.WI 1 322?. 41-4- .421. . (V n L ‘.I .y, .7‘ 21: twat... r 49:1 r' - Thesirtbrrmégmzm :ofEBA ~ ;, i .f MlCHIGANSTATE umvaasm ., _, , . A ‘Dgnglflbfiefipgafi , - .w ,, "1,4“. "V. ." d “mg” r..,-. '0 " "y 4'" ( . .4 M. 54 "’»'”:§7.:.'.; 33-"; "‘ ' . ’ vn‘Wf" p :2 v.’ ,H. . r ,. . A M ‘ , v- »n. 5, J a: '3 ¢ " . .- 4- .'.u1-'v u' l , . A v . , . . _ 1 ‘ n no.- “ ‘v cap- 5. . ,. I,“ s .‘ , "v1.0“ «.3. ch‘it .. 1 g- ‘ 1 “I 1'“- . ‘ ‘ . 1. . . masts 0-169 llflflllllflllfllfll“WNWWHIHHIHHINHM 3 28 1335 This is to certify that the O! thesis entitled THE FORECASTII‘I} PROPERTIES OF ILBIDERS ' 'IRADBACTIOI‘B presented by Donald Lee Rogoff has been accepted towards fulfillment of the requirements for LIBRARY Michigan State University DBA degree in W Administration Date @é/717‘ Maj ofessor It 49 .» s d it. “git... $541”: .fii 5' ABSTRACT THE FORECASTING PROPERTIES OF INSIDERS' TRANSACTIONS by Donald L. Rogoff This study is a test of the hypothesis that corpor- ate insiders, through their purchases or sales of their own company's common stock, forecast its market price. If the hypothesis is valid, the insiders' transactions should tend to lead general public trading. Thus, given the psychologi- cal effect of market trends on expectations and spending patterns, its validity would have an economic significance. It was also the purpose of this study (1) to present infor- mation on the insiders' trading patterns and the price per- formance in each stock; and (2) to compare the performance of the insider-bought and insider—sold stocks. There were insiders' transactions in 98 out of the 100 corporations drawn at random from the 1,065 common stocks listed on the New York Stock Exchange during 1957- 1960. This sample yielded 1,507 monthly observations divi- ded about equally between buying and selling. The insiders- monthly-activity index (hereafter, I.M.A. index) was the num- ber of buyers minus the number of sellers in the stock, and the index of stock performance was the percentage of change in the SEC-BOO-stock-price index six months from the date of Donald L. Rogoff the net buying or selling. Thus, the hypothesis tested whether inside-buying led a six-month advance greater than the general market movement and whether inside-selling led a six-month decline relative to the market. The results of various statistical tests suggest that the hypothesis is valid, but that the correlation be- tween insider transactions and the later market tendency of the stock is not inevitable. In other words, forecasting properties were erratic, and only certain kinds of transac- tions seemed to be dependable as predictors. Generally, insiders timed their purchases more accurately than they did their sales. A company's stock bought by two or more of its executives was more likely to perform better than the mar- ket, and also to outperform stock whose insiders had sold it. While more stocks sold by insiders declined than ad- vanced, the ratio of successful predictions among them to the total number of forecasts was only slightly better than coin—tossing. Despite strong evidence supporting the hypothesis that insiders forecast through their transactions the market price of their own company's common stock, the variation in performance is nevertheless large and the average perfor- mance resulted in such small financial gain that a policy of frequent turnover, with the associated trading costs, is Donald L. Rogoff hardly warranted. The risk of making an incorrect predic- tion is probably never much less than one-third, even when the I.M.A. index has large values--i.e., differs consider- ably from zero. Although over-all correlations were not strong, there is a good indication that insider stock- transactions in companies with permissive policies toward the trading activities of their executives have positive predictive value. Some executive groups were extremely accurate in forecasting, and others were equally inaccurate, about two and a half very successful records for every equally unsuccessful one. It seems probable that a random selection of companies which encourage or tolerate insider buying and selling would yield a stronger relationship between transaction and eventual trend. The evidence is consistent with the hypothesis that insiders have an advantage over outsiders when it comes to trading in their own company's stock. However, the perfor- mance of the stocks is too variable to make their apparent predictive property profitable to the speculator; it might ‘well prove useful to people interested in aggregated eco- ‘nomic values, nevertheless. THE FORECASTING PROPERTIES OF INSIDERS' TRANSACTIONS by Donald Lee Rogoff A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF BUSINESS ADMINISTRATION Department of Accounting and Financial Administration 1964 ACKNOWLEDGMENTS I wish to express thanks to the members of my com- mittee--Robert W. Johnson, Chairman; Roland I. Robinson; and James H. Stapleton--for their unlimited assistance and en- couragement. I am also indebted to Michigan State University and its faculty for the phase of my graduate academic prepara- tion leading to this dissertation. The completion of this research would not have been possible without the generous assistance and cooperation of many persons, among them James Don Edwards. A National Defense Fellowship provided much needed financial assistance during this phase. My family has given me the encouragement and under- standing needed to enable me to bring this study to its con- clusion. Their consideration, forebearance, and sacrifice are invisibly written on each page. ii TABLE OF CONTENTS Page ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . ii LIST OF EXHIBITS . . . . . . . . . . . . . . . . . . vi Chapter I. INTRODUCTION . . . . . . . . . . . . . . . . l The Hypothesis and Its Significance . . 1 Organization of the Study . . . . . . . . 8 II. PRIOR STATISTICAL STUDIES . . . . . . . . . . 10 Aggregate Dow-Jones Transactions Lead Their Average Price at Turning Points . . 10 Selected Portfolio Outperforms the Market 0 O O O O O O O O O I O O O O O I 13 III. INSIDER TRADING--HISTORICAL AND LEGAL BACKGROUND . . . . . . . . . . . . . . . . 19 Insider Trading Prior to 1933 . . . . . . . 19 Section 16 of the Securities Exchange Act of 1934 . . . . . . . . . . . . . . . .21 General Limitations in the Data . . . . . . 25 Ethics and Policies: an Addendum . . . . . 28 IV. THE HYPOTHESIS AND A METHOD OF INVESTI- GATION . . . . . . . . . . . . . . . . . . 33 The Hypothesis . . . . . . . . . . . . . . 33 The Model . . . . . . . . . . . . . . . . . 39 Rules for Collecting and Summarizing the Data . . . . . . . . . . . . . . . . 52 Summary . . . . . . . . . . . . . . . . . . 61 iii Chapter VI. VII. ANALYSIS OF STATISTICAL RESULTS Forecasting the Magnitude of Change Forecasting the Direction of Change Summary . . . . . . . . THE RESULTS OF COMPANY TOLERATION OF INSIDER TRADING FOR PROFIT Method Used in the Study Results . . . . . . . . . . . . . . Sample of Companies with Three or More Net Buyers and Sellers Within Any One Month . . . . . . . . CONCLUSIONS AND IMPLICATIONS Summary of Statistical Tests Implications APPENDIXES . . . . . . . . . . . Appendix A: Frequency and Number of Net Buyers and Sellers in the Monthly Insiders' Transactions Indices, by Company, During 1957-1960 . Appendix B: Dates of Monthly Observations of Stock Prices Appendix C: Insiders' Transactions: Rules for Inclusion Appendix D: SEC Rule 16a-9 (Exemption for Small Transactions) Appendix E: Illustration of Data Summarization: Aldens, Inc. Appendix F: Sign Test on Forty-eight Monthly Correlation Coefficients . iv Page 64 65 92 120 128 129 130 131 136 139 141 159 159 165 167 171 174 181 Appendix G: Appendix H: Appendix I: Appendix J: Appendix K: BIBLIOGRAPHY Page Ranking of 83 Companies by Correla- tion Coefficients Between Their Insiders' Transactions and the Rela- tive Stock Price Lagged Six Months Based on Six or More Monthly Obser- vations During 1957-1960 . . . . . . . 183 Significance Test of Difference Between Means . . . . . . . . . . . . 189 Average Percentage Performance of Insiders' Net Buying and Selling by Company . . . . . . . . . . . . . . 191 A Ranking of the Insiders' Forecast- ing Record in Predicting the Direc- tion of Their Company's Stock Price Relative to the Market as a Whole Six Months Later . . . . . . . . . . . 201 Frequency of Successful Forecasts of Direction in Two Models, One Deflated and One Not Deflated . . . . . . . . . 210 . . . . . . . . . . . . . . . . . . . 213 Exhibit 2- 1 5- 1 5- 2 5- 3 5- 4 5- 5 5- 6 5- 7 LIST OF EXHIBITS Average Performance of Selected Stocks in 1958 Relative to the Performance of the Moody's 125 Stock Average . Monthly Correlation Results Between Performance and Corporate Insiders' Transactions Indices During 1957-1960, Ranked by Size . . . . . . . . . Frequency Distribution of 48 Monthly Correlations Between Insiders' Transac- tions and Relative Stock Price Perfor- mance Lagged Six Months . . . . . . Six Largest and Six Smallest Monthly Correlations . . . . . . . . . . . . . Three Correlation Coefficients Per Month Between Insiders' Transactions and the Relative Stock Price Lagged Six Months and a Background of the General Market Movement . . . . . . . . . . . . . . Frequency Distribution of Companies by the Coefficient of Correlation Between the Insiders' Transactions and the Lagged Stock Price . . . . . . . Relative Performance of Stocks Purchased and Sold by the Insiders-Monthly-Activity Index 0 O O O O O O O I I O O O O O A Summary of the Forecasting Properties of the Insiders' Transactions in 98 Common Stocks Selected Randomly During 1957-1960 vi Page 15 67 69 71 72 78 81 87 Exhibit 5-13 5-16 Companies Ranked by Average Performance (Net Buying and Net Selling) Over the Six Months . . . . . . . . Aggregate Results of Direction-Performance in Corporate Insiders' Transactions . Distribution of Percentage of Success Among the Companies . . . . . . A Frequency Distribution of the Number of Companies by Their Insiders' Forecasting Record . . . . . . . . . . . . . . . . List of Companies Whose Insiders' Transac- tions Were very Successful Forecasting Direction and a Description of Their Trading Pattern . . . . . . . . List of Companies Whose Insiders' Transac- tions Were very Unsuccessful Forecasting Direction and a Description of Their Trading Pattern . . . . . . . . . . . The Highly Successful Series . . . . . . . The Significantly Unsuccessful Series (Based on Six or More Observations) Insiders Forecast Through Their Transac- tions the Price in the Market of Their Own Company's Common Stock: A Summary of Selected Results . . . . . . . . . The Seven Executive Groups of the 98 Tested with Three or More Net Buyers and Sellers in at Least Two Months . Performance of Forecasts at Monthly Peaks of Buying and Selling in Stocks with Three or More Net Buyers and Net Sellers Over Six Months . . . . . . . . . vii Page 89 95 103 107 116 117 122 123 126 132 134 CHAPTER I INTRODUCTION The Hypothesis and its Significance The literature contains studies of the ability of various groups to forecast stock prices, ranging from in- vestment advisors and fire insurance companies to mutual funds. The evidence of successful forecasting by any of them, however, is very scanty. In fact, the evidence shows the reverse--that the "experts" do not in general success- 1 AS far as can be determined, fully forecast stock prices. there are no scholarly studies concerning the abilities of key corporate officials as expert forecasters. This study is such an investigation, and proposes to 1Statistical tests of the best individual records of the "experts" fail to demonstrate that they exhibited skill in investment. The average performance is unsuccessful com- pared either to the average random investment or to the mar- ket performance in general. And the least successful ex- perts perform less well than can be reasonably attributed to chance. In essence, these were the results of the Cowles Commission for Research in Economics over three decades ago. See Alfred Cowles, 3rd., "Can Stock Market Forecasters Fore- cast?" Ec nometrica, I (1933), 309-324. More recently, the same general conclusion was suggested in A Study of Mutual Funds, A Report Prepared for the Securities and Exchange Commission by the Wharton School of Finance and Commerce (washington: Government Printing Office, 1962), p. 17. l 2 test the usefulness of reported stock transactions made by corporate officials--their purchase or sale of their own company's common stock--as a predictor of that stock's value in the time following the transaction. In other words, the hypothesis assumes that insiders forecast through their transactions the market-price of their own company's common stock. The corporate insider should, presumably, know more about his own company than the general public does. If this is the case, and the executive feels free to invest accord- ingly, following profit-maximizing goals, his operations should predict the stock's future value more accurately than general public trading does, and should tend to lead the general public trading. A corollary of the hypothesis is that such insiders' stock transactions have an economic significance, at least in part, because of the importance of stock prices in our economy. Keynes, for example, observed that the level of prices in an organized stock market is influenced by specu- lative considerations having little or nothing to do with the real business outlook. He felt, however, that the level of the market does influence businessmen's expectations and thus their investment or capital-spending decisions; hence the economy is affected. More recently, Granger and Morgen- stern suggested the psychological importance of the stock market on the economy: The stock market is an institution of considerable interest to the public at large and of real importance to students of a nation's economy. The variables which make up a stock market may not directly affect the mechanism of the economy but they certainly influence the psychological climate within which the economy works. To the extent to which the movements of the economy directly affect the stock market, a feedback situation occurs, although there are reasons to suspect that the strength of the feedback is not strong. The stock market produces large amounts of high quality data derived from well-understood variables. Despite these facts, the stock market has attracted surprising- ly little study by professional economists or statisti- c1ans. Alfred Cowles's justification of his analysis of the forecasting efforts of professional agencies has a bearing on this point, as well: It seemed a plausible assumption that if we could dem- onstrate the existence in individuals or organizations of the ability to foretell the elusive fluctuations, either of particular stocks, or of stocks in general, this might lead to the identification of economic theories or statistical practices whose soundness had been established by successful prediction.3 This investigation, like Cowles's, “is a means of testing the success of applied economics in the investment field."4 Possibly this empirical study may have significance (in 2C. W. J. Granger and Oskar Morgenstern, "Spectral Analysis of New York Market Prices," K.k1 s, XVI (March, 1963), 1. 3Cowles, op. cit., p. 309. 4gbid. 4 either method or finding) for the field of security analy- sis.5 Some students of speculative markets believe that the behavior of prices in them can, in part at least, be ex- plained by the existence of some group or entity more know- ledgeable of a stock's likely future in the market than the general participants. It would indeed be useful to find a specific group with this capacity. The importance of insiders' stock transactions can be inferred from the large number of public traders who ap- parently follow their activities as a qualitative or techni- cal indicator of a stock's value in making investment deci- sions.6 One of these reporting services bases its opera- tions entirely on the premise that insiders at any time ought to know earlier than the general public if their company's stock is under- or over-valued.7 While the ob- 5An inductive approach to studying the behavior of stock prices has significance with respect to security anal- 'ysis. See Benjamin Graham, "Towards a Science of Security .Analysis," Financial Analysts Journal, VIII (August, 1952), 96-99. 6Articles regularly appear in the financial periodi- cals. For example, see Roger Bridwell, Barrgn's National Business and Financial Weekly; or Robert B. Shaw, Magazine of wall Street, reports on insiders' stock activities pub- lished quarterly. Several investment services hint that this information is a useful investment guide. See The Valgge Line Investment Survey--Part I, "What Officers and Di- rectors are Buying and Selling," published weekly. 7Bridwell and Company, "The Insider Report--It Pays 5 jectively written and documented investigations of the pub- lic's stock trading do not make blatant claims of the signi- ficance of insiders' transactions, they sometimes report the trading balances of this group as part of the "useful back— "8 This use of the data ground of conflicting influences. suggests that the findings of this study should have not only social but economic implications as well. Finally, be- cause insider trading is an important market phenomenon, its Significance as a market factor is well worth knowing.9 Apprpach The focus in this study is on individual stock pri- ces rather than on the aggregate level of the market. It seems more reasonable to assume some superior capacity of the insiders to recognize the over- or under-valuation of their own company on the market than it is to attribute to them a superior capacity to forecast general market trends. While Corporate profits are interrelated, there is also to Invest with the Insiders," Palo Alto, California; semi- monthly investment advice. 8Securities and Exchange Commission, Stock Trading on the New York Stock Exchange on September 3. 1946, Trad— ing'and Exchange Division (Washington: Government Printing Office, 1947), p. 3. 9More concrete evidence supporting this Opinion is found in Exhibit 2-1, Chapter II. 6 evidence that they do not follow one consistent trend.lo Since earnings per share are an important factor in the price of a stock, there is no reason to expect one stock price cycle to coincide with another. In fact, stock prices of corporations do not follow one consistent trend, although the patterns they follow are not random either. In a study of the general economy it is reasonable to smooth out such price diversity. Until it is determined that there are pre- dictable or unique patterns in the transactions of insiders, the possibility that aggregation might conceal such patterns makes it less hazardous to study the transactions separately within each company. EQQEQ It is beyond the scope of this study to theorize 11 Nor will about the personal motivations of individuals. this study try to assess the impact of the insiders' trans- actions on the prices of the relevant stocks, although any correlation between the transactions and the stock prices 10Thor Hultgren, i iv iti ' t - tapes pf Ipdgatrial ggrpgratigns, Occasional Paper no. 32 (New York: National Bureau of Economic Research, Inc., 1950), p. 11. 11The behavioral assumptions will be explicitly stated in terms of the executives as an aggregated class in (Ihapter IV and subsequently reexamined in the light of the evidence. 7 revealed in the course of the study will be noted.12 The sample.--It was decided to limit observations to transactions in stocks listed on the New York Stock Exchange during 1957-1960. The common stocks of corporations listed on this exchange were given consecutive numbers 1 through 1065.13 Random tables were used and 100 common stocks were selected as the sample. The sample design was simple random . . l4 sampling Without replacement. The sampling fraction of approximately 1/10 appears to have resulted in an adequate sample. The 100 randomly 12The volume of the insiders' transactions was less than 1 per cent of the total volume in these stocks during the month and apparently had little, if any, influence on the price of these stocks. In a few instances the shares bought and sold by the executives represented between 1 and 2 per cent of the total purchases and sales: however, in the majority of cases, the size of this group's transactions represented less than 3/10 of l per cent. 13This listing fairly represents the population. Companies are continually being added or subtracted from the population over time. Mergers are the major cause of change. Over-all, however, these changes are relatively infrequent. The prepared list, as of July 1, 1957, is therefore adequate for our purpose. 14A few very large companies dominate the total mar- ket value of the stocks in the market. Therefore, a cluster sample design was considered so that General Motors, for example, would have had a greater chance of inclusion than one in 1065. If this study were concerned with predicting the level of the total market this might have been an advis- able procedure to assure a representative sample. However, adequate testing of the hypothesis does not appear to re- quire a cluster sample design. 8 drawn companies with common stock issues listed on the New York Stock Exchange are presumed to be typical of the more than 1000 listed companies, and include two of the thirty stocks in the Dow-Jones Industrial Average. The sample ap- pears to be of sufficient size to allow confidence in the stability of its characteristics. The list of companies comprising the sample is provided in Appendix A. Time span covered.--Insider transaction data were collected for the forty-eight months beginning January, 1957, and ending December, 1960. The total time span avail- able for testing is from April, 1935 (the inception date of the SEC insider report publication) to the present. Since many significant changes in the economic structure could have occurred during this twenty-eight year period (as, for example, from the effects of World War II, the Korean War, increased tax rates, or the growth of stock option plans), it was decided to test recent history as being most likely to be useful for present-day purposes. Organization of the Study Chapters II and III will present empirical and legal background information on insider trading. In Chapter IV 'the hypothesis is analyzed and a method derived for testing its validity. Chapter V summarizes the aggregate results of 9 the statistical tests and analyzes the findings. It also examines the forecasting properties of the observed insiders' transactions, company by company. Do the insiders' trans- actions in certain companies consistently forecast reliably, while those in other companies as consistently fail to fore- cast? The hypothesis of the study includes a time relation- ship factor. Are insiders' transactions good predictors of their company's stock corrected for changes in the level of the market? Do insiders' transactions in a particular com- pany's stock serve to predict market movements in those stocks? The results suggest the need for greater selectivity and Chapter VI briefly examines the risk involved in select- ed transactions. These tests will use an appropriate aver- age as the reference index for changes in the level of the stock market. The conclusions and implications of this study are presented in Chapter VII, along with a summary of the recom- mendations drawn from this study. In summary, the figures to be assembled in this study will present, apparently for the first time on a broad scale, information on the predictive value of insiders' transac- tzions and the subsequent price behavior of the market prices (of their own corporation's common stock. CHAPTER II PRIOR STATISTICAL STUDIES Is there any evidence of predictive value in the reported transactions of corporate executives? A search through the literature reveals a dearth of theoretical- statistical studies using the well defined and readily available data. Their contributions are presented below. \ Aggregate paw-gonea Transagtipns Lead Their ver e Price t Turnin Points Apparently the first statistical study investigating the usefulness of the reported insiders' transactions data relevant to the hypothesis that insiders forecast through their transactions the market price of their own company's stock was written by Robert S. Hamada.l In this study he investigated the usefulness of the turning points in an ag— gregate insiders' transactions series in an effort to fore- cast the cyclical turning points in the stock market as a 1"An Analysis of Diffusion Indices of Insiders' Erransactions," (unpublished Master's thesis, Massachusetts Institute of Technology, Cambridge, 1961). 