ABSTRACT AN EMPIRICAL EVALUATION OF THE COMMUNICATIVE EFFECTIVENESS OF CONSISTENCY EXCEPTIONS IN AUDITORS' OPINIONS BY Elba Franklin Baskin The purpose of this research effort is to examine the information content of consistency exceptions in auditors' opinions and, by inference, to provide empirical evidence which will support a judgment as to the effective- ness of the consistency exception communication network. Secondary to this main purpose, certain questions con- cerning the concept of materiality are explored. It is hypothesized and supported by prior empirical work that a causal network among information inputs, covert investor decision-making processes and overt price re- sponses exists. Thus, a price response variable can be utilized in determining the information content of consistency exceptions. The study examines historical common stock price changes, adjusted for market effects during the annual report issuance period, of three samples of New York Stock Elba Franklin Baskin Exchange firms during the period 1965 to 1969. The control sample is composed of 126 firms that made no accounting principle changes. Experimental sample I consists of fifty-eight firms which exhibit consistency exceptions. These fifty-eight firms notify investors of their account- ing principle changes in their annual earnings announce— ments and in their annual reports. The third group of firms, experimental sample II, consists of seventy firms that also exhibit consistency exceptions. But these firms report their accounting principle changes only in their annual reports. Thus, the control sample and experimental sample I disclose no "new" information, ceteris paribus, in their annual reports while experimental sample II reports new information concerning accounting principle changes. By comparing the three samples on the basis of the market-adjusted price response variable, inferences can be made concerning the information content and communicative effectiveness of consistency exceptions in auditors' opinions. Two measures of materiality are also created. A quantitative measure of the magnitude effect on reported net incomes of the firms in the study is used in a re- search design to determine if there is a relationship between this materiality-magnitude measure and price response. This research design seeks to provide evidence Elba Franklin Baskin as to the existence of investor materiality functions as well as aiding in the determination of the information content of consistency exceptions. A qualitative measure of materiality based on the type of accounting principle change made by the firms in the study is also used in various research designs comparing price response among the three samples. These designs also seek to provide evidence as to the existence of investor materiality functions. The results of the analysis of covariance test used to determine the information content of consistency exceptions is that the average price response differences among the three samples are statistically significant at a level of .13 which is greater than the conventional level of .05. Therefore, the conclusion is that consist— ency exceptions in auditors' Opinions appear to possess no information content when the statistical results are judged at the conventional significance level. Secondly, a simple linear regression model is used to regress the price change variable against the quantitative measure of materiality defined in this study. The result of this materiality-magnitude test demonstrates that there is little or no relationship between the price response measure and the materiality measure. Thus, one concludes that investors do not appear to possess Elba Franklin Baskin materiality functions based on the quantitative measure used in this study to operationalize the concept of materiality. Lastly, the results of the materiality test utilizing the qualitative definition of materiality provide very little evidence to substantiate the existence of investor materiality functions. AN EMPIRICAL EVALUATION OF THE COMMUNICATIVE EFFECTIVENESS OF CONSISTENCY EXCEPTIONS IN AUDITORS' OPINIONS BY Elba Franklin Baskin A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Accounting and Financial Administration 1970 0271054 © Copyright by ELBA FRANKLIN BASKIN 1971 ACKNOWLEDGMENTS I wish to express my appreciation to the members of my dissertation committee, Dr. Roland F. Salmonson (Chairman), Dr. Alvin A. Arens and Dr. William H. Schmidt, for their invaluable counsel and aid in completing this dissertation. For the honor of being selected as a fellowship recipient and for the financial assistance thus provided, I take this opportunity to thank the American Institute of Certified Public Accountants, the Earhart Foundation and the National Honor Society of Phi Kappa Phi. To Dr. James Don Edwards, Chairman of the Depart- ment of Accounting and Financial Administration, I express my gratitude for the advice and assistance that aided in my completion of the doctoral program. I also express my appreciation to the members of the Department of Stock List of the New York Stock Exchange for their helpfulness with certain data gathering aspects of this dissertation. To my good friends and colleagues at Michigan State, especially G. Michael Crooch, Hugh A. Hoyt, iii George W. Krull, Robert G. May and William J. Morris, I wish success and express my appreciation for their willingness to engage in interminable discussions on numerous tOpics. And to my wife, Judith, for her love, devotion and patience which have made these doctoral years so wonderful and enjoyable, I express my eternal love. iv TABLE OF CONTENTS LIST OF TABLES O O O O O O O O O O I O I O O O O I 0 LIST OF FIGURES O 0 O O O O O O O I O O O O O O O 0 CHAPTER I 0 INTRODUCTION 0 O O O O O O O O O O O O O O 0 Purpose of the Research . . . . . . . . . The Second Auditing Standard of Reporting . . . . . . . . . . . . . . Definition and General Discussion of the Standard . . . . . . . . . . . . . . . Measurement Rule Changes . . . . . . . . Types of Qualified Opinions Arising as a Result of Accounting Principle Changes . . . . . . . . . . . . . . . The Subjectivity Surrounding Measurement Rule Changes . . . . . . . . . . . . . . Introduction . . . . . . . . . . . . . . Criteria Characterizing a Better Measure . . . . . . . . . . . . . . . Implications for the Area of Accounting Principle Changes . . . . . . Literature Review . . . . . . . Scope and Approach of the Research Summary . . . . . . . . . . . . . II. THEORETICAL CONSIDERATIONS . . . . . . . . . Introduction . . . . . . . . . . . . . . Specification of an Investor Behavior MOde l O O O I O O O O O O O O "Intrinsic Value" Adjustments as Reflected by the Investor Behavior Model . . . . . Description of the Process of Intrinsic Value Adjustments . . . . . . . . . . Implications of the Intrinsic Value Adjustment Process . . . . . . . . . Empirical Evidence Supporting the Information Content Hypothesis . . . . Page viii 10 13 23 25 27 27 27 29 29 30 30 CHAPTER The Information Content of Consistency Exception Opinions . . . . . . . . . . . Introduction . . . . . . . . . . . . . . Investor Response to Consistency Exceptions . . . . . . . . . . . . . . Investor Materiality Function . . . . . Methodological Implications . . . . . . . Summary . . . . . . . . . . . . . . . . . III. HYPOTHESES AND METHODOLOGY . . . . . . . . . Introduction . . . . . . . . . . . . . . . General Hypotheses . . . . . . . . . . . . Principal Hypothesis . . . . . . . . . . Materiality Hypotheses . . . . . . . . . Definition of the Information Stimulus . . Identification of the Universe and Samples . . . . . . . . . . . . . . . . Population Criteria . . . . . . . . . . Sample Selection Criteria and Procedures . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . Formulation of the Test Statistics . . . . Logarithmic Price Relative . . . . . . . Removal of the General Market Effect from the Logarithmic Price Relative . Identification of the Annual Earnings Announcement Date and the Annual Report Issuance Date . . . . . . . . . The Test Statistics . . . . . . . . . . Data Sources . . . . . . . . . . . Statistical Hypotheses and Statistical Tests . . . . . . . . . . . . . . . . . Introduction . . . . . . . . . . . Principal Statistical Hypothesis . . . . Materiality Statistical Hypotheses . . . Estimation O O O O O O O O O O O O O O 0 Significance Level . . . . . . . . . . Limitations of the Methodology . . . . . . Summary . . . . . . . . . . . . . . . . . IV. RESULTS OF THE STATISTICAL TESTS . . . . . . Introduction . . . . . . . . . . . . . . . Results of the Statistical Test Concerning the Principal Hypothesis of the Study . Results of the Statistical Tests Con- cerning the Materiality Hypotheses . . . vi Page 37 37 38 39 41 42 44 44 45 45 47 52 53 53 56 61 62 62 63 69 76 78 79 79 79 86 95 95 96 97 98 98 105 CHAPTER Page Materiality-Magnitude Hypothesis . . . . 105 Materiality-Type of Accounting Principle Change Hypothesis . . . . . 107 Summary . . . . . . . . . . . . . . . . . 115 V. SUMMARY, CONCLUSIONS, IMPLICATIONS AND RECOMMENDATIONS O O O O O O O O I O O O O 117 Summary of the Research . . . . . . . . . 117 Conclusions of the Study . . . . . . . . . 119 Implications of the Empirical Results . . 125 Recommendations . . . . . . . . . . . . . 126 BIBLIOGMPHY O O O O I O O O I O I O O O O O O O O O 128 APPENDICES A. COMPUTATIONAL FORMULAS FOR CALCULATING AUTOCORRELATION COEFFICIENTS AND COEFFICIENTS OF DETERMINATION . . . . . . 133 B. SUPPLEMENTARY FIGURES TO CHAPTER III . . . . 135 vii TABLE 1. 10. LIST OF TABLES Number of Firms in the Population as Determined by Applying the Population criteria 0 O O O O O C O I O C O O O O O 0 Identification of the Number of Firms in the Experimental Group . . . . . . . . . . Determination of the Number of Firms in the Control Group . . . . . . . . . . . . . . Frequency Distribution of the Autocorrela- tion Coefficients of Determination . . . . Frequency Distribution of the Coefficients of Determination . . . . . . . . . . . . . The Category Frequencies of the Price Response Measure Used as Inputs into the Two-Way Analysis of Covariance Research Design . . . . . . . . . . . . . . . . . . The Category Frequencies of the Price Response Measure Used as Inputs into the Two-Way Analysis of Covariance Research Design . . . . . . . . . . . . . . . . . . Results of the Two-Way Analysis of Co— variance Test of the Principal Hypothes is O O O O O O O O O O O O O O O 0 Results of the Two-Way Analysis of Covariance Test of the Materiality-Type of Accounting Principle Change Hypotheses . . . . . . . . . . . . . . . . Results of the Student's t Test Comparing the Average Price Responses of the Type of Accounting Principle Change Categories of Experimental Sample II with the Average Price Response of the Control Sample . . . . . . . . . . . . . . . . . . viii Page 55 57 60 68 7O 85 91 104 112 114 TABLE Page 11. Frequency Distributions for the Value of the Magnitude Measure for the Experi— mental Samples . . . . . . . . . . . . . . 122 12. Frequency Distributions for the Type of Accounting Principle Changes Reported by the Experimental Samples . . . . . . . 123 ix Figure l. 2. 10. 11. 12. LIST OF FIGURES Model of the Measurement Process . . . . . . Annual Report Profile for 254 Annual Reports . . . . . . . . . . . . . . . . . Category Means Found in the Two—Way Analysis of Covariance Research Design . . . . . . Category Means Found in the Two-Way Analysis of Covariance Research Design . . . . . . Annual Earnings Announcement Profile for 254 Annual Earnings Announcements . . . . Annual Earnings Announcement Profile for 126 Annual Earnings Announcements-- Control Sample . . . . . . . . . . . . . . Annual Earnings Announcement Profile for Fifty-Eight Annual Earnings Announce- ments--Experimental Sample I . . . . . . . Annual Earnings Announcement Profile for Seventy Annual Earnings Announcements-- Experimental Sample II . . . . . . . . . . Annual Report Profile for 254 Annual Reports I O O O O O O O O O O O O O O O 0 Annual Report Profile for 126 Annual Reports--Control Sample . . . . . . . . . Annual Report Profile for Fifty-Eight Annual Reports--Experimental Sample I . . Annual Report Profile for Seventy Annual Reports--Experimental Sample II . . . . . 75 102 110 136 136 137 137 138 138 139 139 CHAPTER I INTRODUCTION Purpose of the Research Historically, corporate management has selected the accounting principles employed to measure firm activi— ties and, within certain constraints, has been allowed to change these accounting principles. Management could have multiple motives for changing accounting principles, but presumably any change should always lead to a better measure. At present, however, the determination of what is a better measure is a highly subjective process. The American Institute of Certified Public Account- ants (AICPA) has not developed specific criteria to define a better measure, but has instead relied on the subjective concept of "generally accepted accounting principles" (GAAP). Thus, an accounting change is usually acceptable when the GAAP requirement is met. Coincident with meeting the GAAP requirement, the AICPA specifies that the inde- pendent auditor must conform to the second auditing Standard of reporting which requires disclosure of the occurrence of an accounting principle change and its related effects to users of financial statements. Presumably, upon notification of the change, a user can make the appropriate adjustment(s) in his decision process by using the data disclosed. The purpose of the research presented herein is an attempt to provide empirical evidence demonstrating the effectiveness of the notification and disclosure net- work required by the second auditing standard of reporting. Specifically, does the reporting standard notify the user of the occurrence of accounting principle changes, or does a communication problem exist which the accounting pro- fession must solve? The Second Auditing Standard of Reporting Definition and General Discussion of the.Stafidard To aid in attaining interperiod comparability, the AICPA includes the second standard of reporting among its auditing standards. This reporting standard states that "the report [auditor's opinion] shall state whether such principles have been consistently observed in the current 1 period in relation to the preceding period." The stated objective of-this consistency standard is 1Committee on Auditing Procedure, Auditin Standards and Proceduresé-Stgtements on Auditing Procedure No. 33 (New York: American Institute of Certified Public Accountants, 1963), p. 42. (l) to give assurance that the comparability of financial statements as between periods has not been materially affected by changes in accounting principles employed or in the method of their application; or (2) if comparability has been materially affected by such changes, to require a statement of the nature of the changes and their effects on the financial statements.1 The Committee on Auditing Procedure of the AICPA pointed out that interperiod comparability is affected by changes arising from: (a) A change in accounting principles employed, (b) changed conditions which necessitate account- ing changes but which do not involve changes in the accounting principles employed, (c) changed conditions unrelated to accounting.2 The committee stated that only changes in accounting principles employed (hereafter referred to as type (a) changes) involve the consistency standard; thus, only these changes, if material, require recognition in the opinion paragraph of the auditor's report.3 Any of the situations listed below in which the effect of the change is material lead to an auditor's qualification4 of his opinion as to the fairness of a firm's financial state- ments. lIbid. 2Ibid., p. 43. 31bid., p. 43. 4"Any modifying phrase in the standard short-form opinion paragraph (or sentence) should be considered as qualifying the opinion in some manner. . . ." Ibid., p. 59. '———' l. A change from a generally accepted accounting principle to an alternative generally accepted accounting principle. 2. A change to a generally accepted accounting principle from a principle or practice lacking general acceptance. 3. A change to a principle or practice which lacks general acceptance. (If highly material, such a change may lead to an adverse opinion stating that the financial statements do not present fairly the financial position or results of operations in conformity with generally accepted accounting principles.) 4. A change which is expected to have a material future effect is not disclosed in a note to the financial statements.1 Measurement Rule Changes Type (a) changes are defined as measurement rule changes. As Grady states: The objective in requiring disclosure of the effect upon income of changes in accounting principles [type (a)] is to provide users of financial state- ments with the means of measuring the remaining changes in operations arising from changed conditions, i.e., events or transactions. . . .2 As stated previously, management may have multiple motives in making a measurement rule or accounting policy change, but a change, if made, should lead to a better measure. The difficulty lies in the subjectivity of ascertaining lIbid., pp. 45-48. 2Paul Grady, Inventory of Generally Accepted Accounting Principles for Business Enterprises (New York: American Institute of Certified Puinc Accountants, 1965), p. 182. what is a better measure. Rather than develop criteria to determine a better measure, the AICPA places reliance on the consistency exception and the related disclosure requirements of the second reporting standard or consist- ency exception standard to provide notification of measure— ment rule changes to financial statement users. With this notification, the user can make appropriate adjustment(s) in his decision process. Types of Qualified Qpinions Arising as a Result Of Accounting Principle Changes The standard form of the short-form independent auditor's report which accompanies the basic financial statements is: We have examined the balance sheet of X Company aS-of June 30, 19 , and the related statement(s) of income and retgined earnings for the year then ended. Our examination was made in accordance with generally accepted auditing standards, and accord- ingly included such tests of the accounting records and such other auditing procedures as we considered necessary in the circumstances. In our opinion, the accompanying balance sheet and statement(s) of income and retained earnings present fairly the financial position of X Company at June 30, l9__, and the results of its operations for the year then ended, in conformity with generally accepted accounting principles applied on a basis consistent with that of the preceding year. The first paragraph of the short-form report is referred to as the "scope" paragraph and "consists of a representation Committee on Auditing Procedure, p. 57. as to the work performed. . . ."l The "opinion" or second paragraph is a "representation as to the independent 0 I 2 aud1tor's conclus1ons. . . ." When an accounting principle change occurs, the opinion paragraph of the short—form report is modified as follows if financial information presented for prior years is not restated and the change affected these prior years: . . . in conformity with generally accepted account- ing principles applied on a basis consistent with that of the preceding year, except for the change, (insert expression of approval), in pricing of inventories . . . as described in Note __ to the financial statements. The expression of approval is optional, and obviously, the example given of a change in inventory pricing is just one of many examples that could have been used in the illustra- tion.4 If financial information for prior years is re- stated when an accounting principle change occurs, the short-form opinion is modified as follows: . . . in conformity with generally accepted account- ing principles applied on a basis consistent with that of the preceding year after giving retroactive effect to the inclusion, which we approve, of the Ibid. Ibid. Ibid., pp. 46-47. but)?“ Ibid., p. 47. accounts of foreign subsidiaries as explained in Note __ to the financial statements. Again, the expression of approval is optional, and the example of the accounting principle change illustrated is arbitrarily chosen.2 The illustration of the modification of the stand- ard short-form opinion where financial information pre- sented for prior years is not restated is referred to as the consistency exception type opinion. When the financial information presented for prior years is restated, the modified opinion is referred to as the retroactive type opinion. Firms issuing annual reports containing a retro- active type opinion are considered to be consistently applying their accounting principles just as firms that reported no changes whatsoever are considered to be consistent in their application of the accounting principles. But firms issuing annual reports containing an exception type opinion are not considered to be con- sistent in applying their accounting principles. lIbid., p. 49. 