2022正规essayAccounting Essay Smaple：Accounting Research The Press as a
Accounting Essay ：Accounting Research The Press as a Watchdog for Accounting Fraud
ABSTRACTThis paper investigates the press’s role as a monitor or “watchdog” foraccounting fraud. I find that the press fulfills this role by rebroadcastinginformation from other information intermediaries (analysts, auditors, andlawsuits) and by undertaking original investigation and analysis. Articles basedon original analysis provide new information to the markets while those thatrebroadcast allegations from other intermediaries do not. Consistent with adual role for the press, I find that business-oriented press is more likely toundertake original analysis while nonbusiness periodicals focus primarily onrebroadcasting. I also investigate the determinates of press coverage, findingsystematic biases in the types of firms and frauds for which articles are published.In general, the press covers firms and frauds that will be of interestto a broad set of readers and situations that are lower cost to identify andinvestigate.∗Harvard University. I thank an anonymous referee, Jeff Abarbanell, Mary Barth, SudiptaBasu, Brian Bushee, Fabrizio Ferri, Stu Gilson, Cristi Gleason, Michelle Hanlon, Paul Healy,Jack Hughes, Amy Hutton, Bruce Johnson, Bob Kaplan, Tom Lys, Michael Maher, MaureenMcNichols, Doug Skinner, RossWatts, GregWaymire, and JoeWeber as well as workshop participantsat Boston College, Emory University, Harvard Business School, Massachusetts Instituteof Technology, Notre Dame University, The Ohio State University, University of Iowa, Universityof Michigan, University of Texas-Austin, the 2004 Duke/UNC Fall Accounting Camp, andthe 2003 Stanford Summer Camp, and several anonymous members of the press for commentson earlier versions of this paper. I thank Sarah Eriksen, Anne Karshis, and Kathleen Ryan forresearch assistance. I am grateful for the funding of this research by the Harvard BusinessSchool.
1. IntroductionThis paper examines the press’s role as an early information intermediaryin the public identification of corporate financial malfeasances. Priorliterature regarding the press is limited and presents a conflicting view of itseffectiveness as an information intermediary. On the one hand, there arearguments that the press caters to the lowest common denominator, doesnot provide in-depth research or analyses, and focuses on sensationalizing issuesin order to sell papers (Jensen , Core, Guay, and Larcker ,DeAngelo, DeAngelo, and Gilson [1994, 1996]). On the other hand, thereis evidence that a free media is related to country-level economic growthand that pressure created by press coverage can play an important role inthe corporate governance of firms (Djankov et al. , Dyck and Zingales[2002a], Dyck, Volchkova, and Zingales ). The later studies suggest#p#分页标题#e#the press plays an important informational role, but do not examine howthis is achieved. While a few studies do examine the impact of specific information,they have focused on rebroadcasting of information created bymanagement or other information intermediaries (Huberman and Regev, Dyck and Zingales , Dyck, Volchkova, and Zingales ).Thus, even for studies that suggest the press is an important informationalsource, there is limited evidence regarding how the press fulfills the informationalrole and no evidence regarding whether the press provides newinformation.In this paper I use a sample of firms whose accounting was challengedby the Securities and Exchange Commission (SEC) to examine whetherthe press is involved in the early public identification of an accountingissue. Further, I study whether the press directly identified the issue or isrebroadcasting information provided by more traditional intermediaries(analysts, auditors, and legal institutions). Finally, I examine whether thepress is systematically biased towards covering certain types of firms andfrauds. Combined, these analyses contribute to the literature by providinga more complete understanding of the press’s role as an information intermediaryin financial markets.I use accounting, auditing, and enforcement releases (AAER) to identifya sample of firms that were sanctioned by the SEC for accounting malfeasances.Use of AAER allows me to examine a sample of firms widely believedto have engaged in accounting fraud and provides objective measures of thecharacteristics of the fraud, such as the magnitude and nature of infractions.Mysample consists of 263 firms that have committed a wide range of accountingviolations. I find that the press publishes articles regarding accountingfraud prior to a public acknowledgment by the firm or SEC for 75 (29%) ofthe firms.These 75 articles indicate that the press is involved in the early public disseminationof an accounting issue, suggesting they do fulfill an informationintermediary role. I analyze the content of the articles and use market tests toinvestigate whether the press is providing “original” information to the marketsor simply rebroadcasting information provided by other informationPRESS AS WATCHDOG 1003intermediaries. My results indicate that both occur. My first analysis examinesthe sources used to support the article. Many articles indicate the pressundertook original “investigative” reporting by analyzing a mix of information,often including publicly available documents, to create its ownallegation of accounting misdeeds. However, other articles refer to traditionalinformation intermediaries, such as analysts, litigation/court actions,and auditor changes, suggesting the press treats rebroadcasting as a legitimate#p#分页标题#e#information role. Further supporting a dual role in the press, I findevidence that press outlets and writers that specialize in business reportingare more likely to provide original stories, while local papers and generalistsare more likely to rebroadcast existing information.I next examine whether the articles provide incremental information tocapital market participants. I perform this analysis by informant category: reporteroriginal, analysts, auditors, and