'L “MERGERS :FI‘RM'SWiN mnusmm ‘0}? 0 \ 0!:st CS ARMTIFEREST!~ WA cI‘ATE ANA Th-esi for the Degree 9‘ P“ A MUL HMNCiAL’CH gl-uon ”w’a-u u Ive-v STATE UNIVERSITY - ' mam m 1972‘ Iii-2H“ t“ t DONALD LSTEVENS’ IlfllllmlIll'll‘llllflfllfllflllllflllllllll 5L 3 1293 01009 0227 This is to certify that the thesis entitled A MULTIVARIATE ANALYSIS OF FINANCIAL CHARACTERISTICS OF ACQUIRED FIRMS IN INDUSTRIAL MERGERS presented by Donald L. Stevens has been accepted towards fulfillment of the requirements for Ph.D Accounting & Financial Adm. Jegree in Major professor ‘4 Date June 239 1972 0-7639 l g ,4} ~ > I 1 ABSTRACT A MULTIVARIATE ANALYSIS OF FINANCIAL CHARACTERISTICS OF ACQUIRED FIRMS IN INDUSTRIAL MERGERS by Donald L. Stevens The growth of one firm by the acquisition of another firm has seen increasing importance in the last decade. Although this external investment decision is not unlike the widely treated internal investment decision in the capi- tal budgeting model, little attention has been paid to similarities between these two decisions to expand. Most of the current research in the area of mergers has been directed at post-merger performance of merger— active firms, effects of mergers upon competition, and determinants of exchange ratios in mergers. This study is directed at the acquired firm and analyzes the financial characteristics of acquired firms for the two year period immediately prior to acquisition. Hypotheses are formu- lated as to systematic differences between acquired firms and similar firms which were not acquired. The differences analyzed are financial in nature and relate to the finan- cial variables which are also relevant in the traditional capital budgeting framework. Donald L. Stevens Two samples of industrial firms were taken, each of size 40. The first contained firms acquired during 1966, and the second contained firms available in 1966 but not acquired as of 1970. For each of the firms, financial statement data was collected and a series of ratios were calculated to measure financial variables such as liquidity, profitability, leverage, and activity. In that a great deal of multi-collinearity existed in the data, factor analysis was employed to reduce the ratio set to less than ten potentially useful variables which summarized the total data set and were mutually uncorrelated. These variables were used as inputs for multiple discrimi- nant analysis (MDA). The MDA determined which set of vari- ables best discriminated between the two samples, the rela— tive contribution of each variable to total discrimination, and the significance of the differences between the groups on all of the employed variables. The results of the initial MDA and the classification accuracy of the model were significant and indicated that acquired and non—acquired firms were different with respect to these financial variables. However, the evaluation of the assumptions of the MDA model resulted in a reformulation of the samples using an "attractiveness for acquisition" Donald L. Stevens criterion rather than the acquired and non-acquired group— ings. Factor analysis was employed again to re—align the firms into those mutually exclusive sample groups. MDA was then used and results were very significant at all stages of the analysis. For both models the attractive-acquired firm samples were more profitable, more liquid, had higher asset turnover, and much lower levels of debt. For the natural groups model, dividend payout also entered the dis- criminant function and attractive firms had higher dividend payout. The discriminant model was validated by accurately classifying firms not used in the deve10pment of the model. These findings were consistent with the belief that a finan- cial basis was common to the merger decision similar to that of other asset acquisition decisions. This study also demonstrated the usefulness of a multivariate framework in financial analysis. Empirical research of this kind often faces the problem of multi— collinearity and factor analysis was shown to be a useful device to summarize the total data set without significant loss of information, such that the remaining variables mini- mized the inter-correlation problem. MDA proved to be a useful technique for detecting group differences and indi- cating which variables best distinguished the groups. This Donald L. Stevens application of these tools should provide incentive for 'their future application in research in finance. A MULTIVARIATE ANALYSIS OF FINANCIAL CHARACTERISTICS OF ACQUIRED FIRMS IN INDUSTRIAL MERGERS by Donald L. Stevens A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Accounting and Financial Administration 1972 «ed COpyright by DONALD L. STEVENS 1972 ACKNOWLEDGMENTS I wish to express my sincere thanks to the members of my thesis committee-—Professors Myles Delano, Richard Gonzalez and Hayden Howard--for guidance and direction in this research. In particular, the assistance and encourage- ment provided by Professor Delano during the initial stages of the research was very helpful, not only in limiting and defining the tOpiC, but in providing a basis for future research. During my extended residence at Michigan State Univer- sity, the assistance provided by Professor James Don Edwards was most generous and I along with all of my departmental graduate student colleagues, acknowledge his major contri- bution to our educations. In addition to the financial assistance that was always provided, his foresight with respect to the need for quantitative tools immeasurably im- proved the quality of my graduate program. Much of the research was completed at the University of Illinois and the programming assistance of Michael McLaughlin of the Commerce Computer Laboratory is gratefully acknowledged. Also, Joe Kolman of the SOUPAC staff in the Department of Computer Science provided valuable insight into the multivariate analysis and that phase of the ii programming. Sincere thanks can only partially acknowledge the significant contribution of Professor Jagdish Sheth to this research. No obligation of any kind can explain his generous gift of many hours of consultation which was directly respon— sible for the multivariate portion of this research. His influence and contribution is apparent to anyone familiar with his work. However, where errors remain, they exist in spite of his contribution and are the fault only of the author. Finally, sincere thanks are due to many individuals who have influenced much more than this dissertation. First, to my parents and brothers, for their loyalty and love. Then, to all the fellow students and teachers who have shaped my educational experience. And, most importantly to my wife, Susan, and to Jeffrey, for accepting more than their share of the load while I completed the program. They have provided joy and a sense of humor at every step of the way. iii TABLE OF CONTENTS ACKNOWLEDGMENTS. . . . . . . . . . . . . . LIST OF TABLES . . . . . . . . . . . . . . Chapter I II III IV INTRODUCTION. . . . . . . . . . . Introduction. . . . . . . . . . Purpose of the Dissertation . . Outline of the Dissertation . . MERGER RESEARCH AND FINANCIAL CHARACTERISTICS OF FIRMS . . . . Introduction. . . . . . . . . . Directions of Merger Research . Mergers and Financial Characteristics Summary and Statement of Research Questions . . . . . . Subsequent Analysis . . . . . . SAMPLE DESIGN, DATA ANALYSIS AND FACTOR ANALYSIS. . . . . . . . . Introduction. . . . . . . . . . Data. . . . . . . . . . . . . . Ratio Analysis. . . . . . . . . Factor Analysis as a Multi- variate Tool . . . . . . . . . Factor Analysis in this Study . Application of Factor Analysis. Principal Components Analysis . Analysis of Merger Data . . . . MULTIPLE DISCRIMINANT ANALYSIS. . Introduction. . . . . . . . . . Multiple Discriminant Analysis. The Multiple Discriminant Model Two Groups . . . . . . . . . . Classification Procedure. . . . MDA Model . . . . . . . . . . . Group Differences . . . . . . . iv Page ii vi |—‘ 21 26 28 28 29 35 37 40 45 46 47 60 60 61 64 67 68 72 Chapter V Appendix I The Natural-Groups Assumption of MDA . . . . . . . . . . . Factor Analysis of Firms. . . Discriminant Analysis with Natural Groups . . . . . . . Validation of Results . . . . Summary . . . . . . . . . . . Introduction. . . . . . . . . Financial Characteristics . . Multivariate Framework. . . . Factor Analysis for Data Simplification . . . . . . . MDA & Re-evaluation with Factor Analysis . . . . . . . . . . Assumptions and Limitations . Application and Suggestions for Future Research. . . . . . . Factor Analysis . . . . . . . . BIBLIOGRAPHY . . . . . . . . . . . . . . CONCLUSIONS, EVALUATION, LIMITATIONS AND SUGGESTIONS FOR FUTURE RESEARCH. Page 77 80 86 98 100 101 101 101 105 107 110 112 115 119 125 3.4 3.8 LIST OF TABLES Summary of Financial Ratios Employed. . . Means and Standard Deviations for Twenty Ratios in Three Firm Groupings . . Correlation Matrix for Twenty Ratios on Eighty Firms . . . . . . . . . . . . . Summary of Factor Analysis: Twenty Ratios on Eighty Firms. . . . . . . . . . Summary of Factor Analysis: Orthogonally Rotated Factor Matrix, Varimax Rotation . Orthogonally Rotated Factor Matrix, Varimax Rotation. . . . . . . . . . . . . Varimax Rotation into Ten Space: Summary of Factors and Rotated Factor Matrix. . . Varimax Rotation into Six Space: Summary of Factors and Rotated Factor Matrix. . . Univariate Tests of Significance for Group Means: Observed Groups . . . . . . MDA: Group Differences, Observed Groups. MDA: Scaled Vectors and Discriminant Function, Observed Groups . . . . . . . . MDA: Classification Matrix, Observed Groups . . . . . . . . . . . . . Factor Analysis: Summary of Factor Matrix, First Twenty-one Roots. . . . . . Orthogonally Rotated Factor Matrix, Varimax Rotation: Eighty Firms into 2-Space o o o o o o o o o o o o o o 0 New Classification of Observed Groups . . vi Page 38 48 49 50 52 53 56 58 73 73 74 76 82 84 86 Table Univariate Tests of Significance for Group Means: Natural Groups. . . . . MDA: Group Differences, Natural Groups. . . . . . . . . . . . MDA: Scaled Vectors and Discriminant Function, Natural Groups. . . . . . . MDA: Classification Matrix, Natural Groups. . . . . . . . . . . . Significance for Group Means for Natural Groups and Reformulated Model MDA Group Differences, Reformulated Model. . . . . . . . . . MDA: Scaled Vectors and Discriminant Function, Reformulated Model. . . . . MDA: Classification Matrix, Reformulated Model. . . . . . . . . . Individual Firm Discriminant Scores . MDA: Classification Matrix from Validation of Model . . . . . . . . . Correlation Matrix for Variables in Reformulated Model. . . . . . . . . . vii Page 88 89 90 91 94 95 96 96 97 99 109 CHAPTER I INTRODUCTION A firm seeking to expand has two avenues for growth. The first is internal growth which involves the acquisition of economic resources which are transformed into goods and services within the existing firm. The second is external growth by the acquisition of the assets of existing entities by means of merger. Although both alternatives should lead to the same ultimate goal—-maximization of the wealth of the firm's owners—-the two methods have been treated quite differently by both the business and regulatory communities, and by academicians in economics and finance. Specifically, the area of capital budgeting decision making has been directed at evolving a theory which sets forth the principles for evaluating alternative asset acquisition decisions. Numer- ous financial criteria are develOped for inputs into models designed to measure risk/return relationships and lead to a decision. While both types of growth, external and internal, would intuitively be included in a capital budgeting frame— work, the direction and emphasis in capital budgeting has 1 2 been only with the internal growth of a firm and the invest- ment decisions surrounding that problem. The acquisition of assets may be for any number of purposes within the firm. Marketing, production, and physi- cal distribution projects all compete for limited funds available to the firm at any point in time. The competition of the many projects is evaluated using the capital budget- ing framework, and based on a measure of risk and flows of benefits from the projects. Thus, independent of the specific qualitative benefit of a proposal, such as a more efficient production assembly process versus a faster inven- tory control network, the capital budgeting framework trans— forms the benefits so that they can be evaluated in terms of expected cash benefits from the project. Therefore, projects don't compete on qualitative measures such as "more efficient" and "faster", but rather on the increment of wealth gained from their acceptance. The wealth incre- ment is measured by the present value of expected cash benefits from a project. However, if one were to review the existing merger literature in finance and economics, it would appear that other motives and reasons dominate the decision to grow externally through merger. Little attention has been given to a search for systematic similarities which may 3 exist to subject the external growth decision to the same capital budgeting framework as an internal growth decision. No doubt, the going concern has many additional dimensions which need identifying as part of the investment decision process, but the premise remains that all investment deci— sions implemented with the traditional capital budgeting framework, evaluate financial variables in terms of present values and risk. The merger decision may add new criteria to this framework, or attach different weights to the same criteria, but the merger decision remains an investment decision which can be accepted only if the result is an in- crease in the value of the firm to it's owners. PURPOSE OF THE DISSERTATION The research investigates two areas. First, the relationship between financial profiles of industrial firms and likelihood of acquisition, and the applicability and usefulness of financial ratios and ratios analysis to the measurement of performance and development of financial profiles. Second, the statistical tools of factor analysis and multiple discriminant analysis (MDA) will be employed to financial analysis as analytical substitutes for tradi- tional, judgemental methods. The investigation of the first question will pro— ceed with the hOpe that extensions to the current state of 4 knowledge will accrue in that merger research has not taken this particular approach at this point in time. Although much interest exists as to why firms merge and the economic significance and implication of mergers, the great share of empirical testing has been conducted with respect to either the merger transaction itself of with post-merger perform— ance. The second area of investigation, multivariate analy- sis techniques as tools in this context, will extend and serve as a replication of other successful research using these methods in finance and related disciplines. OUTLINE OF THE DISSERTATION Chapter II will evaluate the direction taken in merger research in recent years. This includes the views of the business community, the direction of regulatory studies, and the emphasis in academic research. From this body of research the specific efforts relating financial criteria to any phase of the merger decision will be dis- cussed. This will be used to formulate, where possible, hypotheses pertaining to a priori belief about financial qualities and their relation to acquisition. Where there is no consistent belief, both sides of the issue will be discussed. Chapter III reports the research design by which the hypotheses are examined and tested. The nature of the 5 firm samples and the type of data collected are described. The tool of factor analysis is introduced for the purpose of simplifying the large set of original data. This re— duced set provides the basis for the actual testing for differences between samples which is performed in Chapter IV. Chapter IV uses multiple discriminant analysis to analyze if financial differences exist between acquired and non-acquired firms. Further, it determines, in the case of differences, which financial variables are most different between the groups. Assumptions of the MDA are examined and the reformulation of the samples leads to further in- sight about financial characteristics and attractiveness for merger. Chapter V reviews the results of the tests with respect to the hypotheses of Chapter II, and draws conclu— sions about differences between the samples. Further, the usefulness of the multivariate techniques is evaluated and limitations of the study are discussed. Finally, sugges- tions for future inquiry are made for using these findings in the area of mergers and for the future applications of the multivariate tools in financial research. CHAPTER II MERGER RESEARCH AND FINANCIAL CHARACTERISTICS OF FIRMS INTRODUCTION This chapter reviews the areas of investigation in merger research which relate both to the current merger movement and to this research. Merger research has taken many directions in the last decade within the fields of economics and finance. The existing literature in these areas contains numerous attempts to explain various facets of the merger movement as well as effects of mergers. There may be as many reasons for the acquisition of one firm by another, as there have beenmergers. No doubt, there are unique reasons in every merger which con- tribute to the decision to acquire or to be acquired. In most cases, however, while the relative importance of these unique reasons in a single merger may be noteworthy, they are not likely to be significant to the merger movement as a whole. However, if reasons and goals do exist which are common to most mergers, learning these goals and reasons would contribute a great deal to the understanding of the 6 merger movement. DIRECTIONS OF MERGER RESEARCH There have been a number of attempts to analyze the corporate merger and what role the merger plays in the growth of a firm. Two of these, quite different in sc0pe, were the Federal Trade Commission's Economic Report on 1 Corporate Mergers, published in 1969, and The Corporate Merger, by Alberts and Segall.2 The latter contains a series of papers presented at a seminar in 1963 by both professional and academic participants, all experienced and knowledgeable in various aspects of mergers. The emphasis of the FTC study was a concern about increasing levels of economic concentration in a relatively few large corporations, and an evaluation of the competitive consequences of the increased concentration. The merger study was in that context and the role of the merger in increasing economic concentration was emphasized. One section of the FTC study discussed the factors encouraging the current merger movement. A number of 1Federal Trade Commission, Economic Report on Cor- ,porate Mergers, Hearings on Antitrust and MonOpoly, Committee of the Judiciary, U.S. Senate, 9lst Congress, lst session: Part 8A, USGPP, 1969. 2Alberts, W. W. and Segall, J. E. The Corporate Merger, Chicago: University of Chicago Press (1966). 8 incentives to grow through merger were identified including market power, efficiencies of various types, economies of scale, and specific tax and accounting advantages. A con- clusion of this section of the study points out the FTC concern. The balance of evidence so far available lends little support to the view that the current merger movement reflects, in substantial mea- sure, efforts to exploit Opportunities to im- prove efficiency in resource allocation. On the contrary, there are abundant indications that certain institutional arrangements involv- ing tax and accounting methods, aided by specu- lative developments in the stock market, have played a major role in fueling the current merger movement. In this context there is little reason to expect significant social benefits to flow from a continuation of current trends. Further reasons for concluding that current merger activity, particularly insofar as leading firms are involved, represents a divergence between private and social benefits are develOped in subsequent chapters. Resultant recommendations centered around more vigorous enforcement of anti-trust statutes, more complete disclo- sure of financial Operations of major corporations, and tax reform to eliminate tax incentives for merger. It would be easy to conclude from the FTC report that the merger movement was dominated by a few giant firms devouring the most profitable and efficient of the competi- tors in numerous industries. 