21bid. The Subjectivity Surroundinngeasurement Rule Changes Introduction In a normative sense, any change in accounting principles should lead to a better measure. But, what characteristics should a "better" measure exhibit? Using a theoretical framework analogous to the one presented by Ijiri and Jaedicke, an attempt can be made to develop criteria that an accounting principle or measurement change should satisfy to be characterized as a better measure.1’ 2 This framework will also demonstrate why, given the present state of knowledge about financial data users, the account— ing profession places reliance on the present, rather subjective, criterion of GAAP and on the consistency exception standard of reporting in the area of accounting principle changes. Criteria Characterizing a Better Measure Assume a particular attribute is to be measured. Also, assume several measurers (accountants) using specified measurement rules (accounting principles) lYuji Ijiri and Robert K. Jaedicke, "Reliability and Objectivity of Accounting Measurements," The Accounting Review, XLI, No. 3 (July, 1966), pp. 474-483. 2Yuji Ijiri, The Foundations of Accountinngeasure- ment (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1917) , pp. 133-146. perform measurements of the attribute. Figure 1 illustrates the results of this measurement process. V’///////[\\V:::\-F.D. of xi E[xi] x* x. = the measurement taken by the ith measurer under a specified measurement system x* = the "true" measure of the attribute F.D. of x. = the frequency distribution of the measurements E[xi] = the expected value or mean of the xi's Fig. l.--Model of the measurement process Ijiri and Jaedicke define the dispersion of the xi values around their expected value, Elxi], as objec- tivity or the degree of consensus among the measurers. The distance of the mean, E[xi], from truth, x*, is defined as bias. Any change in measurement rules should lead to a measure which meets one of the following criteria: 1. Have less bias than the old measurement with the same objectivity (dispersion). 2. Have more objectivity (less dispersion) and less bias than the old measurement. 3. Have more objectivity (less dispersion) than the old measure with the same bias. 10 If the new measure meets one of these conditions, the new accounting policy can be characterized as leading to a better measure. Implications for the Area of Accounting Principle Changes The true measure, x*, has purposefully remained undefined. A metaphysical approach could be taken. But highly abstract or abstruse approaches even though very logical have often led to opposing conclusions and a multiplicity of accounting alternatives in certain areas. The more recent posture in the accounting profes- sion is to define the true measure, x*, to be a function of users' decision processes. Thus, Ijiri and Jaedicke's definition of bias is user-oriented in that truth is defined as an alleged value which is a function of the user's predictive process.1 At this point, an example may aid in illustrating the concept of an alleged value. Assume an investor uses the following model to predict the net income of a firm in time period t+l. = 2EPS EPSt+1 t where: = the predicted earnings per share in time period t+l, EPSt+1 EPS = the reported earnings per share in time period t. leiri and Jaedicke, pp. 474-483. 11 Now, assume that the actual per share earnings number reported in time period t is $2. Based on this informa- tion, the investor predicts that earnings per share next period should be $4. Subsequently, the firm actually reports earnings of $3 per share. Then, given that the investor does not change his decision model, the investor alleges that in time period t reported earnings per share should have been $1.50 rather than $2. Thus, the differ— ence between the alleged value of $1.50 and the reported value of $2 is defined as measurement bias. Obviously, to determine measurement bias, one must know the decision model of the investor. Another example of defining the true measure, x*, to be a function of the user's decision process is supplied by the American Accounting Association's (AAA) Committee on Managerial Decision Models which defines measurement system bias as the difference between the expected value of measurements produced by some set of measurement rules (such as absorption inventory costing) and the corresponding measurement of those values which would produce optimal decisions when intro- duced into a user's decision model (perhaps, in a particular case, direct costing). lAmerican Accounting Association, Report of the Committee on Managerial Decision Models, a supplement to The Accounting Review, XLIV, 1969, p. 50. 12 This definition of bias is also user—oriented, but differs from Ijiri's and Jaedicke's in that it makes a judgment as to the deCision outcome. If the user-oriented definition of "truth" is adopted, supposedly the user and others who know the decision model will be able to determine if the new measure resulting from an accounting principle change reduces measurement bias. It is unknown whether an increase or decrease in objectivity would be perceived by the user since present financial statements contain only point estimates of the attributes under measure. Since very little is presently known about users' actual decision processes, truth, x*, becomes rather difficult to define. This means that bias cannot be measured and, therefore, that measurement rule changes cannot be judged better or worse on the basis of more or less measurement bias. Since accounting principle changes are allowed, one can assume that the AICPA has been willing to trade-off possible management manipulation of financial statement data with hOped for movements to measUrement systems with more objectivity and/or less measurement bias. The possible reasons for the AICPA's posture are the current impossibility of determining measurement bias as well as management's familiarity with the firm's opera- tions and circumstances. However, the AICPA implemented 13 the second auditing standard of reporting in order to notify the users of financial statements of the occurrence of accounting policy or measurement rule changes, thereby allowing the user to make any adjustment in his decision process deemed necessary. Because of the present inability of anyone other than the user to determine the measurement bias effect of accounting policy or measurement rule changes, it is vital that the consistency exception stand- ard be successful in ensuring notice to users of accounting measurement changes. Literature Review Recent accounting literature abounds with tests of user's reactions to differing accounting alternatives. Some of these studies have been empirical while others have taken an experimental approach. These studies have been concerned with whether differing accounting alterna- tives used in the same circumstances by different firms caused different user responses.l 1Examples are: Charles P. Bonini, Simulation of Information and Decision Systems in the Firm (Englewood CIiffs, New fiersey: Prentice-Hall, Inc., I963); Robert E. Jensen, "An Experimental Design for Study of Effects of Accounting Variation in Decision Making," Journal of Accounting Research, IV, No. 2 (Autumn, 1966), pp. 244- 238; W. J. Bruns, Jr., "Inventory Valuation and Management Decisions," The Accounting Review, XL, No. 2 (April, 1965), pp. 345-357;-T. R. Dyckman, 1rOh‘the Investment Decision," The Accountin Review, XXXIX, No. 2 (April, 1964), pp. 285-295; T. R. Dyckman, "The Effects of Alternative 14 Hepworth's classic article demonstrated that financial variables may be subject to manipulation by using accounting principle changes.l Gordon developed a theoretical framework suggesting management's motivation for "smoothing" or minimizing the variance of reported earnings.2 Many studies have attempted empirically to support the smoothing hypothesis.3 Accounting Techniques on Certain Management Decisions," Journal of Accountinngesearch, II, No. 1 (Spring, 1964), pp. 91-107; T. R. Dyckman, flOfi’the Effects of Earnings-- Trend, Size and Inventory Valuation Procedures in Evaluat- ing a Business Firm," Research in Accounting Measurement (Menasha, Wisconsin: George Banta, 1966): pp. 175-185; John L. O'Donnell, "Relationships Between Reported Earnings and Stock Prices in the Electric Utility Industry," The Accounting Review, XL, No. 1 (January, 1965), pp. 135:I43; J6hn L. O'DonneII, "Further Observation on Reported Earnings and Stock Prices," The AccountingReyiew, XLIII, No. 3 (July, 1968), pp. 549-553; Edward L. Summers, "Observation of Effects of Using Alternative Reporting Practices," The Accountinngeview, XLIII, No. 2 (April, 1968): pp. 257F265. 1Samuel R. Hepworth, "Smoothing Periodic Income," The Accounting Review, XXVIII, No. 1 (January, 1953), pp. 32-39. 2MyronyJ. Gordon, "Postulates, Principles and Research in Accounting," The Accounting Review, XXXIX, No. 2 (April, 1964): pp. 251-263. 3Examples are: Myron J. Gordon, B. N. Horwitz, and P. T. Meyers, "Accounting Measurements and Normal Growth of the Firm," Research in Accounting Measurement (Menasha, Wisconsin: George Banta, 1966), pp. 221-231; Ronald M. Copeland, "Income Smoothing," Empirical Research in Accounting: Sglected Studies, 1968, a supplement to Volume VI ofithe Journal of Accounting_Research, pp. 101- 116; Ronald M. Copelandfiand Ralph Licastro, "A Note on Income Smoothing," The Accounting Review, XLIII, No. 3 (July, 1968), pp. 540-545; Ronald M. Copeland and 15 These smoothing studies show mixed results and are not conclusive in demonstrating an all pervasive ex— istence of smoothing behavior. Cushing stated that while his results were consistent with the smoothing hypothesis, he favored an alternative hypothesis that dampened the unethical connotation associated with smoothing behavior and implied by the authors of many of the smoothing studies.l Cushing's alternative hypothesis states that "the decision to make an accounting change may be viewed as reflecting a resolution of two related issues: (1) whether a change should be made and (2) in what time period it should be implemented."2 The first issue is probably related to long-run considerations and probably is very rational and ethical.3 The second aspect of the Joseph F. Wojdak, "Income Manipulation and the Purchase- Pooling Choice," Journal of Accounting Research, VII, No. 2 (Autumn, 1969), pp. 188-195; Barry E. Cushing, "An Empirical Study of Changes in Accounting Policy," Journal of Accounting_Research, VII, No. 2 (Autumn, 1969), pp. I96-203; Jean-Marie Gagnon, "Purchase Versus Pooling of Interest: The Search for a Predictor," Empirical Research in Accounting: _Selected Studies, 1967, a supple- ment to Volume V ofithe Journal of Accounting Research, pp. 187-204; Michael Schiff,fiAccounting Tactics and the Theory of the Firm," Journal of Accounting Research, IV, No. 1 (Spring, 1966): PP. 62—67. 1Barry E. Cushing, "An Empirical Study of Changes in Accounting Policy," Journal of Accountinngesearch, VII, No. 2 (Autumn, 1969), pp. 196-203. 21bid., p. 203. 31bid. 16 decision to change accounting principles may be related to shorter—run considerations in terms of minimizing the effects on earnings.1 While the ethicalness of this short- run behavior is open to question, management is actually reacting rationally within the environmental constraints presented by government regulatory agencies and the accounting profession. Neither group can completely control management behavior, but each group can place restraints on behavior. It is the responsibility of the accOunting profession to offer evidence supporting the effectiveness of the consistency exception standard, the present solution used to notify users of financial data about the occurrence of an accounting measurement rule change. If the communication aspect of the consistency standard proves not to be effective, then the profession must provide alternative solutions. Neumann has reported on the nature and incidence of consistency exceptions while Bird has examined inter- period comparability in financial statements and surveyed financial analysts as to their desires for additional data to aid them in attaining interperiod comparability.2’ 3 lIbid. 2Frederick L. Neumann, "The Incidence and Nature of Consistency Exceptions," The Accounting Review, XLIV, No. 3 (July, 1969): PP. 546-554. 3Francis A. Bird, "Interperiod Comparability in Financial Reporting," Journal of Accountangy, CXXVII, No. 6 (June, 1969), pp. 51-56. 17 Using a computer simulation model, McCosh attempted to determine whether no changes over time in accounting principles, moderately inconsistent sets of changes over time in accounting principles or markedly inconsistent sets of changes in accounting principles over time produced earnings per share results that depicted economic fluctua- tions most accurately.1 In the simulation, McCosh used twelve different sets of economic data and fifteen differ- ent sets of accounting principles with which to report the economic data. The simulation covered fifty report periods. Moderately inconsistent sets of changes in accounting principles were defined as five accounting principle changes over the fifty report periods while markedly in- consistent sets were defined as ten to twelve changes over the time period. No accounting principle changes over time were defined as consistent sets. McCosh's findings were that "cOnsistent application of accounting policies, how— ever diverse these policies may be, will lead to earnings per share results that depict economic fluctuations more accurately and similarily than even moderately inconsistent applications of accOunting policies. . . ."2 1Andrew M. McCosh, "Accounting Consistency-Key to Stockholder Information," The Accounting_Review, XLII, No. 4 (October, 1967), pp. 6934700. 2 Ibid., p. 699. 18 In a recent exposure draft, the Accounting Principles Board (APB) of the AICPA proposed an opinion concerning methods of treating and disclosing changes in accounting principles.1 One of the main recommendations which would be required by the opinion and which differs from the historical treatment of accounting changes is that I accounting changes, . . . should be computed and applied retroactively by restating financial statements presented for any period affected by the change. In the area of disclosure of accounting changes, the APB opinion would require that management give the reason for the change . . . in a manner which makes clear the reasons why the newly adopted method furnishes more useful results than the method previously followed.3 This second disclosure recommendation of the APB will be quite difficult to apply since the APB has not defined criteria for determining usefulness. The criteria developed in this chapter to define a change to a better measure are increased objectivity, less measurement bias or both. Since it is doubtful that management has 1Accounting Principles Board, Proposed APB Opinion: Changes in AccountingMethods ang Estimates (New York: American Institute of Certified Public Accountants, February 16, 1970). 2Ibid., p. 7. 3Ibid., p. 8. l9 knowledge of inveStor decision models, measurement bias cannot be determined. Therefore, based on the criteria for determining a better measure as presented in previous seCtions of this chapter, the implication is that manage- ment should only make an accounting change when the new measure is more objective than the old measure. But, if management followed this policy, measurement bias might increase thus offsetting any objectivity gains. This again serves to illustrate the subjectivity surrounding the area of accounting principle changes, and the crucial nature of the consistency exception network of communica- tion. Until recently, no studies had examined the area of user response when a firm made an accounting policy change, and none has attempted an empirical evaluation of the communicative aspect of the network used by accountants to notify the user of an accounting policy change. In a recent issue of the Financial Analyst Journal, Dean E. Graber reports a study of fourteen firms which had higher average increases in stock prices than the Standard and Poor's Industrial Index (no statistical tests were made of the significance of the difference) in the period subse- quent to accounting policy changes which had the effect of increasing reported earnings.1 Graber implies that higher 1Dean E. Graber, "Real and Illusory Earnings Growth," Financial Analysts Journal (March-April, 1969), pp. 52-54. 20 prices should not accompany what he calls "illusory earnings growth."1 Graber assumes that new information does not result from the change; the old measurement pre- sents better information than the new.2 Graber's results are also rather inconclusive when judged on practical grounds as well as suffering from the questionable assumption mentioned in the previous paragraph. The sample of firms is small and non-random. Graber does not report the prices used for computing the price changes in the period subsequent to the accounting policy change but does for the period of the policy change. Upon exami- nation of these prices, substantially all of the average increase in price reported for that period is accounted for by five of the fourteen firms. Purdy, Smith and Gray examined "the impact which placement and method of disclosure of deviation statements [from Accounting Principles Board Opinions] had upon the retention of information presented in financial statements" lIbid., p. 52. 2In the February, 1970, issue of the Journal of Accountancy, Graber's article is reviewed. The reviewer, a member of the AICPA staff, states "Mr. Graber is, of course, making a critical- and questionable-assumption: He assumes that the earn- ings figures before considering the accounting c anges result in a fairer presentation of the results of the qperations of these fourteen firms than the earnings figures after making the changes." [The emphasis is that of the Journal of Accountancy reviewer. ] 21 and concluded "that placement and method made little 1 The authors cite the test situation research 2 difference." design as a limitation of their study. Many research works contributed to the development of the theoretical and methodological structure of the research presented in subsequent chapters. Their primary contributions are noted at this point. The studies concerning efficient capital markets aided in developing the theoretical model.3 Specifically, these studies provided theoretical formulations and empirical evidence supporting the assumption that stock prices fully reflect all information available to the investors in securities and that the adjustment by inves- tors to new information is rapidly reflected in stock price movements. 1Charles R. Purdy, Jay M. Smith, and Jack Gray, "The Visibility of the Auditor's Disclosure of Deviance from APB Opinion: An Empirical Test," Empirical Research in Accounting: Selggted Studies, 1969, a supplement to Volume VII of the Journal of Accounting Research, p. 12. 21bid. 3Examples are: Eugene F. Fama, "The Behavior of Stock Market Prices," The Journal of Business, XXVIII, No. 1 (January, 1965), pp. 34-105; Paul H. Cootner, "Stock Prices: Random vs. Systematic Changes," in Frontiers of Investment Analysis, ed. by E. Bruce Fredrikson (Scranton, Pennsylvania: International Textbook Company, 1966), Part VI, pp. 489-510. For an excellent and thorough review of the theoretical formulations and empirical evidence con- cerning efficient capital markets see Eugene F. Fama, "Efficient Capital Markets: A Review of Theory and Empirical WOrk," Journal of Finance, XXV, No. 2 (May, 1970), pp. 383-417. 22 King's work concerning the tendency of stock prices to move together as general economic conditions changed enabled this "general market effect" to be identified and remoVed from the test statistic.1 Thus, the probability of attaining significant research results is increased by removing an effect which may have obscured the true rela- tionship. — Beaver's research is of particular value as it demonstrated that investor's reaction to accounting in- formation is rapid thus supporting the efficient capital market assumption.2 Specifically, Beaver found that in weeks in which earnings announcements appeared, significant stock market reactions occurred as compared with price movements in other weeks.3 Thus, Beaver's work suggests that if the time period when the investor becomes aware of an accounting principle change can be isolated, the investor's reaction as reflected by stock price movements may provide evidence as to the effectiveness of the con- sistency exception communication network. 1Benjamin F. King, "Market and Industry Factors in Stock Price Behavior," The Journal of Business, XXXIX, No. 1, Part II (January, 1966), pp. 139—190. 2William H. Beaver, "The Information Content of Annual Earnings Announcements," Empirical Research in Accounting: Selegted Studies, 1968, a supplement to Volume VI of the Journal of Accounting Research, pp. 67-92. 3 Ibid., p. 84. 