10 11 whole. He aggregated the insiders' transactions in the thirty corporations comprising the Dow-Jones Industrial Average, and tested the forecasting correlation of that in- dex with the Dow-Jones Industrial Average during the twenty- five year period 1936-1960. Hamada chose the diffusion in- dex approach to aggregate the transactions of the individual series. Within this concept, he tested different bases for aggregating the transactions and different smoothing inter- vals. Hamada found that insiders' transactions do corre- late significantly with stock prices with a lead.2 He based his conclusions on the general results and implications of the following: regression analysis, tests for randomness, turning-point conformity and timing consistency, Okun's appraisal test for leading indicators, and an objective in- vestment decision rule. The other findings of Hamada's study follow, para- phrased for brevity: (1) There is little difference in the forecasting properties of diffusion indices and indices com- puted from the first-difference of the aggregate index; (2) the performance of the aggregate index differed consistently ‘ 2Ibid., p. iii. As a second general conclusion, JHanada says, "The diffusion index concept provides another nuathod for testing for Correlation between two economic time variablesfl' 12 from that of the other insiders' indices; it therefore can- not be called a good leading indicator; (3) computing indi- ces on different bases from the same raw data makes little difference for forecasting; (4) smoothing an original index eliminates undesired irregular fluctuations that hinder the forecasting performance; (5) a structural change in the early 1950's improved the forecasting properties of the in- siders' indices.3 This study, like Hamada's, tests the validity of the hypothesis that insiders' transactions have an inherent pre- dictive value. The basic difference between the two is the method of testing for correlation and its rationale. Hamada locked at general market movements, while this study ana- lyzes the company's stock price deflated for general market 4 Hamada did not compute the transactions series movements. for each company in his sample before completing the aggre- gation. This study computes and tests the individual cor- poration series; in other words, it looks at the pattern of 3Tbid., pp. 173-174. 4Assuming that Hamada's method is appropriate and his conclusion of a significant leading correlation valid, the question arises whether the insiders are successfully predicting their own company's common stock, independent of the stock market, or forecasting the market as a whole, 'their stock investment in their own company following the general market? This conceptual difference between the two studies is expanded in Chapter IV. 13 trading activity in each corporation and tests its forecast- ing properties. Except for the regression analyses, our tests are quite different. Samples differ in size and char- acter as a consequence of different sampling methods, and there are other differences which do not merit an exhaustive presentation.5 Selected Portfolio Outperforms the Market In 1962 William Parks, John Weiland, and I wrote a paper entitled "Common Stock Selection Design Using SEC In- sider Reports: Analysis and Evaluation." The investment rule formulated and tested involved the investment each month of an equal amount of money in any stock listed during 1958 which met the discriminating criteria defining the "universe." The hypothesis was that stocks selected from the defined universe would, on the average, outperform the general market averages. The universe of common stocks con- sisted of those whose reported insiders' transactions met the following criteria: 1. Shares were purchased by three or more insiders during the same month, except that 5For example, our indices are constructed and ‘Meighted differently. Also, Hamada included purchases of Shares on option, weighted equally with regular purchases, “daile the current study excludes this type of purchase titansaction. l4 2. Option and odd-lot purchases were excluded. Also excluded were an official's first pur- chase, defined as such when the quantity of shares purchased equalled his month-end bal- ance of shares held. 3. Shares were not sold by any insider during the month. 4. Purchase of shares increased the holdings of at least two-thirds of the officials by 10 per cent or more. Over the twelve-month period, forty-five common stocks in forty different corporations qualified for selection. The data were tested by comparing the average results of the selected portfolio with the performance of Moody's 125 Stock Average. Five holding periods were tested. Automatic selling rules were imposed at the subsequent time intervals of one, three, and six months and one and three years. The conclusions are outlined below. The evidence tended to support the hypothesis that the selected portfolio would outperform the average diversi- fied portfolio in a rising market. The six-month holding period, on the average, appeared to be optimal. Generally, the evidence suggested a short-term predictive value for the insiders' purchase transactions. Both of these unpublished studies indicate that sub- :atantial gains would have accrued to a speculator using the itasiders' transactions as a signal during the periods stud- I 15 em.» om.Hm om.mma om.sma “mama mums» m oa.m oa.m oa.ema om.ema “mama Hams H mo.ma Hm.m No.6HH mm.mmH “mama mauaos o em.oa Hm.e mm.m0H oo.mHH “mums mgueos m mm.oa Hm. ma.moa eo.eoa umnma space a oo.ooa oo.ooa mums coauoa UGNflHMDGé ANS mmOOOSm xmch mmmnm>¢ xprH .xum .mxpm mpflmcH emcammuso Mesa urlxm mma m.seooz emuomamm me mxooum a m .Hoonfi .Hou m04mm>¢ MUOBm mNH m.MDOOZ HEB m0 moz¢zm0mmmm HEB OB m>HB¢Qmm mmma ZH mMUOBm QHBUHQmm ho moz¢2m0mmmm m0¢dm>¢ HIN BHMHENM 16 ied. In both studies the rules were followed automatically without regard to other information that might have been available to the speculator. While this procedure was use- ful in the attempt to isolate the value of the insiders' transactions as an investment tool, application need not be restricted to this practice. In other words, qualitative analysis and judgment coupled with the quantitative aspects might improve the value of this tool. Tabulation pf statistics As noted in Chapter I, a search of the literature did uncover several non-theoretical statistical tabulations, although no statistical inferences had been published which relate to the hypothesis under consideration. A rasearch report by the New York Stock Exchange.-- The New York Stock Exchange undertook a detailed analysis to determine who the buyers and sellers were during the extra- ordinary upsurge of volume and the sharp price decline and recovery between May 28 and May 31, 1962.6 The activity of the executives of certain corporations was studied as well. This study provides a clue to the significance of this 6The Stock Market Under Stress: The Events pf May 28, 29, and 31, 1962 (New York: The New York Stock Ex- change, 1963). l7 segment of activity on the total market: Although there was some unusual activity in a few individual issues, over-all it is unlikely that insid- ers had any significant influence on stock prices over the three days. The shares bought and sold by this group represented only 0.6 per cent Of total purchases and sales. . . . Among all stocks listed on the Ex- change, insiders purchased 388,000 shares and sold 70,000 shares during the last three business days in May. . . . Insiders were net buyers on all three days, with the heaviest buying occurring on Monday and Tues- day, May 28 and 29.7 How are these findings relevant to the present study? First, the trough formed during May 28-31, 1962 per- iod, made it a good time to buy stocks for the short term, and the insiders were buying. This suggests that the insid- ers, in the aggregate, timed their transactions exceedingly well with respect to the market level. Second, the research department of the NYSE believed it appr0priate to devise rules of inclusion and exclusion for collecting the data. Their method of collecting data can be seen in the following description: Transactions in a particular common stock by the corporations' directors, officers . . ., and by any stockholder owning 10 per cent or more of the issue are referred to as "insider" trading. Included are transactions for accounts in which these stockholders are considered to have a beneficial interest, such as family member, trust, or personal holding company. For purposes Of this report, "insider" trading does not include acquisitions through stock Options, stock splits or stock dividends; nor disposition through 7Ibid., p. 38. 18 gifts or stock transfers. These items are excluded because they do not involve transactions on the New York Stock Exchange. While the literature does not reveal successful forecasting Of individual stock prices in any statistically significant manner, it does reveal a consensus as to the appropriate statistical technique for measuring the rela- tionship between variables in the stock market. 8;bid., p. 37. CHAPTER III INSIDER TRADING--HISTORICAL AND LEGAL BACKGROUND This chapter will briefly review the background and legislation which brought insider trading data into exis- tence. Next, some Of the limitations are presented, followed by a brief discussion of the ethical aspects of insider trading. Insider Trading Pripr to 1933 The testimony at early congressional hearings during the investigation preceding the adoption of the Securities Exchange Act of 1934 suggests that there were numerous in- stances in which corporate Officials benefited from their advance knowledge of factors significantly affecting the value Of stocks.1 During 1929 alone, there were over 100 cases in which corporate "insiders" participated in pools 1Donald C. Cook and Meyer Feldman, "Insider Trading Under the Securities and Exchange Act," Harvard Law Review, LXVI (1953), 385. The factual background of abuse which led to the enactment of the Securities Exchange Act of 1934 is summarized in Section 2 of the Act. 19 20 trading in the stock of their own companies.2 Having the benefit of advance information on significant changes in corporate financial policy, the insiders were able to reap large profits from stock-market activity at the expense of the stockholders. For example, Cook and Feldman reported: In one of several glaring instances of abuse, the chairman of an executive committee participated with a director in a pool organized to trade in the stock of their company. When the pool was formed the company was paying no dividends. Shortly thereafter, however, they caused the company to declare a dividend, and they continued a policy of irregular dividends, payable at such strategic times that more than twenty-five per cent of what was paid was received by the pool. These dividends were paid in spite of the fact that the com- pany's earnings were not sufficient to cover them and part of its surplus had to be diverted for that pur- pose. In the less flagrant violations Of trust, the executives' testimony (and that of broker-dealers) suggests that profits from "sure-thing" speculation were regarded as one of the normal "emoluments" of Office. Such practices and attitudes led to the enactment of Section 16 of the Securities Exchange Act of 1934. 2Ibid., p. 386. The one hundred stocks subjected to pool operations were listed on the New York Stock Exchange. One pool in a seven-day period bought and sold almost 1,500,000 shares in Radio Corporation of America, at a net profit to the members of the pool of almost $5 million. 3Ibid. 21 Sectign 16 of the Securities Exchange Act of 1934 Section 16 is designed to curb the misuse of corpor- ate information not available to the general public.4 The provisions of this section concern the stock market activity Of insiders in "listed companies" (insiders are officers, directors, and large stockholders in these registered com- panies). The three-pronged approach to control of insider abuses may be described as (1) public disclosure; (2) purg- ing of profits realized from short-term transactions; and (3) the prohibition of certain sales. The important section for purposes of this study is the disclosure provision, al- though all are discussed. Public diaclosura Before discussing the disclosure provision, it is useful to clarify the term "equity security" as it is stipu- lated in the Act. In Section 3(a)(ll), an "equity security" is defined as . . . any stock or similar security; or any security convertible, with or without consideration, into such a security, or carrying any warrant or right to sub- scribe tO or purchase such a security; or any such warrant or right; or any other security which the 4The WOrk of the Securities and Exchange Commission (washington: Government Printing Office, April, 1961), p. 6. 22 Commission shall deem to be of similar nature and con- sider necessary or appropriate, by such rules and regu- lations as it may prescribe in the public interest or for the protection of investors, to treat as an equity security.5 The courts' interpretation of the phrase "or similar securi- ty" is broad enough to include income bonds, voting deben- tures, and options of various sorts including puts, calls, spreads, and straddles.6 Thus, the possession of any Of these privileges should be reported by persons falling with- in any category spelled out in Section 16(a): Every person who is directly or indirectly the beneficial owner of more than ten per centum of any class of any equity security (other than exempted secur- ity) which is registered on a national securities ex- change, or who is director or an officer of the issuer of such security, shall file, at the time of the regis- tration of such security or within ten days after he becomes such beneficial owner, director, or officer, a statement with the exchange (and a duplicate original thereof with the Commission) of the amount of all equity securities of such issuer of which he is the beneficial owner, and within ten days after the close of each cal- endar month thereafter, if there had been any change in such ownership during such month, shall file with the exchange a statement indicating his ownership at the close of the calendar month and such changes in his ownership as have occurred during such calendar month.7 5Cook and Feldman, Op, cit., p. 393, "citing" Sec- tion 3(a)(ll). 6A "call" is an option to buy; a "put" is an Option to sell; a "spread" and a "straddle" each involve a combina- tion of a put and a call, thus giving the holder the choice of buying or selling on the date at the prices fixed. 7Securities and Exchange Apt Of 1934. Section 16(a). Tha Pgblig Utility fiolding Company Act of 1935, Saction 23 Subsection (a) is based upon the theory that dis- closure of insiders' transactions would put a stop to the abusive transactions illustrated above. On the other hand, if the insiders do disclose their transactions, they are en- titled like any other investor to anticipate future trends by "appropriate purchases and sales." What is meant by "appropriate purchases and sales" can be gleaned from a brief summary of other subsections. Pprging pf prpfits Subsection 16(b) provides that any profit realized from short-swing trading Shall inure to the corporation. Hence, profits Obtained by insiders from purchases and sales (or sales and purchases) of such equity securities within any six months' period may be recovered by the corporation or by any security holder on its behalf. This recovery right must be asserted in the appropriate United States District Court. This sanction is based on the theory that if the possibility of profit were removed, short-swing trading by I2(b) and the Investment gompany Act of 1940, Section 20 (f) were subsequently passed by Congress and signed by the President to regulate and reveal to the public the activi- 'ties of insiders in these two specific areas of economic activity. These have similar provisions and are not de- tailed in the text above. 24 insiders would be discontinued. Since the courts utilize a life cost basis (instead Of a fifo or average cost method) to compute the profit realized on transactions within six months, the possibility of profit on short-swing trades is lessened providing there is a plaintiff willing to assert recovery in the interests of the corporation. Prghibition of certain sales Subsections 16(c) and (d) continue the description of lawful and unlawful transactions for insiders. For exam- ple, Subsection (c) imposes an absolute prohibition upon certain sales, making it unlawful for insiders to sell their securities short or to engage in the practice of "selling against the box" (a system whereby a person sells securities and then borrows securities to meet his commitment, hoping to purchase the securities later at a lower price for deliv- ery to his creditor).8 8"Selling against the box" is a short sale Of bor- rowed stock in which the seller has a long position. While “selling against the box" may be prohibited, it can be ar- gued that this would be harmless. Selling securities short which you do not own adds to the supply of shares and may subsequently result in a realized short-term profit for the insider. On the other hand, selling securities short in which you have a long position does not add to the supply Of shares if the insider's alternative is to liquidate his long .position. More important, it can be argued that the insider (does not make a profit in selling against the box. If the Inarket falls, his capital gain in the short position is czancelled by his capital loss in his long position. 25 This brief summary of the disclosure provision in the Act and its interplay with the other provisions has been presented as an introduction to the discussion of the general limitations in the data. Seneral Limitations in the Data Much of Section 16 has been the subject of contro- versy, but there has been little opposition to the publicity requirements contained in Subsection (a).9 Some of the criticisms of this subsection are directed at its failure to require reports more Often than once a month, and its limi- tation to the equity and convertible securities of listed companies. It has been suggested that, in view of the vola- tility of the securities markets, information a month or more Old is not very useful to investors.10 The objection of limited applicability of the section is part of the broader Objection to the irrationality of restricting the Exchange Act safeguards to investors in listed corporations, especially since listing is entirely voluntary. No legisla- tion compels publicly held corporations that are not listed to offer these protections. The Commission has consistently 9Cook and Feldman, pp, cit., p. 392. lOIbig, For purposes of regulation, however, late reports are probably as useful as prompt ones. 26 protested against this discrimination against investors in over-the-counter market securities. The SEC mentions in its foicial Summary that publication of this information (on insiders' transac- tions) is in no sense a representation that the data as filed are correct, complete or genuine. Responsi- bility for accurate reports rests with the reporting individuals and not with the Commission or with any exchange.12 This raises the question of the meaningfulness of the pub- lished transactions data. Do key officials in large and widely held companies purposefully withhold or taint infor- mation on their transactions? This question is beyond the scope of the study; the hypothesis relates to the reported insiders' transactions, not the insiders' transactions by some other definition. If there are informed junior Offi- cials without executive titles who do not publicly disclose their transactions, the reported transactions are simply not completely comprehensive.13 Although the Commission asserts lllbid, The SEC Special Market Study, released in April, 1963, again urged extension of the disclosure re- quirement. lQSecurities and Exchange Commission, foicial Sum- mapy of Security Transactions and Holdings, Washington 25, D. C., published monthly. This publication is hereafter referred to as the Official Summary. 13Another limitation is the possibility that real "insiders," if not overscrupulous, can place "dummies" on the board and trade on the basis of information supplied to them. See Benjamin Graham, David L. Dodd, and Sidney 27 that responsibility for accurate reports rests with the re- porting Officials, there are penalties for false reporting and the enforcement of the disclosure subsection is the responsibility Of the Commission. One Chairman of the Com- mission has observed: It is the Commission's policy to watch the reports closely and request a clarification whenever a dis- crepancy is noted. . . . Discrepancies are . . discovered when, in accordance with the Commission proxy rules, the corporation shows in its proxy- soliciting material that the Officer or director has a different number of shares from that which appears on his latest ownership reports. Some indication as to the validity of the data can be deduced from the number of amended and late reports filed with the Commission. Corrections reported average approxi- mately one out of 700 reporting companies per month. The Commission says that tardy reports are relative- ly infrequent. The data collected in this study confirm this. The relative frequency of late reports was Observed to be approximately 4 per cent Of the total number of trans- actions. Approximately 90 per cent of the late reports ob- served in this study are tardy less than two months, and no one has ever been prosecuted by the Commission solely for Cottle, Security Analysis (New York: McGraw-Hill Book Com- pany, Inc., 4th ed., 1962), p. 681. 14Cook and Feldman, Op, cit., p. 407. 28 late reporting.15 For our purposes, therefore, it seems fair to assume that reported insiders' transactions fall within reasonable limits of accuracy. Ethics and Poligies: an Addendum One of the designs of the Securities Exchange Act is to protect the investing public by maintaining fair and open markets for the purchase and sale of securities. This is accomplished through a measure of regulatory control upon certain organizations and individuals, and upon their re- porting and trading practices. The reporting practices of listed companies or the trading practices Of specialists, brokers, and dealers, for example, overshadow the practices of insiders in importance to the extent that they affect public policy. Nevertheless, the same broad purpose of "fairness" in the Act can likewise be ascribed to Section 16. While it is not possible to judge the propriety of an action by inspecting the short- or long-term dimension within a buy-and-sell cycle, Section 16 seems to distinguish between short- and long-term trades. Section 16, in forcing 15Failure to file insider reports was the specific charge under which stock manipulator Alexander Guterma was arrested. However, the charges under which he was eventually jailed were broader. 0f . "its; 29 disclosure, in effect limits unfair trading to no oftener than twice a year. If the intent were to prevent unfair transactions, the Commission would not be denied the enforce- ment function under Subsection (b), and the recovery clause would not be limited to profits realized within a six-month period. Epfgpcamant lagking Denying the enforcement function to an agency trained in this sometimes technical area sharply reduces the preventive powers of the subsection. If an alert but busy stockholder uncovers an illegal trade, the inconveni- ence of pursuit and the smallness of his pro rata interest in the recovery make it unattractive for him to take action. Thus, enforcement generally rests with the corporation as plaintiff, and litigation is the least attractive remedy. If litigation is started, the corporation is likely to seek 16 The corporation is settlement at a compromise amount. equally unlikely to pursue an appeal of an adverse decision by a trial court. On the other hand, even if the Commission were 16Cook and Feldman report instances where actions taken have neither been settled nor concluded by a judgment of the court. Instead the plaintiff has "voluntarily" dis- continued the action. 