“4'5- 444W” ‘4' 4'”! .'IIJIPI'CI"4& ""4' 440"?" I‘I?‘ 'IIIH' 4'444'I ' I‘:CI4' ".'"'. 'I'”' 4 ‘4I44:'4' IIJ 'I4I4III.4 .I I4'I' m I CfiCIu '4I. 4. 'IIII4III *1. \" III: .4.I'4I '.'4""' I??? III! IIII4C " I "AH' I'I'I’IC |f'IIIIII 'I.""" 11" 4I4I:II{III4 ' 44III' 'III "'IIIC 'I'I'IIII" MI ,II II'I44' MIMI Cums . IIu/M. 'II . -'."fk'/"'"'. Ii‘IIl III CI'4' 4) I III I III I ' 'I""'14 II'4'I ' ' '4 VP .4(4I?I4 IC' """ I" III" ' I'IIIIIIJ'I' 'I! ".'"u'I-I" M IC' 3. ". 3444‘ :"fi? ‘3. '1' ‘34, 44V 4 'h'II'k- [434' 2.14,.34 .'3'” ‘. ' I o ' ' . 193-, I: . , ‘ 4'4'C4II'II" "I 7K'Ixfi'rl4'5‘f. 'L'QIY'IAW: ’4 '2' "3'5" JI'II'IIJI'LN ”'4 5" I'LQ: "'I.:"—.:.d::'. 4 .1 f4» 4 III '4de, 44'3. 444444;. ”$445.44. 4;; 34': $4.4” 4-‘444'.I 4 $144IIZ'IIIEII'IIC vi ‘I'Iu'rl'IfIL'F'I'CCL'? -'-'4'4:!, 'Lf‘ ih’ék'b 35'4" - . I lel‘ f4l3£@;§:j4 ‘ r " I “(-44 "'I " "' '3 ' ' C4'4 ' :31'4'_"' .f'C 4' in”; tjl'h Qti I ' . ‘7' I 4‘4 1'4"” ""'"'CP4'v’I4'I"" “3w. ..,4.:_.~.' 'C" 4 INF; CI." 3%" 'III'" ""‘""' . . 44 44 4 d ‘I 49 4,3. 4444.444; {3" 4145‘1-‘4'4 ' ..4;. 4 . 444.44...r1 44 a}! 'II Chit :[3' CH": 4 :‘f‘j‘kn .':""'"'"Ig‘I'LTI'I'CI 44' I! 4'1”“ Jrfi I4g'fif’ ”gin. '2"; P ‘3)“.034'4'1'4‘?‘ I... 4 4’ :2"? I I 3'5”“. «I! 44"." . i. I J. x C X -:4"4 ' § . ””41 ”I“ 4"474'41" 43;}; ‘ ':.4‘_.4'- "-' '4' \Q' '444441 “’4.“ “44¢. C44! 9.4”:4... - "4 I 44 4 4'3: "."""" fi'hIYI’ .'CIE'I'. '4 1'! .4. l l CC':'1‘C- ‘ 4' I I-‘Iiij ':'-;'.-:j:‘-'"" :5" ”If" 4'“ ' """"" " "' ' I'CII"".'E"'"': 'II'L' ' —w 5%1po'."lv'i""?"" w ' R'K'qu" 44 I: j$I . " "at: .'g I “I I '.."'I." v 2.4 ,4 C“. .< I“. ' 4'14 ”4]”,- 444‘“ "‘1!" "'.-'"'- '"" "'UII" ‘ M4444“ .II‘IIIIF'I«'1‘lim'I .' . " .'K' "' In“ I I a. I 4 . \"" J . '4' r _' ' .~ '. 7:. . .' 4 I' . m' ~ ' .4 [I . _ 0,-:~:':'~'-)-:. {I' ‘. “HI C ‘4“ ‘4" “(II II" “5 'I" gpnov "III“! II III ,ICJ. I‘ :C:'!‘~:.'i5’,".‘.".:.’.-"'!-.,.I ~1!' ' ' .’ICC ' "' 110‘." 4 'I'II'J' HI! I" "'. ' 4 '. IQI'I'UII'JIJIF . . - '1' ICIII'I-II ""*'II" ”445 I441qu! 4"" (:ICICI'I'II '4'L'44'4 4"4_"" 44444 I My}! MICI l- '4'”! 529.1: 444 4'_;. I44‘ "-I4'!III IIIII4IJI‘IIICI THESIS \ &W*h"mw s‘: V! nun-W p“, ‘C This is to certify that the thesis entitled THE RELATIONSHIP BETWEEN SYSTEMATIC OPERATING CASH FLOW RISK AND MARKET RETURNS TO SECURITIES: AN EMPIRICAL STUDY presented by Merle W. Hopkins has been accepted towards fulfillment of the requirements for Ph.D. Business, Accounting Jegree in figéiL—-€ / 1A»cn:j Major professor Date :Q,//~(:/i 2/ 0-7639 RETURNING MATERIALS: ‘IVISSI.J Place in book drop to LIBRARIES remove this ChECkOUt from “- your record. FINES will be charged if book is returned after the date stamped below. - THE RELATIONSHIP BETWEEN SYSTEMATIC OPERATING CASH FLOW RISK AND MARKET RETURNS TO SECURITIES: AN EMPIRICAL STUDY BY Merle Wayne Hopkins A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Accounting 1982 ABSTRACT THE RELATIONSHIP BETWEEN SYSTEMATIC OPERATING CASH FLOW RISK AND MARKET RETURNS To SECURITIES: AN EMPIRICAL STUDY By Merle Wayne Hopkins This research is an empirical investigation of the relationship between corporate cash-flow risk measures and stock market returns. Prior research has generally used accounting income numbers to con- struct accounting-based risk measures. The association between such measures and systematic risk estimated via a market model has been cited as evidence of the usefulness to investors of accounting data. In this study, the explanatory power of cash-flow risk measures is directly tested. The testing procedure employed is that developed by Fama and MacBeth. The objective of their research and the present research is to assess directly the relationship between systematic risk and returns to securities. In their study the measure of systematic risk was formed from historical market returns rather than from historical corporate cash flow data as in the present study. In this study, portfolios of firms with homogeneous cash flow risk levels were formed. The cash flow risk of the portfolio is used as one independent variable among others in an attempt to establish the extent to which market returns (dependent variable) reflect this mea— sure of systematic risk. The principal hypothesis is that systematic cash-flow risk and Merle Wayne Hopkins market returns are related. The test results are not consistent with this hypothesis. The statistical power of the test procedure was reduced by a large degree of instability in the observed levels of estimated corporate cash flow risk. This instability hindered the formation of portfolios with homogeneous levels of cash flow risk at the firm level and led in part to inconclusive results. Several sug- gestions to improve the test design are included to be pursued in future research. DEDICATION This dissertation is dedicated to my family and friends. Their understanding and confidence have made a major contribution to the completion of this study. ACKNOWLEDGMENTS I wish to express my sincere thanks to my dissertation com- mittee of Professor Richard R. Simonds (Chairman), Professor Melvin C. O'Connor, and Professor Kelly Price. Their counsel and encourage- ment have been appreciated at the various stages in the development and completion of this research. This research was partially financed by the Deloitte Haskins & Sells Foundation and the Department of Accounting of the Graduate School of Business Administration at Michigan State University. I offer my hearty thanks to these organizations. Merle Wayne Hopkins iii TABLE OF CONTENTS Page LIST OF TABLES. . . . . . . . . . . . . . . . . . . . . . . . vi LIST OF ABBREVIATIONS . . . . . . . . . . . . . . . . . . . . vii Chapter INTRODUCTION. . . . . . . . . . . . . . . . . . . . viii I. SIGNIFICANCE OF CASH FLOW MEASURES. . . . . . . . . 1 Objectives of Financial Statements. . . . . . . . 1 Tentative Conclusions on Objectives of Financial Statements of Business Enterprises. . . . . . . . . . . . . . . . . . . 2 Statement of Financial Accounting Concepts No. 1 . . . . . . . . . . . . . . . . . 2 Discussion Memorandum on Reporting Earnings . . . 3 Capital Asset Pricing Model . . . . . . . . . . . 4 Market Model. . . . . . . . . . . . . . . . . . . 5 Theoretical Determinants of Systematic Risk . . . . . . . . . . . . . . . . . . . . . . 7 II. LITERATURE REVIEW . . . . . . . . . . . . . . . . . lO Empirical and Analytical Determinants of Systematic Risk . . . . . . . . . . . . . . . 11 Operating Risk and the Cost of Capital. . . . . . 14 Operating Risk, Cost of Capital, and Risky Debt . . . . . . . . . . . . . . . . . 15 Financial Leverage and Systematic Risk. . . . . . 16 Operating Leverage and Systematic Risk. . . . . . 19 Operating Leverage and Financial Leverage . . . . 20 Accounting Measures of Risk and Systematic Risk. . . . . . . . . . . . . . . . . 22 Covariant Forms of Accounting Variables and Systematic Risk. . . . . . . . . . . . . . . 25 Corporate Decision Variables and Systematic Risk. . . . . . . . . . . . . . . 28 Empirical Relation Between Ex Post Risk and Market Returns. . . . . . . . . . . . . 29 Role of the Present Study . . . . . . . . . . . . 30 iv III. MODEL DEVELOPMENT . . . . . . . . . . . Modigliani and Miller . . . . . . . . Haugen and Pappas . . . . . . . . . . IV. CONSTRUCTING THE FIVE CASH FLOW RETURNS THE SYSTEMATIC CASH FLOW RISK MEASURES FOR THE FIRMS O O O O O O C O O O O O 0 Defining the Five Cash Flow Return Measures . . . . . . . . . . . . . . Seasonally Adjusting the Various Return Series. . . . . . . . . . . . Calculating the Variables Necessary in the Present Study . . . . . . . . v 0 TESTING O O O O O O O O O O I O O O O 0 Test Design and Research Hypotheses . Testing Procedures. . . . . . . . . . Market Return Ranked Estimation and Testing Periods. . . . . . . . . Cash Flow Return Ranked Estimation and Testing Periods. . . . . . . . . VI. TEST RESULTS. . . . . . . . . . . . . . $F Relation Between bp Relation Between b$F Definitions 1 and 2. . . . . . . . . and Rmr. . . . . P Rankings on Cash Flow Data versus Market Return Data . . . . . . . . . Relation Between the Risk-Free Rate and the Rgr O O O O O O O O O 0 Relation Between bgr and Rgr When Using Quarterly Returns. . . . . . . Relation Between 3(51) Term and the Observed r in the Regression. . . . VII. CONCLUSIONS AND LIMITATIONS . . . . . . MPENDIX A O O .7 O O O O O O O O O C O O O O O 0 APPENDIX B . . . . . . . . . . . . . . . . . . . BIBLImRAPHY O O O O O O O O O O O C O C O O O O O and Rgr using Cash 36 37 38 46 46 47 49 53 54 56 59 60 63 74 75 77 8O 81 82 84 87 91 98 Table LIST OF TABLES Test Results for Cash Flow Definition 1 (Eight Panels). . . . . . . . . . . . . . . Test Results for Cash Flow Definition 2 (Eight Panels). . . . . . . . . . . . . Test Results for Cash Flow Definition 3 (Eight Panels). . . . . . . . . . . . . . Test Results for Cash Flow Definition 4 (Eight Panels). . . . . . . . . . . . . Test Results from a Replication of the Fama and MacBeth Study Panel A (monthly market return data). Panel B (quarterly market return data). . vi Page 65 67 69 71 73 73 AICPA BKS CAPM CRSP FASB NOI NYSE OLS PIF ABBREVIATIONS American Institute of Certified Public Accountants Beaver, Kettler, and Scholes Capital Asset Pricing Model Center for Research in Security Prices Financial Accounting Standards Board Haugen and Pappas Modigliani and Miller Net Operating Income New York Stock Exchange Ordinary Least Squares Production, Investment, and Financing vii INTRODUCTION This dissertation investigates the empirical relationship between corporate net operating cash flows and market returns to equity holders. The relationship is explored through the use of an analytical model which relates systematic cash flow risk to firm financial leverage. Chapter One discusses the significance to investors of cash flow information. Chapter Two reviews the extant literature in this area of research. Chapter Three develops the analytical model used in this study. Chapter Four defines the various cash flow measures employed in empirically testing the model. Statistical test proce- dures are explained in Chapter Five and the results are presented in Chapter Six. Conclusions and limitations of the present study are detailed in Chapter Seven. viii CHAPTER ONE SIGNIFICANCE OF CASH FLOW MEASURES This chapter summarizes the importance of actual and expected cash flows to the areas of accounting and finance. From the perspective of finance, these cash flows are discussed in relation to firm valuation and systematic risk. Providing information leading to accurate prediction of future cash flows to each firm has been mentioned frequently as a desirable objective of financial reporting. Many prominent references to this objective are discussed below. This desirability originates in the understanding that present and future net cash flows to the firm are the ultimate source of cash flows to the debt and equity investors. The American Institute of Certified Public Accountants (AICPA) appointed a Study Group on Objectives of Financial Statements in April of 1971. This act resulted in a statement of Objectives of Financial Statements (commonly known as the Trueblood Report) issued by the AICPA in October 1973. Among the objectives presented and defended by the Study Group is the following statement: Information about the cash consequences of decisions made by the enterprise is useful for predicting, comparing and evaluating cash flows to the users.1 The Study Group felt that knowledge about current and expected cash flows to the firm was an essential part of an investor's useful micro-information set. The micro-information set is the aggregation 2 of all information available on a single firm. The investor's objective may be to form reliable expectations of future cash flows to himself via dividends or interest from either equity or debt securities, respectively, or from the proceeds of the sale of these debt or equity securities. This objective assumes a link between cash flows to the firm and cash flows to the investor. All valuation models employing cash flows to the firm as a principal element rely on this presumed relationship. The Study Group reported the following among the criticisms of current accounting practice relative to an investor's legitimate need for reliable information on or expectations of cash flows to the firm: Assertions have been made that the present financial state- ments do not provide sufficient information about the liquidity and cash flows to an enterprise.2 The Financial Accounting Standards Board (FASB) issued its Tentative Conclusions on Objectives of Financial Statements of Business Enterprises in December 1976. The importance attributed by the Study Group to the firm's cash flow information was adopted and made more explicit by the FASB in these tentative conclusions. In November 1978, the FASB issued Statement of FinancialAccounting Concepts No. 1, "Objectives of Financial Reporting by Business Enterprises." In this Statement the FASB expressed the following view: Financial reporting should provide information to help present and potential investors and creditors and others in assessing the amounts, timing, and uncertainty of prospective cash receipts from dividends or interest and the proceeds from the sale, redemption, or maturity of securities or loans. The prospects for those cash receipts are affected by an enterprise's ability to generate enough cash to meet its obligations when due and its other cash operating needs, to reinvest in operations, and to pay cash dividends and may also be affected by perceptions of investors and creditors generally about that ability, which affect market prices 3 of the enterprise's securities. Thus, financial reporting should provide information to help investors, creditors, and others assess the amounts, timing, and uncertainty of prospective net cash inflows to the related enterprise.3 An important aspect of financial reporting is providing information aiding in assessments of cash flow prospects to the enterprise as the FASB clearly asserted above. In July 1979, the FASB issued a discussion memorandum on reporting earnings which announced that the resolution of the reporting issues surrounding the relationship between the earnings process and cash flows would wait until the forthcoming project on fund flows and liquidity. The FASB did, however, make the following interesting statement with respect to its current thoughts on the significance of cash flows and the earning process: . . . . it is desirable that financial reports should give information on the differences in timing between current earnings and current cash flows and information to help users