23 chpe and_Approach of the Research Chapter I’is primarily concerned with a statement of the purpose and justification of the research proposed herein. Chapter II scrutinizes certain theoretical con- siderations pertaining to user response to information. Specifically, the theoretical relationship among informa- tion inputs, covert investor responses and overt price responses is examined. From the theoretical considerations presented in Chapter II, a relationship between information inputs and subsequent security price response is hypothe- sized and supported by empirical evidence. The conclusion deduced is that one can observe security price movements in the time period subsequent to the introduction of data stimuli and, thus, make inferences concerning the informa- tion content of the data stimuli. Utilizing the conclusion of Chapter II, the methodology presented in Chapter III identifies when users of financial data, specially common stock investors, could be notified through the consistency exception channel of* communication about the occurrence of accounting principle changes. Then, in Chapter IV, the results of statistical comparisons of price response between a control group I containing firms without measurement rule changes and an experimental group containing firms with accounting 24 principle changes are reported. These statistical results allow inferences to be made concerning the information inpact that can be attributed to the differential data concerning acCounting measurement changes. The research concludes with Chapter V, a summary of the results with attendant conclusions and recommendations. Thus, the research presented herein is somewhat similar to both the Graber and Purdy, gE_§l. studies. That is, the research focuses on user response utilizing price changes as Graber did. But, unlike Graber, general market effects are removed from the price change leaving a residual which reflects factors unique to a particular firm. Also, rather than using a year time period over which to measure the price change as Graber did, specific identification is made of the time period in which the investor could acquire knowledge of the measurement rule change thus isolating the response to the information concerning the accounting principle change. Large sample sizes are also utilized to overcome the practical defects of Graber's research. And, since the intent and construc- tion of this research are quite different from Graber's, his questionable assumption which states that the old measurement rules present fairer results than the new rules is avoided. 25 This research deals with the effectiveness of the communication aspect of an accounting reporting standard as did the Purdy, Smith and Gray study. But this study focuses on providing empirical evidence on the effective- ness of the consistency exception standard rather than on the reporting standard stimulus used by Purdy, gE_gl. The research design is also quite different from the Purdy, eE_§l, design. By using historical price responses which occurred as a result of actual economic decisions and processes with all the attendant conditions of benefit and penalty which accrue to the investor, the limitation of the test situation faced in the Purdy, et_31, study is surmounted. Summary Chapter I states the purpose of this research is an attempt to provide empirical evidence as to the effec- tiveness of the notification network embodied in the consistency exception standard of reporting. From a theoretical framework, Chapter I examines criteria that a measurement change should meet to be characterized as "better." Using this framework and the measurement criteria, the inherent subjectivity of determining whether a new accounting measurement is better than the old measurement is examined. The conclusion reached from the 26 examination is that the second auditing standard of reporting provides a crucial communiCation link in informing users of financial statements of the occurrence of an accounting policy change. Since the acc0unting profession has no knowledge as to the effectiveness of this critical notification bridge, the justification for this study is to provide empirical information as to whether this crucial link serves its purpose or whether a communication problem exists. Chapter I also contains a review of the pertinent literature preceding this research effort as well as a discussion of the scope and approach of the research. CHAPTER II THEORETICAL CONSIDERATIONS Introduction By specifying a theoretical model of the behavior of one group of financial statement users, investors in a company's common stock, a research methodology can be con- structed (in Chapter III) that will relate their behavior to information about accounting measurement changes. But, first, the following model is specified and used to demonstrate possible user reaction when information about accounting policy changes exists and does not exist. Specification of an Investor Behavior Model Most forms of investor models of behavior can be expressed in the following symbolic manner. pvo = f(E[EPSi]) where: PV = present value of a common stock at the beginning of time period one, EPS. = future earnings per share of the firm issuing the security in time period i, i=1, 2, 3, o o 0' n] 27 28 El ] = expected value, f( ) = a specification of the functional relation- ship between the present value and expected future earnings. This model states that the present value of a security bears a specified functional relationship to the expected future earnings stream of the firm issuing the security. For purposes of this research, the general model illus- trated above is a necessary assumption. However, a more explicit model which is a special case of the general model, but which is not a necessary assumption to this research, may prove useful in examining user reaction to information about accounting changes. The special case model is formulated as: E[(DPSi)] n PVO = Z 1 i=1 (1+ri) but, since E[(DPSi)] E[(EPSi(1-bi))], then 2 E[(EPSi(1-bi))] PV 1 i=1 (1+ri) O where: PVO = present value of a share of stock at the beginning of time period one, DPSi = dividends per share in time period i, EPSi = earnings per share in time period 1, El ] = expected value, n = time horizon (infinite, in this case), 29 r. = discount rate (opportunity rate) in time period i, bi earnings retention rate in time period i. This model states that the present value (PV) of a share of stock at the beginning of time period one (1) is equal to the discounted value of the expected earnings stream of the investor which in the model above is an infinite stream of dividends. Since all investors are earning their opportunity rate or greater in equilibrium, the market value (MP) of this share of stock equals the pres- ent value (PV). "Intrinsic Value" Adjustments as Reflected by the Investor BehaVior Model Description of the Process of- Intrinsic Value Adjustments Fundamental economic factors, such as asset and liability structure, management capabilities and earnings of a firm, determine a stock's "intrinsic" or true economic value. But, as new information concerning these fundamen- tal factors becomes known to the market, investors modify or change their expectations as to future earnings streams or modify or change their risk preferences which are re- flected in discount rates or both. Since expectations of future earnings and discount rates serve as inputs to the decision models illustrated in the previous section, this modification process leads to a new present value for a 30 security. The investor compares this reformulated present value with the current market price of the security and, presumably, makes buy, hold or sell decisions based on this comparison. The buy-sell-hold decisions are reflected in the security's price movements, and these price move- ments, based on the new information, reflect adjustment to a new intrinsic value. Im lications of the Intrinsic VaIue Adjustment Process If the description of the intrinsic value adjust- ment process is valid, then a relationship between a specific price movement of a security and an information input exists. If the time period during which the investor becomes aware of the new information is sufficiently isolated, then a causal relationship could exist between this new information and subsequent price movements. Since price movements reflect the results of covert, but conscious, economic decisions by investors, the informa— tion content of a message may be inferred if a price move is identified with a particular information input. Em irical Evidence Supporting the Ingormation Content Hypothesis Random Walk Hypothesis.--The research concerned with the random walk or efficiency of capital markets provides evidence that stock price changes have an 31 appearance of randomness and that prices fully reflect all available information.1 Specifically, the random walk research states that prices randomly fluctuate around an intrinsic value; therefore, price changes are independent in the statistical sense. Symbolically, this random process is: Pt = IVt + et where: Pt = price of a security at time period t, IVt = intrinsic value of a security at time t, et = the random disturbance term at time period t. As new information enters the market, investors revise the inputs into their decision models or make adjustments in their decision models or both and take action based on the results of this process. This process leads to new intrinsic value levels represented symboli- cally as: . IVt+l + e "U II t+l t+l ' where: P = price of a security at time period t+l, t+l IVt+1 = intrinsic value of a security at time ,period t+1, e = the random disturbance term at time period t+l t+1 1Eugene F. Fama, "Efficient Capital Markets: A Review of Theory and Empirical Work," Journal of Finance, XXV, No. 2 (May, 1970), pp. 383-417. 32 Since the random disturbance terms have a mean expectation of zero, the change in price from period t to period t+1 is equal to the change in intrinsic value from period t to period t+1. Symbolically, the change in price from period t to period t+1 is: Pt+1 ' Pt = IVt+1 + et+1 ' (Ivt + et) APt+1,t = (Ivt+1 ’ IVt) + (et+l ' et) E[APt+1,t] = E[(Ivt+1 ' Ivt)] + E[(et+l ‘ et)] but, since E[(e - et)] equals zero, then t+1 AP = IV - IV t+1,t t+1 t where: Pt+l’ Pt’ Ivt+l’ IVt, et+l' et are defined as before, and AP = the change in price between period t t+1,t and period t+1, E[ ] = expected value. This mathematical process supports the implica- tion drawn in previous sections of this chapter. That is, if the time period when an investor becomes aware of new information can be isolated, then conclusions as to the covert reactions of investors to this information can be made by observing the overt price adjustment to a new intrinsic value. 33 Discussion of the Beaver Study.--Several studies could be examined to illustrate support for investor adjustment to new information.1 William H. Beaver's empirical study on the information content of annual earnings announcements is utilized for two reasons.2 One, the methodology used in the Beaver study is quite similar to that proposed in Chapter III of this study. And, secondly, the Beaver study's orientation is the same as that of the study reported here, a concern with the information content of accounting data. Beaver's study "is based upon a sample of annual earnings announcements released by 143 firms during the' 3 Beaver examined both volume years 1961 through 1965." and price adjustments to the news event, annual earnings announcements. Only the price adjustment portion of his research is examined here since those are the most germane findings in providing evidence to support the information adjustment process hypothesized in previous sections. lIbid. 2William H. Beaver, "The Information Content of Annual Earnings Announcements," Empirical Research in Accounting;_ Selegped Studies, 1968, a supplement to Volume VI of the Journal of AccountingResearch, pp. 67-92. 31bid., p. 70. 34 Beaver used the following transformation of weekly price changes as his test statistic:1 R _ 1n Dit + Pit , - - ———T--—- It Pit-l where: Dit = cash dividend "paid" on share of firm i in week t, Pit = closing price for share of firm i at end of week t, Pit-l = closing price at end of week t-l, adjusted for capital changes (e.g., stock splits and stock dividends). Rit is the natural logarithm of a price relative and is the rate of return for a security assuming continuous compound- ing. Next, Beaver regressed these weekly logarithmic price relatives against the Standard and Poor's Industrial Price Index. The following simple linear regression model was used:2 Rit = ai + biRMt + uit where: Rit is computed as shown above, and (SP)t RMt a 1“ (Spit: I = closing value of the Industrial Price Index at end of week t, (SP)t 1Ibid., p. 73. 21bid., p. 78. 35 (SP) a closing value of the index at end of t-l week t-l, a. and b. = parameters that vary from security to 1 1 . . . security wh1ch relate past changes in each firm's security price to past changes in the general market index, u = the price residual which represents the portion of the individual security's price change that cannot be accounted for by effects of market-wide events reflected in the market logarithmic price relative, RMt' it Using the estimates of ai and bi’ Beaver removed the general market effect that King had found to be significant from each weekly logarithmic price relative as follows:1 ujt = Rjt ’ ai ' biRMt where: i = index of the number of firms in the study, i=1, a o 0' 143; j = index of the number of earnings announcements, j=1, . . ., 506, t 2 index of time, t = -8, . . ., +8, where the week of the annual earnings announcement is designated as week zero. The residual, ujt' was squared and divided by the variance of the residual for its firm as follows:2 _ 2 1Benjamin F. King, "Market and Industry Factors in Stock Price Behavior," The Journal of Business, XXXIX, No. 1, Part II (January,Il966), pp. 139-190. . 2Note that by squaring the residual, Beaver gave greater weight to firms with larger price changes. Beaver makes no comment as to the effects on his research that this squaring process might produce. 36 where: i, j and t are defined as above. The above process results in a ratio, th. is found by averaging across j, the number of The average ratio, fit' earnings announcements, for each of the seventeen weeks of the period, i.e., the eight weeks before and after the annual earnings announcement week and the week of the annual earnings announcement. Beaver's a priori prediction was "the mean [ratio] . . . will be greater than one during week 0 [the week of the annual earnings announcement], if earnings "1 Based on the reports possess information content. analysis of his results, Beaver concluded that "the behavior of price changes uniformly supports the conten- tion that earnings reports possess information content."2 The significance of Beaver's study is three-fold. One, evidence is provided demonstrating that intrinsic value adjustments do occur as the result of information stimuli. Secondly, the results of Beaver's study demon- strate that this adjustment process is rapid; therefore, the stock market is reasonably efficient. And thirdly, Beaver's results suggest that the causal link between lBeaver, p. 79. 21bid., p. 32. 37 information inputs and subsequent price responses can be isolated and inferences made as to the information content of accounting data. The Information Content of Consistency Exception Opinions Introduction At the time an accounting policy change occurs, management makes, at least implicitly, a subjective evalua— tion as to whether a better or worse measure would result. The second auditing standard of reporting is used to supply users with notice of accounting policy changes so that the "market" is not blindly dependent on management's judgment. That the second auditing standard of reporting exists is prima facie evidence that the accounting profes- sion feels that users need information about the accounting policy change in order to re-evaluate the firm and to adjust their decision processes if necessary. But the pro- fession does not know if the communication network concern- ing the policy change is functioning. If the conclusions concerning the relationship between information inputs and stock price movements are valid, inferences can be made concerning the effectiveness of the communication network provided by the consistency exception in the auditor's opinion by examining the overt price responses that result from investors' decision-making processes. 38 Investor Response to Consistency Exceptions If the information stimulus, the consistency exception, is not reported or is not perceived by the investor, no re-evaluation of the present value of a security would result; the stimulus necessary for a re- sponse is lacking. Thus, no measurable price response results if notification of the accounting change does not occur. The accounting profession is faced by an apparent communication failure. If the consistency exception network is function- ing, however, two hypothesized user reactions could occur. One, the user is furnished with information which causes revision to the inputs of his decision model. These revisions lead to present values different from those that would result if no notification of the accounting change is received. Based on the comparison between this new present value and the current market price of the security, the investor takes economic actions which produce price movements that differ from those that would occur if notification of the change is not communicated. The second reaction that could occur if the con- sistency exception network is functioning is that the new informatiOn is not significantly inconsistent with previous? investor expectations to.cause revision in the inputs to 39 his decision model. Therefore, no measurable market price adjustment occurs. This second hypothesized user reaction that could occur if the notification network is functioning is the result of the investor possessing a materiality function. Materiality considerations are the subject of the next section. But the discussion of this section concludes with the comment that if the effect of the accounting change is communicated to the investor and is significantly' inconsistent with his expectations, then revision in ex- pectations or discount rates or both would lead to measurable market price responses differing from those price responses that result when the information stimulus is lacking. Investor Materiality Function To attain economy in preparation of financial statements, the accountant must be concerned with the materiality of an item. Kohler defines materiality as the characteristic attaching to a statement, fact or item whereby its disclosure or the method of giving it expression would be likely to influence the judgment of a reasonable person.1 1Eric L. Kohler, A Dictipnaryfor Accountants (Englewood Cliffs, New Jersey: Prentice—Hall, Inc., 1952), pp. 267-268. 40 Hicks defines the essence of materiality as "if it doesn't really matter, don't bother with it."1 Just as the accountant is concerned with the trade- off between cost in terms of preparation of financial statements and the amount of information included in the statements, the investor must also be concerned with the cost in terms of effort expended examining and analyzing the financial statements and the incremental information gained by the effort. The investor exists in an environ- ment containing a large amount of data stimuli. To be efficient in decision-making, the investor may establish a materiality function. If the effect of the accounting principle change relative to some base item does not exceed a certain level, the investor assumes that the effects are minimal and makes no adjustments in the inputs to his decision process even though he is notified of the change. Thus, the a priori outcome of a differential price effect which was hypothesized in the previous section would not occur. But as the materiality of the effect increases, the expectation is that differential price responses would be discernible. Importance of an accounting principle change can be viewed in two ways; the magnitude effect discussed in lErnest L. Hicks, "Materiality," Journal of Accounting Research, II, No. 2 (Autumn, 1964). p.I158. 41 the previous paragraph is one. The second view concerns the nature or the type of accounting principle change. In their decision processes, investors may have different rankings for the importance of an item as a result of the item's particular nature. This interest in an item could exist because of the investor's education, previous experi- ence, and personal biases. Or current financial news stimuli may create an interest in an item. News concerning Accounting Principle Board Opinions such as the opinion on pension costs which led to accounting principle changes by some firms serves as an example of bringing investor's attention to particular types of measurement rule changes. Thus, one might also expect differential price effects if the investor is notified of the occurrence of a measure- ment rule change due to the nature or type of accounting principle change. Methodological Implications The theoretical discussion of this chapter implies a particular methodological approach for testing the notification power of the second auditing standard of reporting. This approach is: 1Accounting Principles Board, APB Opinion No. 8: Accounting for the Cost of Pension Plans (New York: American Institute of Certified Public Accountants, November, 1966). 