law. Prior work has argued that presscoverage is important due to the pressure it places on management (Dyckand Zingales [2002a], Dyck, Volchkova, and Zingales ). If press coverageper se is informative, I would expect a negative response to all articlecategories. Since reporter-generated analysis provides new information, Ipredict a greater response to such articles. I find that the response is greaterfor articles based on reporter analysis than for those that rebroadcast fromother information intermediaries. After controlling for publication type,the response to reporter investigation is approximately −12.6%, while theresponse to rebroadcasting articles is not significantly different from zero.These findings show that in many cases the press develops new informationfor the markets, suggesting an important role as an independent monitoror information intermediary in financial markets. However, it provides nosupport for an informational role when rebroadcasting from other informationintermediaries. Of course, it is still possible that this rebroadcastingplays an important role in attracting the attention of entities that may takeactions (such as regulatory or consumer groups), consistent with politicalcosts and monitoring suggestions in prior literature (Watts and Zimmerman, Dyck and Zingales [2002a]).1Having established that the press plays a role in the early identificationof fraud, I examine how characteristics of the firm and fraud impact thelikelihood of an article alleging fraud. The goal of this analysis is to provideevidence regarding when the press acts as a monitor of firms’ activities. Mypredictions are based on the press trading off the long-term pecuniary costand benefits of acting as a watchdog.I predict that firms with high information visibility are more likely to havearticles written regarding their accounting fraud due to the obvious highinterest in these firms, which suggests there will be interest in a story, and to1 It is important to note that the tests are designed to exclude the period when the rebroadcastedinformation was initially excluded. Thus, my findings are not suggesting that these othersources lack credibility in general or even are less credible than the press as a source for allegations.Rather, the results suggest that there is no incremental reaction when these allegations#p#分页标题#e#are made broadly public through the press.1004 G. S. MILLERthe lower cost of investigating firms with a rich information environment.Consistent with these predictions, I find that firms with a large number ofgeneral press articles or greater market value of equity are more likely tohave their accounting violations first identified in the press. However, I findno evidence that greater analyst following increases the likelihood of anarticle.Audit changes are public events that draw attention to a firm. They alsoprovide the basis for beginning a story regarding the firm’s accounting. Dueto these reduced costs, I predict and find that articles are more likely forfirms with auditor changes.The press industry generates much of its income from advertising revenueand so may be less likely to be critical of firms that are currently largeadvertisers or have the potential to be in the future. However, I find noevidence that the press is less likely to write articles regarding firms in highadvertising industries.I also expect aspects of the fraud will impact the cost of identifying thefraud and the benefits of publishing an article. I use the AAER informationto capture several aspects of the fraud. I find a high association betweenthe number of individuals involved in the fraud and the likelihood of anarticle, consistent with inside leaks reducing the cost of investigating frauds.Frauds with a greater dollar magnitude are also more likely to result in astory, consistent with coverage of more egregious and interesting frauds.Similarly, frauds that involve public misleading statements (such as a pressrelease claiming a large new contract and related accounting malfeasancesto support the press release) are likely to both reduce search costs and beof greater interest to readers of the original disclosure. Consistent with this,I find articles are more likely if the AAER alleges the company provideda material publicly misleading statement or filing. Finally, frauds that involvemanagement profiting due to insider trading, hidden compensation,or plain theft more easily lend themselves to a personalized and controversialspin, suggesting the press may find frauds involving such actions asbeing more newsworthy. Consistent with this, I find that frauds that areaccompanied by such actions are more likely to result in an article.The purpose of this study is to provide evidence on the role of the press asan information intermediary. The evidence indicates that the press facilitatesearlier public knowledge of a fraud both by original investigative reportingand broadly rebroadcasting information from other intermediaries. Whilethe original reporting is informative to the capital markets, rebroadcastingis not. Further, this study shows that the press is systematically biased towards#p#分页标题#e#coverage of high visibility firms and those with interesting frauds. Combined,this suggests that the press can provide an important informational role, butthat its coverage is far from complete. This study combines with other earlystudies on the press to begin to develop a fuller understanding of the roleof the press in financial information flows.The evidence provided in this paper is also of interest due to the increasinguse of the press as a control variable in accounting and economic studiesPRESS AS WATCHDOG 1005(Frost, Gordon, and Hayes , Bushman, Piotroski, and Smith ,Haw et al. ). With a greater understanding of the press, we can moreclearly determine whether these variables effectively capture the constructsthey are meant to measure.This paper proceeds as follows: Section 2 discusses related literature andmotivates the research question, section 3 discusses the sample selectionand collection of