MARKETS Areas of capital market imperfection exist. For instance investors may not possess sufficient knowledge about a firm to accurately access the market price of its stock. This is expected for many of the small banks that are candidates for holding company acquisition. Because the bank may be owned by relatively few investors, the stock may not be ac- tively traded. Estimates of its value would then be difficult to make. Under such circumstances it may be possible for a bank holding company with ast value. would nc Hou an oppor would be sent op; For exa: market. would he Acquisi' essary . return , for the: this ef moron Th is am indica size d milder-a ecOno 64 with astute management to acquire the bank at a price below its true value. Such an occurrence would benefit the holding company owners but would not affect the affiliate's returns or risk position. However, it may be the case that market imperfections may present an opportunity for MBHC's to increase the return of an affiliate. This would be expected when the potential affiliate does not have the invest- ment opportunities necessary to achieve its desired risk-return position. For example, there might be little profitable loan demand in the bank's market. Without a strong correspondent banking relationship, the bank would have to settle for a position below the BDEA line of Figure 4.2. Acquisition by an MBHC could supply the bank with the loan demand nec- essary to move its position to the efficient frontier, i.e., higher return for the level of risk accepted. The two measures used to test for changes in loan and investment return mentioned earlier will capture this effect. ECONOMIES OF AFFILIATION The desire to reduce unit costs through economies of affiliation is another possible reason for bank acquisition. Empirical evidence indicates that significant economies of scale exist for banks whose size does not exceed about $10 million in deposits, but these economies moderate as bank size increases.4 Whether MBHC's can achieve these economies by acquiring small banks is debatable. Scherer states that: ...The bulk of all scale economies in production are realized at the plant level; multi-plant production and physical distribution economies of scale appear to be modest or non-existent. Suppose then that two previously independent plants producing the same product are brought under the same corporate shell. What economies of scale will be realized? The answer must be little or none. The plants are already built; not much can be done to un- build them in order to increase their scale. “‘1 If tion, a bined e to Figu higher novenen risk ha positic Tl measure If the ex{Dense alrigRy been di and/or maXimi. thESe 65 If economies of affiliation do exist as a result of MBHC affilia- tion, an acquisition would tend to increase the holding company's com- bined earnings and dividends and therefore its market price. Referring to Figure 4.2, this would be depicted by a movement from point E to some higher return point lying above E. Diseconomies would be depicted by movement to a position below E. In both of these cases the effects of risk have been ignored. Its presence would cause movement to a final position along a line similar to but lying above BDEA in Figure 4.2. The presence of economies or diseconomies of affiliation will be measured using the following operating ratio: Operatingfexpenses Operating income If the result of an acquisition is an increase in earnings relative to expenses, this measure will capture that effect. SUMMARY In summary, five reasons that explain why MBHC's acquire banks have been discussed. All of the reasons ultimately lead to changes in return and/or risk and are consistent with the hypothesis that mangers seek to maximize shareholder wealth. Three measures will be used to capture these risk and return effects. These dependent measures are: Interest on loans; Total loans Before-tax interest on investments ; and Total investments Operatingiexpenses. Operating income It is cated mom varia latin coupe with an ex of th FEJER affec m£mbe as re diffe “Eats Cash Esmbe than The C and : 66 It is hypothesized that affiliates, compared with independent banks lo- cated in the same market, will experience higher levels for the first two measures and lower levels for the last. INDEPENDENT VARIABLES Presented below is the rationale for including each independent variable found in Equation (1) in the regression model. BANK TYPES X AND Y X and Y have values of O or 1 and represent bank type. By manipu- lating the values of these two variables, a given bank type (affiliate, competing independent, or non—competing independent) can be contrasted with either of the two others. The three comparisons to be tested and an example explaining how X and Y will be used was described on page 51 of this report. FEDERAL RESERVE SYSTEM MEMBERSHIP, M It is necessary to include this variable because membership may affect bank performance. One reason is that Federal Reserve System member banks are much more limited in the types of assets that qualify as required reserves than many non-member banks. A result is that differences in composition of asset portfolios and return on invest- ments is likely among member and non-member banks. For instance, the cash to assets ratios for member banks are higher than those of non- members. Gilbert has estimated that many member banks with assets less than $50 million make relatively little use of Reserve Bank services. The cost of membership for these banks averaged 11.2 percent of profits and 1.8 percent of equity capital in 1976.6 Thus it would be expected di: Di: of ti< of 563 f0: bre de: Am ‘1': ha: lit in 67 that, since most banks acquired by MBHC's have total assets less than $50 million, Federal Reserve membership will cause banks to be less profitable than non-members. STATE BRANCHING LAW, L All states closely regulate the extent to which banks can branch. Some permit no branches, others permit branching only within a limited geographical area, and some states permit statedwide branching. These are described as the unit, limited, and statedwide branching states, respectively. Inclusion of this variable is necessary if the effect of differences among banks due to branching laws is to be held constant. Differences would be expected to arise from factors such as economies of scale that accrue to branches but not to one large bank, risk reduc- tion achieved through customer and geographic diversification, and level of competition, which would be reflected in output prices and types of services offered. In fact, several researchers have found differences in bank per- formance because of state branching law. Bell and Murphy found that branching tends to raise bank costs.7 Schweiger and.McGee found evi- dence that branch banks have higher operating expenses than unit banks.8 And a study by the New York State Banking Department found that banks with branches increased their lending activity relative to unit banks.9 It would be expected then, that banks acquired by MBHC's in unit banking states would have lower values of R than banks acquired in 3 limited branching states. In this study none of the markets included in the sample were from states permitting statedwide branching. There- fore a dummy variable with a value of 1 was used to indicate a state ":1 C) v 68 permitting limited branching and a value of O to indicate a unit branch- ing state. BANK ASSET SIZE, 3 Several researchers have found that bank performance is dependent on bank size measured by total deposits or total assets. Schweitzer found that scale economies exist for very small banks with assets less than 3.5 million dollars, constant returns exist for banks within a 3.5 million to 25 million dollar range, and decreasing returns exist for banks of asset size greater than 25 million dollars. Bell and Murphy state that since branching raises bank costs and most banks with branches are large banks, costs vary with bank size. They also found increasing returns to scale for many bank functions.10 It is necessary therefore to include bank asset size as an independent variable so that its effect on bank performance can be held constant while holding com- pany effect is measured. HERFINDAHL INDEX, R It seems reasonable that an MBHC holding a major share of deposits within a market would react differently to MBHC entry than would a bank which shares the market equally with several other banks. How a bank reacts to holding company entry is therefore likely to be a function of market concentration. A convenient measure of market concentration is the Herfindahl Index: n n - :8: 181 where Si is the market share of the ith firm. When there is only one fin: the CO‘ 69 firm in the market, the index will have a value of 1.00. As the number of firms increases, the value of the index decreases. Thus the higher its value, the greater the market concentration. This continuous varia- ble is therefore present in the regression equation. When the ratio has a high value, the market is concentrated and monopoly profits may exist. If this is the case, R1 may be greater for these concentrated markets than for more competitive