3O empowered to investigate transactions that appear suspect, the more subtle forms of evading disclosure would make profit-detection difficult. For example, an insider with knowledge of the probable exchange ratio in a merger can evade disclosure by having a relative or friend act as the ostensible principal in the transactions. A critical hglding period pf six months Not all short-term trades are improper; some may be explained by "special circumstances." However, in my opin- ion, speculations based On confidential information belong- ing to the corporation are unethical regardless of the length of the holding period. Use of information regarding signi- ficant changes in financial plans and policies (such as dividend changes, splits, liquidations, and mergers) for personal gain is a clear breach of the fiduciary relation- ship between the executive and the stockholder. The Act is remiss in not authorizing recovery rights on transactions Of this type that result in profits realized over more than six months. Executives are usually wealthy men, and present tax laws provide for a differential tax treatment between short- and long-term capital gains. Any improper application of inside information is most likely to result in long-term stock market profits, and hence not come 31 within the compass of the law. Ethical behavior is, in the final analysis, a matter of individual character. Nevertheless, it would seem that directors might well set down a company policy for the guidance of its officers on this delicate matter Of insider transactions. Such a policy should not discourage insider buying, but make it very clear that purchases made on the basis of information not available to other stockholders are unethical and will not be tolerated. Insider selling has obvious disadvantages for the company, and should probably be discouraged under all circumstances. A more conservative policy would perhaps require that all company executives refrain from stock trading of the company's own stock, on grounds that even though the official may not have informa- tion inaccessible to others, the others are likely to assume that he has. In other words, his position makes it impos— sible for him to behave as a private individual when it comes to trading his own company's stock, no matter how good his intentions. This discussion of the ethics Of insider trading is not irrelevant because, it will be remembered, the assump- tion that the executives "feel free" to invest following profit-maximizing goals was part of the reasoning that led to the formulation of the hypothesis that their transactions 32 have forecasting properties. If the insiders regard buying or selling as unethical, or if the corporation does not sanction insider trading, the empirical evidence in a random sample which includes such individuals and companies may fail to support the hypothesis, and we may erroneously re- ject it. Possibly the forecasting usefulness of the data depends upon filtering out the transactions of such insiders or companies. Certainly, if the observed trading patterns differ markedly among companies, one reasonable explanation may be the differences in corporate policies and attitudes that govern insider trading. Policies need not be written, and attitudes may be influenced by differences in tradition, corporate history, personal background, and the personality (including ethics and character) Of the chairman. To sum up, the historical and legal review in this chapter suggests that the insiders may in fact know more than the general public and, in addition, that they may be entitled, legally if not ethically, to anticipate future trends in the price of their stocks if they hold their pur- chases at least six months. CHAPTER IV THE HYPOTHESIS AND A METHOD OF INVESTIGATION If insiders seek maximum profits, their transactions in their own company's shares may forecast the price of their company's common stock; if their transactions and the company's stock price are significantly correlated, the transactions are successfully forecasting the stock price. This is our hypothesis. The purpose of this chapter is to explain its rationale, describe a method of investigating it, and derive values for the variables used in testing it. A key problem, typical of all inductive studies, is one of definition and measurement Of the variables which relate to the hypothesis being tested. W The hypothesis that insiders forecast through their transactions the market price Of their own company's common stock assumes the standard profit maximization hypothesis under conditions of uncertainty. Assuming that executives follow maximum-profit goals does not necessarily mean that 33 34 all such transactions reveal the insider's assessment of his company's prospects. Other factors may motivate his behav- ior. A sale of stock need not mean that the seller believes the shares have attained their full value on the market, just as a purchase need not mean that the purchaser believes the shares are being sold to him at less than their full value. yaripps possible motivations behind a sipgla transaction Insider selling.--An insider with large holdings may sell part of his stock for estate planning or other tax pur- poses, even though he may believe the stock has further po- tential for appreciation. The market for the security may be thin, and he may decide to distribute his holdings gradu- ally rather than make it necessary for his estate (if he dies) to liquidate a large block of stock at a later time. If an executive's holdings represent a major part of his assets, he may be moved by a desire for diversification. The insider may have been paid in stock rather than cash when he sold a business to or merged with the parent company. Perhaps the most common reason for a sale is the need of cash for taxes, consumption or some capital employment. He may need cash to exercise previously granted Options to buy 35 shares under the current market price. Or, an insider may be dissatisfied with new management or with a changed policy of old management; having thus lost his feelings of "loyal- ty," he may be looking for a new company and a new career line where he can feel his talents are more greatly appre- ciated. If transactions based on such suggested motives could be identified and eliminated, the remaining sales transactions might reflect the insiders' view that the shares in question are fully valued by the market. However, data do not reveal motivations behind a sale, and such "misleading" transactions cannot be excluded from the sam- ple. Insider buyipg.--Purchases by insiders are very Often the exercise of Options to buy shares at prices sub- stantially below the current market. The exercise of such Options is not a necessary indication that the insiders consider the shares a good value at the current market price. Fortunately, since these purchase transactions on option are reported separately, they can be excluded from the sample. However, there are other possible motives underlying purchases that cannot be thus identified and eliminated. For example, a purchase may reflect an insid- 36 er's desire to strengthen his position in relation to others who have a sizeable stock holding in the company. A proxy fight may be on the horizon. A purchase may be necessary to qualify an executive to serve on the board of directors. An official may increase his holdings to satisfy feelings of company loyalty, to achieve savings goals, or to accord with corporate mores despite a forecast of price decline or no forecast at all. In sum, one limitation in the available data is the absence of an indication of the motive behind a particular sale or purchase. Generally speaking, a sale Of stock by an insider appears to be subject to a greater variety of inter- pretations than a purchase at the market price. Neverthe- less, since the limitation affects both types of transac- tions, no single transaction can be assumed to reflect the executive's assessment of his company's prospects. Tha advantages Of aggregating transactions in a company The wide variety Of possible motivations behind in- sider transactions, and the inability to deduce motivation from the available data, suggest the need for aggregating the transactions within each company. While the forecasting function may not be true for any single transaction, it is conjectured that the hypothesis may reasonably apply to the 37 transactions of executives as an aggregated class. If the non-forecasting transactions are assumed to be randomly distributed over time, and there seems to be no reason why they should not be, an aggregate measure of the transactions within a company during a month may reduce the influence of the "misleading" transactions. If so, such a measure may indicate the insiders' assessment of their company's pros- pects. A positive relationship.--If an index of the insid- ers' transactions reveals their assessment of the company's prospects in the market, the forecasting hypothesis of a positive relationship between it and the stock's lagged in- dex of price-performance may have some theoretical founda- tion. If the executives are all profit maximizers, the fact that they do not buy and sell in the same month can be "ex- plained" in terms of their risk preference. Risk preferencas.--Executives may be reasonably ex- pected to have different degrees of risk preference. When a number of executives take the same market action it may be presumed that they feel the risk of error is slight and they are therefore minimizing their risk in these particular market transactions. Thus, the greater the number of net buyers or net sellers, the greater is the risk aversion or 38 the more the risk is minimized (and the more important is the inside information). For example, three net buyers in a stock during the month may be assumed to summarize the activity of "risk minimizers, men willing to invest in the security at a time of a larger expected return because of the smaller variation. In contrast, an aggregate company index of the in- siders' transactions denoting one buyer or seller during a month might suggest the least risk aversion (or the largest risk preference). Other things being equal, it may be hypo- thesized that the insider's expectation is one of a larger variation. However, in the long run the expected returns may be smaller because of the larger variation. Theoretically the expected returns (by the index number denoting the insiders' monthly activity) yield a positively sloped regression line. The nature of the "in- side" information (and its utility to the executives) is beyond the defined limits of this investigation. For our purposes it is irrelevant whether or not the information on -which the insiders presumably act is strictly confidential or discounted to some extent in the current price of the stock by thOSe outsiders privy to this "confidential" infor- mation. 39 The Model A linear model Since the Objective Of the study is to find some simple explanation of the assumed relationship, a linear expression of lagged price performance per transactor has been selected.1 Admittedly the real world is not this sim- ple, but a linear model is useful (and workable) and should give a fair indication Of what is going on. The model is used to examine the relationship between the insiders' transactions and the variation in the stock price from the date of the transactions to a date six months later. Thus, Yi,n = Ai + Bi xi,n + Ei,n where Yi n = the index of performance of stock I, begin- ’ ning at month p, over a six-month interval. It is simply the percentage of change in a particular stock's price six months later less the percentage Of change in the Securi- ties and Exchange Commission's composite com- mon stock index for the identical period.2 1I did not experiment with a curvi-linear model. Si, n + 6 I Si,n _ Ei,n + 6 ‘ Ii,n 100 Si,n Ii,n 2 _ Yi,n ' 40 = the number of insiders buying, minus the number of insiders selling the particular stock during the month.3 = a constant (the intercept), the average per- formance of stock Id over six-month inter- vals, if there are no insiders' transactions during a month. = the regression coefficient, a constant (the slope), the average percentage of change of stock I, over six-month intervals, for each insider trading during a month. = the random factor producing the "unexplained" variation Of stock I_at month p_over a six- month interval. These notations may be simplified. For a given where Si,n = i,n n: 3The the stock price of company I,at month p_ — the composite SEC StOck Price Index of 300 common stocks on the NYSE at month p, month of Observation, 1-48, 1957-1960; see Appendix B. value Of this variable is negative when the number of sellers is greater than the number of buyers. A zero value indicates an equal number of buyers and sellers, (usually zero less zero). In this study the term "net" means that we are netting buyers and sellers in a particu- lar company during a given month. (We will not net trans- actors between months in a given company.) 41 company we assume Y = A + BX + E. We suppose E to have a distribution with a mean of zero so that the expected value of Y is given by A + BX. Since A may differ for each com- pany, this term is not combined with the error term, E. The statement that insiders' transactions have no predictive value is, in terms of this model, equivalent to the hypo- thesis that B is equal to zero. Large values of the correla- tion coefficient between variables X and Y will tend to in- dicate that B is not zero. While the correlation coefficients will be analyzed in detail, the slope (i.e., the regression coefficient) and the intercept will not be presented because of the lack of a useful economic interpretation to these parameters. The theoretical interpretations of these terms are discussed below. Tpa intercept and the slope.--The intercept is the estimate of the percentage that a particular stock will ad- vance or decline, relative to the market as a whole, during a six-month period if there are no insiders' transactions. The slope is the estimate of the additional percentage change that the particular stock will undergo for each additional transactor. It is supposed that every company has a unique lepe and intercept. Buyers give the total 42 estimate a positive boost, and net sellers provide a nega- tive "kick" to the price change of the stock. For example, assume that the average slope and intercept for one corpor- ation are 2 per cent and 3 per cent respectively and that there are three more purchasers than sellers (+3) on January 15, 1958. The estimate of the relative price change six months later is 11 per cent, i.e. 2 per cent plus 3 per cent times +3. Also, assume that the slope and intercept for a second corporation are -l per cent and l per cent, and that there are three sellers net of buyers during a month. The estimate of this relative price change is -4 per cent. As a third hypothetical illustration, assume that the A and B terms are estimated to be -2 per cent and 10 per cent re- spectively for a third corporation and that there are zero transactors in January, 1958. Our oversimplified model would estimate the total relative change forecast by the zero transactors to be equal to the intercept, or -2 per cent in this case. In other words, interpretation of the slope and intercept in this model are estimates of relative price change unique to a particular corporation. A company is generally either advancing or declining relative to the market, and the rate of change varies for individual companies. Companies differ in size, location, capital structure, and management, and the price fluctua- 43 tions in their stocks appear to reflect this, reacting dif- ferently to economic and psychological forces (after elimi- 4 It is not nating the influence of the market as a whole). uncommon for a stock to rise in a bear market or to fall in a bull market. For example, during the year 1957 in which all comprehensive stock price indices declined, 801 listed issues declined in price, but 229 rose; and 7 of the 229 5 Observing this phenomenon, rose by 50 per cent or more. one economist stated: The diversity of simultaneous price behavior of indi- vidual stocks is so wide as to render the action of any given stock price index unrepresentative of a variable, but usually substantial segment of the mar- ket.6 The fact of this heterogeneity in the data suggests that there is no reason to assume that a company's stock price will behave as the "average" stock. Therefore, there is no reason to assume that the A term in this model should equal zero. Indeed, on the basis of chance in individual cases the A term would not be equal to zero. 4While it may be reasonable to attempt to group companies according to their "risk class" in an aggregate study, this is beside the point for present purposes. 5Robert W. Storer, "A Critical Evaluation of Stock Market Indexes," American Statistical Associatipn 1952 Pro- geedings: Basiness and Ecgnomics Section, p. 35, citing New Xgrk Stock Ezchanga Fact ngk, 1958. 6Ibid. 44 In the aggregate data it is not possible to inter- pret the regression estimate with reference to the popula- tion. Since the composite tests for correlation did not include the relative performance of those stocks in the sample with zero transactors,7 the resulting intercept is not an estimate of the population. It is the average inter- cept of those stocks in the monthly sub-sample, and this group was not randomly selected. The sub-sample is composed of stocks in which there were insiders' transactions. Ad- ditional reasons for expecting that the intercept will dif— fer from zero are (l) the chance phenomenon that a sample of 100 is not going to reflect the population of 1,065, (2) the performance Of the sample is related to a population of 300 instead of the population of 1,065, and (3) the fact that the monthly sub-sample is not randomly selected. On the other hand, if all the stocks in the sample were tested each month, and if the real world were actually duplicated by the hypothetical, then the intercept should 7Inferences cannot be statistically drawn with ref- erence to the question: Does the absence of insiders' transactions in a given stock indicate that the company is an average performer? While an answer to this question would be interesting (and significant), it is not central to the thesis and, therefore, would not have justified the time and effort involved to compute the index of a stock's per- formance related to the case of zero buyers less sellers in a month. Therefore, the forty-eight months of inside-trading in 100 stocks yield 1,507, not 4,800, Observations. 45 theoretically tend to be zero on the Y axis (except for a technical problem noted below).8 As noted earlier, the oversimplification of this model does not reduce its use- fulness. Six-month lead tested.--The rather arbitrary choice of a lead interval of six months is based upon a theory that if the hypothesized relationship is valid for six-month in- tervals, there is probably also some relationship for inter- vals of four, five, seven, and eight months. The determina- tion of the most useful lead for various purposes is a mat- ter for further study. On the other hand, the available empirical evidence does suggest short-term rather than long-term forecasting properties. For example, during the period from June, 1954, through December, 1960, Hamada found the highest correlation for the index with a four-month lead rather than other lead- lag intervals tested, and the coefficient of correlation diminished as the lead was lengthened.9 For the full period 1936-1960, Hamada found that the median lead was six months 8Even in this case, however, the A term would tend towards the positive side of zero because of the slight bias introduced by the cumulative effect of arithmetically aver- aging successive plus and minus percentage changes. 9 p. 120. Hamada, "An Analysis of Diffusion Indices . . ., 46 from the peak Of the insiders' transactions index to the peak of the Dow-Jones Industrial Average: "The median lead is six months for peaks and five months for troughs, with the standard deviation of the lead in both cases less than its length."10 The earlier study summarized in Exhibit 2-1 found an eight-month interval preferable.11 These varied conclusions do, at least, suggest a lead of less than twelve months as the Optimum holding period. SEC cpmppsite index.--The choice of the Securities and Exchange Commission's composite stock price index as the reference series for deflating the general market move- ments is based on its being an index of conceptually sound 12 It has greater coverage than the more construction. widely known Dow-Jones averages, and its weighting scheme is more consistent over time.13 The stocks selected at random 101bid., p. 143. 11This study suggests a six-month interval, on the average, as the optimum holding period. However, six months in that study is the equivalent of eight months in the cur- rent study, due to the two-month lag from the action date Of the speculator to the action date of the insiders. 12The Standard and Poor's indexes are of equal theo- retical soundness for our purposes. Storer, loc, git., pp. 27-43, and Discussion, pp. 49-50. l3Irwin Friend, F. E. Brown, Edward S. Herman, and Douglas Vickers, A Study of Mutual FundS, A Report Prepared for the Securities and Exchange Commission by the Wharton 47 represent a cross-section of the majority of industry classi- fications. However, since most averages Of the market tend to move together anyway, the choice of one over the other does not appear to be crucial. Rationale for deflating Theoretical backgroand.--The technique adopted is to analyze the executives' transactions and the subsequent share price of the company's common stock corrected for changes in the level of the stock market. It is an attempt to explain the interrelationships between the variables in the model. The first step in the analysis is to divide the variables assumed to be significant into two broad classes: those which describe internal forces in the system, endo- genous variables; and those which describe external or exo- genous variables.l4 While the particular level of a stock's price is presumed to be in part a function Of the indepen- dent external forces acting on the market as a whole, the School of Finance and Commerce (Washington: Government Printing Office, 1962), p. 293. 14The external forces assumed to be reflected in the market averages include such factors as interest rates, tax rates, inflation fears, etc., and other social, econom- ic and political environmental considerations. 48 adjusted stock price deflated for this exogenous force is considered to be the dependent variable, internal to the system. This stock price, free of external market influ- ence, is dependent on such factors as future earnings and dividends, which are themselves a function of profitability, growth, management, location, etc. of that particular com- pany. The model makes no attempt to explain changes in the market as a whole,15 but it does attempt to eliminate the effects Of changes in the level of the market on the par- ticular stock's price. The assumption is that the level of a stock's price is largely determined by forces independent Of the internal forces presumed to be manifested earlier to the insiders. Thus, the model is constructed to consider the market as a whole as the exogenous factor, and the in- dividual company's adjusted stock price (after eliminating the influence of the external forces) as the internal 15The question of whether or not insiders can fore- cast the market as a whole is not addressed in this study. If one does not remove the effect of the market on a particu- lar company's stock price, the analysis relates to the pre- dictive ability of the insiders' transactions with respect to two variables: the change in the market, and the change in the stock itself independent of the market change. If there is predictive ability in the transactions, the question re- mains: Which Of the two variables is most significant? (Are the insiders market-wise or company-wise?) A later study might attempt to predict the variation of both variables. 49 dependent variable. The model can be tested to see if a change in the pattern of insiders' transactions foreshadowed a change in the relative price of the company's stock. Thus, changes in the public's expectations of future earnings and divi- dends are not investigated directly, but are assumed to be reflected in the stock's adjusted price. The statistical argument for deflating is implicit in this discussion. It is simply a means for studying the price variations of a stock which are caused by the factors resulting from the company's idiosyncrasies as they are "explained" by the transactions of the insiders: a relative price that is free Of the variable forces within the market itself that result from changes in the company. In addi- tion, since stock prices were generally rising substantially during 37 of the 53 months ending June, 1961, by deflating, we can expect an approximately equal number of cases of stocks to move up or down relative to the market. This will tend to simplify the presentation and facilitate the penny- tossing analogy for the non-statistician.16 16However, the applicability of the adopted statis- tical tests does not depend on the assumption that the true probability is one-half for a single success. Nor must one assume a normal distribution. While the frequency distribu- tion of successes in a series of penny-tossing games may happen to have a normal distribution, this is not a pre- 50 The deflating process illustrated.--In this model an insider can incur a loss (that is, not make a profit) and still increase his wealth relative to the performance of the "average" alternative stock investment. For example, assume that Mr. Smith buys shares in his company at $10 per share at a time when the market index reads 100. If the stock declines to $9 per share and the market index declines to 60 six months later, Mr. Smith has not maximized profits. He has incurred a 10 per cent loss in capital value. How- ever, despite his loss the situation tends to support the hypothesis. Mr. Smith's loss in capital value does not mean he failed to. forecast the price of his company's stock successfully. On the contrary, his forecast on this score was accurate. His loss only indicates that he failed to forecast the movement of the stock market in general. This case will serve to illustrate the experimental manipulations to be performed on the data: Month 2. SlX months Percentage later of change Stock price (Si,n) $ 10 $ 9 -10.0 Market index (Ii n) 100 '60 -40,0 ’ Stock price percentage of change less the market index percentage of change for six months (Yi,n) + O O requisite of the tests used. Provided the assumptions of the tests are valid, the statistical treatment would be just as appropriate without deflating. 