42 1. Identify a group of firms which have qualified auditors' opinions concerning accounting princi- ple changes. 2. Identify a group of firms without qualified opinions concerning accounting principle changes. 3. For both groups, make observations of price changes during a news event period, the annual earnings announcement, when investors are responding to the information contained in the announcement. Note if the earnings announcement contains information about the accounting principle change. If no information is communicated about the change, hypothesize that there is no difference in investor response between the two groups of firms. 4. Identify the subsequent time period when the investor could become aware of the fact that one group of-firms has made accounting principle changes. Make observations of price changes for both groups of firms during this period. Hypothe- size a differential informational effect, that is, a difference in investor response between the two groups of firms. The methodology presented in Chapter III follows the general structure proposed in steps one through four. Summary Chapter II discusses the theoretical formulations on which the methodology presented in the next chapter is constructed. In Chapter II, a general model of investor behavior is specified, and an examination of price re- sponses as a result of adjustments in an investor's decision process is performed. A causal link among information inputs, covert investor responses, and overt price response is hypothesized, and empirical evidence is 43 offered in support of this hypothesis. This hypothesis' implications for the area of accounting principle changes are presented. Also, consideration is given to the con— cept of materiality, its relationship to an investor's decision process and its implications for the area of accounting principle changes. Lastly, the methodological structureimplied by the theoretical considerations pre- sented in this chapter is discussed. CHAPTER III HYPOTHESES AND METHODOLOGY Introduction Chapter III is concerned with two topics. Firstly, general hypotheses concerning the information content of consistency exception opinions implied by the theoretical formulations discussed in Chapter II are presented. Secondly, the methodological structure of the research reported herein is presented. This structure includes a definition of the information stimulus to which this study is limited, identification of the populations of firms under study, definition of the sample selection criteria and procedures, and the development of the test statistics used to estimate the pOpulation parameters proposed in the statistical hypotheses. A discussion of the data sources used, the statistical hypotheses and the statistical tests employed as well as the statistical significance level specified are also included in the methodological portion of this chapter. The chapter concludes with a discussion of the limitations of the methodological structure. 44 45 General Hypotheses Principal Hypothesis Beaver's research demonstrates that annual earnings announcements cause significant stock market reactions in the week that they appear as compared with price movements in other weeks.l Presumably, annual earnings announcements contain information which causes changes in investor expectations. These earnings announcements are short summaries of financial data such as yearly sales and earnings which will be fully reported in subsequent annual reports. A majority of firms that make accounting principle changes do not report these changes in their annual earnings announcements. Thus, a notification network concerning the measurement rule change which is necessary for the investor to appropriately adjust his decision process does not exist; the stimulus necessary for the arousal of a response is not present. Therefore, the investor reacts to the earnings number reported in the annual earnings announcements as if the entire number reflects the results of real economic transactions. 1William H. Beaver, "The Information Content of Annual Earnings Announcements," Empirical Research in Accounting: Selected Studies, 1968, a supplement to Volume VI of-the Journal of Accounting Research, pp. 67-92. 46 Subsequently, when firms' annual reports are issued, the notification network of the consistency ex- ception standard serves as a communication network by notifying the investor that an accounting principle change occurred. Thus, investors in securities of firms with accounting principle changes who do not receive notifica- tion of the change at earnings announcement date must perform subsequent evaluations upon the receipt of the annual report. If these subsequent evaluations lead to changes in future earnings expectations or discount rates or both, one would hypothesize that price movements for the securities of those firms that notify the investor of the accounting principle changes at annual report date would differ from security price movements for firms that notified investors of their measurement rule changes at earnings announcement date. This hypothesis is consistent with and deduced from the information content hypothesis developed in Chapter II. Firms without accounting princi- ple changes and firms notifying investors of their accounting changes at annual earnings announcement date report no "new" information, ceteris paribus, at annual report issuance date. But firms that notify investors about their accounting principle changes in their annual reports rather than in earnings announcements divulge "new" information. Thus, price adjustments for the group notifying the investor through the annual report medium 47 should differ from price movements for the other two groups of firms. The principal hypothesis of this research is: Null Hypothesis, Hi: The consistency exception communication network in the independent auditor's opinion in the annual report does not notify investors of the occur- rence of an accounting principle change(s). Alternative Hypothesis, Hi: The consistency exception communication network in the independent auditor's opinion in the annual report is effective in notifying investors that an accounting principle change(s) occurred. Materiality Hypotheses Introduction.--The principal hypothesis presented in the previous section represents a broad approach in determining the communication effectiveness of the con- sistency exception standard. In the "real" world, in- vestors, assailed with data stimuli, are likely to develop a materiality function. These materiality functions represent trade-offs between cost in terms of effort expended in acquiring information and the incremental information gained by the effort. Therefore, hypotheses are proposed which provide information concerning the investor materiality function hypothesis and the 48 notification effectiveness of the consistency exception standard. These hypotheses are exploratory as no empirical evidence can be cited in support of an investor materiality function. But the concept of an investor materiality function has an intuitive appeal. And the accountant's concept of materiality is an obvious parallel to the concept of an investor materiality function. Materialitnyagnitude Hypothesis.--Hicks recommends to accountants that magnitude or size effects are not material if these effects are ten percent or less relative to some base item.l Hicks' Opinion is that effects of twenty percent or greater are obviously material.2 These percentages leave a large area in which the judgment of accountants is of importance. Although Hicks' recommenda- tions on the magnitude effects of materiality are offered to the accounting profession, the implication is that investors' materiality levels would be congruent. Hicks' materiality guidelines are used to develop the magnitude measure utilized in this study. Specifically, the magnitude measure is defined as the absolute percentage magnitude effect that the accounting principle change has 1Ernest L. Hicks, "Materiality," Journal of Accounting Research, II, No. 2 (Autumn, 1964), p. 162. 2 Ibid. 49 on current reported net income relative to the simple average of reported net incomes for the previous five years. Mathematically, this percentage is computed as follows: P- m 5 Z Z./5 1=l 1 Y=Z -Z c on co where: P = the percentage effect, ch| = the absolute value of the change in the current year's reported net income caused by the accounting principle change, Zcn = the current year's reported net income computed using the new accounting principle(s), Z = the current year's reported net income computed using the old accounting principle(s), Zi = net incomes of the previous five years, i=1, 0 o o, 50 The reported net incomes of the previous five years are those net incomes presented in the comparative finan- cial summaries of annual reports for the firms in the study. Net income after taxes but before extraordinary items is used if disclosed, otherwise the net income figures shown are used. Firms with negative five year average net incomes are excluded from consideration. 50 In some instances, the independent auditors state that the effect of the accounting principle change on net income of the current period is "immaterial," "insignifi- cant," or has no effect. These instances are arbitrarily assigned a zero magnitude effect. The statement of the hypothesis concerning magni- tude or size effects is: Null Hypothesis, Hg: Investors do not possess materiality functions based on magnitude or size. Alternative Hypothesis, Hi: Investors g2 possess materiality functions based on magnitude or size. Although of an exploratory nature, the test of the magni— tude hypothesis stated above not only provides information concerning investor materiality functions but also allows inferences to be made concerning the notification effective- ness of the consistency exception standard. Materiality-Type of Accountinngrinciple Change Hypothesis.--A stratification design based on the nature or type of accounting principle change may also aid in providing evidence as to investor materiality functions and the notification power of the consistency exception standard. Factors such as education, previous experience, personal biases or current news events may create investor interest in certain types of accounting principle changes. 51 Because of this interest, investors may respond differently to various types of accounting principle changes. To implement the stratification design based on the type of accounting principle changes, it is necessary to exclude from consideration those instances where more than one accounting change occurred. The accounting principle change reported in the annual reports of those firms with one change are classified into the following type of change 1. Change 2. Change 3. Change 4. Change 5. Change 6. Change 7. Change policy 8. Other. categories: in in in in in in in of inventory policy. depreciation policy. investment credit policy. deferred income tax policy. consolidation policy. pension policy. expense, capitalization and amortization various items. The hypothesis concerning type of change effects is: Null Hypothesis, H3: Investors do not possess materiality functions based on the type of accounting principle change reported. Alternative Hypothesis, Hg: Investors do possess materiality—functions based on the type of accounting principle change reported. 52 This type of change hypothesis is completely exploratory; therefore, no a priori expectation is given. But, if the null hypothesis is rejected, evidence will exist supporting both the investor materiality function hypothesis and the notification power of the consistency exception standard. Definition of the Information Stimulus In Chapter I, examples are given of the independent auditor's standard short-form opinion, and modifications to this standard form upon the occurrence of an accounting principle change are illustrated. These modifications to the standard short-form opinion are referred to as the exception type opinion and the retroactive type. This study concentrates on the exception type opinion and excludes the retroactive type from consideration. Two reasons exist for placing this limitation of scope on this research. One, as stated in Chapter I, firms with retroactive type opinions are considered to be con- sistently applying their accounting principles while firms with exception type Opinions are not considered to be consistent in their application of accounting principles. Secondly, by limiting the research to the exception type Opinion, 3 uniformity of stimulus observed by investors is obtained. This uniformity of stimulus in the research is introduced in hOPGS'Of attaining significant results. 53 Identification of the Universe and Samples Population Criteria The universe or population of firms considered in this research meet the following criteria: 1. The firms appear in the following editions of the AICPA publication, Accounting Trends and Tech- niques. a. Twentieth: This edition lists firms with fiscal years ending in the period January 11, 1965 to January 10, 1966. b. Twenty-first: This edition lists firms with fiscal years ending in the period January 11, 1966 to January 10, 1967. c. Twenty-second: This edition lists firms with fiscal years ending in the period January 11, 1967 to January 10, 1968. d. Twenty-third: This edition lists firms with fiscal years ending in the period January 11, 1968 to February 10, 1969. 2. The firms meeting criterion one which are listed on the New York Stock Exchange (NYSE), except that (1) firms leaving the NYSE because of de- listing, merger, or listing on other exchanges must have price data available for four months after the end of their fiscal years, and (2) firms just listed on the NYSE must have price data available for one month or more prior to the end of their fiscal years. Each edition of Accounting_Trends and Techniqnes discloses information concerning the financial reporting practices of a sample of six hundred firms that have fiscal 1American Institute of Certified Public Account— ants, Accountin Trends and Techniques in Published Annual Re orts, 20th-4 rd editions (New York: American Institute of Certified Public Accountants, 1966—1969). 54 years ending during a given period of time. These Accounting Trends and Techniques firms are used because they apparently are representative of the financial reporting practices of American corporations in the opinion of the AICPA. Criterion two restricting the study to NYSE firms is used since the number of American Stock Exchange firms appearing in Accounting Trends and Techniques is relatively small. The price date availability requirements listed in criterion two are to ensure that the annual report issuance date and the annual earnings announcement date fall in the test period. The Securities and Exchange Commission requires annual reports of listed firms to be issued not later than 120 days after the end of the fiscal year.1 Utilizing these criteria, the total population of firms numbers 1,978 or 505 firms with fiscal years ending in 1965, 509 firms with fiscal years ending in 1966, 489 firms with fiscal years ending in 1967, and 475 firms with fiscal years ending in 1968.2 These results are presented in Table l. 1New York Stock Exchange Guide, Volume III--Related Laws and Regulations (New York: Commerce Clearing House, Inc., 1962), p. 6183. 2From year to year, the composition of the 600 firms listed in Accpnntinngrends and Techniques remains fairly stable. Thus, technically, the popfilation does not consist of 1,978 different firms. But, in a statistical sense, Firm A in 1965 can be treated as "different" from Firm A in 1966. 55 maa.a new awe mom mom conumanmom was an aspen mo Hansen apnea an m p m e mapmunuo sunnnnmafim>m pump manna Oceania no: How pmpsHoxm mauwm mo acnesz mma.a has mme «Hm aom pause mmcmnoxm nooum gap» 362 «0 nonssz moe mma mod pm Hm pmumcnsnam manna mmcmcoxm gooum snow 3mz|coc mo umnfisz ooe.~ com com com com mmswnanoma pap pecans manuapooom an pmumna manna no “mass: annoy mumps mums mean span puma Had sauce new» samufimm mmswflcncma can mpcmna msflucsoood flHmmBHmo ZOHfldquOm mmB UZHMQmmd wm.nflZHzmmBma mdpchBmADmOm mmB ZH mZMHh m0 mmmSDZ H mnmdfi 56 Sam 1e Selection Criteria an Procedures Experimental Groups.--From the 1,978 firms in the population, the sample of firms that reported the occur- rence of accounting principle changes in an exception type auditor's opinion in their annual reports is identified by reference to the four editions of the AICPA publication, Accounting Trends and Techniques. This sample of firms is referred to as the experimental group and in total numbers 137, or, as Table 2 illustrates, 34 firms with fiscal years ending in 1965, 23 firms with fiscal years ending in 1966, 28 firms with fiscal years ending in 1967 and 52 firms with fiscal years ending in 1968. The experimental group is further divided by determining the point in time when investors were notified of the occurrence of an accounting principle change. Eng Wall Street Journal Index is consulted to determine the date the annual earnings announcement was issued.1 Then the annual earnings announcement that appeared in 322 Wall Street Journal is examined to determine if the announcement contained information about the occurrence of an accounting principle change.2 Fifty-eight firms of 1Dow Jones & Company, Inc., The Wall Street Journal Index, various volumes for the years I964-1969. 2Dow Jones & Company, Inc., The Wall Street Journal, various issues during the years 1964-1969. TABLE 2 IDENTIFICATION OF THE NUMBER OF FIRMS IN THE EXPERIMENTAL GROUP Calendar Year in Which Fiscal Year Ends Total All 1965 1966 1967 1968 Years Number of firms in the total population 505 509 489 475 1,978 Number of firms in the experimental group composed of those firms exhibiting an exception type auditor's opinion 34 23 28 52 137 Number of firms con— sistently applying accounting princi- 471 486 461 423 1,841 ples the 137 firms in the experimental group notify investors of the occurrence of an accounting principle change in their annual.earnings announcements. These fifty-eight firms are referred to as experimental group I, and the remaining seventy-nine firms that did not notify investors about the occurrence of an accounting principle change at annual earnings announcement date are designated as experimental group II. 58 Control Group.--From the population of firms consistently applying accounting principles, stratified random samples are drawn. The stratification design is based on time with the sample size of each of the stratum determined on a proportional basis. That is, the total number of firms in each stratum is divided by the total number of firms in the population. The resulting percent- age is multiplied by the total control group size to determine the sample size of each stratum. The control group size is arbitrarily determined by setting its size equal to 137, the experimental group size. This stratifica- tion procedure is used when the variability within the strata is approximately constant or when little is known about the variability within the strata; hence, the assumption is that the variability is constant.1 Mathematically, the computation of the stratifica- tion percentage is: where: PC = a fraction or percentage, M.c = the number of elements in each of the strata, 3 j, in the population, c, of firms consist- ently applying accounting principles where j = 1965, 1966, 1967, 1968, Nc = the total population. 1W. A. Spurr and C. P. Bonini, Statistical Analysis for Business Decisions (Homewood, Illinois: Richard D. Irwin, Inc., 1967), p. 324. 59 Table 3 shows the results of this process which is a group of 137 firms that consistently applies accounting principles. This control group of 137 firms is reduced by two firms, one with a fiscal year ending in 1965, and one with a fiscal year ending in 1966, when data concerning the issuance of their annual reports were found to be unavailable. Thus, the control group numbers 135 firms. Elimination of Retroactive Type Opinion Firms.-- The control group is composed of firms that consistently apply accounting principles, that is, firms that made no accOunting changes and firms that made accounting changes but retroactively adjusted their financial statements presented in their annual reports. The annual reports of the 135 firms in the control group are examined in order to eliminate firms exhibiting retroactive type auditor's opinions. Also, the 137 annual reports of the firms in the experimental groups are examined to ensure that the initial identification of this group as firms exhibiting exception type Opinions is reliable. The results of the elimination and verification process are: 1. Nine firms exhibiting a retroactive type auditor's opinion are eliminated from the control group of 135 firms. The remaining 126 firms are referred to as the control sample. 