to project such differences into the future.4 The investor-user group has been emphasized in the foregoing discussion on the merits of cash flow data relative to accrual earnings. Equity security and debt security owners or potential owners are included in this investor-user group. Among other user groups which may be interested in cash flow data are sellers of resources where payment is to be deferred. The decision model approach to accounting theory manifests itself in the specification of cash flows as a variable of interest. This approach to accounting theory holds that the deciSion model specifies the variab1e(s) of interest. In this context cash flows achieve importance because they are the driving force of firms' values and 4 the expected values of firms enable an investor to expand or preserve his utility. The capital asset pricing model (CAPM) based on the two-parameter portfolio theory developed and extended by Markowitz,5 Tobin,6 Sharpe,7 Mossin,8 and others specifies that the only firm-specific parameter necessary to establish the expected return to firm i is Bi’ the systematic risk parameter of firm i in the market portfolio m. The CAPM is presented and described below: Eq' 1 E(Rit) g th + Bim [E(Rmt) I th] where E(Rit) equals the expected return on security i in time t, conditional upon 81m and Rmt' B = cov (R R t)/var(Rmt) which reflects the im it’ m systematic risk of security i in the market portfolio m. E(Rmt) = the return on the market portfolio in time t. th = the riskless rate of return. Note: 8 need not be a constant or known ex im ante despite its appearance in the CAPM. Summarizing to this point: (1) there is an extensive body of literature in finance arguing analytically and intuitively that firms' values are functions of actual and expected cash flows. (2) Accounting information is part of the macro-information set available for use by the market participants. This macro-information set is the aggregate of all information available with or without cost to economic decision- makers. Cash flow information for the firm can be derived from the published accounting data. (3) An analytical theory (CAPM) argues 5 persuasively the significance of Bi as the relevant firm—specific risk parameter under a standard set of assumptions. It is an empirical question whether the analytical and intuitive significance of cash flow information to firm valuation can be used to more accurately estimate the systematic risk measure, 8 than can the use of market 1. return data alone. Expectations within the two-parameter model describe the process through which equilibrium prices are formed in the marketplace. An equation can be expressed describing the relation between the expected return on any security in an efficient portfolio and the risk contribu- tion that security makes to the portfolio. The market model has been commonly used to generate estimates of 81 via ordinary least squares (OLS) for use in testing the CAPM. The market model is presented below: Eq. 2 En = th + bld‘mt - th> + in“. where Rit = the return to security i in time t. th = the riskless rate of return in time t. ~mt = the return on the market portfolio in time t. bi = the OLS estimate of Bi (referred to here as ex post beta). fiit = the random error term. E('lit'R‘mt) g 0' cov (“it’Rmt) = 0. cov (£1 ) = O. t’ui,t+l Note the subtle and crucial stationarity assumption made when using the market model to implement the CAPM. Bi is assumed to be stationary in its estimation period and over the future time horizon of market participants. This assumption may be the source of the naive 6 belief that this parameter is stable over intermediate time spans. The CAPM does not require the assumption of a stationary 81' The expecta- tion of Bi will be referred to here as ex ante beta. When Bi will remain stable and_is expected to do so, the ex post and ex ante betas will be equal. The introduction here of the ex ante beta acknowledges that market participants may modify the ex post betas in view of whatever pertinent information is available at time t or is expected to become available between time t and t+n. In the context of the CAPM and ex ante B, accountants should explore the role of accounting data as a possible source of useful information to investors seeking to revise ex post b in such a way i that the expectation of future systematic risk is closer to the Si which will in fact be observed over the relevant time horizon in the future than is ex post beta.9 The investors may form portfolios to more accurately reflect their tastes for expected risk and expected return if they have better ex ante B estimates. It should not be concluded, however, that the role of accounting data is or should be limited to this objective of seeking better ex ante estimates of true Bi over some future time horizon.10 Two significant implications could result if returns to securities were found to be some function of corporate cash flow data. One, that cash flow data serves as a meaningful determinant of systematic risk. This could be used in a more fully specified market model to perhaps obtain better estimates of ex ante beta than are currently generated using only past return series. Two, that the CAPM might be made into a more complete, richer model through the addition of another variable 7 reflecting the form of the function of cash flow data found to be significant. What are the theoretical determinants of the systematic risk measure? Fama and Miller develop the concept that 81 is a representation for the standardized covariance of expected outcomes of production, investment and financing decisions (PIF decisions) made by the firm relative to PIF decisions made by other firms in the market.11 This covariance of expected outcomes of PIF decisions made by the firm and the market would of course be standardized by the variance of the market return. The standardized covariance relation is based entirely upon expectations. Haley and Schall compare the firm to a money pump.12 The value of the firm is determined in this context by the firm's output of money to its debt and equity owners and the expected future output of money relative to the market's actual and expected output of money. But where do these expectations of future cash flows to the firm and to the market originate? For a moment, consider the basis upon which PIF decisions are made by the managers of firms. The managerial finance literature is founded upon the premise that managerial decisions (PIF decisions) ought to be made via a careful evaluation of future uncertain cash flows over time. Thus, when the PIF decisions of a firm change relative to those decisions made by other firms in the market, we should expect a 81 change. The covariance of the current cash flows to the firm and to the market would be expected to differ from the former covariance relation to the extent relative changes in PIF decisions involve current period cash flows. The variance of current cash flows to 8 the market could be altered as a result of the relative change in firms' PIF decisions. The foregoing directly implies the efficacy of a current cash flow beta. Given the links between cash flows, expected cash flows, firm valuation, and market returns, can a systematic current cash flow risk measure be developed analogous to the market model b and be found 1 to be part of the market return generating process? FOOTNOTES - CHAPTER ONE 1American Institute of Certified Public Accountants, Objectives of Financial Statements (New York: 1973), p. 14. 2Ibid., p. 16. 3Financial Accounting Standards Board, Statement of Financial Accounting Concepts No. 1, "Objectives of Financial Reporting by Business Enterprises" (Stamford, Connecticut: 1978), p. 17. 4Idem, Discussion Memorandum: "An Analysis of Issues Related to the Reporting of Earnings" (Stamford, Connecticut: 1979), p. 95. 5H. Markowitz, Portfolio Selection: Efficient Diversification of Investments (New York: Wiley, 1959). 6J. Tobin, "Liquidity Preference as a Behavior towards Risk," Review of Economic Studies 25 (February 1958): 65-86. 7W. F. Sharpe, "Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk," Journal of Finance 19 (September 1964): 425-442. 8Jan Mossin, "Equilibrium in a Capital Asset Market," Econometrica 34 (October 1966): 768-783. 9bi is the market model's estimate of 81. b1 is generated from observed returns and has not been revised in any direct way for changes in the return prCSpects of the firm nor of the market. 10James C. Boatsman and Lawrence Revsine, "Firm Specific Accounting Data and the Capital Asset Pricing Model: A Reconciliation and Extension," (unpublished paper, 1979), p. 19. 11Eugene F. Fama and Merton H. Miller, The Thegry of Finance (New York: Holt Rinehart and Winston, 1972), Chapter 7. 12Charles W. Haley and Lawrence D. Schall, The Theory of Financial Decisions, 2d ed. (New York: McGraw—Hill Book Company, 1979), p. 25. CHAPTER TWO LITERATURE REVIEW Chapter One outlined in some detail the role the present research plays in identifying the determinants of systematic risk. Numerous previous works, many of which are discussed below, have addressed the problem of identifying these determinants. The purpose of the present chapter is to identify and summarize significant previous research and form an integrated background for the present research. The last part of this chapter is devoted to the integration of previous research papers into a framework which permits the role of the present research to be more fully understood. The ten papers reviewed in this chapter are grouped conveniently into four areas. The first area includes the papers by Modigliani and Miller,1 Haugen and Pappas,2 Hamada,3 and Myers.4 These papers develop the idea that both cyclicality and financial leverage should be determinants of systematic risk. These papers generally present empirical or analytical arguments in support of these determinants. The second area contains papers by Lev5 and Wolfson.6 These papers suggest that operating leverage in conjunction with financial leverage is a determinant of systematic risk. The third area is composed of the papers by Beaver, Kettler, and Scholes;7 Thompson;8 and Bildersee.9 These papers eXplore the association between several variables and the observed level of systematic risk displayed by the firm. The 10 11 predictive power of these variables with respect to future observa— tions of the level of systematic risk displayed by the firm is also explored. The variables analyzed in this area include both accounting variables (risk measures) and corporate decision variables. The fourth area contains only the Fama and MacBeth10 paper. This paper provides the basic procedures for testing the research hypotheses in the present paper. Empirical and Analytical Determinants of Systematic Risk Myers11 presented a review of empirical research in the area of systematic risk determinants. In addition, he developed a model incorporating variables which he analytically showed to be determinants of systematic risk. In his review of significant empirical research, it became apparent that the principal emphasis to date has been on accounting variables. Occasionally, these accounting variables have been used in conjunction with financial market variables. Myers was troubled by the use of financial market risk measures in attempting to understand the determinants of systematic risk. He felt that those financial market measures would reflect not only the real sources of systematic risk but also the variables perceived as proxies of those real sources. For this reason, he believed that studies should exclude financial market measures from the independent variables used as prospective determinants of systematic risk. Myers pointed out that the determinants of systematic risk need not be the same each period, and that the relative roles played by the various determinants of systematic risk need not be the same across 12 all time periods. Thus, he argued that the initial problem would be the identification of the various determinants of systematic risk. With these determinants identified, the relative roles they played could be examined over time. Myers' empirical research summary listed four measures thought to be determinants of systematic risk and which the empirical tests had confirmed as being generally significantly associated with systematic risk: cyclicality, earnings variability, financial leverage, and growth. Cyclicality described the relation between the earnings process for a single firm and the earnings process of the market. If these two earnings processes tend to move closely together, then the cyclicality measure would be high for the firm. Conversely, a low cyclicality measure would reflect an earnings process largely immune from fluctuations in the earnings process of the market. Myers concluded that the significance of earnings variability may have stemmed from its being a surrogate for cyclicality. He felt that financial leverage was well supported in the empirical research literature as a determinant of systematic risk. He noted that the measurement problems surrounding the testing of growth were especially troublesome. Growth exhibited uneven significance as an explainer of systematic risk over the studies he examined in which it had been one of the variables examined. It may be that the special measurement difficulties for this hypothesized determinant make the erratic significance understandable. In the portion of his paper that developed an analytical model explaining the role played by the hypothesized and empirically supported determinants of systematic risk, Myers concluded that cyclicality, l3 earnings variability, and financial leverage ShOuld be determinants of systematic risk. He began by assuming that cyclicality was a determinant. His analytical work did not confirm the empirically irregular significance of growth as a determinant of systematic risk. The principal significant feature of the model he developed is the covariance between the market's earnings and the "surprise" portion of the firm's earnings. Another feature of the model is a term reflecting the elasticity of earnings expectations. This term reflects the covariance of elasticities of earnings expectations for the individual firm and for the market. Specifically, _ .__£;tll. cov (Rm, 6) where B c a h) r = one period rate of riskless interest. n = elasticity of expectations or the ratio of the signal of a changed mean earnings to the sum of this signal plus noise contained in the signal. A = the price of risk as traditionally expressed, or Rm - Rf . 