ABSTRACT THE INFLUENCE OF QUARTERLY EARNINGS ANNOUNCEMENTS 0N INVESTOR DECISIONS AS REFLECTED IN COMMON STOCK PRICE CHANGES by Robert George May Unless specifically exempted, every company whose securities are listed on either the New York Stock Exchange or American Stock Exchange is required to report unaudited earnings summaries every quarter. As a result of the influence of major securities exchanges, the practice of reporting quarterly earnings is well established in our economy and shows no sign of diminishing in spite of the doubts that practicing accountants have eXpressed from time-to-time about the reli- ability of quarterly income measurements. Although accountants recognize that their earnings measurements for any finite period, short of the full life of an entity, are subject to uncertainty, their expressed concern about quarterly measurements is grounded on their knowledge of problems affecting quarterly income mea- surements to which annual measurements are not vulnerable. Furthermore, their anxiety is heightened by the fear that investors do not share accountants' knowledge of the relative limitations of quarterly measure- ments and will therefore be misled by relying too heavily on quarterly accounting data. During the last decade writers have frequently made suggestions that show promise of increasing the reliability of quarterly accounting measurements relative to their annual counterparts. But recent evidence indicates that there is still a great deal to be done and that the pace 1‘ —. \‘u -._ 5s. \ ‘ Robert George May Of improvement has apparently been slow. What has been lacking in dis— cussions of needed improvements in quarterly accounting data is evidence of the influence of quarterly data on actual investment decisions. The purpose of this study has been to provide the needed evidence both to facilitate decisions having to do with improving quarterly data and to provide information about investor sensitivity to the quality (reliability) of accounting data. Examination of investor response to quarterly earnings announce— ments, as reflected by common stock price changes in the week of announce- ments, produced the basic evidence of the study. Average stock price-change responses (corrected for market changes) in weeks of quarterly earnings announcements were compared to average stock price changes in annual announcement weeks and in non-announcement weeks for a sample of 105 American Stock Exchange firms over a three-year period. Generally, it was found that whereas average price-change response to quarterly earnings announcements was significantly greater than the average price change in non-announcement weeks (the difference was highly significant), the response to quarterly announcements was not significantly less than the response to the more reliable annual announcements. That investor response to quarterly data is not significantly different from investor response to annual data led to two conclusions: (1) that the lesser effort invested in quarterly accounting measurement relative to annual measurement is not justified on the basis of greater influence of annual data on investor decisions, and (2) that it is not Clear that investors are aware of or take account of differences in Robert George May quality of quarterly and annual data in making investment decisions. Since quarterly data were found to be highly significant in their influence on investors, the implication is that significant benefits to investors can be realized by accountants and managements who take steps to improve the quality of their quarterly data relative to annual data and to clearly apprise investors of remaining relative deficiencies in their quarterly data. To the extent that investors are made aware of the relative limitations of quarterly data it will become less likely that they will act with a strength of conviction not justified by the reliability of the data on which their decisions are based. On the other hand, since actual investor decisions are clearly influenced by quarterly earnings data, any increases in the reliability of the data presumably will lead to more efficient investor decisions. THE INFLUENCE OF QUARTERLY EARNINGS ANNOUNCEMENTS ON INVESTOR DECISIONS AS REFLECTED IN COMMON STOCK.PRICE CHANGES By Robert George May 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 .. "7/ © Copyright by ROBERT GEORGE MAY 1971 ACKNOWLEDGMENTS I would like to express my deep appreciation to the many people whose help and encouragement contributed to my ultimate success in the Ph.D. program. The consideration and guidance of Professor Roland F. Salmonson (chairman) Richard F. Gonzalez, and Alden C. Olson greatly facilitated the completion of this dissertation effort. The comments of Mssrs. Elba F. Baskin and William J. Morris were particularly helpful in getting the project started. Throughout the program my wife, Carol, has been a steady source of encouragement as well as a willing and capable assistant in the data collection for this project. During these years of study I have greatly benefited from the advice and personal interest of Professors James Don Edwards and Herbert E. Miller. I have enjoyed the generous financial support of the Ford Foundation during the early part of the program and the Arthur Andersen & Co. Foundation during the dissertation stage. Professor Edwards skillfully provided support during any and all periods when support was not available from other sources. I sincerely hOpe that the conduct and achievements of my profes— sional career adequately repay all of these and the many other contrib- utors to my progress to date. ii TABLE OF CONTENTS Page ACKNOWLEDGMENTS. . . . . . . . . . . . . . . . . . . . . . . . . . ii LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . Vi LIST OF FIGURES. . . . . . . . . . . . . . . . . . . . . . . . . . viii Chapter I O INTRODUCT ION I O I O O O O I 0 O O O I O O O O O 0 O O O 1 Institutional Background of Interim Reporting Problems of Interim Income Measurement Institutionalization of the One-Year Period Seasonality Random Fluctuations Improvement in Quarterly Accounting Practice Purpose and Motivation of the Study The Approach and Organization of the Study II. PAST RESEARCH CONCERNING QUARTERLY ACCOUNTING DATA. . . . . . . . . . . . . . . . . . . . 16 Survey of Reporting Practice Evaluations of the Usefulness of Reported Data Surveys of Opinion as to Usefulness of Quarterly Data Empirical Studies of the Influence of Quarterly Data on Actual Investor Decisions III. STOCK PRICES, EXPECTATIONS AND INFORMATION. . . . . . . 33 Information Classes and Correction of Stock Price Changes for "Market Effects" Specific Information Flows and Measured Stock Price Responses The Random Walk Theory Relating Specific Information to Specific Price Changes Selection of Price-Change-Response Period Pinpointing the Moment of First Perception of Announced Earnings Summary iii Chapter IV. VI. VII. Page THE BEHAVIOR.AND SIGNIFICANCE OF PRICE- CHANGE RESPONSES TO QUARTERLY DATA . . . . . . . . . . 48 The Significance of Price-Change Responses to Quarterly Announcements The Response to Quarterly Data Relative to the Response to Annual Data Clarification of the Approach to Measuring Investor Response Summary SAMPLE SELECTION, DATA GATHERING AND RESEARCH FINDINGS. . . . . . . . . . . . . . . . . . . 6O Selection of a POpulation Selection of a Study Period Sample Selection Procedures Finite POpulation Represented by the Sample Selected Data Gathered Correction and Transformation of Prices Testing the Significance of Investor Response to Quarterly Announcements Testing the Significance of the Difference Between Investor Response to Quarterly Earnings and Investor Response to Annual Earnings Summary ANALYSIS OF THE EFFECTS OF CERTAIN ASSUMPTIONS ON THE REPORTED FINDINGS . . . . . . . . . . . . . . . 94 Correction of Price Changes for Market Effects Multiple Announcements and Comparability of Measurements of Response to Annual and Quarterly Earnings Reevaluation of the One Week Response Period Density of Other News Events in Weeks Prior to Earnings Announcements Summary ANALYSIS OF THE EFFECTS OF TIME SERIES AND SAMPLE COMPOSITION ON THE REPORTED FINDINGS . . . . . . . . . 120 Calendar Quarter Time Series of Response Ratio Profiles Evaluation of Differences Between Subgroups of Sample Firms Cash-Dividend—Paying Firms vs. Non- Paying Firms iv Chapter . Page Calendar-Year Firms vs. Non-Calendar- Year Firms Summary of Evaluation of Subgroups of Sample Firms VIII. SUMMARY OF FINDINGS, CONCLUSIONS, RECOMMENDATIONS, AND SUGGESTIONS FOR FURTHER RESEARCH . . . . . . . . . 146 Summary of the Findings of the Study Conclusions and Recommendations APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . 159 L IST OF TABLES Table Page 1. Results of application of sample selection criteria . . . . . . . . . . . . . . . . . . . . . . . . . 7O 2. Sample results of average response ratio measurements for various groupings of announcements . . . . . . . . . . 86 3. Sample results of differences between average response ratios of annual announcements and average response ratios of quarterly announcements. . . . . . . . . . . . . 92 4. Sample results of average response ratio measurements without market correction for various groupings of announcements . . . . . . . . . . . . . . . . . . . . . 100 5. Sample results of differences between average response ratios of annual announcements and average response ratios of quarterly announcements-without market correction . . . . . . . . . . . . . . . . . . . . . . . . 101 6. Sample results of differences between average response ratios of annual announcements and average response ratios of quarterly announcements-incorporating multiple statements. . . . . . . . . . . . . . . . . . . . 106 7. Proportion of price changes in prior and subsequent weeks having the same sign as the price change in the announcement week. . . . . . . . . . . . . . . . . . . 112 8. Frequencies of other news events in the weeks prior to earnings announcements. . . . . . . . . . . . . . . . . 117 9. Frequencies of dividend related news events in the weeks prior to earnings announcements. . . . . . . . . . . 118 10. Distribution of sample firms by payout policy and fiscal year. . . . . . . . . . . . . . . . . . . . . . 127 11. Sample results of average response ratio measurements for various groupings of announcements of dividend— paying firms . . . . . . . . . . . . . . . . . . . . . . . 134 12. Sample results of differences between average response ratios of annual announcements and average response ratios of quarterly announcements of dividend-paying firms. . . . . . . . . . . . . . . . . . . . . . . . . . . 134 vi Table 13. 14. 15. 16. l7. 18. Page Sample results of average response ratio measurements for various groupings of announcements of non- dividend-paying sample firms . . . . . . . . . . . . . . . 136 Sample results of differences between average response ratios of annual announcements and average reSponse ratios of quarterly announcements of non—dividend- paying firms . . . . . . . . . . . . . . . . . . . . . . . 137 Sample results of average response ratio measurements for various groupings of announcements of calendar— year firms 0 O O O C O O O O O O O O O O O O O O I O O I I 141 Sample results of differences between average response ratios of annual announcements and average response ratios of quarterly announcements of calendar-year firms. . . . . . . . . . . . . . . . . . . . . . . . . . . 141 Sample results of average response ratio measurements for various groupings of announcements of non-calendar- year firms 0 o o a o o o o o I o o o o o o o o o o o o o o 142 Sample results of differences between average response ratios of annual announcements and average response ratios of quarterly announcements of non-calendar- year firms . . . . . . . . . . . . . . . . . . . . . . . . 143 vii LIST OF FIGURES Figure 1. Profile of mean price-response ratios for 17 weeks surrounding 506 annual earnings announcement dates of 143 firms for years 1961 through 1965 . . . . . . . 2. American Stock Exchange Price Level Index: Monthly Close January, 1963 - December, 1968 . . . . . . . 3. Pattern of observation of individual firms' earnings announcements during the study period. . . . . . . . . . 4. Frequency distribution of firms' proportions of explained variance due to regression of price-change measurements on price-index changes. . . . . . . . . . 5. Frequency distribution of firms' coefficients of auto— correlation of residual price change measurements. 6. Eleven week profiles of mean average response ratios for various groupings of earnings announcements. . . 7. Frequency distributions of highest average response ratios according to week relative to announcements . . 8. Frequency of incidence of earnings announcements in the weeks of 1966. . . . . . . . . . . . . . . . . . 9. Eleven week profiles of mean average response ratios with market correction and without market correction . 10. Eleven week profiles of mean average response ratios incorporating multiple statements. . . . . . . . . . . . ll. Eleven week profiles of mean price response ratios according to time of earnings announcement . . . . . . . 12. Eleven week profiles of mean average response ratios of the 40 cash-dividend-paying sample firms. . . . . l3. Eleven week profiles of mean average response ratios of the 65 non-dividend-paying sample firms . . . . . . l4. Eleven week profiles of mean average response ratios of the 56 calendar-year sample firms . . . . . . . . . . 15. Eleven week profiles of mean average response ratios of 49 non-calendar-year sample firms . . . . . . . . . . viii Page 44 66 68 82 82 88 89 98 99 105 122 131 132 139 140 CHAPTER I INTRODUCTION This chapter briefly outlines the institutional background of interim accounting reports, the problems associated uniquely with interim accounting measurement, the purpose and motivation of the study, and finally the approach and organization of the dissertation. Institutional Background of Interim Reporting1 In spite of attendant theoretical and practical problems, quar- terly financial statements (or announcements), giving summary sales and earnings data for periods less than a year, are currently well entrenched in our economy and appear to be "here to stay." The stock exchanges and financial analysts have been the primary proponents of interim financial reports. The New York Stock Exchange has advocated interim reports for its listed companies since 1910 and has been the dominant influence in eXpanding the practice of quarterly reporting. The American Stock Exchange, in its listing requirements, revised in 1962, requires its listed companies to make quarterly financial reports to the public. At the time that the revised requirements went into effect some 60% of its listed companies were already conforming to the 1This brief description of the institutional framework of interim reporting essentially paraphrases a fine review given by Robert G. Taylor, "A Look at Published Interim Reports," The Accounting Review, XL, No. 1 (January, 1965), pp. 89-96. .A .1 .. ( .. a _. .. u... \ L. ,. 7.1-. . v..\\ . .. x .u.\ . 2 quarterly reporting requirement. It was largely as a result of pressure from the Financial Analysts Federation that the SEC adOpted a requirement for the brief and unaudited semi—annual income statement 9-K, on June 23, 1955. Practicing accountants and their professional associations have generally been reluctant to endorse interim financial reports and have, in the past, resisted interim reporting requirements by authorities such as the SEC. The American Institute of Accountants noted in its 1936 bulletin that it pointed out to Congress during consideration of the Secu- rities and Exchange Bill that statements for interim periods are likely to mislead investors.1 Carman Blough, then editor of the Journal of Accountancy, editorialized against the 1952 prOposal by the SEC for an interim reporting requirement.2 Although sharing some of the caution of practicing accountants, academic accountants have been much more positive in recognizing the value of interim reports to investors. In 1934 Sanders suggested that investors needed interim information and that accountants should supply it, each reporting firm dealing with seasonality and other problems as best it can.3 In Accounting and Reporting Standards the Committee on IAmerican Institute of Accountants, "Discussion of PrOper Basis for Quarterly Reports," Bulletin of the American Institute of Accountants, No. 144 (April 16, 1936), p. 14, as cited in Taylor, The Accounting Review, Vol. XL, No. l. 2Carman D. Blough (editor), "Some Dangers Inherent in Quarterly Financial Statements," Journal of Accountancy, Vol. XCV, No. 2 (Febru— ary, 1953), pp. 221-222. 3Sanders, T. H., "Reports to Stockholders," The Accounting Review, Vol. IX, No. 3 (September, 1934), as cited in Taylor, The Accounting Review, Vol. XL, No. l. 3 Concepts and Standards Underlying Corporate Financial Statements of the American Accounting Association exhibited a belief in the potential value of interim accounting information in the statement that: ...it endorses the growing practice of supplementing annual reports by statements giving highlights of interim periods Operations and suggests the need for special and prompt recording of unusually significant events. Unfortunately, the SEC has generally been caught in the middle of the positions of these interested groups, first agreeing with one posi— tion, then the other. In 1946 the SEC began requiring quarterly sales data and later prOposed requiring quarterly income statements as well. But resistance to these prOposals was so strong that even the sales requirement was withdrawn in 1953. As was noted above, in response to pressure from security analysts, the SEC in 1955 finally set interim income statement requirements that have endured to the present. The Wheat Disclosure Study, still under consideration by the SEC, has again recommended, among other things, that the SEC require un- audited quarterly reports of companies with registered securities.2 As of the time of this writing, however, "The Commission has neither adopted, approved nor disapproved the report."3 1American Accounting Association, Accounting and Reporting Standards (Columbus, Ohio: American Accounting Association, 1957), p. 46. 2U. S. Securities and Exchange Commission, "Summary of Disclo— sure Policy Study Report Entitled: Disclosure to Investors-~A Re- appraisal of Administrative Policies Under the '33 and '34 Acts." 3Ibid., p. 1. 4 Problems of Interim Income Measurement Although measurement of income for any period shorter than the life of the enterprise presents difficult problems1 there are additional problems of income measurement for periods as short as a fiscal quarter that are not encountered in annual income measurement. Indeed, an exam- ination of some of the additional problems seems to explain partially the historical resistance to interim reporting requirements on the part of some representatives of practicing accountants. The additional prob— lems of income determination can be conveniently grouped by their source. The three major sources are: (1) the institutionalization of the one- year period, (2) seasonality, and (3) degree of aggregation of the effects of random events. Institutionalization of the One-Year Period Institutionalization of the (ne-year period is manifest in the many accounting variables whose measurement is finalized only once each year. Income tax eXpense, pension costs and profit shares are typically finalized at the end of the fiscal year either by fiat or in accordance with contractual provisions. But the measurement of many other account- ing variables is finalized only once a year for reasons of economy or practicality. Examples of the latter are the adjustment of recorded inventory for physical shrinkage and obsolescence, and the variety of other adjustments that may result from the typically once-a-year appli— cation of audit procedures. Institutionalization of the one year accounting period means 1Eldon S. Hendriksen, Accounting Theory (Homewood, Illinois: Richard D. Irwin, Inc., 1965), p. 103. 