60 hmH Hem.a mummm Ham Hmuoe mm human. Hema\mme mmq mpma vm Heomm. Hema\ame Ape span pm mmmpm. Hemfl\pmv was mean mm qwmmm. Hema\aae Has mpma “om x bmav um 02 sofiumasmom mccm new» Hmomflm mnwm mumuum .Mfll 0:» ca musmEmHm sneak cw use» Hmcsmamu macho Houucoo .2 m mgmdfi mDOMU HomBZOU WEB ZH mSMHm m0 mmmEDZ mmB m0 ZOHBdZHZMHBMQ 61 2. No firms exhibiting a retroactive type opinion occur in experimental group I. Thus, the fifty- eight firms in this group are referred to as experimental sample I. 3. Nine firms exhibiting a retroactive type Opinion are eliminated from experimental group II. The remaining seventy firms are referred to as experimental sample II. Summary The previous sections of this chapter identify the population of firms considered in this research as those New York Stock Exchange (NYSE) firms that appear in the 1966, 1967, 1968, and 1969 editions of the AICPA publica- tion, Accounting Trends and Techniques. Also, this section reports the sample selection criteria and procedures used. Application of these criteria and procedures results in experimental sample I containing fifty-eight firms, and experimental sample II composed of seventy firms. Both experimental samples of firms have exception type opinions in their annual reports, but firms in sample I notify investors of their accounting principle changes at the annual earnings announcement date while firms in sample II do not. The control sample of 126 firms has the attribute of no accounting principle changes. Sample statistics from these samples are used to estimate the population parameters proposed in subsequent statistical tests. The results of these statistical tests aid in confirming or rejecting the hypotheses presented in this chapter. 62 Formulation of the Test Statistics Logarithmic Price Relative The theoretical considerations of Chapter II suggest a link between information inputs and subsequent security price movements. It is suggested that the security price response which results from covert economic decisions by investors could be utilized in making infer- ences about a related information input. The price re- sponse measure employed in this study is the natural logarithm of the weekly security price change. Specifi- cally, the logarithmic price relative is: D. + P. PRit = 1n g? It it-l where: PRit = the natural logarithm of the price relative of the 1th firm's common stock at time t, D. = the cash dividend per share of firm i in the it week t that the share went ex-dividend, Pit = the closing price for a share of firm i at the end of week t, Pit-l = the closing price for a share of firm i at the end of week t-l, adjusted for capital changes. PRi is a measure of the price change of a firm's common t stock; it is the rate of return for the week assuming continuous compounding. 63 Removal Of the General Markgt Effect from the Logarithmic Price Relative Introduction.-—The uncorrected logarithmic price relative reflects price changes caused by factors unique to the firm, by-industry factors and by general economic conditions. King's evidence demonstrates that the general economic or market effect is significant.1 Therefore, a simple linear regression model which has been employed quite effectively in other empirical works is used to eliminate the general market effect.2 King has found that on the average only eleven percent of a firm's security price change was explained by its related industry.3 Since this amount relative to the average market effect and the unique firm effect is minor, the industry effect is not removed from the logarithmic price relative. The Regression Mode1.--The linear regression model used to determine and remove the general market effect is: PRit = ai +‘biIRmt + eit 1Benjamin F. King, "Market and Industry Factors in Stock Price Behavior," The Journal of Business, XXXIX, No. 1, Part II (January, 1966), pp. 139-190. 2Examples are: William H. Beaver, "The Informa— tion Content . . . "; Eugene F. Fama, et al., "The Adjustment of Stock Prices to New Information," Inter- national Egonomicgeview, X, No. 1 (February, 1969), pp. IF21; Raymond Ball and Philip Brown, "An Empirical Evaluation of Accounting Income Numbers," Journal of Accounting_Research, VI, NO. 2 (Autumn, 1968), pp. 159-178. 3King, p. 156. 64 where: PRit = the natural logarithm of the price relative of the ith firm's common stock at time t, ai and bi»: parameters that vary from security to security which relate past changes in each firm's security price to past changes in the general market index, (SP) t . IR = 1n I__7__—- ; the natural logar1thm of the mt SP t- market price relative at time t, (SP)t = the closing value of Standard and Poor's Industrial Price Index at the end of week t, (SP)t_1 = the closing value of Standard and Poor's Price Index at the end of week t-l, eit = the logarithmic price relative residual or random distrubance term. The error or disturbance term, is expected to eit' conform to the following assumptions of the linear model. 1. eit has zero expectation. The periods in which earnings announcements and annual reports were issued are excluded in the regression because the expectation of eit during these periods is not zero. 2. The standard deviation of eit is the same for all values of IRmt; this is the property of homoscedasticity. 3. The eit are independent of each other, i.e., no autocorrelation. From the regression analysis, estimates of the parameters, a1 and bi' for firm i's security result. weekly price changes are used in the regression for a 1W. A. Spurr-and C. P. Bonini, pp. 564-565. 65 period of sixty-five weeks. The annual earnings announce— ment period is designated as week zero, and the regression period includes the quarter in which the earnings announce- ment is issued, the previous two quarters and the subsequent two quarters. This time period is arbitrarily selected. The relationship between any security and the market index cannot be expected to be stationary over time; therefore, the regression time period chosen is that which is most current in terms of the time period under study. Data gathering limitations are also considered. By using the estimates of the parameters, ai and bi’ and the actual value change in the Standard and Poor's Industrial Price Index for the periods in which earnings announcements and annual reports are issued, the general market effect can be deleted from the actual price change for firm i's common stock. In equation form, this can be represented as: eit = PRit ' ai ' biIRmt where: the natural logarithm of the price relatives of firm i's security and the §E§ndard and Poor's Industrial Price Index, respectively, at time period t, PRi and IRm t t time period t = either the end of the period the earnings announcement is issued or the end of the period the annual report is issued, a. and b. = the estimates of the parameters, ai 1 and bi. 66 The remaining residual, eit, presumably includes only the effects of firm i's stock price that are unique to firm i. In this study, it also includes an industry effect. But, as previously stated, King has found that on the average only eleven percent of a firm's security price change was eXplained by its related industry.1 Since this amount relative to the average market effect and the unique firm effect is minor, the industry effect is not removed. That is, the increase in exactness does not outweigh the in- creased costs of data gathering and computational problems. The reasons for selection of the Standard and Poor's Price Index are: 1. Weekly values of the index are easily obtainable. 2. Fama, et al. have found that the results using the Standard and Poor's Price Index in conjunction with the above linear regression model provided results that were in agreement with those using other types of indices for removing the market effect. Evidence Supporting the Use of the Regression Mode1.—-Strict conformance to the regression model's assumptions is not necessary since the model is robust. Yet, as a test check, the autocorrelation coefficient of determination which is the ratio of the variability lKing, p. 156. Eugene F. Fama, et al., "The Adjustment of Stock Prices to New Information," International Economic Review, X, No. 1 (February, 1969), pp. 1-21. 67 explained by the independent variable to the total variability of the dependent variable is computed.1 Under the assumption of no autocorrelation, these coefficients should be zero. The average autocorrelation coefficient of determination, r:, for the control sample is .027934, for experimental sample I is .024997, and for experimental sample II is .027601. These coefficients are sufficiently close to zero and provide assurance that no autocorrelation exists. The frequency distributions of the autocorrelation coefficients of determination are shown in Table 4. Findings on the Magnitude of the General Market Effect.--King's study of the relationship between general market movements as evidenced by a general market index and individual security price movements reveals that, on the average, thirty-one percent of the variation in an individual security's price change can be explained by the change in value of the market index.2 For the samples used in this study, the general market effect as shown by the coefficient of determination, r2, which is the ratio of the variance explained by the independent variable (the market lThe autocorrelation coefficients of determination are computed by lagging the logarithmic price relative residuals Of each firm by one week. For the specific computational form, see Appendix A. 2King, p. 151. 68 m Hopamo. ammemo. «mmnuo. «a mo osam> mmmum>a on mm ONH fiance o o q mmmm. I on. H m a mama. I oa. mm mm HNH ammo. I oo. mosmsvmum mosmsvmum mosmsvmum MM HH madfipm H mamsmm panama Hopscoo .ucmaonmmmoo map Hmucmawuomxm amucmfiwummxm mo mosHm> was mmcmm ZOHBdZHmeBmQ m0 mBZMHUHmKMOU ZOHB¢AMMMOUOBD¢ HEB m0 ZOHBDmHMBmHQ HUZmDOmmm v mnm4fi 69 index) to the total variance of the dependent variable is 1’ 2 Specifically, the average coeffi- not as significant. cient of determination for the control sample is .19770, for experimental sample I is .17881,.and for experimental sample II is .18397. Table 5 provides information on the frequency distributions of the coefficients of determina- tion for the three samples. Identification of the Annual Earnings Announggment Date and the Annual Report Issuance Date Annual Earnings Announcement Date.--The Wall Street Journal Index is used to identify the date that the annual earnings announcement is issued for each firm in the samples. The issue of the Wall Street Journal in which the earnings announcement appears is examined to verify 1See Appendix A for the specific computational form for the coefficient of determination. 2A possible reason explaining the differences in the average coefficient of determination reported by King and found in this study is that King used a monthly measure of price change while this study uses a weekly measure. Other studies utilizing a weekly measure of price change also report less market effect than that effect reported by King. For example, Beaver reports an average coefficient of determination of .06 for the New York Stock Exchange firms in his study (Beaver, p. 80). Also, May reports an average coefficient of .11 for the American Stock Exchange firms in his study (Robert G. May, "The Influence of Quarterly Earnings Announcements on Investor Decisions as Reflected in Common Stock Price Changes" (unpublished Ph.D. dissertation, Department of Accounting and Financial Administration, Michigan State University, 1970), p. 81)). 70 ammma. Hmmaa. oaema. «u «0 msam> mmmum>a on mm ONH annoy o o H mmmm. I om. o o h mmmm. I ow. ma OH mH mmmm. I om. ma «a on mmmm. I ON. ON ma mm mmma. I OH. ma ma . Hm mmmo. I oo. mosmsvoum mocmsvmum mocmsvmum an HH mamemm H mamemm mamsmm Honucoo .uGOAOHmmmoo ms» Hmucmswummxm HousmeHmmxm mo mmsHm> can manna ZOHBflZHSMMBMQ m0 mBZMHUHmmmou HEB ho ZOHBDmHmamHQ wozmboth m mnm<8 71 this date. The date found is used to identify the week designated as week zero in the regression analysis, and the general market effect is removed from the price change for-that week. The remaining residual, 8. includes 1t' effects unique to firm i and can be related to the informa- tion input, the annual earnings announcement. Beaver's research into the information content of annual earnings announcements demonstrates that price adjustment at earnings announcement occurs instantaneously when instan- taneous is defined as within the week of the announcement.1 Annual Report Issuance Date.--New York Stock Exchange (NYSE) firms are required to notify the Department of Stock List of the NYSE of the date that annual reports of listed firms are mailed to investors.2 For firms that make accounting principle changes, but do not notify investors of the changes in the earnings announcement, the annual report provides initial notification and information to the investor concerning the accounting principle change. Obviously, leakages of information may occur. But, with Securities and Exchange "insider ruleS" and NYSE lBeaver, pp. 67-92. Appendix B offers empirical evidence concerning market adjustment at annual earnings announcement week. 2New'York Stock Exchange, New York Stock Exchange Company Manual, p. A-69. 72 requirements concerning information releases, these leakages should be minimal and can be ignored. The Department of Stock List of the NYSE keeps a permanent card file of the date that the annual report of NYSE firms is sent to stockholders. Specifically, the Department records the date it receives a firm's annual report minus two days.1 This date is used to identify the week in which annual report data could have first reached investors. This week is referred to as the annual report issuance week. The price response to the annual report data is expected to be more diffused than the response to the annual earnings announcement. The method of identifying the annual report issuance date reported in the previous paragraph is partially responsible for the a priori ex- pectation of a diffusion effect. But two other reasons are primarily responsible for expecting a diffused price reaction to the annual report. One, many investors receive the annual report of the firm in which they have invested at a "street" address. That is, the annual report first goes to the brokerage firm through which the investor bought the stock, and, subsequently, the annual report is forwarded to the investor thus leading to a time lag 1Interview with Mrs. Adele Williams, staff member of the Department of Stock List, New York Stock Exchange, June 1, 1970. 73 between issuance and receipt. The second reason for eXpecting a diffused price reaction is that the annual report contains a greater quantity of data relative to the annual earnings announcement and, thus, requires greater time in evaluation before final investor decisions are formed and investor actions taken. In order to determine whether the price response to annual report data is diffused as suggested above, an annual report profile of price response is calculated in the following manner: I R -2 u. ;t=-8,-7’ o O o, 0, o o 0' 7,8 t 1.1 it I . a. uitaM 1t ;t=-8’-7, o o o, 0’ o o 0, 7,8 2 a. 1 = 1,2, . . ., I T31 f 1T M where: Rt 8 the average price change ratio in week t, “it a the ratio of the absolute value of the logarithmic price relative residual of the ith firm during week t and the average absolute value of the logarithmic price relative residual of the 1th firm, [git = the absolute value of the logarithmic price relative residual computed by using a simple linear regression model, I a the total number of firms or annual reports in the study; I a 254, 74 M a the total number of logarithmic price relatives for the ith firm excluding the annual earnings annOuncement week and the annual report issu- ance period. M is equal to fifty—eight if the annual earnings announcement and the annual report do not coincide, but M equals fifty-nine when the periods are coincident, 1., an index of the number of annual reports, T a an index of the number of residuals for the ith firm, t = an index for the number of weeks in the profile where the annual report issuance week is designated as week zero. In the absence of any new information in a given time period, one would expect the price change ratio of a given firm's common stock to be equal to the average price response for all comparable time periods for the same firm. Mathematically, where: Rt»a the average price change ratio in period t, l a the expected value of the average price change ratio, e a the random error term in period t. t The random error term in non-news periods is expected to have a mean of zero and be normally distributed. But, if new information stimuli occur in a particular time period, an adjustment to a new intrinsic value presumably occurs, and, thus, the actual price change ratio for the period when news events occur is expected to be greater than one (1). 75 Figure 2 presents the annual report profile for the 254 annual reports in this study.1 Zero week is the week identified by using the annual report issuance date obtained from the New York Stock Exchange. 1.25m 1.20d 1.15j 1.10- 1.051 1.004 .95H .90d .851 .BOT -8 -7 -6 -5 -4 —3 —2 -l 0 l 2 3 4 5 6 7 8 week Fig. 2.--Annual report profile for 254 annual reports The price change ratio, Rt’ for week zero and several weeks after week zero is greater than one but not significantly different from one in a statistical sense. But, since it is known that the annual report is issued during this period, the assumption is that this diffused price response occurred as the reSult of the information in the annual 1Appendix B offers empirical evidence concerning market adjustment during annual report issuance period for the three samples in the study. 76 reports. In the absence of new information, one would expect to find five consecutive weeks greater than one only three times in 100 trials if the price change ratio's error term, et, has a mean of zero and is normally dis- tributed.1 The five week period defined by zero and the four subsequent weeks shown in Figure 2 is defined as the annual report issuance period. This procedure is based on the a priori arguments for expecting a diffused price response to annual report data and on the statistical evidence that the occurrence of five consecutive price change ratios greater than one should rarely occur. The general market effect is removed from the price change for each of the five weeks of the annual report issuance period, and the remaining residuals, éit, are summed algebraically and can be related to the informa- tion input, the annual report which contains audited financial statements. The Test Statistics Transformations of the logarithmic price relative residuals, git' for the annual earnings announcement week and for the annual report issuance period are used to 5 (.55) (.5 . .03. ( iProbability of five consecutive weeks greater than one - ) 5 77 estimate the population parameters of the statistical hypotheses presented in subsequent sections. Specifically, the absolute values of these residuals are used and are formulated as: and, A E' eite 1te = where: Eite a the absolute value of the logarithmic price relative residual of the 1th firm's common stock at time t, eit = the logarithmic price relative residual of the ith firm's common stock at time t, t = the annual earnings announcement week, e - identification for the annual earnings announcement residual, where: 5 A E. = 2 e. 1Ta tal 1ta EiTa =='the absolute value of the 10garithmic price relative residual of the ith firm's common stock at time T, e = the logarithmic price relative residual of 1ta the ith firm's common stock at time t, T = the annual report issuance period, t = the index for the weekly periods over which the weekly residuals are algebraically summed; tal is the annual report issuance week identified by utilizing the date obtained from the New York Stock Exchange, a a identification for the annual report issuance period residual. 78 Data Sources Several of the data sources are mentioned in previous sections of this chapter. For completeness, however, a summary of the various sources is supplied at this point. Earnings data, information about accounting princi- ple changes as to type and magnitude and other miscellaneous data are gathered from the annual reports of the firms in the study.1 Earnings announcement dates appear in the Wall Street Journal Index; the earnings announcements 3 appear in the Wall Street Journa1.2' Weekly price in- formation for New York Stock.Exchange firms used in the regression is taken from the ISL Daily Stock Price Index.4 The weekly closing value of the Standard and Poor's 425 Industrials-Price Index used in the regression analysis is found in Barrens.5 The New York Stock Exchange (NYSE) 1Annual reports of the firms in the study for years 1964-1969. 2Dow Jones & Company, Inc., The Wall Street Journal Index, various volumes for the years l964-l969. 3Dow Jones & Company, Inc., The Wall Street Journal various issues during the years 1964-1969. 