markets because banks in less competitive markets may be able to charge higher interest rates on loans than other- wise. COUNTY INCOME, I Structure within a market is a function of supply and demand fac- tors and the level of competition. In this research, where the interest is in the commercial banking structure within a county, variables affect- ing structure which are of no research interest will be held constant by including them in the model to be tested. The Herfindahl Index specifies market concentration or the level of competition. County personal income, I, is included as a measure of demand and supply. The higher the personal income within a county, the greater the demand for such banking services as deposits, loans, etc. Further, bank output is a function of county personal income. The greater is I, the more loans a bank can make. Hence county personal income is included to hold market structure among counties constant. County income data were compiled from various issues of the "Survey of Current Business," U.S. Department of Commerce/Bureau of Economic Analysis. DEPOE a ma: anal savi serv the serw p13: ban d6! th th 70 DEPOSITS AT COMPETING INSTITUTIONS, D The demand and supply of substitutes for banking services within a market is another structure variable that must be included in the analysis to hold constant differences among markets. Savings banks, savings and loan associations, and credit unions all offer substitute services for commercial banks. However, in the case of credit unions, the services are not usually available on a countydwide basis. Rather services are available only to members who share a common bond such as place of employment. Therefore D includes deposits at only savings banks and savings and loan associations. Sources of deposit data were: Savings Banks FDIC Summary of Accounts and Deposits in All Mutual Savings Banks, various editions. Savings and Loan Associations FHLBS Summary of Savings Accounts, Member Savings and Loan Associations of FHLBS, Washington, D.C., various editions. The last two variables, I and D, are meant to hold supply and demand constant among markets. These two will not completely describe the supply or demand functions for each market. This is not the intent. Rather the intent is to capture the supply and demand effects. These two independent variables will capture the important components of those effects. SUMMARY The objective of this chapter was to describe the research model that will be used to test for significant differences in bank perfor- mance. A multivariate multiple linear regression model was hypothe- sized which included three dependent variables (return on loans, before- tax return on investments, and operating expenses to operating revenue) 71 and nine independent variables (calendar year, Federal Reserve System membership, state branching law, bank asset size, market concentration, countydwide personal income, deposits at competing financial institu- tions, and two variables indicating bank type). A detailed description was given of each variable. The remainder of the chapter was an expla- nation of the rationale for selecting the variables that appear in the model. The next chapter describes the research sample that will be used to test it. ofI ina :22 Res WIFUIZ‘.“ IL. I511 72 FOOTNOTES T0 CHAPTER IV 1Income figures are available from The Consolidated Report of Income which was issued annually prior to 1976. 2The effects of increased financial leverage will be discussed in a subsequent section. 3Frederick M. Scherer, Industrial Market Structure and Indus— trial Performance, (Chicago: Rand McNally, 1970): 116. 4For example, see the following Board Orders in the Federal Reserve Bulletin: --State Street Boston Financial Corporation, July, 1973, pp. 526—528. --United Jersey Banks, March, 1972, pp. 296-297. --Banc0hio Corporation, April, 1972, pp. 415-416. -4Valley Bancorporation, May, 1972, pp. 470-471. --Depositors Corporation, January, 1971, pp. 36-37. --First Holding Company, Inc., February, 1971, pp. 139-140. --Dominion Bankshares Corporation, March, 1970, pp. 307-309. 5For instance, see Thomas R. Piper, The Economics of Bank Acqui- sitions byiRegistered Bank Holding Companies, (Boston: Research Report to the Federal Reserve Bank of Boston, No. 48, 1968), and Eugene Rotwein, "Bank Mergers and the Bank Concentration in California in the Postwar Period," unpublished paper, Federal Reserve Bank of San Francisco, 1964, pp. 9-10. 6R. Alton Gilbert, "Utilization of Federal Reserve Bank Services by Member Banks: Implications for the Costs and Benefits of Membership," Review, Federal Reserve Bank of St. Louis (August, 1977): 3. 7Frederick W. Bell and Neil B. Murphy, Costs in Commercial Bank- ing: A Quantitative Analysis of Bank Behavior and Its Relation to Bank Regulation, (Boston: Research Report to the Federal Reserve Bank of Boston, No. 41, 1968): 179. 8Irving Schweiger and John S. McGee, "Chicago Banking," The Journal of Business, (July, 1961): 208-209. 9Postwar BankingTDevelopments in New York State--A Summary Report, (Albany, New York: New York State Banking Department, 1958). 1°1b1d., Bell and Murphy, p. 47. selec of t? ing a coma: firs tion sele inc] and cha] in tie CHAPTER V THE SAMPLE [The objective of this chapter is to describe how the sample was selected and to describe several of its characteristics. A discussion of these topics will be helpful in understanding the statistical test- ing and results described in Chapter VI. To examine the response of commercial banks to the initial market entry by MBHC's, markets must first be defined. Discussion therefore begins with the market difini- tion used in this study. Next is a discussion of how the sample was selected. Finally, several sample characteristics are described. These include the sample size, geographical location of the sampled banks, and the relative success achieved in pairing banks of equal size. The chapter ends with a summary. MARKET DEFINITION Prior to selecting the sample of banking markets to be included in this research, it was necessary to define a "banking market". Coun- ties are used as the definition. While arguments against this definition can be raised, several reasons exist for its use.1 These are: 1. State branching laws are frequently based on county lines; 2. Geographic barriers that were used to establish county lines also constrain banking markets; 3. In a study of prices in different markets in Florida, Salley found that 3 county definition worked as well as any other definition; 4. County boundaries do not change over time; 5. In many counties commerce revolves around the county seat which is located near the geographical center of the county; and 73 6. ll year pe will nc Income end of MBHC‘ are p achie initi Incl tric Peri as ( tho: com aft in 74 6. Much of the data is readily available only by county. The sample will also be constrained so that it encompasses the eight- year period from 1/1/69 through 12/31/76. Data prior to this time span will not be used because some of the Report of Condition and Report of Income categories needed for the dependent measures were changed at the end of 1968 and therefore are not comparable with data from later years. SAMPLE SELECTION An explanation of how markets experiencing initial entry by an MBHC were identified and the criteria used to select individual banks are presented in this section. The next section describes the success achieved in matching same size bank pairs. Eight states were surveyed to find those counties which experienced initial MBHC entry after 1/1/69. They were: Florida Ohio Iowa New Mexico Michigan Texas Missouri Wisconsin Included are all states in the Chicago and Dallas Federal Reserve dis- tricts that permit MBHC's and most of the other states in the U.S. ex- periencing greatest MBHC activity after 1968.3 Counties to be included in the sample were selected by surveying Summary of Deposit data by county for all FDIC insured commercial banks as of 6/30/76. Counties with and without MBHC's were identified. For those counties with.MBHC's, the date of acquisition of each holding company bank was determined. If the initial entry of an.MBHC occurred after 1969, the Federal Reserve Board Orders describing each acquisition in that county were inspected to determine whether a holding company re- lationship existed between the MBHC and the new affiliate prior to 75 acquisition. If not, the county was included in the research sample. De novo banks and MBHC lead banks were excluded because they are not representative of typical acquired banks. De novo banks are recently created banks. High "startup" costs and the absence of an established customer clientele make de novo banks atypical banks during their first years of operation. MBHC lead banks are large banks which are typically the initial acquisition of an MBHC. Their acquisitions represent organizational changes rather than policy changes. As stated earlier in this report, Piper found that, on average, lead banks were twenty times larger than their affiliates. They are therefore not representa— tive MBHC affiliates. Altogether 83 counties were found which experienced initial market entry by an MBHC after the beginning of 1969 and met the other selec- tion criteria. In each of these counties two banks were selected for inclusion in the sample. One bank was the new'MBHC affiliate, i.e., the first MBHC affiliate in the county. The other was an independent bank at the time of MBHC entry which remained independent through 1976. In the statistical testing these two banks will be paired with each other to check for changes. Criteria used to select these independent banks were as follows: 4 1. Bank deposits The independent bank closest in size to the MBHC affiliate was the overriding selection criterion. In this way dif- ferences between the two banks caused only by size (bank deposits) would be minimized. Unless an independent bank could be found which was no more than 50 percent larger or smaller than the affiliate, the county was omitted from the sample. At the very worst then, the affiliate 76 was paired with a competing independent bank 50 percent larger or 50 percent smaller than itself. 