1FTC, op. cit., 159. 9 While the FTC report points out the concern of that regulatory body of the role of the merger in increasing concentration of assets into the control of a few large companies, these mergers do not dominate the merger move- ment. The same FTC report estimated that over 15,400 mergers were consummated during the period 1948-1968, and only 1,276 of these involved acquired firms with assets in excess of $10 million at the time of acquisition.1 The question remains as to what motivations cause firms to seek out other firms for acquisition, when the parties to the mergers are not giants in their respective industries. The discussion and analysis of numerous aspects of the merger decision was the direction taken in The Corporate Merger. Advantages and dis-advantages of external growth by merger and internal growth were analyzed. Sources of gain through merger were discussed and each phase of the merger transaction received attention. In a concluding and summariZing paper, Alberts attempted to assimilate the thoughts of the other partici- pants. Merger gains could be expected, he concluded, only in the presence of synergy and Operating economies without relaxing the assumptions of efficient pricing and valuation 1FTC, Op. cit., Table 1—1, 43: and Appendix Table 1-3, 667. 10 of the respective firms in the merger. However, if the pos- sibility of "bargains" prices is introduced, additional sources of gain arise. The measure of the importance of potential bargains is that they should be relatively common, be detectable by merger active firms, and be exploitable. After examining numerous potential bargain sources and measuring these against the above criteria, Alberts reduced the bargain set to the following group: 1) tax bargains: 2) mis-management bargains: 3) cost-of—capital bargains; and, 4) forecast bargains. Tax bargains provided Opportunity due to tax statutes which allowed gains upon merger in certain circumstances. This was also noted by the FTC study. Mis-management bar- gains existed in those firms which were Operated at less than maximum efficiency, and upon acquisition, the improved management could increase the value of the combined firm. The cost-of—capital bargain existed where the market systematically applied different capitalization rates to certain classes of firms. Thus, upon acquisition, the re— alignment of the capitalization rate would increase total combined value. Forecast bargains occurred when the acquisi— tion was made of a firm whose owners had been overly bearish in their estimates of the future performance of the firm 11 and had undervalued the stock. The mis-management bargains notion has been directly discussed by some authors and implied by others indirectly. Manne1 argued that mis-management bargains were common and many mergers were made to exploit them. Additional explora- tion of the mis-management bargain idea will be developed later in this chapter. The cost-of—capital bargain is dependent upon dis- crepancies in the market valuation mechanism. If small fimms are systematically capitalized at a higher rate than are large firms, large firms could acquire small firms generally and achieve higher total value of the combined firm. In addition, for firms who are small and have infre- quent access to the capital markets and do not have active secondary markets for their common stock, the lack of "marketability" would give ample Opportunity for cost—Of- capital bargains. The forecast bargain, noted by Alberts, would relate to this situation also. If the management of a firm and/or the market, is inaccurate in its forecast of the future performance of a firm, then another firm that could detect this and accurately revalue the firm, could lManne, Henry, G. "Mergers and the Market for Corporate Control," JOurnal of Political Economy, LXIII (April, 1965), 110-120. 12 realize a gain. The existance of gain in the case of market valua- tion discrepancies was explored by Gort.l He believed that economic disturbances such as new growth and technological change, resulted in basic changes in the firm causing valu- ation problems. The past for the changed firm is no longer a good indication of the future and the result would be greater dispersion in the market valuation of the firm. The discrepancies allowed for gains by the acquisition of those firms. While this would, no doubt, account for some of the mergers in the current merger movement, it is not at all likely that the movement could be generally accounted for by this phenomenon. Lewellen Observed that "valuation errors of the scale and frequency required to explain much of the level of conglomerate activity in the 1960's would connote a degree of market imperfection, or a pattern of investor perversity, that most investigators nowadays would be unwilling to grant."2 Such is the case with most of the individual explanations of why firms merge. Other authors 1Gort, Michael. "An Economic Disturbance Theory of Mergers," The Quarterly Journal of Economics, LXXXIII (November, 1969) 624-642. Lewellen, Wilbur G. ”A Pure Financial Rationale for the Conglomerate Merger," The Journal of Finance, XXVI (May, 1971) 521-537. 13 have also enumerated the same items in different form. Joel Dean listed the same basic traits as characteristics which make a company a tempting target for takeover.2 MERGERS AND FINANCIAL CHARACTERISTICS While the explanation of why firms merge in par- ticular situations are numerous, and many kinds of explana— tions are offered for incentives to merge, little has been achieved in reducing these to one or a few criteria which would hold for the majority of mergers. To this end, an obvious point of departure would be the standard theory of the firm framework. If one assumes firms make decisions to maximize the wealth Of their owners, merger decisions must compete with other investment decisions before the firm. These would include at any point in time, both internal investment alternatives and other merger alternatives. When a merger path is chosen, the benefits from the merger must have been recognized as being superior to the other alternatives. In the capital budgeting framework, the future expected benefits from the acquisitions (whether lConglomerate Mergers and Acqgisitions: «Opinions and Analysis, 44 St. Johns Law Review (Special Edition, 1970L Dean, Joel S. "Causes and Consequences of Growth by Conglomerate Merger: An Introduction," Conglomerate Mergers and Acquisitions: Opinion and Analysis, 44 St. John's Law Review, (Special Edition, 1970). 14 they are the result of Operating efficiencies, marketing advantages, or some other reason) as estimated by future cash flows, must have greater expected present value. If this is the case, the investigation of the characteristics of the acquired firms themselves and the acceptance of the capital budgeting framework for decision making would allow a technique whereby the possible gains from merger might be reflected. The following section reviews additional research with the goal of relating mergers to this criterion of potential gains from a merger being reflected in the finan— cial flows and criteria of the acquired firm. This develOp- ment will be the primary emphasis of the dissertation. The existing research will be used to develOp a set of criteria of financial characteristics of acquired firms which either directly measure potential gain from merger, or indirectly reflect gains. The research will be discussed in a section on merger transaction studies and a section on financial criterion studies. Merger transaction studies have analyzed the rela- tionship between financial variables of the parties to the merger and the exchange ratio in stock exchanges. Also, the use of various other financial instruments has been studied with respect to the post-merger earnings effects 15 from different financing methods. Dellenbarger used multiple regression techniques to study the determinants of exchange ratios in industrial mergers.1 He concluded that earnings per share, cash divi- dends per share, book value per.share, and market price per share of the firm's stock were the relevant variables be- tween the firms in the determination of the actual exchange ratio. Another type of merger transaction study was that of Pinches, who studied the effect of convertible preferred stock as a financing instrument in mergers.2 He tested to see if this instrument altered the post—merger earnings of the combined enterprise, and resulted in higher earnings than if other financing methods had been used. He was un- able to reject a null hypothesis although differences were observed in the direction in which he would have predicted. However, he did note in separate testing, that market prices and earnings Of merging firms were the dominant factors in exchange ratios. lDellenbarger, Lynn, E. Jr. Common Stock Valuation in Industrial Mergers, Gainesville: University of Florida Press, 1966. 2 . . Pinches, George E. "F1nanc1ng Corporate Mergers and Acquisitions with Convertible Preferred Stock," Unpublished Ph.D. Dissertation, Michigan State University, 1968. 16 Previous studies by Bosland,1 Weston and Brigham,2 and Reilly,3 also evaluated determinants of exchange ratios with similar variables taking on importance in the respec- tive works. Weston and Brigham did, however, emphasize that unique circumstances of individual mergers often dominated other variables. The research question raised by these studies is as follows: If financial qualities such as profitability, dividend policy and price/earnings ratios are the dominant decision variables in the merger transaction, do they serve a similar role in the search for acquisition candidates? To the extent that they do, certain types of firms should emerge as favored over other firms due to their profiles on these variables. The financial criterion studies have been attempts to relate merger motives generally to a financial bases. 