4Investment Statistics Laboratory, Inc., ISL Daily Stock Price Index: New York Stock Exchange (Palo Alto, California: Investment Statistics Laboratory, Inc.), volumes.for the years 1964-1969. 5Dow Jones & Company, Inc., Barron's, various issues during the years 1964—1969. 79 keeps a permanent card file on the date that the annual report of NYSE firms was sent to stockholders, and this source is used to identify the annual report issuance date. Statistical Hyptheses and Statistical Tests Introduction This section presents the statistical hypotheses which operationalize the general hypotheses proposed at the beginning of this chapter. Also, the statistical models employed in testing the hypotheses are discussed. Principal Statistical Hypothesis Introduction.--The principal hypothesis of this research is: Null Hypothesis, H1: The consistency exception 0 communication network in the independent auditor's opinion in the annual report does not notify investors of the occur- rence of an accounting principle change(s). Alternative Hypothesis, H1: The consistency exception a communication network in the independent auditor's opinion in the annual report is effective in notifying investors that an accounting principle change(s) occurred. 01 [H "1 80 To operatiOnalize this general hypothesis, a statistical hypothesis which tests for differences in price responses among the three samples during annual report issuance period is prOposed. A two-way analysis of covariance model is utilized which employs both the absolute value of the logarithmic price relative residual, Ei , of the Ta annual report issuance period and the absolute value of the logarithmic price relative residual, Ei , of the te annual earnings announcement week. The independent variables in the two-way analysis of covariance model are an information effect variable and a time period variable. Test of-Price Response During the Annual Report Issuance Period.--The following statistical hypotheses are proposed to test average price response during the annual report issuance period: 1. Test of treatment effect j: la _ -. Ho . cj - 0 for all 3 la la Ha . not Ho 2. Test of treatment effect k: 1b Ho . Bk lb 1b Ha : not Ho = 0 for all k 81 3. Test of interaction effects: 1c _ . Ho . ij - 0 for all jk 1c 1c Ha . not H0 The null hypothesis, Hia, states that there is no treatment effect j where j is defined as new information or no new information. In this research, j = 1 denotes a population composed of firms that report no accounting principle changes, and j = 2 denotes a population composed of firms notifying investors at earnings announcement date of their accounting changes. These two populations of firms report no "new" information in their annual reports, ceteris paribus. But the pOpulation denoted by j = 3 and composed of firms notifying investors of their accounting changes through the consistency exception in the auditor's opinion which appears in the firms' annual reports does report "new" information. Thus, this population is expected to differ from the two populations which report no "new" information. The null hypothesis, Hib, is concerned with deter- mining if there is a treatment effect k where k is defined as the time period or year. Since investors should be cognizant of significant information without regard to tbme period, the null hypothesis is not expected to be rejected. 82 The null hypothesis, Hie, is concerned with the interaction effect between treatments j and k. No inter- action effect is expected. Two-Way Analysis of Covariance Design.--If, as hypothesized, the consistency exception network notifies investors about the occurrence of an accounting principle change, the expectation is that systematic differences exist between the control sample-experimental sample I and experimental sample II during the annual report issuance period. The control sample and experimental sample I divulge no new information in their annual reports, thus, ceteris paribus, no "effect" or difference in investor response is expected. Experimental sample II does present new information, thus, an "effect" is expected. If a relationship between information input and price response really exists as is hypothesized and supported by empirical eVidence in Chapter II, then a difference in mean or average price reSponse is an indicator of a statistical relationship between the independent variable, information input, and dependent variable, average price response. This conclusion assumes that all possible other factors are controlled or that the influence of uncontrolled factors can be regarded as random error. To test the statistical hypotheses presented above, a two-way analysis of covariance model is utilized. 141.4 83 The model for the two—way analysis of covariance design is:1 = u + p(Xjk - X) + aj + Bk + ij + e EjkTa jk where: fijkTa = the average absolute value of the logarith- mic price relative residual for the ith firm's common stock; this is the category mean for category jk, u = the grand or over-all mean, NI jk = the covariate mean for category jk, XI = the grand or over-all mean of the covariate, p the population linear regression coefficient, a. = the effect associated with the particular treatment population j, Bk = the effect associated with the particular treatment population k, Y'k = the interaction effect created by the combination of treatments j and k, ejk = the random error term, j = a treatment population where treatment is defined as new information or no new informa- tion; j = .1, 2’ 3’ k = a treatment population where treatment is defined as the time period; k = 1965, 1966, 1967, 1968, T = the annual report issuance period, a = identification for the annual report issuance period price relative residual. 1B. J. Winer, Statisnical Pringiplee in Experi- mental Desi n (New York: McGraw-Hill Book Company, 1962), pp. - . Also, see Winer for the computational forms for the two-way analysis of covariance. 84 The covariate used in the above test is the absolute value of the logarithmic price relative residual, E at annual earnings announcement week. The two-way ite' analysis of covariance design requires the regression of the dependent variable, E. on the independent variable, 1Ta’ Bite' The estimates of the regression parameters are used as weighting factors to adjust the category mean, fijkTa' thus removing any differences among the samples on the covariate before comparing the samples on the dependent variable. The purpose of the regression is to remove any effects caused by initial non-homogeneity among the three samples at annual earnings announcement week. After adjustment of the category means as well as the error variance, a two-way analysis of variance is performed to test for treatment and interaction effects. The two-way analysis of covariance design essentially requires a stratification by type of information input and by year of the logarithmic price relative residuals, EiTa Table 6 demonstrates the two-way analysis of covariance deSign for the dependent variable, EiTa' used in this study. The values in Table 6 are the number of price relative residuals in each of the categories. If the null hypothesis that there is no difference in average price response among the control sample, experi- mental sample I and experimental sample II is true, then 85 TABLE 6 THE CATEGORY FREQUENCIES OF THE PRICE RESPONSE MEASURE USED AS INPUTS INTO THE TWO-WAY ANALYSIS OF COVARIANCE RESEARCH DESIGN Year Sample 1965 1966 1967 1968 Total Control 34 31 31 30 126 Experimental I 14 10 8 26 58 Experimental II 20 13 13 24 70 Total 68 54 52 80 254 the information effect variable, aj, will equal zero. The year effect, Bk, is expected to be zero as well as the interaction effect, ij' Thus, the grand mean, u, is the best estimate of any individual logarithmic price relative, EiTa' But if the causal network between information input and subsequent price response exists, then the information effect variable, aj, for experimental sample II will not equal zero. Thus, a difference in price response should exist between experimental sample II and the control- experimental sample I. The outputs of the two-way analysis of covariance procedures are F statistics relating to the three effects, information input differences, year differences, and inter- action differences. These F statistics can be tested for 86 significance, and if they are significant at a pre- designated significance level, the conclusion is that the three samples differ in average investor response as measured by the absolute value of the logarithmic price relative residual during the annual report issuance period. The assumptions underlying the use of the two-way analysis of covariance model are:1 l. The random error terms, e. , are normally dis- tributed with expectation of zero for each treatment-combination population jk. 2. The standard deviations of the random error terms are homogeneous. 3. The random error terms are independent within and across treatment combinations. 4. Treatment effects and regression effects are additive, and the proper form of regression equation is fitted to the data. Materiality Statistical Hypotheses Introduction.--The principal hypothesis of this research represents a broad approach to determining if the notification network of the consistency exception standard is effective. If the investor possesses a materiality function, the designs proposed below may provide informa- tion about these hypothesized investor materiality func- tions, and, in addition, may provide information concerning the effectiveness of the notification network. 11bid., p. 586. 87 Materiality-Magnitude Hypothesis.--This hypothesis states: Null Hypothesis, Hg: Investors do not possess materiality functions based on magnitude or size. Alternative Hypothesis, Hi: Investors dg_possess materiality functions based on magnitude or size. The statistical hypothesis used to operationalize this general materiality-magnitude hypothesis is: H23: b = 0 o a Hza: b >0 3 a where: ba = the regression coefficient for the population of firms that notify investors of their accounting principle changes only in their annual reports. 2a The null hypothesis, Ho , states that the population regression coefficient is equal to zero while the alternate 2a hypothesis, Ha , states that the population regression coefficient is positive. To test the statistical hypothesis, Hga , a simple linear regression model is used, and an estimate of the population regression coefficient, ba' is obtained. The regression model is: E. = a. + b.P. + e 1 1 1 1Ta iT 88 where: EiTa = the absolute value of the logarithmic price relative residual for the ith firm's common stock at time T, ai and bi = parameters that vary from security to security which relate past changes in each firm's security price to past changes in the independent variable P., 1 Pi a-the magnitude measure defined as the absolute percentage magnitude effect that an accounting principle change made by the ith firm has on current reported net income relative to the simple average of reported net incomes for the previous five years, e. = the random error term, 1T T a the annual report issuance period, a‘- identification for the annual report issuance period price relative residual. The sample regression coefficient, bi, is used to estimate the population regression coefficient, b. The assumptions underlying the use of the simple linear regression model are: 1. 2. 3. The eiT have zero expectation. The standard deviation of eiT is the same for all values of Pi’ The eiT are not autocorrelated.1 In essence, this statistical model is used to test the relationship between the dependent variable, price change, and the independent variable, the absolute value of the magnitude effect. The magnitude effect measure is 1W. A. Spurr and C. P. Bonini, pp. 564-565. F”. r! 89 an attempt to operationalize the concept of materiality. The a priori expectation is that there is a positive relationship between the dependent and independent vari- ables. That is, as the value of the magnitude effect measure increases, the value of the price change measure increases. The sample regression coefficient for experi- mental sample II will provide evidence concerning the relationship between the measure of the magnitude effect of accounting principle changes and the measure of price response. Materiality-Type of Accounting Principle Change Hypothesis.--This hypothesis states: Null Hypothesis, Hg: Investors do not possess materiality functions based on the type of accounting principle change reported. Alternative Hypothesis, Hz: Investors do possess materiality functions based on the type of accounting principle change reported. To operationalize this general materiality-type of accounting principle change hypothesis, a two-way analysis of covariance model is utilized. The inputs to the model are the absolute values of the logarithmic price relative residuals, EiTa' during the annual report issuance period for the firms in experimental sample I and experi- mental sample II. Essentially, the two-way analysis of 90 covariance model requires a stratification of the price response measure according to type of accounting principle change and type of information where firms in experimental sample I report no new information, ceteris paribus, in their annual reports while firms in experimental sample II report.information about their accounting principle changes for the first time. The categories for the types of accounting principle changes are: 1. Change in inventory policy. 2. Change in depreciation policy. 3. Change in investment credit policy. 4. Change in deferred income tax policy. 5. Change in consolidation policy. 6. Change in pension policy. 7. Change in expense, capitalization and amortization policy of various items. 8. Other. Table 7 illustrates the inputs into the two-way analysis of covariance design for the price response variable, E. . The values in Table 7 are the number of 1Ta price relative residuals in each category. Firms with more than one accounting change are excluded from this analysis, and in the actual test, categories four and eight of the type of accounting princi- ple change are excluded as the category.frequencies of experimental sample I are zero. 91 TABLE 7 THE CATEGORY FREQUENCIES OF THE PRICE RESPONSE MEASURE USED AS INPUTS INTO THE TWO-WAY ANALYSIS OF COVARIANCE RESEARCH DESIGN Type of Change Sample 1 2 3 4 5 6 7 8 Total Experimental Sample I 2 33 5 0 2 2 3 0 47 Experimental Sample II 12 16 5 5 9 7 2 4 60 Total 14 49 10 5 11 9 5 4 107 The model for the two-way analysis of covariance test is:1 EjkTa = u + p(xjk - X) + “j + Bk + ij + ejk where- fijkTa = the average absolute value of the logarith- mic price relative residual for the ith firm's common stock; this is the category mean for category jk, u = the grand or over-all mean, ijk = the covariate mean for category jk, i a the grand or over-all mean of the covariate, p = the population linear regression coefficient, a. = the effect associated with the particular treatment population j, = the effect associated with the particular 8 k treatment population k, lWiner, pp. 578-606. Also, see Winer for the computational forms for the two-way analysis of covariance. 92 Y'k a the interaction effect created by the 3 combination of treatments j and k, a the random error term, j = a treatment population where treatment is defined as new information or no new informa- tion; j = l, 2, k a a treatment population where treatment is defined as the type of accounting principle change; k =_l, 2, 3, 5, 6, 7, T = the annual report issuance period, a = identification for the annual report issuance period price relative residual. The covariate used in this study is the magnitude measure, Pi‘ Pi is defined as the absolute percentage magnitude effect that an accounting principle change made by the ith firm has on current reported net income relative to the simple average of reported net incomes for the previous five years. The two-way analysis of covariance design requires the regression of the dependent variable, E on the independent variable, Pi' The estimates of iTa' the regression parameters are used to adjust the category mean, fijkTa’ thus removing any differences between the samples on the covariate before comparing the samples on the dependent variable. After adjustment of the category means and the error variance, a two-way analysis of variance is performed to test for treatment and inter- action effects. 93 The specific statistical hypotheses used to test the general materiality-type of change hypothesis are: 1. Test of treatment effect j: 3a _ . Ho . aj - 0 for all 3 H33: not H3a a o 2. Test of treatment effect k: 3b Ho . Bk H3b: not H3b a o = 0 for all k 3. Test of interaction effects: 3c _ . Ho . ij — 0 for all jk 3c 3c H : not H a o Hypothesis 3a tests the differences in average price response for treatment j where j is defined as no new information or new information at annual report issuance. Hypothesis 3b tests the differences in average price re- sponse for treatment k where k is defined as the type of accounting principle change while Hypothesis 3c tests the interaction effect of these two treatments. Essentially, Hypothesis 3a is redundant in that the test of the principal hypothesis of the study is made to determine the information effect. Hypothesis 3b is a test which provides evidence concerning investor materiality functions based on type of accounting principle changes 94 while Hypothesis 3c provides evidence concerning the interaction effects between investor response to type of information and type of accounting principle change. The a priori expectation is that the null hypotheses will be rejected. The outputs of the two-way analysis of covariance are F statistics associated with the treatment and inter- action effects. These F statistics can be tested for their probabilities of occurring by chance. The assumptions underlying the use of the two-way analysis of covariance model are:1 l. The random error terms, e‘k, are normally dis- tributed with expectation of zero for each treat- ment combination pOpulation jk. 2. The standard deviations of the random error terms are homogeneous. 3. The random error terms are independent. 4. Treatment effects and regression effects are additive, and the proper form of regression equation is fitted to the data. The type of accounting principle change category means for experimental sample II are also compared to the mean price response of the control sample utilizing the Student's t test for differences between population means.2 The significance level of the computed t statistic for lIbid., p. 586. 2Hays, pp. 320-323. Also, see Hays for the computational forms for the Student's t statistic. 95 each comparison can be determined and compared with a pre-designated significance level. This test provides additional evidence about the existence or non-existence of investor materiality functions based on type of accounting principle change. Estimation For the actual computations made in this study, the parameters of the analysis of covariance model are reparameterized to reflect the contrasts of interest among the different groups that the a priori theory presented in Chapter II suggests. If the statistical results are significant at the chosen a priori significance level, confidence intervals can be placed on the least squares estimators of the population parameters that result from the reparameterized model. Significance Level The actual significance levels found in the study will be reported so that the reader may draw his own con- clusions if he so desires. But, in terms of recommending action, if necessary, a significance level of .05 will be required of all statistical tests. Thus, Type I error, the probability of rejecting the null hypothesis when in fact it is true, is equal to .05. Type II error, the probability of not rejecting the null hypothesis when it (n 96 is false, cannot be known without formulating a specific alternative distribution which is beyond the scope of this research. The specified level of Type I error was determined judgmentally and is consistent with the conventional level of significance used in the social sciences. The account- ing profession has a social obligation to investors to report information in a timely and effective manner. And, since the profession feels that investors should be aware of the occurrence of accounting principle changes, then the significance level should be set to ensure the detec- tion of a problem if one exists. Limitations of the Methodology_ Beaver's study demonstrates that instantaneous price adjustment occurs at annual earnings announcement date when instantaneous is defined as within the week of the announcement.1 The study disclosed herein presents several a priori reasons for expecting the price response adjustment to annual report data to be diffused. Also, this study presents empirical evidence which tends to support this diffusion effect. No problem exists in identifying the starting point of the annual report issu- ance period, but a problem in isolating the termination lBeaver, pp. 67-92. 