2. Distance between banks If more than one independent bank of the same approximate deposit size as the af— filiate existed in the county, the one selected was the one nearest the affiliate. However, in very few counties was there more than one independent bank of the same deposit size. Thus this second criterion was seldom applied. In fact, many counties that met all other selection criteria were omitted from the research sample because no single independent bank was present in the county whose deposits were within t 50 percent of the affiliate's. Finally, for each state a list of all counties which had no MBHC or one-bank holding company affiliates during the 1969-1976 period was pre- pared. An independent bank from one of these counties was then select- ed for pairing with the MBHC affiliate. The only selection criterion was deposit size. In general there were more counties without holding companies than counties experiencing initial MBHC entry. Therefore it was possible to more closely match deposit size of the competing inde- pendent-noncompeting independent bank pairs than the affiliate-compet- ing independent bank pairs. Two final points should be made regarding sample selection. First, no large metropolitan area (cities with populations exceeding 100,000) are represented in the sample. Although these areas were not explic- itly excluded, the sample selection criteria prevented any of them from being selected. In the states surveyed, initial acquisitions in large metropolitan areas either occurred prior to 1969, were acquisitions of lead banks (merely an organizational change), or were only formal rec— ognitions of a holding company relationship that existed prior to 77 acquisition. As a result, the findings of this research cannot be gen- eralized to MBHC acquisitions occurring in large metropolitan areas. The second point is that for many of the counties included in the sample, other MBHC's entered through acquisition after the initial en- try. Hence some counties contained only one MBHC affiliate during the 1969-1976 period while others contained several. These subsequent ac- quisitions should provide more pressures for change, especially among remaining independent banks which watched their competitors change from independent banks to MBHC affiliates. Because the major research in- terest is to find whether the hypothesized changes occurred because of MBHC entry, these subsequent entries will have the effect of making whatever changes might occur more evident. In summary, the research sample consists of 83 three-bank pairings: affiliate--competing independent bank--non-competing independent bank. same county different county These three-bank pairings are designated in this research as "triplets". Tables 5.1 and 5.2 on the following two pages present the sample size (number of triplets) by state for each year relative to the acquisition year and for each calendar year, respectively. SAMPLE CHARACTERISTICS AVERAGE BANK SIZE A brief discussion of selected sample characteristics is presented here. Two characteristics are described. First, the average asset size 'of the three bank groups by state is described. Then the degree of suc- cess achieved in pairing same size banks is presented. Table 5.3, which follows on page 80, gives the average size of each of TABLE 5.1 NUMBER OF TRIPLETS BY STATE FOR EACH YEAR RELATIVE TO ACQUISITION YEAR* 78 seam. Florida Iowa Michigan Missouri NeW’Mexico Ohio Texas Wisconsin Total 11 .2 46 10 19 11 _7 64 10 23 3 14 11 _§ 77 _Q 3 5 1o 25 4 16 11 _3 82 fl 3 5 9 24 4 15 10 _8 78 :2. 1 4 9 21 4 15 9 .13. 71 +3 13 14 +4 13 10 loo YEAR RELATIVE TO ACQUISITION +5 10 16 Io_ta;. 14 35 63 168 29 112 75 560 *A triplet is a paired group of three banks: Affiliate--Competing independent-~Non-Competing independent different county same county NUMBER OF TRIPLETS AVAILABLE FOR TESTING 79 TABLE 5.2 BY STATE AND CALENDAR YEAR STATE Florida Iowa Michigan Missouri New'Mexico Ohio Texas Wisconsin Totals 10 11 loo 10 14 loo 14 15 0‘ km CALENDAR YEAR 26 15 10 loo 79 10 27 16 11 10 27 14 11 °° I 1— CD 10 27 14 11 loo 10 27 13 11 '00 Total 14 35 63 168 29 112 75 S60 F] Ic 80 TABLE 5.3 AVERAGE ASSET SIZE OF BANK GROUPS DURING THE ACQUISITION YEAR AVERAGE ASSET SIZE, ($1,000's) Competing Non-Competing STATE Affiliates Independent Banks Independent Banks Florida 17,716 24,613 19,668 Iowa 19,992 20,560 19,387 Michigan 22,864 23,863 24,286 Missouri 15,086 13,599 14,455 New Mexico 26,075 26,382 24,941 Ohio 28,255 23,992 26,961 Texas 47,875 42,096 47,619 Wisconsin 18,544 20,505 18,649 81 the three bank groups in the study. The data shown represent group averages of the banks in each triplet for the year in which the affili- ate was acquired. For instance three Florida triplets exist represent- ing three Florida counties experiencing initial MBHC entry after 1/1/69. Acquisitions occurred in 1971, 1974, and 1975. The affiliate bank size shown in Table 5.3 for Florida was calculated by summing the 1971 affiliate's 1971 size with the 1974 affiliate's 1974 size and with the 1975 affiliate's 1975 size and then dividing this sum by three. The same was done for the other two Florida groups and for groups in the other states. From this table it is evident that the non-competing independent banks more closely match the affiliates in asset size than do the com- peting independent banks. The reason is that there were more non-com- peting than competing independent banks. A same size non-competing bank was therefore easier to find than a same-size competing bank. It is also evident from this table that average bank size by group varies considerably from state to state. There are several reasons for this. ‘Most important is that the calendar year of most acquisitions in some states was more recent than for other states. Hence average bank size for banks acquired in 1973 or 1974 would be larger than aver- age bank size for banks acquired in 1969 or 1970 because of the growth of average bank size as time passes. Most Wisconsin acquisitions oc- curred during the early years of the time period studied while most Michigan and Texas acquisitions occurred during the early or middle 1970's. It would be expected that average size of Michigan and Texas banks would be larger than Wisconsin banks, and they are. Another reason that average bank size varies is that the average 82 size of all commercial banks in a host state varies from state to state. In heavily industrial states commercial banks are larger than in less industrialized states. Banks in triplets from industrialized states (Michigan and Ohio) will therefore be larger on average than banks in other states (Iowa and Missouri, for example). ACCURACY OF PAIRING TECHNIQUE An indication of the accuracy achieved in matching banks by size in each triplet is presented in Table 5.4. Shown for the acquisition year are absolute size differences between the bank groups in each pair and the percentage difference. The percentage difference is defined as Percentage I Mean of absolute size differences between pairs Difference Average bank size of affiliate Average bank size of c:;peting independent bank. The denominator of this expression depends on whether the affiliate-- competing bank pair or the competing--non-competing bank pair is of in- terest. Averages of bank size and percentage difference for all states together are: Average Size Percentage ($1,090's) Difference Affiliates 24,320 Competing independent banks 22,840 6.1 Non-competing independent 23,991 5.0 banks The largest absolute differences in dollars between the bank pairs exists for the Texas triplets. However, because the Texas banks are the largest banks in the research sample, the percentage difference for 83 TABLE 5 . 