5 that for interim purposes the values of many variables must be estimated with a lesser degree of certainty than their annual counterparts. To the extent that the values of particular variables are incorrectly esti- mated for the first three quarters of a fiscal year, their patterns may potentially mislead investors in forming eXpectations of the annual values that ultimately supercede them.1 In a research study that will be described in the next chapter, Gale E. Newell confirmed that this potential is reflected in the actual patterns of reported quarterly accounting information.2 Seasonality Seasonal fluctuations in various activities of business enter— prises can lead to substantial within-year mismatching of expense with related revenue unless considerable additions to the procedures for 1A classic example of the potential of quarterly values for mis- leading investors in forming their expectations of annual values, due to a typically once-a—year measurement cycle, was reported by Gordon Shillinglaw in "Concepts Underlying Interim Financial Statements," Accountingrgeview, Vol. XXXVI, No. 2 (April, 1961), p. 228. Relating the results of a survey of the interim accounting practices of six com- panies the author noted that: "...Each of the companies makes an equal provision for year- end bonus in each of the first three quarters, based on the bonus paid in the previous year, despite the fact that in every case the amount of the annual bonus is based on the Operating results of the current year. Any change in the annual bonus is concentrated in the fourth quarter state- ments only." For another classic example having to do with annual inventory adjustments see discussion of the case of Kaiser Frazer vs. Otis and Company in SEC AccountinggPractice and Procedure (2nd edition; New York: Ronald Press Co., 1963), by Louis H. Rappaport. 2Gale E. Newell, "Published Quarterly Financial Data: Their Adequacy for Investment Decision.Making" (unpublished doctoral disser— tation, Michigan State University, East Lansing, Michigan, 1968). 6 allocating eXpense amggg years are employed in allocating expense within years. A well-known and widely adOpted additional within—year allocation practice with respect to production costs is the accrual of vacation pay throughout the year. Many manufacturing and nondmanufacturing expenses, however, are not clearly amenable to within-year allocation, although they vary with both the season and production. For instance, employers' shares of F.I.C.A. tax, heat, light and similar expenses which may be incurred at uneven within—year rates unrelated to production rates, cause variation in incurrence of factory overhead that cannot be attributed to production or efficiency levels. Thus interim overhead variances will in part stem from the usual causes, efficiency and activity levels, and in part from seasonal fluctuations in expense incurrence. As Shillinglaw points out, if each variance is totally deferred at interim dates, adjustment of income for deviations from budgeted levels of activity and efficiency will be felt in the fourth quarter, whereas if each vari- ance is totally charged or credited to income and/or inventory" ...fluc- tuations in reported income may be either damped or intensified depending on the relationship between seasonal patterns in costs [incurred], in production, and in sales."1 A more refined treatment requires charging or crediting the efficiency and activity-related portions of variances to income and inventory accounts, while deferring the season-related portions.2 But such treatment requires more sophisticated observation and analysis of cost patterns on the part of management. Just as some factory costs may vary with the time of year as 1Shillinglaw, The Accounting Review, Vol. XXXVI, No. 2, p. 226. 21bid., pp. 226-227. well as with production levels, some variable non—manufacturing costs may bear different relations to sales volume in different seasons. Further- more, some non-manufacturing costs that do not vary in any discernable way with sales may nevertheless vary with the time of year, e.g., the cost of heat and light in administrative office buildings. As with the season-related factory costs the effects of such non-manufacturing cost patterns can be accommodated by management or allowed to "flow through" to interim income.1 In any case the problems caused by seasonality of both manufacturing and non-manufacturing costs are not avoided. Their burden may simply be transferred to investors to a greater or lesser degree through more or less SOphisticated cost allocation procedures. Random Fluctuations In the researcher's experience, discussions in the literature of the problems of interim reporting merely mention greater random fluctu- ation as one limitation of interim measurements relative to annual measurements. The reasons for and nature of the limitation are left to the experienced reader's intuition or imagination. The reason for greater relative variability; of interim measurements is their lesser 1Ibid., p. 227. 2The notion that the degree of within-year allocation of season- related cost elements merely regulates the portion of season—related income determination problems that the investor must cOpe with was first encountered by the researcher in "On Criteria for Judging Accounting Earnings Estimators," adapted from John W. Kennelly, "An Empirical Investigation of Interim Earnings Reports" (unpublished Ph.D. disser— tation University of Chicago, in progress). 3Relative variability for present purposes is intended to mean the range of chance variability of a variable compared to the average size of that variable. 8 degree of aggregation (over partially chance-determined events) relative to annual measurements. Since the relation between relative variability and aggregation may not be intuitively obvious to every reader, an illus— tration is provided in the Appendix.1 A reader who is not really secure in his understanding of the nature of the relationship between aggrega- tion and variability is encouraged to read this Appendix before going on. The present discussion will continue with consideration of the limitation imposed on interim accounting measurements by their greater chance vari— ability. Chance variability in measured variables essentially tends to obscure the non-chance-determined portion of a given measured level of the variable. As a result, inferences (including estimates or predic— tions) that an interested person wishes to draw from partially chance- determined variables are not perfectly efficient.2 In fact, the effi- ciency of inference generally decreases as the chance variability in available observations increases. That accounting variables are used for inferences is a well established belief among accountants.3 Their traditional concern that 1The illustration has been omitted from the body of the discus- sion because it is necessarily abstract and requires knowledge of the properties of random variables on the part of the reader. 2Efficiency of estimate in statistics, the discipline devoted to accommodating chance variability in making inferences from observations, is defined as the variance of estimated values with respect to the true value of interest. Perfect efficiency may be thought of as zero vari- ance of estimate. 3Eldon S. Hendriksen, AccountinggTheory (Homewood, Illinois: Richard D. Irwin, Inc., 1965), p. 103. 9 investors will be misled by interim reports due to random fluctuations reflects at least an intuitive understanding of the effects of chance variability on efficiency of inference. For if investors fail to recog- nize the expected lesser efficiency of inferences based on more vari- able interim data, they will presumably act on those inferences with a strength of conviction that is not justified. Unlike the effects of institutionalization of the one year period and seasonality, there is nothing accountants can do about chance variability itself. They can only attempt to improve investors' understanding of the different degrees of chance variability inherent in accounting measurements covering accounting periods of different lengths. Improvement in Quarterly AccountinggPractice As was pointed out in the discussion above, the limitations on quarterly accounting measurements imposed by seasonality and chance vari- ability cannot be completely eliminated by the efforts of accountants. But the limitations imposed by institutionalization of the one year period can be substantially avoided, in some cases, by changing the law or traditional contract provisions, e.g., the cases of federal income tax expense and pension costs, in other cases by simply devoting addi— tional resources to the measurement process, e.g., by making counts of physical inventory quantities quarterly rather than only annually. In addition, investors may be spared some avoidable effects of seasonality by more sophisticated cost allocation procedures than are required for making allocations of costs between annual periods. Furthermore, accountants could conceivably take it upon themselves to better apprise users of the implications of the greater chance-variability of interim 10 measurements relative to their annual counterparts. Many specific procedural suggestions have been made in the liter- ature for improving quarterly accounting data relative to annual data. Some of the more widely suggested and more promising are: l. 2. Always defer the season related portion of factory vari- ances at interim dates.l Separate administrative and selling expenses into fixed and sales related portions. Allocate the former to interim periods in equal lump sums, deferring a charge or credit for the difference between the amount incurred to date and the amount allocated. Allocate the latter in proportion to sales at a standard rate for the whole year with deferrals for only season-related differences in the cost-to-sales relation through the year.2 In all cases base quarterly estimates of once-a-year measurements on management's carefully formed judgement of the expected annual level of the measurement, not on the prior years' level or some other value unrelated to current year-to-date experience.3 Instigate at least limited surveillance of interim reports by independent auditors.4 Convey to the public the differential lack of precision inherent in quarterly reports by informative commentary or footnotes and/or by reporting quarterly income (and annual income) as falling in an interval rather than as a single value.5 The sizes of ranges given for quarterly 1 See Shillinglaw, The Accounting Review, XXXVI, No. 2, p. 227. 2Ibid., pp. 227-230. For a still more comprehensive approach to seasonality see David Green, Jr., "Towards a Theory of Interim Reports," Journal of Accounting;Research, Vol. II, No. 1 (Spring, 1964), pp. 35-49. 3 Shillinglaw, The Accountigg Review, XXXVI, No. 2, pp. 227-230. hsee for example: Lee J. Seidler and William Benjes, "The Credibility Gap in Interim Financial Statements," Financial Analysts Journal, Vol. XXIII, No. 5 (September-October, 1967), pp. 114-115; Robert G. Taylor, "The Published Interim Report and the CPA," Journal of Accountancy, Vol. CXX, No. 3 (September, 1965), p. 57; and Gale E. Newell, 1968, p. 155. 5Newell, pp. 153 and 154. 11 and annual data should, of course, reflect the differ— ential precision in measurements covering periods of different lengths. That current accounting reporting practice does not include wide-spread adOption of most of the above suggestions is apparent to most students of interim reporting. It is reflected in the findings of research studies specifically concerned with the quality of reporting practice. For instance, consider the following statement from an offi- cial summary of the SEC's recent "Wheat Disclosure Study": ...The Study carefully examined a significant sample of quar— terly financial reports and releases provided by the two [national securities] exchanges. It was readily apparent (and acknowledged by representatives of the exchanges) that they varied from extremely useful to extremely poor and un- informative.1 and this statement from the conclusions of Newell's study of the pat- terns of quarterly accounting data: The evidence presented in this study indicates that quar— terly data are often inaccurate and suggests that the poten- tial mis-advising from the use of such data is significant. As this paper has indicated, reported quarterly net income is often unreliable and therefore many of the items that are used in its determination must also be unreliable.2 Two things that seem to stand in the way of wide-spread improve- ment in the quality of quarterly accounting data are (l) the motivation of accountants and managers and (2) the high cost of implementing Specific improvements in measurement procedures for interim reports relative to the cost of solving other problems facing accountants. The motivation lU.S. Securities and Exchange Commission, Disclosure to Investors: a Reappraisal of Federal Administrative Policies Under the 33 and 34 Acts (New York: Commerce Clearing House, Inc., 1969), p. 39. 2Newell, p. 155. As was promised earlier, this study's findings will be reviewed in greater detail in Chapter II. 12 of accountants and managers will, of course, be directly related to the‘ cost-benefit relations inherent in each problem that demands their atten- tion. But whereas accountants and managers may have some definite grasp of the probable costs of improvements in quarterly accounting measure— ment, the potential relative benefits have not been well established. Certainly information concerning the degree of use of quarterly data by investors and/or the significance of the influence of quarterly data on investors would be useful to accountants and managers who must decide whether to devote themselves to greater quality of quarterly reports or to other demanding accounting problems. Purpose and Motivation of the Study Although the major securities exchanges have been the prime source of organized demand for quarterly accounting data, they have not been particularly active in assessing the significance of the reports they require to the investing public. Accountants have made attempts to assess both the usefulness of quarterly data and the degree to which they are used by or influence investors. Their major efforts will be reviewed in some detail in the next chapter. But it will be seen that while the investigations into the usefulness of quarterly data may have been fruitful, the investigations into the significance of their use and/ or influence have been unsatisfactory. The purpose of this study is to provide information and motiva- tion to accountants and managers who have to make decisions affecting the resources devoted to improving quarterly accounting measurements by attempting to answer the following two empirical research questions: 1. Do quarterly accounting data, in the form of public earnings announcements, have a significant effect ... 13 on investor decisions as reflected in market price changes? 2. Does there appear to be a significant difference between the influence on investors of quarterly and annual earnings announcements? Does the difference, if any, reflect investor awareness of the lesser quality of measurement inherent in quarterly income? In addition to providing information relevant to the specific issue of improvement in quarterly accounting measurement, the second research question shows promise of contributing to the much broader issue of investor sensitivity to the nuances of the accounting measurement pro- cess in general. Knowledge of the sensitivity of investors to the dif- ference in quality of annual and quarterly accounting measurements is a positive (if limited) contribution to our understanding of investor be— havior. The central importance of understanding of investor behavior to future development of accounting theory is clearly expressed in the following segment from A Statement of Basic AccountingiTheory: Because of the great value of accounting information to external users, and because we have some knowledge of many users' needs, it is possible to develop significant account- ing information even though the precise and total needs of each user for each decision are unknown. This is so because even crudely measured and only generally apprOpriate infor- mation may be of considerable use to external users in view of their highly uncertain situation. It follows that it is not necessary to develop a detailed list of all user needs in advance. On the contrary, until much more is known of the behavioral characteristics of external users, accounting in— formation must be developed from a broad and imprecise under— standing of the informational needs of external users. When and as the results of fundamental research on the informa- tional needs of external users bear fruit, the structure of accounting theory and reporting based upon it can logically be expected to expand. lAmerican Accounting Association, A Statement of Basic Accountipg Theory (Evanston, Illinois: American Accounting Association, 1966), p. 20. 14 The Approach and Orggnization of the Study The approach of the study is to infer from measured price changes, immediately following earnings announcements, the relative effects of quarterly earnings announcements on investors' expectations. Other ways of measuring the significance of accounting data to investors, namely questionnaire or interviewbbased methods, have been rejected because of their dissociation from actual decisions.l Price changes on the other hand, while perhaps not perfect reflections, are believed to reflect changes in expectations or at least the resulting decisions to buy, sell or hold particular securities. In addition, price changes have economic significance quite apart from any theory relating them to investor expec- tations. As will be seen in Chapter II, other attempts to use price changes to assess the significance of quarterly data have not been particularly successful in answering the research questions of interest in this study. They are reviewed in Chapter II, along with other empirical research con— cerning interim accounting data, to provide background and a point of departure for develOpment of more promising methodology. Because of the experience of prior research efforts and the high level of so-far uneXplained variability in stock prices, a major portion of this study is devoted to careful elaboration of a measurement system that will satisfy the two research questions of interest. Chapter III will be devoted to developing a basic expected relationship between 1One of the major weaknesses in questionnaire and interview techniques is that the subject is not being observed under actual operating conditions but is only talking about his thought process and actions under those conditions. Thus, unrestricted by operating pres- sures and constraints, he is free to answer queries as he wants and may be influenced by what he believes the questioner wants to hear. ~q If! 15 earnings announcements as specific news events about a firm and carefully measured stock price responses. Chapter IV will employ the relationship developed in Chapter III to construct measurements and tests that will satisfactorily answer the two research questions of interest in terms of sample data. Chapter V describes the selection of a population of firms for the study, the further selection of sample firms, data gathering and transformation, the results of the tests developed in Chapter IV, and finally the preliminary findings of the study with respect to the research questions. Before reaching final conclusions based on the sample results, however, certain assumptions and potential procedural weaknesses of the methodology will be reviewed in Chapter VI, in light of the information contained in the sample data. Chapter VII will examine, gx_pp§£, the effects of sample composition on the findings of the study. Chapter VIII will be devoted to final conclusions and recommenda- tions. CHAPTER II PAST RESEARCH CONCERNING QUARTERLY ACCOUNTING DATA Research to date concerning quarterly accounting data can be classified into the following categories: 1. A survey of reporting practice (to shareholders). 2. Studies of the utility of quarterly data for various purposes. 