press coverage, section 4 provides results, and section 5concludes.2. Related Literature, Motivation, and Research Questions2.1 THE PRESS IN THE INFORMATION PROCESSExisting research on the relation between the press and commerce islimited (Zingales ). The literature that does exist often has dramaticallyconflicting findings. While many of the studies suggest that the pressimpacts public perception in some manner, one group of studies indicatespress coverage lacks in-depth research and tends towards sensationalism.For example, Jensen  argues that the press has become a form ofentertainment and that articles are written to appeal to the lowest commondenominator. Consistent with this, Core, Guay, and Larcker  studypress coverage of compensation and conclude that the press sensationalizestheir coverage by focusing on large ex post stock gains rather than compensationexpense to the company. They conclude that this press coveragehas no impact on compensation behavior. Using a clinical study approach,DeAngelo, DeAngelo, and Gilson [1994, 1996] suggest that simplistic presscoverage of junk bonds may have skewed economic behavior of customersand regulators. Dyck and Zingales [2002b] suggest that the press may encouragefinancial bubbles by adopting a company’s spin in return for privateinformation. As a group, these studies question the validity of the press asan important information intermediary in the economy and even suggest itmay play a negative role.In contrast, other studies suggest that the press is an important componentof the information environment in society. There is a growing literature thatuses cross-sectional variation in the national characteristics of the press toinvestigate the role of the press in corporate governance, governmentalactions, and development (Dyck and Zingales [2002a], Stromberg ,#p#分页标题#e#Djankov et al. ). This literature generally concludes that a strongpress can have a positive impact on the political and economic makeup ofa country. This impact is presumed to be driven by the oversight functionof the press. However, these studies have not examined the mechanismthrough which the press exerts this oversight.A few studies have shown that the press shapes public opinion by packagingand rebroadcasting information that is already available (Hubermanand Regev , Dyck and Zingales , Dyck, Volchkova, and Zingales). These studies indicate that press coverage per se impacts the1006 G. S. MILLERpublic’s response to information and the action of firms being covered.Since these studies focus on rebroadcasting of information, they do notprovide any insight regarding whether the press actively investigates andprovides original information.The goal of this paper is to add to the literature regarding the press byexamining a highly technical situation in which the press is likely to facevarying incentives to perform an investigative role. This allows me to examinewhether the press can participate as an early information intermediaryin technical situations, whether this participation includes original analysis,the relative informational value of original analysis and rebroadcasting, andwhether the press skews coverage of frauds consistent with a preference forsensationalism.http://www.ukassignment.org/Accounting_Essay/2.2 THE PRESS AS A WATCHDOG FOR ACCOUNTING FRAUDIn this paper, I focus on one aspect of press reporting—acting as a monitoror “watchdog” for the public by assisting in the early identification ofaccounting impropriety.2 The watchdog role is frequently referred to asone of the most important functions of the press (Serrin and Serrin ,Islam , Djankov et al. , Dyck and Zingales [2002a]). The watchdogprocess often includes combining public and non-public informationwith an analysis that highlights potential problems. Watchdog journalismin business reporting is unique in that SEC filings and other publicly availableinformation provide a rich starting point for the process. However, thereporter still must identify the issue, improvise to collect supporting information,synthesize the information, frame the issues, and disseminate to thegeneral public (Keller ).My study investigates the pres’s role as a watchdog by examining whetherthe press publishes an article alleging accounting irregularities prior to apublic admission by the company or announcement of an SEC investigation.In some situations, the press provides the first public indication thatan accounting issues exists, serving as the original information analyzer. Inothers, it may pick up indications from another information intermediary#p#分页标题#e#and rebroadcast the information. Being the first entity to identify an accountingissue is a clear example of a monitoring role, and likely containsmore informational value than the subsequent coverage. However, even ifthe potential issue was identified by another public information intermediary,the press can still provide an important function by publishing an2 The term “watchdog” refers to a journalist alerting the public to an issue through presscoverage just as a canine watchdog alerts others to a danger by barking. As with many jargonterms, “watchdog reporting” has a loose definition. All of the working definitions include theneed for critical thought and question asking and many also restrict the definition to cases inwhich the reporter is one of the first entities to address an issue to the broader public. Othersexpand watchdog reporting to the follow-up that occurs after an issue is initially identified.While examination of that follow-up role would be interesting, it is beyond the scope of thispaper.PRESS AS WATCHDOG 1007article that synthesizes this concern with other information regarding thefirm (Dyck and Zingales [2002a]). This validates the original concern andmakes the entire story more accessible to the general public.Accounting fraud is an important event in evaluating companies and thusalso an important news event. The often extreme actions, tensions, and personalitiesinvolved in accounting fraud create a compelling story, consistentwith sensationalism (Jensen ). For example, one popular journalismtext, Jamieson and Campbell [2001, p. 41–52], defines a “newsworthy event”as having five characteristics: (1) can be personalized, (2) dramatic, violent,and conflict filled, (3) actual and concrete, (4) novel and deviant, and (5)an issue of ongoing concern. Accounting fraud is one of the few businessstories that meets all five criteria.However, countervailing pressures may reduce the pres’s activities aswatchdogs over firms. The nongovernment-owned press are market participantsthemselves. Many of their interests are aligned with the market,making it less likely that they desire to cause concern regarding the marketor face scrutiny themselves (Herman ).3 These apprehensions can beintensified by affiliations with parent companies or advertisers who may beharmed by investigative reporting (Herman , Jamieson and Campbell). The press also may censor stories in an attempt to keep ongoingsources available for future information (Jensen , Herman ,Dyck and Zingales [2002b, 2003]). Individual reporters may be concernedabout retaining their job and/or future employability if representatives ofcompanies complain to their editors (McNair ).4 Finally, the press#p#分页标题#e#has to be concerned about offending its readers, who are participants in themarket themselves (Jamieson and Campbell ).5 These forces suggestthat the press is likely to face strong pressure against reporting accountingimproprieties.6 Accordingly, the press may choose not to provide initial orearly information regarding the problems, but rather to wait until a fraud hasbeen publicly acknowledged by the company and then use ex post articlesto fulfill the need for interesting business stories.As the above discussion shows, the watchdog role could be an importantcommerce-related function of the press. However, we have no3 Jamieson and Campbell [2001, p. 62–63] quote Peter Silverman, the business and financialeditor of the Washington Post: “Newspapers themselves are among the most secretive and themost protective about the facts and figures of their own business. They are not likely to askothers to do what they are unwilling to do themselves.”4 McNair  presents several examples of attempts by businesses to silence reporters.While many were not successful, in at least one case a reporter was fired shortly after a companycomplained about coverage (McNair [2002, p. 14]).5 BillWasik, an editor of Harpers, argues “When a company’s fortunes seem poised to enlargeindefinitely, the interests of all potential sources—the company, the analysts, large investors—are aligned not only with one another but with the interests of the reader , who is assumed to be ashareholder or potential shareholder.” (Wasik [2003, p. 84]) (Italics in original).6 It is interesting to note the similarity between these forces and those studied in the analystand auditing literatures.1008 G. S. MILLERunderstanding of if or how this occurs. Specifically, how often does the pressparticipate in the early public identification of an accounting fraud? Is thisdone primarily through reporter investigative analysis or as a rebroadcastingof other information intermediaries? Does the market view either action asinformative? Is there systematic bias in the types of firms or frauds coveredin the press? The next section develops specific research predictions andmethods for each of these issues.2.3 RESEARCH QUESTIONS2.3.1. Press Participation in Early Allegations of Accounting Malfeasances. Ifirst examine the degree to which the press fulfills the early identificationrole. As I discuss more fully in the data section, I include firms in my samplebased on SEC AAER that allege accounting violations.7 Examining theproportion of these firms identified by the press prior to the official announcementof an accounting issue provides evidence regarding the press’sability to identify and its willingness to publish stories regarding suspected#p#分页标题#e#fraud.Next, I examine whether the articles more broadly disseminate informationcreated by other intermediaries versus providing an original analysisand synthesis of new information. I address these questions using both anexamination of the sources stated in the articles and a returns-based eventstudy. I first examine the articles and code them as either attributing theallegations to another information intermediary (analysts, auditors, or legalfunctions) or investigative analyses by the press. These investigative analysesare based on a wide range of activities such as review of financial statements,channel checking with customers, investigation of announced contracts, etc.Of course, isolation of sources based on the ex post article is difficult, as thepress may initially receive information from one entity but in the processof writing the article rely on subsequent sources. Assuming no systematicbiases in such switches, this portion of the study still should provide insightinto the process by which the press collects information.Reporter-generated analysis of accounting is a technical skill that is morelikely to exist within specialized business publications and reporters. Thus,I expect that reporters and publications with these skills are more likely togenerate articles with original content, while those in less specialized outletsare more likely to rebroadcast the views of other information intermediaries.I investigate this within-press variation by examining the relation betweensources of information in an article and type of press/reporter.I use a market return event study as a final method of examining originalinformation content versus rebroadcasting of information. If the press is7 I use the term fraud in a broad sense to imply any type of systematic accounting misstatement.AAER are the SEC’s version of the actions that led to an accounting misstatement withinthe firm. Firms settle AAER without admitting or denying guilt and thus still may argue thatthey were not guilty of accounting fraud or misstatement. However, the existence of an AAERsuggests that the SEC believes there were sufficient problems to