4 ACCURACY OF BANK PAIRINGS BANK PAIRING AFF - CPTG CPTG - IND Mean of Absolute Mean of Absolute Differences Percentage Differences Percentage §gél§_ ($1,000's)* Difference** (§1,000's):, Difference++ Florida 7629 43 4943 20 Iowa 4129 21 3805 19 Michigan 4690 20 6194 26 Missouri 3426 23 3253 24 New Mexico 2845 11 3191 12 Ohio 5101 18 4648 19 Texas 10303 22 11472 27 Wisconsin 3893 21 3520 17 *Mean of absolute differences between banks in this pair. ean of absolute differences between banks in pairs expressed as a percent of the affiliates' average asset size. Mean of absolute differences between banks in this pair. Mean of absolute differences between banks in pairs expressed as a percent of the competing independent banks' average asset size. 84 the two groups of pairs was only a little above the average for all banks together. Largest differences in pairing banks occurred in the Florida pairs. The reason is that only three counties in Florida are included in the sample and in all of them the competing independent bank closest in deposit size to the affiliate was near the I 50 percent selection criterion. SUMMARY The research sample was selected from eight states: Florida, Iowa, Michigan, Missouri, New Mexico, Ohio, Texas, and Wisconsin. For a va- riety of reasons, counties were used as banking markets. Those counties experiencing initial market entry by an MBHC through acquisition between 1/1/69 and 12/31/76 were identified by surveying Summary of Deposit data for all FDIC insured banks. De novo banks and acquisitions of lead banks were omitted from the sample. Although 83 counties were indenti- fied that met all sample selection criteria, Missouri and Ohio contrib- uted the greatest number of counties; Florida, New Mexico, and Iowa contributed the least. Each new affiliate was paired with an independent bank located in the same county. The independent bank closest in deposit size and lo- cated nearest the affiliate was chosen. From a list of all counties in the host state containing neither one-bank nor multibank holding companies, a same size independent bank was selected. Thus for each county experiencing initial MBHC entry, three banks were indentified: affiliate, competing independent, and non-competing independent. Discussion concluded with a brief description of sample character- istics. The average asset size of each bank group by state was pre- sented. Reasons for variations in bank size by state were explained. 85 Finally, the degree of success achieved in pairing same size banks was described. Affiliates and non-competing independent banks were both slightly larger on average than competing independent banks. 86 FOOTNOTES TO CHAPTER V 1For instance, see Richard W. Stolz, "Local Banking Markets and the Relation Between Structure, Prices, and Non-Prices in Rural Areas," Dissertation for the Degree of Ph. D., Michigan State University, 1975. Stolz found that while many counties adequately represent banking markets, many do not. He therefore recommended that market definitions be based on economic and demographic factors. 2Charles D. Salley, "Concentration in Banking Markets: Regulatory Numerology or Useful Merger Guideline," Monthly Review, Federal Reserve Bank of Atlanta (November, 1972): 190-191. 3Every county in states which are located in two Federal Reserve districts was surveyed. For example, every county in Michigan (part of Michigan is in the Chicago district and part is in the Minneapolis dis- trict) was surveyed. CHAPTER VI TESTING AND RESULTS This chapter describes both the analysis of the research data and results from the statistical testing. A discussion of characteristics of the continuous variables is followed by a graphical analysis of the research data. Plots of the behavior of each dependent variable for each bank group and of the Herfindahl Index for markets with and with- out holding company affiliates are examined. The next section then covers the statistical testing and describes the comparisons made and the results of each. Signs and magnitudes of the estimated coefficients are then examined. CHARACTERISTICS OF THE RESEARCH VARIABLES Twelve variables appear in the statistical model (Equation IV-l). Of these, seven are continuous, four are dummy variables with values of 0 and 1, and one, calendar year, is an integer variable. This section begins with a description of several statistical characteristics of each of the seven continuous variables. A discussion of the correlations among all twelve variables then follows. Material presented in this section will be helpful in understanding the results that are described in later sections of the chapter. CHARACTERISTICS OF THE CONTINUOUS VARIABLES On the following page Table 6.1 presents selected results of the SPSS CONDESCRIPTIVE computer program used to compute data characteristics 87 88 TABLE 6.1 STATISTICAL CHARACTERISTICS OF mmmmmvnmmm VARIABLE MEAN MINIMUM MAXIMUM County-wide Personal Income 190. 10.0 1800. ($1,000's) Competing Deposits 36.4 0 475 ($1,000's) Bank Assets 27,263 1800 263,297 ($1,000's) R1 7.7% 4.6% 12.4% R2 5.8% 1.4% 12.5% R3 79.9% 34.8% 160.7% Herfindahl Index .284 .06 .81 STANDARD w 227.2 54.4 89 for the research sample. Since characteristics of the dummy and integer variables are of little interest or meaning, only the seven continuous variables are described. Values presented are computed from the total of all 560 triplets shown in Table 5.1. Because there are three different banks in each triplet, 1680 separate bank observations during the 1969-1976 research period are represented (560 X 3 = 1680). It is important to note that all banks in each of the three bank groups over the 1969-1976 period are included in the results of this table. No distinction was made among acquisition years, calander years, bank type, branching law, or Federal Reserve System membership status. Instead, all data was ana- lyzed as a single group to obtain the reported results. It may be noted that county-wide personal income, competing depos- its, and bank asset size all have rather large standard deviations com- pared with their respective means. This result was caused by inclusion in the sample of MBHC acquisitions in counties with very low levels of economic activity as well as counties with much higher levels. For in- stance, countydwide personal income varied from 10 million to 1.8 bil- lion dollars, the latter figure being 180 times greater than the former. Large differences in the ranges of competing deposits and bank asset size similarly exist. These large ranges caused by the great size dis- parity among markets included for study explain the large standard devi- ations that exist. On the other hand, the three dependent performance measures have relatively small standard deviations compared with their means. This would indicate that individual banks are unable to perform much dif- ferently than average. Results for R1 and R2 suggest that bankers have 90 only a narrow range in which they can vary rates charged on loans (R1) or earn interest on investments (R2). Reed et a1. state that it would be unusual for banks to make loans at rates much higher than 2.5% above their prime lending rate.1 Similarly, interest rates on mortgage loans and consumer loans for items such as automobiles tend to vary over a relatively small range. Thus a large standard deviation on the average bank lending rate would not be expected. A similar situation exists for the average return on investments. A spread of only about 3% be- tween the safest investments (T-Bills) and the riskiest (perhaps mu- nicipal securities) exists. Before-tax return on investments would not be expected to vary greatly around the average. Its relatively low standard deviation in Table 6.1 supports this expectation. Neither does the operating expenses to operating revenue ratio, R3, vary greatly for banks within the asset range studied. Evidently bank- ers are limited in their ability to achieve economies through mecha- nization, to achieve economies of scale or affiliation, or to change greatly the prices paid or received for services relative to other banks. Because prices received consist of fees and interest rates, and because the largest cost component is interest expense, it is apparent that the small variation in interest rates discussed above contributes to the uniformity among banks in operating efficiency. Table 6.1 also indicates that for the 83 counties studied from 1969 through 1976, the Herfindahl Index averaged 0.284. Consequently, a market with an index value greater than this would be more concen- trated (less competitive).and one with a lower value would be more com- petitive than average. (n Ir—I 91 CORRELATIONS AMONG THE VARIABLES Correlations among all the research variables were calculated on data which included all banks in all time periods. Results are pre- sented in Table 6.2. The discussion that follows explains the corre- lations found in this table that exceed 0.40. Calendar Year and Acquisition Year. A relatively high positive cor- relation exists between these two independent variables. This is a logical result since as time (calendar year) progressed, year relative to acquisition year increased. Calendar Year and R%. Calendar year also has a high positive cor— relation with R1, return on loans. Bankers can rapidly adjust the re- turn received on their loan portfolios in response to rising interest rates in the economy because a large proportion of bank loans have short-term maturities. As interest rates rise, bankers adjust loan rates upward. Since interest rates in the economy rose secularly during the eight years covered by this study, the relatively high positive correlation appearing in Table 6.1 was expected. County-Wide Personal Income and Both Deposits in Competinnginancial Institutions and Bank Asset Size. County-wide personal income has a relatively high positive correlation with both deposits in competing financial institutions and bank asset size. This was also expected be- cause as county income rises, banks and competing financial institu- tions can attain a larger scale of operation. Deposits at financial institutions would increase with increases in county income. As more funds become available, banks could increase dollars of loans and in- vestments. Since banks count these as assets, bank size would increase. 