1Bosland, Chelcie. "Stock Valuation in Recent Mergers: A Study of Appraisal Factors," Trusts and Estates (June, July and August, 1955) 516—669. 2Weston, J. Fred, and Brigham, Eugene F. Managerial Finance, 2nd Edition, New York: Holt,Rinehart & Winston, Inc. (1966) 661—682. 3Reilly, Frank K. "What Determines the Ratio of Exchange in Corporate Mergers," Financial Analysts Journal, l7 Lintner1 analyzed the post—1950 merger movement and con- trasted its characteristics with the merger waves of earlier periods. He identified Opportunities for gain by merger due to l) reductions in lenders' risk of bankruptcy losses; 2) scale diseconomies in credit investigation of smaller firms, in floatation expenses in public issues and in in— vestor information (marketability): and, 3) changes in investors' assessments of future prospects per share when price/earnings ratios differ. With respect to the leverage gains Lintner notes that even if the merging firms have Optimal capital struc- ture, two sources of gain exist through leverage. First, borrowing costs decline as firm size increases and, second, reduced lenders' risk increases debt capacity. Because the earnings streams of the merging firms are not generally perfectly correlated, the joint probability of failure in any period is less than the sum of the separate probabilities for the firms independently. Levy and Sarnat also noted this and further pointed out that limits exist to the diversification effect, and, to the extent that firms become very large and well Lintner, John. "Expectations, Mergers and Equili- brium in Purely Competitive Securities Markets," American Economic Review, LXI, No. 2 (May 1971) 109—114. l8 diversified, financial risk may become insignificant. Lewellen addressed this same question of "whether it is possible for firms, by merging, to produce gains for their stockholders in the absence of any Opportunities for Operating efficiencies--to create increments in wealth out of 'pure' financial combination of enterprises."2 If the possibility exists then when other sources of gain also are present due to increased efficiency and better management, the merger will be that much more desirable to the firm's shareholders. Lewellen's argument for the purely financial ration— ale for merger is that the increased size of the combined enterprise and the lower joint probability of failure in— creases debt capacity and lowers lenders' risk. This accrues independent of any increased debt capacity of the individual firms. Thus, market value increases in the com— bined firm are obtainable even in the absence of market imperfection. These authors jointly establish financial rationale for mergers and the increased debt capacity, lower lender lLevy, Hiam and Sarnat, Marshall. "Diversification, Portfolio Analysis and the Uneasy Case for Conglomerate Mergers," The JOurnal of Finance, XXV, No. 4 (Sept. 1970), p. 801. 2Lewellen, Op. cit., 522. 19 risk benefits of mergers, are Operative even in the case of Optimal capital structure of the merging firms. Other possibilities can be generalized from their contributions in the case of less than Optimal capital structures. First, in the case of an acquiring firm with excess debt, the mer- ger would lower financial risk and move the new firm in the direction of an Optimal capital structure. Second, in the case of the acquired firm with little or no debt in its capital structure, considerable debt capacity is gained. First the unused portion of debt capacity of the acquired firm independently is available, plus the additional capacity from the combination of the firms. Therefore, debt capacity and capital structure should be an important financial variable when searching for merger targets. Even if it is accepted that gains accrue with Optimal capital structures, to the extent that acquired firms tend to be conservative employers of finan- cial leverage, they should be more attractive. A similar line of argument can be advanced for the case of firms with less than optimal levels of current assets. Gains from acquisition of firms of this type could arise independent of earnings expectations. The acquiring firm, by reducing the excess levels of the overly liquid acquired firm, would gain a one—time release of funds 20 which could be subsequently reinvested. This type of gain would be the result of a mis—management bargain of the type described by Manne.1 The cash benefits from this process would be immediate and relatively risk free, compared to cash benefits from typical capital budgeting projects of the firm. The present value of these funds would be their capitalized value, using the firm's cost of capital as the capitalization rate. The concept of excess liquidity and attractiveness is much more apparent in the business community, than the debt capacity arguments. A number of sources are available which give collective insight into how management of merger active firms searches and screens acquisition candidates.2 Interesting insight into the acquisition process was pro— vided by Hutchinson,3 who edited a series of presentations lManne, op. cit., 110-120. 2Hutchinson, G. Scott (ed.) The Business of Acquisitions and Merggrs. New York: Presidents Publishing House, Inc., 1968. Mace, Myles L., and Montgomery, George G., Jr. Management Problems of Corporate Acquisitions. Boston: Graduate School of Business Administration, Harvard University, 1962. McCarthy, George D. Acquisitions and Mergers. New York: Roland Press, 1963. Short, Robert A. Business Mergers: How and When to Transact Them. Englewood Cliffs: Prentice-Hall, 1967. 3Hutchinson, Op. Cit. 21 by executives active in acquisition programs. Most firms used similar criteria in searching for acquisition candi- dates. Important financial criteria included expectations of earnings, growth, and rate of return on assets. Gen— erally, these were established with reference to the per— formance of the acquiring firm itself. The role of liquidity was consistently stressed as an important criterion in the eyes of the acquisition executive. "In my Opinion, the most important thing to do in reflecting on an acquisition is to study the history in terms of cash flows."1 And, by another executive: "...the determining factor in the success of all subsequent deals has been the effective nonuse of cash."2 In fact, liquidity oriented methods such as payback and cash recovery period were Often used to evaluate the proposed acquisition. SUMMARY AND STATEMENT OF RESEARCH QUESTIONS All of the preceding studies suggest a diversity in belief and illustrate the lack of a consistent and cohesive collection of thoughts for testing with empirical data. There were, of course, conclusions drawn from all of the studies, but no attempt to assimilate. The process of 1Hutchinson, Op. cit., p. 181. 2Hutchinson, Op. cit., p. 232. 22 assimilation raised questions as to if there existed systematic differences between firms chosen for acquisition and those discarded. In a capital budgeting sense, an acquisition will take place if the present value of the expected benefits is greater than the present value of the expected outlays. Thus, even if the primary motive for an acquisition does not appear to be financial, it must ultimately be weighed for its financial impact upon the stockholders' wealth in the combined enterprise. The success of all investment deci— sions must be measured for the effect upon the value of the firm. However, the existing empirical research, with empha- sis upon the performance of merger active firms, and post- merger performance has not been able to deal with pre— merger conditions which would have motivated the merger over some other alternative. Certainly each merger deci- sion was weighed against alternatives which would include the acquired firm against other potential firms which were acquisition candidates at that point in time. An important question not answered in the current literature is what qualities existed in the acquired firm which resulted in their selection over other similar firms? The actual target firms should exhibit measureable 23 differences in order to attract acquisition minded firms. If these are directly measureable financial qualities or even indirectly measureable qualities which are reflected in a firm's financial posture, then an analytical framework may be possible to attempt to isolate the differences. With this framework, it may be possible to discover the differences and make judgments or formulate new hypoth- eses pertaining to why firms merge. It may be that differ— ences between firms which could lead to merger of one and not the other are differences unique only to the two firms involved and could not be detected by analysis not recog- nizing the unique conditions. No doubt, mergers do occur for reasons unique to the merging parties but, the frequency of transactions of this type should be small. In the section which follows, an attempt will be made to assimilate both rationale for merger and the finan- cial qualities which may possibly reflect conditions attracting merger initiatives. These financial qualities will be used as the primary independent variables to test for differences between firms acquired and firms not acquired at a point in time. The implied null hypotheses which will be tested are that, for each of the financial qualities or variables, no differences exist between acquired and non-acquired firms. The rejection of these 24 null hypotheses would allow conclusions to be drawn as to the specific differences between these firms, and would be the basis for the formulation of new hypotheses concerning the types of variables which are most systematically differ— ent in firms which are acquired. This would allow definite insights as to which of the motives for merger enumerated in the literature are actually found in practice. Each of the financial qualities or variables will be stated and discussed as to how acquired firms would be expected to differ from non-acquired firms on this variable. LIQUIDITY: To the extent that mis—management bargains occur, as discussed by both Alberts and