51 Relative to the market, this stock increased 30 per cent. It is not known what percentage of change Mr. Smith forecast, if any, but a relative increase of 30 per cent resulted. If Mr. Smith had sold instead of purchased,our method would say he had forecast in error, despite superior stock market timing due either to skill or chance. This deflating process illustrates the dependent variable in this study. The hypothesis to be tested has no concern with whether or not insiders can forecast the market as a whole. The study is restricted to whether or not in- siders forecast the movement of their own company's stock price, defined as the deflated or relative performance over six months. A behavioral argument fpr deflating.--Another argu- ment for deflating rests on a behavioral assumption. If insiders thought the market in general was going to rise, there would be no special reason to buy their own company's stock unless they thought it would perform notably better than the market. In view of the requirements for reporting insiders' transactions and the limitations on the nature of those transactions, it would seem more likely that execu- tives who expected general business and stock price im- provement would buy stock of other companies than their own. 52 Therefore, there would have to be a reason for buying their own stock strong enough to outweigh whatever disadvantages might accrue from the disclosure Of such transactions to the public. Rules for Collecting and Summarizing the Data Insiders' transactions The basic information on the insiders' transactions is published monthly in the Official Summary.17 This month- ly summary lists the information given in the reports of the insiders filed during the month preceding its publication: information on the company, type Of security traded, name of _corporate official, his affiliation with the company, char- acter of the transaction, nature Of ownership, month of transaction, and month-end holdings of security traded. In general, rules were necessary to sift out of the mass of data presented by the SEC only what could be used in formulating an aggregate company measure of the insiders' transactions. The criterion is implicit in the question: might this transaction in and of itself have a forecasting value, i.e. does it signify a conscious decision on the part 17Securities and Exchange Commission, Official Sum- mary of Security Transactions and Holdings, Washington, D. C., published monthly. 53 of the insider relevant to the hypothesized over- or under- valuation of the particular stock at its current market price? The rules adopted here for the collection of the transaction data are presented in Appendix C. Briefly, common stock acquired by insiders by the purchase of new right issues, by the exercise of conversion Options attach- ing to other securities or through compensation or stock Option plans have been excluded from insiders' market pur- ichases. These procedures provide a clearer indication of insiders' Open market transactions and are logically con- sistent with the hypothesized predictive value of these transactions and the subsequent short—run price behavior of the company's stock. The sales data, on the other hand, include the Open market transactions of all shares ac- quired. Thus, while insider acquisitions may very well con- genitally exceed dispositions, in this study the Opposite may appear true because of the inclusion—exclusion rules for collecting the data. Another self-imposed rule which tends to reduce both purchases and sales is the exclusion of the disclosed Off-the-board transactions in listed secur- ities. Occasionally noted private transactions between parties were excluded. The major category excluded from the purchase 54 transactions is the exercise Of stock Option grants. As (noted earlier, such purchases were excluded because they represented bargain purchases at the date of exercise. On the other hand, key executives frequently sell a part Of their holdings in order to finance the exercise of Options.18 Thus, although it was thought to be appropriate to exclude Option purchases, it was not feasible to exclude sales of option-acquired stock. Theoretically a sale of stock by an insider with a stock Option, which is specifically trans- acted to provide the funds to meet the option price in order to take up the grant, should be excluded. Since these were not and could not be excluded, this particular limitation could quite possibly diminish the expected correlation.19 Late reports.--As noted in Chapter II, approximate- ly one out of every twenty-five transactions is reported late. While 90 per cent of these are tardy by one or two months, the remaining 10 per cent are late by as much as two years. It was necessary to formulate a rule to handle this 18Henry Rothchild, "Financing Stock Purchases by Executives," Harvard Business Review (March-April, 1957), p. 136. 19Incidentally, a Commission rule exempts the sale of Option-acquired stock, under certain conditions, from the scope of Subsection 16(b). This accounts for the alleged frequency of occurrence of this manner of financing option purchases. 55 limitation in the data, and it was decided that the late re- ports would be included in their proper transaction month as if they had been reported on time. This procedure has the effect of improving the accuracy of the reported data. These current omissions, transactions known only by subse- quent report, reduce the data available to the outside fore- caster. This would be an important disadvantage if one wanted to beat the market, but since beating the market is not the purpose of this study, the late reports are included in the data tested.2O Selected indax of transactions.--While there are numerous methods for deriving an index number for the in- siders' transactions variable, the selected measure is based on the number of transactors. It is simply a monthly sum- mary of the number of insiders buying, less the number of insiders selling the stock. A transactor who increases his holdings on balance during the month is assigned a weight of plus one. A minus one is assigned to each transactor who reduces his holdings on balance during the month. The sum Of these assigned numbers represents the value assigned to 20The reader interested in stock market speculation should keep in mind that, in addition to late reports, there is a lag in gathering and publishing the data of approxi- mately one and a half months. 56 the insiders' transactions variable for a certain company during a given month. For example, an Observation with a weight of plus three tells the reader that there were three purchasers more than there were sellers of the stock during the month. A monthly observation weighted minus three indi- cates that there were three more sellers than there were purchasers. A zero weight indicates that the number of buyers was equal to the number of sellers. In some in- stances a zero weight summarizes activity of an equal num— ber of buyers and sellers; however, in the vast majority of cases it indicates no activity and, as previously noted, the monthly Observations weighted zero were not tested. Among the Observations tested, the sign of the index number indi- cates the direction of the forecast (in the model) and the number, itself, expresses the relative importance of this particular forecast of advance or decline in the company during the period studied. This technique weights each insider's monthly trans— actions equally, and weights each monthly Observation in the time series of the company according to the number of buyers less the number of sellers. Thus, each monthly observation in a particular company may have a positive, zero or nega- tive value. And the forty-eight monthly values in a com- pany's time series may reveal a consistent, cyclical, or 57 irregular trend or pattern. Since insiders have various motives, the inference of profit maximization is more likely the larger the number of observed transactors who behave alike. The greater the number of net buyers or net sellers in any given month, the greater the confidence one may presumably place on the magnitude of the forecast (but unknown) percentage change of advance or decline, respectively. Another useful econom- ic interpretation, drawn from the utility maximization hypo- thesis, is that the inference that action taken on the basis of inside information carries relatively little risk, the larger the number of net buyers or net sellers. For exam- ple, the greater the proportion of buyers over sellers, the more favorable was the information available to the insiders and the more likely this stock to advance. Similarly, the greater the number of net sellers, the more Significant was 6 the adverse information available to the insiders and the more likely the stock to drop in price. An accurate inter- pretation Of the available data is impossible; nevertheless, the adopted technique is simple to compute, and the result- ing index numbers are easily understood (in terms of the model). Evalpation of index of transactions.--Deriving a 58 value for the transactions variable presents the problem of interpreting the importance a transaction had in the mind of the transactor, and the problem of finding a common denomi- nator by which separate transactions can be compiled into one number. Does the purchase Of 1,000 shares by one official have ten times the forecasting significance as does a pur- chase Of 100 shares by another? Or, is it ten times as im- portant only if each executive originally held the same num- ber of shares? Purchases or sales of dissimilar quantities may be weighted equally or in accordance with size or some other criterion. Because individual executives differ from one another in terms of financial resources and risk prefer- ences (data which are not known), it was decided to give each purchaser or seller the same weight, regardless of the size of the transaction or its relationship to the insider's holdings a month earlier. Consequently, there could be a situation where five insiders purchase a total of 1,000 shares while one official sells a total of 5,000 shares. While the value assigned to these insiders' transactions is plus four, the officers are actually selling the stock, on balance, during the month. The executive who sells 5,000 shares may consider this an insignificant item in his total wealth, whereas there are five executives presumably "voting" that the stock price 59 will advance; there is no knowledge, really, as to the im- portance of these transactions to the wealth-income position of the transactors. While a $3,000 transaction is weighted equally with a $100,000 transaction, the adopted method does not equally weight an insider's transaction of less than $3,000. The Securities and Exchange Commission's rules for reporting generally exempt small transactions of less than $3,000 from the disclosure requirement.2 The weighting technique does have limitations. The index number of buyers less sellers gives the same weight to a transaction which reduces an insider's holdings by 10 per cent as it does to another's transactions which reduce his holdings by 100 per cent. If the selling insider did, in fact, forecast a relative price decline and felt it was im- minent, he would probably feel a greater urgency and sell a larger quantity of his personal holdings. However, if these two transactions were the only two in a particular company occurring in different months, our system would weight the forecasts equally. Or, the executive who felt strongly that the price was going to advance might buy 500 shares, whereas if he were less confident that the price would rise, he 21See Appendix D, SEC Rule l6a-9. 60 might buy only 100 shares. On the other hand, all the bases for index construction that were considered seem to ignore the possibility that to a junior Officer 100 shares may be more important (telling) in terms of his personal wealth than a transaction of 100 shares would be to a director or senior Officer. Stock price and stock market rallies. The values of both the stock prices, S and the i,n’ stock market averages, Ii n’ are one month apart and the 5 tested intervals of a stock's relative performance, Yi,n’ are approximately six months apart. A mid-month Friday is the point in time at which the closing stock price and mar- ket average are Observed and recorded. The insiders' trans— actions are assumed to be made at that moment and price. Actually, the transactions are made throughout each month at varying prices, but the exact date of a transaction was not given in the Official Summary for the years tested.22 Averaging the daily prices or the monthly high and low would have increased the number of calculations. The alternative of computing an average price for the month was discarded in 22In 1963 the Commission resumed the practice of reporting the day of each monthly transaction and thus remedied this particular limitation. 61 favor of a specific price at a specific point in time. Since this was the method Of recording the prices of the stocks, it was the appropriate method for recording the readings of the SEC Composite Index. Adjusting stock prices for stock dividends, splits, app,--The stock prices were adjusted to make proper correc- tions to offset the effect of changes in the number of shares outstanding as a result of the issue of rights, stock dividends, splits, etc. For example, forty-eight Of the issues in the sample had stock dividends during the four and one-half years ending June 30, 1961. The ex-dividend dates were determined, and appropriate adjustments in the price of the stock were calculated. Appendix E is provided to illustrate and thereby summarize the mechanics of extracting values from the raw data. Aldens Inc., the first company in the sample, is the example used. Summapy A method of investigating the correlations between the insiders' transactions and the price of the company's stock has been described. While there are other and pos- sibly better techniques, the one developed for this study is believed to be both sound and workable. This model, summar- 62 ized in notation form earlier, provides the plan for the study. Since the transactions are reported by month, the values for this variable are monthly. The various tests employed will be limited to a leading time of six months. In general, positive values in the insiders' transactions index summarize the intensity of the insiders' net buying, and are interpreted as a forecast of an increase in the price of a stock that is relatively greater than the general market movement. Similarly, negative values summarize net selling and a forecast of decline in a stock's price rela- tive to the market as a whole. The statistical groundwork and collection procedures have been summarized in the text and illustrated in the appendix. The study's focus is on the company and its monthly transactions. Since there are 100 companies and forty-eight months during 1957-1960 being considered, there is a maximum possible number of 4,800 observations. How- ever, the sample actually consists of ninety-eight companies and 1,507 Observations, because the executives in two of the 100 companies never traded, and the executives in the other companies executed transactions about once every three months. This average conceals, however, the diversity of the activity, which ranged from one to thirty—nine Observa- tions per company during the forty-eight month period. 63 Each observation has a pair of hand-calculated values: the index of the insiders' transactions and the index of the stock's performance six months later. These two values, along with the identifying company and month number, were punched on 1,507 machine tabulated cards for further detailed analysis. 2 . . . . . 3The summariZing statistics listed below were cal- culated on a Control Data Corporation l60-A computer. A program was written to calculate sums, sums of squares and cross products, means, standard deviations, variances, co- variances, correlation and regression coefficients between the two variables. CHAPTER V ANALYSIS OF STATISTICAL RESULTS This chapter will test the thesis of this study: whether the corporate insider forecasts through his trans- actions the future market price of his company's stock, independent of the market trend. If his stock purchases are followed in six months by a better-than-average market price and his sales of stock by a worse—than-average market price, we can conclude that the insider's transactions do provide forecasts. This chapter is organized into two major sections. The first examines the magnitude and direction of the lagged-price performance. The second section examines the direction of change without regard to extent of percentage of change. Each section is generally organized with the aggregate results (usually forty-eight monthly composites) preceding the company results over time. The company re- sults are usually summarized in the text and detailed in the appendixes. The essential techniques of description are correlations and mean-performance statistics in the first section and binomial probabilities in the second. Tests are 64 65 run on the sign of the correlations and the difference be- tween sampled means. Chi-square goodness-of-fit tests are run on the sampling distributions in the final section. The sub-hypotheses proposed in the segments of each section are generally stated in the introductory remarks to that sub-section. Forecasting the Magnitude of Change The hypothesis assumed in the linear model (see Chap. IV) states that the larger the net number of trans- actors, the larger the rate of advance or decline Of the stock after six months, relative to the market. Correlation analysis is first employed to test this hypothesis.1 Posi- tive correlation coefficients support the validity of the hypothesis. Therefore, the larger the coefficient, the stronger the linear relationship between the transactions and performance indices. The data are first grouped by month Of Observation, and the correlation coefficient is computed for each group. Thus, during 1957-1960, there are forty-eight correlation 1The product moment correlation coefficient will hereafter be referred to as the correlation. It is calcu- lated by using a linear regression analysis of the stock performance index on the insiders' transactions index. Be- cause Of the six-month overlap among some of the Observa- tions in the company data, these correlations through time cannot be statistically tested. 66 coefficients between the two indices based on 1,507 observa- tions. These results, ranked by size of correlation, are shown in Exhibit 5-1. If the forty-eight correlation co- efficients seem to yield the same result consistently, it will be substantial evidence tending to confirm or deny the hypothesis that insiders forecast through their transactions the market price of their own company's common stock six months later. . Support for the foracasting hypgthesis The evidence suggests that, in general, the correla- tion is probably positive, but it does not appear to be large. The median correlation and the average correlation were both found to be +0.132. While the linear relationship appears to be weak, the frequency distribution of the forty-eight results in Exhibit 5-2 is striking. Since forty of the forty-eight correlation coefficients have positive signs, the forecast- ing hypothesis is supported and the slope of the regression coefficient is very likely positive.2 The statement that insiders' transactions have no predictive value is, in terms of this model, equivalent to the hypothesis that the proba— 2The average regression coefficient for the sample was found to be +1.2 per cent per transactor. 67 EXHIBIT 5-1 MONTHLY CORRELATION RESULTS BETWEEN PERFORMANCE AND CORPORATE INSIDERS' TRANSACTIONS INDICES DURING 1957-1960, RANKED BY SIZE Correlation of Net Number of Buying and Net Selling Observations Observations and the per Month Relative Price Performance Six Months Later 22 +.5016 3O +.4923 29 +.4722 31 +.4422 33 +.3352 37 +.2999 31 +.2872 30 +.2812 36 +.2735 35 +.2597 36 +.2546 30 +.2420 34 +.2368 40 +.2237 44 +.2108 29 +.l979 33 +.l766 32 +.1637 36 +.l621 28 +.1438 33 +.1436 29 +.1387 32 +.1348 34 +.l346 30 +.1290 32 +.l242 32 +.1150 68 EXHIBIT 5-l--Continued Correlation of Net Number of Buying and Net Selling Observations Observations and the per Month Relative Price Performance Six Months Later 37 +.1077 36 +.103l 27 +.O977 28 +.O68O 37 +.O615 35 +.0518 31 +.0502 43 +.O49l 36 +.O384 21 +.0298 31 +.0280 23 +.O234 34 +.0024 27 -.Ol42 23 -.0147 23 —.0596 25 -.0959 31 -.1272 31 -.l466 31 -.2293 19 -.2806 Total 1507 69 EXHIBIT 5-2 FREQUENCY DISTRIBUTION OF 48 MONTHLY CORRELATIONS BETWEEN INSIDERS' TRANSACTIONS AND RELATIVE STOCK PRICE PERFORMANCE LAGGED SIX MONTHS Range of Correlation Coefficients Frequency Between +.500 and +.599 1 Between +.400 and +.499 3 Between +.300 and +.399 1 Between +.200 and +.299 10 Between +.100 and +.199 14 Between .000 and +.099 11 Between -.100 and -.001 4 Between -.200 and -.101 2 Between -.300 and -.201 2 Total 48 70 bility of obtaining a positive correlation is equal to the probability of obtaining a negative correlation. This hypo- thesis of randomness Can be rejected at the .00001 level of confidence.3 On the other hand, a linear model results in variable correlations and a weak correlation on the average. Variation in monthly correlations.--Only six of the 48 monthly tests, considered individually, yielded statisti- cally significant results.4 In Exhibit 5-3 the six signifi- cant correlations are listed along with six negative corre- lations as a further indication of the range. The dates of the net buying and net selling Observations are also pro- vided. The reasons for the curiously large variation in the correlations now need to be considered. One possible explanation is that the general behav- ior of the market was a factor, despite efforts to eliminate its influence. In an attempt to discern a pattern in the correlations, Exhibit 5-4 is presented. However, inspection shows that there is no consistent pattern in the correlations 3The binomial test with respect to the Sign Of the correlation coefficients is provided in Appendix F. Statements of statistical significance Of correla- tion coefficients computed from linear approximations are based on Table V‘A Of R. A. Fisher, Statistical Methods for Research Workers as reproduced in D. B. Owen, Handbook of Statistical Tables (Reading, Massachusetts: Addison-Wesley Publishing Company, Inc., 1962), p. 510. 71 EXHIBIT 5-3 SIX LARGEST AND SIX SMALLEST MONTHLY CORRELATIONS Month and Year of - bs r - orr 1 - ' ' ° Net Buying and Net 0 e va C e a Significance Selling Observations tions tion Level Mar., 1957 22 +.5016 Yes at .01 July, 1960 30 +.4923 Yes at .005 Apr., 1958 29 +.4722 Yes at .01 Dec., 1958 31 +.4422 Yes at .01 Aug., 1957 33 +.3352 Yes at .05 June, 1960 37 +.2999 Yes at .05 Aug., 1958 23 -.0596 NO at .05 July, 1958 25 -.0959 NO at .05 Sep., 1958 31 -.1272 No at .05 Oct., 1959 31 -.l466 No at .05 Oct., 1958 31 -.2293 No at .05 Oct., 1960 19 -.2806 No at .05 72 ssmo.+ mmmo.+ evmo.u m.mm + .nmm samo.+ smva.+ whoo.+ m.mm + mmma ..amn moam.+ hmmfl.+ ammo.+ o.mm + .omn mapo.+ some.- mmmH.+ e.am + .>oz mmsm.+ omma.- oaam.+ H.Hm I .uoo smmm.+ mHmH.- memo.+ m.om . .mmm mmmm.+ omso.u omao.+ m.mm . .ms< svao.- omom.- memo.+ m.mm + mass omem.+ samo.u msm4.+ m.nm + mass omaa.+ Homm.n ommo.+ 6.6m + mm: mpsa.+ aooo.+ ammo.- m.mm + .um< odom.+ mamo.+ mmoo.+ s.mm + .umz omea.+ mmmm.- Hema.- m.sm . .nmm momo.+ emam.+ evao.u m.mm smma ..cmn ....wwm.wmmo ..MM%.. ..MM“. ..... .0....... mGOHum>HmeO mo mango: me pmmmmq mosmEuomumm moaum Mooum O>Humawm map can macapommcmua .mnwpflncH mo coflumamunou HMO Gm GO Aooa u mmusmmac w 6 an a xwuaH muflmomeou 0mm Ezmzm>02 BHMMflS AflmmZmO mmB m0 QZDOMOMUdm 4 QZ< mmBZOS NHm Qmwwfid mUHmm fiUOBm m>HB¢AmM mmB Q24 mZOHBU¢m2HOMHOm mzu tam .mnmpflmGH mo coflumHmHHou xmch spamomsoo 0mm omoo.+ maoa.+ mmm0.+ s.mHH I .>oz opea.u oaom.u HmeH.+ m.oHH + .uoo mmea.+ oemm.+ mmmo.+ H.4HH . .mmm mmmo.+ Hoao.+ mmov.+ 4.0ma + .m:« smma.+ omsm.+ mvmm.+ m.oma + mass pema.+ Hmaa.u enao.- m.pHH I mess smma.+ moom.+ mmmH.+ m.saa + was Hmoa.+ oemm.+ memm.u m.saa + .nms ommo.+ Hmmm.+ «Hom.u b.4HH + .umz Hmoa.+ mmmm.+ oemm.+ 0.0HH . .nmm momm.+ mmem.+ oomm.- m.mHH + mmma ..amn mmev.+ smao.+ mmeo.+ m.soa + .omn mama.+ mamm.+ mmm0.u m.soa + .>oz momm.- mooo.+ momm.u m.moa + .uoo msmH.- mama.u 4600.- 6.5m + .mmm ommo.n meHH.n mmoo.+ m.mm + .ms¢ ammo.u mmmm.+ mmmm.u R.Hm + mass 4moo.+ ammo.u mooa.+ 6.0m + mesa omma.+ moea.u .«.z m.sm + was mmse.+ ommm.u momm.+ 0.6m + .um< mwmo.+ mspfl.+ moea.u m.mm + mmma ..umz mcoflum>ummno mcaaamm msflhsm Hm>mq GOHOOOMHQ #62 Has #62 662 . . macapmSHmmno MO mnpsoz.xflm Ummmmq Aooa u mmlhmmav “mow can nucoz UmficflfifimUIIOIm BHmHmNm 74 .Hmmsnlumclmco mum? momma Add .muwaamm Hm>o mumhsn mo Hones: won on“ CH GOaumHHm> OHON ou map OHQMHHm>m Roz u .