92 TABLE 6.2 CORRELATION MATRIX OF ALL VARIABLES IN THE STUDY L C I A D M S R1 R2 R3 H 2 Y X L 1.00 C .06 1.00 I -.08 .10 1.00 A .17 .69 .20 1.00 D -.06 .14 .87 .28 1.00 M -.07 -.06 .12 .12 .12 1.00 S —.28 .18 .48 .30 .51 .27 1.00 R1 -.14 .50 .09 .33 .09 -.01 .19 1.00 R2 -.04 .19 .01 .11 .01 -.02 .02 .15 1.00 R3 .08 .14 .22 .15 .16 .06 .14 .13 .02 1.00 H .02 .07 -.34 -.07 -.24 -.02 -.04 .10 .00 -.17 1.00 2 .00 .00 -.20 .00 -.16 -.07 .01 -.02 -.02 —.11 .42 1.00 Y .00 .00 .10 .00 .08 .05 -.02 -.01 -.03 .07 -.21 -.50 1.00 X .00 .00 .10 .00 .08 .02 .02 .03 .04 .04 -.21 -.50 -.50 1.00 L represents state branching law. C represents calendar year. I represents county personal income. A represents year relative to acquisition year. D represents countydwide deposits in competing financial institu- tions. M represents S represents R1 represents return on loans. R2 represents before-tax return on investments. R3 represents operating expenses to operating income. H represents the Herfindahl Index. 2 represents independent banks located in counties with no holding Federal Reserve System membership. bank asset size. company affiliates. Y represents banks located in the same county as holding company affiliates. X represents bank holding company affiliates. 93 Deposits at Competing Financial Institutions and Bank Asset Size. The fact that these two variables have high positive correlation sug- gests that all competing financial institutions changed size together during the eight year period investigated. Herfindahl Index and Markets with no MBHC Affiliates. The relatively high positive correlation found between these two variables suggests that market concentration is higher in markets without MBHC affiliates than in markets with affiliates. That is, there is less competition in markets that have only independent banks. Perhaps the efforts of the Federal Reserve to approve holding company acquisitions that increase competition and to deny those that do not (unless there are offsetting benefits) contribute to the difference in concentration between these two types of markets. Correlations Among the Three Bank Groups. The negative correlations among the three bank groups result from the pairing technique and re- search design. They have little research meaning. For instance, if a bank is an affiliate (X), one paired bank would be an independent com- peting bank (Y) and one would be a non-competing independent bank (Z). Because these banks were represented by dummy variables with values of 0 and l, correlations would, by design, have a value of -.50. GRAPHICAL ANALYSIS In this section the behavior of the three dependent performance measures and the performance of the Herfindahl Index for each year rel- ative to the acquisition year are examined graphically. For the three dependent variables, the graphs show performance differences for each bank group. Although the graphs do not indicate whether the differences 94 that appear are statistically significant, they do highlight relative performance differences. The graph of the Herfindahl Index versus type of market (with or without an MBHC affiliate) is included to determine whether banking concentration in markets void of MBHC affiliates is af- fected by the initial entry of an MBHC. Each of these graphs is explained in more detail in the remainder of this section. The next section will discuss whether differences among bank groups on the three dependent measures are statistically significant. DEPENDENT MEASURES Figures 6.1, 6.2, and 6.3 present the behavior of R1, R2, and R3, respectively. Each figure is described in the discussion that follows. Return on Loans,734, Figure 6.1 shows the response of R1. Return rises slowly for all three groups during the 10 years relative to ac- quisition year shown. Loan return for the group of MBHC affiliates is lower than return for either of the other two groups prior to the ac- quisition year. However, R1, becomes greater than the other groups following acquisition and stays greater throughout the next six years. No clearly identifiable differences between the competing independent and non-competing independent bank groups exist. This increase in loan return for the MBHC affiliates supports the research hypothesis stated earlier in this report that holding companies restructure the loan portfolios of their new affiliates to riskier, but higher yielding loans. Yet McLeary and Johnson and Meinster found that loan revenue to total loans fell for affiliates compared with independ- ent banks. Perhaps the sample selection criterion used in the research that limits the sample to only initial market acquisitions is respon- sible for this difference. .09 .08 .01 95 MBHC affiliates Competing independent I . ; Non-competing ~..‘ ,-— independent banks o 07‘ .06: , ”I \\ affiliates Year Relative to Acquisition Year L 1 1 1 I r T T T l -3 -2 -1 0 1 2 3 4 5 6 FIGURE 6.1: Behavior of R , Return on Bank Loans, for Bank Groups Relative to Acquisition Year. Competing independent banks \ Non-competing independent banks Year Relative to Acquisition Year -3 -2 -1 0 1 2 3 4 5 6 FIGURE 6.2: Behavior of R2, Return on Investments, for Bank Groups Relative to Acquisition Year. 96 Before-Tax Return on Investments,_R2. Figure 6.2 shows the behavior of R2 by bank group. As with R1, return rises slowly for the two inde- pendent bank groups without discernible differences between them during the ten years shown. However, unlike with R1, return on investments for the group of affiliates, after rising rapidly during the three years prior to acquisition, fell slightly during the years following acquisi— tion. Further, for the affiliates, R2 is higher than the other groups prior to acquisition and equal to or lower than the other groups after acquisition. It seems that acquisition makes a considerable difference in R2 for affiliates both relative to the other two bank groups and relative to the acquisition year. This behavior is not consistent with the research hypothesis that MBHC's shift funds of their affiliates into higher yielding securities after acquisition. A possible explanation is that affiliates may pay lower tax rates than independent banks. This might occur because op- erating expenses are relatively high for affiliates compared to inde- pendent banks (see the discussion of R3 below). If affiliates do invest in riskier tax-exempt securities, but are not in the highest income tax brackets, before-tax yields may not be as high as on safer investment securities. In this case return would fall even though affiliates did shift into riskier investments compared with independent banks. Operating Expenses to Operating Revenue, R,. The behavior of R3 J is shown in Figure 6.3. Most notable is that the two competing banks groups-€MBHC affiliates and competing independent banks--experienced higher operating expenses per dollar of operating revenue than the non- competing independent bank group. There are no major differences be- tween affiliates and competing independent banks. This finding is 97 R .95» 3 Competing independent banks .83” MBHC affiliates ...—'T r“; 3“ \ Non-competing .7 independent banks Year Relative to . ‘ Acquisition Year T I v I I l I I ' ' -3 -2 -1 0 1 2 3 4 S 6 FIGURE 6.3: Behavior of R3, Operating Expenses to Operating Revenues, for Bank Groups Relative to Acquisition Year. 98 consistent with past research and suggests that if economies of af- filiation exist, they are offset by increases in operating expenses. Affiliates mdght experience higher operating expenses arising from extended customer service hours, additional services offered, an added layer of management compared with independent banks, and so forth. Although R3 was lower for non-competing independent banks than for either of the two other groups, it was lower both prior to and after acquisition by approximately the same amount. Thus holding com- pany entry does not appear to affect either of the two competing bank groups relative to each other or relative to non-competing independent banks. In summary, graphical analysis suggests that, in response to ini- tial market entry by MBHC's, 1. MBHC affiliates achieve higher return on their loan port- folios, lower return on investments, and no significant change in operating expenses to operating revenues compared with both competing and non-competing independent banks, and 2. Competing independent banks do not behave much differently than non-competing independent banks on any of the three dependent measures 0 HERFINDAHL INDEX Figure 6.4 contrasts market concentration in counties containing MBHC affiliates with counties void of holding company affiliates. Two points are of interest. First, markets without MBHC affiliates are much more concentrated (less competitive) than markets containing an MBHC af- filiate. Perhaps this is caused by the reluctance of the Federal Reserve to approve holding company acquisitions in already concentrated markets .30 " .20 - 99 R 040 '1? Markets void of MBHC's .30 ‘ MBHC \/ markets .20 A» Year Relative to Acquisition Year -3 -2 -1 0 l 2 3 4 5 FIGURE 6.4: Changes in the Herfindahl Index Relative to Acquisition Year for Markets with and without MBHC Affiliates. 