Manne, acquired firms should be more liquid than non—acquired firms. Liquid assets are easily transferred to the acquiring firm which may need liquidity for current Operations and/or additional acquisitions. Excessively high levels Of liquid assets indicate, generally, inefficient management of assets or lack of investment Opportunities. FINANCIAL LEVERAGE: Acquired firms should have lower levels of debt than their non-acquired counterparts. This adds to the attractiveness of a firm in that not only is an incre— ment of debt capacity added from the increased size of the combined firms, but also the unused debt capacity of the acquired firm will further add to the borrowing capability 25 of the merged firm. Additionally, the conservative manage- ment which allows high levels of liquidity and other ineffi- ciencies to occur, would, predictably, not be aggressive users of financial leverage. ACTIVITY & TURNOVER: This is another measure of efficiency type qualities in a firm in that the more efficient a firm's management operates, the greater the turnover of assets. If firms were conservative and mis-managed levels of liquidity and leverage, turnover would be expected to be lower than aggressive firms. PROFITABILITY: In a univariate sense, it is difficult to make a priori statements about the profitability of acquired firms. If they are conservative or mis—managed, profits may still be satisfactory, but not as large as they would be with better management. Arguments exist which would allow prediction either way. The FTC implied that acquired firms were the most profitable in their respective industries, but bargains would be expected to occur more frequently with conservative firms who were undervalued by investors. DIVIDEND POLICY: Again, dividend payout ratios by them— selves could be predicted to be either higher or lower by current theory. It would be expected to correlate highly with other financial qualities of the firm. For example, a highly liquid firm, with low levels of debt, and 26 conservative management would be expected to pay good divi- dends. Thus, high dividend payout and acquired firms could aling with one another. However, if mergers occur from stockholder dis-satisfaction, low dividend payout would be a good indicator of stockholder dis-satisfaction, along with poor earnings. Thus, as in the case of the profitability factor, the univariate information content of these qualities is low, whereas their joint behavior with other qualities could be very enlightening. IEBICE/EARNINGS RATIO: This ratio will be a good measure of the cost-of—capital bargain potential which exists according to Alberts, et a1. If the market systematically undervalues small firms, or certain other types of firms, and undervalu- ation can be reflected in low PE ratios, then these firms will be acquired in order to benefit from the market adjust— ment. In this context acquired firms would be expected to have lower price earnings ratios than non-acquired firms. SUBSEggENT ANALYSIS The next section, translates the research questions and hypotheses into a form for formal analysis of differ- ences. Financial qualities such as those listed above, will be measured by alternative ratios and these will be used to develop a formal framework for the analysis of differences between the acquired firms and non-acquired 27 firms. Finally, conclusions will be drawn with respect to these hypothesized differences. CHAPTER III SAMPLE DESIGN, DATA ANALYSIS AND FACTOR ANALYSIS INTRODUCTION The process of gathering and transforming the data into a form suitable for analysis involved a series of com- promises between theoretical rigor tO minimize sampling and non-sampling error, and unavoidable reality always encoun- tered in empirical research. In this chapter the data and ratio analysis sections report the sampling of firms and development of financial qualities hypothesized to be different between the samples. The remaining sections introduce factor analysis as an analytical technique for data simplification. First, the technique is develOped and described, followed by the actual employment of the tool with the data. The two related multi- variate studies (Altman, and Monroe and Simkowitz) relied upon vaguely specified judgemental procedures in reducing the input ratios to a much smaller set for the final dis— criminant model. The use of factor analytic procedures in this study represented an attempt to improve and extend the 28 29 contributions of these researchers' methodologies within the multivariate context. w A primary problem in any research involving mergers and financial information for firms which have been acquired, is that of accurate reasonably complete data. Especially in the case of acquired firms, data is available in only a few cases and is often not complete. Reid, in his book Mergers, Managers and the Economy, noted the limited research on mer— gers reported in the literature.1 The data problem may ex- plain the limited inquiry with respect to the character— istics of acquired firms. Reid notes: "We do not know the pertinent financial data about most of the acquired firms, such as assets, market price of stock (if traded), profits, and sales volume." The data source for listings of firms involved in mergers was the Federal Trade Commission's Bureau of Economics. The FTC annually publishes a listing, Large Mergers in Manufacturing and Mining, containing names of acquiring and acquired firms in mergers beginning with 1Samuel Richardson Reid, Mergers, Managers and the Economy, New York: McGraw-Hill BoOk Co. (1968) 20. 2 Reid, Op. cit. 21. 30 1948.1 Additionally, date of acquisition, asset value and type of merger, are listed. Large mergers were defined as those with asset size at the time of acquisition of no less than $10 million. The deficiencies of the FTC data are well documented in most of the existing empirical research on mergers.2 The primary problem is the extent of coverage. Mergers involving small firms are not recorded with any great frequency, and even the listings of large mergers are not totally comprehensive. The FTC estimated that its listing of large acquisitions included 70.0% of all mergers in the $10.0 million and above asset class. However, in response to an inquiry for listings of mergers involving firms with less than $10 million of assets, the FTC esti- mated their coverage for the smaller classes would be only about 10% of the total number of mergers. The principal data sources used by the FTC are Moody's Industrials, Standard & Poor's Corporation Records, The Wall Street Journal, and other newspapers and prospectuses filed with the Securities and Exchange Commission. 1Federal Trade Commission, Large Mergers in Manu— facturing and Mining 1948-1969, Statistical Report #5, Bureau of Economics, March 1970. 2Gort, op. cit., 632: Pinches, op. cit., 48; Reid, 02. Cite, 20-210 3Federal Trade Commission, Large Mergers, Op. cit., 2. 31 In consideration of these data deficiencies, and the nonexistence of alternative data sources, the SCOpe of the study was limited to industrial firms with assets of at least $10 million at the time of acquisition. Another problem was discovered in develOping the sample of acquired firms. Firms which were included in the FTC listing did not always have complete financial statement data available in Moody's manuals. Moody's was used as the source of balance sheet and income statement data for both samples. In the cases where required data was not complete, the firm was deleted from the sample. Of the original list- ing of 69 firms by the FTC for the sample year, 40 were found with complete data, and these firms composed the sample of acquired firms. The design of the second sample, that of firms which had not been acquired, was a stratified random sample of industrial firms. The stratification was with respect to the size distribution of assets, and was made with the intention of having two samples of firms with the same rela- tive distribution of firms by size of assets. The logic behind this decision was that there is a relationship be- tween the size of a firm and its likelihood of being acquired. For example, firms with asset sizes in excess of $250 million are much more susceptible to Justice 32 Department antitrust action to block the merger than are small firms. The same data availability requirements were imposed upon the second sample. The following criteria were met by the 40 firms included in the second sample: 1. The firm was in existence in 1966 (the sample year) and had not been acquired subsequent to that year, and as of January 1, 1970. 2. The firm had financial data available in Moody's for the years 1961—1966 (the test period). 3. The firm had assets valued at no less than $10 million in 1966. 4. The firm was listed on either the New York Stock Exchange or the American Stock Exchange. The first criteria assured that firms in the second sample were in existence and "available" as candidates during the sample year. The second and third criteria were the same as those met by the first sample. The decision to require firms in the second sample to be listed firms was made to attempt to eliminate some of the firms which may have been attractive merger targets but unavailable for acquisition because they were closely held lFTC, Economic Report on Concentration, Op. cit., 13. 