The Power of Pen and Executive Compensation

ARTICLE IN PRESS Journal of Financial Economics 88 (2008) 1–25 www. elsevier. com/locate/jfec The power of the pen and executive compensation$ John E. Corea, Wayne Guaya,A, David F. Larckerb a The Wharton School, University of Pennsylvania, Philadelphia, PA 19104, USA b Graduate School of Business, Stanford University, Stanford, CA 94305, USA Received 28 October 2005; received in revised form 20 March 2007; accepted 4 May 2007 Available online 5 December 2007 Abstract We examine the press’ role in monitoring and in? uencing executive compensation practice using more than 11,000 press articles about CEO compensation from 1994 to 2002.Negative press coverage is more strongly related to excess annual pay than to raw annual pay, suggesting a sophisticated approach by the media in selecting CEOs to cover. However, negative coverage is also greater for CEOs with more option exercises, suggesting the press engages in some degree of ‘‘sensationalism. ’â€℠¢ We ? nd little evidence that ? rms respond to negative press coverage by decreasing excess CEO compensation or increasing CEO turnover. r 2007 Elsevier B. V. All rights reserved. JEL classi? cations: G32; G34; J33; M41Keywords: Press; Media; Executive compensation; Corporate governance 1. Introduction With the possible exception of major accounting frauds (e. g. , WorldCom, Enron, etc. ), there are few topics that are more pervasive and produce bigger headlines in the business press than executive compensation. The debate about executive compensation tends to focus on the overall level of compensation (e. g. , relative to workers in the US or to executives in other countries), the rate of increase (e. g. , relative to in? ation or stock price returns), and the form of payment (e. . , stock options). Although there is extensive academic research on the determinants of executive compensation, there is little empirical evidence on the role of the popular and business press as a poten tial monitor of executive pay (e. g. , see Zingales, 2000; Bebchuk and Fried, 2004). The objective of our study is to provide insight into three questions: (1) What decision model does the media use to select chief executive of? cers (CEOs) for coverage about their compensation, (2) What determines the proportion of that coverage that is negative-toned, and (3) Do ? ms and managers ? nd this attention $ We thank Greg Miller, seminar participants at Stanford University, and an anonymous referee for their helpful comments. We also thank Jihae Wee for excellent research support, and appreciate ? nancial support from the Wharton School of the University of Pennsylvania and the Graduate School of Business at Stanford University. ACorresponding author. E-mail address: [email  protected] upenn. edu (W. Guay). 0304-405X/$ – see front matter r 2007 Elsevier B. V. All rights reserved. doi:10. 1016/j. j? neco. 2007. 05. 001 ARTICLE IN PRESS 2 J. E.Core et al. / Journal of Financial Ec onomics 88 (2008) 1–25 suf? ciently costly that they respond by making changes to their compensation or employment practices? Empirical evidence on these research questions provides insight into the role of the press in monitoring and in? uencing executive compensation practice. We examine a large sample of ExecuComp CEOs and an extensive collection of more than 11,000 press articles about CEO compensation from 1994 to 2002. Using an iterative key word search procedure, we partition the press articles based on whether they have a negative tone.Thus, for each CEO, in each year, we obtain a measure of the number of compensation articles and the fraction of these articles with a negative tone. We use this data to provide evidence on the press’ decision model and on the effect of press coverage on ? rms’ actions. Not surprisingly, the press chooses to cover CEOs with high total annual pay. We also ? nd that in deciding which CEOs to cover, the press does not appear to discriminate between CEOs that receive high expected pay versus CEOs that receive high excess pay, where excess pay is the residual from an expected compensation model that controls for standard economic determinants.Further, CEOs at large ? rms and ? rms with poor operating performance are also more likely to be selected for coverage. Conditional on the press deciding to cover a CEO’s compensation, we ? nd that negative coverage is more strongly related to measures of excess total annual pay than to raw total annual pay. We interpret this result as evidence that the press uses a relatively sophisticated approach when writing negative articles about CEO compensation. On the other hand, we also ? nd that negative coverage is related to the CEOs’ proceeds from option exercises. This latter ? ding is consistent with Holmstrom and Kaplan’s (2003) concern that one of the reasons the press portrays executive pay as a ‘‘runaway train’’ is t hat it misinterprets the payoff from exercised options as being a component of annual pay. In fact, the grant date value of options, not the payoff at exercise, is widely considered the more appropriate measure of option pay. 1 We ? nd little support for the hypothesis that the press serves as a catalyst or change agent for CEO compensation practices. Speci? cally, there is no consistent evidence that total compensation decreases after CEOs receive negative press coverage, and we ? d no evidence that negative press coverage of CEO compensation is related to CEO turnover. Thus, our results do not corroborate recent evidence that the media exerts an important in? uence on corporate governance choices (e. g. , Dyck and Zingales, 2002, 2004; Louis, Joe, and Robinson, 2004). The remainder of the paper consists of four sections. Section 2 provides a literature review and develops our research questions. Section 3 describes the sample selection and measurement choices. The results are pres ented in Section 4, and summary conclusions are provided in Section 5. 2. Background and research questions . 1. Determinants of media attention about CEO compensation Although there is considerable discussion about the role of disclosure and transparency in monitoring managerial behavior, the precise mechanisms for disclosing and disseminating information have received limited attention in the academic literature (Zingales, 2000). Dyck and Zingales (2002) argue that this limited attention stems from the small role that the diffusion of information plays in agency models. 