make a public case.PRESS AS WATCHDOG 1009playing an informative role there will be a negative reaction to the articles.Conversely, if the press is simply repeating widely known information and/oris not considered a reliable source of analysis, then there will be no reactionto the articles. Based on this, in the cross-section I expect a larger reactionto articles that rely on reporter-generated information than on those thatrebroadcast information from other intermediaries. Several studies arguethat the press changes expectations and actions by rebroadcasting information(Huberman and Regev , Dyck and Zingales , Dyck,#p#分页标题#e#Volchkova, and Zingales ). Consistent with these studies, I expect anegative reaction even for rebroadcasted articles.82.3.2. Characteristics of Firms and Frauds That Influence the Likelihood of anArticle. The press may be an effective monitor in some situations, but ineffectivein others due to biases in coverage (Mullainathan and Shleifer). Thus, I examine the firm and fraud characteristics that lead thepress to identify and publish an allegation of accounting fraud in some instanceswhile missing or staying silent in others. As with all economic agents,I expect reporters will choose articles based on maximizing benefits and reducingcosts. That is, the press prefers articles that are interesting to a largegroup of readers, resulting in higher subscription revenues and a largerreader base to offer potential advertisers (Stromberg ). Further, thepress will attempt to increase future readership by publishing stories readersfind memorable (Mullainathan and Shleifer ). The press must alsoconsider the cost of undertaking such investigations. Firms with a richer informationenvironment can be analyzed more easily and some aspects of thefrauds may be easier to detect, both of which reduce the cost of investigation.As is often the case, it is difficult to separate the impact of cost and benefitsin much of my empirical analyses. Accordingly, I discuss the empiricalimplementation of the predictions provided by costs and benefits together.I begin by considering three firm-level characteristics.I expect that firms with a high number of stakeholders present an opportunityto attract a large reader base. Further, stakeholder demand forinformation should lead to greater demand for information directly fromthe firm and an increased number of other information intermediaries followingthe firm, resulting in a richer information environment for thesefirms. Thus, relative to other firms, the benefits of coverage of these highvisibilityfirms are greater and the cost of analyzing these firms is lower. Iproxy for high visibility by using market value, analyst following, and overallpress coverage (details of these items are discussed more fully in the sampleand results sections).8 Dyck and Zingales [2002a] and Dyck, Volchkova, and Zingales  are more in thespirit of studying the impact of a group of articles rather than a single article. Accordingly, mystudy of individual articles does not directly address their monitoring hypothesis. Instead, itexamines whether rebroadcasting creates an immediate informational update.1010 G. S. MILLERAuditors are primary analyzers of firm financial information. Changes inauditors may serve to draw attention to a firm’s accounting and even provideleads to disputed accounting issues. This reduces the cost of investigating#p#分页标题#e#the firm. Thus, I expect that firms that experience an auditor change aremore likely to receive an article.A firm’s advertising spending also may impact the likelihood of a criticalarticle. Advertising is an important source of revenue for the press andpublishers may be hesitant to print articles that will offend large currentor prospective advertisers (Reuter and Zitzewitz ). Articles alleginginappropriate accounting have a high potential for upsetting affected advertisers.I predict firms that have the potential to be large advertisers are lesslikely to be the subject of unfavorable articles. I proxy for advertising statususing an indicator variable that is coded as 1 if the firm is in an industrythat was in the top 15 advertising industries (according to Advertising Age)in each of 1985, 1990, 1995, and 2000.Next, I expect that characteristics of the fraud may impact the press’sability to frame the story in a way that will appeal to a broad set of readersand/or the ease with which the press can detect the fraud. Thus, thesecharacteristics will influence both the costs and benefits that determinewhether an article is written regarding a particular fraud. I use informationfrom the AAER to collect various aspects of the fraud.9By nature, frauds are designed to be concealed from outsiders. Thus,identification of a fraud can be costly and the outcome highly uncertain atthe beginning of an investigation. An internal leak, or at least indicationof the potential for an issue, can have a strong influence on reassuring thereporter that there is a story to be pursued and may even point the reportertowards the sources needed to confirm the fraud. The more people involvedin the fraud, the greater the likelihood of a leak of the activities to outsiders.Accordingly, I predict that the likelihood of an article is increasing in thenumber of people involved in the fraud.Frauds that involve a large dollar magnitude may be deemed as morenewsworthy, increasing the benefit of publishing an article. I investigatewhether the dollar magnitude of the fraud is related to the likelihood of anarticle by including the total magnitude of the fraud.I expect the press to be more likely to identify a fraud that involved amaterially misleading public statement. Such statements are likely to attractscrutiny and communicate items the company expects will catch thepublic’s attention, suggesting broad appeal to consumers of the press. Further,they allow the reporter to frame the story as having caught managementattempting to fool the public, creating a personal element as well asshowing conflict and deviant behavior. This item both increases the benefitsand decreases the costs of investigating and publishing a story on thefirm.