33 and the majority stockholders did not want to merge. While it is recognized that listed firms exist which have small groups of stockholders which constitute a majority of the voting stock, it was felt that a sample of listed firms would be less likely to include these firms. To the extent that firms of this type were included in the second sample, this would decrease the discriminating ability of the model. As a further precaution, during the data collection, if information was available with respect to "large" stock— holders, this was used to exclude the firm. In these cases, if a firm had one or a small number of stockholders who owned more than 20% of the voting stock, the firm was ex- cluded. The 20% cut-off was arbitrary but made with the belief that substantially less than 50% of the voting stock was needed to control a firm which meets listing require- ments, because many stockholders never exercise their voting rights. The selection of 1966 as the test year was based on several factors and represented a compromise. First, a recent year was desired for the test year in order to give as much relevance to the results as possible. However, in order to establish that firms in the second sample were not about to be acquired, it was felt that there should be a time interval subsequent to the test year to observe whether 34 any firms in the second sample had been acquired. Another restriction existed because much of the FTC information for the most recent years was preliminary. Finally, years earlier than 1966 were ruled out because of the small number of mergers listed by the FTC. The 1961—1966 test period was also chosen after con- sideration of several criteria. Of primary importance was that the test period should immediately precede the acquisi- tion, as this period would be used by acquiring firms in their analysis and valuation of merger candidates. When evaluating the level of borrowing of an acquisition candi- date, for example, the relevant data is the current level of outstanding liabilities, not the twenty year history of borrowing. The five year period was chosen to assure the collection of enough data to assure flexibility with respect to the data and its measurement by ratios, some of which are subject to extreme values. Also a five year period would facilitate initial screening of the data for errors in collection and translation to analytic form. To summarize, the original sample consisted of two groups of 40 firms, one group of firms acquired in indus- trial mergers, and a second group which had not been acquired. The test year was 1966 and balance sheet and income statement data were collected for the 1961-1966 35 test period. RATIO ANALYSIS The role of traditional ratio analysis in the area of finance is well known and the wide use of ratios as a tool for analysis in numerous contexts needs no review here. It is equally apparent that "traditional” ratio analysis has lost its place as a tool for analysis, and has been replaced in the current literature by more analytically oriented methodologies using statistical techniques as the basis for models and hypothesis testing.1 Academic research efforts, with the assistance of the computer, have turned to more powerful tools for measurement and testing. Altman re-introduced ratio analysis in a multivariate context and developed a discriminant model with respect to likelihood of bankruptcy based upon a ratio profile.2 The ability of the model to classify firms with over 90% accuracy up to three years prior to bankruptcy illustrates one instance in which the use of ratios with a statistical tool such as discriminant analysis, was very powerful. The success of the Altman model prompted the use of 1Edward I. Altman. "Financial Ratios, Discriminant Analysis and The Prediction of Corporate Bankruptcy," Journal of Finance (September 1968), 589-609. 2Ibid., 590-593. 36 a similar technique in this study. It was decided that financial ratios would be employed with both samples to measure various financial qualities. There exists an almost endless combination of financial data, any of which could be called a ratio. In order to reduce the set of available ratios for financial analysis, a search was made of the mer- ger literature and the work of other researchers in related areas with similar tasks. Due to the number of Significant ratios found to be useful in other studies, a list of 20 potentially helpful ratios was used. Ratios listed were classified into several classes. These classes were profit- ability, leverage (financial), liquidity, and activity. Table 3.1 is a summary of the ratios used, by class, in the final series of analyses. This group formed the data matrix which was factor analyzed. Other ratios chosen initially, or added during the analysis, were subsequently discarded dur to lack of complete and reliable data for all of the firms. For example, cash flow information for other than large firms was often not available in Moody's for the entire test period. In addition, during the analysis, ratios were deleted and changed in an attempt to find those which best measured the above financial qualities and were least subject to extreme variations. Table 3.2 contains pertinent statistics for the samples including means and standard 37 deviations for each of the ratios. FACTOR ANALYSIS AS A MULTIVARIATE TOOL The matrix of correlations among the ratios is pre- sented in Table 3.3 and indicates the high level of col- linearity among the variables. Research problems of this type typically face this situation at some stage. The prob- lem was to choose a subset of the original group of ratios such that the ratios were not highly correlated among one another, and, at the same time, the subset contained as much of the information of the entire set as possible. Such a problem can be solved with the use of factor analysis. Factor analysis is a powerful multivariate method which enables the researcher to simplify and summarize a large data matrix into a smaller one without appreciable loss of information. This technique is primarily concerned with the resolution of a set of observed variables with linear transformation, to form new derived variables (factors), and considerable parsimony, or simplification, is attained. Kendall classifies statistical relationships into two categories; analysis of dependence and analysis of interdependence.1 Analysis of dependence includes both lKendall, M. G. "Factors Analysis as a Statistical Technique," Journal of the Royal Statistical Society, Vol. 12 (1950), 60-73. 38 TABLE 3.1 SUMMARY OF FINANCIAL RATIOS EMPLOYED Class Number Ratio Liquidity 12 net working capital/total assets 18 net working capital/sales Profitability 1 EBIT/total assets 6 gross profit/sales 7 EBT/sales 8 net income/sales 9 EBIT/sales 10 net income/net stockholders equity 11 net income/total assets Leverage 4 long term debt/market value equity LT liabilities/mkt. value equity 14 LT debt/net stockholders equity l9 LT debt/total assets 20 total liabilities/total assets Activity 17 sales/total assets 16 cost of goods sold/inventory 15 sales/(current assets—inventory) "Other" 13* interest/(cash + marketable securities) 2 cash dividends/net income 3 price/earnings NOTE: The distinction between LT debt and LT liabilities was that LT debt included only long term bonds and similar obligations whereas LT liabilities included all entries of a long term nature . *This ratio behaves similarly to LIQ and LEV. 39 analysis of variance and regression analysis. These tools have had wide application in research in the area of finance in recent years. It has appeared to some that, in some cases, their use may have unduly limited the research in which they were used, as well as the conclusions drawn from these studies.1 Within the last several years, however, other tools have been introduced into the literature which may prove to be equally as powerful and more applicable in many instances. These methods come under Kendall's classi- fication of analysis on interdependence. This class in- cludes various types of correlation analysis, cluster analysis, factor analysis, and multidimensional scaling. The important distinction between the two categories is that the techniques under analysis of dependence all require the designation of one or more of the total variables in the analysis as dependent variables. The techniques under analysis of interdependence focus attention on relationships among the total set of variables without singling out any of . . . . 2 them for spec1al conSIderation as dependent variables. lMichael Keenan. "Models Of Equity Valuation: The Great SERM Bubble," Journal of Finance, XXV (May, 1970), 244-260. 2 Jagdish Sheth. "The Multivariate Revolution in Marketing Research," Journal of Marketing, Vol. 35 (Jan., 1971), 14. 40 FACTOR ANALYSIS IN THIS STUDY Kendall's categories are relevant to the use of fac— tor analysis in this study. The specific problem was a large number of intercorrelated ratios with information about a group of firms. The task was to reduce the number of variables without loss of information. The total set of variables was of the essence; no dependent variable existed in this context. The principal aim of factor analysis is the principle of scientific parsimonyl——a construct or a model should be simpler than the data upon which it is based—-which will be illustrated in the following section in terms of the problem at hand. Suppose we have a data matrix, X, of size 20x80, in which each element in represents variable i's measurement of a quality on the jth firm (i=1,2,3,...,n; N=80 and n=20 in this study). For any one of the characteristics x the j: system (le,xj2,xj3,...,xjN) of N real numbers can be con- sidered a point in N-dimensional space. However, by con- sidering each system as a vector, we simplify the configura- tion to 20 dimensions, although these 20 dimensions must 1Jagdish Sheth and Douglas Tigert, "Factor Analysis in Marketing," unpublished paper presented at AMA Workshop in Multivariate Methods in Marketing, January, 1970, 41 pages. 41 be considered as imbedded in the original 80 dimensional space. Factor analysis reduces this n-dimensional space (n=20) even further to r-space (r .NHHH44420000H00 0.0 000Humassdo mocmaum> unwoumm mocmfium> Houomm muwaaum 1000.0 0N0.01 00H.01 00H.0| H 0« 0H Hqul-HON I-lr-tv—Iv-lv-{Hr-l HNMQWONQO‘S NHMH 57 described by two measures of liquidity. Factors 5 and 6 respectively were uniquely identified by dividend payout and price/earnings ratios. A comparison of Tables 3.7 and 3.8 shows that several ratios identified with the last five factors of Table 3.7 did not enter any of the new factors in Table 3.8 as repre- sented by their respective coefficients. This confirmed what was apparent--the first five factors did not describe all of the variance in original space and there were ratios Which contributed uniquely to the reduction of variance. However, they tended to account for only small portions of the variance. The factor analysis had served its purpose well in that from an original space of 20 dimensions, less than ten could be used to represent the great portion of the variance of the total space. The exact number of dimensions used remained flexible so as to allow the discriminant analysis some alternative input matrices. In this application, the usefulness of the ratios for group separation was important, and the number of dimensions in the sample-space was made dependent upon the results of the subsequent discriminant model. The method of arriving at ratios, given alternative factors and factor loadings was simple and straightforward. 