0(fim) p = the correlation between a security's rate of return and the market's rate of return. 5 = (expected cash flow at time t)t_1 - (actual cash flow at time t)t' This suggests intuitively the relevance of an ex ante beta as the variable of interest based upon those elasticities of earnings 14 expectations. Myers concluded that it is questionable whether cyclicality can be measured by a single statistic. He acknowledged that an important problem existed in the comprehensive measurement of cyclicality for use in empirical studies. Operating Risk and the Cost of Capital Modigliani and Miller12 claimed in their Proposition 11 that the expected return to equity investors (r ) was a linear function of the j capitalization rate (pk) appropriate for the firm's operating characteristics, the risk premium originating in the risky operating characteristics (pk - r), the risk-free rate of return (r), and the debt/equity ratio (Dj/Sj): D. j=°k+(pk'r)‘s‘:f Eq. 4 r A firm has an inherent operating risk level associated with the operating characteristics of the firm. This risk level cannot be changed by financial leverage within MM's risk classes. MM argued that financial leverage can only change the systematic risk levels borne by the various classes of debt investors relative to the systematic risk experienced by the equity investors. It follows directly that financial leverage should be a determinant of the systematic risk level experienced by the equity investors. Conversely, financial leverage should not be a determinant of the systematic operating risk levels displayed by the firms within the risk—class construct utilized by MM. Before MM's Proposition II, the traditional view had been that the required rate of return to equity investors in the absence of taxes was invariant with respect to the debt ratio employed by the firm until some critical debt level had been surpassed. MM argued that 15 investors were sensitive to even low levels of financial leverage because of the effect such leverage had on the distribution of returns to equity investors. The significance of the work by MM is that financial leverage should be positively correlated with the required rate of return across all levels of financial leverage. This viewpoint is consistent with the CAPM regarding the role of borrowing or lending upon the return required by the equity investors. Operating Risk, Cost of Capital, and Risky Debt Haugen and Pappas13 analytically linked the theory of investor behavior toward risk with the theory of corporate finance. Essen- tially, they showed that Modigliani and Miller (MM) were correct in the Net Operating Income construct for the cost of capital. Haugen and Pappas showed that the cost of capital for the firm was invariant with respect to the extent of financial leverage undertaken by the firm. Moreover, Haugen and Pappas demonstrated this result without the restrictive use of MM'S homogeneous business risk classes. Haugen and Pappas's analytical results suggest that one can use the systematic risk level of the firm in conjunction with the systematic risk level of the firm's risky debt securities and the firm's level of financial leverage to obtain an alternative specifica- tion of the systematic risk level of the firm's equity securities. The analytical framework developed by them provides the basis for present study. See Chapter Three of the present research for the details of their analytical framework as it has been implemented here. 16 Financial Leverage and Systematic Risk Hamada14 examined the relationship between levels of financial leverage and systematic risk. One motivation for this research was to confirm the MM theory, which argued that the financial leverage would increase the rate of return required on equity securities. Hamada assumed MM to be correct and proceeded to construct both levered and unlevered return measures. The latter is the levered return measure which has been adjusted both in the numerator and denominator to represent what the unlevered return would have been. These unlevered returns are not directly observable since virtually all firms utilize some degree of financial leverage. First, the levered systematic risk measure and then, the unlevered systematic risk measure as Hamada developed them: cov (~ , R ) Eq. 5 8levered = RBt mt ozdz ) mt cov (R , R ) Eq. 6 Bunlevered = t mt o2 mt ~ (X-I)(l-T)t-Pt+AGt where RBt = 5 ° Bt-l I = interest payments. P = preferred stock dividends. SBt-l = the market value of the firm's common shares at the beginning of the period. X (l - 1')t + AGt At SAt-l '21: ll SAt-l = the market value of the firm's common shares if the firm had no debt or preferred shares. 17 X! It the firm's earnings before taxes, interest, and preferred dividends. T = the corporate tax rate. AG = the change in the firm's future growth opportunities occurring since t-1- 02(Rmt) = the variance of the return to the market. ~ mt = the market rate of return at time t. These levered and unlevered return measures were constructed from annual CRSP and Compustat data. The return measures were separately regressed onto a market return index. The market return index was the NYSE arithmetic rate of return where the dividends had been reinvested. The study included 304 firms which had complete data on both CRSP and Compustat tapes over the 20-year span, 1948-1967. The regressions were run using 20 annual observations for each measure. Hamada reported results consistent with the hypothesis that financial leverage is associated with a higher level of systematic risk. He claimed to have explained 20—24 percent of the mean systematic risk levels through the use of his levered and unlevered return measures. Thus he argued that financial leverage must be a determinant of system- atic risk because it alone had explained a significant proportion of the value of the mean beta. Hamada acknowledged problems in the measurement of the unlevered return measure. However, he claimed support for the role of financial leverage as a determinant of systematic risk in spite of the measurement problems. Unless the direction of the measurement errors can be assessed, the results obtained may be because of, rather than in spite 18 of, those measurement errors. However, measurement errors need not necessarily work against a significant finding. Hamada did conduct further regressions to check the apparent adequacy of the unlevered systematic risk estimates. He modified the unlevered systematic risk measure and proceeded to regress the modified risk measure on the traditional market-return determined systematic risk measure. The slope coefficients observed in these regressions were remarkably near 1.00 when using a 20-year average for the ratio of the equity's market value in the levered case (numerator) and in the unlevered case (denominator). This ratio was the device used to modify the unlevered systematic risk. A slope coefficient of 1.00 would have indicated a perfect positive relation between the two risk measures. 0f significance to the present research is Hamada's conclusion that financial leverage is a determinant of systematic risk. This conclusion has been regarded as valid. For example, Myers later stated that Hamada's empirical results "justify substantial confidence in the link between financial leverage and the stock beta."15 Myers even excluded financial leverage from his analytic model because he felt no further analytical justification was required in defense of financial leverage as a determinant of systematic risk.16 Hamada's investigation of the relationship between financial leverage and systematic risk could have been conducted at the portfolio level. This might have reduced the role of measurement errors inherent in the measurement of the return to the unlevered securities. If those measurement errors were random, they could be eXpected to be l9 reduced if portfolios were formed either randomly or using some ranking on the observable level of systematic risk. Operating Leverage and Systematic Risk Lev17 hypothesized and then analytically demonstrated that the firm's degree of operating leverage has a positive relation to the systematic risk level displayed by the firm's equity securities.18 This analytic conclusion was consistent with the theory put forth by Fama and Miller19 that the production, investment, and financing decisions made by a firm were collectively reflected in the systematic risk measure. The production and investment decisions made by the firm relative to similar decisions by other firms would influence the firm's relative operating results in reSponse to relative changes in the demand for the goods and services produced by the firm or by the market. Thus, intuitively, operating leverage and its determinants should have an influence on the systematic risk level of the equity securities. Lev regressed the total cost of production onto the levelcflfouput achieved by the firm measured in relatively homogeneous units. He used the resulting slope coefficient as his surrogate for the variable cost per unit of output. In the next step of his study, he regressed the firm's systematic risk measure onto the variable cost per unit in order to assess the strength of that relation. The results of regression con- firmed his hypothesis and analytic conclusion that there is a positive and significant relation between operating leveragezuuisystematic risk. In selecting firms for inclusion in his study, Lev restricted himself to industries whose units of output were relatively homogeneous so he could cross-sectionally consider many firms within the same industry. Industries selected were electric utilities, steel 20 production, and oil production. There were two test periods, 1949- 1968 and 1957-1968. These periods of overlapping years, but of differing lengths, reduced the sensitivity of the conclusions to changes in the production technology within these industries. Lev's results confirmed the relation between the firm's operating leverage and systematic risk of the firm's equity securities. Operating Leverage and Financial Leverage As many authors have shown, financial leverage is considered to be a determinant of systematic risk. In his dissertation, WolfsonzO has shown empirically that financial leverage and operating leverage are inversely related to each other. Moreover, he has shown that both are directly related to the systematic risk level of the firm's equity securities. Thus, these sources of risk are complementary in their effect on the overall systematic risk level of the firm's equity securities. In establishing this link on an intuitive plane, Wolfson used agency cost arguments. Agency cost is the difference in the firm's market value caused by the necessity to use professional management personnel rather than owners and lenders in management. He argued that when a firm exhibits any degree of financial leverage, maximizing the value of equity may not be consistent with maximizing the value of the firm. Managers acting for equity owners may select investment oppor- tunities which are sub-optimal with respect to maximizing the firm's value. Their justification for making these decisions may be a desire to abide by the indenture agreements putzhmplace by the debt security owners. These creditors have perhaps sought to restrict the kinds of Operating decisions the managers could make so the risk level of the 21 firm could not be substantially increased. A drastic increase in the operating risk level of the firm could cause a wealth transfer from the debt owners to the equity owners. The sudden increase in the operating risk could result from entrance into a very risky but profitable venture where the debt securities prices would fall. In his empirical testing, Wolfson used monthly time series of several variables which were taken from the monthly CRSP tapes, or which were constructed from both CRSP and Compustat data. His primary effort was to estimate the systematic operating risk, overall systematic risk, variability of the operating return measure, and the degree of financial leverage, each for adjacent 30-month periods within a 60-month span. The systematic operating risk measure was obtained by regressing a market-value-weighted return to holders of various classes of debt and equity of a single firm onto the return to the equally weighted market index. Through the use of a Chi-square test, he found that the interrelationships he hypothesized were present in the variables he constructed. The significance of Wolfson's work is that consideration of operating risk in the absence of consideration of financial leverage may produce erratic results. Moreover, both kinds of risk are positively related to the firm's overall cyclicality risk level. In light of Wolfson's findings, it is not prudent to explore the role played by cash flow risk without controlling for the degree of financial leverage employed by the firm. If systematic cash flow risk were the same as either Lev's operating leverage or Wolfson's operating risk, than Wolfson's findings would strongly argue for the simultaneous inclusion of financial 22 leverage into the consideration of either as a determinant of systematic risk. Systematic cash flow risk, however, is not the same as either Operating risk or operating leverage. Whether one of these can be a surrogate for the other is an interesting issue, but not relevant to the present research. Accounting Measures of Risk and Systematic Risk Beaver, Kettler, and Scholes21 (BKS) conducted one of the earliest studies in the testing of the contemporaneous relation between accounting measures of risk and the market-determined measure of risk (ex post beta), and the usefulness of accounting measures of risk to predict the relevant systematic risk parameter (ex ante beta). This study can be interpreted as (1) an attempt to explore certain traditional accounting-based risk measures, looking for evidence that one or more of these accounting risk measures contemporaneously reflects a deter- minant of systematic risk, or as (2) an attempt to establish a causal relation between the accounting risk measures and the market risk measures. The authors felt that an accounting system could be constructed or defended in terms of the decisiondmaking criterion if either interpretation of their first objective was supported. The accounting risk measures were developed from a review of the traditional security analysis and valuation literature. No analytic theory was used in defense of these specific accounting measures of risk. One accounting risk measure, however, is substantially exempt from these comments. Specifically, the accounting beta has an intuitively and conceptually persuasive body of reasoning behind its selection for use in the BKS study. 