97 point of the issuance period and, thus, the duration of the issuance period does exist. Therefore, the empirical results reported in Chapter IV are subject to this limita- tion. Another limitation is that the results of this study are strictly applicable only to the New York Stock Exchange firms appearing in the 1966, 1967, 1968, and 1969 editions of Accounting Trends and Techniques. But, to the extent that the New York Stock Exchange firms in these editions are representative of all New York Stock Exchange firms, this limitation is not critical. Summary Chapter III presents the general hypotheses con- cerning the information content of consistency exception opinions. Then these general hypotheses are operation- alized through statistical hypotheses. Appropriate statistical tests are prOposed to test the statistical hypotheses. Also, in Chapter III, the population and samples used in the research are identified, and the data sources used are disclosed. Finally, the chapter concludes by citing the limitations of the methodology. CHAPTER IV RESULTS OF THE STATISTICAL TESTS Introduction Chapter IV reports the outcomes of the tests of the statistical hypotheses proposed in the previous chapter. Also, conclusions based on these outcomes are reported. Results of the Statistical Test Concerning the Principal Hypothesis of the Study The principal hypothesis of this study is concerned with determining if the consistency exception opinion is effective in notifying investors that a particular firm has changed its accounting principles. Specifically, the principal hypothesis states: Null Hypothesis, Hi: The consistency exception communication network in the independent auditor's opinion in the annual report does not notify investors of the occur- rence of an accounting principle change(s). 98 99 Alternative Hypothesis, Hi: The consistency exception communication network in the independent auditor's opinion in the annual report is effective in notifying investors that an accounting principle change(s) occurred. A two—way analysis of covariance model is used to test the principal hypothesis of this study. The specific statistical hypotheses are: 1. Test of treatment effect j: Ho : 03 = 0 for all j Hla: not H1a a O 2. Test of treatment effect k: 1b Ho : Bk = 0 for all k Hlb: not Hlb a o 3. Test of interaction effects: lc ch: not ch a O = 0 for all jk Hypothesis la tests the differences for the three samples in average price response for treatment j where j is defined as no new information or new information at annual report issuance. The firms in the control sample and experimental sample I report no new information, ceteris paribus, in their annual reports while the firms in 100 experimental sample II report their accounting principle changes for the first time. Hypothesis lb is concerned with determining whether a time period effect exists where k is defined as the years 1965, 1966, 1967, and 1968. Hypothesis 1c provides evi- dence as to any interaction effect between information and time period. The inputs into the two-way analysis of covariance research design are the absolute values of the logarithmic price relative residuals, E. , during annual report issu- 1Ta ance period and the absolute values of the logarthmic price relative residuals, E. during annual earnings announce- 1te’ ment week for the three samples. The price change measure, E. , is used as a 1te covariate in the research design in order to remove any effects of initial non-homogeneity among the three samples. Essentially, the price change variable for the annual report issuance period is regressed against the price change variable for the annual earnings announcement week . 1 in order to remove any non-homogeneous effects. 1Utilizing the price change measure, Eite, in a one-way analysis of variance design, a test of differences in price response among the three samples at annual earnings announcement week was performed. The F statistic value of .764 has a probability of occurring by chance of .47, thus, the sample differences are not statistically significant at the conventional level of significance of .05. Yet, the analysis of covariance model allows one to remove the effects of these minimal differences. 101 This regression results in a value of the sample regression coefficient of .2253 which has a probability of occurring by chance of less than .0005. Thus, a signifi- cant statistical relationship exists between the two price change measures. But the coefficient of determination has a value of only .0507 which suggests that while the statis- tical relationship is significant the size of the relation— ship is small. Using the estimate of the regression weight, the category means, , are adjusted, and a two-way analysis I:jjkTa of variance is performed to test for treatment and interac- tion effects. Figure 3 presents the unadjusted category mean price response found for each of the three samples, control, experimental I and experimental II, by year. This figure demonstrates that in three out of four years the average price response during the annual report issuance period found in experimental sample II is greater than the average price response found in experimental sample I. Also, in two out of four years the average price response in experimental sample II is greater than the average price response in the control sample. These results accord with the a priori predictions. When time periods are dis- regarded, the average price response found in the control sample is .0669, in experimental sample I is .0510, and in experimental sample II is .0730. Again, these over-all 102 smflmmp noncommu moccaum>oo mo mammamcm mmslozu may CH pcsom momma whommumoll.m .mflm E 6353 H Begum 635m Hmucmefluwmxm HmvcoEwummxm Houucou . T. a 0.0 no.H I.I-I $3 . . ..... aoma o N .1... 33 Ill. 33 uo.m T04 -o.m no.0 1:. uo.m $6. L13 'enIeA eantosqv ueew OIX‘g z- 103 results are as predicted. But the statistical significance of these results is not in the conventional realm of rejec- tion of .05 for the null hypothesis, Hia. That is, the F statistic associated with the differences in average price response among the three samples is 2.06 and is only significant at the .13 level which is greater than the conventional significance level of .05. Therefore, the null hypothesis, Hia, is not rejected at the conventional significance level, and the conclusion is that there is no statistical difference in average price response among the three samples. That is, the "new" information pre- sented by firms in experimental sample II about their accounting principle changes does not cause price response to be significantly different from the price response found for firms which do not present "new" information, ceteris paribus, in their annual reports. Thus, the principal null hypothesis of this study, H2, is not rejected. By infer- ence, the consistency exception type auditor's Opinion possesses no information content for most investors. Table 8 presents the statistics and significance levels associated with the information effect reported above, year or time period effect, and the interaction effect for these two treatments of information and time period. These results demonstrate that there is no year or time period effect and no interaction effect. This accords with the a priori expectations. 104 .uouuw msflpcsou HousmEoom smoo. avm cflnufiz Hes. mum. pace. 0 xfl .uommmm cofluomumucH mac. mom.~ choc. H “mamas 0cm cacao .o mom. mmm.a mvoo. a Daumucmso .n mom. «Ho. moooo. H Ramada .m Uoflumm mafia .x poommm ucmsummua oma. Hmo.m mmoo. N coau8590mca .m pommmm ucmfiummna mocmno an oaumwumum m mumsvm Eopmmum coaumaum> mcfiuusooo oaumwumum m man «0 muflaflnmnonm can: mo momummo MO 00H50m mHmmmfiomfim AdeUZHmm HEB m0 Emma WUZ¢HM¢>OU m0 mHmNA¢Z¢ w¢3IOSB WEB m0 mBADmmm m mam¢9 105 Results of the Statistical Tests Concerning the Materiality Hypotheses The following sections report the outcomes of the statistical tests associated with the two materiality hypotheses. The first hypothesis is concerned with the concept of materiality utilizing a magnitude measure while the second hypothesis utilizes the type of accounting principle change to operationalize the concept of materi- ality. Materiality:Magnitude Hypothesis This hypothesis states: Null Hypothesis, H2: Investors do not possess O materiality functions based on magnitude or size. Alternative Hypothesis, Hi: Investors do possess materiality functions based on magnitude or size. The statistical hypothesis used to operationalize this general materiality-magnitude hypothesis is: 2a _ Ho . ba — 0 2a Ha . ba > 0 where: ba = the regression coefficient for the population of firms that notify investors of their accounting principle changes only in their annual reports. 106 A simple linear regression model is used to regress the absolute values of the logarithmic price relative, EiTa' during the annual report issuance period for sixty-nine firms of experimental sample II against the measure of magnitude, Pi“ Pi is a percentage computed by dividing the absolute value of the change in current reported net income caused by a change in accounting principle by the simple average of the previous five years' reported net incomes. One firm of the seventy in experimental sample II is excluded because of a negative average five years' re— ported net incomes. For the remaining firms in the sample the values of the magnitude measure, Pi' are: l. Twenty- seven firms with values of Pi ranging from 0.0 to .0499 2. Eighteen firms with values of Pi ranging from .05 to .0999 3. Eleven firms with values of P. ranging from .10 to 1499. 4. Two firms with values of Pi ranging from .15 to 1999. 5. Eleven firms with values of P. equal to or greater than .20 The estimate of the population regression coeffi- cient resulting from the least squares procedure is -.00335. The sign of this value is not in the direction predicted, and the probability that a value of this magni- 2a tude could occur by chance when the null hypothesis, HO , is true equals .978. Thus, the null hypothesis which 107 states that the population regression coefficient equals zero is not rejected. There appears to be little relation— ship between investor response as measured by the price response variable and the materiality—magnitude variable. Therefore, investors do not appear to possess materiality functions associated with magnitude effects in the area of accounting principle changes. Materialitynype of Accounting Principle Change Hypothesis This hypothesis states: Null Hypothesis, H3: Alternative Hypothesis, Hg: Investors do not possess materiality functions based on the type of accounting principle change reported. Investors dg_possess materiality functions based on the type of accounting principle change reported. A two-way analysis of covariance research design is used to test the above hypothesis. The specific statistical hypotheses are: 1. Test of treatment effect j: 3a _ . Ho . aj - 0 for all 3 H33: not H3a a o 2. Test of treatment effect k: 108 - = f 11 k Ho . Bk 0 or a H3b: not H3b a O 3. Test of interaction effects: 3c _ . Ho . ij - 0 for all jk H3O: not H3° a o Hypothesis 3a tests the differences in average price response for treatment j where j is defined as no new information or new information at annual report issuance. Hypothesis 3b tests the differences in average price re- sponse for treatment k where k is defined as the type of accounting principle change while Hypothesis 3c tests the interaction effects of these two treatments. The value of the F statistic for the test of information effects which compares the average price response during the annual report issuance period between the experimental samples is 2.97 which is significant at the .088 level. This significance level of .088 is greater than the conventional significance level of .05; therefore, the null hypothesis, Hga, is not rejected. These results merely confirm the findings for the principal hypothesis of this study. That is, the information content of the consistency exception auditor's opinion appears to be zero when the statistical results are judged at the conventional level of significance of .05. 109 Figure 4 presents the values of the average price response during the annual report issuance period for each type of accounting principle change category for both the experimental samples. The type of accounting principle change categories used in the comparison of price response between experimental sample I and experimental sample II are: 1. Change in inventory policy. 2. Change in depreciation policy. 3. Change in investment credit policy. 4. Change in deferred income tax policy. 5. Change in consolidation policy. 6. Change in pension policy. 7. Change in expense, capitalization and amortization policy of various items. 8. Other changes in policy. As predicted, the average price response found in experi- mental sample II is greater than the response found in experimental sample I in four of the six categories.1 But the F statistic value of 1.75 is significant only at the .132 level which is greater than the specified level of .05. Therefore, the null hypothesis, Hgb , is not rejected. Investors do not appear to possess materiality functions based on the type of accounting principle change reported. 1Categories four and eight are not used in the actual test since there were no occurrences of these types of changes in experimental sample I. 110 cmwmmp aoummmmu mocmwum>oo mo mamaamcm mm3|o3u may cw mason mammfi muommum01|.v .qflm mmcmzu mamflocflnm mswucsooom mo mama m m m m H I:I.IHH meEMm amusmEHuomxm H mamfimm Hmucmfifluwmxm \\ To.o 1o.m Io.v to.m no.m 1o.m To.o.m Io.HH IT , '3 enIeA eqntosqv ueew P z_0I x 111 The interaction effect between type of information and type of accounting principle change reported is not significant; therefore, the null hypothesis, H30, is not rejected. The results of the tests of the materiality-type of change statistical hypotheses are summarized in Table 9. The average price response found in the control sample is also compared with the average price response of the type of accounting principle change categories of experimental sample II. The type of accounting principle change categories are: 1. Change in inventory policy. 2. Change in depreciation policy. 3. .Change in investment credit policy. 4. Change in deferred income tax policy. 5. Change in consolidation policy. 6. Change in pension policy. 7. Change in expense, capitalization and amortization policy of various items. 8. Other changes in policy. Each category mean price response is compared to the mean price response of the control sample utilizing the Student's t test. The value of the observed or calculated t statistic is compared with the value of the t statistic at the significance level of .01 in an attempt to determine if average price response differs between the type of 112 maoo. mm nflnuflz 8mm. am.H mmoo. m xfl .uommmm cofluomumucH mma. m>.H mmoo. m mmcmno wamwocflum mcfiucsooom mo mam» .x pommmm ucmsummne mmo. hm.m «moo. a cowumahomcfl .m uommmm usufiummua mocmno an caumflumum m mumsvm Eocmmnm GOflamfiHm> mcfluusooo oaumwumum can: mo mmmummo mo monsom m was no upwaflnmnoum mmmmmfiomwm m024m0 manUZHmm UZHBZDOUUd m0 HANBIMBHflfiHmmfiflz mmB m0 EmmB @024Hm¢>00 ho mHmMAmq oaumfiumum u meEmm muommumo mmcmno OOGMOAMHGmwm msu mo msHm> Houucoo mnu mo .meflm on“ no .mawm mo maze mumEflxoummd .mmcommom moanm wmmum>¢ .mmcommom unflum mommm>¢ m0 mmHmoomadv mwzmmu manUZHmm UZHBZDOUU< m0 mmWB Ema m0 mmmzommmm monm m0¢mm>¢ WMB UZHMflmZOU Emma u m.BszDBm mmB ho mBADmmm mam2¢ mm? SBHS HH mamZflm AdBZMSHmmmNm OH mam¢9 115 Summary The principal hypothesis of this study is that if the consistency exception auditor's opinion possesses information content, then average price response between the control sample and experimental sample I should not differ since the firms in these samples report no new information, ceteris paribus, in their annual reports. But a difference in average price response is expected to exist between these samples and experimental sample II which consists of firms reporting new information in their annual reports concerning their accounting principle changes. The results of the statistical test performed to determine the information content of the consistency exception type auditor's Opinion are that the difference in average price response among the three samples is in the direction predicted and statistically significant at the .13 level. This significance level is greater than the conventional level of .05 so the null hypothesis is not rejected, and the conclusion drawn is that the con- sistency exception auditor's opinion appears to possess little information content. It is also hypothesized that investors possess materiality functions which aid in their decision processes by allowing the investor to make rational trade-offs be- tween information gained and effort expended in examining 116 data. Two operational measures of materiality were defined with one measure being quantitative while the other is qualitative. The materiality-magnitude hypothesis states that as the quantitative measure of materiality increases, investors' materiality functions become operative, and, therefore, one would expect to find larger price changes to be associated with higher levels of materiality. The results do not substantiate this hypothesis. The sample regression coefficient is -.00335 which is not signifi- cantly different from zero. Also, the results of the materiality test using the qualitative measure which is defined as the type of accounting principle change made by a firm did not substantiate the existence of investor materiality functions. Thus, investors do not appear to possess materi- ality functions based on the operational measures of the concept of materiality used in this study. CHAPTER V SUMMARY, CONCLUSIONS, IMPLICATIONS AND RECOMMENDATIONS Summary of the Research The purpose of this research effort is to examine the information content of the consistency exception type auditor's opinion and, by inference, to provide empirical evidence which will support a judgment as to the effective- ness of the consistency exception communication network. Secondary to this main purpose, certain questions con- cerning the concept of materiality are explored. It is hypothesized and supported by prior empirical work that a causal network among information inputs, covert investor decision-making processes and overt price re— sponses exist. Thus, a price response variable can be utilized in determining the information content of the consistency exception type auditor's opinion. The study examines historical price response during the annual report issuance period of three samples of New York Stock Exchange firms during the period 1965 to 1969. The control sample is composed of 126 firms that 117 118 made no accounting principle changes. Experimental sample I consists of fifty-eight firms which exhibit consistency exception type auditors' opinions. These fifty-eight firms notify investors of their accounting principle changes in their annual earnings announcements and in their annual reports. The third group of firms, experimental sample II, consists of seventy firms that also exhibit consistency exception type auditors' opinions. But these firms report their accounting principle changes only in their annual reports. Thus, the control sample and experimental sample I disclose no "new" information, ceteris paribus, in their annual reports while experimental sample II reports new information concerning accounting principle changes. By comparing the three samples on the basis of the price response variable, inferences can be made concerning the information content and communicative effectiveness of the consistency exception type auditor's opinion. Two measures of materiality are also created. A quantitative measure of the magnitude effect on reported net incomes of the firms in the study is used in a research design to determine if there is a relationship between this materiality-magnitude measure and price response. This research design provides evidence as to the existence of investor materiality functions as well as aiding in the determination of the information content of the consistency 119 exception auditor's opinion. A qualitative measure of materiality based on the type of accounting principle change made by the firms in the study is also used in various research designs comparing price response among the three samples. These designs also provide evidence as to the existence of investor materiality functions. Conclusions of the Study The result of the statistical test to determine the information content of the consistency exception auditor's opinion is that while the average price response differences among the three samples are in the predicted direction, the statistical significance of .13 is greater than the conventional level of .05. Therefore, the con- clusion is that the consistency exception auditor's opinion appears to possess no information content when the statis- tical results are judged at the conventional level of significance of .05. Secondly, the result of the materiality-magnitude test utilizing the quantitative measure of materiality defined in this study demonstrates that there is little or no relationship between the price response measure and the materiality measure. Thus, one concludes that investors do not appear to possess materiality functions based on the quantitative measure used in this study to operationalize the concept of materiality. 