¢.z m.ama I mane m.vma + was m.ema + .umm e.oma + .umz «.mma + .nmm m.oma + Homa ..:m6 64mm.+ mm60.I mamm.+ m.maa + .omo mmmo.+ smom.+ oes¢.I m.maa + .Soz momm.I omem.+ mmeo.I m.HHH I .uoo meHo.I m66¢.+ veom.I o.mHH I .mmm msma.+ Omea.+ smvm.+ m.mHH + .m:¢ mmm¢.+ oeom.+ seem.+ m.eaa I mass mmmm.+ «mem.+ mmno.I m.saa + mesa mamm.+ vsmH.+ ommo.+ m.mHH I mm: memH.+ somo.I mmoo.I «.maa + .umm msmm.+ omso.+ ommm.+ «.maa I .umz Hmeo.+ «Hoo.I mmmm.I H.mHH I .nmm Shoa.+ Hmaa.+ smvH.I m.mHH I coma ..qmn smmm.+ msmo.I mmsm.+ m.oma + mmma ..umn mcoflum>ummbo mcflaamm mcfihsm H0>mq GOHDOOHHQ #OZ HH¢ #OZ #OZ . . mGOH#m>H®mQO MO mango: me pmmmmq monmfinomnmm moflum xOOOm O>Humamm Tau tam mcoflpommcmua .mumpfimcH mo coflumamuuoo Hmw CM CO Aooa u mmIsmmHV w 6 an 2 xousH muflmomaou 0mm UmmnfifladUIlfllm BHmHmNm 75 as the market advanced or declined. Another possibility is that either the net buying observations or the net selling observations, but not both, are related to the stock's performance, and that the other dilutes the total correlation. For example, 31 of the 1,507 observations of executive trading occurred in January, 1957, and correlated with their stock's price performance at +0.0502. The observations in each month were dichotomized .into net buying and net selling Observations and separately correlated (hereafter called sub—correlations). For exam- ple, out Of the 31 observations in January, 1957, 19 were net buying observations and 12 were net selling observations. Correlation of the 19 net buying observations, with their stock's performance (-0.0244) does not support the hypothe- sis, while correlation Of the 12 net selling Observations (+0.2134) does tend to support the sub-hypothesis that in- siders' selling transactions forecast a relative decline of their company's stock price. However, this and other pat- terns of the two sub-correlations are not consistent, and do not help explain the variation in the 48 monthly correla- tions between the insiders' transactions and their company's stock price in the last column of Exhibit 5-4. The preponderance of positive correlations between rust buying and the rate of advance, and net selling and the 76 rate Of decline, seems to indicate a real relationship, al- though for most months there is not enough evidence to demonstrate it statistically. If we ignore this limitation (among the sub-correlations), one pattern does emerge when we compare the proportion Of positive signs to the total months in each sub-correlation column and in the total cor- relation column. Only 27 out of the 47 correlations between net buying and performance, and 29 out of the 48 correlations between net selling and performance, were positive. Compar- ing these two frequencies with the 40 positive correlations out of the 48 monthly tests suggests that an alternative hypothesis might be more accurate--the hypothesis that in- siders forecast through their transactions the direction but not the magnitude Of their company's stock price.5 This hypothesis is tested in the other major section of this chapter. However, the direction-only-model also furnishes only a weak relationship for most companies. Variations among performance of insiders, by pom- 5For example, if the true correlations for the popu- lation are +0.05, +0.05, and +0.25 between net buying, net selling, and net buying and selling, respectively, and per- formance, this would suggest the hypothesis that insiders forecast through their transactions the direction but not the magnitude of their company's stock price. In other words, the correlation (+0.25) would seem to depend more on the difference in performance of insider-bought stock com- pared with insider-sold stock rather than the forecasting properties of either the net number of buyers or sellers. 77 ,papy,--The most plausible explanation for both the variation in the 48 monthly correlations and the concentration of four of the six negative correlations in the July to October, 1958 period is that the executives in some companies success- fully forecast their company's stock price and those in other companies do not.6 This variation by company has its influence on the monthly correlations, depending on the distribution by month and year of the trading activity in the successful and unsuccessful company series. The performance of the insiders' transactions in companies with more than five observations over the 48 months are ranked by correlation coefficient in Appendix G. This ranking attempts to compare the degree of forecasting success of the companies' executives in terms of the hypo- thesis Of a linear relationship between monthly activity and price performance. The frequency distribution of these ranked correlation coefficients is summarized in Exhibit 5-5. 6Still another question is raised by the non-random- frequency Of three negative correlations occurring in three of the four October tests. DO the insiders' transactions reveal a seasonal pattern? While the empirical evidence in this study suggests a negative answer for the sample as a whole, the possibility of a seasonal influence in the trans- actions in some companies does exist. For example, a person- al letter from an executive of a listed company said, ". my personal finances are such that I have found it possible and desirable (regardless Of market price) to purchase Sim- mons stock during December for each of the last several years." THE INSIDERS' TRANSACTIONS AND THE LAGGED STOCK PRICE 78 EXHIBIT 5-5 FREQUENCY DISTRIBUTION OF COMPANIES BY THE COEFFICIENT OF CORRELATION BETWEEN Coefficient Between Correlation NO. of Companies in Which the Insiders' Transactions Correlated with Their Stock's Lagged Price Performance +.8000 - +.8999 l +.7000 - +.7999 1 +.6000 - +.6999 7 +.5000 - +.5999 4 +.4000 - +.4999 7 +.3000 - +.3999 10 +.2000 - +.2999 5 +.1000 - +.l999 16 .0000 - +.0999 8 -.0001 - -.0999 10 -.1000 - -.l999 5 -.2000 - -.2999 2 -.3000 - -.3999 3 -.4000 - -.4999 1 -.5000 - -.5999 0 -.6000 - -.6999 1 Unclassified 2 Total no. of companies with 6 or more monthly Observations 83 79 This exhibit prompts two comments. First, a linear model, at least in the past, was a fairly good description of the forecasting properties of the transactions in approximately 19 executive groups in the sample. Second, the evidence seems to indicate that company-by-company experience is vastly different from the average correlation mentioned above in the discussion of Exhibit 5-1. The correlation results alone do not adequately describe the relationship between the proportion of buyers over sellers, or vice versa, and the rate of advance or decline of stock price relative to the market six months later. The average regression coefficient of +1.24 per cent per net transactor for all companies in the sample provides a clue to the rate of the relative percentage of change among the companies. Another clue is provided by comparison of the average performance of the stocks bought with that of the stocks sold, by company. The problem implicit in the rest of this chapter is: what explains the low average cor- relation? Spperiority of insider-bought stgcks over insider-sold stocks as fgrecasters A second test is desirable to substantiate the con- clusion that the forecasting hypothesis is tenable. The 80 model assumes the hypothesis that the population of insider- bought stocks out-performs the population of insider-sold stocks. A test Of the difference between the two sampled means will corroborate the results of the binomial test on the signs Of the 48 monthly correlations Of composite re- sults. A significant difference between the two sampled means will allow the inference that the two population means are different from one another. The mean performance of the stocks in the aggregate is presented in Exhibit 5-6, classified by the index denoting the intensity of the monthly trading activity. The exhibit shows that the 1,507 observations were divided about equally between net buying and net selling.7 It also suggests that brisk monthly trading by any executive group is relatively infrequent. In fact, 1,135 monthly observations Show only one net buyer or one net seller. This particular sub-sample represents 75 per cent of the total. It will be shown later in the section on forecasting direction that this sample is 7During 1957-1960 the market experienced two stock- price cycles and the economy went through one business cycle. Despite two periods of falling stock prices the market was rising substantially over the period studied. In fact, the SEC Stock Market Index advanced about 50 per cent from January, 1957 to June, 1961. It is interesting to note that insiders were selling on balance during this period. In terms of observations the balance is 776 to 731, but in 'terms of net transactors the balance is 1,111 to 984. 81 EXHIBIT 5-6 RELATIVE PERFORMANCE OF STOCKS PURCHASED AND SOLD BY THE INSIDERS-MONTHLY-ACTIVITY INDEX Net NO. of Buyers (+) No. of Average Performance or Sellers (-) Monthly Relative (6 Mos. Lag) Per Stock per Month Observations to the Market (Col. 1) (Col. 2) (Col. 3) +7 3 - 1.45% +6 3 +1l.37% +5 6 + 2.63% +4 12 +15.26% +3 22 + 5.94% +2 116 + 5.20% +1 569 + 4.00% Total no. of net buying Observations 731 Mean performance Of insider-bought stocks, weighted by column (1) + 4.94% Mean performance of insider-bought stocks, not weighted by column (1) + 4.45% 82 EXHIBIT 5-6--gontinued Net NO. of Buyers (+) No. of Average Performance or Sellers (-) Monthly, Relative (6 Mos. Lag) Per Stock per Month Observations to the Market (Col. 1) (Col. 2) (Col. 3) -l 566 + 2.17% -2 137 + 3.37% -3 46 - 8.01% -4 15 + 4.49% -5 5 - 7.30% -6 3 -l6.50% -7 3 - 8.43% -9 l -19.00% Total no. of net selling Observations 776 Mean performance of insider-sold stocks, weighted by column (1) + .59% Mean performance of insider-sold stocks, not weighted by column (1) + 1.62% Difference between sampled means, a weighted 4.35% Difference between sampled means, a not weighted 2.83% aHighly significant at the .01 level. 83 the least significant and, by itself, would not allow us to reject the hypothesis of independence at the .01 level be- tween the transactions and price performance variables. Method of averaging.--The measures of performance in Exhibit 5-6 are arithmetical averages. The cumulative effect of arithmetically averaging percentages is a slight bias toward the positive side. However, this limitation does not weaken the validity (or significance) of the cor- relation or chi-square results in the two major divisions of this chapter, since these tests deal with individual cases, not averages of grouped Observations through time.9 DeSpite this Operational defect, average-performance measures have some utility. Any bias will appear in both samples of stocks purchased and sold.10 A comparison 8This finding and the two means for the one-net- buyer and one-net-seller sub-samples of +4.0 and +2.2 re- spectively suggest that the average intercept for the total sample may be approximately +3.1 per cent. In other words, the lOO-stock sample may have out-performed the market dur- ing 1957-1960 by as much as 3.1 per cent. 9This bias partially explains why the average inter- cept for a given company tends to be greater than zero in the model. 10 . . . . . This limitation suggests that a mean of zero is an inappropriate standard for the average-relative performance of the market as a whole in this study. Thus, statements that a given average semi-annual performance is better or poorer than the market in general may be suspect. Experi- mentation, however, does not reveal a biased result and does 84 between the two percentages of mean performance in both the aggregate and company data does provide evidence of a fore- casting difference between insider net buying or net selling. T g weighting technigpes.--One weighting scheme fol- lows the linear model used in the regression analysis. Thus, each observation is weighted by the net number of buyers or sellers. If there are four more buyers than sel- lers of a particular stock during a month, this stock's per- formance would be weighted in the over-all average four times as much as another observation in which the index of the insiders' transactions denotes one net buyer. This is equivalent to investing four times as much money in the former. The computation of the average performance of the insider-sold stocks could be similarly weighted in accor- dance with the net number of sellers in each Observation. This scheme results in an average performance for the insid- er-bought stocks of +4.94 per cent, compared to the average of the insider-sold stocks of +0.59 per cent. The differ- ence between the two sample means is 4.35 per cent and highly significant at the .01 level. A second weighting method is consistent with the g riot support this suspicion. Consequently, these statements Inay be reasonable. 85 chi-square tests and the binomial model which will consider each Observation equally. Thus, if an equal amount of money is invested in each observation regardless of the number Of net buyers or net sellers that month, the performance of these two samples relative to the change in the market as a whole would have been +4.45 and +1.62 per cent respectively. The difference between the two sample means, weighting each Observation equally, is 2.83 per cent and also highly signi- ficant at the .01 level.11 Both the weighted and unweighted averages of stocks bought and sold resulted in sample estimates that were sig- nificantly different at very high levels of confidence. The smaller difference of 2.83 per cent between the sampled means of insider-bought and insider-sold stocks could be attributed to chance less than one time in a hundred. (The difference of 4.35 per cent is, of course, even more signi- ficant.) Thus we can, with reasonable confidence, reject the hypothesis that the true means of these two populations are equal. Since we do not have a representative standard for -the average semi-annual record of the stock market's per- formance relative to itself (which theoretically should be 11See Appendix H for the formulas used. 86 zero), we cannot infer that the insider-bought stocks showed a semi-annual record on the average which was 4.94 per cent better than the "average" common stock. Nor can we con- clude that the insider-sold stocks achieved a semi-annual result which was 0.59 per cent better, on the average, than the market. But we can state that in terms of the model the insider-sold stocks achieved a six-month result that was on the average poorer than the insider-bought stocks by 4.35 per cent. If the 1,135 less important Observations--transac- tions by one net buyer or one net seller--were eliminated, the results would suggest that, on the average, Observations showing more material net buying or net selling forecast the relative advance or decline of the stock price lagged six months. Thus, it would appear that the sub-hypothesis that insiders forecast through their sales transactions the 'price decline of the stock relative to the market is valid only for observations showing three or more net sellers during the month. Exhibit 5-7 has a bearing on this point. It presents estimates of the amount of dispersion in the data.on the basis of seven defined random samples during 1957-1960. Measures of dispersion in aggregate data.--Despite 87 som.a mamuoa o.m ¢.ma m.ml hm on or flu m.m o.mH o.m| we ml m.a o.Nm v.m+ bma mu m. m.mH N.m+ mom HI n. m.mH o.v+ mom H+ O.H m.mH m.m+ @HH m+ m.m m.ma o.m+ we n+ ou m+ ”MI? x m Asm u we newumwwm umxums may Aruaoz_mrp magnum cum o 0 cu O>Hpmamm xooum m OH .AIV muwhsm 2 SD. M OOGMEHOM Hmm ASV H0>O mHOHHOm .HO was no acauma>mo . HODGCMIHEwm ONHm A+v mumaawm Hm>o muowsm .HOHHM mo ucwo Hum OHQEMm mo coflpnomoum usev pumpamum was and: mmamfimm cm>mm oomfllhmma UZHMDQ MHZODZflM Qmeumqmm mMUOBm 202200 mm ZH mZOHBUp< oumaaom no Hohnn um: Tao mo mommo mma.a mo mHmEmmIQsm v.vm m om v.@WI Towpnmouwm mmmoosm orb HMh Hmaoe mom pom COGHHOOQ mmm hmg COOCC>C¢ nmcoflum>uomno sarpaoe som.H no wamemm Hmnoa mcflaawm mcflmsm Ahaamswod xoowm mmmuswuumm msu mmEHB mo .02 onev mmwoosm umz Dwxumz map Op O>Humamm mOflocwskum CmHmEmm mo :oHumHHome Hamnnm>o mo mommu msusoz xflm Commmq mo .02 one wocmEHOMHmm Hmsuod mmZOHBU¢m24MB .mmeHmZH m9¢m0mm00 ZH moZdSMOmmmleOHBUmMHQ m0 MBADmmm m9¢0mm00¢ mlm BHmHmNm 96 .HmSEH Ho.mnp um Ammunmncmnm p020 Hmsma go. was up remunmncmAmn "mum3 mmflonmsvmum mmmsp ummp mundvmlflso m mGHmD .umMHmE map on m>flpmamn GOHHOOHHC mo mpmmomuom HsmmmmOOSm mum mcoflpm>ummno mOGHEImSGHE no mDHmImSHm Mano .mscafilmscfle paw .msHm ImsGHE .mscHEImsam .msamlmsam mum mcoflumcflnfioo OHQHmmom “Dom one .mscfle Ho msam moaned on umse mosam> 03¢ map mo Oman map .msna .uwpma mCuGoE xflm pmxumfi esp on O>HOMHOH monmfi luomumm OOHHQ m.MOOum Oman mcfluocop ODHC> map Cam £#:OE map mcHHSC Mooum cm>flm m :H hufl> Ifipom .mnmCHmCH map mCflNHumEEST m5am> map "mosam> mo Mama m mm: coflum>ummno Comma H.mm m.mm hTOMI mmmucooumm mmmoosm firm .mldfl H309 NHH em Cmaflaomn mm moa Cmocm>p< ancofi m CH mumaamm Ho mummsn um: whoa no 03¢ mo mommo mum mo OHQEMmIQDm mcflaamm maflhsm Ahaam5po¢ xooum mmmucmonmm may moEHB mo .02 Trev mmmOODm uoz poxnmz may on O>Humamm mmwocwnvmum Cmamamm mo coaumfluommn HHMIHT>O mo mommu mnucoz xflm Commmq mo .02 one mocmEuomumm Hmsuum Concepcoollmlm BHmmem 97 direction Of the price change than the record Of any single transactor. Exhibit 5-9 shows that the 1,507 Observations in the sample are composed of 731 net buying and 776 net selling observations. This means that on 731 occasions insiders in a given month bought more frequently than they sold and, therefore, in terms of the general model, their buying would forecast a rise in the price of that stock six months later. In fact, six months later, on 427 occasions the stodks had gone up and on 304 they had gone down, relative to the mar- ket. This is a 58.4 percentage of success, or about four successes out of seven trials. On 776 occasions more insid- ers were selling, thereby forecasting a decline in their company's stock price relative to the general market move- ment. The stocks actually declined only 393 times for a percentage of success of slightly more than 50. While the transactions summarized in this study were statistically significant predictors Of direction, the net selling results are not much better than coin-tossing. Thus, selling by corporate insiders may not signal worse-than-average perfor- mance over the next six months. Prpfit maximization.--In the previous chapter were listed possible motivations consistent with rational buying 98 or selling by an individual executive but not consistent with the interpretation that the executive under- or over- valued the shares. We reasoned that by aggregating the transactions of all the executives during a month, a clearer indication of profit maximization consistent with the logic of the hypothesis would emerge. In other words, the greater the excess of buyers over sellers or sellers over buyers, the more likely is the forecasting deduction a correct inter- pretation. This notion that we can identify and eliminate the "misleading" transactions can be tested by the same tech- nique. This filtering operation reduces the size Of the sample by 75 per cent, as only 372 out of the 1,507 Observa- tions showed either two or more net buyers or net sellers. The remaining 1,135 observations summarize monthly transac- tions in a stock by one net buyer or net seller. A chi- square test on the distribution of these observations was not significant at the .01 level. In a large majority of the 1,135 observations sum- marizing one net buyer or seller, the gross activity is also one buyer or one seller that month. Thus, while the +1 value assigned to the insiders' transactions in a stock dur- ing a month could theoretically summarize three purchasers and two sellers, in fact it usually summarizes one purchaser 99 and no sellers. Similarly, the net selling observations frequently summarize the gross insider activity in the stock as well. During the few months when the index of monthly trading activity for a few companies was relatively large (for example, five net sellers) the number of sellers before subtracting the number of buyers was invariably five or Six. In short, brisk activity by a company's executives during any one month is relatively infrequent, but when it does occur they generally conform to the same pattern--either buying or selling--that month. Exhibit 5-9 also summarizes the more meaningful observations of two or more net buyers or net sellers of a stock during a month in terms of profit maximization, and the distribution is highly significant at the .01 level. It tends to support the hypothesis that insiders forecast the relative direction of their company's stock six months later. The exhibit reveals that 162 of the 731 net buying observations summarize monthly activity of two or more buy- ers than sellers in a stock. Six months later, the relevant stock actually advanced 108 times and declined only 54 times for a two to one ratio or a 66.7 success percentage. And 210 of the 776 net selling observations summarize monthly activity in a stock by two or more sellers than buyers. 100 Only 53.3 per cent of the time did the stock actually de- cline six months later relative to the market, the price going down 112 times and going up 98 times. Several remarks seem warranted. First, the statis- tical significance of the entire sample appears to rest heavily on the successful forecasts of two or more net buy- ers or sellers of a stock, and Observations containing two or more comprise less than 25 per cent of the sampled Obser- vations. Second, the probability of a successful forecast of direction of price change of a stock seems to be related to the number of its net buyers or sellers in a month. .Monthly activity in a stock by at least two such net buyers or sellers improves the average chance of success in fore- casting from about 54.4 per cent to 59.1 per cent. Finally, when two or more insiders have predicted that a stock's price will go up, it is more likely to advance relative to the market than a stock predicted, by two or more insiders, to decline. As mentioned earlier, executives sell their com- ,pany's stock for many reasons other than a conviction that its price is about to decline. One of these reasons is to finance the purchase of additional stock at Option prices. It seems not unreasonable to estimate that much executive selling is done for precisely this reason. The data studied 101 include all sales but exclude purchases on Option, however, which may Obscure the clarity Of the attempted classifica- tion between net buying and net selling; if so, it may well explain why net selling is less successful as a prediction of price change than net buying in the sample. This may be a reasonable explanation if stocks of corporations with stock-Option plans perform better-than-average. More pre- cisely, it is conjectured that stocks of corporations whose executives are actively exercising stock Options are prObab- ly better than average performers relative to the market. An additional explanation of the relatively inferior perfor- mance of the selling Observations is simply that executives may be liquidating holdings for various reasons consistent with their expectations and risk preferences, and their plans to maximize their income after taxes. ggmpany experienca and trading patterps The remainder of this chapter sets aside the conclu- sion (based on empirical fact regarding net purchases, at least) supporting the interpretation of profit maximization in monthly operations by two or more net transactors. Since there were relatively few of these monthly Observations in each company, an advantage will accrue by subjecting the total sample of 1,507 Observations to analysis rather than 102 the more significant sub-sample of 372 Observations. There has been some evidence that trading intensity in a stock during a month is a useful discriminator of forecast- ing performance. This segment of the study will compare the trading frequencies and patterns among the companies for possible clues to improve performance. One purpose of this sub-section is to attempt to detect the companies in which executives are allowed to operate largely with personal profit maximization in mind. In order to filter out those companies that might discourage the practice, activity is defined as the number of monthly observations out of a possible 48, ignoring the number of net buyers or sellers in any one month. For this reason, the summaries of direction-perfor- mance by company in the text of this study are the details of the total sample of 1,507 observations shown earlier in Exhibit 5-9, while Appendix J details by company the 820 successful forecasts out of 1,507 Observations. Their aggregate percentage Of success is 54.5 per cent. Exhibit 5-10 summarizes the distribution of the percentage of success 14Binomial probabilities will be the principal lnethod of analysis and the larger the number of observations ,per company, the closer is the binomial distribution (among the companies) an approximation of a normal distribution. Selective samples on the basis of the conclusions in this chapter are tested in the next chapter. 