23 The BKS study analyzed the financial statement data of 307 firms whose annual Compustat data and monthly CRSP data were complete for the 1947-1965 period. The l9-year period was broken into two time periods so tests could be run assessing the stationarity of the accounting and market-based risk estimates between the two periods. The contemporane— ous association tests were conducted via correlation analysis at both the individual security level and the portfolio level. The correlations in both time periods were statistically significant for four accounting risk measures: payout ratio, financial leverage, earnings variability, and accounting beta. This significance held at the individual security level and at the portfolio level. The tests of prediction of future systematic risk involved the use of certain accounting risk measures from the first period (1947— 1956) to form an estimate of the systematic market risk to be observed in the second period (1957-1965). These estimates were contrasted against the naive systematic risk forecast, where the systematic risk level in the first period is used without modification as the forecast of the systematic risk in the second period. Multiple Stepwise regression was used to select the accounting variables used in making the forecast of the systematic risk to be observed in the second period. Multicollinearity was a concern, and ultimately only three accounting variables were selected for use in making the forecast of systematic risk: payout ratio, growth, and earnings variability. Note that neither financial leverage nor the accounting beta were used in generating the forecasts of the systematic risk level to be observed in the second period. The BKS forecast of beta through accounting risk measures slightly outperformed the naive 24 forecast of beta in explaining variation between the levels of systematic risk observed. BKS suggested two explanations of their results: (1) investors do use the accounting risk measures in making their decision, and (2) the accounting risk measures reflect some of the same underlying processes which influence the behaviors of investors. One cannot conclude that the accounting risk measures are a causal force in this relationship. The scant improvement obtained through the use of the accounting risk measures over the use of a naive forecast of ex ante systematic risk does not serve as impressive evidence that these accounting measures are causal factors. Moreover, the authors made no analytic argument for there being such a relationship. Only for the accounting beta and financial leverage can one develop a plausible argument that these accounting risk measures could be a vital force in the relationship. If there is any other accounting variable which is correlated with the accounting income measure, then this other variable is a possible source of confounding. At most, this other accounting variable could be the force driving the systematic risk generation process. Cash flows to firms should be correlated with the accounting income measure. Accountants would suspect that their income measure is the driving force in this relationship. We must remember that the accounting income measure is the result of applying a myriad of accounting rules. Accounting income is an empty economic concept except that it may, under certain unusual circumstances, represent an approximation to economic income. Income measured in cash flows may provide a better 25 understanding of the risk-return relation than does the measurement of income via the accounting income concept. BKS could have included a cash flow risk measure in their correlational studies. Inclusion of that accounting variable might have contributed to our understanding of the role of net operating cash flows to the firm in the determination of the systematic risk exhibited by the equity securities of the firm. Covariant Forms of Accounting Variables and Systematic Risk Thompson conducted another test of contemporaneous association between the various forms of accounting variables and the ex post level of systematic risk.22 The principal contribution of this study was his emphasis on the form of the accounting variables themselves. Specifically, he investigated the relation between the mean, trend, covariance, and variance forms of the accounting variables selected relative to the observed systematic risk level. Thompson developed a model which suggests that covariance forms of some accounting variables may exhibit a higher association with the systematic risk level observed ex post. The model is presented below with time subscripts deleted: cov [(di + e1 + k1), (dm + an + km)] Eq. 7 B = 2 ~ ~ ~ ’ o (d + e + k ) m m m where d, e, and k represent dividends, accounting earnings, and earnings multiple measures, respectively. Thompson also used the separate covariance/variance components in combination to form dividend betas, earning betas, and earnings multiple betas. These betas were also used in the correlation results he reported. Thompson conducted the 26 tests on individual firm data, and on data from portfolios which had been formed to minimize the impact of measurement errors in the accounting variables selected.23 Thompson identified 290 firms which had complete data available for at least one of the two testing periods, 1951-1959 and 1960-1968. To be included, the firm had to have complete data available for construction of the accounting variable being correlated with the systematic risk level. The composition of the samples investigated differed for each accounting variable examined and for each of the two testing periods for these reasons. The actual sample sizes ranged from 193 to 282 firms, depending upon which variable and which testing period was examined. The correlation coefficients at the individual security level were generally statistically different from zero. However, the sizes of the correlation coefficients were not economically impressive. Thompson concluded that the covariant forms of the explanatory variables were more closely correlated to the levels of systematic risk than were the mean and variance forms during the 1951-1959 period. He did not conclude that the covariant forms were more closely correlated with systematic risk during the second testing period 1960-1968. The correlation coefficients at the portfolio level were nearly all statistically different from zero. More important, the size of most of the coefficients was impressively large, with many of them being higher than .70. An accounting risk measure was investigated at the portfolio level only if at the individual security level it hadthigh correlation with the systematic risk leve1,had the correct sign, had consistent implications for both testing periods, 27 and had been found to be correlated in previous analytic or empirical work. Perhaps the high degree of correlation observed at the portfolio level can be more easily understood because of these rigorous selection criteria. Tests at the portfolio level were run only on data from 1960-1968 because of the need to use data from the earlier period to rank the firms for portfolio formation. Thompson then ran multiple regressions using the firm's systematic risk as the independent variable. The independent variables generally were the accounting risk measures which had been identified as being highly associated with the systematic risk level when observed at the portfolio level. Thompson's objective was to see which combinations of accounting risk variables seemed to be better explainers of the observed level of systematic risk. When he used the covariant forms for dividends, earnings, and earnings multiples, he obtained an r2 of .846 (standard error of .063). In the regressions using the mean dividend payout, earnings variance, and the earnings multiple variance, he obtained an r2 of .658 (standard error of .093). Based on these results, he concluded that the use of the covariant forms did add noticeably to the ability of these variables to explain differences in levels of systematic risk. Thompson appears to have empirically justified the use of the covariant forms of the variables which are being explored as possible determinants of systematic risk. His results were impressive. Moreover, they are encouraging for the use of covariant forms of other accounting variables. This is especially true if an analytic framework exists or can be developed defending the use of a particular accounting 28 variable in some form. In order for significant empirical results to be compelling, there must be an analytic theory explaining why the results should be as they are. In the absence of such an analytic theory, the significant findings seem to be the result of chance. Corporate Decision Variables and Systematic Risk Bildersee24 demonstrated that a higher adjusted r2 in a multiple regression could be obtained if corporate decision variables were added to the set of independent variables commonly used to explain the systematic risk exhibited by common equity shares. Bildersee interpreted his improved explanatory power to imply that the information added to the multiple regressions by the decision variables must be non-overlapping with the information contained in the accounting variables. His conclusion appears plausible. The sample was limited to those firms having both traded common and traded non-convertible preferred shares. Ninety-eight such firms were included in this study. Bildersee looked at the dividend decisions on the firm's common and preferred shares. From these dividend decisions he structured dummy variables. These dummy variables were used in conjunction with accounting variables traditionally used in systematic risk association studies. One decision variable did not deal with dividend policy. He called it the diversification variable and it reflected changes in the firm's SIC code in comparison with the firm's Compustat industry number. Bildersee reported an adjusted r2 of .598 for multiple regressions of systematic risk onto independent variables comprised of accounting 29 risk measures and decision variables. These results were obtained at the individual security level. No results were reported from regressions of the same variables measured at the portfolio level. Fama and Miller25 had previously argued that the observed systematic risk level should be a summary measure of the impact of the production, investment, and financing decisions made at the firm level relative to those made by firms comprising the market index. Building directly upon that idea, Bildersee demonstrated that decision variables (largely related to dividend policy) can be added as independent variables to multiple regression analysis in order to improve the explanatory power of the association. The significance of Bildersee's work for the present research is its emphasis that the scope of the variables which can be legitimately considered for inclusion in a systematic risk association study is much broader than merely looking at accounting variables. Empirical Relation between Ex Post Risk and Market Returns Fama and MacBeth26 developed procedures for testing the strength of the association between ex post beta (bi) and the market returns to the related equity security. Their work supported the CAPM as an. adequate explainer of the risk—return relation. The details of the testing procedure used in Chapter Five of the present research are essentially those developed by Fama and MacBeth. The hypotheses tested by Fama and MacBeth were: 1. Is the CAPM properly specified when it contains no~ term~reflecting a non-linear relationship between R and Rm-Rf? Specifically, they tested whether the relationship was non-linear in bi by using a biz term as an independent variable in addition to a bi term. 30 2. Are there systematic effects from non-B risk? Specifically, does the average residual term from the bi estimating regressions prove to be useful in explaining observed market returns to security i? Fama and MacBeth concluded they could not reject the testable hypotheses with respect to the two-parameter model when using ex post betas as the estimate of 81' The results of their study have established the legitimacy of these testing procedures in the attempt to more fully understand the risk-return relation. Given the results of their study, it is an empirical question whether other Specifications of the systematic risk parameter can be used within their testing procedures to achieve the same relationship between systematic risk and market returns to securities. Role of the Present Study All three papers in the first area empirically or analytically support the theory that financial leverage is a determinant of systematic risk. Myers concluded that cyclicality is also a determinant of systematic risk and is subject to major measurement difficulties. The research reported in this dissertation measures cyclicality through the estimation of an operating net cash flow beta. The analytic model develOped later explicitly integrates various levels of financial leverage into the systematic risk measure which is constructed for each firm and each portfolio. This is an innovative empirical contribution of the present research. The papers in the second area deal with operating leverage as a determinant of systematic risk. Wolfson incorporated both financial leverage and operating leverage. Operating leverage may seem closely related to the systematic operating cash flow risk measure which will 31 be develOped later. For that reason, understanding the differences between these measures becomes important. These measures may be related, however. Operating leverage expresses a relation between the fixed operating expenses and variable Operating expenses. These expenses may be cash or non-cash. Depreciation, for example, would be a non-cash operating expense which typically is fixed in nature. Systematic cash flow risk as used in the present research expresses a relation between the net operating cash inflows to the firm relative to the net operating cash inflows to the group of firms which comprise the market index. The operating leverage is a relation between two kinds of intra—firm expenses. The operating cash flow risk is the systematic net operating cash inflow relation between the firm and equivalently measured market index. The relationships express greatly different firm-specific characteristics. If there is a correlation between the net operating