3. Surveys of opinion as to the usefulness of quarterly data. 4. Empirical studies of the impact on actual investor deci- sions of quarterly accounting data. This chapter will simply review several studies that fall into the first three categories. But because of their similarity of purpose and approach to this project the two studies making up the fourth cate- gory will be evaluated as well as reviewed. Survey of Reporting Practice Rebert Taylor conducted a survey of reporting practices in quar— terly reports sent to shareholders during the early nineteen-sixties, following the tradition of the A.I.C.P.A. surveys of annual reporting practice.1 Briefly, some of his detailed findings for the 600 companies surveyed were: 1Taylor, The Accounting Review, Vol. XL, No. l. l6 17 1. Comparative summary income statement figures were found in 97% of the cases. 2. Cumulative figures, i.e., covering the full period from the beginning of the fiscal year, were found in 38% of the cases. 3. The median number of items displayed in the income state- ments was nine. 4. Approximately 20% of the reports contained balance sheets. 5. The greatest variety in practices was found in the textual material included in the reports. 6. The evolution of reporting practices was observed to be in the directions of more frequent interim reports (from semi— annual to quarterly reports), more reporting directly to stockholders, and special reports to the financial community. In the summary of his survey findings Taylor noted that the quar- terly reports were more extensive than he had eXpected and that they were "....constantly being changed and worked on."1 Evaluations of the Usefulness of Reported Data With one exception the past research efforts that have focused on the usefulness of quarterly data for investment decision purposes have generally explicitly or implicitly used the criterion of ability to pre— dict annual earnings as the determinant of usefulness. The exception, a study by George Staubus, employed the criterion of ability of comparative earnings levels to predict discounted future cash flows (measured in Ibid., p. 92. l8 retrospect) to holders of common stocks. Most explicit in the use of prediction of annual earnings as a measure of usefulness have been the forecasting studies of Green and Segall2 and Brown and Neiderhoffer.3 G & S began what was to become a series of three related studies by attempting to determine whether naive but plausible forecast- ing models using first-quarter as well as prior years' earnings data would better predict current year annual earnings than would models using only prior years' earnings data. In a study of 50 NYSE firms for years 1959 through 1964 they computed three measures of success, the percent difference between actual and forecast earnings, the absolute percent difference, and the squared percent difference. Based on these measure- ments the authors tentatively concluded that first-quarter data did not promise any improvement in forecasting ability. Finding these results rather remarkable, the authors replicated the study for an additional year using the same sample of firms and an additional sample of 44 firms. The replications produced substantially the same results and conclusions as the original study. 1George J. Staubus, "Earnings Periods for Common Share Analysis," Journal of Business, Vol. XLI (October, 1968), pp. 472-476. 2David Green Jr. and Joel Segall, "The Predictive Power of First Quarter Earnings Reports," Journal of Business, Vol. XL, No. 1 (January, 1967), pp. 44-55; "The Predictive Power of First Quarter Earnings Re- ports: A Replication," Empirical Research in Accountipg: Selected Studieg, 1966, a Supplement to Volume 4 of the Journal of Accounting Research. 3Phillip Brown and Victor Neiderhoffer, "The Predictive Content of Quarterly Earnings," Journal of Business, Vol. XLI, No. 4 (October, 19 B & N, apparently stimulated by the findings of the G & S studies dealing with first—quarter data, undertook to apply the G & S methodology to all three quarters' data. Their results for first-quarter data, using the same measurements but a much larger sample of firms, were mildly con- tradictory to the G & S findings. But as would be hoped, they found that forecasts of annual earnings improved progressively as second and third quarter data were included in the inputs to the forecasting models. Nevertheless, the average absolute forecast errors experienced by the two best models, using all three quarters data in forecasting a given annual datum, ranged from 12.9% for 1963 using one of the models to 9.3% in 1965 using the other model. Additionally, although not recognized by B & N, their tables of error measurements seem to indicate that forecasts based on quarterly as well as annual data tend to be more consistently conser- vative (forecast earnings lower than actual) than forecasts based on annual data alone. Newell approached the usefulness of quarterly earnings reports from a different point of view than did the forecasting studies.1 He evaluated the extent to which the problems associated with quarterly accounting income measurement were reflected in actual patterns of reported quarterly and annual earnings figures. Working with 87 American Stock Exchange companies, Newell found that fourth quarter ratios of net income to net sales deviated most from the annual results more frequently than would occur by chance one time in one hundred thousand. Furthermore, 1Gale E. Newell, "Published Quarterly Financial Data: Their Ade- quacy for Investment Decision Making" (unpublished doctoral dissertation, Michigan State University, East Lansing, Michigan, 1968). 20 his finding that the fourth quarter results tended to deviate more fre- quently on the high side is consistent with the B & N results that indi— cated that forecasts based on first three quarters' data tend to be con— servative. Newell concluded that "...these results verify that the problems that are inherent in quarterly reporting do affect the reported results and that quarterly statements are limited in their reliability."1 Although, in another analysis using ratios of income tax to net income before taxes Newell did not find any conclusive evidence that management "managed" quarterly income by exercising discretion over quarterly income tax charges, he did indicate that there was some evi— dence that taxes were charged to interim earnings so as to state interim earnings-after-taxes conservatively. Additionally, after evaluating instances where quarterly data had been revised, it was concluded that revision did not make the quarterly data more predictive (as measured by the methodology of the forecasting studies). Indeed, as with other areas of the analysis, it was found that restatement of one year's quar- terly earnings data tended to make the next year's data appear more con- servative by comparison. Staubus attempted to determine which of several lengths of obser— vation period of firms' past earnings proved best in predicting perform— ance of common stocks over subsequent periods of one to twelve years. Performance was measured by discounting the actual cash flows of common stocks over the performance periods at nine percent. For each of several llbid., pp. 103-104. 2Staubus, Journal of Business, Vol. XLI. 21 samples, the average correlation of the performance of the sample common stocks with their related aggregate historical earnings numbers was com- puted for several decision dates. Conclusions were based on the number of times earnings aggregated over an historical earnings observation period of a particular length achieved the highest average correlation with subsequent performance of the stocks in a given sample and decision date. Two of the historical periods chosen by Staubus were the most recent quarter (prior to the cut-off date) and the most recent half year. The conclusion with respect to the two interim periods was that they did not show up well in comparison to observation periods of one, two, three and four years. Surveys of Opinion as to Usefulness of Quarterly Data In addition to the survey of reporting practice reviewed above, Taylor surveyed both executives of companies listed on the American Stock Exchange and financial analysts.1 He reported that 84% of the companies responding indicated strong positive feelings on the usefulness of quar— terly reports. Representative responses from financial analysts indicated that they also strongly felt that quarterly financial data are useful. Like Taylor, Newell surveyed financial analysts as to their feelings about quarterly data. He asked the specific question: "How use- ful do you consider published quarterly reports to be in.yppr_analysis of 2 investment quality of a firms' securities?" The responses that he received were as follows: 1Taylor, The Accounting Review, Vol. XL, No. l. 2Newell, p. 163. 22 peggee of usefulness percent of total response very useful 45% quite useful 36% of limited usefulness 19% of no use 0% The inference drawn was that, "As these analysts feel quarterly data are useful in their analysis, it is presumed that they g§e_these data in their analysis."1 Except for the work of Newell the research reviewed up to this point provides neither information about general investor behavior nor information that would be useful in making decisions aimed at improving quarterly accounting measurements. Whereas the forecasting studies per— haps established that quarterly data can aid in forecasting annual re- sults, they provided no idea of how the performance of quarterly data measures up to potential. Whereas the opinion surveys as to usefulness of quarterly data indicated that significant segments of the financial and corporate communities, when asked, say that they feel quarterly data are useful, they do not indicate that quarterly data are significant in actual investment decisions. Newell's investigation is significant in the sense that it estab- lishes that the potential differences in quality between quarterly and annual accounting measurements actually influence the pattern of reported data in a significant way. This result adds to the conviction of the researcher that significant improvements in quarterly accounting measure- ment can be achieved. In addition, since they suggest that quarterly data 1Ibid., p. 144 (emphasis added). 23 have significant potential to mislead, Newell's findings should tend to heighten interest in the question of how significantly quarterly data influence actual investor decisions (the first research question that will be undertaken in this project). The discussion may now turn to research efforts that approached (with very limited success) the empirical questions of concern in this study. Empirical Studies of the Influence of Quarterly Data on Actual Investor Decisions The studies by George Benston and Brown and Kennelly, that will be briefly described and evaluated below, are given exceptional treatment because they are the only large-scale studies to date that attempted to measure empirically the significance to investor decisions of quarterly accounting data.1 However, as will be seen in the discussion, Benston's efforts to explain investor response to quarterly data, limits the appli- cation of his findings to the research questions of interest in this study. A similar limitation results from Brown and Kennelly's specifi— cation of the kind of news, i.e. good or bad, contained in accounting data and measuring consistency of response with the type of news. Any mismeasurement of the type of news contained in accounting data leads to a mismeasurement of investor response to that data. 1George J. Benston, "Published Corporate Accounting Data and Stock Prices," Empirical Research in Accounting: Selected Studies,_l967, 8 Supplement to Volume V of the Journal of Accounting Research and Phillip Brown and John W. Kennelly, "The Information Content of Quar- terly Earnings: A Clarification and an Extension," forthcoming in the Journal of Business. Thanks to the generous cooperation of Professor Kennelly the researcher has had access to an early draft of the latter paper. The evaluation in the following pages may not apply to the later, published version of the article. :- I h 24 While it may appear to the reader that the discussion in the next few pages is rather negative, no expression of dissatisfaction with the overall efforts of other researchers is intended. It is simply nec- essary to evaluate negative points in order to improve on the approaches taken in earlier work. Furthermore, in all fairness to Benston and Brown and Kennelly it should be pointed out that they did not set out to answer the exact research questions of this study. Hence it is no sur- prise that their findings have not already answered them satisfactorily. Using changes in market prices of common stocks as a reflection of investor decisions concerning the stocks, Benston hypothesized a rela- tionship between the changes in stock prices and the measured rate of change in an accounting variable. His complete relationship, however, included variables other than accounting variables and may best be sum- marized (in the manner that Benston found most significant) in the following functional form: A Pt = F(ARt , QRt , AD , I , U) 3 F indicates that the variable on the left is a function of the variables in brackets. AP is the measured price change (corrected for market-wide changes) around time T. AR is equal to the change in the annual level of an accounting variable, measured as a rate of change. QRt is equal to the rate of change in the same accounting variable, measured for the first three quarters of the current and past year. AD is a variable measuring any rate of change in dividends. I is a dummy variable identifying the industry, j, to which a particular observed firm belongs. U is the residual, unexplained factor, thought to behave randomly. 25 To test the significance of the accounting variables Benston tested the above, hypothetical relationship using the multiple regression equation: APt = a + a ARt + aBQRt + a AD + a I + U. 1 2 53: Using this equation form as well as several more complicated 4 ones, regression analyses were performed for each of several different accounting variables, e.g. net income, net sales, net Operating income plus depreciation, etc. As was noted above the equation with the account- ing variables in the simple rate of change form was most significant, but certain of Benston's conclusions pertain to all forms and variables. The rate of change in dividends was found to be a significant variable in less than half of the regressions and the industry identity in less than 20%. The quarterly data were statistically significant only for the accounting variable "sales". In the case of some Of the income variables the addition of the quarterly data variable, although not significant itself, seemed to increase the significance of the annual data variable. Even with the quarterly data variable, however, and the best functional form, the annual accounting data exhibited a very weak rela- tionship to price changes. In Benston's own words: "...the effects (as measured here) of published accounting data on stock prices are not very great, especially when one considers that the market is capitalizing future expected changes in income."1 But the researcher is cautious in accepting Benston's results. The assumption implicit in all the regression equations that Benston 1Benston, Empirical Research in Accounting: Selected Studies, 1967, p. 22. 26 tried was that annual and first three-quarter measurements pf_the same accounting variable are independently related to end—of-year stock price changes. That is, the rate of change in, say, sales between the current year and last year would have one effect on the stock price around the end of the year, while the rate of change experienced between the first three quarters of this year and last year in the same variable would have a separate and independent effect on the stock price around year—end. This seems to be a highly implausible assumption, indeed. It denies any correlation between quarterly and annual levels of the variables or any cumulative effect of "readings" on the same variables in interim periods. But Benston makes the point himself: ...a shortcoming of the study may be that insufficient atten— tion was given to the specification of quarterly data. A comparison of the final quarter's data with those of the previous three quarters may have proved more fruitful than the comparison made between the third quarter and annual results of succeeding years.1 "...to The intent of the research of Brown and Kennelly was clarify the previous results [of the forecast studies and Staubus' work] and to extend the empirical knowledge of current interim reports in a limited sense."2 As in the forecast and Staubus studies, ability to pre- dict was the criterion selected to judge the usefulness of quarterly earnings data. But the choice of the events predicted and the concept of prediction used were quite different from those employed in the fore- cast and Staubus studies. Hence the researcher feels that it is more 1Ibid., pp. 25-26. 2Brown and Kennelly, "Information Content of Quarterly Earnings," 27 appropriate to view the work of Brown and Kennelly, not as an attempt to clarify the forecast and Staubus studies, but rather as an inquiry into the nature of the impact of quarterly earnings data on common stock prices. In the manner of a previous study, Brown and Kennelly define the information content of an earnings-per-share number relative to the dif- ference between the number and a forecast or expectation of the number.1 If the actual number is larger than the forecast the actual number is considered good news; if the actual number is less than the forecast it is considered bad news; and if there is no difference the number is con— sidered to have zero information content. To test whether this is a good model of the information content of earnings numbers, an hypothetical situation is established wherein a simulated investor is given the sign of the difference between forecast and actual earnings for each of 94 common stocks 12 months in advance of the date that actual earnings become known to the market. If the sign is positive the investor buys the security at the advance date; if negative he sells short; if neither positive nor negative no action is taken. His cumulative monthly investment perform— ance in excess of general market performance is then mapped from the advance date to a date several months subsequent to the time when the actual earnings number first becomes generally available in the market. Performance in excess of market performance is determined by first regressing a firms' monthly rate of return for some sample period on the monthly rate of return of a market index for the same period to 1The methods employed by Brown and Kennelly were first used in Ray Ball and Phillip Brown, "An Empirical Evaluation of Accounting Income Numbers," Journal of Accounting Research, Vol. VI, No. 2 (Autumn, 1968), pp. 159-178. 28 establish an average relationship. Then the excess or unexpected rate of return for the month of interest is computed as the difference between the actual rate of return and the rate of return predicted by the regres- sion relationship, given the actual market rate of return for the month. The justification for this technique is strongly supported empirically, and will be fully elaborated in a later chapter. If the excess return is positive and the hypothetical investor is holding the security or if the excess return is negative and the in— vestor has sold the security short, he has gained and conversely. The cumulative performance of the investor from the advance date, T, to the end Of any subsequent month, M, is the product: M C = (1+rt) t=T where rt is the excess rate of return for the month t for a particular security with its sign adjusted according to the strategy of the investor. The more frequently the investor experiences large positive rt's the larger C will be. That is, the more often the signs of the unadjusted excess returns of months up to and including the month of announcement of the actual earnings number are consistent with the sign of the difference between forecast and actual earnings, the greater will be the cumulative performance of the strategy that is determined by the sign of that difference. 