#p#分页标题#e#9 The SEC often publishes multiple AAER regarding a single infraction. I collected informationfrom all AAER when this occurred. Given the difficulties in identifying AAER, it is possiblethat some AAER were missed, introducing noise into the tests.PRESS AS WATCHDOG 1011Many accounting frauds occur concurrent with management misappropriationof funds, either through illegal insider trading, manipulation ofbonus plans, or pure theft. While the cost of identifying these actions is unclear,they will make the story more compelling by increasing the degree towhich the article can be personalized, is dramatic, has conflict, and coversdeviant behavior.3. Research Design and Sample3.1 SAMPLE CREATION AND AAERTo investigate the press’s actions as a watchdog for accounting malfeasances,I must identify firms that have committed accounting violations.Consistent with prior work, I use the SEC AAER to build a sample (Dechow,Sloan, and Sweeney , Bonner, Palmrose, and Young , Beneish[1999a, b]). There are differing opinions regarding why AAER are issued.While some believe they are meant to address current trends observed bythe SEC (Feroz, Park, and Pastena ), others believe they are issuedfor high-profile cases that will enhance the stature of the SEC (DeFond andSmith ). In any case, there is wide agreement in the literature thatAAER generally represent egregious violations of the generally accepted accountingprinciples (GAAP) standards of reporting and disclosure. As such,they provide the basic criteria needed to develop my sample: an observablesample of firms that have committed accounting malfeasances. AAER havethe added advantage of providing detailed information, such as discussion ofthe violations, summaries of findings, and a timeline of the violation. Thisin-depth characterization is helpful in developing variables to investigatehow the type of fraud impacts press coverage.Use of AAER has disadvantages. First, AAER likely represent extremeviolations. If these firms have characteristics that vary systematically fromfirms that commit fraud but do not receive an AAER, then the results maynot generalize into the population in general. Second, Feroz, Park, andPastena  report that an SEC official indicated that approximatelyone-third of its leads come from the financial press. It is not clear whetherthat entails only articles that allege fraud, or that the press is simply onesource for collecting information. However, it is possible that a portion ofthese firms would not have been identified without the press. Assumingthat all such articles are read by the SEC and that the SEC would neveridentify the fraud absent press coverage, the findings in this paper wouldlikely overstate the proportion of frauds in the general population that are#p#分页标题#e#identified by the press. That is, there may be many frauds missed by boththe SEC and the press.AAER are identified using the SEC Web site.10 AAER are issued for awide range of violations. While accounting fraud is a primary reason for10 As a robustness check, the AAER identified were compared to those in the October 2002General Accounting Office document Financial Statement Restatements, Trends, Market Impacts,Regulatory Responses , and Remaining Challenges. There were no discrepancies noted.1012 G. S. MILLERTA B L E 1http://www.ukassignment.org/Accounting_Essay/Accounting and Auditing Enforcement Release Firms by End of Year of FraudYear Number of Firms Percentage of Sample Cumulative Percentage1987 1 0.4 0.41988 1 0.4 0.81989 2 0.8 1.61990 6 2.3 3.91991 7 2.7 6.61992 21 8.0 14.61993 29 11.0 25.61994 28 10.6 36.21995 22 8.4 44.61996 30 11.4 56.01997 32 12.2 68.21998 27 10.3 78.51999 22 8.4 86.92000 24 9.0 95.92001 9 3.3 99.22002 2 0.8 100.0Total 263 100 100
A firm is included in the sample if an AAER was identified on the SEC Web site and that release isrelated to an accounting fraud perpetrated by the firm.an AAER, they also frequently include insider trading, violations of SECrequirements by auditing firms, and other illegal acts. For the purposesof this study, only AAER that include a substantial accounting fraud areretained.11 Subsequent analyses rely heavily on data from the Factiva newsservice, such as identifying the date a firm issues a press release that it isunder investigation or restating earnings. The two primary services for firmpress releases are PR Newswire, with coverage beginning January 2, 1985 andBusiness Wire, with coverage beginning July 28, 1998. Accordingly, AAERwith violation periods that began prior to January 2, 1985 are excluded.12 Asshown in table 1, this results in a sample of 263 AAER with violation periodsending between 1987 and 2002.13Information from eachAAERis coded for use in subsequent tests. Untabulatedanalyses show that the AAER in this sample are consistent with those inprior studies. For example, revenue manipulation is the most common formof violation (49% of the sample) (Feroz, Park, and Pastena ). Assetoverstatement is the second most common violation (35% of the sample).11 For example, there were several AAER that involved payments that violated the ForeignCorrupt Practices Act. These AAER also noted minor accounting violations related to how the#p#分页标题#e#payments were recorded.12 Robustness tests indicate the results are similar if AAER with violation periods prior toJuly 28, 1998 are excluded.13 The number of frauds in the beginning years of my study is low due to the requirementthat the fraud occur no earlier than 1985 and the generally long lag time between a fraud andAAER. Similarly, the number of frauds in the last two years of my study is low due to the samelag and the data collection period in early 2003.PRESS AS WATCHDOG 1013Other common violations include providing materially misleading publicstatements/filings (26%) and manipulation of reserves (18%). These sumto greater than 100% because the SEC frequently identifies more than oneviolation per firm. In fact, the number of violations noted in the AAER varyfrom one (16% of the sample) to seven