cash flow to the market and the number of units produced by a single firm, then it would be reasonable to expect the two risk measures described above to be related. Alternatively, if there is a correlation between the total costs of a single firm and the net operating cash inflows to the market, it could be expected there would be a relation- ship between the two risk measures. No relationships between operating leverage and Operating cash flow risk need exist for purposes of the current research. This research focuses on the hypothesized relationship between an alternative specification of the systematic risk measure and the returns to securities. This relationship entirely avoids the necessity of a 32 link between systematic net operating cash flow risk and operating leverage. Papers in the third area represent efforts to express the correlation between or the predictive relation of two or more risk measures. These papers examined the extent to which certain accounting or corporate decision variables could be found to be determinants of systematic risk. Significant correlational results were generally interpreted to indicate a contemporaneous relation between the accounting and decision risk measures being examined and the surrogate for the true level of systematic risk. In the predictive ability studies, a significant relation between the accounting (or decision) variables and the systematic risk level observed in the subsequent period would have been thought to confirm the role of these variables as determinants of systematic risk. There was relatively little success reported in these predictive ability tests. One significant potential contribution of the present study will be to directly observe the relation between the new systematic risk measure and the observed market returns. Earlier studies relied on correlations between risk measures in generating test results. This direct test is not as sensitive to non-stationarity in the market-based systematic risk measures as were the three studies included in the third area. If the true systematic risk level changes, the observed returns ought to reflect the new level of systematic risk, the market being assumed efficient in its assessment of the ex ante systematic risk levels. Non-stationarity in the accounting-based systematic risk level remains an issue, however. All previous 33 association or prediction studies had two sources of non- stationarity to contend with: (1) non-stationarity in the accounting (or decision) variables' systematic risk measure, and (2) non- stationarity in the market-based systematic risk measure. Testing the strength of the risk-return association directly, rather than measuring the associations between the two risk measures, appears to make sense. If an alternative risk measure can be developed which alone or in conjunction with the more traditional risk measure can be more closely related to observed returns, then this alternative risk measure would add to our understanding of the risk-return relationship. Indirectly, at least, the new measure would be a determinant of systematic risk. FOOTNOTES - CHAPTER TWO 1Franco Modigliani and Merton H. Miller, "The Cost of Capital, Corporation Finance, and the Theory of Investment," American Economic Review (June 1958): 261-270. 2R. A. Haugen and J. L. Pappas, "Equilibrium in the Pricing of Capital Assets, Risk-Bearing Debt Instruments, and the Question of Optimal Capital Structure," Journal of Financial and Quantitative Analysis VI (June 1971): 943-953. 3R. Hamada, "The Effect of the Firm's Capital Structure on the Systematic Risk of Common Stocks," Journal of Finance 27 (May 1972): 435-452. 4S. C. Myers, "The Relation between Real and Financial Measures of Risk and Return," in Risk and Return in Finance, ed. Irwin Friend and James L. Bicksler (Cambridge, MassaChusetts: Ballinger Publishing, 1977), pp. 49-100. 5B. Lev, "On the Association between Operating Leverage and Risk," Journal of Financial and Quantitative Analysis IX (September 1974): 627. 6Mark A. Wolfson, "Toward the Understanding of the Complementary Nature of Security Price and Non-Security Price Information in Relative Risk Parameter Estimation," (unpublished Ph.D. dissertation, University of Texas at Austin: 1977). 7William Beaver, Paul Kettler, and Myron Scholes, "The Association between Market Determined and Accounting Determined Risk Measures," Accounting Review (October 1970): 654. 8Donald J. Thompson II, "Sources of Systematic Risk in Common Stocks," Journal of Business (April 1976): 173. 9John S. Bildersee, "The Association between a Market-Determined Measure of Risk and Alternative Measures of Risk," Accounting Review (January 1975): 81-98. 10Eugene F. Fama and James D. MacBeth, "Risk, Return, and Equilibrium: Empirical Tests," Journal of Political Economy 18 (May- June 1973): 607-636. 34 35 11Myers, "The Relation between Real and Financial Measures of Risk and Return," pp. 49-100. 12Modligiani and Miller, "The Cost of Capital," p. 270. 3Haugen and Pappas, "Equilibrium in the Pricing of Assets," l['Hamada, "The Effect of the Firm's Capital Structure," p. 435. 5Myers, "The Relation between Real and Financial Measures of Risk and Return," p. 64. 16Ibid., p. 76. l7Lev, "0n the Association between Operating Leverage and Risk," p. 627. 18 Operating leverage is the ratio of the relative change in income for a relative change in the units of production. It essentially expresses the relation between the firm's fixed and variable expenses. 19Fama and Miller, The Theory of Finance, Chapter 7. 20Wolfson, "Toward the Understanding of the Complementary Nature of Security Price and Non-Security Price Information." 21Beaver, Kettler, and Scholes, "The Association between Market Determined and Accounting Determined Risk Measures," pp. 654—682. 2Thompson, "Sources of Systematic Risk in Common Stocks," pp. 173-188. 23See Chapter Five of the present research for the details of this common approach. 24Bildersee, "The Association between a Market-Determined Measure of Risk and Alternative Measures of Risk," pp. 81-98. 25Fama and Miller, The Theory of Finance, Chapter 7. 26Fama and MacBeth, "Risk, Return, and Equilibrium," pp. 607-636. 36 CHAPTER THREE MODEL DEVELOPMENT Accountants have, or should have, a particular interest in the role played by accounting data within the macro-information set. This macro-information set is the aggregate of all information available with or without cost to economic decision-makers, including investors. A relevant parameter of interest to this group has been shown to be the ex ante systematic risk measure of securities in the marketplace. Some argue that the ex ante systematic risk measure is the only relevant parameter of interest in a CAPM world. Beaver and Manegoldl (1975) believe this systematic risk parameter to be relevant if it is merely one of the measures of security risk. Various accounting risk measures have been shown frequently to be contemporaneously related to the ex post beta measure. As has been mentioned above, the link between accounting data and ex ante beta becomes tenuous because of the important explicit role played by investor expectations in the ex ante beta measure. To the extent there are accounting variables correlated with accounting income, accountants cannot feel comfortable that accounting income EE£.§S is the accounting variable of interest in this contemporaneous relation— ship. Accrual income has often been described as a good predictor or measure of recurring cash flows. An intuitive justification of the relevance of the firm's cash flows to investor decision-making 37 typically followed. No study has been reported where cash flow cyclicality measures have been explored as a contemporaneous explainer of ex post systematic risk. Empirically testing the hypothesized relationship between cash flows to the firm and ex post market returns on the security would be a natural first step. This must be limited to current cash flow measures since cash flow expectations are not available for empirical analysis. The objective of this research is to establish empirically the relationship, if any, between current cash flow risk measures and returns on the related securities. The broader goal of this type of research is to establish the relationship, if any, between current accounting data and ex ante beta. If measuring the ex ante beta were not an intractible task at the present time, the broader goal would already have been explored at length. Lenders comprise one subset of the investor group. The concept of systematic cash flow risk is especially relevant to these users because of their emphasis on repayment prospects. The concept of systematic cash flow risk is also relevant to equity investors to the extent a link can be established between cash flow risk and equity returns. Haley and Schall2 support the essence of the preceding statements when they state that the firm's value (the value of debt plus the value of the equity) is a function of the firm's output of money and the firm's investment policies. Their comments reflect the expectation of future outflows of money from the firm to both Classes of investors, either directly or indirectly. In 1958, Modigliani and Miller3 set forth two now famous propositions. MM evaluated the interrelationships of security 38 valuation, financial leverage and the cost of capital. They made three noteworthy assumptions: 1. The stream of profits to a firm has a known and finite mean and observations on this stream are randomly distributed subject to a probability distribution. Within homogeneous risk classes, the stream of profits is assumed to be capitalized at aconstant rate (pk). This represents the required rate of return to the firm. pk reflects only the operating risk characteristics of the firm and none of the financial risks of the firm. Debt securities are issued to yield r -- the capitali— zation rate for sure and constant streams. Thus, this debt is riskless in nature. The propositions set forth by MM are paraphrased below. After each proposition, explanations are made to facilitate an intuitive grasp of their significance in relation to later work by Haugen and Pappas4 Eq. 8 Eq. 9 in 1971. Proposition I: firm ‘1: vequity+vdebt, ll <2 0) H C (D where V >_ o in Eq. 45 through Eq. 48 above. The t, are the means of the regression coefficients averaged across the 20 portfolios for each of the models (Eq. 45 through Eq. 48). These regression coefficients are originally calculated for each port- folio separately. The s(;i) terms are the averages of the portfolio's component firms' residuals from the b2: estimating regressions. The (bgFt-l)2 terms are the averages of the portfolio's component firms' 9 big which have been squared separately. Thus, the (bgFt_l)2 term is 9 mislabeled in a strict sense. Tests of the three research hypotheses answer the following questions: 1. Is there a linear relation between systematic cash flow risk and security returns? Or can some non-linear relation between these variables be discovered? 2. To what extent, if any, does the residual risk from the b2: estimating regressions explain the returns on securities? 56 Alternatively stated, does the variability of the error III terms [32(egt)] exhibit some relation to Rii? $F 3. Is the average multiple regression coefficient on bp t-l 9 (V2) statistically greater than zero? If the general CAPM relation between market returns on securities and systematic operating cash flow risk is found to be significant, there would be support for the hypothesis that systematic net operating cash flows are a surrogate for the firm's cyclicality measure. Testing Procedures Tests of the research hypotheses involve the means of Y2’ Y3, and Y4- Many of the testing procedures used here were developed in Fama and MacBeth4 to test the relation between risk and return in the CAPM mr where 81 was estimated from the return series Rit various BiF estimates formed as described earlier, the relationship and Rm. With the mt between market returns to portfolios and these various risk measures has been assessed. In order to determine the extent to which non- systematic risk or the non-linear systematic risk measure is related to observed market returns, V3 and J, [in equations 45, 46, and 47 as indicated] have been evaluated to determine if those coefficients are statistically different from zero. The approach here has been to form portfolios using the firms included in the study in order to yield relatively homogeneous biFand bit within each portfolio. Portfolios were formed to allow for the offsetting of measurement errors in the estimating of 81 from the return series data. These measurement errors were assumed to be random. Estimates of Bi have been found to be quite volatile over short measurement periods.5 57 If measurement errors occur in biF as they do in bmr’ the formation of i portfolios should have permitted a reduction in the standard error of the estimate of the systematic risk measure to be used in the CAPM. This volatility in Bi estimates has been assumed here and elsewhere to result from measurement errors while the underlying true beta was relatively more stationary over longer periods. Thus, in the formation of equally weighted portfolios random measurement errors were assumed to offset ll each other. These measurement errors were assumed to be uncorrelated I with B or with Rmr. ; 1 mt L By assigning securities to portfolios based upon a ranking of ( b?r, the intent was to form portfolios of relatively homogeneous biF and bmr within those portfolios. This would occur if there were a 1 high correlation between biF and bEr. The primary objective in this assignment to portfolios was to test the risk-return relation across a wide span of systematic risk levels. By maximizing the diversity SF between the various bp and bgr across the portfolios, a clearer understanding of the relation of the proposed systematic risk estimator to observed market returns on securities can be obtained. $F p used to rank the securities and subsequently form portfolios, the bSF If portfolio betas (b and bgr) are estimated from the same data mr and bp are likely to be biased toward 1.0 in the testing periods which follow. This phenomenon is called regression-toward-the-mean and has long been recognized as a potential problem in portfolio formation situations (Blume;6 Fama and MacBeth;7 and Foster8). The regression phenomenon originates when the B1 are measured with error and the firms mr are ranked according to the observed b . As indicated above, b$F (bmr) i p p were formed to permit random measurement