1The upper-case greek letter pi, H, is used to represent iterative multiplication in the same way that the upper-case sigma, Z, is used to represent iterative addition. nub Tut p, 29 Besides measuring the aggregate cumulative performance C, Brown and Kennelly noted the frequency of consistency between (1) the signs of monthly excess rates of return and (2) the signs of annual earnings fore- cast errors for the investment strategy based on foreknowledge of annual earnings forecast errors. They then constructed an analogous strategy based on foreknowledge of the signs of forecast errors for the three quarterly earnings numbers. It essentially permits a switching of posi- tions in each stock starting with the month after each quarterly earnings announcement based on prior knowledge of the error in forecasting the next quarterly earnings number. Their hypothesis was essentially that if the quarterly numbers have information content, then strategies exploiting foreknowledge of that content would do better than strategies exploiting foreknowledge of the content of the annual number only. Their results confirmed their hypothesis. Based on the cumulative excess returns their conclusion was that "...interim reports increase apparently by some 30-40 per cent, the value Of information contained in annual EPS."1 As did Ball and Brown in an earlier study, the authors found that for strategies based only on annual earnings forecast errors, "...the markets's anticipation of annual EPS is sufficiently accurate that its release does not appear to cause any unusual jumps ...in the announcement month."2 In addition, when quarterly switching of strategies was permitted the authors found that the months of the first, second, and third quarter 1Brown and Kennelly, "Information Content of Quarterly Earnings," p. 10. 21bid., p. 11. 3 . TI" 1 1.. . ‘ . “.51. I.-.‘ s-g. :‘v. ‘- v._ 30 earnings announcements were months of remarkable excess returns whereas the month of the fourth quarter (annual) was not nearly as remarkable. In addition, the frequency of agreement in sign between forecast errors and excess rates of return for months of quarterly announcements was more highly significant than for months of annual announcements. They "...suggests again that annual EPS when it is conclude that this pattern finally released is not usually newsworthy, although the previous three quarterly reports may be of interest to investors."1 Although Brown and Kennelly's findings establish at least that the direction of investor response to quarterly announcements is signifi— cantly related to the direction of forecast error, the researcher is cautious in accepting their conclusion on the comparative newsworthiness of quarterly and annual earnings announcements. There is at least one limitation of the Brown and Kennelly design that renders such a conclu- sion tenuous. As the authors put it: "It is clear that the validity of the exercise depends largely on how well earnings reports are classified into 'good', 'bad', and 'indifferent.'"2 If their forecasts of earnings are poor characterizations of market forecasts, then the signs of their forecast errors which govern the classification of earnings reports could be poor estimates Of the kind of information contained in the reports, i.e. good news, bad news, etc. Thus their measurement of infor- mation content is subject to measurement error due to misspecification of forecast earnings. Because of the importance of this limitation of their lIbid. 2Ibid., p. 4. 31 design, discussion of the three specific forecast models used by Brown * and Kennelly has been postponed until now. One of the forecast models employed, a regression relation with a market earnings number, is analogous to the model described above for computing excess monthly rate of return of common stocks. However, the authors give little credence to its use. In their own words the model was included "...more to show the specification and estimation weaknesses than for any other purpose."1 In addition to the regression model the authors use two models that they call "naive". The first consists of forecasting current period earnings equal to the earnings of the comparable period of the prior year. If the Object of forecast is annual earnings, the forecast will be the prior years' earnings. If quarterly earnings is the object of interest, the earnings of the same quarter of the prior year is the forecast amount. The second naive model makes the same type of forecast but adds an his— torical average change in earnings of the comparable period of the prior year. In order for Brown and Kennelly's conclusions regarding the rela- tive newsworthiness of quarterly and annual earnings numbers to be valid, these forecast models must be equally good characterizations of market expectations of both quarterly and annual earnings numbers. That the forecasts of annual earnings numbers do not incorporate any of the infor- mation contained in the quarterly numbers ignores the widely held belief that investors use quarterly data to forecast annual results. Although nonrigorous, this reason alone is enough to cause the reviewer to doubt 1Ibid., p. 6. ..I . ‘ v; .n nov- .1. I'll. II II‘I. a : 29;: 0. to. u... H . ..4, . ' ’H‘M~I ’ w ~ua.“ v ‘-~._.‘~ ..v{‘ 5.5 "E-. 'u. I . ~-..-". g "‘ I‘D. .: ; ‘1 .f'sc. . . 32 the homogeneity of Brown and Kennelly's measurements with respect to com- parisons of newsworthiness of quarterly and annual earnings numbers. Thus the reviewer agrees that Brown and Kennelly's results sup- port a conclusion that their forecast models were remarkably good esti— mates of market expectations of quarterly earnings numbers but not of annual earnings numbers. Their results do not support the conclusion that "...annual EPS when finally released is not usually newsworthy, although the previous three quarterly reports may be of interest to investors."1 The purpose of this study is to attempt to measure investor response to quarterly and annual earnings announcements in a way unclouded by potential specification error of the types present in both the work of Brown and Kennelly and Benston. The next chapter begins the elaboration of a measurement methodology intended to serve this purpose. 1Ibid., p. 11. CHAPTER III STOCK PRICES, EXPECTATIONS AND INFORMATION The starting point in studying the impact of quarterly financial data on investors' decisions must be a theory that relates financial in- formation to those decisions. A widely accepted theory of stock prices provides a basis for assuming a cause-and-effect relationship between financial information flows, in the form of quarterly data, and stock price changes. The theory holds that the price of a common stock at a point in time is equal to the present value of the expected future cash flows to the holder of a share of the stock, discounted at the expected Opportunity rate of return for the expected level of risk attendant upon the flows. Each investor forms his own expectations about future flows, risk, and the opportunity rate of return and arrives at his own price. If his price is different from the price at which he can buy and sell he will presumably change his holdings. In the aggregate the buying and selling activity of individuals whose valuations differ from a particular market price will change the price in the direction of the difference. A change in the price of a common stock may therefore be caused by a change in expectations regarding any or all of the elements of the 1Benjamin Graham, David L. Dodd, and Signey Cottle, Security Analysis: Principles and Technique. (fourth ed.; New York: McGraw— Hill Book Company, 1962), p. 450. 33 /f 34 theoretical relationship. Expectations change in response to new stimuli that are not perfectly consistent with expectations just prior to their perception. That is, new information that is not exactly as expected, has the power to change expectations with respect to related events that are yet to occur. Therefore any new bit of information about events or condi- tions related to the elements of the theoretical basis of a stock's price has the power to change expectations with respect to those elements, and hence, to change the stock's price as well. Thus, in theory at least, one can guage the effect of a particular bit of new information by measuring the change in a stock's price that re- sulted from it. In practice, however, it is very difficult to attribute a particular change in a stock's price to a particular bit of new information. Nevertheless, this is the approach that will be taken in this study. It is considered the most relevant approach to the study of the impact of quar- terly financial data on investors' decisions, because changes in stock prices are considered to be the best reflection of actual investor decisions. Fortunately, the researcher is not without reported research aimed directly at the Operational problem of relating a specific bit of new infor- mation to a specific market price change as well as methodology suggested by similar studies that had to deal with the problem. In particular, the methodology arrived at in this chapter and the next to satisfy the general research questions of this project will be an elaboration (with similar results) of the methodology develOped and used by Beaver in his study of ‘what he termed the "information content" Of annual earnings reports. 1William H. Beaver, "The Information Content of Annual Earnings .Announcements," Empirical Research in Accounting: Selected Studies, 1968, a supplement to Volume 5 of the Journal of Accountinngesearch, pp. 67-92. 35 Rather than review that work in detail and then proceed to the research questions at hand, though, the researcher prefers to elaborate the metho- dology in a manner that he believes will be better understood by the reader. Information Classes and Correction of Stock Price Changes for "Market Effects" Benjamin King attempted to isolate the effects of different classes of information on the stock price of a firm.1 He started with the following proposition concerning information inflows to the stock market: The stock market is subjected to a steady flow of infor- formation, much of which will have an effect on the set of anticipations that determine the price of security j. This does not mean, however, that all of the information affecting j affects j only. In fact, it is intuitively appealing to think of incoming information as falling into various classes according to the SCOpe of its effects on the market. King proceeded by factor analysis to determine to what extent market-wide and industry "effects" explained the variances in stock price changes for 63 firms listed on the NYSE from 1927 through 1960. Some of his findings that are of interest here are that: 1. The average prOportion of the variance of each security's price changes explained by the mean change of the other securities in the sample (representing the market as a whole), from August, 1952 through December, 1960, was .307. 1Benjamin F. King, "Market and Industry Factors in Stock Price Behavior," Security Prices: A Supplement, Journal of Business, Vol. XXXIX, No. 1, Part 2 (January, 1966), pp. 139-190. 21bid., p. 140. 36 2. The proportion of variance explained by the SEC two-digit industry classification is .113. The implication of King's findings is that in trying to relate a bit of new information, specific to a firm, to a market price change, it would be fruitful to first correct for the portion of the change in the stock price that is attributable to factors affecting all stocks in the market and, perhaps, all stocks in the same industry as well. Such a correction for market-wide factors will be made, following a method first suggested by Sharpe,1 and evaluated and found to be "... a satisfactory method for abstracting from the effects of general market conditions as monthly rates of return on individual securities"2 by Fama, _ep.§l. The method has been adopted in several studies related in metho- dology to this study.3 It consists of the following steps: 1. First, all price observations for a particular firm to be included in the study are corrected to an equivalent capital basis. Essentially this means adjustments will be made retroactively for any stock dividends or splits that occur during the periods from which observations will be taken. For reasons that will be discussed later, obser— vations of weekly closing prices are contemplated for the study. 1William F. Sharpe, "A Simplified Model for Portfolio Analysis," Management Science, Vol. IX, No. 2 (January, 1963), pp. 277-293. 2Eugene F. Fama, e£_§i,, "The Adjustment of Stock Prices to New Information," International Economic Review, Vol. X, No. 1 (February, 1969). pp. 1-21. 3See, for instance, Beaver, 1968 Supplement to Journal of Account- ing Research, VI, Ball and Brown, Journal of Accounting Research, VI, NO. 2, or Benston, 1967 Supplement to Journal of Accounting Research, V. 37 2. Second, parameter estimates, aj and b , for the follow- J ing relationship will be obtained by regression of observed values of the variables over the period of observation: P. = A, + B,M + R, Jt J J t Jt Pjt is the difference between the natural logarithm of the equivalent closing price1 of the jth stock at the end of week t and the natural logarithm of the closing price at the end of t - l ( = 1n (pjt + Dj ) - 1n (pjt-l) ). Pjt t Mt is the difference between the logarithm of the closing value of a market index at the end of week t and the logarithm of the closing value at the end of week t - 1 (Mt = 1n m - 1n ). Logarithmic measurements will be m t t-l used because they have been found by other researchers to better satisfy the assumptions of the linear regression model.2 Rjt is a term representing the non—market related portion of Pjt' Rjt is assumed to be a random term in the above relationship. Aj and Bj are the intercept and slope, respectively, of the assumed underlying relationship. (a J and b are sample estimates of Aj and B , respectively. J j 1D is the amount of cash dividend, if any, that became the legal righltof owners of stock j during week t. It is added to the closing price of week t to correct it to a basis comparable to the closing price of week t-l. 2Fama, et al., International Economic Review, X, No. 1, p. 4, note 8. 38 3. After a3. and bj have been found for each stock, the above relationship will be reversed to give the values of the non-market-related portions of the price change measure- ments: R, = P, - (a, +b_M Jt Jt J J t) The Rjt will then represent a set of weekly price change measurements for each firm that have been "corrected" for factors that have affected all stocks in general. Each stock's price change measurements will be corrected according to its uniquely estimated relationship to the market in general, represented by its unique aj and bj. Because of the expected high cost (effort) and low return in addi- tional accuracy, no correction is planned for industry effects. Such a correction would require considerable effort. Whereas indexes of market prices are readily available, the researcher would have to construct in- dexes of two-digit industry classes. In order to economize on this effort, an undesirable restriction of observations to relatively few industries might be necessary. However, since King found that the proportion of stock-price— change variance eXplained by industry effects was only .113 for price changes measured over month—long periods, it is eXpected that the degree of precision lost by not correcting for industry effects in weekly price changes is negligible. Price changes measured over shorter periods can be expected to display much less systematic relation to industry (and market) effects and much more randomness of fluctuation than price changes measured over longer monthly periods. However, the relatively 39 strong relationship found between monthly stock price changes and the changes in the total market suggests that a significant relationship may exist even on a weekly basis. Hence, the correction for market movements will be made even though it is expected that much less than 30% of the variance of the weekly stock price changes will be explained by the correction. Specific Information Flows and Measured Stock Price Responses After correcting for influences that affect the changes in all stock prices, the remaining or residual price changes may be attributable to new bits of information specific to the individual firm plus random factors. (Industry effects will be assumed to be incorporated in this random element.) There still remains the problem of relating specific new bits of information to specific price changes. To solve this problem some of the basic elements of the "random walk" theory of stock price behavior will be relied upon. The Random Walk Theory The random walk theory is based on the assumption that the major security exchanges approximate "efficient" markets.1 An efficient market is one in which at any point in time security prices already reflect in- formation about relevant events that have occurred and relevant events that the market expects to take place in the future. In an efficient ‘market the price of a security at any point in time will approximate its intrinsic value. (Intrinsic value is used here in the sense of equilibrium 1Eugene F. Fama, "Random Walks in Stock Market Prices," Financial Iknalysts Journal, Vol. XXI, NO. 5 (September-October, 1965), pp. 55-59. 40 value) The price will diverge from intrinsic value but the actions of many informed and intelligent market participants will prevent the price from deviating from intrinsic value (by more than commission costs) in any way other than by chance. Hence, actual prices will wander about intrinsic values randomly. Intrinsic value may change in response to new information of all kinds. Theoretically, in an efficient market the transition to a new intrinsic value in response to new information will be "instantaneous." In practice, instantaneous means at least two things: 1. The market overadjusts as often as it underadjusts, i.e., the expected value of the new price is the new instrinsic value. (The expected value of the price at any point in time before the new information became available was the prior intrinsic value.) 2. The time lag between the first perception of the changed condition and complete adjustment is, itself, a random variable. This implies that successive price changes will be independent. No period's change in stock price can be predicted from the preceding period's change. Even if the moment of first perception of a new bit Of information is known, the shift in intrinsic value will not occur at a predictable moment (except that with intelligent and informed market participants the maximum delay will be short). Ibid. 41 Relating Specific Information to Specific Price Changes Ideally, the way to relate a particular new bit of information to the attendant shift in intrinsic value is to pinpoint the moment of first perception and observe the stock price movement over a span of time long enough to include the maximum lag before response, but short enough to exclude responses to other bits of new information specific to the firm. On the average, measurements over such periods would be without bias. That is, the expected value of the errors in measurement would be equal to zero. To illustrate, consider a model representing the features of the random walk theory discussed above. The price Ob- served at any moment, t, is a function of the intrinsic value at t, IVt, plus a random term, rt, with an expected value of zero. P = IV + r t t t The change in price over a given period can therefore be represented as AP(t,t+1) = IVt+1 + rt+1 - IVt ' rt' The expected value of this difference is E[AP] = E[IVt+l — IVt] + E[rt+l - rt] = E[IVt+1 - IVt] + 0 = E[Ivt+1 - Ivt] = IVt+1 - Ivt which is the difference in the intrinsic values. It is clear that as a practical matter no period will completely satisfy the two conditions mentioned above, i.e. sufficiently short to 42 exclude shifts in intrinsic value due to other bits of information and yet long enough to include the lagged response to the bit of information under study. Of great help in this choice, however, are the corrections that have already been suggested to abstract from influences of information affecting the whole market. They cut the number of bits of significant information down to those that pertain only to the individual security whose price is being observed. (Among these relatively few bits of new information are the quarterly announcements that are the subject of this study.) Even with the benefit of this shrinking of the number of bits of information that could interfere with the shift in intrinsic value that we wish to associate with a quarterly earnings announcement, the choice of a period of measured response is still largely arbitrary. But, with the two conditions mentioned above in mind, a satisfactory, though arbitrary response period can be selected. Selection of Price-Change-Resppnse Period The period selected for measuring price changes attributable to quarterly announcements is the week in which the announcement is made. The attributes that make this particular length of period a good choice are listed below: 1. The week is not so long as to create insurmountable difficulties in generally finding observations wherein the only new specific bit of significant information entering the market during the response period is the quarterly or annual earnings announcement. 