2 They argue that the media is one vehicle through which information is aggregated and credibly communicated to the public (and across ? ms). Thus, the media can play a substantial role in reducing the costs of contracting parties for collecting and evaluating information, and in shaping the reputation of contracting parties. In order to provide insight into these questions, it is necessary to identify the objectiv e function of the media. As suggested by Jensen (1979), the approach to modeling the media industry is similar to any industry and begins with analyzing the demand faced by news producers (e. g. , newspapers, magazines, etc. ) and the 1 It is possible that the press justi? bly writes negative articles about CEOs with large realized option payoffs if the magnitude of option exercises re? ects a measure of cumulative excess compensation over a period of time. 2 In the accounting literature, diffusion of information plays a large role in research on the quality of accounting information disclosed by management to its shareholders, or in theoretical agency models incorporating channels of communication. However, there is little work on intermediaries, such as the press, that ? lter ? rm disclosures and disseminate information to the general stockholding public. ARTICLE IN PRESSJ. E. Core et al. / Journal of Financial Economics 88 (2008) 1–25 3 supply of news received by these pro ducers. Dyck and Zingales (2002) and Miller (2006) argue that there is a consumer demand for the investigative reporting role of the media, and Zingales (2000) hypothesizes that readers rely on this reporting to form opinions only when they believe the information provided to be accurate and reliable. In contrast, Jensen (1979) takes a more skeptical view of the media and suggests that most of the demand for news services derives not from a demand for information, but from a demand for entertainment.Since the news media’s competition under this scenario is sitcom television and tabloids, the media is expected to sensationalize news stories. Jensen further argues that the media will tailor news stories to take a negative tone about individuals that are out of favor with public opinion (e. g. , CEOs who are paid much more than their peers, or who have laid off large numbers of employees). Miller (2006) provides some initial empirical results that are broadly consistent with bot h of the above sources of demand for publicity.He examines a sample of 263 cases of Securities and Exchange Commission (SEC) Accounting and Auditing Enforcement Releases to investigate whether the press is a watchdog for accounting fraud. Consistent with information provision, Miller ? nds that the media provides the public with information about accounting fraud. However, consistent with sensationalism, he also ? nds that the media is more likely to ? ll the watchdog role for ? rms with a larger public following, ? rms with a richer information environment, and where the story is more likely to be sensational and interesting to the public.Miller also examines whether coverage is less negative for ? rms that do more advertising, but his results do not support this interesting proposition. Media coverage of executive compensation potentially satis? es both of the demand functions identi? ed above. Multi-million dollar pay packages, and the potential scandals surrounding the wealthy i ndividuals who receive high pay, can be very entertaining. For example, there were repeated references, and many negative references, in the press about Tyco International’s purchase of a $6,000 shower curtain for CEO Dennis Kozlowski’s corporate apartment.Similarly, there were repeated references, and many negative references, about the extensive perquisites paid to General Electric’s CEO, Jack Welch, that were disclosed in divorce proceedings after his retirement. On the other hand, if readers of the press demand media coverage about executive compensation that provides reliable information about potential governance problems, we expect that the media will identify and cover individuals who have ‘‘excessive’’ pay. That is, under this hypothesis, the media will not focus simply on large pay.Nor will it focus on large single components of pay such as stock option grants and cash payouts from bonus plans, or on large option exercises. Ex cess pay, de? ned as observed compensation less a measure of expected compensation derived from standard economic determinants, is known to be a sign of poor governance (e. g. , Core, Holthausen, and Larcker, 1999), and poor governance is clearly an important issue for shareholders, employees, suppliers, and society at large.Under this hypothesis, the media will not focus simply on large total pay (or option exercises) because it recognizes that large pay packages are optimal in settings where they re? ect the quality, performance, or bargaining power of the CEO. Thus, we predict that the media makes adjustments to a given CEO’s pay level to control for ‘‘normal’’ or ‘‘reasonable’’ pay, and that coverage of excess pay will primarily have a negative tone. We test this prediction with the following hypothesis: H1. Negative media coverage of CEO compensation is positively related to excess pay.However, if the primary source of demand is not from consumers seeking reliable information, but instead from consumers seeking entertaining news about highly paid executives, we expect that the media will sensationalize its stories. The press may satisfy this demand by writing negative articles about executives with high pay, regardless of whether circumstances are such that the high pay is reasonable. In this case, we view the negative coverage as ‘‘sensationalism,’’ and predict that negative press coverage is positively related to total pay without making adjustments for an expected level of pay given the CEO’s ability and performance.This sensationalism viewpoint provides a contrasting perspective to the ‘‘informing the public’’ notion underlying Hypothesis 1. Speci? cally, the press is predicted to provide negative coverage of high total pay (which is composed of expected pay given ?rm and CEO characteristics, plus excess pay). We propose the following hypothesis to test the ‘‘sensationalism’’ prediction: H2. Negative media coverage of CEO compensation is positively related to total pay (i. e. , related to both expected pay and excess pay). ARTICLE IN PRESS J. E. Core et al. Journal of Financial Economics 88 (2008) 1–25 4 Economists generally view the grant value of stock options as a more appropriate measure of CEO optionbased pay than ex post realized proceeds from multi-year grants. For example, consider a CEO who is granted stock options each year for ? ve years. If this CEO chooses to exercise all of these options in the ? fth year, it would be inappropriate to infer that the CEO received no option compensation in the ? rst four years when the options were granted, and substantial option compensation only in the ? th year when the options are exercised. However, exercise proceeds are a simple-to-understand, and easy-to-compute measure of the value realized by executives from options. And, in f act, a measure of total payout that includes option exercises rather than option grants is frequently cited in pay surveys in the ? nancial press (e. g. , see Forbes’ annual ? rankings of highest paid CEOs). 3 A sensationalism perspective (or possibly just na vete) suggests that the press may not discriminate between the CEO’s annual pay and large dollar proceeds realized by CEOs from options.To examine this hypothesis, we test the following: H3. Negative media coverage of CEO compensation is positively related to large dollar amounts realized from stock option exercises. In addition to our analysis of negative coverage of CEO compensation, we also examine general press coverage of compensation in order to distinguish the decision of the press to cover a story from the choice to produce a story with a negative tone. We do not formulate speci? c hypotheses about general coverage of pay, but rather include these results to provide descriptive vidence on how the press cho oses which CEOs to cover. We view the role of non-negative coverage of compensation as being somewhat unclear. For example, general coverage of total pay (both expected and excess compensation) might be informative for corporate governance purposes by providing benchmarks against which to compare CEO pay across ? rms. However, general coverage of total pay might be consistent with sensationalism, where readers ? nd articles about wealthy CEOs to be entertaining, and are not particularly concerned about whether their pay level is expected or excessive. . 