120 The results of the materiality tests utilizing the qualitative definition of materiality provide very little evidence to substantiate the existence of investor materi- ality functions. Several caveats should be mentioned at this point. One, the results of this study are strictly applicable only to the New York Stock Exchange firms which are listed in the AICPA publication, Accounting Trends and Techniques, for the edition years of 1966, 1967, 1968 and 1969. Also, since statistical models are utilized, this study is subject to the limitation of statistical error. No statistical study can provide absolute certainity of its conclusions. Yet, if one conforms to the probability assumptions on which the statistical models are con- structed, then the actual significance or probability levels found allow one to assess this uncertainty. A third warning concerns the significance levels reported for the tests of the information content hypothe- ses of the consistency exception auditor's opinion. For the principal information content hypothesis, the signifi- cance level reported is .13. For the secondary information content test of the materiality-type of accounting princi- ple change hypothesis the significance level is .088. These significance levels are close to the conventional level of .05. But if one examines the average price 121 response reported for the three samples, it appears that most of this significance is contributed by the difference in means between the two experimental samples.l Several descriptive features of these two experimental samples are examined and reported at this point to provide informa- tion concerning possible unusual characteristics of the samples. In Chapter III, the value of the average market effect for each of these samples is reported. These values are .17881 for experimental sample I and .18397 for experi- mental sample II. Thus, the samples do not appear to differ greatly on this characteristic. The frequency distributions of the values of the market effect for the two samples which are reported in Table 5 of Chapter III also appear to be similar. Table 11 reveals the frequency distributions for the values of the magnitude measure for the two samples. Again, the samples appear to be similar when judged on this basis. But the two experimental samples appear to be dissimilar when judged on the type of accounting principle change reported. The frequency of occurrence of deprecia- tion changes in experimental sample I is larger than the 1The average price response of the control sample is .0669, for experimental sample I is .0510, and for experimental sample II is .0730. 122 TABLE 11 FREQUENCY DISTRIBUTIONS FOR THE VALUE OF THE MAGNITUDE MEASURE FOR THE EXPERIMENTAL SAMPLES Range and Values of Experimental Experimental the Magnitude Sample I Sample II Measure Frequency Frequency .00 - .0499 16 27 .05 - .0999 8 18 .10 - .1499 12 ll .15 - .1999 7 2 .20 or greater 15 11 Total 58 69 frequency of occurrence of such changes in experimental sample II. Table 12 supports this observation. The totals are greater than the number of firms in each sample as some firms reported more than one change. This preponderance of depreciation changes in experimental sample I is caused by the fact that the entire steel industry changed.its method of depreciation during 1968, and these steel industry firms tended to report their depreciation accounting principle changes in both their annual earnings announcements and their annual reports. 123 TABLE 12 FREQUENCY DISTRIBUTIONS FOR THE TYPE OF ACCOUNTING PRINCIPLE CHANGES REPORTED BY THE EXPERIMENTAL SAMPLES Experimental Experimental Sample I Sample II Type of Change Frequency Frequency 1. Inventory 3 15 2. Depreciation 42 21 3. Investment credit 8 6 4. Deferred income tax 3 5 5. Consolidation 4 l4 6. Pension 5 9 7. Expense, capitalization and amortization 6 4 8. Other 0 7 Total 71 81 Even though there is a higher frequency of occurrence of depreciation changes in experimental sample I than in experimental sample II, the difference in average price change between the two samples for this depreciation category is small. The average price change in the depre- ciation category for experimental sample I is .048714 and for experimental sample II is .050200 as shown in Figure 4 of Chapter IV. But Figure 4 does indicate that the average 124 price responses in the consolidation-type of accounting principle change category are quite different for the two experimental samples.l Thus, the difference in average price response in the consolidation category appears to be a primary contributor to the over-all significance level of .088. Table 10 of Chapter IV shows that the average price response of the consolidation category in experimental sample II is different than the average price response of the control sample. But these differences in average price response appear to be scant evidence for supporting the existence of the information content of the consistency exception communication network or for supporting the existence of investor materiality functions when materiality is defined as the type of accounting principle change reported by a firm. 1The average price response in the consolidation category for experimental sample I is .034560 and for experimental sample II is .104449. The two consolidation policy changes that occurred in experimental sample I are stated as being consolidations of previously unconsolidated subsidiaries. Neither change has any material effect on net income in the opinion of the auditors. Of the nine consolidation policy changes that occurred in experimental sample II, six are changes from the cost to the equity method while three are consolidations of previously un- consolidated subsidiaries. Utilizing the magnitude measure defined in this study, five of the nine changes in experimental sample II have zero effect on net income, three changes have an effect of .02 or less, and one change from the cost to the equity method has a magnitude effect of .1088. 125 Implications of the Empirical Results Two plausible explanations can be formulated to explain the lack of information content of the consistency exception auditor's opinion. One, the consistency excep- tion network is not effective in notifying investors of the occurrence of an accounting principle change. Or, secondly, the fact that firms make accounting principle changes is not a significant information variable for investors in their decision-making processes. This second explanation may be due to the inability of investors to incorporate the information into their decision models or because of covert decisions by investors to ignore the information. The AICPA places the reporting of an accounting principle change among its ten auditing standards which implies that professional accountants believe that informa- tion concerning accounting changes should be of importance to investors. If the consistency exception network is not effective in communicating the occurrence of an accounting principle change, apprOpriate groups must develop a solu- tion to the communication problem. If information con- cerning the occurrence of an accounting principle change is not a significant variable in the investors' decision— making processes, then different solutions must be formulated. An educational process is involved if 126 investors possess an inability to incorporate the informa- tion into their decision processes. If investors choose to ignore the information, this is their value judgment, and only through education as to the importance of the information concerning the accounting principle change can this behavioral pattern be revised. The empirical findings concerning the lack of investor materiality functions are also consistent with these explanations. If investors do possess materiality functions and the consistency exception network is not effective in notifying investors of the occurrence of an accounting principle change, one could not expect the materiality functions to be Operative. Alternatively, if information about the occurrence of an accounting principle change is not a significant variable in the decision processes of investors, one could not expect materiality functions to exist in this area. Recommendations The research results presented herein fail to demonstrate that the consistency exception network of communication possesses the information content that pro- fessional accountants have attributed to the network. This failure to demonstrate information content may result because of one or all of the following reasons: 127 l. The research is subject to statistical error. 2. The consistency exception Opinion is not effective in notifying investors of the occurrence of accounting principle changes. 3. Information about accounting principle changes is not a significant input into investors' decision processes. Replication of this research and additional testing utilizing theoretical models of investor behavior can aid in overcoming any possibility of statistical error. Interest in how financial statement users utilize financial statement data has burgeoned in the past decade. The results of the research effort reported herein provide impetus to the continued exploration of the effectiveness of communication methods utilized by accountants. Also, it is obvious that much additional work is needed to determine how investors utilize financial data. The materiality concept used by accountants also needs thorough examination. The results of this study suggest that the auditor's assumption that investors possess materiality levels which are congruent with the auditor's does not seem justified. BIBLIOGRAPHY BIBLIOGRAPHY Books Bonini, Charles P. Simglation of Information and Decision Systems in the Firm. Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1963. Grady, Paul. Inventory of Generally Accepted Accounting Principles for Business Entepprises. New York. American Institute of Certified PuBlic Accountants, 1965. Hays, William L. Statistics. New York: Holt, Rinehart, and Winston, Inc., 1963. Ijiri, Yuji. The Foundationsfi of Accounting Measurement. Englewood Cliffs, Newfi Jersey: Prentice-Hall, Inc., 1967. Investment Statistics Laboratory, Inc. ISL Daily Stock Price Index: New York Stock Exchange. Palo Alto, CalifOrnia: Investment Statistics Laboratory, Inc., volumes for the years 1964-1969. Kohler, Eric L. A Dictionary for Accountants. Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1952. Spurr, W. A., and Bonini, C. P. Statistical Analysis for Business Degisions. Homewood, Illin01s: Richard D. Irwin, Inc., 1967. Winer, B. J. Statistical Principles in Experimental Design. New York: McGraw-Hill Book Company, 1962. Articles and Periodicals Ball, Raymond, and Brown, Philip. "An Empirical Evaluation of Accounting Income Numbers." Journal of Accountin Research, VI, No. 2 (Autumn, 1968), pp. 150—1 128 129 Beaver, William H. "The Information Content of Annual Earnings Announcements." Empirical Research in Accounting: Selected Studies, 1968, a supplement to Volume VI of the Journal ofiAccounting Research, pp. 67-92. Bird, Francis A. "Interperiod Comparability in Financial Reporting." Journal of Accountancy, CXXVII, No. 6 Bruns, W. J. Jr. "Inventory Valuation and Management Decisions." The Accounting Review, XL, No. 2 Cootner, Paul H. "Stock Prices: Random vs. Systematic Changes." Frontiers of Investment Analysis. Edited by E. Bruce Fredrikson. Scranton, Pennsylvania: International Textbook Company, 1966, Part VI, pp. 489-510. Copeland, Ronald M. "Income Smoothing." Empirical Research in Accounting: Selected Studies, 1968, a supplement to Volume VI of the Journal of Accounting Research, pp. 101-116. , and Licastro, Ralph. "A Note on Income Smooth- ing." The Accounting Review, XLIII, No. 3 (July, 1968), pp. 540-545. , and Wojdak, Joseph F. "Income Manipulation and the Purchase-Pooling Choice." Journal of Account- ing Research, VII, No. 2 (Autumn, 1969), pp. 188-195. Cushing, Barry E. "An Empirical Study of Changes in Accounting Policy." Journal of Accounting Research, VII, No. 2 (Autumn, 1969), pp. 196-203. Dow Jones & Company, Inc. Barron's, various issues during the years 1964-1969. . The Wall Street Journal, various issues during the years 1964-1969. . The Wall Street Journal Index, various volumes for the years 1964-1969. Dyckman, T. R. "On the Investment Decision." The AccountingReview, XXXIX, No. 2 (April, I964), pp. 295-2 5. 130 . "On the Effects of Earnings--Trend, Size and Inventory Valuation Procedures in Evaluating a Business Firm." Research in Accounting Measure- ment. Menasha, Wisconsin: George Banta, 1966, pp. 175-185. . "The Effects of Alternative Accounting Tech- niques on Certain Management Decisions." Journal of Accounting Research, II, No. 1 (Spring, I9645, pp. 91-107. Fama, Eugene F. "The Behavior of Stock Market Prices." The Journal of Business, XXVIII, No. 1 (January, 1965), pp. 34—105. . "Efficient Capital Markets: A Review of Theory and Empirical Work." Journal of Finance, XXV, No. 2 (May, 1970), pp. 383-417. , et al. "The Adjustment of Stock Prices to New Gagnon, Gordon, Graber, Information." International Economic Review, X, No. 1 (February, 1969), pp. 1-21. Jean-Marie. "Purchase Versus Pooling of Interest: The Search for a Predictor." Empirical Research in Accounting: Selected_Studies, 1967, a supple— ment to Volume V of the Journal of Accounting Research, pp. 187—204. Myron J. "Postulates, Principles and Research in Accounting." The Accounting Review, XXIX, No. 2 ; Horwitz, B. N.; and Meyers, P. T. "Accounting Measurements and Normal Growth of the Firm." Research in Accounting Measurement. Menasha, Wisconsin: George Banta, 1966, pp. 221-231. Dean E. "Real and Illusory Earnings Growth." Financial Analysts Journal (March-April, 1969), pp. 52-540 Hepworth, Samuel R. "Smoothing Periodic Income." The Accountinngeview, XXVIII, No. 1 (January, 1953), pp. 32-39. Hicks, Ernest L. "Materiality." Journal of Accounting Research, II, No. 2 (Autumn, 1964), pp. 158-171. 131 Ijiri, Yuji, and Jaedicke, Robert K. "Reliability and Objectivity of Accounting Measurements." Th3 Accounting Review, XLI, No. 3 (July, 1966), pp. 474-4 3. Jensen, Robert E. "An Experimental Design for Study of Effects of Accounting Variation in Decision Making." Journal of Accounting Research, IV, No. 2 (Autumn, 1966), pp. 224-238. Journal of Accountancy. Review by AICPA staff member of Dean E. Graber's article, "Real and Illusory Earnings Growth." Financial Analysts Journal (March-April, 1969), pp. 52-54} Journal of Accountancy, CXXIX, No. 2 (February, 1970), p. 93. King, Benjamin F. "Market and Industry Factors in Stock Price Behavior." The Journal of Business, XXXIX, No. 1, Part II (January, 1966), pp. 139-190. McCosh, Andrew M. "Accounting Consistency-Key to Stock— holder Information." The Accounting Review, XLII, No. 4 (October, 1967), pp. 693-700. Neumann, Frederick L. "The Incidence and Nature of Con- sistency Exceptions." The Accounting Review, XLIV, No. 3 (July, 1969), pp. 546-554. O'Donnell, John L. "Relationships Between Reported Earnings and Stock Prices in the Electric Utility Industry." The AcCounting Review, XL, No. 1 (January, 19657, pp. 135-143. . "Further Observation on Reported Earnings and Stock Prices." The Accounting Review, XLIII, No. 3 (July, 1968). PP. 549—553. Purdy, Charles R; Smith, Jay M.; and Gray, Jack. "The Visibility of the Auditor's Disclosure of Deviance from APB Opinion: An Empirical Test." Empirical Research in Accounting: Selected Studies, 1969, a supplement to Volume VII of the Journal of Accountinijesearch, pp. 1-18. Schiff, Michael. "Accounting Tactics and the Theory of the Firm." Journal of Accounting Research, IV, No. 1 (Spring, 1966), pp. 62-67. Summers, Edward L. "Observations of Effects of Using Alternative Reporting Practices." The Accounting Review, XLIII, No. 2 (April, 1968), pp. 257-265. 132 Reports American Accounting Association. Report of the Committee on Managerial Decision Models. A supplement to the Accounting Review, XLIV, 1969, pp. 43-76. Committee on Auditing Procedure. Auditing Standards and Procedures--Statements on Auditinngrocedure No. 33. New York: American Institute of Certified Public Accountants, 1963. Other Sources American Institute of Certified Public Accountants. Accounting Trends and Techniques in Published Annual Reports, 20th-23rd editions. New York: American Institute of Certified Public Accountants, 1966-1969. Annual reports of the firms in the study for the years 1964-1969. Accounting Principles Board. APB Opinion No. 8: Accounting for the Cost of Pension Plans. New York: American Institute of Certified Public Accountants, November, 1966. . Proposed APB Opinion: Changes in Accounting Methods and Estimates. New York: American Institute of Certified Public Accountants, February 16, 1970. May, Robert G. "The Influence of Quarterly Earnings Announcements on Investor Decisions as Reflected in Common Stock Price Changes." Unpublished Ph.D. dissertation, Department of Accounting and Financial Administration, Michigan State University, 1970. New York Stock Exchange. New York Stock Exchange Company Manual. New YorkIStock Exchange Guide, III--Re1ated Laws and Regulations. New Yofk: Commerce Clearing House, 1962. Personal interview with Mrs. Adele Williams, staff member of the Department of Stock List, New York Stock Exchange, June 1, 1970. APPENDICES APPENDIX A COMPUTATIONAL FORMULAS FOR CALCULATING AUTOCORRELATION COEFFICIENTS AND COEFFICIENTS OF DETERMINATION The formula used in computing the autocorrelation l O I I 2 O coeff1c1ents of determ1nat1on, ra, 13: T 2 ( t =1 eiteit+l) T 2 * ((E=1eit) /T) the autocorrelation coefficient of determina- tion for the ith firm, H II 'the residual logarithmic price relative for the ith firm at time t, (‘D II it = the residual logarithmic price relative eit+1 for the ith firm at time t+1, T = the number of time periods, T = 1, . . ., 63. 133 134 The formula used in determining the coefficients 0 0 2 O of determ1nat1on, r , is: 2 r. = 1 r ‘2 T T T 2 (PR. IR )_ - ((2 IR )3 PR. )/T) _t=1 1t mt =1 mtt=1 1t _ T 2 T 2 7 "T 2 T 2 Z (I ) - ((2 IR )/T) 2 (PR.) - ((2: PR. ) /T) where: r: = the coefficient of determination for the 1th firm, PRit = the logarithmic price relatiVe for the ith firm at time t, IR t = the logarithmic market index relative at time t, T = the number of time periods; T=1, . . ., 58 when the annual earnings announcement week and the annual report issuance period do not coincide but T=1, . . ., 59 when these dates coincide. APPENDIX B SUPPLEMENTARY FIGURES TO CHAPTER III The figures presented in this appendix are profiles of price changes that occur for the eight week period before and after the annual earnings announcement week and the annual report issuance week. One can note that as the sample size or level of aggregatiOn increases the random fluctuations subside, and the profiles become sharper. For the annual earnings announcement profiles, the appearance of non-randomness subsequent to the earnings announcement week (zero week) can be attributed to the issuance of annual reports. Likewise, the appearance of non-randomness prior to the annual report issuance week (zero week) in the annual report profiles can be attributed to the issuance of annual earnings announcements. 135 136 1.25- 1.204 .1.15- 1.101 1.05- 1.00- .954 .90q .854 .80+ -8 -7 -6 -5 -4 -3 -2 -1 O 1 2 3 4 5 6 7 8 week Fig. 5.--Annua1 earnings announcement profile for 254 annual earnings announcements 1.25- 1.20- 1.15q 1.104 l.05~ 1.00fi .95- .90- .85~ .80- -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 week Fig. 6.--Annua1 earnings announcement profile for 126 annual earnings announcements-- control sample 1.40- 1.354 1.30- 1.25- 1.20- 1.154 1.10- 1.05- 1.00- .95- .90- .85- .801 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- .80- 137 -8 -7 -6 —5 Fig. 7.-—Annua1 earnings announcement profile for fifty-eight annual earnings announcements-- -1 0 1 2 3 week -4 -3 -2 experimental sample I 4 5 6 7 8 -8 Fig. 8.-—Annual earnings announcement profile for seventy annual earnings announcements-- -7 -6 -5 -4 -3 -2 -1 O l 2 3 week experimental sample II 4 5 6 7 8 138 1.25- 1.20- 1.15- 1.10- 1.05% 1.00- .95- .90- .854 .804 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 week Fig. 9.--Annua1 report profile for 254 annual reports 080‘1 -8 -7 -6 -5 -4 -3 -2 -1 o 1 2 3 4 s 6 7 8 week Fig. 10.--Annual report profile for 126 annual reports--control sample 139 1.25- 1.20- 1.15- 1.10- 1.05- 1.00% - - - - - ———————————————— .95s .904 .85- .804 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 week Fig. 11---Annua1 report profile for fifty-eight annual reports--experimental sample I 1.30% 1.254 1.20- 1.15- 1.10- 1.05- 1.00- - - - ------- — - - _______ _ .95- .90- .85- .80-1 -8 -7 -6 -5 -4 -3 -2 -1 .0 1. 2 3 4 5. 6 7 8 week. Fig. 12.--Annual report profile for seventy annual reports--experimenta1 sample II