or greater (1% of the sample).143.2 PRESS COVERAGECoverage is collected over several time periods, with searches designed tocreate different variables. All press searches are performed on Factiva, whichis Dow Jones’s replacement for the Dow Jones News Retrieval service. Factivacovers nearly 8,000 sources of information including all of the major wireservices (e.g., PR Newswire, Business Wire, Dow Jones, Rueters, AP), majorU.S. business publications (The Wall Street Journal, Barron’s, Forbes, Fortune,Business Week), national and regional newspapers (The New York Times, WashingtonPost, Los Angeles Times, St. Petersburg Times), and trade publications(Computergram, Boating Industry).15The first search examines press coverage over the period that the accountingfraud occurred (i.e., the infraction period). This information is used toassess the intensity with which the press examined the firm while the fraudwas occurring. Searches are performed using multiple variations of the firmname within one search (full and corporate diminutives). Because the goalis to identify press coverage, the wire services that directly reprint managerialnews releases (PR Newswire and Business Wire) are excluded from thissearch. Similarly, summaries of a list of firm names and an item (such asa large traded volume during the day) are excluded.16 Searches cover theentire text of all documents. The resulting number of press articles is thendeflated by the length of the infraction period (in months) to standardizepress coverage across firms.17 The untabulated distribution of the numberof articles per month varies, with 42 of the firms averaging one or fewer articlesa month, and 18 included in more than 100 articles per month duringthe infraction period. However, most firms have between 2 and 20 articlesper month.I search for articles that identify the accounting fraud from the beginningof the infraction period until the date of AAER issuance. However, articles#p#分页标题#e#14 As a robustness check, all variables are interacted with the year of the violation to assurethere are no trends over time that may impact the use of these variables in later analyses. Nosuch trends are noted.15 As a robustness check, several firms are randomly selected and press searches are run onFactiva, Dow Jones, and LexisNexis. The results find Factiva to be a superset of Dow Jones.While a few items appear on LexisNexis that were not on Factiva, they are not from majorpublications and do not include information excluded by sources on Factiva.16 These summaries include the DJ Highlights, News Highlights, V-Alert, P-Alert, CaliforniaSummary, Southeast Summary, Southwest Summary, Dow Jones Corporate Economic NewsSummary, Recap of Dow Jones Special Reports, The Wall Street Journal Earnings Summary, andthe International Calendar of Corporate Events.17 All reported results are similar if the undeflated variable is used.1014 G. S. MILLERare excluded if written after a public announcement by the firm of either anSEC investigation or accounting restatement. There are often several yearsbetween the infraction period and issuance of the AAER so a search stringis designed to comprehensively identify articles that question the firm or itsaccounting while reducing the number of spurious articles to be read.18 If anarticle is identified, only the first article is retained. Data on that article arecollected, including the publication, author, type of article, and cited sourceof information. The decision of whether an article “caught” an accountingfailure is obviously judgmental.19 Coding bias was avoided by employing aresearch associate who was not aware of any of the hypothesized relations.In the case of uncertainty, a second research associate reviewed the article.4. Results4.1 ANALYSES OF FREQUENCY AND SOURCES OF ARTICLES WRITTENThe first analysis examines the proportion of firms the press identifiesas having questionable accounting. Table 2 shows that 75 of the 263 firms(approximately 29%) are identified by the press prior to the SEC or firm’spublic announcement. An interaction of the percentage of firms caughtand the last year of the fraud finds no discernable pattern. Thus, the press’sactions as a watchdog of accounting fraud appear to have remained constantover the time of my study.It is possible that the press frequently writes negative articles regardingfirms’ accounting and thus the 29% finding is consistent with the level ofpress questioning of all firms. In that case, the press’s inability to sort outmalfeasances from legitimate accounting reduces their value as monitors.18 The search includes both the formal and shortened versions of the firm name appearingin the same paragraph as any variation of: accounting, audit, fraud, illegal, illicit, insider#p#分页标题#e#trading, investigate, overstate, understate, probe, quit, quits, quitting, resign, restate, revenue,revenue recognition, rumor, law and suit within two words of each other, opinion and withdrawwithin three words, short and sell or stock within three words. It also included the companyname and any words from the following group: adjustment, compensation, doubt, dubious,financial, number, officer, recognition, record, reserve, Securities and Exchange Commission,and SEC combined with any words from this group: corrupt, conceal, credible, debacle, deteriorate,difficult, discrepancy, dishonest, failure, false, falsify, fear, improper, inconsistent, ills,irregular,misappropriate, mislead, misrepresent, negative, offense, question, sell-off, shortfalls,skeptic, suspicious, trouble, unexpected, unsupported, violated, weak, woe, worried, and writeoff.Several companies with no caught article identified by the primary search are searchedwithout the limiting string. No additional caught articles are found.19 In fact, even the press cannot agree on whether a specific article effectively identifiesmisdeeds at a firm. For example, in the summer 2002 Neiman Reports there are three articlesregarding press coverage of Enron. Madrik  contends that reporters did not look skepticallyat Enron, and in fact helped