2. In? uence of the media on CEO compensation Dyck and Zingales (2002) argue that there are at least three ways in which media attention can affect the reputations of ? rms and their of? cers and directors, and play a role in corporate governance. First, media attention on ? rms with weak corporate governance can drive politicians and regulators to enact legislation to reform or enforce corporate law, especially if they believe that failure to do so would hurt their political careers or cause public outcry.The recent media attention given to stock option backdating, and the consequent regulatory interest, could be thought of as an example of this type of activity. 4 Second, negative media attention on managers and directors can call into question whether these individuals are good decision makers who attend to the interests of their shareholders and employers. Fama and Jensen (1983) make a similar argument that the value of managers’ and directors’ human capital depends primarily on signals about their performance as decision makers within corporations.Thus, if negative media attention damages managers’ and directors’ reputations, it can reduce the value of these individuals in the labor market. Finally, Dyck and Zingales (2002) argue that negative media attention can hurt the reputations of managers and directors within their communities and impose social costs on both them and thei r families. 5 Dyck and Zingales (2002, 2004) also provide evidence in an international setting that the media plays a role in corporate governance and in? uences ? rms’ behavior. Their primary ? ndings are that the private bene? s of control are smaller and the responsiveness of the private sector to environmental issues is greater in countries with larger newspaper circulation. 3 Executive bonuses are generally measured in compensation studies at payout values rather than ex ante values. Ideally, one would measure both option pay and bonus pay at the grant date expected value of the pay. However, although data are readily available to estimate grant date option values, it is dif? cult to estimate the expected value to the executive from a given bonus plan. 4 For example, see Heron and Lie (2007).Also see The Wall Street Journal online at: http://online. wsj. com/public/resources/documents/ info-optionsscore06-full. html, which lists corporations that have come under SEC and Justice Department scrutiny for possible option backdating. We last accessed this website on February 23, 2007. 5 In our study, we do not distinguish between these three channels of media in? uence. For our purposes, it is only important that negative media attention about CEO compensation can impose costs on ? rms and their CEOs. ARTICLE IN PRESS J. E. Core et al. Journal of Financial Economics 88 (2008) 1–25 5 Two additional papers are related to our research question. Johnson, Porter and Shackell (1997) examine changes in compensation from 1993 to 1994 for a sample of 186 CEOs to investigate whether CEO compensation is sensitive to stakeholder pressure. They ? nd that the existence of a negative tone article in any one of ? ve leading periodicals is associated with a smaller increase in total CEO pay from 1993 to 1994 and an increase in the sensitivity of cash pay to ? rm performance. However, as we demonstrate in Section 4. 2, this ? ding is confounded by strong mean reve rsion in pay among the general population of highly paid CEOs (i. e. , when a CEO has high pay in year t, there is a natural tendency for pay to be lower in year t+1). Moreover, highly paid CEOs are also more likely to receive media attention. Therefore, CEOs that draw media attention are more likely to experience mean reversion in pay, but this relation may not be causal. Finally, Louis, Joe, and Robinson (2004) provide some evidence that negative Business Week coverage regarding institutional investors’ assessment of board effectiveness in? ences boards’ actions. In particular, the boards identi? ed as worst are more likely to replace CEOs and board chairs, to separate the CEO and chair functions, and to increase the number of outside board members. However, it is not clear from these ? ndings whether the boards’ actions are due to media coverage or due to pressure from unsatis? ed institutional investors. If negative media coverage damages the reputations and human capital of managers and directors, ? rms will respond to this negative coverage by taking steps to avoid further coverage in the future.However, the nature of the responses that the ? rms might take is not clear. If the media acts as a good watchdog over executive pay, and if its negative coverage primarily serves to provide investors and the public at large with reliable information about excess pay, we expect ? rms to respond by reducing excess CEO pay. 6 An even more severe response would be to terminate the CEO to avoid future negative media coverage of that CEO and his compensation. To gain insight into the outcomes of negative media coverage, we test the following hypotheses: H4.CEO compensation declines following negative media coverage. H5. CEO turnover increases following negative media attention. As noted above, it is also possible that the media’s coverage of CEO pay serves to entertain readers with sensational stories. In this case, we expect that ? rms eit her take no action (and bear the brunt of any reputation damage) or make ‘‘cosmetic’’ adjustments to avoid negative media attention in the future. An example of a ‘‘cosmetic’’ change would be for the CEO to alter the pattern of his stock option exercises.If the media sensationalizes compensation stories by including the proceeds from option exercises in the computation of executive pay, CEOs may avoid exercising options for a few years or ‘‘smooth out’’ option exercises after the negative publicity. We test the following research hypothesis: H6. Option exercises decline following negative media attention. 3. Sample selection and variable measurement Our initial sample consists of all ExecuComp CEOs from ? scal years 1993 to 2001. For a CEO to be included in the ? nal sample, we require that we can match the ? m to the Center for Research in Securities Prices (CRSP) database, that CEO tenure is available in ExecuComp, and that the CEO is in of? ce at the end of the ? scal year. Second, we require non-missing data on CEO compensation and on the variables that we use to estimate our model for excess compensation and press coverage (described below). Finally, we require that the ? rm name and CEO name can be matched to the Factiva news source database. 7 These data requirements yield a sample of 12,090 CEO-year observations from 1993 to 2001.The sample contains 3,126 different CEOs at 2,052 different companies. The summary results in Table 1 show that the number of CEOs in the sample grows slightly over time (as ExecuComp coverage increases). Consistent with other ? ndings using ExecuComp data (e. g. , Hall and Murphy, 2002), we ? nd that CEO total compensation increases substantially over the period, and at a greater percentage growth rate than ? rm sales. In addition, there is a monotonic increase in the average level of total 6 As we discuss below, ? rms will respond to unanticipated negative coverage by reducing future pay.To the extent that ? rms anticipate the costs of negative media coverage, they will reduce current pay to avoid these costs. 