2. Weekly price changes, Observed without regard to the 43 effects of specific bits of new information, have been found to behave in very nearly random fashion. Serial correlation tests of successive weekly price changes conducted by Cootner and others have yielded sample correlation coefficients extremely close to zero. 3. Beaver conducted a study of the impact of annual finan— cial data on stock prices.2 Judging from his results, a week is a sufficiently long period to pick up responses that lag behind the first perception of the annual announcements. The graph shown in Figure 1, typical of Beaver's results, bears out this judgment. It represents the average of a ratio based on changes in prices, after adjustment for market effects, in the week of the annual earnings announcements, week 0, and each of the eight weeks before and eight weeks after. The averages were computed over 143 sample firms and several cross-section years. The remarkable average price change ratio in week zero is consistent with with the discussion above and with the predicted behavior pattern of price changes if annual earnings announcements in general have sufficient information content to alter the intrinsic value of securities. 1Paul H. Cootner, "Stock Prices: Random vs. Systematic Changes,‘ Industrial Management Review, Vol. III, No. 2 (Spring, 1962), pp. 24—45. 2Beaver, 1968 Supplement to the Journal of Accounting Research, VI. 44 U = ratio of squared weekly price 1.6 movement to average for weeks outside 1.5 the 17-week period 1.4 1.3 Average squared weekly l 2 price movement outside ° 17 weeks period 1.1 1.0 ________________ - .90 .80 Source: William 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. Figure l.--Profile of mean price-response ratios for 17 weeks surrounding 506 annual earnings announcement dates of 143 firms for years 1961 through 1965. Pinpointing the Moment of First Perception of Announced Earnings Regardless of the choice of length Of period for observation of price-movement reSponses, the measurements that will be used in the study still depend on recording price-change responses in periods imme— diately following the markets' first perception of quarterly and annual financial data. The day of first perception is to be pinpointed by reference to the date of publication of the earnings numbers in the 45 Eastern edition of the Wall Street Journal.1 In choosing this method of dating the information, reliance will be placed on general conformity to the disclosure requirements of the major exchanges and on the publica— tion policies of Dow Jones & Co., Inc. The American Stock Exchange, for instance requires: As to procedure, information of this type should be issued as quickly as circumstances permit, to as broad an audience as possible, on an "immediate release" or anp held'I basis. It should be distributed to one or more New York City newspapers which regularly publish financial news, to the newsticker ser- vices operated by Dow Jones & Company, Inc. and Reuters Eco- nomic Services and to the news-wire services of the Associated Press and United Press International. When, in spite Of proper precautions, inaccurate information or rumors, true or false, are circulated, companies are expected to cla ify the situ— ation promptly through a public announcement. According to a telephone interview with Mr. Ralph Doddridge, Head of Research, Dow Jones and Company, Inc., the flow of earnings announcements to the public through the media supplied by his firm is as follows: 1. Earnings announcements are received by Dow Jones and Company, Inc. via telephone, wire, mail or public relations newswire service. 2. If received indirectly, the information contained in the announcement is confirmed by direct telephone conver— sation with a company officer. Questionable information, from direct or indirect sources, is clarified by the same means. 1The Wall Street Journal Index (New York: Dow Jones and Company, Inc., Issued Monthly with annual cumulations). 2American Stock Exchange. ”Listing Form L." (Revised September 15, 1966). 46 3. Information of satisfactory accuracy is released over the "broad tape" immediately and appears in the next issue of each edition Of the Wall Street Journal. According to Mr. Doddridge, Dow Jones is rarely "scOOped" in the publication of corporate earnings information. However, if a particular announcement is cleared for release at a time of day too late to make that day's issue of the Eastern edition of the ES; it will be printed in that edition on the following day. Since the Eastern edition is the edi— tion that will be used in this study to determine the week of first per— ception of an earnings announcement, the time basis of associating price- change responses to earnings announcements described earlier might occa— sionally be thwarted, e.g. an announcement appearing on the "broad tape" or in the Midwestern edition of the W§J_on Friday of one week might not appear in the earlier Eastern edition until Monday of the following week. Depending on just how "instantaneous" the market response is to bits of new information, the methodology of the study is vulnerable to error in pinpointing the time of response to earnings announcements selected for study. But since the Wall Street Journal Index is only available for the Eastern edition, there is no practical way for the researcher to refine his dating methodology, e.g. by pinpointing the first publication of an earnings announcement in any edition of the Wall Street Journal. Instead an attempt will be made to gauge the degree of "leakage" in the methodology from all sources, including the one just described, using measurements to be developed in a later chapter. Summary This chapter has been devoted to showing in general terms how the effect of specific bits of new information on investors decisions can be 47 observed by careful measurement of the residual price response in a period immediately following the market's first perception of the bit of information. Sharpe's method of abstracting from the effects of information affecting the market in general is adopted. Portions Of individual firms' price changes not explained by changes in an appropriate market index, will be assumed to be composed of changes attributable to specific information related only to the firm, plus random changes. The moment of first market perception of the quarterly earnings numbers of firms will be determined by the dates of publication of the earnings numbers in the Wall Street Journal. It is expected that the "instantaneous," non-random response, if any, to the quarterly earnings numbers will take place in the weak in which first perception occurs. The next chapter will be devoted to Operationalizing the general research questions of the study in terms of measured residual price changes in weeks of quarterly earnings announcements. CHAPTER IV THE BEHAVIOR AND SIGNIFICANCE OF PRICE—CHANGE RESPONSES TO QUARTERLY DATA The first research question mentioned in Chapter I calls for a determination of the significance of price-change responses in weeks of quarterly earnings announcements. Questioning the significance of a response implies a standard of comparison. The standard of comparison for price responses to quarterly earnings announcements will be the average price response for all weeks of the year, excluding weeks of earnings announcements. The first principal hypothesis of this study is that quarterly earnings announcements have a significant effect on investor expectations. Since changes in expectations lead to changes in intrinsic value, sus— tained (non-random) price changes should take place in periods when quarterly earnings announcements are made. On the average, then, price movements should tend to be greater in the weeks of quarterly earnings announcements than in other weeks of the year. Other weeks in the year may experience bits of information that can by chance have a significant effect on investor expectations. But every week of a quarterly earnings announcement may be considered to have a bit of new information with such potential, if the principal hypothesis is correct. Thus to test the significance of the effects of 48 49 quarterly earnings numbers on investor expectations, it seems reasonable to test the significance of price changes in weeks of quarterly earnings announcements relative to other weeks in the year. To facilitate such a test the standard of comparison will be directly incorporated in the measurements employed. Recall that the weekly price changes for each firm will already be converted, in the manner described in the previous chapter, to residual price changes, Rjt’ free of estimated effects of market-wide influences. The transformation of these measurements into a form well suited to the satisfaction of the first research question will be accomplished in three steps: 1. Since the research question is not concerned with the direction of changes in investor expectations but only their magnitude, the Rjt will be converted to their absolute values, letl-l 2. The average value of the 'Rjt' will be computed for each firm, excluding 'Rjt' of the weeks of earnings announce- ments, i.e. [Egg] = fi-E 'Rjt| where t takes on the num- bers of all weeks in the study period except weeks of earnings announcements. 3. Finally, the ratio IR l + IRth will be computed for each jt To accomplish the same purpose Beaver reported using a measure of magnitude of price response equivalent to R t2 in his study of the price-change response to annual earnings announcements appearing in the 1968 Supplement to the Journal of Accounting Research. The choice of IRth in this study was determined by a desire to minimize the effects on average measurements of a possible few large price change responses occurring in weeks of earnings announcements. Since squaring a number gives disproportionate weight to size as well as eliminating the sign, Beaver's method was not adopted. 50 week in the research period. For convenience the ratio will hereinafter be denoted RRjt' Note that the average of these ratios over all the weeks of the study period except weeks of earnings announcements, will be 1.0. The Significance of Price-Change Resppnses to Quarterly Announcements The average ratio RRj: for the weeks of all the quarterly announcements of a firm may be considered a random variable for which observations may be gathered over the firms in a sample. For each firm, the ratio measures the average relationship between the price change in the week of the announcement and the average weekly price change that the firm experiences throughout the study period. If there were nothing unusual about the price changes in weeks of quarterly earnings announce— ments for individual firms, then these ratios would have an expected value of 1.0, the average of the ratios of all other weeks. Thus, the significance of the price changes that firms experience in the weeks of quarterly earnings announcements can be tested in the form of the follow- ing set of hypotheses: 1. Null hypothesis, H0: The mean of the average ratios, RR.*, jt of quarterly earnings announcement weeks is less than or equal £p_l.0. 2. Alternative hypothesis, H1: The mean of the average ratio is greater than 1.0. Since the sample of firms is intended to be large, the Central Limit Theorem may be relied upon in choosing the '2 test" to be applied 51 to the sample mean of the average ratios.1 The null hypothesis will be rejected if the "z" statistic is greater than the critical value, 2*, for the level of significance that will be chosen (The level will be at least as significant as .05). The "z" statistic will be computed as follows: 2 = (2-1) where R is the sample mean of the average ratios, i.e. — l x =._ n HMS E rim jl 3 s is the square root of the sample variance, n is the number of firms in the sample, and N is the finite number of firms from which the sample is drawn. If the null hypothesis cannot be rejected, then it cannot be con— cluded on the basis Of this test that the price changes firms experience in weeks Of quarterly earnings announcements are, on the average, signifi- cantly different from the price changes experienced by the same firms in other weeks. Although it is important to know the significance of price re- sponses in the weeks of quarterly earnings announcements on an average basis, such averages may be distorted by very large price changes eXpe- rienced by relatively few firms. It is therefore useful as an additional test of the significance of quarterly earnings numbers, to determine the significance of the frequency with which "significant" price changes are 1John E. Freund, Mathematical Statistics (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1962), p. 263. 52 experienced in the week of the earnings announcement by individual firms. The researcher is not aware of a technique that can be used to test both the significance of the individual price changes and the sig— nificance of the frequency of their occurrence in "zero—week" for indi- vidual firms at the same time. (The term "zero—week" is used to refer to the week of an earnings announcement for a particular firm.) Hence the significance of zero-week price changes will be a matter of defini- tion and the frequency of significant zero-week price changes will be subject to test. Within each firm the significance of average zero-week price changes will be determined by rankings relative to surrounding weeks. The 11 weeks including zero-week and the five before and after will be compared according to the size of their average ratios, RRjI . An average zero-week ratio will be considered significant if it is the largest, i.e. highest ranking, ratio in the 11 week profile of average ratios. To test the significance of the frequency of significant week zero price changes across firms, a test based on the hypergeometric dis- tribution will be used.1 If ranks were assigned to the 11 weeks on a uniformly random basis, the probability of any firm's zero week receiv- ing the highest rank would be 1/11. 1. The null hypothesis, H0, is therefore that the probability of significant zero-week price changes is p : 1/11. 2. The alternate hypothesis, H1, is that p > 1/11. 1Freund, pp. 70-71. 53 If the probability of obtaining the sample count of x or more highest zero-week ranks in n firms, assuming p = l/ll, is sufficiently low (as low as .05) the null hypothesis will be rejected and it will be concluded that p i§_greater than 1/11, i.e. that there is a significant (non—random) rate of occurrence of significant average price responses to quarterly earnings announcements among firms. The probability, of O 0 O O O O l Obtaining x or more Significant ranks Will be computed as follows: 8 Cd N-d W = X r n-r r=x C 520 where d is the number of occurrences in the field eXpected under the null hypothesis, x is the actual number of occurrences in the sample, n is the sample size, and N is the size of the finite population. r is the hypothetical number of occurrences in the sample starting with x, the number actually eXperienced, and going to e, the minimum of n and d (the maximum number of occurrences that could possibly be experienced.) 1Notation of the form CX indicates the number of unique combina- tions of y number of items that can be formed from x available items. Computationally x X! C = -———————— y y!(x-y)! where "I" is said "factorial" and x! means x-(x-l) -(x—2) --—-(l). 54 The Response to Quarterly Data Relative to the Response to Annual Data The second research question mentioned in Chapter 1 calls for a comparison of the degree of price-change response to quarterly and annual earnings announcements. An investor who is "knowledgeable" should be aware of the limita- tions of the accounting information he uses. He should be aware for instance of the problems associated with assigning a particular earnings numbers to a particular year. The more limitations that a particular bit of accounting information is subject to, the less should the knowledge— able investor rely on that bit of information. Quarterly earnings figures are subject to considerably more limitations than annual earnings numbers.1 In other words, quarterly earnings numbers may be considered to be subject to more measurement error than annual earnings numbers. That accountants are aware of this difference is unquestionable.2 Their awareness of these limitations is largely responsible for their resistance to early SEC prOposals to adopt quarterly reporting requirements.3 That at least some members of the financial community are aware of the differences between quarterly and annual data (in measurement error potential) is reflected in the following concluding remarks from a recent article appearing in the 1A brief discussion of additional limitations Of quarterly earnings relative to annual earnings was given above in Chapter I. For example, see Blough, Journal of Accountancy, VC, NO. 2, and Newell, p. 123. 3Taylor, Accounting Review, Vol. XL, No. 1, pp. 89-96. 55 Financial Analysts Journal: Certainly, there is a substantial difference between the quality of information in the annual report and in interim statements. It would seem wise to regard interim reports, at a minimum in the same manner that one views direct statements from the company's management and not in the same way that one might view audited financial statements. If a large percentage of investors are truly aware of the differ— ence in reliability between quarterly and annual earnings numbers, then quarterly earnings numbers should have less potential for changing in— vestors' expectations than annual earnings numbers. Less potential to change expectations, on the average, should lead to smaller average price changes in weeks of quarterly earnings announcements than in weeks of annual earnings announcements. This essentially is the second principal hypothesis Of the study. Just as the standard of comparison for the first research ques- tion, average price changes in non—announcement weeks, was built into the measurements, a single measure incorporating a comparison of the average responses to quarterly and annual earnings numbers can be readily constructed. The average ratio, RR * , within each firm, can be computed jt separately for all quarterly and all annual earnings announcement weeks in the study period. The average quarterly ratio can then be subtracted from the average annual ratio to arrive at a difference measure for each firm. If indeed there is no difference between investor response to quar— terly earnings numbers and investor response to annual earnings numbers, the differences between average ratios will be purely chance determined 1Seidler and Benjes, Financial Analysts Journal, Vol. XXIII, NO. 5, p. 114. 