58 TABLE 3.8 VARIMAX ROTATION INTO SIX SPACE: SUMMARY OF FACTORS AND ROTATED FACTOR MATRIX Summary Factor Variance Percent Variance Cumulative Percent l 4.76 28.89 28.89 2 4.32 26.19 55.08 3 3.15 19.09 74.18 4 1.78 10.84 85.02 5 1.29 7.86 92.89 6 1.17 7.10 99.99 Ratio 1 2 3 4 5 6 1 -0.157 0.908 -0.185 0.134 0.104 -0.018 2 —0.157 0.074 -0.083 -0.001 0.811— -0.039 3 -0.028 -0.056 0.232 -0.008 0.028 -0.896- 4 0.888 -0.095 -0.032 0.021 0.007 0.027 5 -0.127 0.458— 0.509 -0.304 —0.131 -0.124 6 -0.227 0.787- 0.492 —0.023 -0.161 0.141 7 -0.174 0.667- 0.570 -0.103 -0.235 0.167 8 -0.151 0.781- 0.525 -0.056 -0.196 0.097 9 0.176 0.784- 0.009 0.108 0.320 -0.059 10 -0.l76 0.951— -0.052 0.067 0.121 0.018 11 -0.169 0.079 -0.135 0.842- 0.206 0.218 12 0.609- -0.116 -0.098 0.098 -0.433 -0.100 13 0.933— -0.072 0.008 -0.053 -0.017 0.018 14 0.080 0.017 -0.794— -0.122 -0.056 0.214 15 0.044 -0.045 -0.078 -0.717- 0.214 0.263 16 -0.088 -0.032 -0.850- 0.137 0.039 0.177 17 -0.022 0.108 0.693— 0.564- -0.002 0.231 18 0.937 -0.079 -0.047 -0.079 0.030 0.096 19 0.927- -0.058 0.021 —0.195 -0.006 0.002 20 0.815- -0.145 -0.272 -0.111 -0.300 -0.145 59 It has been widely used and involved choosing the variable with the highest loading from each of the significant factors.1 In that for several of the factors more than one ratio existed with similar high loadings, some flexibility was afforded. The ratios carried forward from the factor analysis are listed belowlby class. The first ratio listed had the highest coefficient on that factor, while others listed for the same factor were ratios with slightly lower coefficients. (1) 1everage--ratios 13 and 18, 19 (2) profitability-~ratios l and 10 (3) activity--ratios l4 and 15 and 16 (4) liquidity—-ratios 11 and 17 Additionally, from Table 3.7 it was noted that ratios 2, 3, and 12 each had contributed somewhat to the variance reduc- tion. This information provided the basis with which a dis- criminant model was develOped to test for group differences with respect to the above qualities. 28Sheth and Tigert, Op. cit., 24. CHAPTER IV MULTIPLE DISCRIMINANT ANALYSIS INTRODUCTION The result of the factor analysis performed upon the original data was to summarize and simplify the data for further use. The reduced data set was the basic input into a discriminant analysis which tested group differences and attempted to predict group membership from the discriminant function which best separated the groups. This chapter is divided into four parts: Part I develops the statistical tool of multiple discriminant analysis (MDA), and relates other applications of the tool. Part II develOps a model based upon the original observed samples. The quality of the model is then evaluated. Part III evaluates basic assumptions of the discriminant model and concludes that the assumptions can better be satisfied with a re-formulation of the samples into more naturally defined, mutually exclusive groups. This was accomplished with factor analysis. Part IV constructs and tests both a comparative model to that in Part II and an 60 61 improved model based on natural groups, as defined. This model was evaluated and then validated as to its ability to accurately identify firms not included in the develOpment of the model itself. At each step, the financial variables used in the models are analyzed. They are evaluated with respect to group differences in a univariate context and then their individual contributions to the total discriminant function in a multivariate framework is studied. Finally, the results are summarized with respect to how the financial qualities differ between the two groups. MULTI PLE DISCRIMINANT ANALYS I S Multiple discriminant analysis (MDA) is a multivariate technique for the classification of a set of objects, by a set of independent variables, into two or more exclusive and exhaustive categories. With respect to the Kendall classifi- cation discussed in the preceding chapter, MDA would be con— sidered as part of analysis of dependence. MDA is an alter- native to multiple regression when the dependent variable is nonmetric (qualitative).l The primary objective of MDA is to correctly classify entities into mutually exclusive groups by the statistical lSheth, Op. cit., 14. 62 decision rule of maximizing the ratio of among—group to within-groups variance—covariances on the profile developed by the independent variables. In addition, the discriminant analysis reveals which of the specific variables employed accounted for the largest portion of the intergroup differences. One of the first known applications of MDA was by M. Barnard in 1935 to date a series of Egyptian skulls.1 R. A. Fisher then applied the method in 1936,2 and the develOpment of the tool was extensive from that time and widely used in the areas of biometrics and psychometrics. In the business related areas, MDA has been used in market— ing for a number of purposes,4 and has had application in finance. One early use of MDA in finance was Walter's study 1M. M. Bernard. "The Secular Variations of Skull Characters in Four Series of Egyptian Skulls," Annals of Eugenics, VI (1935), 352—371. 2R. A. Fisher. "The Use of Multiple Measurements in Taxanomic Problems," Annals of Eugenics, VII (1936), 179-188. 3D. V. Tiedeman. "The Utility of the Discriminant Function in Psychological and Guidance Investigations," Harvard Educational Review, XXI, No. 2 (Spring, 1951), 71-80 0 4Sheth, op. cit., 14-17. 63 l of price/earnings ratios of large corporations and several applications of the tool are apparent in the literature. The recent study by Altman,3 was an attempt to develOp finan- cial profiles of firms for the purpose of using a discrim- inant model to predict likelihood of bankruptcy. The pre— dictive accuracy of Altman's model, 90% accuracy up to three years prior to bankruptcy, suggested the potential of both MDA and the multivariate framework in financial analysis. 1J. E. Walter. “A Discriminant Function for Earnings Price Ratios of Large Industrial Corporations," Review of Economics and Statistics, Vol. XLI (February, 1959), 44—52. 2H. Myers and E. W. Forgy. "DevelOpment of Numerical Credit Evaluation Systems,” Journal of American Statistical Association, Vol. 50 (September, 1963), 797-806. R. J. Monroe and M. A. Simkowitz. "Investment Characteristics of Conglomerate Targets: A Discriminant Analysis," Unpublished paper given at Southern Finance Association meetings, Fall, 1970. 3Altman, Op. cit., 589-609. 64 THE MULTIPLE DISCRIMINANT MODEL FOR TWO GROUPSl Given two mutually exclusive groups with profiles measured by a set of n independent variables, the discrim— inant function is of the form Zi = leli + bzxzi +oool+ anni (l) where b1' b2,..., bn are discriminant coefficients and Xli'.Xzi"°" Xni are independent variables. In the two group case there exists only one discriminant function. The maximum number of Zi in the general case is the smaller of G-l or n, where G is the number of groups. In the two-group case the index, i, is unnecessary as only one discriminant function exists. Equation (1) is a linear combination of the vari- ables, X1' X2...., Xn' which maximally differentiates among the groups. The task of the discriminant analysis is that 1A more rigorous derivation of the tool for the general case of n groups can be found in either of the references listed below. This develOpment is standard as put forth in this chapter and benefited from Tatsouaka and Morrison. M. M. Tatsouaka. Discriminant Analysis: The Study of Group Differences, No. 6, Champaign, Illinois: Institute for Personality and Ability Testing (1970). J. G. Bryan. "The Generalized Discriminant Function, Mathematical Foundation and Computational Routine," Harvard Educational Review, XXI, No. 2 (Spring, 1951), 90-95. D. G. Morrison. "On the Interpretation of Discriminant Analysis," Journal of Marketing Research, Vol. 6 (May, 1969), 156-163. 65 of deriving the set of bi's. Thus, given n variates x1, x2...., xn representing measurements on G groups (two groups of firms in this con— text), there are Ng (g=l,2,...,G) firms with group 9. If there are N firms in group g, the measure of the jth vari— able Of the pth firm in the 9th group would be ngi' Each variate, xi, has a group mean for group j of the form i . = l- 3 x . g: N p 993 (2) Equation (1) with respect to an individual firm is written y = lel + v2x2 +...1+ Van (3) and the group mean of y is =— 4 Y9 N p ng ' () and the sum of squares among groups for y is 2 ’3 N ' (5) g yg g and the sum of squares within groups is Z _-2 p19 (ng Y9) The problem is to determine the coefficients, v1,v2, ...,vn so that the ratio A of the among groups sum of squares to the within groups sum of squares is maximized. Expressed in terms of matrix notation, the among groups sum of squares can be written as a symmetrical matrix (6) 66 = .. =Z -.—. ..=.. .. A [alj] g Ngxglxgj and a1] a31 (1,3 1,2,...,n) (7) and the within groups sum of squares as a symmetrical matrix - = Z .—_ .-—_ ..= -- W — [wij] p,g (ngl Xgi)(xpgj xgj) and W13 W31 (8) and the column vector V = [v.] (9) + 3 Then, equations (5) and (6) respectively become V'AV and V'WV. Their ratio is the criterion maximized with respect to the v. 1 Y: AV, V'WV + + x = (10) Upon setting the partial derivatives of x, with respect to the vi's equal to zero, we arrive at (11'wa - (11'wa = 0 <11) Dividing through by V'NX and collecting terms we get (A - AW)V = O (12) .). which simplifies to (R - XI)V = 0 (13) + -l where I is the identity matrix and R = W A An equation of the form of (13) where R is a matrix of size nxn, with known elements, 2 is an unknown n—dimen— sional column vector, and A, is an unknown scalar, is a well 1 . . problem in applied mathematics. The derivation of the lTatsuoaka, Op. cit., 31. 