7 Factiva is a joint venture between Dow Jones and Reuters. ARTICLE IN PRESS J. E. Core et al. / Journal of Financial Economics 88 (2008) 1–25 6 Table 1 Trends in CEO compensation and compensation-related press coverage Year N 1993 1,203 1994 1,250 1995 1,305 1996 1,316 1997 1,327 1998 1,392 1999 1,389 2000 1,443 2001 1,465 Percentage change from 1993 to 2001 Total compt (thousands) SalestA1 (millions) Number of articles per CEOt+1 Percentage ofCEOs with coveraget+1 Fraction of CEO compensation articles with negative tonet+1 (%) 1,176 1,345 1,378 1,605 1,859 1,972 2,248 2,578 2,632 124% 883 859 872 950 959 936 1,058 1,061 1,162 32% 0. 27 0. 35 0. 47 0. 85 1. 01 1. 05 0. 98 1. 12 2. 23 724% 0. 09 0. 12 0. 13 0. 21 0. 22 0. 24 0. 23 0. 26 0. 38 302% 43 32 37 31 34 32 30 28 31 A28% The data consist of ExecuComp CEOs from ? scal years 1993 to 2001. The articles on CEO compensation are obtained from the Factiva database for the year after pay was earned, that is years 1994 to 2002. N is the sample size for that year.Total Compt is the sample median salary, bonus, long-term incentive plan payouts, the value of restricted stock grants, the value of options granted during the year, and any other annual pay (in $000s) in the ? scal year shown. SalestA1 is the sample median ? rm sales for year tA1. Number of Articles per CEO is the sample average total number of articles written about the CEO’s compensation in the Factiva database in the ? scal year t+1 after pay was earned. Percentage of CEOs with Coveraget+1 is the percentage of CEOs for whom the press covers CEO compensation.Fraction of CEO compensation articles with negative tonet+1 is the total number of negative articles written about the CEOs’ compensation (using the algorithm described in the text to measure negative tone) as a percentage of the total number of articles written about the CEOs’ compensation. press coverage of CEO pay and in the proportion of CEOs who receive coverage. However, conditional on receiving coverage, the proportion of coverage that is negative is relatively constant over time (we describe the measurement of these publicity variables below). . 1. Measurement of press coverage and negative press coverage We measure publicity about CEO compensation by gathering all articles related to the CEO’s compensation from the Factiva database in the ? scal year after the compensation was earned (for example, for a ? rm with a ? scal year ending June 30, 2001, where CEO compensation is typically disclosed in the proxy statement in August or September of 2001, we would match articles published during the next ? cal year ended June 30, 2002). We include all major news and business publication sources on Factiva with the exception of the press release wires through which ? rms initi ate the release of information, such as PR Newswire, FD Newswire, and Business Wire. Similar to Francis, Huang, Rajgopal, and Zang (2004), we use the company identi? er in Factiva to locate articles covering a speci? c ? rm. We then locate articles written about the CEO’s compensation through the following search: CEO NAME or CEO NAME’S) near20 (compensation or salary or bonus or option* near10 grant or option* near10 receiv* or option* near10 exercis* or restricted stock or (pay near5 00) or (was paid near5 00) or (pay near5 million*) or (was paid near5 million*)) and (CEO NAME or CEO NAME’S) same (compensation or salary or bonus or option* near10 grant or option* near10 receiv* or option* near10 exercis* or restricted stock or (pay near5 00) or (was paid near5 00) or (pay near5 million*) or (was paid near5 million*)) The objective of this free text search is to identify all articles in which the CEO’s compensation is described in either a positive, nega tive, or neutral fashion. We count each article as a single observation, regardless of the number of times a CEO’s name or compensation is mentioned in the article. 8 As described in the Factiva Inside-Out Reference Guide, ‘‘near20’’ locates words within 20 words of the CEO’s name and ‘‘same’’ locates words in the same paragraph as the CEO’s name. ARTICLE IN PRESS J. E. Core et al. / Journal of Financial Economics 88 (2008) 1–25 7 To measure negative publicity about CEO compensation, we iteratively develop a Perl program to process the text of each article about CEO compensation to assess whether the article has a negative tone. The input into the Perl program consists of a set of negative tone keywords and phrases.This set of keywords and phrases was developed from manually reading approximately 200 articles about CEO compensation, where the articles included both randomly selected ? rms and ? rms widel y known to have received negative publicity (e. g. ; Tyco international and Citigroup). 9 In order to validate and improve the Perl algorithm, we applied the search string to articles for a random sample of 50 CEOs, and we allowed the algorithm to classify the articles as having either a negative or non-negative tone. We then read these same articles and manually assigned each as having either a negative or non-negative tone. To identify errors in the Perl algorithm, we compared the two sets of coded negative tones using a contingency table of manual partitioning versus computer partitioning. Based on the classi? ation errors, we adjusted the keyword search to improve the ? t of the search string within this 50 CEO sample. To check the validity of these adjustments, we applied the improved negative tone Perl algorithm string to articles for an independent random sample of 50 CEOs. We again read and partitioned the articles for this second random sample and constructed another contin gency table to assess accuracy. This manual partitioning identi? ed 18% (82%) of the articles as negative tone (non-negative tone). The automated Perl keyword search correctly identi? ed 75% of the non-negative tone articles and 54% of the negative tone articles. Further, the manual partitioning identi? ed 25% (75%) of the ? m-years as having at least one negative tone article (no negative tone article). The Perl algorithm correctly identi? ed 63% of the ? rm-years without negative tone articles and 77% of the ? rm-years with negative tone articles. The fact that the classi? cation rates are less than 100% con? rms that there is measurement error in our search string (in Section 4. 