56 and will have a long—run expected value of zero. Thus the significance Of the difference between investor response to quarterly data and investor response to annual data can be tested in the form of the following set of hypotheses: l. The null hypothesis, Ho’ is that the difference, among firms, between the average price change ratio in weeks of annual announcements and the average price change ratio in weeks of quarterly earnings announcements is less than or equal to zero. 2. The alternate hypothesis is that the difference is greater than zero. Since we are referring to the same large sample of firms as before, the Central Limit Theorem may again be relied upon in satisfying the assumption of a normal distribution required for the "2 test" to be applied to the sample mean difference in average ratios. The null hypothesis will be rejected if the "2" statistic is greater than the critical value, 2*, for the desired level of significance. For this test the "2" statistic will be computed as follows: >ma muwum mwsmzoxm Mooum amOHuma02 .oomH mo mxoos Ono GH muGoEmocsoccm mwcHaumm mo mocmvHoaH mo >0sm:voumnl.w mustm Boo . mmm . OD¢ . wHDh . MZDN . M6H no. man up ooom. span upsapEA Napamoamaawam RA .H6>6H mo. 6:6 up ooom. spas pmmsma Smucpoamsawam « quH mas ace mom NNm Ham mcomapmasoo swam mo monesz esemez qum. kNwmm. 000m. wQOm. NHNm. omHm. m+ meow. mNom. MNHm. qmmq. ooom. oHom. q+ Hon. onq. mums. «Nomm. mamm. mqu. m+ Hmmq. HNNm. kkmoqq. mem. «NHm. Hmmq. N+ qomq. mums. qum. qqu. coma. Nwwq. H+ meme. «com. Nwmq. mm0m. qmmq. onq. H1 HNom. qoqq. wwwq. mmmq. NHNm. mNNm. NI «mem. ammo. «mam. mNmm. mmHm. NHHm. mu «Nomm. qmmq. Noon. «omm. moan. «mNnm. q: «maNm. «meow. mmNm. boom. ooom. «mNom. mu mocHnEoo unassumum Hmscc< umuumso smuumso swuumno ucmemocsoscm ucoEmOcsocc< mHaHuHsz mum ch umH ou O>HumHOm HH< x663 ucmEmossocc< wo mums x663 Basemucsoccm was aH omemau MOHRN was mm cmHm mama on“ mcH>mn meOB ucmsvmmnnm mam HOHHQ CH mwmcmno OOHMQ mo cowuuoqoumal.m mqmuOmno mo muonesz How mmuomuuoo mwmum>< meOOB Nm mm NN 0N mN mmmum>< xHxOOB mumstouma< NM. H % mum m... T oN 0N MN «N mN NI Nm Hm NN Nm Nm MI N« om om wN Nm «I mN mm «N HN mm mI oN ow Hm 0N MN oI cm 00 mN om N ml Hmscc< mumupmso umuumso smuumno umuumpo momsmoosoes< HH< mum oaN umH ou m>HumHmm ucoEmocsocc< mo make xmmz mocmemocsosem mwchsmO ou uOHun mxmm3 Ono SH moeo>o m3os sonuo wo mOHosODvOHmII.w mqmummno mo moonEsz pom mmuoouuoo ommum>< Nmemz a NN m N o owmoo>< NHNOOB oumeonsam< IN. .mIIN. NIH mII m” T o mH « N « NI 0 Hm m HH HH MI «H NN w o w «I « «N m N m mI m ON N N 0 0I m «H o N H NI Hmscc< mHOonso umusmso smuumso smuumso ucoemocsoss< HHa pom pom omH cu O>HumHom ocmemocsocc< mo meme xmmz mucmemocsoscm mwchumO Ou HOHHQ mxmo3 onu CH mocm>m mama moomHou ocmmH>Ho wo mOHOGOJUOHMII.m mqmmHmII.HH Ouamwm NooH oomH mooH f , >75)ij «:6 <. E a: 2» fie/75?; menu—«QNGGEIN HMHH—JF; 123 For simplicity the week zero in each profile is identified simply by a large dot. Figure 11 does not represent a plot of response ratios averaged over successive calendar weeks. Rather, it is the result of lining up the surrounding weekly price response ratios for all earnings announce— ments occurring in any week of a particular calendar quarter. The tails of the profiles of individual earnings announcements that occurred early or later in a calendar quarter actually lie outside the calendar quarter, but are combined with other announcements' profiles completely contained within the calendar quarter. But because of the great amount of chance variability in individual price responses, some method of aggregation was necessary to bring out the contrast in investor response to earnings over time. For the same reason, only the middle three (complete) calendar years were included on Figure 11. Although the observation of price changes began as early as the third quarter of 1964 for some firms and ended as late as the second quarter of 1968 for others, depending on fiscal year—end and other considerations, few firms overlapped at both (ands. Thus the profiles for the last two quarters of 1964 and the first two quarters of 1968 are based on too few announcements to be very mean- ingful. Furthermore the 1964 and 1968 profiles differ with respect to variables other than time, e.g. with respect to composition of firms contributing to the profiles. Because of the nature of the response ratios employed in the study an evaluation of Figure 11 must proceed in two stages. Recall that the demoninator of the response ratios for each firm is the average 124 absolute value of the weekly residual price responses for the firm over the whole study period, excluding weeks of announcements. The numerators are the absolute values of the individual weekly residual price responses. Thus any aggregate pattern in the absolute value of firms' residuals over the study period will Show up in their aggregated response ratios for "other weeks” in the profiles. That is, if residual price responses were generally smaller in magnitude than average early in the study period and larger later, the "other weeks" average ratios in the profiles would tend to be lower than the 1.0 line in early quarters and higher later. Some evidence of such a pattern exists in the low profiles of the second and third quarters of 1965 and the slightly higher profiles of some of the later quarters in Figure 11. Furthermore, there appears to be a gentle oscillation within years, i.e. the second and third calendar quarters' profiles are generally lower than the first and fourth (but only slightly in years other than 1965). The implication of such patterns, of course, is that if one desired to make detailed comparisons of investor response to earnings between periods as short as a year or shorter, some refinement in computing reSponse ratios would be necessary. One possibility would be to compute the average absolute values of residual responses (the base of the response ratios) over shorter, more localized periods. But since some of the profiles appear to run "up hill" or "down hill" within calendar quarters, it is not altogether clear that even a period length as short as a quarter would be wholly satisfactory. 125 Perhaps what is more certain is that whatever period is chosen for computing the bases of measurements similar to those employed in this study, some variability between aggregates over subperiods is to be expected. Comparisons between subperiods should be avoided, therefore, unless "base" measurements are computed separately for each subperiod. Since the comparisons between time—subperiods made in this chapter are not intended to be very rigorous, no adjustment in ratios is considered necessary. In addition, recall that in the major analyses of the study the ratios of all earnings announcements of a particular kind were averaged within a firm for the whole study period, before being averaged over firms. Thus the major analyses of the study already conform to the above admOnition that the period of aggregation be the same as the period of computation of "base" measurements. Because refinement of the measurements will not be attempted, the next step in evaluating the time series of profiles depicted in Figure 11 is slightly more complicated than it otherwise would be. To assess the effects of different quarters on the reported findings of the study, one must visualize each calendar quarter's contribution to the overall average week zero ratio, relative to its contribution to the overall ratio for each of the "other weeks" in the eleven week profile. In other words, it is the contrast within a quarter's profile that determines its contribu- tion to the overall contrast found in the study between residual price changes in weeks of earnings announcements and residual price changes in other weeks. On the basis of within-profile contrast, the years of the study period are remarkably consistent. The first and second quarter profiles 126 Show consistently dramatic contrast and the third quarters show a consistent lack of contrast. Only the fourth quarters are mixed——but with two out of three years' fourth quarters having contrast comparable to first and second quarters. The apparent within—year seasonality of relative price response will be examined further in the next section of this chapter. The im- plication of the consistency of within-quarter contrast between years is very simply that there is some evidence that the relationship between price changes in weeks of earnings announcements and price changes in other weeks is somewhat stable over time. But since the evidence has been examined in retrospect,a pps£_nee conclusion is not warranted. The experience of this study can only properly support an hypothesis that the relative magnitude of price-change response to earnings announce- ments is stable over time (and perhaps seasonal within years). The hypothesis must be tested with respect to sample data other than the data that suggested it, before it can properly be accepted. In the meantime, the researcher is able to draw conclusions from and make recommendations on the basis of the findings of the study with more confidence, knowing that the substance of the findings does np£_a11 stem from a pattern of extreme investor response to earnings in only one year of the study period with virtually no response in the other years. Evaluation of Differences Between Subgroups of Sample Firms For reasons that will be discussed below, this section is concerned with sample firms that differ with respect to dividend payout and firms that differ with respect to timing of their fiscal years. A breakdown 127 of the total sample of 105 firms on the basis of payout policy and fiscal years appears in Table 10 below: TABLE 10.--Distribution of sample firms by payout policy and fiscal year Dividend Non- Paying Paying Total Calendar-Year Firms 22 34 56 Non-Calendar—Year 18_ 31 49 Total 40 65 105 For the most part, Since the subsample sizes are rather small when firms are classified on the basis of both characteristics, the evaluation of this section will deal with one characteristic at a time. There are essentially two reasons why it is felt that a dis- cussion of the influence of cash-dividend-paying firms on the sample findings of the study is appropriate at this point. First, the sample of firms included in the study is biased against cash-dividend-paying firms. The reader will recall from the discussion in Chapter V that firms that frequently announced cash dividends in the same weeks as earnings announce- ments were eliminated from the sample. This, of course, reduced the sample proportion of cash dividends firms relative to the level that would have been experienced had the restriction not been imposed. Assuming that the cash-dividends—paying firms eliminated and those included in the sample are similar in all respects other than timing of dividend announcements, 128 the effect on the sample measurements of reducing the proportion of dividend-paying firms can be appraised by comparing the separate sample measurement results of the dividend paying firms included in the sample with those of all other firms included. A second reason for comparing dividend-paying firms with all other firms is to assess the effects on investor response to earnings of the availability of information contained in dividend policies (in particular, policy changes). It is especially important that an attempt be made to evaluate any differential effects that dividend information might have on measured investor response to quarterly and annual earnings. The reason for directing interest to a comparison between calendar- year and non-calendar-year firms is the potential effects on the sample measurements of the apparent seasonality of contrast between price changes in weeks of earnings announcements and price changes in other weeks. The reader will recall from the preceding section that the within—profile contrast between zero—weeks and other weeks was fairly consistently dramatic for the first, second and fourth calendar quarters of the cross— Section years. However the profiles of the third calendar quarters (summers) of all three cross-section years were consistently lacking in contrast. A time pattern of this sort would not affect the findings of the study if each type of earnings announcement had a more—or-less equal like— lihood of occurring in any calendar quarter--but of course they do not. Of the 105 sample firms used in the study, 56 operated on a calendar year basis. Whereas the other 49 sample firms as a group had earnings announce- ments of all kinds occur in the third calendar quarters, the calendar—year 129 firms generally reported only second fiscal quarter's earnings in the third calendar quarter. Therefore, measured investor response to calendar—year companies' second quarter announcements might be consid— erably less pronounced than measured response to their other quarterly announcements or their annual announcements, for whatever reason is responsible for the apparent Slump in investor response to earnings announced in the third calendar quarter. Since the measured response to all quarterly announcements combined would tend to be pulled down relative to the measured response to annual announcements, it would be useful to examine any possible tendency of the inclusion of calendar-year firms' second quarter reports to bias the findings of the study. For purposes of comparing firms on the basis of payout and fiscal year characteristics the now familiar eleven week profiles will be used. Figures 12, 13, 14 and 15 give eleven week profiles by announcement types for each subgroup of sample firms. Cash-Dividend-Paying Firms vs. Non-Paying Firms A comparison of cash-dividend-paying firms with non-cash—dividend- payers involves Figures 12 and 13. An examination of the profiles of the two subgroups reveals two things: 1. The first and second quarters' profiles and the profiles of all quarters of the dividend payers exhibit more contrast (relative to the contrast in the annual announcement profile) than the non—payers. 2. The third quarter profileci the dividend payers is strik- ingly different in pattern than any other profile in either Figure 12 or 13. 130 The first Observation might lead to a tentative conclusion that diviend information tends to damp investor response to annual earnings relative to first and second quarter earnings. However the second observation mitigates against such a conclusion. The highly anomalous third quarter profile of the dividend payers suggest that the differences between dividend-paying and non-dividend-paying firms is largely reflected in investor response to third quarter announcements. A reexamination of the pattern of dividend announcements tends to confirm this suspicion. In the preceding chapter it was noted that there was no remarkable difference in the density of dividend-related announcements in the weeks before quarterly and annual earnings announcements. This is largely attributable to dividend paying practice. Most firms in the sample that paid dividends paid them quarterly. Thus dividend announcements tended to occur quarterly. However, dividend changes (generally in form of stock splits or stock dividends) tended to be first announced more frequently in the fourth fiscal quarters of firms' Operating years than any other quarter. Now it happens that third quarter earnings announcements are made in the middle of the fourth fiscal quarter, i.e., generally toward the end of the interval from two to six weeks after the end of the third fiscal quarter. Thus it happens that the third quarter announcement is often made at the time when the market has just received or perhaps is antici- pating the most newsworthy dividend announcement of the year. One might speculate that there would be two effects of the approximate coincidence of third quarter announcements and announced changes in dividend payout. First, the importance of dividend information might damp investor interest in and response to third quarter earnings, when lst Quarter Announcements -5—4-3-2-l 0 l 2 3 4 5 week 3rd Quarter Announcements 1.30- 1.25— 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- —5-4-3-2-l O l 2 3 4 5 week All Quarterly Announcements 1.30* 1.25- 1.20- 1.15- 1.10- 1.05— 1.00 ----- .95- .90- .85- -5-4—3-2-l O 1 2 3 4 5 week 131 2nd Quarter Announcements 1.30— 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- -5—4-3-2-1 0 1 2 3 4 5 week Annual Announcements 1.30- 1.25— 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85— -5-4-3-2-1 0 1 2 3 4 5 week All Announcements 1.30“ 1.25- 1.20- 1.15- 1.10- 1.05- -5-4-3—2—1 0 1 2 3 4 5 week Figure 12.--Eleven week profiles of mean average response ratios of the 40 cash-dividend-paying sample firms. lst Quarter Announcements 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- -5-4-3—2-1 O l 2 3 4 5 week 3rd Quarter Announcements 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- -5-4-3-2-l O l 2 3 4 5 week All Quarterly Announcements 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- —5-4-3-2-1 0 l 2 3 4 5 week 132 2nd Quarter Announcements 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- -5-4-3-2-1 O l 2 3 4 5 week Annual Announcements 1.30- 1.25- 1.20— 1.15- 1.10— 1.05- 1.00— — - - .95- .90- .85- -5-4-3—2-1 0 l 2 3 4 5 week All Announcements 1.30— 1.25- 1.20- 1.15- 1.10- 1.05- .95- .90- .85- -5-4-3-2-1 0 l 2 3 4 5 week Figure l3.—-E1even week profiles of mean average response ratios of the 65 non—dividend-paying sample firms. 133 announced. Second, significant investor responses to dividend (change) announcements should be reflected in large "other week" ratios of the third quarter profile. The first expectation appears to be realized in the low week zero ratio of 1.145 of the third quarter announcements of dividend—paying firms. Indeed, the Z value from Table 11 associated with the 1.145 ratio is not significant at the .05 (1.64) level. It should be noted, however, that the investor response to third quarter announcements of individual firms is not so low as to be significantly different from response to the firms' annual announcements, i.e. the Z value in Table 12 associated with the mean difference in response ratios between third quarter and annual announcements is less than 1.64. In other words, the average measured price-change response to third quarter announcements of dividend- paying firms falls in a "grey zone" between the average magnitude of price changes in non-announcement weeks and the magnitude of changes in the weeks of annual announcements. The second expectation appears to be realized in the high aggregate values of non-zero—week ratios in the third quarter profile. However, the presence of large aggregate ratios is not enough. In order to confirm the expectation properly, the large ratios should be traceable to individual news announcements occurring in the weeks in which the large aggregates are observed. To this purpose the researcher looked for significant news events in the third week following and the second week prior to third quarter earnings announcements that had ratios greater than 2.00 in the +3 and -2 weeks of their individual profiles. It was hoped that this procedure would explain the large aggregate ratios of the weeks +3 and -2 in the 134 TABLE ll.—-Samp1e results of average response ratio measurements for various groupings of announcements of dividend-paying firms Mean Sample Average Variance of Sample RRjt Average RRjt Z 1. First Quarter Announcement 1.330 .2825 4.4692 2. Second Quarter Announcements 1.297 .4759 3.0990 3. Third Quarter Announcements 1.145 .4331 1.5860 4. Annual Announcements 1.314 .544 3.0644 5. All Quarterly Announcements 1.251 .1252 5.1061 6. All Announcements 1.272 .1218 5.6100 TABLE 12.