67 solution will not be shown here. The derivation is shown in Tatsuoaka for the present context, and Morrison for the 1 general case. The solution of (13) for the vi is determined by the latent vectors of R and their corresponding latent roots. In the two group case, only one latent vector provides a set of combining weights v ,...,vn such that the resulting 1V2 linear combination as expressed in (l) and (3) has the largest possible discriminant criterion value among all linear combinations of the n predictor (independent) . 2 variables. CLASSIFICATION PROCEDURE For the purpose of classification of firms using the discriminant model, the following procedure was employed.3 Let each individual's discriminant score, Zi' be a linear function of the n independent variables. Thus, as in equa- tion (1), and for this context, i=l,2,...,80, and Zi = bO + blxli + V2X2i +...,+ anni (14) The classification procedure is as follows if Zi>’zcritical' classify firm i as belonging in group 1 classify firm i as belonging in if Zi‘n, which seems contradictory to the principle of parsimony which underlies factor analysis. This apparent contradiction arises from the assumption that the communality equals the total variance of a variable (unity). The unique factors are not derived from the resolution of the data matrix, but are calculated as residuals to account for the difference between communality and unity. Alternatives exist to either include total variance of the variables in factor analysis, or only that part of it which represents communality. If total variance is explicitly included in the factor analysis, this assumes total variance will in- clude the unique portion and the total will be approximated by the factor matrix. In this case, the model is as follows: Zj = ale1 + asz2 +...+ aerr (11) r and S? = I a? J m=1 3m This solution is known as principal components analysis and was the method employed in this study. The latter solution, including only that portion which repre- sents communality, is known as classical factor analysis. Harman has noted that the principal components solution is that generally employed is empirical research. If the classical solution is desired, then the com- mon factors will approximately reproduce the communality of 124 the variable (h: <1), and a unique factor will be necessi— tated to account for total variance. While the two models, as expressed by equations (3) and (11), are distinctly different, the procedures are essentially the same in the derivation of factor loadings and factor scores matrices. The only difference is that while total variance (8: = 1) is retained in principal com- ponents analysis, an estimate of communality is substituted 2 for total variance (S. = hi) in the data matrix before factoring in classical factor analysis. SELECTED B I BLIOGRAPHY Books Alberts, William W. and Segall, Joel E. (ed) The Corporate Merger, Chicago: University of Chicago Press, 1966. Beveridge, W. I. B. The Art of Scientific Investigation, New York: Random House, 1957. Cooley, William W. and Lohnes, Paul R. Multivariate Procedures for the Behavioral Sciences, New York: John Wiley & Sons, Inc., 1962. Dellenbarger, Lynn E., Jr. Common Stock Valuation in Industrial Mergers, Gainesville: University of Florida Press, 1966. Dixon, Wilfrid J. and Massey, Frank J. Jr. Introduction to Statistical Analysis, New York: McGraw-Hill Book Co., 1969. Farrar, Donald E. The Investment Decision Under Uncertainty, Englewood Cliffs: Prentice-Hall, 1962. Federal Trade Commission. Economic Report on Corporate Mergers. Hearings on Antitrust and MonOpoly, Committee of the Judiciary, U. S. Senate, 9lst Con- gress, lst union, Part 8A, USGPP, 1969. Freund, John E. Mathematical Statistics, Englewood Cliffs: Prentice—Hall, 1962. Hadley, G. Linear Algebra, Reading, Mass.: Addison Wesley, 1961. Harman, Harry H. Modern Factor Analysis, Chicago: Univer— sity of Chicago Press, 1967. Hutchinson, G. Scott (ed.) The Business of Acquisition and Mergers, New York: Presidents Publishing House, Inc., 1968. Mace, Myles L. and Montgomery, George G. Jr. Management Problems of Corporate Acguisitions, Boston: 125 126 Graduate School of Business Administration, Harvard University, 1962. McCarthy, George D. Acquisitions and Mergers, New York: Roland Press, 1963. Moody's Industrial Manual. New York: Moody's Investors Service, Inc., 1961-1970. Morrison, Donald F. Multivariate Statistical Methods, New York: McGraw-Hill BOOk Company, 1967. Reid, Samuel Richardson. Mergers, Managers and the Economy, New York: McGraw-Hill Book Company, 1968. Short, Robert A. Business Mergers: How and When to Transact Them, Englewood Cliffs: Prentice-Hall, 1967. Siegal, Sidney. Non—Parametric Statistics for the Behavioral Sciences, New York: McGraw-Hill Book Company, 1956. Simon, Julian L. Basic Research Methods in Social Sciences, New York: Random House, 1969. Tatsuoaka, Maurice M. Discriminant Analysis: The Study_of Group Differences, Champaign, Ill: Institute for Personality and Ability Testing, 1970. Thurstone, L. L. Multiple Factor Analysis, Chicago: Univ- ersity of Chicago Press, 1947. Van Horne, James C. Financial Management and Policy, Englewood Cliffs: Prentice—Hall, 1968. Weston, J. Fred and Brigham, Eugene F. Managerial Finance, New York: Holt, Rinehart & Winston, 1969. Articles and Periodicals Altman, Edward I. "Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy," Journal of Finance, XXIII (Sept. 1968) 589-609. 127 Beaver, William. "Financial Ratios as Predictors of Failure," Empirical Research in Accounting: Selected Studies 1966, Supplement to Volume 4, Journal of Accounting Research (1966) 71-122. Bernard, M. M. "The Secular Variations of Skull Characters in Four Series of Egyptian Skulls," Annals of Eugenics, VI (1935) 352-371. Bosland, Chelcie. "Stock Valuation in Recent Mergers: A Study of Appraisal Factors," Trusts and Estates (June, July, August 1955) 516—669. Bryan, J. G. "The Generalized Discriminant Function, Mathematical Foundation and Computational Routine," Harvard Educational Review, XXI, No. 2 (Spring 1951) 90—95 0 "Conglomerate Mergers and Acquisitions: Opinions and Analysis," 44 St. John's Law Review (Special Edition, (1970). Current Trends in Merger Activity. Federal Trade Commis- sion, Bureau of Economics, 1969. Fisher, R. A. "The Use of Multiple Measurements in Taxanomic Problems," Annals of Eugenics, VII (1936) 179-188. Gort, Michael. "An Economic Disturbance Theory of Mergers," The Qparterly Journal of Economics, LXXXIII (Nov. 1969) 624-642. Hogarty, Thomas F. "The Profitability of Growth Through Merger," Journal of Business, XLIII (June 1970) 312-327. Keenan, Michael. "Models of Equity Valuation: The Great SERM Bubble," Journal of Finance, XXV (May 1970) 244-260. Kendall, M. G. "Factor Analysis as a Statistical Technique," Journal of the ngal Statistical Society, Vol. 12 (1950) 60-73. Large Merggrs on Manufacturipg and Miningygl948-l969. Statistical Report #5, Federal Trade Commission, Bureau of Economics. 128 Levy, Hiam and Sarnat, Marchall. "Diversification, Port- folio Analysis and the Uneasy Case for Conglomerate Mergers," The Journal of Finance, XXV, No. 4 (Sept. 1970), 795-807. Lewellen, Wilbur G. "A Pure Financial Rationale for the Conglomerate Merger," Journal of Finance, XXVI (May 1971), 521-537. Lintner, John. "Expectations, Mergers and Equilibrium in Purely Competitive Securities Markets," American Economic Review, LXI, No. 2 (May 1971), 101-111. Manne, Henry G. "Mergers and the Market for Corporate Control," Journal of Political Economy. LXXIII (April 1965), 110-120. Morrison, Donald G. "On the Interpretation of Discriminant Analysis," Journal of Marketing_Research, Vol 6 (May 1969), 156-163. Mueller, Dennis C. "A Theory of Conglomerate Mergers," Quarterlleournal of Economics, LXXIII (Nov. 1969), 643-668. Myers H. and Forgy, E. W. "Development of Numerical Credit Evaluation Systems," Journal of American Statistical Association, Vol. 50 (Sept. 1963), 797-806. Reilly, Frank K. "What Determines the Ratio of Exchange in Corporate Mergers," Financial Analysts Journal, XVIII (Nov., Dec. 1962), 47-50. Rulon, Phillip J. "Distinctions Between Discriminant and Regression Analyses and a Geometric Interpretation of the Discriminant Function," Harvard Educational Review, XXI, NO. 2 (Spring 1951), 80-89. Sheth, Jagdish. "The Multivariate Revolution in Marketing," Journal of Marketing, Vol. 35 (January 1971) 13-14. . "Using Factor Analysis to Estimate Parameters," Journal of American Statistical Society, Vol. 64 (Sept. 1969), 808-822. Tiedman, D. V. "The Utility of the Discriminant Function In Psychological and Guidance Investigations," 129 Harvard Educational Review, XXI, No. 2 (Spring 1951), 71-80. Walter, James E. "A Discriminant Function for Earnings Price Ratios of Large Industrial Corporations," Review of Economics and Statistics, XLI (February 1959), 44-52. Williams, W. H. and Goodman, M. L. "A Multivariate Analysis of Corporate Financial Data," Journal of American Statistical Association, Vol. 56 (1969), 884-898. Woods, Donald H. and Caverley, Thomas A. "DeveIOpment of a Linear Programming Model for the Analysis of Merger/ Acquisition Situations," Journal of Financial and Quantitative Analysis, Vol. 5 (Jan. 1970), 627. Unpublished Material Monroe, Robert J. and Simkowitz, Michael A. "Investment Characteristics of Conglomerate Targets: A Dis- criminant Analysis." Paper read before Southern Finance Association, November 1970 (Mimeographed). Pinches, George B. "Financing Corporate Merger and Acquisitions With Convertible Preferred Stock." Unpublished Ph.D Dissertation, Michigan State University, 1968. Sheth, Jagdish N. and Tigert, Douglas J. "Factor Analysis in Marketing." Paper read before AMA WorkshOp on Multivariate Methods in Marketing, January 1970 (Mimeographed). Weston, J. Fred. "Conglomerate Firms and Economic Efficiency." Preliminary draft for comment January 1970 (Mimeographed). "‘111111111111111“