1, we show in sensitivity analysis that this measurement error does not appear to affect our inference). We use the revised search string to identify negative tone articles for the full sample of CEO compensation articles (‘‘NEGATIVE’’). Appendix A shows our ? nal negative tone s earch string.In order to provide some descriptive information about our search string, Appendix B contains excerpts from two articles about the 2001 compensation package for E*Trade Financial Corporation’s CEO, Christos Cotsakos. Both articles were published on May 1, 2002. The ? rst article from The New York Times reports the salary, bonus, equity, and other components of Cotsakos’ pay package without taking a view as to whether the pay package is excessive or unreasonable. We classify this article as having a non-negative tone. The second article from The Wall Street Journal also reports the components of Cotsakos’ pay package but takes a negative tone by calling the compensation an ‘‘outsize package’’ and referring to Cotsakos as the ‘‘highestpaid CEO on Wall Street. ’ The keyword ‘‘outsize’’ within a few words of ‘‘salary’’ and/or ‘‘bonus,’â₠¬â„¢ and the keyword ‘‘highest’’ within a few words of ‘‘pay’’ are both triggers for our keyword search that classify this article as having a negative tone. However, note the title of the second article, ‘‘No Discount: E*Trade CEO Gets Pay Deal of $80 Million. ’’ Although this title clearly has a negative tone, the ‘‘play on words’’ nature of the text prevents us from ? agging this title as negative tone with our Perl search string. In this case, the body of the article is suf? cient to categorize the article as negative tone. We acknowledge that it is dif? cult to construct a completely accurate search string and that our negative tone classi? cation inevitably measures true negative tone with error. However, a sensitivity analysis summarized below in Section 4. suggests that our inference using the negative publicity measure in the full sample is not induced by measurement erro r. The time-series statistics on the number of compensation-related articles for our sample CEOs over the period 1994 to 2002 is reported in Table 2 (Panel A). 10 The number of compensation-related articles grew rapidly from 325 to 3,263 (Column 3), an increase of about 900%. However, at the same time, the total number of articles across all topics grew from 216,677 to 825,887 (Column 1), an increase of about 280%. Similarly, the number of news sources covering CEO compensation grew from 62 to 470 (Column 2), a rise of 9 As illustrated in Appendix A, the ? al negative tone search string consists of approximately 150 keywords and phrases, such as ‘‘high pay,’’ ‘‘excess pay,’’ and ‘‘generous options. ’’ For most of the phrases, we allow for the possibility that the keywords do not immediately precede or follow each other, and may be several words apart in the text. We also allow for different characterizatio ns of the same word (e. g. , ‘‘large bonus,’’ ‘‘larger bonus,’’ ‘‘and largest bonus’’). 10 Since our sample data on CEO compensation covers the time period from 1993 to 2001, the articles for the year following the compensation are collected from 1994 to 2002. ARTICLE IN PRESS J. E. Core et al. / Journal of Financial Economics 88 (2008) 1–25 8 Table 2Annual data on the source of articles on CEO compensation Panel A. Trends in Articles about CEO Compensation Year Number of Number of articles–all sources–CEO topics compensation articles (1) (2) 1994 1995 1996 1997 1998 1999 2000 2001 2002 Percentage change from 1994–2002 216,677 196,032 178,378 233,665 303,850 543,058 514,747 542,096 825,887 281% 62 97 131 234 244 279 308 323 470 658% Panel B. Major sources and tone of coverage Type of source Source and Their Sources Number of CEO compensation articles (3) 325 439 609 1,117 1,346 1,465 1,362 1,616 3,263 904% Fraction of CEO compensation articles with negative tone (4) Number of WSJ articlesFraction of WSJ articles with negative tone (5) (6) 43% 32% 37% 31% 34% 32% 30% 28% 31% A28% 58 74 112 104 122 149 44 81 210 262% 48% 45% 43% 38% 39% 38% 25% 25% 40% A17% Number of CEO compensation articles Number of negative tone CEO compensation articles Fraction of CEO compensation articles with negative tone Newswire AP Dow Jones Reuters Sub-total 235 717 1,271 2,223 75 137 279 491 32% 19% 22% 22% Newspaper Chicago Sun-Times Financial Times New York Times The Globe And Mail The Washington Post USA Today Wall Street Journal Sub-total 110 252 260 190 123 49 954 1,938 29 99 88 49 49 22 367 703 26% 39% 34% 26% 40% 45% 38% 36% Magazine Barron’s Business WeekForbes Fortune Sub-total 44 43 43 40 170 27 21 15 17 80 61% 49% 35% 43% 47% The sample consists of ExecuComp CEOs from ? scal years 1993 to 2001. The articles on CEO compensation are obtained from the Factiva data base for years 1994 to 2002, including the source of each article. Number of articles—all topics is the total number of articles for all sample ? rms for each year. Number of sources—CEO compensation articles is the total number of different publications that printed an article about CEO compensation for each year. Number of CEO compensation articles is the total number of articles about CEO compensation for all sample ? rms for each year.Fraction of CEO compensation articles with negative tonet+1 is the total number of negative articles written about the CEOs’ compensation (using the algorithm described in the text to measure negative tone) as a percentage of the total number of articles written about the CEOs’ compensation. Number of WSJ articles is the total number of The Wall Street Journal (WSJ) articles on CEO compensation for our sample, and fraction of compensation articles with negative tonet+1 is the percentage of WSJ articles with negative tone (using the algorithm described in the text to measure negative tone). Number of negative tone WSJ articles is the number of articles where negative tone is assessed using the algorithm described in the text. ARTICLE IN PRESS J. E.Core et al. / Journal of Financial Economics 88 (2008) 1–25 9 about 660%. To explore whether the growing number of compensation-related articles is primarily due to the growth in the number of articles and sources, we present time-series data for The Wall Street Journal, one of the largest sources. As might be expected, The Wall Street Journal released a growing number of compensation-related articles over this period. The total number of articles for this source was 210 in 2002 compared to 58 in 1994 (Column 5), an increase of about 260%. Thus, the increase in articles does not appear to be simply caused by the increase in sources covered by Factiva.The fraction of negative tone compensation articles across all sources has remained a fairly constan t fraction of total articles, with a yearly average of about 33% (Column 4). The last column in Table 2 (Panel A) shows that a somewhat larger fraction of the compensation articles written by The Wall Street Journal are negative, with a yearly average of about 38%. This suggests that some news agencies, as a matter of strategy or reporting orientation, are more likely than others to publish compensation articles with a negative tone. To explore compensation coverage across news sources, we tabulate article counts separately for many of the major sources in Table 2 (Panel B). We classify major news sources as newswires, newspapers, or magazines.The main newswires, Associated Press, Dow Jones and Reuters, provide the greatest number of compensation-related articles, but have the lowest fraction of negative tone compensation articles, at about 22%. This latter ? nding is perhaps not surprising given that newswires tend to capture company press releases. The major newspapers (The Wall S treet Journal, The New York Times, Financial Times, etc. ) supply the second highest fraction of negative tone articles, at 36%. The largest fraction of articles with a negative tone, at about 47%, is written by magazines (Fortune, Business Week, etc. ). This ranking of negative tone coverage potentially re? cts a greater tendency by the papers and magazines to sensationalize stories in order to sell copies, presumably due to differences in their subscriber base and marketing techniques. In the ? rst two rows of Table 3, we provide descriptive data on compensation-related articles by CEO-year. In this table, and in our data analysis in Tables 6–9, we mitigate the in? uence of outliers by setting the upperand lower-most percentiles for our variables equal to the values at the 1st and 99th percentiles in each year, respectively. Media coverage is skewed, with the median CEO receiving no articles about his compensation in a given year. In 21. 