errors to have minimal impact 58 mr i to form portfolios, on the results. However, because of ranking on b mr there is reason to suspect the measurement errors in high and low bi mr firms are not random. High bi firms probably contain non-random measurement errors biasing that measurement away from 1.0 relative to the true 81’ and vice versa for low bin. firms. Consequently, estimates of 8p would be formed reflecting these non-random measurment errors. $ Therefore, the b:r (pr) would be expected to be less extreme in the subsequent testing period when these non-random measurement errors have an expected value of zero. In summary, a mechanism must be used to cause these non-random measurement errors to have minimal $F $F P P CAPM during the testing periods. impact on the b (bgr) before using the b (bgr) estimates in the The common mechanism employed for this purpose is to use data from one period to form estimates of 81 to permit a ranking of these firms. This ranking provides a basis for portfolio formation, but not $ for the purpose of pr (bgr) estimation. Fresh data from a subsequent :F (bgr) and these values are used to estimate 8p, where the biF (b?r) have been equally weighted within period are used to form b their respective portfolios. The cash flow data used to form biF estimates are available on a quarterly basis. Portfolios were formed initially from firms ranked on b?r which have been computed from monthly market return series in the July 1965-June 1970 period. Quarterly Compustat data from July $F p O procedures for obtaining b:F were explained in Chapter Four. 1970-June 1975 were used to initially compute b Details of these The bill. estimates were generated from the monthly returns on the CRSP tapes. The firm had to be continuously listed from July 1965 59 through December 1978 on the CRSP tapes in order to be included in the sample of firms. Firms were also required to have complete Compustat data in the time period July 1970 through December 1978 to be included in the sample. The ranking, estimation, and testing periods are shown and described below. Initial Initial Initial Ranking Estimation Testing Periods Periods Periods July 1965-June 1970 July 1970-June 1975 July 1975-December 1978 Sir estimation BEF estimation Test Period Sgt estimation er estimation Test Period mr SF The estimates of bi,t+l (bi,t+l) were updated quarterly by dropping $F the earliest data entering into the estimation of bmr 03 it it ) and adding the next quarter's data in chronological sequence. Thus, the ranking $F i . t+j mr i,t+j 0’ periods for b )were all 20 quarters in length. This updating process was also used to obtain new b$F and bmr estimates which i,t+j i,t+j SF and bmr estimates as described earlier on p.t+j p.t+j $F mr page 50. These market return ranking procedures yield bp,t+j and p,t+j $F mr i,t+j and bi,t+j which have enter into the b estimated from their component firms' b been previously ranked using only market return data from the five-year period ending at time t+j-20. In addition to the ranking, estimation and testing procedures SF described above, portfolios were formed from a ranking on bi in order SE to assure that the conclusions formed about the relation between b1 and RTr are not adversely affected by a low correlation between bi and b?r in the initial estimation period. Recall that, in assigning 60 firms to portfolios, the rankings had been on bgr for both the cash flow portfolios and the market return portfolios. This procedure permits relatively longer testing periods but at the risk of not attaining the desirable degree of homogeneity of biF within portfolios. SF 1 9 This additional By additionally forming portfolios from rankings on b a check was mr i O permitted on the earlier decision to rank on b information was obtained at the expense of shortening the ranking and estimation periods to two and one-half years each (ten quarters) to reflect the shorter time period over which Compustat data was available. This reserved three and one-half years of data for testing. This testing period was the same length as that which had been used during the initial ranking for portfolio information on the basis of market return data. $F These tests on bi rankings have ranking, estimation, and testing periods as follows: Initial Initial Initial Ranking Estimation Testing Periods Periods Periods July 1970-December 1972 January l973-June 1975 July 1975-December 1978 The ranking on Compustat data yielded bgFt+j estimated from 9 bSF where these b$F have been previously ranked using only cash i,t+j i,t+j flow data from the ten-quarter period ending at time t+j-10. $F mr These bp,t+j and bp,t+j Obtained from both initial ranking procedures were used in equations 45 through 48 as indicated. Monthly market returns for the portfolio were used to provide three test $F b erv tion for ea h b . o s a s c p,t+j The test period was 14 quarters in length with each quarter yielding three monthly test observations. 61 The regressions described in equations 45 through 48 were run for each mr SF and b estimates. ortf 11 for all 17 of the b p ° ° p.t+j p.t+j Then,via t-statistics,the significance of V3, V4, and Y2 were assessed. These t-statistics were computed as follows (illustrated only for research hypothesis 1): Y3 - E(Y3) , su’p N—T-I Eq. 49 t(y3) = where T is the number of test observations (42) in each series. The statistical significance of the research hypotheses was assessed where the alpha (0:) level was 5 percent for all three research hypotheses. Rejection regions for Research Hypotheses 1 and 2 are two-tailed. The critical value indicating statistical significance for these two hypotheses is f 2.000. Research Hypothesis 3 has a one-tailed rejection region and consequently has a critical value of + 1.671. FOOTNOTES - CHAPTER FIVE 1Fama and MacBeth, "Risk, Return, and Equilibrium," pp. 607-636. 2Beaver, Kettler, and Scholes, "The Association between Market Determined and Accounting Determined Risk Measures," pp. 654-682. 3Fama and MacBeth, "Risk, Return, and Equilibrium," p. 622. 4 . Ibid., pp. 607-636. 5Nancy Jacob, "The Measurement of Systematic Risk for Securities and Portfolios: Some Empirical Results," Journal of Financial and Quantitative Analysis (March 1971): 815-834. 6M. E. Blume, "Portfolio Theory: A Step toward Its Practical Applications," Journal of Business 43 (April 1970): 152-173. 7Fama and MacBeth, "Risk, Return, and Equilibrium," p. 615. 8George Foster, "Asset Pricing Models: Further Tests," Journal of Financial and Quantitative Analysis (March 1978): 39-53. 62 CHAPTER SIX TEST RESULTS Rejection of HO would indicate that a statistical relationship 3 does exist between BEF and R?r analogous to what has been demonstrated elsewhere for the relation between BEr and R2r. The evidence presented here generally suggested that the hypothesized relation between the systematic operating cash flow risk of firms and market returns to securities of those firms had not been captured by the model or the testing procedures used here. Rejection of H01 would have permitted the conclusion that the non-linear relation between systematic operating cash flow risk and market returns was not statistically significant. Similarly, rejection of H02 would have permitted the conclusion that specific SF risk, as measured by the size of residuals in the Bei estimating regressions, was not significantly related to the market returns of the securities of those firms. The evidence presented here made it possible to conclude that non-linear systematic operating cash flow risk was not statistically related to the market returns of those securities. The evidence did suggest that the average $F 1 size of the residuals in the Be estimating regressions was positively related to the market returns of those firms' securities. 63 64 The tables of statistical results presented below include summary statistics as follows: the regression coefficients for each term in the regression equations 45 through 48 averaged across the 20 portfolios for market and cash flow data. 1. §3= 2. s(;:): the standard deviation of each of Ehose average 3 regression coefficients about the Yj' 3. The t-score of each of the average regression coefficients where there are 41 degrees of freedom. The formula is: _, Y “11) = emf—7F” - 4. The r2 of each regression equation has been averaged across the 20 portfolios and is presented along with the standard deviation of this r2. Tables 1 through 4 contain statistical results for the tests included here. The table number represents the number of the cash flow definition whose results are being presented (see page 46). Panels A through D in each table contain the results for those port- folios ranked initially on bgr, while Panels E through H in each table contain the results for the portfolios initially ranked on biF. $F Panels A and E contain the results for those portfolios whose bi SF SF it and Rmt [see (a) on page 48]. Panels B and F contain the results for portfolios whose biF was i but using unseasonally adjusted were estimated from unseasonally adjusted R estimated from seasonally adjusted R$ i R3: [see (b) on page 48]. Panels (3 and G contain the results for SF 1 series of both RE: and R3: [see (c) on page 48]. Panels D and H SF 1 portfolios whose b were estimated from the seasonally adjusted contain the results for portfolios whose b were estimated from the $F it of R3: [see (d) on page 48]. Table 5 presents the results of a unseasonally adjusted series of R but the seasonally adjusted series 'Hd‘ _.7RAR—? .4374-2 -.3599~7 .4102-7 .2219-2 .5599-2 .3426-3 .5715-2 -.3265-2 .2462-2 -.2l69-2 .5362-2 i2. -.1166-2 -.4973-3 -.2866-3 .1038-2 12 -.5345-2 -.4995-2 -.lB92-2 -.I34l-2 I; .74l9-2 .4868-2 -.23|9-2 -.6402-2 .8895-2 .2092-2 °.9977-3 -.557I-2 I *‘ U .1711-4 .2486-4 4| .4881-3 0‘988‘3 :2 .1636-2 .3900-2 12 .1034-l .5753-2 (55 'TAUBIJE 1. IN ESTIMATING SYSTEMATIC RISK 4| .9 .6373 .6960 4| .- .3847 .541! I: .4333 .4487 .5246 8(Vl7 0 31 3"! 03627-1 .3103-1 .3440-1 0(71) 62562-1 03628-1 .2013-1 .3832’1 .3507-1 .2745-1 .4043-1 .2879-1 .3570-l .2932-1 03990-l STATISTICAL RESULTS USING CASH FLOW DEFINITION l Panel A 15:31 .(935 .(5,) 617,) :(72) ct?,) :(;,) .1366-1 .7503—3 3.0219 -.545 -.661 .166 1.350 .1631-1 .7574-3 .017 -.195 .209 .92:6-2 2.9029 -.763 -.200 1.50: .1266-1 .766 .525 Panel B 0(72) -(?3) 0t?,) ct?!) t(72) t(?3) t(7‘) .1347-1 .1000-2 3.1368;' .696 -.256 1.663 .705 .1369-1 .2530-2 .900 -2.372 1.259 .9615-2 3.1732 .070 -r.260 1.092 .1015-1 .955 -.066 Panel C 669,) 15321 -(?,) t(71) t(§}) tth) ttfl) .7601-6 .3376-1 3.0626 -.271 .625 1.660 .912 .7760-1 .5438-1 .666 .602 .659 .2816-1 2.9770 .001 -.527 .965 .3102-1 .063 -1.200 Panel D arvz) 11121 -(§;) :(t,) c(?,) c(§,) :07.) .6760-1 3657-1 2.6506 -.726 .063 1.011 1.155 .7501-1 5942-1 .662 .177 .620 .2814-1 2.7150 -.474 -.227 1.237 .2866-1 .061 -I.245 2 r .3559 .2030 .2723 .1157 2 r .3616 .1856 .2770 .0604 2 r .3350 .1013 .2627 .0541 2 r .3241 .1831 .2535 .0502 8(r2) .2177 .1488 .2249 .1217 I(r2) .2105 .1220 .2239 .0710 0(r2) .2165 .1434 .0728 0(r2) .2174 .1433 .2202 60700 (I .6336-2 ,6535-2 ,3709-2 _5939‘2 l «I .6007-2 ,7101-2 .6599-2 71 o4534-2 .7163-2 .6560-2 .6527-2 21 ,5o|5-2 .7273-2 .4817-2 .6700—2 i; -.1526-2 -.6155-3 .1611-3 _5460-3 2 -.1318-3 -.1601-3 -.9580-3 -.1370-2 Y2 i3802§2 -.4632-2 -.386602 -.1105-2 .2225-5 -.3035-2 .1809-2 -.1207-2 -.2944-2 -.7177-3 .4673-3 -.4274-3 .1002-3 _7123—6 .1017-2 -.3012-3 (56 TABLE 1 (Continued) .4 U1 .1700 .3706-1 ,3931-1 .2245 .3744-1 .3932-1 .2 :5. 0(7!) .0792 .3541-1 .3911-1 -.0936 .3489-1 64073-1 .2 <0 .9 -3569 .3282-1 .3805-1 -.0286 .3389-1 .3958-1 0(1.) .3166-1 .3791-1 .1465 .3301-1 .4056-1 Panel E «$21 «173) .6.) a?!) «$21 :63) 15‘: 5: .9903—2 .6692-3 1.7897 .749 -.986 .959 .615 .2765 ,1177-1 ,6166-3 1,066 -,335 .740 .2136 .6680—2 1.7561 .634 -.154 .010 .2070 .9197-2 .967 .380 .1365 Panel F 607,) £31 61?.) 607,) :62) 667,) :07.) :2 7.1003-1 .1909-2 2.9706 1.086 -.602 -.969 .170 .26f5 .1113-1 .1269-2 1.131 -.076 -.808 .1369 .5825-2 1.7903 1.303 -1.053 -.335 .1945 .6783-2 1.037 -1.296 .0722 Panel C «52) 12 197.1 «9,; :62) «73116.7 _r: o4288-1 o1504-1 2-6767 -884 ~568 -1-972 ~35‘ ~2527 .4668-1 .1141-1 1.202 -.530 .620 ’ .1700 .2944-1 1.961 1.241 .000 -.094 .1985 .3954-1 1.056 -.492 .1008 Panel B .792) :12 um N71) c1721 c1173) :6") i .4708-1 .1336-1 1.7575 1.014 .246 -.578 .785 .2412 .4835-1 .1303-1 1.229 -.390 -.353 .1471 .2333-1 1.7125 .934 .128 .548 .1665 .2418-1 1.058 -.113 .0677 .(r2) .1562 .1750 .1471 0(r2) .1004 .1430 .1701 .1273 0(r2) .1315 .1201 .0805 71 -.2765-2 .4662-2 -.3622—2 .4434-2 '11 —_ .1802-2 .5203-2 .2750-3 .5590-2 <1 .2716-3 .3051-2 .8679-3 .5200-2 1’1 -.2014—2 .2764-2 °.1672-2 .5245-2 (77 IMXBHLEZ 2 STATISTICAL RESULTS USING CASH FLOW DEFINITION 2 IN ESTIMATING SYSTEMATIC RISK 72 73 -.1337-2 .1936—6 -.9215-3 .3675-6 -.3663-3 .5612-3 :23; -.3951-2 22421-3 -.359o-2 .2700-3 -.1501-2 -.1170-2 32.33 .6905—3 .6111-2 .1267-2 .1965-2 -.3066-2 —.5301-2 3213 .3536-2 .5000-2 .6000-3 .2061—2 -.2636-2 -.5130-2 .6855 .7383 T6 0(9‘) .3007-1 .3511-1 .3504-1 C(71) .4180 .2824-1 0 3580“ .5623 .2798-1 41 a- .3562 .4314 41 t .4406 .5089 I(f1) 0 3553-1 0 2766-1 C(71) 0 2933-1 0 3525-1 .2924-1 .3963-1 Panel A .1172) 3:2: .6.) 007,) 607,) :63) :07.) _r: on”) .um91 .6u00 30n0 n5n -JM9 .1» L4" .un7.2u2 .1163—1 .6266-3 .069 .516 .555 .2090 .1527 .7065-2 3.0100 -.753 -.332 1.566 .2750 .2276 .9371-2 .0103 .3035 .1152 .1201 Panel B «172) -_G_,)_ n6.) 107,) 6072) :63) 0G,.) :2 0(r2) .1050-1 .1123-2 3.2161 '.609 -2.609 1.301 .033 .3712 .2111 .1032-1 .1666-2 .931 -2.227 1.216 .1962 .1299 .7607-2 3.2377 .063 -1.203 1.112 .2021 .2263 .7u24 .9n -3M6 .0u0 Ann Panel C 0(72) 6193) .07.) 06-") 6072) 6G,) -(7‘) i: .(z’) .5605-1 .1673-1 3.1692 .062 .079 1.576 .726 .3616 .2173 .5507-1 .2510-1 .550 .165 .500 ’ .1069 .1699 .2300—1 ' 3.0662 .201 -.065 .907 .2696 .2153 .2270-1 .030 -1.490 .0661 .0002 Panel D 2 . 6072) 0(73) «7‘5 007,) 117,) «73) :07.) 3.. -(r‘_7_ .5171-1 .1750-1 2.7369 -.660 .630 1.056 1.031 .3292 .2102 .5533-1 .2692-1 .502 .079 .010 .1009 .1512 .2u04 2J6” -¢n2 new 1.n7 Jun .nus .2160-1 .060 -1.521 .0603 .0772 71 .7964-2 .6781-2 .7570-2 .6142-2 41 .6257-2 .6478-2 .6264-2 .6439-2 11 .5029-2 .6704-2 .6617-2 .6329-2 :1 .4249-2 .6856-2 .4697-2 .6484-2 Y2 .1175-4 .4941-3 .6038-3 .2675-3 '2 .1613—2 -.7672-5 .6056-3' .9061-3 .1058-2 y2 .4632-2 -.2771-2 -.2756-3 .2776-2 .5643-3 1’2 .5233—2 -.8231-3 .9648-3 -.4347-3 .1245-2 -.3250-3 _3 12 .4414-4 -.l430 .3514—4 .3. L9 .0156 .2607-5 -.OO69 _3 19. .3425 o 8000'3 -.3788 .11 1’3. .2773 .1562 623 TABLE 2 (Continued) .3475-1 .3421-1 .3971-1 0 3659-1 .3973-1 0 3‘68-1 .4075-1 .3379-1 .3482 {(711 .3358-1 .3799-1 .3444-1 Panel B 0(92) :332: .0969-2 .6356-3 .9500-2 .6260-3 .5025—2 .6970-2 Panel F 0(72) :1221 .9007-2 .0266-3 .9160-2 .6636-3 .5296-2 .5030-2 Panel C 0(72) :3321- .2214-1 .0053-2 .2792-1 .5676 .1239-1 .2672-1 Panel H 01721 0193) .2960-1 .7606-2 .3551-1 '.0799-2 .1701-1 .2307-1 0(7‘) 107,) :62) H73) 01?.) :2. 