103 EXHIBIT 5-10 DISTRIBUTION OF PERCENTAGE OF SUCCESS AMONG THE COMPANIES Percentage Range of the Number of No. of Successes to Com anies Total Forecasts per Companya p Greater than 70% 18 31% to 69% 73 Less than 30% _1, Total |l‘° oo aIf you extend the two bordering percentage inter- vals by 5 per cent each, the distribution from top to bottom is 27, 60, and 11 companies, and the conclusions are the same. 104 among the companies. It shows that the executive groups of most companies experience roughly average success, and that consistently successful forecasts are more than twice as likely as consistently unsuccessful observations.15 However, a listing Of percentages of successful prediction has limited utility if what is wanted is a test of the hypothesis that frequency of trading is related to successful prediction. It does not show the magnitude of the number of successes in each company. A detailed rank- ing Of the 98 companies by the success percentages is limit- ed because the total number of Observations in each case varies from 1 to 39 among the companies. Since the total number of successful and unsuccess- ful observations per company varies, a more useful technique may be the estimated probability Of forecasting as well or better. For example, assume that the executives of three companies have forecasting records of monthly successes out 15This two-to-one ratio of success is based on both purchase and sale Observations and differs from the same ratio shown earlier in Exhibits 5-8 and 5-9, which were per- formance summaries of purchases only. Even these two earli- er exhibits differ in sample construction. In Exhibit 5-8, this ratio summarizes the average percentage of performance of the total sample of 731 observations of net purchases distributed among 94 companies, whereas in Exhibit 5-9, this significantly successful aggregate ratio of direction-per- formance emerged in the sample of 162 Observations of net purchases by two or more buyers than sellers. 105 of monthly transactions as follows: 1 out of l; 4 out of 5; and 14 out of 20. These success percentages of 100, 80, and 70 result in a ranking which is contrary to the significance of these three performance records. The coin-tossing analo- gy makes it easily seen that the probability of doing as well or better as a matter of chance would be .50, .19, and .06 respectively. In other words, the ranking Of the three companies according to the significance of their forecasting records is exactly the reverse of their ranking by success percentage. If there is no relationship between insiders' net buying and selling observations and relative price change lagged six months, the probability should be one-half that the value of the stock will move in the direction indicated by the transactions. If the stock price and the transac- tions are independent, the number Of successful net buying and net selling observations in a company-series will have a binomial distribution with p equal to the total number of Observations and p equal to one-half. In other words, these assumptions of a null hypothesis suggest that an appropriate description of achieved success for each company is the estimated probability of forecasting as well or better. The calculation of these probabilities is facilitated by using the normal-distribution approximation of the binomial 106 distribution. The companies are ranked by their binomial probabil- ities in Appendix J, and the sampling frequencies are summar- ized in Exhibit 5-11. The probabilities derived from the binomial tables are approximate because the observations are not wholly independent. There is some dependence in them because of the Six-month overlap, but this dependence does not increase the number of Observed successes.l6 The effect of the dependence is to make the true prObability smaller in some cases and larger in others than the calculated pro— bability, depending on the trading behavior of executives, who might always buy or always sell. In short, the proba- bilities are rough approximations. However, the method is appropriate, and it does give some notion of the magnitude of the number of successes over what would be expected from chance alone. The first and last ranked companies in Appendix J will serve to illustrate the summary in Exhibit 5-11. The executives Of American Water Works, ranked first, engaged in transactions with their company's stock as net buyers or net sellers in 23 months. Nineteen of their 23 forecasts proved 16A Success is a pair of matched Signs per observa- tion. 107 4m lam lam ImIM mm 23.8. Inn. I43. INN. Inn. .IMHI oo.H I om. m o H 0 OH om. I om. m w s 0 HH om. I on. m o m m OH on. I om. H H m m o om. I om. m H m H b om. I ov. H m H H m os. I om. m s m H OH om. I om. N N O O NH om. I 0H. m e H m mH OH. I oo. meHV AvHIoHC mmIHm O>HHOMCH O>Hu0¢ HomIvHv AO>HHOM mHm>HumHmm mHmumnmCOS O>Huo¢ Mum> ouoomm mchmmomHom HOBOH mmHHuumso CH2 Hmmmb mmaocwsomnm Hmuawm no Hmsvm .Hmuoe no mo momma SRHHHnmnonm m . Ame HHHmmom n mo #50 mummomuom MHCBCOS mo Ozv oowwEHumm one mpH>HpO¢ mcHomuB hHsuGOS mo TOHHHHMDO >3 coHuanuumHQ mcHHmEmm Hmuoa OCH CH mchmmEoo mo Hmnfisz one QMOUmm OZHBmfiommOm .mmmnHmZH MHmmB Mm mmH24mZOU m0 mmmzbz HEB ho ZOHBDmHMBMHQ NUZMDOMMM d HHIm BHmmem 108 correct. Nineteen times the stock either advanced or de- clined six months after net purchases or sales, respective- ly. The estimated probability of being as successful or more successful than that is .0013. Thus, this first ranked company is one of 15 executive groups whose chance probabil- ity of achieving such a successful forecasting record is .l or less. Ignoring the intensity of the monthly activity (the number of buyers less sellers), American Water Works, with insider trading in 23 out of the 48 months, is one of the eight successful companies which is also classified as very active. The executives of Patino Mines and Enterprises traded in their company's stock in only one month, as buy- ers. Six months later the stock had declined relative to the market. The probability of doing as well or better than zero successes is 1.00, and this company is one of 12 com- panies in the Total Frequencies column and one of four in the Relatively Inactive Quartile column in the .90 - 1.00 probability interval. Tptal sample.--The distribution of frequencies in the Total Frequencies column Of Exhibit 5-11 is not statisti- cally significant.17 The expected (frequency) distribution 17This lack Of statistical significance can be 109 for each probability interval is approximately ten compan- ies.18 Since the actual frequencies are not significantly distributed in support of a forecasting hypothesis, this is inconsistent with earlier evidence (based on the same data of 820 successes out of 1,507 forecasts).19 This evidence suggests that executive groups tend to be either very successful or very unsuccessful; fewer executive groups than expected enjoy "average forecasting success."20 This apparently contradictory evidence suggests influences not yet empirically detected in this study. For example, if some corporations tolerate stock trading by intuitively seen in the distribution of 49 groups with re- cords better than .5 and another 49 groups performing worse than this. 18 . In a random sample of 98 forecasters, each haVing the same probability of a successful forecast, we could ex- pect that approximately 9.8 of the 98 observed frequencies Of successful forecasts would be such that the probability of achieving such a frequency or greater is less than .10. Similarly another 9.8 companies could be expected to have unsuccessful forecasting records such that the probability of doing better is .9 or more. 9Since the chi-square test on the aggregate data (in Exhibit 5-9) was Significant at the .01 level, these negative results were not expected. This inconsistency will be explained later in the "weighting," i.e. the number of observations in each company. O"Average forecasting success" is here defined as the chance probability Of between .4 and .6 of equalling the insiders' forecasting record of p_successes out of so many forecasts. 110 their executives, the frequency of trading in these stocks Should be higher than in those of other corporations. The sub-hypothesis is that actively trading executive groups forecast the market-price Of their company's common stock. In other words, are the companies in which the insiders are infrequent traders in that stock a diluting factor? The distribution of frequencies tends to support the sub—hypothesis that insiders in some companies have fore- casting records better than what they could have achieved by pure chance and, at the same time, the least successful re- cords are worse than could reasonably be attributed to chance. A corporate policy that encouraged its executives to trade in its stock might shed some light on the superior records, but a policy forbidding buying or selling or both does not logically explain the exceedingly poor forecasting records, although it might explain the relative inactivity in trading in a company's stock. However, it is still reasonable to expect the forecasting records of companies ‘whose executives' trading is influenced by policy consider- ations to be randomly distributed among the probability intervals, not concentrated in the higher probability ran- ges. Activity classification.--The quartile distribution 111 of the total sample by the activity in the stock during the four years is both suggestive and puzzling. The frequency distribution in the very active (or upper) quartile is the only activity classification found to be significant at the .01 level using a chi—square goodness-of—fit test. (The expected distribution is 2.5 companies for each probability interval.) This would suggest that the statistical signifi- cance of the chi-square test on the data in Exhibit 5-9 (the distribution of the 820 successes out of the 1,507 allowed us to reject the statement that insiders' transactions have no predictive value) rests on performance in relatively few companies in which the executives were active traders. However, the record of the very active executive groups (in the upper quartile) is marred by the least successful groups in the distribution. Since six of the twelve least success- ful groups in the total sample were in the very active quar- tile, any judgment on the significance of activity in a group must be reserved. To generalize about an activity classification that does not distinguish between net buying and net selling nor the pattern of the trading activity is limited. In other words, it appears that monthly activity of two or more net buyers or sellers per observation among .aIT,the companies explains the significance of the relation- ship in the total sample better than the activity classifi- 112 cation of the number of Observations per company over the four years, ignoring the intensity of the monthly activity as shown earlier in Exhibit 5-9. As noted above, it is very difficult, on the basis of possible corporate sanctions against insider trading, to explain why as many as six of the 25 companies in the very active quartile performed so poorly. Since six companies are -2.33 standard deviations away from the expected number of companies in this interval, it is probably not explainable by chance. Similarly, the eight best records in the quartile (+3.67 standard deviations) are not likely to be due to chance. One may reasonably explain the superior forecasting records on the basis of the logic of the forecasting deduc- tion. But if the forecasting hypothesis is tenable, how does one explain significantly inferior performances? It is possible to rationalize a given company's ex- perience, but such conjectures are not empirically founded. For example, one Of the six companies in the upper quartile whose executives were least successful in forecasting direc— tion was Federated Department Stores. The executives of this company were selling on balance in thirty-one months and buying in one month for a total of thirty-two observa- tions during 1957-1960. Since the price of this stock was generally rising faster than the market as a whole in almost 113 every month, the executives were correct in only four of the thirty-two forecasts. One can forecast that the executives of Federated will continue in the future to be predominantly sellers of their company's stock. This statement is akin to the conjecture that these transactions are independent of the future price, because one cannot logically forecast the future price of this stock relative to the market. If this stock were included in future study and it continued to out- perform the market, it is likely that it would appear un- successful. If the stock's price continually performed poorer than the market, the binomial probability would sug- gest that the insiders were successful. However, in this case, if the hypothesis of independence is tenable, a suc- cessful or unsuccessful label would be inappropriate, as the variables are apparently unrelated. If the forecast of continued net selling by the executives of Federated is plausible, it is based on (conjectures of) the relatively high concentration of stock ownership held by management and their families, and the company's stock option plan. If we decide to eliminate the unsuccessful observa- tions in Federated from the sample, it would seem that the successful observations in Sears Roebuck (and American Water Works) should also be eliminated. Since the executives of Sears were predominantly buyers of their company's stock, 114 and the stock generally out-performed the market, this suc- cessful performance may be very similar to Federated's un- successful performance, i.e., an independent result. If we exclude unsuccessful Observations and include successful ones, we will be guilty of statistical manipulation and gen- eralization will be limited. In the next chapter, a test is performed on selected companies from the random sample. The companies that are relatively inactive and those that are active but predominantly buying or selling will be elimi- nated from the sample. Thus, companies like Patino Mines, Federated, and Sears will be excluded from the tests. Activity patterns.--This discussion serves to intro- duce a classification system for the executives' trading patterns over the four years that takes into account the intensity of the monthly activity, its sign, and the fre- quency of the monthly observations. The four patterns in the time series Of the insiders' transactions are classified by visual inspection of the data as either cyclical, irreg- ular (or erratic), predominantly buying, or predominantly selling. (A company time series of the insiders-monthly- activity indices is classified as cyclical if there are consecutive months of net buying and consecutive months of net selling.) 115 The total sample summarized in Exhibit 5-ll suggests that the records of 15 executive groups were significant at the .10 level, and those of 12 groups were equally unsuccess- ful. The most and least successful companies are listed in Exhibits 5-12 and 5-13, along with information on the trading pattern in each company. Perhaps the cyclical pattern is consistently related to successful prediction. While the hypothesis supporting this notion is formulated and tested in the next chapter, a clue that this pattern-selection approach will fail to lead to more positive results is indicated in Exhibit 5-13, where four of the 12 poor performers also had cyclical transactions series. Executives of American Water Works and Sears appeared to be consistent buyers, and the Officials of White Motor and General Dynamics were consistent sellers. While these 1For example, the executives of International Tele- phone were buying in 1957-1958 and selling in 1959-1960. Since prices were higher in the second period, they probably realized profits, but the measures of the variables in terms of the model indicate poor timing in general. There are some elements of success, however, in this executive group. For example, the month of the most active buying (four net buyers) occurred in December, 1958, and led a six-month rise of 19.38 per cent greater than the market. On the other hand, the trough in the transactions series (three net sel- lers) occurred in April, 1960, and unsuccessfully led a 4.64 per cent rise in the stock relative to the market six months later. 116 .HMHDOOHHH Ho OHHMHHO H mm “AcoHuommmsH HCDmH> >3 om>uomno mmv HCOHH0%O n O u n m can “OGHHHmm thsmcHEoomum H mm umcHth mHunmcHEoooum ms coflpmuflmammmflo anmpu m mm o O m UHupomHm Cam mmw mmmcmx mH O OH O OH mOHOHm cmHmsz HmmHO pmuHCD OH mm OH m HH EscHumHm Com CHOO OMOHHOEN Busom mH O NH > m mmHuHHHHD mmw omumoHHOmcou NH O NN HH HH Ammxmav HHO GEOHHOEN Hmnmcow HH 0 NN HH HH mHmHumez_ocm HMOHEOSO GOHGD OH mu b O h OHHHUOHM Com mmo mHosHHHH m3OH m o om mH s coaumuomuoo anmo Hmumamo m mm mm Nm m mOHEmoma Hmuoomw n mm mN mN O .OO mono: manz o O OH m o MOHHOE< mo uHfiomHm OOHHGD m 0 mm ON mH OHHHOOHm Hmumcmw v mm ON m HN .OO Com Mosnoom mummm m U mN OH mH mwcHHHH< GMOHHOEN N mm mN s OH mxuoz Hmpmz GMOHMOEN H Hmuoa OCHHHom mcHhsm cumppmm #02 mo msucoz mo .02 mo muooncH Maud >pn>fluo< 2mmfifi mmmz mZOHBO¢m24MB .mmmnHmZH mmomz mmHZNmSOU m0 BmHH NHIm BHmHmNm 117 H mm “AsoHpommmsH HODmH> ma oo>uomno may HMOHHomo u U u .HCHOOOHHH no OHumHHm n m can “OOHHHmm thcmcHfioomnm H mm umchsn mecmcHEoomum 0M GOHHMOHMHmmmHU cumuumm mm H O H wmcHz ooHpmm mm mm N O N GOHHmnomHoo CCMHMO>O mm mm Nm Hm H monoum Homauummon omumuooom mm 0 mN mH m HomEmHDOm conHowum Hmuocow mm o om MH s pompouam mounuuoz Om mm ON NN N monoum OOmcHQom GOmmmuO mm 0 ON mH mH .HOB Cam .HOB HmooHumcHoucH No 0 OH n m .muoo monsuOHm pcsofimumm Hm mm O H m .06 pumo mansmam .m In om mm am am m sumac sHumnsAx mm mm mm 0 ON .00 A.m .qv ppmunmpm mm m m m o .oo mommMHoz CCOHHoEm hm HOHOB mcHHHom mchsm ouovumm mo unmonoH Manx uwz mo mango: mo .02 suH>Huo¢ ZMHBBdm GZHQNMB mHmmE m0 ZOHBmHMUme N QZN ZOHBUHMHQ OZHEmdvmmom HDmmmmOUDmZD Nmm> WMMS mZOHBOm mouoom OOQMEHOHHON msHpmmoouom mEBZOS 03B BmEHH B4 ZH mEHHHHm GEE mmHMDH BHZ HEOS m0 HHEEB EBH3 HHBmHB mm HEB HO mmDOEO H>HBDUHXH ZH>Hm HEB Hlo BHHHEXH 133 reasonable success in forecasting direction, though not in the linear relationship originally hypothesized. The execu- tives of the remaining three companies did not seem to fore- cast successfully at all. However, there is an unusually consistent aspect of successful prediction if we consider the individual monthly forecasts that formed the peaks and troughs of the relevant transactions series. Exhibit 6-2 shows these observations, their indices of activity and per- formance, and other pertinent data. Multiple (or identical) "peaks" or "troughs," as defined by the value of the insid- ers' transactions index, are all listed in the exhibit in chronological order. First, eight out of the nine forecasts of advance (89 per cent) actually advanced relative to the market. The only unsuccessful forecast, in fact, can be explained by external events which counteracted whatever inside informa- tion the executives of General Electric acted upon in Sep- tember, 1960. It is unlikely that the five net buyers of this stock foresaw the price-fixing scandal Of late 1960, within the subsequent six-month period. General Electric and others were convicted of violating the antitrust laws in the sale Of heavy electrical equipment; General Electric stock declined from a high of 80 to a low of 60 after the scandal broke. AHCmmmOOOCmCCV Humowuom on OHSHHmm C numMOOCOM Hommmmoosm 134 mm.s I + oo.mH mm.mH mI coma .sHsn mm ms.HHI + ma.m sm.m mI mmmH ..>oz mm mm. + + sm.O ms.O O+ smmH ..umz mm so.OH+ + 00.65 oo.Ho OI coma ..ms¢ em ms.mm+ I mm.mm ms.6O 6+ mmmH ..omn hm 56.O + I om.oO oo.OO mI coma ..um¢ 0O mO.mH+ + ms.6s ms.oo O+ mmmH ..oma 0O mH.omI I mo.Hm om.mp mI oomH ..umz mm mm.OH+ + mH.sm mm.mm O+ smmfl ..nmm mm sm.omI + om.so mm.ms m+ oemH ..Cmm om mm.mH+ I ma.mm om.os m+ mmmH ..mom om mO.o I I om.6s om.mm RI mmmH ..nmE om m6.H + I sm.Hp ms.O6 m+ smmH ..Cmm om mm.m I + 6m.sm sO.mm mI mmmH ..uuo mH mo.mH+ + ms.mm oo.mm m+ smmH ..uuo ma mm.mHI I sm.om ms.Hm SI mmmH ..CCO O mm.mm+ + m6.0mw oo.me O+ mmmH ..AC< O uoxumz OCH OCH COCOS HOHMH EHCH>Huo< ou O>HHMHOm XHm mCHHCQ mCuCOZ XHm mxmom pd MMNHMMHWW MOON OCm COCOS HIV mmOH Ho umxumz mo . A+v CHCO X COHuOOHHQ HIV OCHHHOm Ho OOHHN ooumsnod H+v mamsm xmmm mEBZOZ XHm EH>O mEHHHHm BHZ HEN mEHMDH BHZ HEOE E0 HHEEB EBHS mEUOBm EH OZHHHHm HEN OZHBEH HO mEde BHEBZOE BE mBmdonOH HO HUZESEOHEHH Nlo BHHHEXH 135 Six of the eight forecasts of decline were success- ful in terms of the model. One of the two unsuccessful per- formances did predict the general market trend, but was not similarly successful in forecasting the value of its own stock--it advanced a half of a point during the decline in the market. In general, this method of selection, based on the cyclical nature of the executive trading in some com— panies, gave highly successful results. Fourteen successes out of 17 forecasts (82.3 per cent) is strong evidence for the validity of the alternative hypothesis introduced in this chapter—-that insider trading predicts market trends in companies which tolerate (and possibly encourage) trading by executives. CHAPTER VII CONCLUSIONS AND IMPLICATIONS The purpose of this study was to ascertain the pre- dictive property of a phenomenon which seemed reasonably likely to forecast stock price movements. A hypothesis that insiders forecast through their transactions the market price of their own company's common stock was tested by studying insider trading in a random sample of corporations listed on the New York Stock Exchange. There were 100 com- mon stocks randomly drawn from a population of approximate- ly 1,065 such stocks during 1957-1960. Of the 100, only 98 reported instances of insider trading. Having defined the scope of the investigation, brief summaries of empirical and legal background material were provided. A search of the literature did not discover a similar statistical study. A model was derived to test the hypothesis. It was contended that the behavioral assumption that executives attempt to maximize profits is reasonable for company offi- cials as an aggregated class. Rules were established for collecting and summarizing the raw data, and the tests were 136 137 limited to short-run forecasts defined as six months. The method chosen was to construct an insiders' transactions index for each company for each month, and to analyze the monthly values in this index (other than zero) with an index of that stock's performance during the subsequent six-month period. The index of monthly company activity was the number of its own buyers minus the number Of its own sellers. The index of a stock's performance was its percentage of change minus the percentage of change in the Securities and Exchange Commission's composite stock price index over six- month intervals from the date of the insider buying or selling. Thus the hypothesis, in general, tested whether inside-buying led a six-month advance greater than the general market movement, and whether inside-selling led a six-month decline relative to the market. There were 1,507 monthly Observations divided about equally between buying and selling transactions in the stocks of the 98 companies. While the average frequency of trading activity was 16 observations per company or four per year, the activity among the companies ranged from one to 39 out of the 48 months studied. It appears that the greater the disproportion of monthly buyers over sellers in a stock, the more likely it is to out-perform or under-perform the market in the ensuing 138 six-month period, and the rates of advance or decline, relative to the market, tend to increase with this monthly proportion of net transactors. However, the low average correlation of +0.13 in 48 monthly composite tests suggests a weak relationship. On the other hand, the Sign of the correlation coefficient is probably positive. A Sign test suggests that if there were no relationship between the two variables (indices of insider-monthly-activity and a stock's performance relative to the market during the next six months), the chance is one in 100,000 that there would be this many or more positive correlations. The analysis of the performance of insider-bought and insider-sold stocks suggests that insiders' buying and selling transactions are associated with the respective advances and declines in the stock-prices, as related to the market averages over six-month intervals. A company's stock bought by two or more of its executives in a month advanced relative to the market two out of three times on the average. Insider-sold stock declined below the market average more frequently than otherwise, but the ratio Of successful forecasts to the total was not as strong as that of the insider-bought stocks. The insider-sold stocks had a poorer record of successful forecasting than the insider- 139 bought stocks by 4.35 per cent on the average. The possi- bility that these average differences were due to chance alone is much less than one in a hundred. Summary pf Statistical Tests 1. In order to test the hypothesis that insiders forecast through their transactions the market price of their own company's common stock, the statistical null hypothesis of no relationship was rejected at the .00001 level. 