6% of the CEO-years, at least one ar ticle was published about the CEO’s compensation, and the 10% of the CEO-years with the greatest media coverage received at least two articles.Negative media coverage is skewed to an even greater extent, with only 10. 0% of the CEO-years receiving at least one compensation article with a negative tone. In 1% of the CEO-years, at least four negative tone articles were written about the CEO’s compensation. For the 2,607 observations in which the CEO has some coverage of his compensation, 47% of the CEOs have at least some negative-toned coverage, and 28% of the compensation articles have a negative tone. 3. 2. Control variables and model of expected press coverage Our main objective is to better understand the determinants of press coverage about executive compensation, and in particular, negative coverage about executive compensation.The results in Tables 1 and 3 reveal that only a subset of CEOs attracts press coverage on their reported compensation. Among the CEOs tha t attract coverage, there is substantial variation in the degree of negative comments about their pay, as proxied by the proportion of the coverage that is negative. To address this empirically, we ? rst model the media’s choice of whether to cover a CEO with the following probit model: E? Prob? Coverageit? 1 ? F? go ? g1 Compensationit ? g2 Controls?. (1) For those CEOs who receive coverage, we model the proportion of the coverage that is negative with the following general linear model: E ? % of Negative Articlesit? 1 jCoveraget? 1 ? ? G? bo ? b1 Compensationit ? b2 Controls?. (2) The dependent variable in Eq. 2) is a fraction bounded between 0 and 1. We follow Papke and Wooldridge (1996) and estimate Eq. (2) using a general linear model (GLM) in which the link function is logistic. Papke ARTICLE IN PRESS 10 J. E. Core et al. / Journal of Financial Economics 88 (2008) 1–25 Table 3 Descriptive statistics Variable Mean Std Dev P1 Q1 Median Q3 P90 P99 Number of article st+1 Coveraget+1 Number of negative articlest+1 % of negative articlest+1 Number of ? rm articlest+1 Total compt Total payoutt Tenuret S&P500t SalestA1 Bk/MkttA1 RETt ROAt 0. 81 0. 22 0. 23 0. 28 293. 85 3,746 3,122 7. 60 0. 33 3,280 0. 65 0. 20 0. 04 2. 79 0. 41 0. 92 0. 37 669. 29 6,237 6,587 7. 45 0. 47 6,296 0. 7 0. 61 0. 10 0. 00 0. 00 0. 00 0. 00 1. 00 189 117 0. 08 0. 00 17 0. 11 A0. 75 A0. 37 0. 00 0. 00 0. 00 0. 00 52. 00 904 659 2. 17 0. 00 353 0. 44 A0. 13 0. 01 0. 00 0. 00 0. 00 0. 00 116. 00 1,758 1,246 5. 33 0. 00 980 0. 66 0. 11 0. 05 0. 00 0. 00 0. 00 0. 50 244. 00 3,822 2,736 10. 58 1. 00 2,989 0. 86 0. 38 0. 09 2. 00 1. 00 1. 00 1. 00 599. 00 8,334 6,675 16. 92 1. 00 8,775 0. 98 0. 76 0. 14 16. 00 1. 00 5. 00 1. 00 3,856. 00 32,909 37,109 35. 92 1. 00 34,654 1. 20 2. 34 0. 25 This table presents descriptive statistics for the variables used in the subsequent analyses. The sample consists of 12,090 observations for ExecuComp CEOs from ? cal years 1993 to 2001. The a rticles on CEO compensation are obtained from the Factiva database for years 1994 to 2002. Number of Articlest+1 is the total number of articles written about the CEO’s compensation. Coveraget+1 is an indicator variable for whether the press covers CEO compensation. Number of Negative Articlest+1 is the total number of negative tone articles written about the CEO’s compensation, where negative tone is assessed using the algorithm described in the text. % of Negative Articlest+1 is Number of Negative Articlest+1 divided by Number of Articlest+1. This variable is tabulated only for the 2607 observations with Coveraget+1 greater than zero.Number of Firm Articlest+1 is the number of articles (all topics) written about the ? rm during year t+1. Total Compt is salary, bonus, long-term incentive plan payouts, the value of restricted stock grants, the value of options granted during the year, and any other annual pay for the CEO in year t. Total Payoutt is salary, bonus, long- term incentive plan payouts, the value of restricted stock grants, the proceeds from options exercised during the year, and any other annual pay for the CEO in year t. Tenuret is the CEO’s tenure in years at the end of year t. S&P500t is one if the ? rm is in the S&P500 at the end of year t, and zero otherwise. SalestA1 (in millions of dollars) is ? rm sales for year tA1.Bk/MkttA1 is (book value of assets)/(book value of liabilities+market value of equity) at the end of year tA1. RETt is the ? rm’s return for the year t. ROAt is income before extraordinary items divided by average total assets for the year t. and Wooldridge show that this estimator is consistent when the dependent variable is a proportion ranging from 0 to 1, and when there may be a mass of observations at 0 and 1. 11 We note that in these models, coverage and negative coverage are measured in the ? scal year t+1 following determination of compensation in year t. 12 This lessens the chance of a simulta neity bias, in which realized negative coverage causes reductions in realized pay. However, as we discuss further below, if ? ms anticipate that future negative coverage can be very costly, they may reduce current pay in order to avoid future coverage. We expect that publicity about CEO pay derives not only from the magnitude and components of CEO pay, but also from general determinants of press coverage. Therefore, we control for the determinants of publicity that are not directly related to CEO compensation. To our knowledge, an accepted model for the expected level of press coverage related to CEO pay does not exist. As a starting point, we include log(Number of Firm Articles) as a control variable for general ? rm-speci? c press coverage across all topics, where log(Number of Firm Articles) is measured for each ? m-year as the natural logarithm of the total number of articles that mention the ? rm across all major news and business publication sources on Factiva, excluding newsw ires that primarily carry company-initiated 11 We obtain the same inference if we instead estimate a linear model for the fraction using ordinary least squares (OLS). If we estimate an OLS model for the fraction and include a Heckman (1979) correction for the predictability of the coverage decision in Eq. (1), we obtain the same inference. The Heckman correction is not signi? cant in any of our models, which suggests that results are robust to ignoring the selection in the second-stage model. 12Base salary, option and restricted stock grants, and the majority of compensation are determined and paid during the ? scal year. The one exception is cash bonuses, which are determined early the next ? scal year after results are known. However, the bonus amounts tend to be small compared to option and restricted stock grants. ARTICLE IN PRESS J. E. Core et al. / Journal of Financial Economics 88 (2008) 1–25 11 disclosures. 13 We also expect that ? rm size is a key determinant of publ icity (see Jensen, 1979; Miller, 2006). Press coverage of large ? rms will have broader appeal as these ? rms are more likely to be household names and to have larger customer and shareholder bases. At the same time, large ? rms may e able to impose costs on media ? rms that cover them in a negative light. These costs may come in the form of withholding valuable news stories or withholding advertising dollars. 