1.9945 1.468 .008 .649 -.459 .2869 1.104 -.330 .531 .2084 1.9213 1.408 .664 -.402 .2039 .990 .246 .1198 15,; :6.) 0672) :63) H7") r: 2.7120 1.095 -1.136 -.057 .037 .2589 1.044 .339 .025 .1219 1.8275 1.159 -1.192 ’.024 .1941 1.012 -1.161 .0681 am) a?!) :62) «9,1 1'5.) _r: 2.7100 .953 1.340 -2.004 .809 .2809 1.104 -.052 .009 ’ .1664 1.9415 1.216 1.434 '1.249 .2125 1.008 .146 .0902 615.) «9,7 0072) 617,) 26?.) 2’ 1.6335 i.010 1.136 —.606 1.007 .2600 1.156 .224 -.237 .1588 1.616 .073 .367 .620 .1690 1.026 -1.207 .0811 0(r2) .1708 .1658 .1611 .(r’) .1962 .1220 .1757 .0873 0(02) .1462 .1167 .1375 .0969 69 TABLE 3 STATISTICAL RESULTS USING CASH FLOW DEFINITION 3 IN ESTIMATING SYSTEMATIC RISK Panel A - 2 '2 {7, 12 i 9,. 667,) «721 .173) 66;.) H7.) :(721 1173) «7.7 _r_ .(2) .9231-3 -.1145-2 -.1780-4 .5033 .2949'1 .1501-1 .5233-3 22.1652 .200 -.488 .218 1.488 .3157 .1731 .6061—7 .3922-3 .1002—6 .3495-1 .1576-1 .6063-3 .740 .159 .162 .1951 .1601 .6076-3 ~.7070-3 .5500 .2993—1 .9768-2 2.1117 .007 -.169 1.696 .2669 .1799 .3676-2 .1032-2 .3565-1 .1127-1 .662 1.061 .1300 .1639 Panel B — a — - .— - c - CI ' O 6- 2 2 2;_ :2 :2 :1 9(71) 8(72) 0(73) 0(7‘) t(71) t(72) t(73) t(7‘) :_ 0(r ) .1161—2 -.41°4-2 .1493-3 .6613 .3007-1 .1905-1 .1877-2 3.5902 ..247 -1.410 .511 1.179 .3427 .2047 .5355-2 -.4471-2 .2608-3 .3756-1 .1545-1 .1066-2 .913 1.853 1.566 .1797 .1298 .7778-3 -.127fi-2 .5968 .3038-1 .1223-1 2.5746 .164 -.668 1.484 .2448 .1919 .4059-2 .1435-2 .3746-1 .1248-1 .694 .736 .0915 .1069 Panel C d a- - G I. a I — . 2 2 :— 7.1, L3 1,; 8G,) 0112) ‘ “7,1 067,.) tfi'l) N12) “'73) 9G.) :— I(r ) .1122-3 -.8320~2 .5958-3 .5597 .3031-1 .5961-2 .6241-2 3.3748 .024 -.894 .611 1.062 .3070 .2035 .4481-2 -.1717-1 .1526-3 .3827-1 .5749—1 .4518-2 .750 -1.913 .216 . .1328 .1113 -.1534-3 -.1449-2 .6795 .2938-1 .1615-1 3.0225 -.033 -.575 1.440 .2299 .1987 .6957-2 -.1755-2 .3860—1 .1666-1 .822 -.770 .0569 .0662 Panel D - - I, ..- ‘- y~ - " 5-. 2 2 :1 :2. ‘I_ :5. 0(71) 0(72) 0(73) 017‘) £171) 1(72) t(73) t(7‘) r 0(r ) -.732?-3 -.9276-2 .5139-3 .5102 .2993-1 .5818-1 .5518-2 2.9513 ;.050 -1.021 .596 1.107 .3062 .2006 .4345-2 -.1663-1 .8929-4 .3775-1 .6111-1 .4516-2 .737 -1.743 .126 .1375 .1158 -.8257—3 -.1533-2 .6295 .2856-1 .1556-1 2.8206 -.185 .631 1.429 .2304 .2002 .4904-2 -.1699-2 .3852-1 .1455-1 .815 .075 .0581 .0671 70 TABLE 3 (Continued) .979 Panel B '1 72 '3 7,, .1711 .0721 0(173) .67.) 667,) 107,) 11737 11?.) .2 .021 .6192—2 -.913l-3 .1103-3 .0630 .3834-1 .8696-2 .5001-3 1.8411 1.034 -.687 1.412 2.192 .2487 .1527 .5917~2 .2973-3 .2485-4 .4072-1 .8157-2 .4249-3 .931 .233 .375 .1437 .1121 .5707—2 .6328-3 .0452 .3594-1 .4476-2 1.7094 1.017 '.905 .169 .1547 .1445 .616342 .1023—2 .6051-1 .6537-2 .971 1.005 .0670 .0765 Panel F :1 i2. :2 f3 “71) "72’ fl IG‘) 051) 6G,) 0(7'3) 06") 22 0(02) .4969-2 -.8861-3 .3792-4 .1874 .3528-1 .1220-1 .1790-2 1.8020 ..902 °.465 .136 .666 .2344 .1693 .6005-2 -.7m.5-3 .5310-6 .6003-1 .1155-1 .1355 .961 -.391 .003 .1576 .1459 .6966-2 -.1637-2 .1922 .3555-1 .7366-2 1.6967 .096 1.269 .725 .1056 .1625 .6495-2 -.1183-2 .4095-1 .7555-2 1.016 1.003 .0630 .0697 Panel C I; I; i 1,. .0717 .6721 51.3.) .67.) 067,) .172) «+31 “176’ i 0(22) .6736-2 -.1710-2 -.2692-3 -.0101 .3486-1 .3260-1 .4928-2 2.2407 1.237 -.336 -.349 -.003 .2772 .1496 .6599-2 -.2629-2 -.4750-5 .3962-1 .4126-1 .2526-2 1.067 -.408 -.001 ' .1736 .1295 .5759-2 .8261-3 .0672 .3511-1 .1690-1 1.9805 1.050 .313 .217 .1928 .1627 .6266-2 .1793-2 .4013-1 .1502-1 1.000 .764 .0512 .0746 Panel H 1, $2 3?, ‘7. «7,7 .072) 667,) «7.) 617'.) 1172) 1(7)) .1971 :2 60’) .2673-2-.1o66-2 -.0265-3 .6959 .3760-1 .2130-1 .6007-2 1.9966 '.623 .313 -1.295 1.592 .2220 .1517 .6079-2 -.2699-2 .6620-3 .6109-1 .2260-1 .2007-2 .967 -.769 1.025 .1156- .1001 .2798—2 .1176-3 .6173 .3667-1 .1306-1 1.0012 .691 .050 1.620 .1701 .1660 .6236-2 .5626-3 .4080-1 .1059-1 .340 .0530 .0737 :1 .9069-3 -.4293-2 .6030-3 .6064-2 :1 .1988-3 .4817-2 .9011—3 .4236-2 .3792-2 .1143-3 .4826-2 lgf' .3657-3 .3810-2 -.5699-3 0‘790-2 12 -.1158—2 .7596-4 -.7294-3 .9058-3 '2 -.9697-3 .9743-3 -.6668-2 -.8416-3 —.1133-2 72 “02202-2 .7213-2 .1056-1 .9037-3 7]. TYKBJ5EZ 4 STATISTICAL RESULTS USING CASH FLOW DEFINITION 4 12. 08860‘5 0161‘-“ Y3 -.2024-2 -.5191-4 -.2345-2 -.8187-4 19:! oZB‘S‘J .3996-4 12 02712-3 03062-‘ I: .5168 .5688 Y6 .8578 .5891 41 g. .4924 .6837 41 0- .4345 .6340 H171) 03012-1 .3598-1 .3024-1 .3601-1 .2885-1 .3772-1 .3045-1 03786-5 0(71) .3027-1 .3845-1 .2960-1 .3881-1 .2993-1 .3845-1 .2879-1 .3875”! IN ESTIMATING SYSTEMATIC RISK Panel A 0(72) 1572 0(7‘) 0171) H172) 0(1’3) 007:) i .9020-2 .2026-3 2.1332 .212 -.022 .200 1.551 .3167 .9000-2 .1957-3 .766 .050 .520 .1966 .6319-2 2.0023 .230 -.739 1.769 .2676 .7365-2 .725 .066 .1321 Panel B -1?2) .1112: 61?.) :17.) 117;) .193) :17.) 5: .107331 .7272-3 3.6236 '.066 -1.200 -.657 1.516 .3339 .0056-2 .6766-3 .010 -1.695 -1.105 .1065 .0196-2 2.6696 .109 -.750 1.527 .2670 .0636-2 .716 .739 .0935 Panel C .172) :1fé) .17.) 117,) :17.) 117,) 117;) 1: .3209-1 .2936-2 3.6293 .161 -1.300 .620 .069 .3126 .3002-1 .1525-2 .063 -2.173 .160 .1600 .nsp2 LOMG 4n5 n6“ L4” .zu9 .7m6q :n6 m9” .0n7 Panel D .175) .173) .1?;) 617;) 61%,) 11?.) 11?.) r’ .3370-1 .2077-2 3.1691 '.070 -1.371 .603 .070 .3130 .3170-1 .1531-2 .035 -2.120 .120 .1610 .0663-2 2.0171 -.127 -.606 1.660 .2271 .J%52 .7” ~4u5 Jun 0(02) .1453 .1843 .1433 0(r ) .2036 .1261 .1920 .0615 61.2) .2037 .1297 .1956 .0618 '2 Y1 I; «(\ .. .7814-2 -.1325-2 .8415-4 -.3314 .5470-2 -.1190-2 . .6768-2 .6412-3 .6666-2 .6677-3 31 :2 .5584-2 '. .6677-2 -.5102-4 .6655-2 -.1421-2 ,6426-2 ~.1302-2 I. '12 .4912-2 .4637—2 .6372-2 .2023-2 .5537-2 .1452-2 .6461-2 .1336-2 5 32 .2444-2 .5997-2 -.8765-3 .3068-2 -.1072-2 .6228-2 -.10°4-2 5838—4 .12 6226-3 -.4455-4 .2 0‘736-5 .7066-4 .2 .6086-3 -.1476-3 .8173-4 41 D .2670 .0069 72 TABLE 4 (Continued) 0(71) .3952-1 0‘07"’1 0(71) .3571-4 .3892-1 .3532-1 .4046-1 .3623-1 .4029-1 .3626-1 .4035-1 9(11) .3612-1 .4007-1 .4046-1 Panel E .172) .193) .17.) :1?!) ':175) :1131113‘) 1: .5610-2 .1256-3 1.7335 1.266 —1.510 6.296 -l.224 .2619 .Mfibi .ru90 .8fi>-L5u 2Jn1 .1u5 .3920-2 1.7051 1.122 1.065 -.139 .1762 .6371-2 1.013 .605 .0070 Panel P 0- —- - C" r I- a 2 0(12) 0(13) 0(7‘) t(vl) £172) 0(73) t(v‘) :_ .7100-2, .7160-3 2.0621 1.001 -.561 .390 .597 .2003 .6675-2 .5512-3 1.099 -.069 .215 .1652 .6336—2 1.0661 1.136 -2.099 .026 .2130 .6093-2 1.017 -1.706 .0690 Panel C .17.) .173) .17.) :17.) :173) :173) :15.) 5: .1370—2 .1293-2 2.3637 .060 2.167 .023 .763 .2565 .1309-1 .7396-3 1.013 .932 .612 .1323 .7066-2 1.0601 .970 1.106 .652 .1006 .0351-2 1.025 1.025 .0691 Panel H .172) 0073) 017‘) 01?.) :19.) 0193) :19‘) :: .1267-1 .1075-2 1.7062 '.633 .300 -.000 1.010 .2100 .Jflhd .Jumq .mu —JM) ans gum .6090-2 1.7601 .566 -1.127 1.675 .1532 .5266-2 .906 -1.331 .0567 61.2) .1410 .1095 .1278 .0732 0(r2) .1656 .1593 .1716 .0964 0(02) .1854 .1234 .1735 .0805 73 TABLE 5 STATISTICAL RESULTS USING-MARKET RETURN DATA IN ESTIMATING SYSTEMATIC RISK Panel A - c D — o C d - O . u 2 I. 12- :_ :5 0(Yl) a(?}) 0(13) 011‘) t(Yl) C(12) t(131 t(Y‘1 :- -.6667-2 .5022-1 -.2567-1 -.0611 .5600-1 .1666 .1330 .6699 .550 1.953 -1.2200 .5600 .3007 -.5599-2 .2026-1 -.3210-1 .3616-1 .1695 .1291 -1.050 .060 -1.596 .3195 -.2996-2 .1902-1 -.0351 .5199-1 .5107-1 .6601 -.3690 2.3676 .5019 .3601 -.6776—2 .1536-1 .3339-1 .7105~1 - -29156 1.3609 .2676 Panel B 3. :2 i; ’15 :17.) .172) .173) .17.) :1?!) :192) :193) .174) i .3362-1 —.6010-1 .5713-1 -.1090 .1227 .2310 .1399 1.1317 1.755 -1.090 2.616 -1.070 .6260 .2127—1 —.5960-1 .6607-1 .0717-1 .2317 .1200 1.562 -1.669 2.370 .3166 -.snan-z .2666-1 -.0961 .5907-1 .6661-1 1.0696 —.639 2.369 -.566 .3552 -.7958-2 .1825-1 0‘459-1 08540-1 .10163 10369 I .2599 0(02) .2401 .2328 .2575 .2496 74 replication of the Fama and MacBeth study using only those firms meeting the selection criteria described in Chapter Five. Panel A of Table 5 presents the results based on monthly market return data and Panel B presents the results based on quarterly data. In the tables which follow, it was convenient to generally present four digits to the left of the decimal point and to indicate the number of places the decimal point needs to be moved to the left to interpret the size of the number being presented. For example, .5678-3 is equivalent to .0005678 throughout the tables which follow. The test results permitted several comparisons and conclusions about the research hypotheses which were posed in this research. $F The Relation between bEand Rmr P SF The t-scores on the bp terms were not statistically significant under either ranking procedure 52; under any definition of operating cash flow used in this study. While the principal research hypothesis has been rejected, there are observations which should be made. The following comments may overlap other comments made within this section and should not be interpreted in isolation from these other observations. For operating cash flow definitions 3 and 4 both in unseasonally adjusted format for REF and 121le and for both initial ranking procedures, the t-scores were around +1.0 in the fourth regression where the $F independent variables were the risk-free rate and the bp [see Table 3 Panels A and E, and also Table 4 Panels A and E]. These 75 t-scores indicate a positive relation between the two variables of principal interest in this study. When using the cash flow ranking procedure in conjunction with $F cash flow definition 4, where both R1. and REF were in seasonally adjusted format (Table 4 Panel C), the t-scores on the bSF term were all positive and near +1.0 except in the first regression where the t-score was +2.167. When using the other three formats for seasonally adjusting the return series (or not adjusting the return series), the t-scores were inconsistently positive. These results were encouraging for cash $F and Rs 1 mF had been seasonally adjusted and where the firms luui been ranked on cash flow data as described earlier. flow definition 4, where R When using the cash flow ranking procedure and cash flow definition 3, the t-scores on the sz were all positive in regressions 3 and 4. In these regressions, the independent variables were the $F $F risk-free rate, bp , and s(e;); and the risk-free rate and bp reSpectively. These comments hold regardless of which pattern of seasonal adjustment was employed (Table 3 Panels E through H). The $F t-scores on bp were larger in Panels E and F than in Panels G and B. It should be noted here, however, that in the Fama and MacBeth replication the t-scores were significant at the a=.05 level for the bgr term in regression 3 but not for regression 4 (see Table 5 Panels A and B). The Relation between bgF andR:r Using Cash Flow Definitions 1 and 2 From the results obtained in the regressions with only two S p for the hypothesized relation between systematic operating cash flow independent variables (Rf and b F), there seemed to be little support 76 risk as surrogated in this study and the market returns to portfolios. At a minimum, support for the hypothesized relationship would require that the sign of the t-scores on the b:F term would be consistently or generally positive. Using these two definitions of cash flow and market return rankings for portfolio formation purposes, six of eight $F t-scores on the b1) termwere negative. Similarly, using only cash flow data for the initial ranking of firms prior to portfolio formation, five of eight t-scores were negative. Of the positive t—scores in the regressions with only two independent variables from either ranking procedure and from either cash flow definition, the largest positive coefficient was +.525 and is found in Table 1 Panel A. Since the other cash flow definitions provided more support for the hypothesized relationship between bgF and 11:11., it was not necessary to conclude that the hypothesized relation motivating this study was non- existent. Specifically, cash flow definition 3 provided t-scores whichwere appealing with respect to sign and size. The conclusion which did seem warranted by the t-scores in the regressions containing only two independent variables was that cash flow definitions 1 and 2did not provide surrogates of systematic operating cash flow risk which were related to market returns of their respective portfolios regardless of which of the two ranking procedures were used. A priori, it was not known which, if any, of the definitions of cash flow might provide significant or encouraging results. Moreover, it was not clear which ranking procedure would be best suited to provide portfolios with a wide spectrum of bgF values. 77 Rankings on Cash Flow Data versus Market Return Data The tests of the hypotheses were conducted on portfolios formed from firms which had been initially ranked on both market return data and cash flow return data. When using the cash flow ranked portfolios (Panels B through H) observe that the t-score obtained for the risk- free rate's contribution to the multiple regression relation was always positive and often was almost significantly positive at the .05 level. This result was comparable to what Fama and MacBeth observed over a much longer period of time when using market return betas.1 The t-scores obtained on the Rf rate when using portfolios initially ranked on market return data (see Panels A through D) were rarely as large as +1.0 when positive, and were some- times negative. It was not expected that the risk-free rate should be negatively related to the observed end-of—period market returns to the portfolios. Therefore, the cash flow ranked portfolios appeared to pro— vide t-scores on the risk-free rate which suggested this ranking method waspreferable to ranking on market return betas to form portfolios. The t-scores on the independent variables other than the risk- free rate in these various regressions did not permit reference to an expected outcome as did the t-scores on the risk-free rate. The reason for this was that the other independent variables had not been used in previous empirical research and thus, there were no prior empirical results against which a comparison could be made. Since the t-scores on the risk-free rate in regressions involving portfolios initially ranked on systematic operating cash flow risk measures were positive and often approached significance at the .05 level, the conclusion was 78 drawn that the ranking procedure which was based on cash flow betas in the initial ranking periods was superior to the ranking procedure based on market return betas. Only in the t-scores for one other independent variable, 313;), SF ei estimating regressions, was there a the average residual from the b distinct and recurring difference between the two methods of ranking the firms for the purpose of portfolio formation. The t-scores on this measure of residual risk were always positive and approached sig- nificanceanzthe .05 level when using portfolios formed from initial rankings on market return betas (Panels A through D). When observing the t-scores on the same independent variable, but in the regressions where the portfolios had been formed from cash flow betas (Panels E through H) in the initial estimation period, there werefinine negative t-scores while six of these values were larger than +1.0. Intuitively, these t-scores should have been positive because the residual risk term $F reflects (l) a measure of the extent to which the estimation of bei did INN: reflect the cyclicality of the firm's earnings process and (2) random error. A more complete discussion of the first point follows later in this chapter. $F The t-scores on the b1) and (b31532 terms did not provide any basis for assessing the aptness of either the cash flow ranking procedurecnrthe market return ranking procedure. The sign and significance levels of $F $F 2 the t-scores on the bp and (bp ) terms in the various regressions in Tables 1 through 4were not sufficiently different from the two firm ranking procedures to provide a basis for a conclusion as to which ranking procedure was better. 