100 unless the acquisition will strengthen the acquired bank and increase market competition. In that case, MBHC's would tend to acquire banks in markets with lower Herfindahl Index values. The second notable point in Figure 6.4 is that acquisition does not appear to affect market concentration. Patterns of change in market concentration for both types of markets are similar. In summary, relative to markets void of holding company affiliates, acquisition neither increases nor decreases concentration in markets with MBHC affiliates. Because the research objective was to determine whether affiliates change relative to inde- pendent banks, a test for statistical differences between markets in the Herfindahl Index was not made. STATISTICAL TESTING The primary objective of this section is to describe the tests that were made and to explain the results. Three comparisons were tested: affiliates versus competing independent banks, affiliates versus non- competing independent banks, and competing independent banks versus non- competing independent banks. After a discussion of these three compar- isons, the multivariate and univariate test results on the dependent measures are reported for each. The necessity of multivariate instead of only univariate testing is covered next. Finally, signs and coeffi- cients of the independent and dependent variables are described. COMPARISONS In order to find whether banks in markets void of MBHC affiliates change their operating performance after the initial entry of an MBHC through acquisition, three separate comparisons were made. Each MBHC affiliate was paired with a similar size independent bank located in the 101 same market. The group of affiliates was then compared with the group of competing independent banks to test for significant difference be- tween them. This comparison will indicate whether affiliates change compared with competing independent banks. Another comparison looked at the performance of the independent competing banks versus a group of non-competing independent banks located in markets void of affiliates. If, because of market entry by an MBHC, the former group of banks change in ways similar to the MBHC affiliates, the first comparison may not reveal a difference, but the second comparison would. The final com- parison contrasts the group of MBHC affiliates with the group of non— competing independent banks. This comparison serves as a further check to determine whether similar changes occurred to both competing inde- pendent banks and MBHC affiliates. RESULTS Version seven of the SPSS computer program was used to test the research hypothesis that the beta coefficients for the independent variables representing bank type were equal to zero. Although Version seven will perform a multivariate analysis of variance (MANOVA), it will not perform a multivariate linear regression analysis. Therefore the multivariate testing was done using an MANOVA. Unfortunately, output from the MANOVA analysis does not include standard errors of the esti- mated coefficients or the constant term. To get this information on the univariate tests, linear regression was used. Both multivariate and univariate tests were made for each of the 10 acquisition years (-3 through +6) investigated. Three tests for each acquisition year were made--one for each bank group comparison described previously. There- fore a total of 120 tests were performed giving either multivariate or 102 univariate results. Discussion of the results begins with an expla- nation of statistical significance found. Next is a discussion of the need for a multivariate instead of a univariate analysis. This is fol- lowed by a discussion of estimation using raw regression coefficients. Signs of the independent variables are then briefly discussed. Finally, this section concludes with an analysis of the beta coefficients of the three dependent variables. Statistical Significance. Tables 6.3 and 6.4 on the next pages pre- sent results of the 120 tests for significant differences between bank groups. Table 6.3 shows the F-ratios for each test while Table 6.4 ex- presses these ratios as the degree of confidence that statistically sig- nificant differences between bank groups exist. The reported confidence values in Table 6.4 are 1.00 minus the probability of Type I error. High F-ratios and confidence values indicate that significant differ- ences did exist. If, at the 95 percent confidence level, a statisti- cally significant difference exists, a circle has been drawn around the F-ratios or percentages in Tables 6.3 and 6.4, respectively. All other values are not significant at this level. Both of these tables present multivariate results and then uni- variate results of tests for differences. Inspection of the reported values shows that at the 95 percent confidence level, only one of thirty multivariate tests indicate that a statistically significant difference existed: In the acquisition year a significant difference between af- filiates and competing independent banks was found. Yet at this confi- dence level one or two tests (5 percent x 30 8 1.5) would be significant by chance alone. Thus the null hypothesis cannot be rejected based on these multivariate tests. That is, the multivariate hypothesis that 103 TABLE 6 . 3 RESULTS OF F-TESTS FOR SIGNIFICANT DIFFERENCES BETWEEN GROUPS COMPARISON TESTS IND-AFF Multivariate (Z - X) Univariate :1 R2 3 AFF-CPTG Multivariate (X - Y) Univariate :1 R2 3 CPTG—IND Multivariate (Y - Z) Univariate Tabled F-Ratio at 95% Confidence Multivariate Univariate F - RATIOS -3 -2 -1 o 1 2 3 4 5 6 1.60 2.12 2.44 1.55 .56 .73 1.95 1.31 1.64 .91 2.47 .1 .8 1.07 .69 1.23 3.45 1.61 2.55 .05 49-3.06 .94 .8169 .57 3.76 .14 .99 . 2 . 0 .58 .17 .88 1. 0 .17 .61 .31 .39 .59 1.92.1.00 .78 .73 1.23 2.25 1.08 .00 1.07 1 1 .0 2.60 .23 .44 2.17 :4 .31 1.12 .43“ .17 1.65 1.85 .26® .74 .00 .04 .10 3.08 .00 .20 .00 .58 .o1 .12 1.40 1.88 1.37 .62 .41 .85 .63 .53 .26 1.44 2.35 2.94 3.32 1.10 .43 .44 .01 .23 .18 1.28 .49 2.89 .46 .43 .34 .08 1.14 .92 .00 1.31 .88 .03 .49 .72 .15 1.79 1.21 1.23 .45 .75 2.68 2.68 2.68 2.68 2.68 2.68 2.68 2.68 2.76 2.92 3.92 3.92 3.92 3.92 3.92 3.92 3.92 3.92 4.00 4.08 *The F-ratios differ from year to year because the sample size varies as year relative to acquisition year changes. exceeding the tabled values at 95% confidence are circled. F-ratios equal to or 104 TABLE 6.4 RESULTS OF TESTS FOR SIGNIFICANT DIFFERENCES BETWEEN GROUPS CONFIDENCE THAT SIGNIFICANT DIFFERENCES EXIST* Year Relative to Acquisition Year . COMPARISON TESTS -3 -2 -1 0 1 2 3 4 5 6 IND-AFF Multivariate .81 .90 .93 .80 .36 .46 .88 .73 .81 .55 (Z - X) Univariate R1 .88 . . . - .70 .59 .73 . .93 .79 .88 R2 .17 @@ .92 .67 .63 Q) .55 .94 .29 R3 .68 .27 . 2 .55 .32 .65 .32 .56 .42 AFF—CPTG Multivariate .24 .38 .87 Q .60 .49 .46 .70 .91 .63 (X - Y) Univariate R1 .02 .70 . .0 .89 .37 .49 .86 ..4 .42 R2 .71 .49 .32 .80 .82 .39 .61 R3 .04 .16 .00 .92 .02 .34 .01 .55 .09 .26 CPTI3-IND Multivariate .75 .86 .75 .40 .25 .53 .40 .33 .14 .75 (Y — Z) Univariate R1 .87 .91 .93 .71 .49 .49 .08 .37 .33 .73 R2 .51 .91 .50 .49 .44 .22 .71 .22 .04 .74 R3 .65 .13 .51 .61 .30 .82 .73 .73 .50 .61 \ * Numbers reported are 1 - probability of a Type I error. Confidence Values equal to or greater than 95% are circled. 