14 We use two variables to control for ? rm size and likelihood of broad appeal: the logarithm of each ? rm’s sales revenues (‘‘Sales’’) and membership in the S&P 500 (‘‘S&P500’’). Jensen (1979) argues that the media is more likely to write a negative article when the individual under scrutiny has lost popularity with the public. We include recent ? rm performance in our regressions to control for the possibility that the CEO has fallen out of favor with the public.We measure ? rm performance using contemporaneous and lagged stock returns obtained from CRSP (‘‘RET’’) and accounting performance obtained from Compustat (‘‘ROA’’) which is computed as net income before extraordinary items divided by average assets. To allow for the possibility that press coverage is more sensitive to negative performance than to positive performance, we include separate variables for negative (‘‘NEG’’) and positive (‘‘POS’’) stock return and accounting performance. 15 We also include CEO tenure (‘‘Tenure’’) as a control variable because we expect that it may take time for the press to become interested in covering a new CEO.Finally, we expect that press coverage varies across different calendar years and sectors of the economy. To capture this effect we include indicator variables for two-digit SIC code and calendar year in our model. 3. 3. Measurement of compensation variables and excess compensation As described in Section 2, we expect that publicity may be in? uenced by total annual compensation. We measure Total Comp as the sum of salary, bonus, long-term incentive plan payouts, the value of restricted stock grants, the value of options granted during the year, and any other annual pay. This is the most common measure of total pay in the academic literature.We hypothesize in Hypothesis 3 that press coverage could also be affected by realized option exercise proceeds as opposed to option grant value. To test this hypothesis, we construct a measure of total realized payouts to the CEO, Total Payout, computed as the sum of salary, bonus, long-term incentive plan payouts, value of restricted stock grants, proceeds from options exercised during the year, and any other annual pay. This measure of total realized payout is common in the media (e. g. , see Forbes’ annual rankings of highest paid CEOs). 16 We obtain our compensation data from ExecuComp. Descriptive statistics for the compensation variables are presented in Table 3. The mean Total Comp is $3. 7 million, and the mean Total Payout is $3. 1 million.However, the values in the extreme percentile of Total Payout are somewhat greater than those for Total Comp. In addition to these raw compensation variables, we also construct a measure of excess CEO compensation to investigate whether the media appears to make adjustments for a ‘‘normal’’ level of compensation when writing an article with a negative tone. We measure excess compensation as actual compensation minus expected compensation. Our benchmark model for expected compensation follows prior research in this area (e. g. , Smith and Watts, 1992; Core, Holthausen, and Larcker, 1999; Murphy, 1999), and is obtained by regressing the natural logarithm (Log) of compensation on proxies for economic determinants of CEO compensation, such as ? m size, growth opportunities, stock return, accounting return, and indu stry controls: Log? Compensationit ? ? a ? xit b ? uit 13 (3) In the 68 ? rm-years with no articles on Factiva, we set Number of Firm Articles equal to one to avoid losing the observations. The costs of withholding valuable news from the press may apply not only to large ? rms but also to growing ? rms with rich information environments that are engaging in substantial investments, acquisitions, or product developments. At the same time, growth ? rms may also have broader appeal to the public than stable or declining ? rms. Our regressions are robust to including book-to-market as a control variable for ? rms’ investment opportunities. 15Dial and Murphy (1995) raise the possibility that unpopular operational decisions draw media attention. For example, in their case study of General Dynamics, the press strongly criticized the CEO for receiving a bonus payout after the stock price responded positively to his decision to lay off thousands of employees. We examine this possibili ty in Section 4. 3. 16 Total Payout also has preferable econometric properties as compared to using only the proceeds from option exercises. Speci? cally, an option exercise variable has a large mass at zero, whereas Total Payout has a positive value for all cases. 14 ARTICLE IN PRESS J. E. Core et al. / Journal of Financial Economics 88 (2008) 1–25 12 here Compensationit is Total Comp or Total Payout as described in Section 3. 3, and xit consists of Log(Tenure)it, Log(Sales)itA1, S&P500itA1, Book-to-marketitA1, RETit, RETitA1, ROAit, ROAitA1, and Industry controlsit. Book-to-market is (book value of assets)/(book value of liabilities+market value of equity), and the other independent variables are de? ned above. We estimate Eq. (3) using OLS. We estimate Expected Compensation by exponentiating the expected value of Eq. (3). We compute Residual(Compensation) by estimating expected Compensation and subtracting it from Compensation: Residual? Compensationit ? ? Compensationit A Expected Compensationit . (4) We compute %Residual Compensation as: Residual? Compensationit ? ? log? Compensationit ? A log? Expected Compensationit ?. (5) Although we estimate Eq. (2) using annual cross-sectional regressions, in the interest of brevity, we present the results of a pooled cross-section, time-series estimation of Eq. (2) with year indicators in Table 4. Consistent with prior research, we ? nd that all measures of compensation exhibit the expected positive associations with ? rm size, growth opportunities, and stock returns. The coef? cient estimates for the annual regressions are substantively similar to those reported in Table 4. Table 4 Regressions for compensation variables Dependent variableIndependent variable Log(total compt) Log(total payoutt) Log(tenure)t A0. 02 (A0. 80) 0. 42*** (17. 96) 0. 12** (2. 30) A0. 99*** (A9. 76) 0. 27*** (12. 84) 0. 16*** (6. 71) A1. 00*** (A5. 87) A0. 45** (A2. 07) 0. 4290 0. 13*** (6. 93) 0. 40*** (18. 74) 0. 14** (2. 83) A0. 6 9*** (A6. 80) 0. 31*** (11. 64) 0. 26*** (19. 23) 0. 40* (1. 98) A0. 51* (A1. 72) 0. 4274 Log(sales)tA1 S&P500t Bk/MkttA1 RETt RETtA1 ROAt ROAtA1 R2 This table presents results of pooled cross-sectional OLS regressions for the logarithms of two measures of CEO compensation and the economic determinants of compensation. The sample consists of 12,090 observations for ExecuComp CEOs from ? cal years 1993 to 2001. Total Compt is salary, bonus, long-term incentive plan payouts, the value of restricted stock grants, the value of options granted during the year, and any other annual pay for the CEO in year t. Total Payoutt is salary, bonus, long-term incentive plan payouts, the value of restricted stock grants, the proceeds from options exercised during the year, and any other annual pay for the CEO in year t. Log(Tenure)t is the logarithm of the CEO’s tenure in years at the end of year t. Log(Sales)tA1 is the logarithm of ? rm sales for year tA1. S&P500t is one if the ? rm is in th e S&P500 at the end of year t, and zero otherwise.Bk/MkttA1 is (book value of assets)/(book value of liabilities+market value of equity) at the end of year tA1. RETt is the ? rm’s return for year t. RETtA1 is the ? rm’s return for year tA1. ROAt is income before extraordinary items divided by average total assets for year t. ROAtA1 is income before extraordinary items divided by average total assets for year tA1. Fixed effects for year and 2-digit SIC codes are included in the regressions, but not tabulated. T-statistics using Huber-White robust standard errors are presented in parentheses below coef? cient estimates. *, **, and *** indicate two-tailed statistical signi? cance at 10, 5, and 1 percent levels, respectively. ARTICLE IN PRESS J. E. Core et al. Journal of Financial Economics 88 (2008) 1–25 13 3. 