79~ One factor must be remembered in the interpretation of the $F 1 the bgF estimation periods were of different lengths between the two results from these two rankings, the initial b estimation periods and ranking methods. In the testing procedure where the initial ranking had been on the market return betas, the estimation periods for biF and b:F were each 20 quarters in length. Recall that the estimation of biF for this ranking purpose had been via market return data and not $F i from fresh cash flow data over the 20 quarters in the portfolio by systematic operating cash flow data. The b here had been formed estimation period. Alternatively, when the initial ranking of firms had been via $F i only ten quarters in length due to availability of cash flow datacxflqr cash flow risk measures, the b and bgFestimation periods had been baCk through JUlY 1970 for enough firms to enable a satisfactory sample size to be obtained for the study. For the reasons given above and subject to the limitations noted above, it seemed the cash flow beta ranking procedure (Panels E through H) provided a better basis for ordering the firms for entrance into their reSpective portfolios than did the market ranking procedure. The consistently positive and relatively large t-scores obtained on the risk-free rate when using the cash flow ranking procedure provided credibility for that ranking procedure despite the $F smaller number of observations available for the estimation of b1 in both the ranking and portfolio estimation periods. 80' Relation between the Risk-Free Rate and the Rgr As noted in the previous section, only when the portfolios were $F formed on the basis of bi estimated over ten quarters of cash flow data and ranked on this measure, were the t-scores on the risk-free rate of the expected sign (+). If the risk-return relation should be positive ex post, then the owner of a portfolio of risky assets should have earned a rate of return higher than if that owner held the risk- free asset over the same period. While these expected results would not be borne out consistently in ex post results, it could be expected that over a relatively long time period the expected relation would emerge in the ex post results. This expected relation did occur consistently when firms luui been ranked for portfolio formation purposes using systematic operating cash flow risk measures rather than using systematic risk measures obtained from market return data assuming-these two measures were strongly correlated. The conclusion that the t-scores on the risk-free rates have the expected sign was contingent upon which ranking procedure was used. In an earlier section of this chapter, the conclusion was drawn that the cash flow ranking procedure was marginally better than the market return ranking procedure. These conclusions were not, therefore, inde- pendent of each other. $F it provided results whichwere farther from being statistically significant SE). This ‘ 0 Seasonally adjusting the return series (R and R13: or both) generally than did the unseasonally adjusted return series (RE: and Rm applied to both methods of ranking the firms prior to portfolio formation. In regression 4 (with only two independent variables) the signs on the SF p termwere always positive in Panels A and E in coefficient for the b 81 each table. Some of the signs on the coefficient for the bgF term were positive in other panels where some seasonal adjusting had taken place. Except for Table 3 there was no general pattern of positive signs in the other panels. In all cases the r2 of the regressions involving only two independent variables were higher when neither return series had been seasonally adjusted. The explanatory power of the models used in the testing here has not been enhanced by the efforts to seasonally adjust either return series or both. The Relation between bgr and Rgr when Using Quarterly Returns In Tables, Panels A and B, the results were presented respectively for the replication of the Fama and MacBeth study where the estimation and testing had been done (1) on monthly returns over a time period corresponding to the periods used in this study, and (2) on quarterly returns instead of monthly returns over these corresponding time periods. In the first two regressions listed in Table 5,Panel B, the t—scores on the bgr terms were negative. The t-scores on the same independent variable, but in the third and fourth regressions in the same table were positive. The point is that negative t-scores on the b3]: term in the first regressions were insufficient to judge the extent to which the general CAPM relation between risk and return was fulfilled when using quarterly returns. The observed result on market data and for quarterly returns may make it easier to understand or interpret $F similar negative t-scores on the bP terms in the corresponding regressions while using a cash flow definition in conjunction with one of the ranking procedures. guy-“sauna ." .1 .7, . 82 The Relation between the 915;) Term and the Observed r2 in the Regressions Throughout the results obtained, the presence among the indepen- dent variables of the s(ei) termwas related to a greatly improved degree of explanatory power in the regression (r2). The s(e;) term reflected a $F ei term did not capture the cycli- measure of the extent to which the b cality of the firm. Thus, the s(e;) term could be thought of as a measure of other unspecified cyclicality measures as well as a measure of random error. These other unspecified cyclicality measures would be reflected in the observed REI if we assume the market can properly interpret all the information coming to it from whatever sources. This explains why the 9.161) terms were positively and almost significantly related to the market return to the portfolios at the .05 level. Moreover, it helps to explain why the observed r2 in these regressions was enhanced by the inclusion of the s(e.i) term in the regressions. FOOTNOTES - CHAPTER SIX 1Fama and MacBeth, "Risk, Return, and Equilibrium," pp. 607-636. 83 CHAPTER SEVEN CONCLUSIONS AND LIMITATIONS This study has failed to establish the hypothesized link between systematic operating cash flow risk as surrogated here and the market returns to portfolios. The value of the present research of an embyonic nature should be judged on its contributions to research which may follow, rather than on the results of this one study alone. The limitations of the investigation conducted include several factors. First, the cash flow return definitions may not reflect the most appropriate return measures. Perhaps another untried surrogate return measure for operating cash flow would provide a systematic operating cash flow risk estimate that is more closely related to market returns. Improvements in surrogation might be obtained in either the numerator or denominator of the cash flow return measure. The FASB has not concluded its project on funds flow and liquidity. Conclu- sions of that study may provide alternative useful ways of specifying cash flows. A second limitation of this study may be the time span over which the cash flow data were available for the various estimation and testing periods. The systematic operating cash flow risk measures actually used during the testing periods were intended to represent a wide range of those systematic risk levels. Because of considerable $F instability in those measures, however, the bp used in the testing do 84 85 3F not represent the b which could have been expected if these system- atic risk measures were stable at either the firm or the portfolio level (see Appendix A). This limitation may be the result of estima- tion periods and portfolio periods which are not long enough to per- mit a stable measure to be obtained. A third limitation involves the selected analytical model pro- vided by Haugen and Pappas.1 There are other related analytical models which, if implemented, might yield results more consistent with the motivation of this study. Rubinstein2 and Myers and Turnbull,3 among others, have developed such models which could provide a basis for further testing. This study has sought to establish the existence of a direct con- temporaneous one-period relationship between observed cash flow returns and observed market returns to securities. Financial theory suggests there is a multi-period relationship between expected future cash flows and current security market returns. Such a lead relationship has been ignored in this study. If a model could be developed incor- porating actual cash flows and expected future cash flows, the likeli- hood of observing a relationship between cash flows and market returns would appear to be improved. FOOTNOTES - CHAPTER SEVEN 1Haugen and Pappas, "Equilbrium in the Pricing of Capital Assets," pp. 943-953. 2M. E. Rubinstein, "A Mean-Variance Synthesis of Corporate Financial Theory," Journal of Finance 49 (March 1973): 167—181. 38. C. Myers and Stuart M. Turnbull, "Capital Budgeting and the Capital Asset Pricing Model: Good News and Bad News," Journal of Finance (May 1977): 321-333. 86 APPENDIX A As noted in the limitations portion of Chapter Seven, there was a $F large degree of instability in the b1) which were used in the 42 test periods throughout the 14 calendar quarters used in the testing period. In this appendix the bgF which were used in the test periods will be presented for one of the cash flow definitions as an illustration of this instability. For this purpose, the first cash flow definition was selected and neither the firms' nor the market's return series were adjusted for the seasonality factor. Recall the procedures used in arriving at these values. There were two alternative ways of obtaining a ranking of firms for the purpose of assigning those firms to their portfolios. One of these methods was a ranking on market return betas (bmr) and the other method was a ranking on cash flow betas (biF). Regardless of the method used to rank these firms, the bEF used in the test periods exhibited a large degree of instability. This instability may have prevented this study from presenting a defini- tive statement about the risk-return relation where risk is considered to be systematic operating cash flow risk. Additionally, recall, the cash flow return measures used here excluded expected cash flows from their computations. Table 6,Panel A contains the bEF as they were used in the 14 quarters representing the test periods. These firms were ranked $F initially on bmr for the purpose of forming portfolios. The b1 were then measured in the subsequent 20 quarters and equally weighted to 87 tar“, “379.492..ng 9.96.0.2»..- . . 88 obtain bgF. This process was updated each quarter to provide the b$F p t-l which was the surrogate for the systematic Operating cash flow 9 risk measure at the outset of each quarter. Table 6, Panel B contains SF P The firms here were ranked initially using ten quarters of cash flow the b as they were used in the same 14 quarters in the testing period. return data. The biF were estimated again in the subsequent ten SF 1,0-1 used in the testing periods to surrogate the ex ante concept of quarters and equally weighted in order to obtain the b which were systematic operating cash flow risk as it has been described above. As might be expected, the longer the period of observation of bi for the purpose of ranking firms on the systematic risk estimate, the $F less extreme are the subsequent measures of bp used in the testing periods. This longer period of observation, however, requires the use of market return data for that ranking process. Alternatively, as Table 6, Panel B indicates, when ten quarters of cash flow return data $F P which appear in the testing periods are more extreme. It seems that are used for the initial ranking of firms, the result is that the b this is a result of the regression model's inability to efficiently measure the systematic risk of a firm when ten observations are used in that estimation. Each table contains 20 portfolios. These portfolios are presented here starting with the lowest bgF to the highest bSF portfolio and progressing portfolio. Because of the instability discussed earlier and again above, that progress from low to high may not be apparent. 89 ~09.9 n05.9 n09.9 005.9 0n5.9 555.9 505.9 099.9 ~00.9 5m9.9 me9.9 mwm.9 900.9 000.9 999.9 999.9 N5m.9 000.9 900.9 990.9 9mm.9 n~9.9 cm0.9 950.9 099.9 505.9 000.9 N05.9 5~.m 500.9 000.9 0n0.9 550.9 990.9 000.9 emm.9 ~50.9 99m.9 999.9 a0~.9 099.9 59N.9 000.9 H we eee.e eae.e 0H5.9 ~0¢.9 999.9 9N0.9 090.9 555.9 550.9 500.9 995.9 099.9 -m.9 099.9 0M9.9 900.9 nu9.~ ~0m.9 Nnn.9 9N0.9 aaa.e .ea.e 999.9 555.9 500.9 955.9 505.9 m~9.9 mmaumw mmaum a» a» 000.9 905.9 000.9 095.9 eaa.e nea.e ~90.9 mnm.9 900.9 090.9 0N0.9 N90.9 ~Nm.9 990.9 990.9 990.9 Nan.9 900.9 000.9 990.9 999.9 000.9 mem.e mne.e ~90.9 950.9 a.e.e mee.e mama. Luna. awe a». e to waves. H335 < Amz