2. The average performance of the stocks forecast to increase was +4.45 per cent greater than the average common stock equities listed on the New York Stock Exchange. The average performance of those stocks forecast to decline was also +1.62 per cent better-than-average. Statistically, these two means from large samples were significantly dif- ferent at the .01 level. 3. Despite the positive means of the stocks fore- cast tO decline, a chi-square test found that the net buying and selling transactions were significant predictors of direction. 4. An individual insider's transaction in a month not reinforced by another executive's similar transaction in the stock that month was not a good predictor at the .01 140 level of significance. 5. A stock sold by executives is more likely to decline than advance relative to the market. However, the probability of a stock advancing following net purchases is greater than the probability of a stock declining following net sales. 6. Stocks bought and sold by three or more insiders in a month had mean semi-annual performance records and estimates of the standard error of the mean percentages that were significantly consistent with the hypothesis stated in item 1 above. 7. The analysis of the forecasting records of the individual executive groups revealed that some groups were highly successful, and some were equally unsuccessful. There were approximately two and a half very successful individual records for every equally unsuccessful record. 8. Although over-all correlations were not strong, there is a good indication that insider-stock transactions in companies that tolerate such trading by their executives have positive predictive value. It seems probable that a random selection Of companies which encourage or tolerate insider buying and selling would yield a stronger relation- ship between transactions and eventual trend. 141 The results of the statistical tests suggest that insiders do forecast the price of their company's stock relative to the market six months later. Despite convincing evidence supporting the validity of this hypothesis, the variation is nevertheless large and the expected or average gain from insiders' transactions is relatively small. Thus, in the sense that the risk of failure to forecast remains relatively high, the relationship is not strong. The remaining pages contain various suppositions whose validity will require further testing. Implications Macro-egpnpmigs If valid, the hypothesis Of this thesis was presumed to have an economic significance. While this study has produced evidence to support the proposition that insiders' transactions forecast their company's stock price, it has not demonstrated the economic significance of the reported data and hence the economic utility of the foicial Summary publication. The evidence has not led to the rejection of the l"Relationship" refers to both the rate of change as a function of the net number of transactors and, the more general relationship of the direction of change as a function of net buying or selling. 142 plausible theory that the men who run American corporations may be the first to know what is ahead for the nation's publicly-held companies. But there is ample reason to be- lieve that the relationship between their transactions in their company's stock and that stock's price performance in the short run (Six-month future) is not strong. Thus, the likelihood of converting the information on insiders' trans- actions into a useful economic indicator is Sharply reduced. Susiness expectatipns.--The value of the foigial Summary in forecasting may be conceived as a survey of general business expectations. Direct surveys of business expectations have value because they provide good estimates of psychological attitudes. A survey of current capital budgeting estimates would be useful. The expectations of businessmen are in part responsible for the decisions they make. For example, asking the executives of American cor- porations what they think is the most favorable geographical region for locating a new plant would provide some basis for predicting the regions the corporations will select for new plant construction. However, direct surveys of business anticipations have several disadvantages. First, they are costly. Se- cond, they take too long to compile for current forecasting 143 purposes. Third, the responses may be biased. On the other hand, the Official Summary might be considered an indirect survey Of businessmen's expectations. The data published by the SEC is inexpensive and promptly available. The disad- vantage is that expectations, like motivations, must be inferred. This study suggests that inside-buying by two or more executives reveals optimistic expectations. Inside selling by less than three insiders probably does not war- rant an inference that these businessmen have pessimistic expectations. On the other hand, intensive inside selling by three or more insiders probably reveals a negative out- look and a signal for worse-than-average performance of their company's stock in the future. This study suggests that these cases of intensive monthly net buying and selling occur infrequently. Average performange.--Despite the large variation, this dispersion may not present a serious problem for uses and users to which measures of aggregate economic values are essential. Consequently, the average relationship among the stocks studied, although not central to this thesis, pro- vides a concise summary of the linear model. The "average" stock has an intercept of +3.1 per cent and a regression 144 coefficient (or slope) of +1.24 per cent per net transactor. It was estimated that the average performance in observations showing a zero number of net monthly transac- tors is about +3.l per cent relative to the market over six- month periods. This is the average of the least significant sample of 1,135 cases showing either one net buyer or one net seller. Possibly the 3,293 (4,800 minus 1,507) Observa— tions which showed no activity in a stock during a month, if tested, would have performed, on the average, about 3 per cent better than the market as a whole. At first, this estimate of a positive 3 per cent for the average intercept of the sample seemed likely to be the extent of the positive bias. Subsequent calculations, however, suggest that this is probably due to the chance variation of the selected companies among all companies listed on the New York Stock Exchange. The average price of the 100 stocks in the random sample (including the two for which no transactions were Observed) advanced 51.2 per cent from January, 1957 to June, 1961. During this same four and one-half year period, the SEC stock average advanced 48.4 per cent.2 Thus, the 2While it appears that chance led to a sample which out-performed the market, it should be noted that during this four and one-half year period 19 out of 100 stocks were at lower prices at the end Of this bullish period compared 145 selected stocks out-performed the selected index of the mar- ket by 2.8 per cent. This variation appears to explain much of the intercept of +3.1 per cent. Discussion of aggregates.--Focusing attention on the performance of the individual stocks rather than the average performance of monthly samples has provided a clearer indi- cation Of the extent of the variation among stocks. This variation would probably have been obscured in a study of aggregates.3 It is intuitively expected that the variation in a stock's price will generally exceed the variation in an aggregate measure of many such stocks because of the cancelling effect of the variation among the components in an average. In other words, once a stock makes a signifi- cant move, up or down, the extent of its price change is probably greater than that of the market change during the to their prices in January of 1957. These prices are ad- justed for stock-splits and stock dividends, and the fact that 19 per cent of the sample declined in dollar value during a general stock market advance of 48.4 per cent, is convincing evidence that the stock-market is really a mar- ket of stocks. 3While both the average and median correlation were found to be +0.13, the correlation between the 96 monthly average net buying and net selling indices and their stocks' average performance six months later (i.e. aggregates not included in the discussion in the text) was calculated to be higher, or +0.25. 146 same period, and a successful forecast will probably remain successful whether the performance measure is deflated or not. More Often than not, intensive buying or selling by insiders signal these important price movements. Support for this idea can be seen in Appendix K, a comparison of the relative-performance measure (used in this study) with dollar-performance measure in a random sample Of monthly trading by two or more net transactors. In View of the general market behavior Of the period 1957-1960, the result is not surprising for the cases of net buying, but it is somewhat curious in the cases of net sel- ling. For example, Appendix K shows that all seven months in which there was net selling by six or more corporate insiders preceded by Six months dollar declines (not de- flated) as well as percentage declines deflated for the market trend. Admittedly, the sample Of net selling by six or more insiders is small; nevertheless, the evidence suggests that the hypothesis that net monthly selling by many insiders forecasts a downturn in the price of that stock is reasonable. The evidence in Appendix K, showing dollar perfor- mance, does favor net buying. However, when the general advance of the market over the whole period is considered, the advantage of buying is not as apparent in the figures 147 as might have been expected. Generally, the ratios of de- flated successes to total observations closely parallel the ratios of undeflated (or dollar-performance) successes to total observations by transaction-index-number, the ratios tending to get smaller in both models as the intensity of the monthly trading activity diminishes. An aggregate leading indicator.--The possibility that insiders' transactions could be used as an aggregate leading indicator of the market as a whole is still a real one. In order to base an index on this theory, there would have to be more companies which encourage their executives to engage freely in inside trading. If insiders' transac~ tions do lead the market, a more selective sample should have a superior forecasting performance and a moving average Of the monthly transactions (or smoothing an aggregate index and its inherent delay or lag) might not be needed. Future study must determine if an insiders' transac- tions index reliably foreshadows the cyclical changes in the market. The evidence presented in this study suggests that a random sample will probably provide more false leads than a more selective sample based on intensive insider trading in companies that permit buying and tolerate selling by their executives. This study also suggests that it is 148 probably a good idea to construct an insiders' transactions index that takes into account the net number of transactors. Qausas and ggnjectures.--The positive relationship between insiders' transactions and a stock's price behavior in the market seems unlikely to result from pure chance. While the identity of causative factors must await further study, several possibilities should be noted. A change in stock prices is never the result of a single cause but of a complex of causes which constantly vary in intensity and of which only a part can be adequately measured by statistics. Chronology does not imply causal- ity. It is possible for insiders' transactions to occur in advance of an economic change without being one of the causes of the event. The observed success in forecasting may result from the Officials' inside information, their skill in determining whether publicly-known events will affect their company's future favorably or unfavorably, or from some other cause. For example, it is a well-known stock-market phenomenon that initial activity generates additional activity in the direction of the initial change in prices. 4Friend, Brown, Herman, and Vickers, A Study of Mutual Funds, p. 359. See also Securities and Exchange 149 The volume of shares traded by the insiders was found to be an insignificant part of the total shares traded during the month. Even if the volume of the insiders' transactions had no significant influence on the stock price as the evidence suggests, it is still possible that the Observed correlation may be systematic. For example, one possible explanation is that in their transactions the insiders may themselves cause the fulfillment of their own expectations. This would not happen because of the volume of their transactions or because of any inside knowledge they might have of a share's worth. Rather, the volume precipitated by the insiders' transactions is probably the influence which sends the stock price in the predicted direction. If the six-month interval is arbitrarily divided into three equal parts, the effect of the increase in volume hypothesized to result from others' following the lead of the insiders, whether knowingly or unknowingly, could be causally related to the Observed success in fore- casting. For example, it is reasonable to believe that there is some trickle of unreported volume associated with Commission, Stock Trading on the New Tork Stock Exchanga gn Saptembar 3, 1946 (Washington: Government Printing Office, 1947), p. 11. 150 the transactions of friends and relatives of the insiders. The insider's broker knows of these transactions before they are publicly disclosed. He has other customers and fre- quently recommends to them "favored stocks" on some basis which may well include his knowledge of the insider's views. During the second two months, public disclosure may stir up additional demand or supply. A substantial and growing number of citizens subscribes to the stock exchange releases and the later but more detailed Official Summary publication. The articles regularly appearing in the financial press strongly suggest that the insiders' transac- tions may have a large market following. Some investment advisors have prospered on the empirically plausible but possibly fallacious assumption that insiders trade on the basis of important and confidential information. The hypothesis that a change in price may result from shifts in demand or supply during the first four months may be reasonable. If the section of the trading public which follows the insiders' transactions acts in such a way as to have an effect on the price in the hypothesized direction, it is probable that this will be noted by specu- lators. There is evidence that a significantly large number of stock-traders follows price and volume movements at the technical level. The investment behavior of chartists could 151 have a snow-ball effect tending to produce the observed correlation. If these supposed relationships and feedbacks are strong enough to cause price changes, the economic signifi- cance of the relationship may be useless or short-lived. In other words, if insiders do n9t,act on the basis of confi- dential knowledge, there may be no correlation between the insiders' transactions and their company's trend in earnings or dividends. These are questions that require further study and analysis. In sum, successful forecasts in the aggregate may be due to the official's inside knowledge, superior investment skill in general, or something else. This thesis has only shown that the forecasting relationship is probably not due to chance alone. Social and behavioral For many years there has been a difference of opinion about the usefulness of reported stock transactions of corporate officials as an investment guide. Most of the claims on each side have come from investment services in their effort to explain the inclusion or exclusion of the information in their reports. Most of the claims, however, have been based upon limited observation, and on neither 152 side has convincing experimental evidence supported the contentions. Basic to the claim that insiders' transactions are useful as an investment guide is the assumption that key corporate officials are expert forecasters of their com- pany's stock price for two reasons: they have important "inside" information, and they are profit maximizers not hampered in this pursuit by law or corporate policy. The arguments supporting the contention that the reported stock transactions are useless for forecasting purposes include the notion that insiders have various motives when they buy or sell their company's stock, and that since insiders are not investment analysts, their judgment of the merits of a stock is distorted by their involvement in its destiny. Whether either set of assumptions is valid is irrelevant, for this study does not directly test the use- fulness of the reported data as an investment tool. Never- theless, the evidence seems to support the first set of assumptions (basic to the hypothesis of the study) that there is some utility in the reported insiders' transactions --providing, of course, that the inferred favorable or adverse information (on which the insiders trade) is not discounted in the market price during the one-to-two month 153 lag between the insiders' transactions and their public disclosure. Forecasting for profit.--A second closely related implication deduced from the evidence (and not inconsistent with the hypothesis) is that insiders do enjoy an edge in trading. However, testing the ability of insiders' transac- tions to forecast for profit was not the purpose of this study. For this reason such factors as taxes, commissions, lags, and cash dividends have been ignored. Nevertheless, if forecasting properties exist, their use in stock-market strategies should be possible. The weak relationship be- tween insiders' transactions and the market trend found in this study casts some doubt on the notion that insiders' transactions can be used as an investment guide. The great risk of making an incorrect decision would be compounded by the increased trading costs attendant on frequent turnover. On the other hand, if greater selectivity of sample reveals a stronger relationship, it is possible that the application of quantitative rules could improve portfolio allocations and the performance of the average diversified portfolio. Perhaps such a relationship could be converted into some kind of timing formula that, in addition to mini- mizing the investor's subjective reactions to current 154 events, would enable him to avoid the emotional errors of others. At least, analysis of insiders' transactions might encourage further investigation of securities where there is a large net number of buyers and sellers. fiegurity analysis Brokerage house analysts, advisory services and other security analysts frequently select and recommend common stocks whose situation seems to favor a near-term market advance. The usual assumption is that if earnings increase or the dividend is raised, the price will improve. Thus, the process consists essentially of locating and recommending companies likely to increase their earnings or dividends in the near term. The hazards inherent in the method are that the expected improvement does not take place, that it is already discounted in the current price, and that for some other reason or for no known reason the price does not move as anticipated. Diversification and aggregate results.--Because of these hazards, diversification is a principle of sound in- vestment, and investors are encouraged to diversify their holdings. In spite of these hazards it may be possible to obtain worthwhile results on the average from competent 155 short-term analyses and predictions. Benjamin Graham suggests that there are similarities between the predictions of the life insurance actuary and those of the security analyst, noting that "diversification is of the essence in actuarial science." The life insurance actuary makes predictions . . . based on . . . carefully analyzed past experience. . . Out of these predictions, with the aid of mathematical techniques, he fashions suitable premium schedules for various types of insurance. What is most important for us about his work and his conclusions is that he deals not with individual cases but with the probable aggregate result of a large number of similar cases.5 Graham concludes the analogy by saying that security analysis would benefit from the use of established statisti- cal tools to investigate the "successful" approach to stock market forecasting:6 We should collect the studies and recommendations of numerous analysts, classify them in accordance with their Objectives . . . and then do our best to evalu- ate their accuracy and success. The purpose of such a record would not be to show who is a good security analyst and who is a poor one, but rather to show what methods and approaches are sound and fruitful and which ones fail to meet the test of experience. . . . In view of the importance of this analytical work, in terms of time, energy, and money cost, it might not be 5Benjamin Graham, "Towards a Science of Security Analysis," Financial Analysts Journal, VIII (August, 1952), 97. 61bid., p. 99. 156 a bad idea to subject it to a thorough-going evalua- tion.7 The analogy is probably weak because of the greater number of psychological variables in the field of inveStments compared to that in actuarial science. Nevertheless, the point is clear: the statistical limitations of heterogen- eity (among companies), non-randomness (among insiders' transactions), and interdependence (among internal and external events), for example, are not insurmountable bar- riers to a more scientific approach to the study of secur- ity prices. APPENDIXES APPENDIX A FREQUENCY AND NUMBER OF NET BUYERS AND SELLERS IN THE MONTHLY INSIDERS' TRANSACTIONS INDICES, BY COMPANY, DURING 1957-1960 APPENDIX A FREQUENCY AND NUMBER OF NET BUYERS AND SELLERS IN THE MONTHLY INSIDERS' TRANSACTIONS INDICES, BY COMPANY, DURING 1957-1960 Frequency Peak No C32? Company 3‘ _§, 3 g 'S.‘ ".3 8. :1 a 3 a a la Aldens, Inc. 8 4 2 2 Central Illinois Light 3 O 1 0 3 Allied Laboratories 7 11 1 2 4a American Airlines 15 10 4 7 5 American Investment 2 l4 1 5 6 American Machine and Metals 14 2 3 1 7 American Molasses Company 6 3 l 1 8 American News Company 15 2 2 1 9 American Water WOrks 19 4 4 l 10 Barker Bros. Corporation 7 5 l 2 11a Beech Aircraft 5 9 2 3 12 Brooklyn Union Gas Company 4 1 2 1 l3 Buckeye Pipeline Company 1 1 l 1 14 Central Ill. Public Service 2 1 2 1 15a Cerro de Pasco Corporation 12 3 3 3 16 Chicago Rock Island & Pacific 7 4 1 1 17a American Metal Climax ll 2 6 18 Colorado Fuel and Iron 8 8 2 1 19a Commercial Credit 14 16 2 2 20a Consolidated Electrodynamics 5 18 2 4 21a Consolidated Gas Utilities 5 7 2 2 159 160 APPENDIX A--Continged Frequency Peak No C133? Company a g 35;; E H a 3 $ £3 3% 22a Dover Corporation 6 12 3 2 23 Ekco Products Company 12 10 7 1 24 Federated Department Stores 1 31 1 4 25 Fifth Avenue Coach Lines 10 3 1 26 Gar Wood Industries 7 1 1 27a General American Oil of Texas 11 11 7 2 28a General Cable Corporation 7 13 2 3 29a General Dynamics Corporation 5 32 2 7 30a General Electric 18 2O 5 7 31 General Precision Equipment 18 l 3 323 General Tire and Rubber CO. 8 15 4 3 33a Georgia Pacific Corporation 12 9 3 2 34 Grayson-Robinson Stores, Inc. 0 8 O 3 35 Great Western Sugar Company 1 1 l 36 Heller (Walter E.) and Company 2 22 1 2 37 Hercules Motor Corporation 5 1 2 l 38 Hoffman Electronics Corporation 2 l2 1 2 39 Hussman Refrigerator 11 2 4 1 40a International Tel. and Tel. 13 15 4 3 41 Iowa Illinois Gas and Electric 7 3 0 42 Kansas Gas and Electric 9 O 1 O 43 Kimberly Clark Corporation 3 21 1 2 44 Lane Bryant, Inc. 4 l 2 l 161 APPENDIX A--Continued Frequency Peak No 05 0) Code Company g‘ _§ 3 3 NO' -a '4 m H % r4 >: H 5 m s w m U) m U) 45 Lehman Corporation 4 O 2 0 46a Lerner Stores Corporation 20 5 2 2 47 Liggett and Myers Corporation 0 l7 0 4 48 McCrory Stores Corporation 8 3 l 1 49 Mercantile Stores CO., Inc. 7 2 1 50 Midland Steel Products Company 13 O 2 0 51 Midwest Oil Corporation 4 9 l l 52 Minnesota and Ontario Paper 12 O l O 53 .Monsanto Chemical Corporation 5 18 1 4 54 .Monterey Oil Company 5 2 l 55 Myers (F. E.) and Bros. Co. 2 3 l 3 56a National Distillers & Chemical 12 7 2 2 57 National Lead Company 10 1 2 58 National Tea Company 4 5 l 1 59 Newberry (J. J.) Company 10 5 7 1 60a North American Aviation 4 10 2 9 61a Northrop Aircraft 7 13 2 2 62 Overland Corporation 2 O 1 0 63a Paramount Pictures Corporation 3 7 2 2 64 Parmelee Transportation Co. 1 6 l 1 65 Patino Mines and Enterprises 1 O 1 O 66 Pennsylvania Glass Sand COrp. 4 O 2 0 67a Pennsylvania Power and Light 4 7 2 3 162 APPENDIX A--Continued Frequency Peak No nge Company g g a): :3: H H a a; a a: 68 Philadelphia and Reading Corp. 4 13 l 2 69 Phillips Petroleum 6 10 1 2 70 Public Service Co. of Indiana 2 1 1 1 71a Publicker Industries, Inc. 5 34 2 3 72 Real Silk Hosiery Mills, Inc. 0 10 O 2 73 Reynolds Metals Company 2 17 1 1 74 Rhodesian Selection Trust Ltd. 0 O 0 O 75 St. Joseph Light and Power 6 2 l 76 San Diego Gas and Electric 15 2 3 1 77a Seaboard Airline Railroad 3 l6 2 3 78 Sears Roebuck and Company 21 5 5 l 79 Smith (A. O.) Corporation 1 4 1 1 80 South American Gold & Platinum 11 3 2 l 81 Spencer Kellogg and Sons 13 1 2 1 82 Starrett (L. S.) Company 26 0 4 O 83 Stewart and Warner 9 1 2 84 Stix Baer and Fuller Company 0 11 O 3 85 Storer Broadcasting Company 3 1 1 1 86a Sunray Mid-Continent Oil 11 10 2 3 87a Talcott (James), Inc. 22 9 6 4 88a Tennessee Corporation 10 3 4 2 89a Twentieth-Century Fox Films 6 l2 2 2 90a Twin City Rapid Transit 16 l 3 2 91a Union Chemical and Materials 11 11 3 2 163 APPENDIX A- -Continued Frequency Peak No C1332 company g E g g 92a United Biscuit of America 6 8 2 2 93 United Engineering & Foundry 8 3 l 1 94 U. S. Playing Card Company 3 1 1 1 95a United Cigar Whelan Stores 10 4 4 3 968‘ Van Norman Industries 21 10 2 3 97 Van Raalte Company, Inc. 11 10 2 1 98 Western Pacific Railroad CO. 13 2 2 1 99 Wheeling and Lake Erie Ry. Co. 0 O O 100 ‘White.Motor Company 6 23 l 2 aCompany in 33-company selected sample. A P P E N D I X B DATES OF MONTHLY OBSERVATIONS OF STOCK PRICES APPENDIX B DATES OF MONTHLY OBSERVATIONS OF STOCK PRICES The prices of the stocks and the level of the SEC Composite Index were recorded as of the close of trading on the following dates: n Date n Date n Date 1 1-18-57 19 7-18-58 37 1-15-60 2 2-15-57 20 8-15-58 38 2-12-60 3 3-15-57 21 9-12-58 39 3-18-60 4 4-18-57 22 10-17-58 40 4-14-60 5 5-17-57 23 11-14-58 41 5-13-60 6 6-14-57 24 12-12-58 42 6-17-60 7 7-12-57 25 1-16-59 43 7-15-60 8 8-16-57 26 2-13-59 44 8-12-60 9 9-13-57 27 3-13-59 45 9-16-60 10 10-18-57 28 4-17-59 46 10-14-60 11 11-15-57 29 5-15-59 47 11-18-60 12 12-13-57 30 6-12-59 48 12-16-60 13 1-17-58 31 7-17-59 49 1-13-61 14 2-14-58 32 8-14-59 50 2-17-61 15 3-14-58 33 9-18-59 51 3-17-61 16 4-18-58 34 10-16-59 52 4-14-61 17 5-16-58 35 ll-13-59 53 5-12-61 18 6-13-58 36 12-18-59 54 6-16-61 165 A P P E N D I X C INSIDERS' TRANSACTIONS: RULES FOR INCLUSION APPENDIX C INSIDERS' TRANSACTIONS: RULES FOR INCLUSION The rules adopted in this study for the collection of the insider transaction data are presented below. Ex- planations are offered in the Discussion section following the list of rules where the reason for the rule is not Obvious. Rules fgr Inclusion. 