4. Illustrations from the sample Table 5 (Panel A) lists the ten CEOs with the greatest amount of coverage (i. e. , greatest number of articles) about their compensation in any given year during our sample period. The compensation and ? rm characteristic variables are provided for the year prior to the press coverage variables (thus, the year t+1 designation on the press variables). These CEOs had between 87 and 320 compensation-related articles, as well as very substantial negative press coverage, as measured by either fraction of articles that are negative, or number of articles that are negative.The percentage of negative articles in this group of CEOs ranges from 32% to 73%, whereas the sample average is 28% (see Table 3). CEOs with a large number of compensation-related articles tend to manage large, poor performing ? rms. Seven out of the ten ? rms have market capitalization of $20 billion or more, and three-year market-adjusted returns are negative for all of the ten ? rms. Dennis Kozlowski of Tyco International received the most compensation-related articles in 2002 with 320, as well as the most negative articles (57% or 183 neg ative articles). His total compensation in 2001 was $77. 8 million with substantial estimated excess compensation. 17 Five of these ten CEOs had positive excess total pay in the year prior to the publicity.However, excess compensation during the prior year was not the obvious instigator of the press coverage for some of these CEOs. For example, Sanford Weill, CEO of Citigroup, received 178 compensation-related articles (40% of which were negative), but had negative excess total pay. At the same time, Mr. Weill had a combination of fairly large raw compensation at $16. 6 million, substantial option exercises, poor three-year market-adjusted stock return performance (A44%), and a history of prior media attention for being among the higher paid CEOs. Similarly, Carly Fiorina, CEO of Hewlett Packard, received 168 articles in 2002 (32% of which were negative), but had lower than expected pay in 2001.However, although she had negative excess compensation, Ms. Fiorina was the recipient of considerable criticism about Hewlett Packard’s sub-par performance as evidenced by Hewlett Packard’s market-adjusted stock return of A68% from 2000 to 2002. Another interesting example is Thomas Siebel of Siebel Systems, Inc. , who drew 132 articles and 65 negative articles about compensation in 2003, and yet received no pay in 2002. However, Mr. Siebel exercised a substantial dollar amount of options in 2002 (as well as in 2001), and also received a very large grant of new options in 2001. Siebel Systems also had extremely poor three-year market-adjusted stock price performance at A123%.Table 5 (Panel B) lists the ten CEOs with the greatest percentage of negative articles in any given year during our sample period (i. e. , the number of negative articles about compensation divided by the total number of articles about compensation). We restrict our attention to ? rms that have at least four articles on CEO compensation, because there are many CEOs with only one or two compensationrelated articles, and where 100% of these articles are negative. The ? rms in Panel B are generally much smaller than those reported in Panel A which suggests that the total volume of press coverage is related to ? rm size. The results suggest a mixture of explanations for a high percentage of negative articles.The CEOs at Bear Stearns, EOG Resources, and Warnaco Group received very large total and residual compensation, and the CEO of Micron received a large stock option payout in the year of negative press coverage. The CEOs of Hillenbrand Industries, Nike, and Federal-Mogul received relatively modest levels of total compensation, and the negative press coverage seems to be due to their large negative marketadjusted returns. The explanations for Delphi Financial and Manpower are not clear, as both of these companies have low relative total compensation, no stock option payouts, and reasonable market-adjusted returns. It is also useful to examine some features of negat ive publicity for CEOs selected on the basis of large excess compensation.For example, an examination of the ten CEOs in 2001 with the greatest excess total direct compensation indicates that eight out of the ten CEOs received some negative publicity in 2002 (not tabulated). Interestingly, some of these excessively paid CEOs received no media attention. Greg Reyes, CEO of Brocade Communications, received about $370 million in total direct compensation, primarily due to a grant of more than 10 million stock options. However, even though he received the greatest amount of excess pay, Mr. Reyes received no negative publicity (although he was subsequently accused of 17 In Table 4, we do not winsorize any of the variables being shown. ARTICLE IN PRESS J. E. Core et al. / Journal of Financial Economics 88 (2008) 1–25 14Table 5 Panel A. CEOs with greatest number of articles Company name CEO last name Year Number of articlest+1 % of negative articlest+1 (%) Tenure as CEO (years)t Tot al compt Residual (total comp)t Total payoutt Three-year mkt-adj stock returnt (%) Market value of equity ($mil)t Tyco International AMR Corp. Citigroup Inc. Hewlett-Packard Co. Siebel Systems Delta Air Lines Qwest Commun. Disney (Walt) Co Disney (Walt) Co. Disney (Walt) Co. Kozlowski Carty Weill Fiorina Siebel Mullin Nacchio Eisner Eisner Eisner 2001 2002 2002 2001 2002 2002 2001 1996 1997 1998 320 250 178 168 132 121 116 109 88 87 57 54 40 32 49 73 35 50 55 38 9. 2 4. 6 4. 9 2. 3 . 4 5. 3 4. 9 12. 0 13. 0 14. 0 77,767 10,171 16,556 18,121 0 14,039 74,115 202,185 10,654 5,768 55,390 2,484 A18,313 A11,533 A6,994 6,901 57,349 192,527 A233 A9,244 42,177 1,109 13,367 1,248 34,586 4,870 101,995 8,654 10,654 575,596 A85 A111 A44 A68 A123 A120 A106 A7 A45 A74 88,064 1,030 180,901 32,633 3,600 1,493 23,506 42,631 54,099 52,552 Panel B. CEOs with greatest percentage of negative articles Company name CEO last name Year Number of articlest+1 % of negative articlest+1 (%) Tenure as CEO (years) t Total compt Residual (total comp)t Total payoutt Three-year mkt-adj stock returnt (%) Market value of equity ($mil)t Hillenbrand IndustriesNike Inc. Delphi Financial Grp. Bear Stearns Federal-Mogul Micron Technology Bear Stearns EOG Resources Inc. Manpower Inc. Warnaco Group Inc. Hillenbrand Knight Rosenkranz Cayne Miller Appleton Cayne Hoglund Fromstein Wachner 1998 1997 2002 1995 2000 1996 1998 1994 1995 1996 12 10 9 7 6 5 5 4 4 4 100 100 100 100 100 100 100 100 100 100 9. 6 29. 3 15. 6 1. 9 0. 3 1. 9 4. 9 7. 3 6. 9 9. 3 3,887 1,679 1,500 8,472 880 4,251 27,176 13,365 3,726 20,490 A51 A4,256 A1,327 4,649 A1,960 1,074 16,260 11,477 A559 17,301 2,936 1,679 1,500 9,384 426 4,847 27,176 20,114 3,726 9,434 A102 A56 A1 A26 A149 4 28 4 28 A32 3,793 13,202 783 2,508 163 4,750 6,663 3,000