——Samp1e results of differences between average response ratios of annual announcements and average response ratios of quarterly announcements of dividend-paying firms Sample Mean Sample Difference in Variance of Sample Average RRjt's Differences Z 1. First Quarter vs. Annual Earnings -.0158 .5917 -.1756 2. Second Quarter vs. Annual Earnings .0178 1.072 .1470 3. Third Quarter vs. Annual Earnings .1696 1.04 1.4222 4. A11 Quarterly Earnings vs. Annual Earnings .0636 .6303 .6850 135 aggregate third quarter profile. The attempt was largely unsuccessful. Few of the large individual ratios were traceable to specific news events. The conclusion was reached that an examination of the interaction between specific dividend and earnings information must await a research effort specifically designed for that purpose. In the absence of a clear understanding of the interaction between dividend and earnings information, the posture of this papm'will be to suppose for the sake of argument that, in spite of at least weak evidence to the contrary, the presence of div'rlend information tends to reduce measured investor response to annual earnings more than to quarterly earnings. On the basis of this supposition the reported finding of the study that there is no significant difference between measured investor response to quarterly and annual earnings may be attributable to the supposed influence of dividend information on what would otherwise be a Significantly greater investor response to annual earnings. In the absence of "hard" evidence that the supposition is false, the only way to eliminate ambiguity Short of an additional large-scale study, is to eliminate all effects of cash-dividend-paying firms from the sample measurements, then re-test the hypotheses of the study using the measurements for only the 65 non—dividend-paying firms in the study. The results of this procedure are shown in Tables 13 and 14. The E values in Table l3»indicate that we would judge the differences between price— change responses in the weeks of any group of earnings announcements and price changes in other weeks to be highly significant for non-dividend- paying firms, i.e., all of the E values are considerably greater than 1.64. More importantly, the E values of Table 14 indicate that we would 136 judge the difference between price-change reSponse to the quarterly and annual earnings announcements of these firms to be pg£_significant, i.e., none of the E values is greater than 1.64. Thus it has been shown that for firms without the potential damping influence of dividend information, the findings of the study are the same: that there is no statistically significant difference between price-change response to quarterly and annual earnings announcements. Although it is felt that most of the damping influence of dividend information is reflected in investor response to third quarter rather than annual earnings, this last analysis gives assurance that even if the reverse is true, the findings of the study needn't be equivocated in either case. TABLE l3.—-Samp1e results of average response ratio measurements for various groupings of announcements of non-dividend-paying sample firms Mean Sample Average Variance of Sample RR, Average RR. Z Jt Jt 1. First Quarter Announcement 1.226 .5152 2.9176 2. Second Quarter Announcements 1.154 .3614 2.3738 3. Third Quarter Announcements 1.216 .3519 3.348 4. Annual Announcements 1.299 .5472 3.7455 5. All Quarterly Announcements 1.207 .1846 4.4644 6. All Announcements 1.240 .1475 5.7906 137 TABLE 14.--Samp1e results of differences between average response ratios of annual announcements and average response ratios of quarterly announcements of non-dividend-paying firms Sample Mean Sample Difference in Variance of Sample Average RRjt's Differences Z 1. First Quarter vs. Annual Earnings .0727 .9494 .7798 2. Second Quarter vs. Annual Earnings .1447 .8761 1.6158 3. Third Quarter vs. Annual Earnings .0520 1.069 .5216 4. All Quarterly Earnings vs. Annual Earnings .0921 .7065 1.1453 Calendar-Year Firms vs. Non-Calendar-Year Firms Whereas in the face of the uncertain influence of cash dividend information on measured response to earnings announcements the concern of the analysis was that the sample measurements might tend to understate the true significance of annual earnings announcements relative to quarterly earnings, the opposite possibility is the concern with calendar-year firms. It was noted in the time series analysis that price responses to earnings numbers announced in the July-September quarter were consistently not dissimilar to price changes in surrounding, non-announcement weeks. Since calendar-year firms report only quarterly earnings (generally second quarter) in the summer quarter, their quarterly earnings would tend to be less significant, certeris paribus, than their annual earnings. And since calendar-year firms make up Inore than half (56 out of 105) of the total sample their behavior will tend 138 to influence the overall sample results. To evaluate this possibility, Figures 14 and 15 were prepared to show the response ratio profiles separately for calendar-year firms and non-calendar—year firms, respectively. It is immediately noticeable in Figure 14 that the measured investor response to second quarter announcements for calendar—year firms is quite low, as expected. In fact the week zero ratio of the second quarter profile contrasts sharply with the week zero ratio of the annual profile. The in— fluence of the second quarter profile is also evident in the unusually large difference between the week zero ratios of the all quarter profile and the annual profile. The low level of the second quarter is also brought out by comparison with the second quarter profile of non-calendar—firms appearing in Figure 15. To assess the effects of the observed "depression" in measured response to the second quarter announcements of calendar-year firms the sample measurements were all recomputed for the calendar-year firms only. The sample results appear in Table 15 and 16. Not surprisingly, the 2 values of Table 15 indicate that the week zero ratios of all announcement groups are highly significant with the ex- ception of the ratio of the second quarter announcements. The Z values of Table 16 suggests the type of effect that the measured investor response to the second quarter earnings of calendar-year firms might have on the findings reported for the whole sample. The large Z values (greater than 1.64) of the average differences between response ratios of second quarter and annual and all quarters' and annual earnings, indicate mixed results. Had the study only been concerned with calendar—year firms the findings would have been that: while the measured investor response to first and third quarters' earnings is not significantly less than to annual earnings, response to second quarter lst Q 1.30— 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- 3rd Q 1.30— 1.25— 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- All Q 1.30- 1.25- 1.20— 1.15- 1.10- 1.05- 1.00— .95- .90- .85— uarter Announcements -5-4-3-2-l 0 1 2 3 4 5 week uarter Announcements -5-4-3-2-l O l 2 3 4 5 week uarterly Accouncements 2nd Q 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- Annua .30- .25- .20- .15- .10- .05- 1.00- .95- .90— .85- h‘h‘h‘k‘h‘h‘ uarter Announcements -5-4-3-2-1 0 1 2 3 4 5 week 1 Announcements —5-4-3—2-l 0 l 2 3 4 5 week ‘ A11 Announcements ~5—4-3—2-l 0 1 2 3 4 5 week -5—4-3—2-1 0 1 2 3 4 5 week . Figure l4.--E1even week profiles of mean average response ratios of the 56 calendar-year sample firms. lst Quarter Announcements 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- —5-4-3—2-1 0 l 2 3 4 5 week 3rd Quarter Announcements 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00 ----- .95- .90- .85- -5-4-3-2-1 o 1 2 3 a 5 week All Quarterly Announcements 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- r .95- .90- .85— -5-4-3-2-1 0 1 2 3 4 5 week 140 2nd Quarter Announcements 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- —5-4-3-2-1 0 1 2 3 4 5 week Annual Announcements 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- - .95- .90- .85- -5-4-3-2—1 0 1 2 3 4 5 week All Announcements 1.30- 1.25- 1.20- 1.15- 1.10- 1.05- 1.00- .95- .90- .85- -5-4-3—2—l 0 1 2 3 4 5 week Figure 15.--E1even week profiles of mean average response ratios of 49 non-calendar-year sample firms. 141 TABLE 15.--Samp1e results of average response ratio measurements for various groupings of announcements of calendar—year firms Mean Sample Average Variance of Sample RRjt Average RRjt Z 1. First Quarter Announcement 1.225 .4368 2.9014 2. Second Quarter Announcements 1.106 .3512 1.5244 3. Third Quarter Announcements 1.222 .3186 3.3219 4. Annual Announcements 1.344 .5624 3.9093 5. A11 Quarterly Announcements 1.196 .1444 4.3958 6. A11 Announcements 1.245 .1492 5.4057 TABLE l6.--Samp1e results of differences between average response ratios of annual announcements and average response ratios of quarterly announcements of calendar-year firms Sample Mean Sample Difference in Variance of Sample Average RRjt's Differences Z 1. First Quarter vs. Annual Earnings .1188 .729 1.3353 2. Second Quarter vs. Annual Earnings .2372 .7312 2.6622 3. Third Quarter vs. Annual Earnings .0867 1.077 .7946 4. A11 Quarterly Earnings vs. Annual Earnings .1474 .5990 1.8278 142 TABLE 17.--Sample results of average response ratio measurements for various groupings of announcements of non-calendar-year firms Mean Sample- Average Variance of Sample ' RRjt Average RRjt 8 Z 1. First Quarter Announcement 1.313 .4174 3.8698 2. Second Quarter Announcements 1.325 .4506 3.8673 3. Third Quarter Announcements 1.151 .4552 1.7877 4. Annual Announcements 1.260 .5236 2.8701 5. A11 Quarterly Announcements 1.255 .1815 4.7810 6. All Announcements 1.260 .1251 5.8717 earnings was significantly less. As one would expect from these results and observation of Figure 15, a somewhat opposite effect is experienced when the sample measurements are recomputed for non-calendar-year firms only. The Z values in Table 17 indicate that the measured responses to all groups of announcements of non-calendar-year firms are significant, all.but the response to third quarter announcements being highly significant. Furthermore, the mean differences in average response to annual earnings annd quarterly earnings (Table 18) indicate that, not only are measured Inesponses to first and second quarter announcements not significantly less ttuan to annual announcements, they are slightly greater. The Z value for the 143 third quarter announcements indicates that although response to annual announcements is greater than to third quarter announcements it is not significantly greater. Both the mean difference between average response to all quarterly announcements and to annual announcements, and its related Z value, indicate that there is virtually no difference between measured investor response to all quarterly announcements and to annual earnings announcements of non—calendar-year firms. Thus, as with non-cash-dividend firms, no equivocation of the general findings of the study appears necessary in applying them to non-calendar-year firms. As was true of the effects of dividends information on investor response to the earnings of dividend paying firms, the reason for the apparently seasonal response to calendar-year firms' second quarter earnings is a research challenge that falls outside the sc0pe of this study. It would TABLE 18.--Sample results of differences between average response ratios of annual announcements and average response ratios of quarterly announcements of non-calendar-year firms Sample Mean Sample Difference in Variance of Sample Average RRjt's Differences Z 1. First Quarter vs. Annual Earnings -.0522 .8994 -.5094 2. Second Quarter vs. Annual Earnings -.0645 1.160 -.5543 3. Third Quarter vs. Annual Earnings .1090 1.044 .9874 4. A11 Quarterly Earnings vs. Annual Earnings .0056 .7572 .0596 144 be particularly interesting, however, to determine if the apparent seasonality is actually due to some lack of tuna information in the earnings announced in the July-September quarter or whether some seasonal condition results in little investor interest in what would otherwise be newsworthy earnings data. Summary of Evaluation of Subgroups of Sample Firms One of the two general findings of the study reported in Chapter V is that the measured investor response to annual announcements was slightly greater than measured response to all quarterly announcements combined, but the difference is not significant. This chapter has largely been devoted to questioning the degree to which this finding is applicable to subgroups of sample firms. In particular it was found that the finding applies, but not unambiguously, to the third quarter earnings announcements of dividend-paying firms, and that it does not apply to the second quarter earnings announcements of calendar-year firms. Otherwise, however, it was determined that the general finding applies. Indeed, the measured response to the announcements of certain quarters of subgroups of firms was greater than the measured response to those firms' annual announce- ments (although not significantly greater). The evaluation was conducted for only one characteristic at a time because a two-way classification would have divided the total sample into subsample groups too small (the smallest would be only 18 firms) to be relied upon to overcome the amount of random variability inherent in price- change measurements. Thus the non-dividend-paying firms included both calendar—year and non-calendar-year firms, the non—calendar-year firms 145 included both dividend-paying firms and non-dividend—paying firms, etc. This limits the evaluations of this chapter, but only in the way that any study is limited to capabilities for which it is designed and for which adequate amounts of data have been gathered. Given the scope and purpose of this study the conclusions and recommendations will not be greatly affected by the fact that investor response to annual earnings numbers was significantly greater than inves- tor response to one quarters' earnings of a certain group of firms. It is more important that the evidence produced by the study did not indicate that quarterly earnings generally have significantly less average influence on investors than annual earnings. CHAPTER VIII SUMMARY OF FINDINGS, CONCLUSIONS, RECOMMENDATIONS, AND SUGGESTIONS FOR FURTHER RESEARCH This study has been an inquiry into the degree to which quarterly accounting data influence actual investor decisions as reflected in market price changes. The data on which the findings of the study are based pertain to 105 firms whose common stocks were listed on the American Stock Exchange during the full study period of three fiscal years between July 1964 and June 1968. All of the 105 firms included in the sample reported quarterly earnings throughout their study periods and, at most, only occasionally announced a divi- dend payment in the same week that earnings were announced. It was estimated that the 105 sample firms are strictly representative of from 300 to 400 total firms that conform to the above specifications. The findings and conclusions of the study that will be given below are intended to apply strictly only to the population represented by the sample. Because of the limited population and sample chosen for study, the recommendation of most immediate importance that will be made in this paper is not one of the suggestions of appropriate action for accountants that will appear after the conclusions section of this chapter. Rather, it is considered most important that this study be replicated and extended to more time periods and types of firms. Since the methodology of the 146 147 study withstood the ex pg§£_evaluation described in Chapter VI, extension and replication would largely consist of extensive data gathering and preparation, primarily clerical efforts. Extension and replication will ensure that before undertaking the expense of implementing the recommen- dations that will be made in the following pages, accountants and managers will be certain that the findings and conclusions of the study are not simply the result of an extreme and unusual sample and that they apply more widely than to the somewhat limited population from which the sample was drawn. Furthermore, the implications of the effects of dividend information and the observed seasonality of investor response to earnings described in Chapter VII deserve more rigorous examination than has been possible in this study. Summary of the Findings of the Study The inquiry into the influence of quarterly earnings data on investors consisted of two basic comparisons. The first was a comparison of the magnitude of price-change responses of the market in the weeks in which quarterly earnings were announced to the average magnitude of market price changes for the sample firms in all other weeks of the study period. The second was a comparison of the relative magnitude of price-change responses in weeks of various kinds of quarterly announcements with the relative magnitude of price-change responses in weeks of annual announce- ments. The results of the comparisons are as follows: 1. In general the magnitude of price-change responses in weeks of all types of earnings announcements was greater than the average price change for non-announcement weeks. With the exception of a particular quarter's announcements 148 of each of two subgroups of firms, the magnitude of price— change responses in the weeks of announcements was significantly greater than the average for other weeks. In the cases of the two exceptions, namely the second quarter announcements of calendar-year firms and the third quarter announcements of dividend-paying firms, the lack of significance appears to be a result of the special environmental or contextual conditions in which these par- ticular quarterly announcements are made, rather than evidence of any general lack of influence of quarterly announcements on investor decisions. 2. Generally, the relative price-change response to quarterly earnings was less than response to annual earnings, but, with one exception, not significant1y_less. The single exception was the second quarter announcements of calendar- year firms. As above, this exceptional result is attributed to an apparently seasonal slump in which second quarter earnings of calendar-year firms are announced rather than any generally lesser degree of influence on investors of quarterly earnings. Indeed this position is somewhat secured by the observation that the investor response to certain quarterly earnings announcements of some subgroups of firms were actually greater (although not significantly greater) . 1 than the response to the annual announcements of the same firms. 1Referring to Table 16 the reader will note that the measured response to first and second quarter earnings of non-calendar-year firms is greater than response to annual earnings. 149 Conclusions and Recommendations A conclusion that there is significant demand for quarterly accounting data to be used by investors in actual decisions seems to be justified by the first finding of the study, i.e. that price changes in the weeks of quarterly earnings announcements are significantly greater than average price changes. But the second finding of the study, that relative price-change responses to quarterly earnings are not significantly less than responses to annual earnings, leads to the conclusion that investors may be unaware of, or unable to take account of the difference in quality (reliability) of quarterly and annual accounting data. The implications of these two conclusions for accoun- tants seem clear: 1. Any significant improvement in the quality of quarterly data themselves will lead to significant social benefits since the data will then provide an improved basis for actual investment decisions. 