1. Examine only insider transactions in the 100 common stocks of the companies that make up the sample. 2. Include officers, directors and beneficial owners other than corporations. 3. Exclude trustee or voting-trustee transactions unless the person was also either an officer, director or beneficial owner. 4. Exclude bequests or inheritances. 5. Exclude compensation or bonuses. 6. Do not include other forms of distribution of shares by or to corporate officials. 7. DO not include exchange or conversion, exercise of rights, etc., or gifts. 167 168 8. Do not include redeemed, called, matured or retired transactions. 9. Do not include stock dividends or stock splits. 10. Do not include transferred or accrued transac- tions. 11. DO not include stock-Option purchases or shares otherwise acquired through the exercise of option agreements. 12. Do not include private transactions not com- pleted in the open market. All other transactions, no matter what the nature of the insider's ownership, were included. Thus, transactions by insiders' wives were included, as were transactions of securities, whether beneficially owned or not, made indirect- ly through a holding company, partnership, trust or other form of intermediary. Also included were transactions made and held directly by the insider, but which he disclaimed as not being the beneficial owner of the securities. Discussign of Rules As noted above, rules are discussed only where reasons for their formulation are not immediately obvious. No. 2. The transactions of beneficial owners, other than corporations, were included. Individual large share- holders and investment companies were included because it 169 was assumed that they may have more information than some of the directors. On the other hand, some industrial, rail and utility corporations may either consistently add to their holding month after month or dispose of their entire beneficial interest at one time. Corporations were excepted from this beneficial owner inclusion rule on the assumption that they were in the process of changing their complex of subsidiaries, and that this long-run goal negated the ra- tionale of assigning short-run forecasting properties to these transactions. As the data later revealed, these cor- porate transactions were infrequent. When they did occur, the purchases were in lots that were small in relation to the prior month's holdings. The sales transactions usually consisted of the entire beneficial interest. No. 13. Private transactions not completed in the open market were excluded because meaningful price data were not available. A P P E N D I X D SEC RULE l6a-9 (EXEMPTION FOR SMALL TRANSACTIONS) APPENDIX D SEC RULE 16a-9 (EXEMPTION FOR SMALL TRANSACTIONS) Under the authority contained in Section 3(a)(l2) of the Securities Exchange Act of 1934, the Commission has adopted rules exempting certain securities or transactions from the provisions of Section 16(a). One of these rules is of sufficient importance to this study to warrant its pre- sentation. Since this study will derive an insiders' transactions index based on the number of buyers less the number of sellers, ignoring the quantity of the transaction, it is important to establish the fact that the small trans- action is exempted from the reporting requirements. Thus the reported transactions individually exceed $3,000 and are for the most part material in dollar amount. Rule l6a-9 exempting small transactions from disclosure states: (a) Any acquisition of securities shall be exempt from Section 16(a) where-- (1) the person effecting the acquisition does not within six months thereafter effect any dispo- sition, otherwise than by way of gift, of securities of the same class, and (2) the person effecting such acquisition does not 171 (b) (C) 172 participate in acquisitions or in dispositions of securities of the same class having a total market value in excess of $3,000 for any six months period during which the acquisitions occur. Any acquisition or disposition of securities by way of gift, where the total amount of such gifts does not exceed $3,000 in market value for any six month period, shall be exempt from Section 16(a) and may be excluded from the computations prescribed in paragraph (a)(2). Any person exempted by paragraph (a) or (b) of this rule shall include in the first report filed by him after a transaction within the exemption a state- ment showing his acquisitions and dispositions for each six month period or portion thereof which has elapsed since his last filing. A P P E N D I X E ILLUSTRATION OF DATA SUMMARIZATION: ALDENS , INC . APPENDIX E ILLUSTRATION OF DATA SUMMARIZATION: ALDENS, INC. The mechanics of collecting and compressing the raw data as outlined in Chapter IV are illustrated in this Appendix. Table E-l shows the transactions of four selected months after filtering in accordance with the rules of inclusion and exclusion. Table E-2 reveals the monthly indices of transaction activity (x) and stock-price perfor- mance (y) for this series as well as the continuous stock price series. Table E-3 illustrates one computation of the index of a stoCk's performance (y) from Table E-2. 174 175 OOH OHm OHOmHucmum .O .m O OOH OmO.H OHOmmDm .O .O OmOH .smz OOm mOO.OH ngucmmom .m .m OOH SHm smHuoz .m .o H+ OOH OOH amOHOmz .O .O OmOH ..omO H+ OOH O mmouo mcHammmOO smOH ..omO OOH SHH mmHnos .m .0 OOm mos OHOmHuamum .O .m OON OOH.H mHnmmum .O .O O+ oom SmH xoo .m .o BmOH ..Doo NOOGH o umxnmz Hm OOOMHB mmumsm .mcmua mmwmm Ommmnuusm mo mmcHOHom mchHommm HmHUHmmO wumn .mumUHmcH 3m mmumzm nucoz maHachmm mmmthmmH mmBZOZ QMBUMHmm m0 mZOHBU¢m2¢mB .mmeHmZH .ozH .mzmnHH Him mam O Test N = 48 X H the number of positive correlation coefficients = 40 By using a normal approximation (including a correc- tion for continuity) one can calculate the probability of getting 40 or more positive correlation coefficients out of 48 when P II m + . - z=(x 5) NP=4.77 Thus, the probability of randomness is approximately .00001. 181 A P P E N D I X G RANKING OF 83 COMPANIES BY CORRELATION COEFFICIENTS BETWEEN THEIR INSIDERS' TRANSACTIONS AND THE RELATIVE STOCK PRICE LAGGED SIX MONTHS BASED ON SIX OR MORE MONTHLY OBSERVATIONS DURING 1957-1960 APPENDIX G RANKING OF 83 COMPANIES BY CORRELATION COEFFICIENTS BETWEEN THEIR INSIDERS' TRANSACTIONS AND THE RELATIVE STOCK PRICE LAGGED SIX MONTHS BASED ON SIX OR MORE MONTHLY OBSERVATIONS DURING 1957-1960 Company No. of Net Buying Code and Net Selling Correlation NO- Observations coeffIClent 75 7 +.8631a 81 14 +.7732a 92 14 +.6963a 21 12 +.6720a 84 11 +.6469b 1° 12 +.6384b 41 7 +.6375 72 10 +.6226C 8° 14 +.ezoga 9 23 +.554oa 91 22 +.5385a 17 16 +.5352b 4 25 +.524oa 14 +.4773C 64 7 +.4744 76 17 +.4693C 183 184 APPENDIX G--ggntinued Company No. of Net Buying Correlation Code and Net Selling Coefficient No. Observations 49 12 +.4688 34 8 +.4257 46 25 +.4156C 60 14 +.4098 28 2o +.3871C 50 13 +.3848 22 18 +.3805 73 19 +.3732 16 11 +.3493 3o 38 +.3455C 67 11 +.3260 100 29 +.3095 27 22 +.3043 38 14 +.3005 93 11 +.2901 48 11 +.2718 39 13 +.2377 20 23 +.2351 96 31 +.2200 58 9 +.1971 53 23 +.1961 98 15 +.1933 33 21 +.1769 15 15 +.1721 88 13 +.1631 185 APPENDIX G--Continued Company No. of Net Buying Correlation Code and Net Selling . . No. Observations CoeffICIent 69 16 +.1553 95 14 +.1522 68 17 +.1502 71 39 +.1381 78 26 +.1338 86 21 +.1284 11 14 +.1203 47 17 +.1096 25 12 +.1069 5 l6 +.1044 3 18 +.0859 23 22 +.0849 18 16 +.0775 6 16 +.0396 26 9 +.O294 51 13 +.009l 24 32 +.0066 82 26 +.0063 29 37 -.0066 83 12 -.0069 8 17 -.0094 40 28 -.0129 87 31 -.0225 90 17 -.0280 56 19 -.0512 77 19 -.O628 186 APPENDIX G--C0ntinued Company No. of Net Buying NO- Observations 19 30 -.0706 43 24 -.0773 97 21 -.1150 89 18 -.1216 59 15 —.1292 32 23 -.1423 36 24 -.l487 37 6 -.2241 54 12 -.2459 61 20 -.3l26 63 10 -.3164 31 23 -.3929C 7 9 -.4714 57 11 -.6272b 42 9 ____3 52 12 ____3 1See Appendix A. 2By definition we will call companies more successfuL in this ranking, the larger the correlation coefficient. Thus, we can approximately identify the relatively more suc- cessful and least successful companies in terms of the cor- relation coefficients of a linear relationship. However, tables of statistical significance assume that the number of observations are completely independent. The correlations through time are not completely independent because of the overlap in the six month intervals of some of the observa- tions. Thus statements of probability of a chance occur- rence are not appropriate. Despite this limitation, the standard tables of statistical significance are useful for 187 identifying the relatively more consistent transactions in terms of a linear relationship. The companies identified as a, b, or c, are defined as successful (or unsuccessful) on the basis of a .05 level of significance. a = .01 level; b = .025 level; and c = .05 level. 3Two of the 83 series have an undefined correlation due to a zero variation in the number of buyers less sellers for each observation. The nine and 12 monthly observations in companies numbered 42 and 52 respectively were always one net buyer in each month. A P P E N D I X H SIGNIFICANCE TEST OF DIFFERENCE BETWEEN MEANS APPENDIX H SIGNIFICANCE TEST OF DIFFERENCE BETWEEN MEANS The formula for this test when the means are drawn from large samples is: Xl-X2 t: S? + SE E; E; If t is > 1.96 or < -l.96 then the means are said to be significantly different at the .05 level. If t is > 2.58 or < -2.58 the means are significantly different at the .01 level. t = 3.042 highly significant at the .01 level. The formula used to compute the variance of the sample with one degree of freedom is: 2Y2 - (2W2 3 = N N - 1 189 A P P E N D I X I AVERAGE PERCENTAGE PERFORMANCE OF INSIDERS' NET BUYING AND SELLING BY COMPANY APPENDIX I AVERAGE PERCENTAGE PERFORMANCE OF INSIDERS' NET BUYING AND SELLING BY COMPANY In this appendix the companies are ranked according to the average magnitude of success achieved by the insiders in buying and in selling. Each observation is the percen- tage change in the stock's price less the percentage change in the market over six month periods from the month of the insiders' net buying or net selling, The first table sum- marizes the 731 net buying observations and the second table presents the average results of the 776 net selling observa- tions. Presumably, the plus sign exemplifies performance superior to the market and the minus sign denotes a poorer performance relative to the market. 191 TABLE I-l AVERAGE PERCENTAGE PERFORMANCE OF INSIDERS' NET BUYING BY COMPANY Ideniiggzgiion NO' Of Monthly RZI::i::rP::I:3:a::e Number Ob:::V::;::: Of Of the Insider-Bought (See Appendix A) Stock (Per Observation) 64 1 +64.65 73 2 +34.98 100 6 +33.84 68 4 +27.65 1 8 +25.26 21 5 +23.82 38 2 +23.30 92 6 +22.40 55 2 +18.76 42 9 +18.56 87 22 +15.77 14 2 +15.58 2 3 +13.04 l9 +12.58 41 7 +11.94 88 10 +11.74 80 11 +10.57 32 8 +10.54 4 15 +10.20 ll 5 + 9.63 33 12 + 9.57 35 1 + 8.94 78 21 + 8.82 40 13 + 8.19 192 193 TABLE I-l--Continued Company. No. of Monthly Mean Percentage of Identification Observations of Relative Performance Number Net Bu in of the Insider-Bought (See Appendix A) y g Stock (Per Observation) 76 15 + 8.05 75 6 + 8.03 20 5 + 8.01 96 21 + 7.86 71 5 + 7.48 17 5 + 7.19 95 10 + 6.94 83 3 + 6.60 46 20 + 6.57 36 + 6.39 66 + 6.06 51 + 5.73 8 15 + 5.57 58 4 + 5.39 49 5 + 5.19 12 4 + 4.12 53 5 + 3.95 10 7 + 3.75 91 11 + 3.46 69 6 + 3.39 44 4 + 3.36 48 8 + 3.27 61 7 + 2.94 85 3 + 2.89 97 11 + 2.71 194 TABLE I-1--Continued Company. No. of Monthly Mean Percentage of Identification Observations of Relative Performance Number of the Insider—Bought (See Appendix A) Net Buying Stock (Per Observation) 30 18 + 2.60 27 11 + 2.55 89 6 + 2.50 19 14 + 2.12 60 4 + 1.82 98 13 + 1.70 86 11 + 1.58 90 16 + 1.37 50 13 + .98 52 12 + .93 54 7 + 78 81 13 + 75 23 12 + 73 24 1 + 54 22 6 - .30 70 2 - .34 43 3 - 43 67 4 - .71 93 7 - .90 5 2 - .94 39 11 - 1.50 94 3 - 1.64 28 7 - 1.96 59 10 - 2.65 56 12 - 2.84 195 TABLE I-l--QQBLiEEGQ Company. No. of Monthly Mean Percentage of Identification Observations of Relative Performance Number Net Bu in of the Insider-Bought (See Appendix A) y 9 Stock (Per Observation) 18 8 - 2.93 26 7 - 3.17 37 5 - 3.48 13 1 - 3.91 7 6 — 4.14 29 5 — 4.37 3 7 — 4.44 79 1 - 4.84 82 26 - 5.45 63 3 - 5.64 6 14 — 5.70 15 12 - 5.93 16 7 - 6.71 25 10 - 7.35 45 4 - 8.69 31 5 - 9.69 57 1 — 9.81 77 2 -1l.78 62 2 -23.11 65 1 -26.51 196 TABLE I-2 AVERAGE PERCENTAGE PERFORMANCE OF INSIDERS' NET SELLING BY COMPANY Company. No. of Monthly Mean Percentage of Identification Observations of Relative Performance Number Net Sellin of the Insider-Sold (See Appendix A) 9 Stock (Per Observation) 10 5 -25.90 35 1 -21.49 80 3 -20.85 90 1 -19.65 85 1 -16.62 16 4 -15.34 81 1 -15.02 13 1 -14.27 27 11 -12.72 75 l -12.14 91 11 -11.67 76 2 -11.00 15 3 -10.77 22 12 -10.15 12 1 - 9.58 92 8 - 8.97 29 32 - 8.70 70 1 - 8.57 17 11 - 8.30 18 8 - 7.42 48 3 - 7.15 28 13 - 6.32 30 2O - 6.22 197 TABLE I-2--Continued Company. No. of Monthly Mean Percentage of Identification Observations of Relative Performance Number Net Sellin of the Insider-Sold (See Appendix A) 9 Stock (Per Observation) 93 3 - 5.85 98 2 - 5.70 4 10 — 5.17 67 7 - 5.04 96 10 — 4.95 6 2 - 4.95 3 11 - 4.79 86 10 - 4.32 26 2 - 4.24 46 5 - 3.16 21 7 - 2.94 39 2 - 2.92 9 4 — 2.68 57 10 — 2.65 56 7 - 2.46 49 7 , - 1.33 5 14 - 1.32 20 18 - .80 59 5 - .66 78 5 - .54 58 5 - .47 72 10 - .28 69 10 - .13 23 10 - .07 25 2 + .33 198 TABLE I-2--Continued Company. No. of Monthly Mean Percentage of Identification Observations of Relative Performance Number Net Sellin of the Insider-Sold (See Appendix A) 9 Stock (Per Observation) 84 11 + .39 77 16 + .78 53 18 + .79 47 17 + 1.48 83 9 + 1.70 73 17 + 2.06 60 10 + 2.48 55 + 2.80 95 4 + 2.89 38 12 + 2.96 33 9 + 2.98 11 9 + 3.65 7 3 + 4.00 71 34 + 4.01 14 1 + 4.01 89 12 + 4.19 19 16 + 4.49 43 21 + 4.73 100 23 + 5.01 97 9 + 5.02 63 7 + 5.36 51 9 + 5.50 94 1 + 5.66 37 1 + 7.08 79 4 + 7.23 199 TABLE I-2--antinued Company. No. of Monthly Mean Percentage of Identification Observations of Relative Performance Number Net Sellin of the Insider-Sold (See Appendix A) 9 Stock (Per Observation) 1 6 + 7.55 54 5 + 7.93 24 31 +10.28 31 18 +10.33 40 15 +10.79 88 3 +11.12 36 22 +11.91 44 l +12.50 8 2 +12.87 87 9 +15.76 61 13 +16.6l 34 8 +18.59 68 13 +18.73 32 15 +22.06 64 6 +25.62 APPENDIX J A RANKING OF THE INSIDERS' FORECASTING RECORD IN PREDICTING THE DIRECTION OF THEIR COMPANY'S STOCK PRICE RELATIVE TO THE MARKET AS A WHOLE SIX MONTHS LATER 201 MHMHHOumz paw HUOHEOno GOHGD ammo. mm mH mmmo. n m UHHuomHm 0cm mmw MHOGHHHH MBOH hhmo. om OH GoHpmuomnoo OHQMU Hmuwcww mwmo. mm mm coHpmHomuoo OOHEmchn Hmumamw OmHO. mm Hm mammsoo nouoz mquz mOOO. mm OH mmcHHnHm cmoHu08< mOOO. OH NH OOHHOEH O0 uHsomHm OOHHOO mmoo. mm mm OHHHOOHm Hmumaww OHoo. mm Hm NCMQEOO Cam Husnmom mummm MHoo. mm mH mxuoz nmumz cmuHH084 mpmmomuom mo .02 Hmuoe Gnu mo mGOHgm> mgmmumnom use mommmuosm Inmmno thucoz mo .02 Hmumwuw mHnucOS Hammmmoosm Ho Hmnvm am we SUHHHHODOHO UmumEHumm one mo umnssz Hmuoa mo mHOOHmcH mmadfl mmBZOZ XHm mqomz ¢ mfl BMMMfiS HEB OB m>HB omNH. m m #anH MHOGHHHH Hmupcmu mmHH. mH NH menoumuonma OOHHHH mmHH. mH NH GoHumnomuoo Hm>on mmHH. HH m hammfioo Ummq HmaoHumz mmmo. a O OHupome cam mmw mmmcmx mmmo. OH OH monoum COHwnz HmmHU OOHHGD mmmo. OH OH EOGHpmHm UGO UHOO GMOHHOE< nusom omho. NH m mmHuHHHuD mam OmumoHHOOCOO ammo. NN mH mmxmB mo HHO GMUHHOEN Hmumcmw mummumuom mo .02 Hmvoa may mo mcoHum> mvmmomuom use mommwoosm Inmmno mHnHGOS mo .02 Hmummuw hHfipaoS Hammmmuosm no Hmsvm am no SDHHHDOHOHO OmpmsHumm 0:9 no 808852 Hmuoa MO mHQUH mCH UOSGHHGOUIIb NHQZmQMN 203 OOON. HH O usmHH Osm Hm3om mHsm>H>mssmm mmmN. m e mOHHumsOsH U003 HMO oemN. N N Osmmfioe Hmmsm sumummz ummue NONN. 0H OH AXMEHHU Hmuwz GOOHumfidv EssOOQNHOS meHHO eeNN. O m Hm3om Osm uanH CQOOOO .um NNHN. mN mH sOHumHomnoe mmuoum HOGHOH eNHN. OH O soHumuemuou moHsouuomHm smEumom ONHN. OH O msom Osm OOOHHOM umoswmm eNHN. OH O .OGH msOOH¢ mmmH. NH m GoHumuomuoe .moum nOxumm OHmH. HN mH soHumnomHoe OHNHOON mHmuOOO mOmH. m O Osmmfioe mme soHsD sthooum NeeH. OH HH hsmmfioo OBOZ GOOHHOEH mummomuom mo .02 HmuOB Gnu mo msoHum> mummomnom use mmmmmoosm Immune thusoz mo .02 HOumOHO mqusoz Hsummwoosm no Hmswm am no ODHHHnmnoum UmumEHumm One mo 868892 Hmuoe mo OHOOHOGH CODGHuCOUIIb XHQmem¢ 204 ooem. OH O GOHumuomuoe msHemOm Osm OHsmHOOmHHnm ooom. OH O sOHumuomuoe mum»: Osm uummmHH oeem. OH OH msmmEoe mHmumz meHoshmm ooom. mN NH soHumuomuoe HOUHEOQO Ousmmsoz 000m. mm 0N .OsH .mOHuumsOsH HOHOHHQsm ooom. HH 0 muessom Osm msHuOOsHmsm OOuHsD mHOO. OH 0 GOHH cam Hmsm OOMHOHOO mHOO. OH O .OGH .mHmqu cam OGHCOMS GOOHuwfid NOmm. NH O .osH .msmmfioe mmuoum OHHusmoqu mmmm. mN mH MOHEmsheouuome OmumOHHomsoe mNHm. O m mammfioo msHummovmonm Hmuoum emom. mH m OsmmEOU emeHHHmm OHMHOON sumumm3 momN. MH m hammEeU HeumHOmHHumm smEmmsm mummomuom mo .02 Hmuoa Onu mo msoHum> mummomuom use mmmmwuosm lummne thusoz mo .02 Hmummue mHnusoz Hsummmoosm no Hmsvm sm 60 ODHHHHOHOHO OmumEHumm was no 868852 Hmuos we OHOUHOGH @ODGHHGOUIIU NHQmem¢ 205 mmmO. OH O uHmcmue OHmmm OuHo OHze NeOe. mH m OOOHHHMN OGHHHHN Osmonmwm NeOe. NH 0 HOOHEmne Usm mHOHHHumHQ HmsoHumz Nmee. HN OH HHe usOsHusoeIOHz Omussm NHee. mN HH Osmmfioe HOQQsm Osm OHHB Hmumsme Oeme. N O .OGH .mmuoum semsHQemlsommmue OOoe. OH O seHumH>¢ GOOHHOEN nuuez OHOm. em OH uHOOHO HOHOHOEEOU eeem. m N msmmfioe OOH>uOm OHHQsm mHosHHHH Hmnusme oeom. m N msmHvsH mo OGOQEOU OOH>HOm OHHQsm eoem. m m .OGH .usmhum OGDH eeem. HH e OHuHomm Osm UGOHOH xuom OUOOHQU ooem. mH O soHumuomuoe Ommmmssma mummomuom mo .02 HmuOB Onu mo msOHum> mummuwuom use mwmmmousm lummne mHnusoz mo .02 umummue thusoz Hsummmoosm no Hmsum am no ODHHHHOHOHO OmumEHumm 0C9 mo umnssz Hmuoa MO mHmmVHmcH .memHmmdwnuO xHOZOOOO 206 OOOO. OH O ummHOHH¢ nommm OmOO. O m Osmmfioe soHumuHommsmHB OOHOEHON ONOO. OH O AmHosHHHHV usOEumO>sH GOOHH08¢ ONOO. OH O EsOHOHuOm mmHHHHsm eomO. N H msmmEoo OGHHOQHN wwwxosm HOOO. O O hammfioe MOE HmsoHumz OmNO. HH m soHumnomHoe mmnoum MHOHOOS mmeO. mH O hsmmfioe muesOoum Hmmum OGOHOHS OOOO. mH O soHumHemnoe HHe ummszz OOOO. OH R Onmmsoo H.O .OO Ouuwgamz OOmO. OH O coHumHomHoe oommm OO OHHOO mOOO. O N sOHumuomuoe Osmm OOOHO OHsm>HmmssOm mOmO. O N soHumHemuoe GOESOH mummomuom mo .02 Hmuos Tau we msoHum> mummomuom use mwmmwuusm summne mHsusez mo .02 Hmumwue thust Hsmmmmuosm Ho Hmswm am no suHHHnmnoum OmumsHumm OOH mo umnssz Hmuoe mo OHOOHOCH COUCH uGO OI 1h. NHQmean 207 OOHO. OO O Ocmmsoo A.m .Hv sumuumum NeHm. m m Osmmfioe mwmmmHOE GMUHHOE< Oemm. O N GOHumuomuoe nouoz mmHsouOm OOOO. HH O msmmfioo HOHHsm Osm Hmmm xHum HHOO. OH O mEHHm xom mususme nuOHusmza Hmmm. OH O .OOH .mHHHz OuOHmom HHHO Hmmm mNHm. m N soHumHomHoe A.e .dv suHEm mNHO. m N Osmmfioe .moum osm A.m .mv whoa: mwem. HN m .osH .hsmmfiou OuHmmm sm> NOeO. NH O Hmmmm OHHmuse Osm muommsst NOom. NH O Osmmsoo HHe hmumusoz NOom. NH O Hmsumz Osm uumzwum NOOO. NH O OOOHH numoo mscm>¢ BumHm mummomuom mo .02 Hmuoa Onu mo msoHum> mummomuom use mOmmOOOsm lummne mHsusoz mo .02 HOumOHO MHnusez Hsmmmmoosm so Hmsum cm mo ODHHHsmsoum OmumsHumm was mo umnssz Hmuoa MO mHOUH mCH UOEGHuGOUIIb NHQZWhHfi ’\ 208 muH>HuOm oz 0 o .msm .E< umsHB sOHuOOHOm COHmOOozm huH>Huum oz 0 e Osmmfieo hm3HHmm OHHm OMOH Osm msHHOm£3 eeeo.H H o mOmHHmHOusm Osm mOsHS osHumm eeeo.H N e sOHumuomuoe OsmHuO>e eeeo.H Nm O mmuoum usOEuummOQ OOuMHOOOm OOmm. mN m usmemHsUm sOHmHUOHm Hmumsmw HOmm. eN m ummuouH4 mouzuuoz OOOO. Om n mammsoo Oam H.m umuHOSO anHmm Ommm. ON 0H .HOB Osm .HOB HmsOHumGuOusH mmOm. 0H m GOHumuomuoo mmusuOHm ussOEmHmm mOmm. O H msmmfioe Oume msHmmHm .m .D NONm. ON m GOHumuomuoe MHOHO OHHOQEHM mummumuom mo .02 HmuOB Osu mo OGOHum> mummomuom use mmmmwoosm Inmmne thusoz we .02 HmumOHO no Hmsvm am no OuHHHsmnouO OmumeHumm was anusoz Hsmmmmousm mo nmnssz Hmuoa mo muOOHmsH .mwmmflmmmwllb NHQmemd " A P P E N D I X K FREQUENCY OF SUCCESSFUL FORECASTS OF DIRECTION IN TWO MODELS, ONE DEFLATED AND ONE NOT DEFLATED 210 mmw .mmfl .MMfl mseHumsuwon A+V we .0: HmuOB .MM .mm .MHH N+ OH mH NN m+ OH OH NH O+ m m O m+ m m m O+ H N m O+ HOmumHumO uoc HOwumHmmO OOGOEHOMHONV mosmEHomummv msoHum>uOer nusoz m msHHsQ mumHHOO Hmooz was anucoz Hooum O OH HIV manHOm no we mEHOB CH soHuOOHHQ we mummomuom Hsmmmwuusm mo .02 mo .02 Hmuoe H+v mumhsm mo .02 uwz "XOUGH mCOH #Ummgm .HH. .muOOHmsH Qmfifidhmfl 802 $20 QZ¢ QMBfiHhmfl WZO .mHmQOZ.OBB ZH ZOHBUMMHQ ho mfimdvmmom HDmmmmUUDm m0 Nuzmbommm M NHQmemd 211 O‘ In In H H °‘ In mummomuom Hmuou eu mmmmmousm Hmuou mo Ommusmoumm OON ONN NOm msOHum>HmeO mo .02 Hmuoa MN! .mwm mam. msoHum>uOmno HIV mo .02 Hmuoa NHI. HII HII. OI m m m OI m m m OI O m m mI O O OH OI ON mm OO MI Om OO OmH NI AOOumHmOQ uos AOmumHmOQ TOGOEHOOHONV OusmEHemuOmv mseHum>HOmne nusoz m msHHsQ mumHHon HOOoz was anunoz Huoum O OH HIV mumHHmm no H+O mummsm mo .02 umz mo mEuOu sH sOHuUOHHQ mo mummomuom Hsmmmmuosm we .02 . m me 02 H ueB "NOOGH msoHuummsmHB .mHOOHmsH OODGHHGOUIIM XHQmem4 BIBLIOGRAPHY BIBLIOGRAPHY Bgoks Graham, Benjamin, Dodd, David L., and Cottle, Sidney. Segur- ity Analysis. 4th ed. New York: McGraw-Hill BoOk Company, 1962. Owen, D. B. Handbook of Statistical Tables. Reading, Mass.: Addison-Wesley Publishing Company, 1962. Securities and Exchange Commission. The Work of the Secar- ities and Exchange Commission. Washington: Government Printing Office, 1961. Artigles and Periodicals Cook, Donald C., and Feldman, Meyer. "Insider Trading Under the Securities and Exchange Act," Harvard Law Review, LXVI (January, 1953), 385-422, 612-641. Cowles, Alfred, 3rd. "Can Stock Market Forecasters Fore- cast?" Econometrica, I (July, 1933), 309-324. Graham, Benjamin. "Towards a Science of Security Analysis," Financial Analysts Journal, VIII (August, 1952), 96-99. Granger, C. W. J., and Morgenstern, Oscar. "Spectral Analy- sis of New York Market Prices," Kyklos, XVI (March, 1963), 1-27. Rothchild, Henry. "Financing Stock Purchases by Executives," Harvard B siness Review, XXXV (Marck-April, 1957), 136- 144. Securities and Exchange Commission. Official Summary 9f Security Transactigns and Holdings. washington: Gov- ernment Printing Office. Published monthly. Storer, Robert W. "A Critical Evaluation of Stock Market Indices," American Statistigal Association 1959 Pro- 213 214 ceedings: Business and Ecgnomics Section. 1959. Reports Hultgren, Thor. Cyclical Diversities in the Fortunes of Industrial Corporations. Occasional Paper No. 32. New York: National Bureau of Economic Research, Inc., 1950. New York Stock Exchange. The Stock Market Under Stress: The Evants of May 28II 29, and 31, 1962. New York, 1963. Securities and Exchange Commission. Stock Trading on the New York Stock Exchange on September 3, 1946. A Report Prepared by the Trading and Exchange Division. Wash- ington: Government Printing Office, 1947. A Study of Mutpal Funds. A Report Prepared by the Wharton School of Finance and Commerce. Washing- ton: Government Printing Office, 1962. Unpublished Material Hamada, Robert S. "An Analysis of Diffusion Indices of Insiders' Transactions." Unpublished Master's thesis, Massachusetts Institute of Technology, 1961. "IIIIIIIIIIIIIII