105 there are no significant differences between bank groups cannot be re- jected. The implication is that initial market entry by MBHC's does not result in significant performance differences among banks. Only 6 of the 90 univariate tests were significant at the 95 per- cent confidence level. All of the significant univariate differences appear for the tests on R2, before-tax return on investments. However, these differences are distributed both before and after acquisition without any pattern related to MBHC entry. It therefore does not ap- pear that statistically significant differences between bank groups on R2 was caused by MBHC entry. Further, these 6 significant differences are not much greater than the 4 or 5 that would be expected from chance (5 percent x 90 = 4.5). These results are consistent with the multi- variate findings. These conclusions are not much different even at the 90 percent significance level. Only 4 multivariate tests (10 percent x 90 - 3 expected) and 13 univariate tests (10 percent x 90 = 9 expected) were significant. Nine tests were significant for return on investments, but only three occurred during the years after MBHC entry. In this entire analysis only two notable patterns occur. First, there is a lack of significant differences on either R1 or R3 for any bank group. In fact, there were no significant differences on these measures at 95 percent confidence and only 3 of 60 tests at 90 percent were significant. The other pattern is the concentration of statistical significance in R2. While 6 of 30 univariate tests on R2 were significant at the 95 percent confidence level, 9 of 30 were significant at 90 percent. Additionally, considering only the two comparisons involving MBHC affiliates, 6 of 20 and 8 of 20 tests were significant at the 95 and 90 percent levels, res pen hat by ca St fe EV 8E 106 respectively. Clearly MBHC affiliates perform differently than inde- pendent banks. Yet because the significant differences were distributed before and after acquisition, the differences do not seem to be caused by holding company entry. All other reported results in these two tables are not statisti- cally significant, even at the 90 percent level. Probably the most striking result of this testing then, is that very few significant dif- ferences between any of the bank groups caused by MBHC entry were found. Evidently the initial market entry by MBHC's does not greatly affect either the new affiliates or competing independent banks relative to each other or relative to non-competing independent banks. Need for Multivariate Analysis. One of the major objectives of this research was to determine the appropriateness of multivariate statistics in the analysis. Most research into the effect of holding company ac- quisitions has used only univariate testing. What little multivariate testing that has been done indicated that univariate tests do not give the same results as multivariate tests. In this testing the need for multivariate statistics was investi- gated by testing the null hypothesis using both multivariate and uni- variate statistics. If both types of statistics give the same results, then the use of multivariate is not necessary in determining whether the null hypothesis is to be accepted or rejected. Results indicate that there is no evidence that the univariate model distorts the findings. The estimated coefficients of the bank type variables were not significantly different from zero in either the univariate or multivariate model. Results from neither statistical technique caused the null hypothesis to be rejected. 107 Estimation. As part of the statistical testing, raw regression coefficients for each independent variable were computed. They give an indication of the effect each variable had on differences in perfor- mance between the bank groups in the three comparisons tested. Appendix A consists of ten tables of these coefficients--one table for each year relative to the acquisition year. In each table the three bank group comparisons are denoted as follows: IND - AFF represents the average performance of (Z - X) MBHC affiliates (X) subtracted from the average performance of non-competing banks (Z), AFF - CPTG represents the average performance of (X-Y) competing independent banks (Y) sub- tracted from the average performance of MBHC affiliates (X), and CPTG - IND represents the average performance of (Y - Z) non-competing independent banks (Z) sub- tracted from the average performance of competing independent banks (Y). All of the dependent performance measures used in this research are ratios of information found on bank call reports. It must therefore be cautioned that because the raw regression coefficients represent differences in performance measures which are ratios, the exact meaning of a significant difference found in a ratio is difficult to interpret. Coefficients for each independent variable and the constant term, biO’ are presented for each univariate test. Below the estimated coefficients 108 are the standard errors. Again, the corresponding coefficients for the multivariate tests are not shown because Version 7 of SPSS does not print them. Calculating the elasticity of the dependent variables with respect to the independent variables will give the relative importance of each independent variable in explaining variations in the dependent perform- ance measures. In other words, Elasticity = beta coefficient (X)/(Y), where'X’and'Y represents the mean of the independent and dependent var- iables, respectively. This type of analysis indicated that, in general, calendar year was the most important variable explaining differences in all three dependent measures. The Herfindahl Index was also important in explaining all three performance measures while state branching law was important in explaining differences in R1 and R3, but not R3. Federal Reserve System membership was the least important independent variable. Deposits at competing financial institutions was also rela- tively unimportant. Direction of Effect of Independent Variables Compared With the Hypoth- esized Effect. Appendix B consists of seven tables, one for each in- dependent variable excluding the two bank variables. Each table shows the direction of effect that an individual independent variable had on the three dependent measures--R R2, and R3--for each bank comparison 1’ tested. Signs are presented for each dependent measure by year relative to acquisition year (-3 through +6). A plus sign indicates that, as the independent variable increases, differences between the performances of 109 the two bank groups increases. An inverse relationship is denoted by a negative sign. It must be noted that the signs indicate direction of effect, but not whether the effect was statistically significant. As would be expected, movement from unit to limited branching laws increases R3 and decreases both R1 and R2 for all three bank compari- sons in Table B.1. R3 probably increases because of higher operating expenses associated with branch operation. The decline in R1, return on loans, falls perhaps because banks with branches are generally larger than banks with none. Larger banks can make larger loans. The rate of interest charged on large loans is generally smaller than the rate charged on small loans because of differences in the risk of the bor- rowers. Large banks may also make a smaller proportion of risky agri- cultural loans and more relatively safe business loans. Before-tax return on investments, R2, is also lower for banks with branches. Per- haps theSe banks require more liquidity than unit banks. Because liq- uid securities generally yield less than other investments, the greater liquidity requirements of branch banks could cause overall return on investments to fall compared with unit banks. In none of the comparisons does MBHC entry make a difference in the direction of effect. The next table, 3.2, shows the effect on the dependent measures of increasing levels of countydwide personal income. As county income rises, R3 rises. Perhaps customers demand more low profit services from banks as income rises. Such services would include credit cards, trust services, and so forth. Also, the use of checking services would in- crease as personal income increases. Most banks do not charge their customers sufficient fees to compensate for the expenses necessary to handle checks. The effect would be an increase in R3. 110 Interestingly, for all three comparisons, differences in both R1 and R2 fell as personal income rose prior to the acquisition year but rose with personal income from that point on. In other words, MBHC entry seems to have made a difference on how R1 and R2 respond to in- creases in personal income. Although this was not a statistically sig- nificant finding, Figures 6.1 and 6.2 suggest that there was a tendency for affiliates to earn higher returns on loans and lower returns on in- vestments beginning with the acquisition year. Perhaps then the hold- ing company effect dominated the effect of personal income. Beginning with the acquisition year, increases in R1 and decreases in R2 caused by holding company entry would lead to larger differences (positive sign) for the two Comparisons involving affiliates. Evidently the re- sponse of competing independent banks to MBHC entry was great enough to cause similar large differences for the comparison of the two independ— ent bank groups. The response of the dependent measures to deposits at competing financial institutions is presented next in Appendix B. Differences in R2 for the three comparisons do not appear to follow any pattern since plus and minus signs occur equally often in a nearly random pattern. While there do appear to be recognizable patterns in the behavior of R1, and R it must be remembered that competing deposits was shown in Ap- 3. pendix A to be one of the least useful variables in predicting the de— pendent measures. The patterns of signs that are present are therefore not significant and possibly were caused by some other confounding effect. Differences in R consistently fell as competing deposits rose. 3 Possibly as competing financial institutions become larger, differences 111 between bank groups on this measure decrease, causing negative coef— ficient signs. Differences might decrease