2. Any effort on the part of accountants that succeeds in unambiguously conveying to investors the lesser reliability of quarterly data, will contribute to the prevention of potentially significant market inefficiencies, i.e. under or overvaluation of securities in the period between market adjustments to quarterly earnings numbers and subsequent adjustments to the superceding, more reliable annual earnings numbers. In view of the recommendations made by past writers, outlined in Chapter I of this paper, there appears to be no difficulty finding starting 150 points for improvements in quarterly accounting practice. What appears to be most needed is additional research concerned with problems of implementation and questions of specific benefits to investors of the more sophisticated techniques recommended.1 For certain relatively un- sophisticated recommendations, though, accountants and managers can move immediately toward improvement of quarterly data. In particular, improvements that would stem from straight—forward application, on a quarterly basis, of efforts comparable to those now being applied only once per year, do not present any great barriers to feasibility. That they are costly seems to be the only barrier to their implementation. Although the findings of this study do not indicate that such additional costs would be worthwhile in any absolute sense, they do clearly indicate that the sometimes great differential in effort (cost) expended on annual accounting measurements relative to quarterly measurements is not justified on the basis of the importance to investors of annual data relative to the importance of quarterly data. Thus the researcher recommends that the following steps be taken by individual firms and, where applicable, by firms and their independent auditors together: 1The researcher is aware of current research in progress that is addressed to the questions of the cost relative to benefits to be obtained for investors of different degrees of accounting effort to compensate for the effects of seasonality on revenues, costs, income, etc. The source of the awareness is "On Criteria for Judging Accounting Earnings Estimators," adapted from John W. Kennelly, "An Empirical Investigation of Interim Earnings Reports" (unpublished Ph.D. dissertation, University of Chicago, in progress). 151 l. Wherever possible shift from typically once—per—year measurements to quarterly (or more frequent) measurements. 2. Extend audit surveillance of at least limited scope to quarterly accounting information. Whereas it may appear at first glance that the researcher is suggesting a total measurement effort four times as extensive as currently expended, such is not the case at all. If measurements are conducted at more frequent intervals, present levels of reliability obtained in annual measurements might be obtained throughout the year with much less measurement effort expended each quarter than is now expended once per year. A case in point is reconciliation of perpetual inventory records to physical count data. It is not clear that all inventory items need be counted every quarter to achieve the same level of reliability as is now achieved by some firms once per year. For even if only certain segments were fully counted in each successive quarter, the perpetual records would have less opportunity to "drift" away from physical reality from one reporting date to the next. Thus each quarterly (and annual) report might be just as reliable as current annual reports with respect to inventory valuation with a full count of only, say, one-fourth of all inventory classifications each quarter, augmented perhaps by representative samples of other classifi- cations. Of course the amounts of quarterly effort suggested here have no basis in experience but rather are simply suggested for illustrative purposes. Hence, as a corollary recommendation, it must be suggested that 152 firms arrive at proper levels and types of effort to achieve quarterly levels of reliability now experienced only annually, by individual research and experimentation. It is also suggested that general guide- lines might emanate from the naturally more extensive research efforts of professional organizations. Although there appears to be no great difficulty in finding starting points for improving quarterly accounting data, it is not immediately clear how accountants might unambiguously convey to investors the relative degree of unreliability of quarterly data. Even if all possible efforts to improve quarterly data were expended, quarterly data would still be less reliable than annual data due to residual seasonal and chance-variability effects; thus the problem is of great long-run importance to accounting. A minimal starting effort would be a requirement by authorities that the qualifying language of the form: "unaudited, subject to year-end adjustments", that sometimes appears in quarterly reports to stockholders, also appear conspicuously in the more widely circulated releases in the financial press. But the researcher harbors little hope that such a measure could be anything more than a possibly face-saving gesture for accountants, permitting an "I told you so" response in the event of a large-scale criticism involving investor experience with quarterly data. The evidence indicates that the investor would still have to bear the major burden of quantifying the degree of unreliability inherent in any given quarterly datum, a burden made extreme by the wide variety of quality of quarterly data among firms. To appreciate the task that the investor now faces and would still face if the efforts of accountants -153- stopped with merely labeling quarterly data as unreliable, one need only consider the statement quoted in Chapter I from the Wheat Disclosure Study: ...The Study carefully examined a significant sample of quarterly financial reports and releases provided by the two [national securities] exchanges. It was readily apparent (and acknowledged by representatives of the ex- changes) that they varied from extremely useful to extremely poor and uninformative.l and the experience of Gale E. Newell, related as follows: This study had indicated that the reported quarterly data of certain firms...appear to fluctuate much more than expected and the pattern of such fluctuations lead [sic] one to question the reliability of the data. Other firm's [sic] reported quarterly data fluctuate very little and the discrepancy between the fourth quarter's data and the data of the other three quarters is insignificant... It does not appear that such fluctuations occur within certain industries but rather that the fluctuations occur in certain firms in many different industries. In view of the variety in the quality of quarterly data among firms, it is little wonder that even though some investors are undoubtedly aware of the general lesser quality of quarterly data, the actions of the market as a whole do not reflect that the knowledge enters into actual investment decisions. The problem facing accountants is how to ensure that if the same observation holds in the future, it is not due to a misperception on the part of investors of the degree of reliability inherent in specific quarterly data. What is needed is a technique that is capable of conveying the degree of reliability of quarterly data uniquely for every reporting firm. 1U.S. Securities and Exchange Commission, Disclosure to Investors: a Reappraisal of Federal Administrative Policies Under the 33 and 34 Acts., (New York: Commerce Clearing House, Inc., 1969), p. 39. 2Newell, pp. 156-157. -154- Such a technique has been suggested by Newell for specific application to the relative-reliability dilema that is presented to accountants by quarterly data.1 It consists of stating accounting data as intervals rather than single values. Although the technique is particularly suited to solution of the specific problem of conveying relative reliability of quarterly data to investors, it has great future potential for improving the presentation of accounting data generally, a prediction supported by the following quotation from A Statement of Basic Accounting Theory: ...Another aspect of multiple valuations involves the use of non-deterministic measures or quantum ranges with or without probabilistic measures. In view of uncertainties surrounding business activities and the measurement of their impact, the use of such non-deterministic measures is likely to become a part of an expanded accounting discipline of the future. What makes the range technique particularly suited to the problem of conveying differing degrees of reliability in quarterly and annual data of different firms, is that the size of the range can be altered according to the unique characteristics of the measurements and conditions that produced the data. Furthermore the range technique is completely flexible over time and compatible with other recommendations of the study since, when applied properly, it would convey only the degree of reliability that applies to a given set of data for a particular firm. That is, as individual firms improve their quarterly measurements relative to annual measurements, they would automatically convey the improvement of relative reliability of the quarterly data by reporting smaller ranges, ceteris paribus. This study therefore strongly recommends that ranges-of-value lIbid., p. 154. 2American Accounting Association, 1966, p. 65. -155- reporting be adopted in the long run as a solution to the problem of conveying the lesser degree of reliability of quarterly data to investors. That the range—of—values solution will prove difficult and expensive to apply properly is unquestionably true. But the difficulty that accountants will have in quantifying the relative reliability of quarterly and annual data only serves to emphasize the plight of the investor who has less knowledge of the unique conditions existing in particular firms' cases and perhaps less knowledge of accounting nuances as well, but who must now perform this function for himself. APPENDIX An Illustration of the Nature of the Effects of Degree of Aggregation on Relative Chance Variability of Aggregates. To see the potential effects of aggregation on income computa— tions for varying period lengths, assume a very simplistic situation in which every event (transaction) of the enterprise creates a small incre- ment in income that is partly a function of the systematic efforts of the enterprise and partly a function of the chance combination of conditions that existed at the time of performance of the particular event. In the long run the income increments of the individual events will average out to the amount due to the enterprise's systematic efforts but individual events will vary above and below average, depending on the chance factors. Assuming that the events take place continuously over time at a uniform rate their individual income increments can be represented as: it = s + ct where it represents the income increment from the event taking place at instant t, 8 represents the portion due to the systematic efforts of the enterprise, and c is the portion due to chance causes, which for the t sake of discussion is assumed to be a normally distributed random vari- able with a mean of zero and a variance of 02. The income increment will therefore also be a normally distributed random variable with a variance of 02 but with a mean of s. If a fiscal quarter can be thought of as containing n events then the income for the quarter will be the sum: MD 156 157 which is also a normally distributed random variable with a mean of ns and variance of n02. If a fiscal year may be thought of as containing four fiscal quarters, each containing exactly n events then the income for the fiscal year will be the sum: 4n = Z ' Ia t=1 it which, like quarterly income is a normally distributed random variable but with a mean of 4ns and a variance of 4no2 The reader may recall that one of the prOperties of the normal distribution is that approximately two—thirds of the values of normally distributed random variables will, in the long run, fall in the range of plus or minus one standard deviation of the mean.1 Further, the ratio of one standard deviation of a random variable to its mean is an index of variability that is convenient for comparison purposes because of its independence of the scale of the variable. For normally distributed ran- dom variables, it measures the limit of the size of most (two-thirds) deviations of individual observations of the variable, relative to their long-run average. Thus we have a means of comparing the relative vari- ability of the normally distributed aggregates described above. The means of the income increment for one event, for the events of a quarter, and for a year were 3, ns, and 4ns, respectively. The related standard deviations (square roots of the variances) are 0, /no, and 2/52 The indexes of variability are 0/3, Vho/ns, and 2/no/4ns. A 1The standard deviation is by definition the square root of the variance of a random variable. 158 glance at the indexes in this simple progression of aggregation from the basic event to the hypothetical year, demonstrates the fundamental relationship between aggregation and chance variability. Whereas the denominator of the index, the eXpected value of the aggregate, increases with the number of basic events included in the aggregate, the numerator increases disproportionately less i.e., with the square root of the number of events. Thus the relative variability of the aggregate decreases with the degree of aggregation. In the example the index of relative variability of the quarterly income aggregate, /no/ns, is twice as large as the index of variability of the annual income aggregate, 2/50/4ns. Although the situation constructed in the preceding paragraphs is extremely simple, the parallels to the aggregation over events (trans- actions) inherent in accounting income determination is clear. Indeed, the notion that each event can be associated with a Specific income increment (or decrement) is not completely foreign to accounting, e.g., computation of gross margin under a specific identification approach to inventory and cost of sales valuation. Although a difficult and time consuming task, the reasoning employed could be extended to a much more clearly representative model of the accounting income determination process (as practiced.) However, it is felt that the discussion above has been adequate to illustrate the nature of the difference in chance variability to be eXpected in interim accounting measurements relative to their annual counterparts. BIBLIOGRAPHY Books American Accounting Association. Accounting and Reporting Standards. Columbus, Ohio: American Accounting Association, 1957. American Accounting Association. A Statement of Basis Accounting Theory. Evanston, Illinois: American Accounting Association, 1966. Arkin, Herbert. Handbook of Sampling for Auditing and Accounting. New York: McGraw—Hill Book Company, Inc., 1963. Berlo, David K. The Process of Communication. New York: Holt, Rinehart and Winston, 1960. Freund, John E. Mathematical Statistics. Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1962. Graham, Benjamin, David L. Dodd and Sidney Cottle. SecurityAAnalysis: Principles and Technique. Fourth edition; New York: McGraw- Hill Book Company, 1962. Hendriksen, Eldon S. Accountinngheogy, Homewood, Illinois: Richard D. Irwin, Inc., 1965. Mendenhall, William. Introduction to Probability and Statistics. 2nd edition; Belmont, California: wadsworth Publishing Com- pany, Inc., 1967. Rappaport, Louis H. SEC Accounting Practice and Procedure. 2nd edition; New York: Ronald Press Co., 1963. Siegel, Sidney. Nonparametric Statistics For the Behavioral Sciences. New York: McGraw-Hill Book Company, 1956. Solomon, Ezra. The Theory of Financial Management. New York: Columbia University Press, 1963. U.S. Securities and Exchange Commission. Report of Special Study of the Securities Markets. Chapter IX, "Obligations of Issuers of Publicly Held Securities," washington: United States Government Printing Office, 1963. U.S. Securities and Exchange Commission. Disclosure to Investors: A Reappraisal of Federal Administrative Policies Under the '33 and '34 Acts. New York: Commerce Clearing House, Inc., 1969. 159 160 Articles American Institute of Accountants. "Discussion of PrOper Basis for Quarterly Reports," Bulletin of the American Institute of Accountants, No. 144 (April 16, 1936), p. 14. Ball, Ray and Phillip Brown. "An Empirical Evaluation of Accounting Income Numbers," Journal of Accounting Research, Vol. VI, No. 2 (Autumn, 1968), pp. 159—178. Beaver, William H. "The Information Content of Annual Earnings Announcements," Empirical Research in Accountipg: Selected Studies, 1968, a Supplement to Volume VI of the Journal of Accounting Research. Benston, George J. "Published Corporate Accounting Data and Stock Prices," Empirical Research in Accounting: Selected Studies, 1967, a Supplement to Volume V of the Journal of Accounting Research. Blough, Carman G. (editor). "Some of the Dangers Inherent in Quarterly Financial Statements," Journal of Accountancy, Vol. XCV, No. 2 (February, 1953), pp. 221-222. Brown, Phillip and John W. Kennelly. "The Information Content of Quarterly Earnings: A Clarification and an Extention," forth- coming in the Journal of Business. Brown, Phillip and Victor Niederhoffer. "The Predictive Content of Quarterly Earnings," Journal of Business, Vol. XLI, No. 4 (October, 1968), pp. 488-497. Canadian Chartered Accountant. "Value of Interim Statements," LXXVII, No. 5 (November, 1960), pp. 379-380. Chambers, R. J. "PrOSpective Adventures in Accounting Ideas," The Accounting Review, Vol. XLII, No. 2 (April, 1967), pp. 241—253. Cohen, Manuel F. "The SEC and Accountants: Co-Operative Efforts to Improve Financial Reporting," Journal of Accountancy, Vol. CXXII, No. 6 (December, 1966), pp. 56-60. Cootner, Paul H. "Stock Prices: Random vs. Systematic Changes," Industrial Management Review, Vol. 111, No. 2 (Spring, 1962), pp. 24-45. Fama, Eugene F. "The Behavior of Stock-Market Prices," Journal of Business, Vol. XXXVIII (January, 1965), pp. 34-105. "Random walks in Stock Market Prices," Financial Analysts Journal, Vol. XXI, No. 5 (September—October, 1965), pp. 55-59. 161 Fama, Eugene F., Lawrence Fisher, Michael C. Jensen and Richard Roll. "The Adjustment of Stock Prices to New Information," International Economic Review, Vol. X, No. 1 (February, 1969), pp. 1—21. Green, David Jr. "Towards a Theory of Interim Reports," Journal of Accounting Research, Vol. II, No. 1 (Spring, 1964), pp. 35—49. Green, David Jr. and Joel Segall. "The Predictive Power of First Quarter Earnings Reports," Journal of Business, Vol. XL, No. 1 (January, 1967), pp. 44-45. . "The Predictive Power of First Quarter Earnings Reports: A Replication," Empirical Research in Accounting: Selected Studies, 1966, a Supplement to Volume IV of the Journal of Accounting Research. ~ King, Benjamin F. "Market and Industry Factors in Stock Price Behavior," Security Prices: A Supplement! Journal of Business, Vol. XXXIX, No. 1, Part 2 (January, 1966), pp. 139-190. Sanders, T. H. "Reports to Stockholders," The Accounting Review, IX, No. 3 (September, 1934). Schoomer, B. Alva Jr. "The American Stock Exchange Index System," Financial Analysts Journal, Vol. 23, No. 3 (May — June, 1967). Seidler, Lee J. and William Benjes. "The Credibility Gap in Interim Financial Statements," Financial Analysts Journal, Vol. XXIII, No. 5 (September - October, 1967), pp. 109-115. Sharpe, William F. "A Simplified Model for Portfolio Analysis," Management Science, Vol. IX, No. 2 (January, 1963), pp. 277—293. Shillinglaw, Gordon. "Concepts Underlying Interim Financial Statements," The Accounting Review, Vol. XXXVI, No. 2 (April, 1961), pp. 222- 231. Staubus, George J. "Earnings Periods for Common Share Analysis," Journal of Business, XLI (October, 1968), pp. 472-476. Taylor, Robert G. "A Look at Published Interim Reports," The Accounting Review, XL, No. 1 (January, 1965), pp. 89-96. "The Published Interim Report and the CPA," Journal of Accountancy, Vol. CXX (September, 1965), pp. 55—58. 162 Other American Stock Exchange. "Listing Form L." (Revised September 15, 1966). ISL Daily Stock Price Index: American Stock Exchange. Palo Alto, California: Investment Statistics Laboratory, Inc., published quarterly. Newell, Gale E. "Published Quarterly Financial Data: Their Adequacy for Investment Decision Making," Unpublished doctoral disserta— tion, Michigan State University, East Lansing, Michigan, 1968. U.S. Securities and Exchange Commission. "Summary of Disclosure Policy Study Report entitled: Disclosure to Investors -— A Reappraisal of Administrative Policies Under the '33 and '34 Acts." Avail- able on request from the Securities and Exchange Commission. The Wall Street Journal Index, New York: Dow Jones & Company, Inc., Issued monthly with annual cumulations. R155 1 M'TITI'ITILRHJMI]!31101151111113]!