because independent banks with relatively low R3 values may become more like affiliates (addi- tional services, possibly at low or negative profits) in their attempt to remain competitive. A reduction in differences between bank groups would therefore occur. R1 appears to change from positive to negative beginning with the acquisition year for the two comparisons involving affiliates. Prob- ably the tendency for affiliates to achieve higher loan returns compared with other banks explains this behavior better than changes in deposits at competing financial institutions. Table B.4 implies that differences in all three dependent measures rose as time (calendar year) progressed. A general increase in interest rates during the time span investigated would cause R1 and R2 to rise. Accelerating inflation during most of the time span would be reflected in higher R values, i.e., operating expenses increased faster than op- 3 erating revenues. Holding company entry does not affect these findings. Being a Federal Reserve member seems to lower differences in R1 and R2 as shown in Table B.5. Return on loans falls probably because most member banks are large banks which make relatively large loans. As al- ready explained, larger loans generally carry lower interest rates than smaller loans because large borrowers are safer. Return on investments would be expected to fall because, as large banks, members may need more liquidity than non-member, smaller banks. Thus relatively more invest- ments would be in low yielding but liquid securities. Also, members may hold more low yielding government securities for collateral against de- posits by governmental units. R3 increases with membership, 112 substantiating Gilbert's findings described earlier in this report that membership is costly to banks. It is also possible that the relatively large member banks offer more marginally profitable services than the relatively small non-member banks. MBHC acquisition does not appear to be related to the effect membership has on the performance differences. Table B.6 shows that, in general, performance differences between bank groups increase as bank size increases. R3 might have increased because of the tendency of large banks to have more branches and offer more marginally profitable services than smaller banks. Furthermore, researchers have found that if economies of scale exist, they are ex- hausted by the time banks attain 10 million dollars of assets.2 Thus as small banks grow, operating expenses to operating revenue stops fall- ing and possibly begins to increase. Patterns for R1 suggest that loan rates rose as bank size increased. Yet the positive signs probably re- flect rising interest rates more than increases in bank size. Previous discussion has emphasized that interest rates increased secularly during the years investigated. Further, during the ten years investigated, av- erage bank asset size increased secularly. Because a variable measuring the general level of interest rates was not included in the research model and because interest rates and bank size rose together, R1 would appear to increase because of asset increases. More than likely how- ever, rising interest rates in the economy caused the changes. Return on investments generally fell as bank size rose. Perhaps the reason cited earlier that larger banks need greater liquidity than smaller banks explains this effect. Finally, Table B.7 presents the effect of market concentration on bank performance. No pattern in the signs is identifiable for R But 2. 113 differences in R1 increase and differences in R3 decrease as market concentration increases. As banks experience less competition (in- creased concentration), loan rates and therefore differences in loan rates, measured by R , would be expected to rise. The ratio of oper- l ating expenses to operating revenue fell as fewer and fewer banks shared a market. Possibly in the absence of much competition bank officers feel little pressure to branch, extend existing services, or offer new servies. Low values of R3 would then be expected. Also, these banks would be relatively homogeneous on R3 compared to banks in very competitive markets. These latter markets could contain unit banks, both small and large, and banks with several branches. A rela- tively large difference between banks on R in these markets might oc- 3 cur. Thus the relatively low values of R3 and the relative heteroge- neity of banks in concentrated markets compared with competitive markets is a possible reason why differences between bank groups decreased as market concentration rose. In summary, seven tables were presented which showed signs of raw regression coefficients obtained from the statistical testing. In gen- eral, signs were either consistently positive or negative, with very few instances of random patterns occurring. Where patterns existed, explanations were presented. ANALYSIS OF BETA COEFFICIENTS Beta coefficients for the dependent performance measures are pre- sented in Table 6.6. These coefficients represent differences in bank performance between bank groups in each comparison after holding con- stant the effects of the 7 other independent variables. In other words, the coefficients represent performance differences between bank groups .mmmmnu .moanmauo> unopcwampcfi co>om mas «o muommmm onu ucmumuoo wafimaon noumw osu now oucmfiuowuoa anon a“ moo:muowm«p oaomouamu muaowofimmmoo wouuoaoM« lacuna :« szocm mum muouum venomoum comaumaaoo comm :« masouw xomn oau 114 Anemo.v Ammmoiv Aeeao.v Aemac.v x~mao.v Ameao.v Asmao.v Aoeao.v Aaaao.v Aaaao.v «memo. cameo. emaao. «memo. amoNo. assoc. cacao. NNaao.u cameo. cameo. me Aamoo.v Ammoo.v Ammoo.v Amoco.v Amaao.v Amaoo.v Aaaoo.v Aeaoo.v Amamo.v Ammoo.v emcee. Sacco. eeooo.u Nomoo.- aamoo.- acaoo.u aNHoo.- canoe. cameo. eeaoo.u Ne Assoc.v AoNoo.v Am~oo.v Aeaoo.v Amaoo.v Aeaoo.v Amaoo.v Amaoo.v Aaaoo.v Amaoo.v a oomoo. emooo. ~aaoo. waooo. aoaoo. eaaoo.u emaoo. sewoo. NmNOC. omNoo. e fie . we azH . uaao Amomo.v Aeamo.v Auaao.v Aeeao.v Asmao.v Aeaao.v Am~ao.v Ameao.v Aaeao.v Aeaoo.v emoHo.- ammoo. cameo.u oHooo.u waeoo.n «some. aaHNo.u eoooo.- Namoo. aaooo. mm Amnoo.v Aomoo.v AHNoo.V Aeuoo.v Amoao.v Aeaoo.v Aeaoo.v Amaoo.v Aomoo.v Awaoo.v emeoo.- eeeoo.u acaoo.u ammoo.n ommao. oeooo.- wmeoo. Nance. mNEoo. Names. Na Aamoo.v Awaoo.v ammoo.v Amaoo.v Aeaoo.v Aeaoo.v Aeaoo.v Amaoo.v Amaoo.v Aeaoo.v cameo. cameo. cameo. waooo. mecca. mmNoo. ~oooo.u aeaoo.u amaoo.u «coco. am A» . xv came 1 sea flammo.v Ammmo.v Aeaao.v Aemao.v Ammao.v Aeeao.v Aeneo.v Aoaao.v Amaeo.v Amaao.v eeaao.u cameo.u maeoo.- eaeao.u waeao.- mmeoo.- anode. 6~aao. mmaoo.- Neaao.- mm Aoaoo.v Ammoo.v Ammoo.v Aaaoc.v Amaao.v Amaoo.v Aaaoo.v Aeaoo.v a-oo.v aamoo.v N m-oo.s omaoo. Nance. ceeoo. aficio.- eeaoo. mmmoo.u Naeoo.u waeoo.- aeooo.u m Ameoo.v Aomoo.v Am~oo.v Aeaoo.v Amaoo.v Aeaoo.v Amaoo.v Amaaa.v Aeaoo.v Ameoo.v waeoo.- amNoo.- w~eoo.- eaaoo.- oe_oo.- eeaoo.- emaoo.- omaoo.u mmaoo.- emuoo.n Hm Ax . NV am< n nzH a m a m N a o a- N- m- away zomHmamzco pom» coeufimwovo< ou o>auMHmm you» «mHzmuonmmoo 49mm mmum Hzmnzmmmn mom mHZMHUHhhmou HZMQZMAMQZH mHmMH mom mHzmHonhmoo onmmmmomm 34m ~.< xHozmmm< ZOHHHmHDOU< OH onmm m QmH=H NEH 20mm HzmnzmmmnzH ZOHHHmHDOo< OH monm m azmazmamnzH onHHmHDOU< OH MOHmm m HZMQZNMflQZH oneHmHomoa mo meow one some «Ham zo omaz memoe mom maononmooo zmHmmommom 34m e.< mezoam< 134 Hm1mHom.v He1ommo.v HH1omam.v Hm1mmoe.m HH1mNNH.v HN1meHa.v HH1mNeH.m HH1mmeH.m Hmom.v N m1memm.1 m1mmaH. oNH.1 m1maNN. H1mmHN. H1oaHH. N1maam. N1mmem. H1maom.1 m He1mOHm.v Am1meam.m “N1mmHm.v Hm1mmNe.m HN1mNNH.m Hm1omNe.v HN1mHmH.m HN1maeH.m HmHm.m N m1mmom.1 m1mHHH. N1mmNH. N1mmNN. m1mNNH.1 N1oomN. N1mmNH.1 N1omHH.1 eNH.1 m He1mmaN.m Hm1mmmm.m HN1mNmm.o Hm1omam.m HN1mmmH.o Hm1mmoe.v HN1mmeH.m HN1oNeH.V HH1mmaN.m H e1onN.1 m1ommm. N1mmmm. N1mamH.1 N1mmam.1 N1omNN. N1maHH. N1ommN. mNH.1 m HN 1 mm mzH 1 mmmo HN1MHom.m Ae1mmmm.m HH1oemm.v Hm1oame.o HH1mNNH.V HN1omHe.o HH1oNeH.V HH1oemH.m Hmmm.m m m1oamm.1 m1mmaH. NNH.1 o1moeN. H1mmHN. H1omHH. N1mHmm. N1mmHN.1 H1oemm.1 m Ha1oon.v Hm1meam.m HN1mNnm.m Am1mHNe.o HN1mNNH.V Hm1mmNa.m HN1mHmH.m HN1mNmH.m HmHm.o N m1o0Nm.1 m1omHH. N1mmme. N1mHoN. m1mmmH.1 N1mNmN. N1mmoH.1 N1oNeH. eNH.1 m He1ommN.v Hm1mHmm.m HN1oHNm.V Am1mHoe.m HN1ommH.v Hm1mama.v HN1meeH.m xN1memH.v HH1mHmm.m e1meNN.1 m1moma. N1maHe. N1mmmH.1 N1ommm.1 N1mmmN. N1maHH. m1mmmH.1 mNH.1 Hm Am 1 xv mmmo 1 mm< Hm1oHom.v Ae1mmmm.m HH1ommm.v Hm1maoe.m HH1mNNH.V aN1oeHe.m HH1mNeH.m HH1mmeH.m Hmmm.m m1mmmm.1 N1omaH. HNH.1 m1oHNN. H1mmHN. H1mmNH. N1mmmm. N1mmmN. H1mmmm.1 mm Ha1mon.v Hm1omam.m HN1mmHm.m HN1oHNe.V HN1oNNH.o HmummNa.v HN1mHmH.V HN1oomH.v HmHm.m N m1mmmm.1 m1ommH. N1mHeN. N1moNN. m1oeeH.1 N1mmeN. N1omNH.1 e1HONm.1 eNH.1 m He1meaN.v Hm1ommm.v HN1onm.v Hm1mmam.m HN1mmeH.m Hm1mmme.o HN1oNeH.V HN1mNeH.V HH1oaaN.m H e1oemN.1 m1ommm. N1mNeN. N1moNH.1 N1mamm.1 N1onN. N1mNmH. N1m~aH.1 mNH.1 m Ax 1 No mm< 1 92H m H m m H o : ummwlxaam oaaomaoo mama zomHmamzoo m4m HZMQZMAMDZH ZOHHHmHDOU< MMHm< m HzmnzmmezH onHHmHDOU< Mm9h< m HzmnzmmeZH oneHmHomo< mmem< mame omHme use some HzmnzmmezH ZOHHHMHDOU< mmHm< M HZHQZMAMQZH ZOHHHmHDOU< mm9m< mmmw mahHm mmH 20mm HszZMMMQZH oneHmHmoo< mmem< m