to perpetuate many of its practices. On the other hand, Behr argues that negative coverage was minimal, but points to a March 2001 Fortune articleas raising questions regarding Enron. Steiger [2002, p. 10] (a managing editor at The WallStreet Journal) argues that Enron’s misdeeds were uncovered in October of 2001 by “relentless,careful, intelligent work of two Wall Street Journal reporters.”PRESS AS WATCHDOG 1015TA B L E 2Firms “Caught” by Press Article Presented by Last Year of Accounting FraudNumber of Number of Firms Percentage of CurrentYear Firms with “Caught” Article Year with “Caught” Article1987 1 1 1001988 1 0 01989 2 0 01990 6 1 16.71991 7 1 14.31992 21 4 19.11993 29 9 31.01994 28 12 42.91995 22 10 45.51996 30 9 30.01997 32 7 22.01998 27 6 22.21999 22 5 22.82000 24 6 25.02001 9 3 33.32002 2 1 50.0Total 263 75 naPercent of Total 100% 28.5%A firm is included in the sample if an AAER was identified on the SEC Web site and that release isrelated to an accounting fraud perpetrated by the firm. A firm is considered “caught” by the press if anarticle appears questioning the firm’s accounting prior to public disclosure by the firm (or SEC) that anaccounting problem exists.To address this issue, 75 firms are randomly identified from the Centerfor Research in Securities Prices (CRSP) database. The only criteria imposed#p#分页标题#e#are that the year matches the time frame of the AAER sample andthat the firm continues to exist for at least two years after the initial yearof inclusion. The same research associate employs the identical searchstring used for the AAER firms over a two-year period, finding only onearticle alleging accounting malfeasances.20 A χ2 test indicates this rate of1.3% is significantly lower than the 29% found in the AAER sample (0.001level).Next, I use the information in articles alleging fraud to gain a better understandingof whether the press generates original information or primarilyrebroadcasts views of other intermediaries. Table 3, panel A classifies articlesinto source categories meant to aggregate articles that suggest a similarunderlying information-gathering process: reporter-generated information,analyst, legal cases, and auditor resignations. Articles in the first categorysuggest the press is the first information intermediary to publicly identifythe accounting issues. Articles in the last three categories make it more likely20 A second article is identified that the research associate does not believe alleges fraud,but does challenge the explanation of legitimate accounting provided by the company. If it isincluded as an article, the rate increases to 2.7% and the difference with the AAER sample isstill statistically significant at the 0.001 level.1016 G. S. MILLERTA B L E 3Information Sources and Types of Publication for Watchdog ArticlesPanel A: Information sources cited by the pressNumber of Percentage of CumulativeInformation Source Articles Caught Sample PercentageReporter-generated information 27 36.0 36.0Analyst 22 29.4 65.4Legal cases 15 20.0 85.4Auditor resignation 11 14.6 100.0Total 75 100 100Panel B: Types of publicationsNumber of Percentage of CumulativeType of Publication Articles Sample PercentageNational business 29 38.7 38.7Local market 23 30.7 69.4Electronic business 14 18.7 88.1Trade publications 7 9.3 97.4National nonbusiness 2 2.6 100Total 75 100 100Source is based on the source that is attributed as providing the primary information or to haveprovided the information that initiated the article. Reporter-generated information includes articles thatstate sources as publicly available SEC documents or financial statements, analysis of public statementsmade by the firm, tips from customers, industry insiders, or anonymous sources, or article is written basedon analysis from such information without stating another external source. Analyst includes attribution toan individual who is a sell-side analyst or buy-side analyst, to a professional investing newsletter, or to shortsellers (named or anonymous). Legal cases include articles that reference information either from court#p#分页标题#e#proceedings or from participants in court proceedings (civil and criminal). Auditor resignations references8-K are of auditor changes. Industry/customer information includes attribution to either industry insidersor customers of the firm (named or anonymous).National business publications are publications that cover the national (or global) areas. This categoryincludes The Wall Street Journal, Fortune, etc. Local market publications are generally regional papers,such as The San Francisco Chronicle, Chicago-Sun Times, etc. Electronic business publications consist almostentirely of the Dow Jones News Service with one article from Bloomberg. Trade publications are basedon covering a specific industry in depth. Examples include Boating Industry and Business Insurance. Thenational nonbusiness publications are USA Today and The New York Times.that the press is not the first to provide identification, but rather played themore limited role of rebroadcasting. Still, this highlighting of an issue is animportant function in the overall watchdog process.Articles based on reporter-generated information are most common, with27 articles (36% of the sample). This includes articles using reporter-basedanalysis of public information (such as financial statements) and articlesrelying on information from sources that are not normally available to thegeneral public (such as customers or industry insiders).21 In each of thesearticles, it is the reporter making the case for accounting impropriety based21 Many of these articles simply refer to the financial statements without providing exactsources (i.e., they just say the latest quarter or annual report). However, others provide adetailed explanation of the information used. As a few examples, seven of the articles refer tocustomers or industry insiders, two to anonymous tips, one to Web-based information, four tospecific SEC filings (by filing number and page), etc.PRESS AS WATCHDOG 1017on analysis of public and private information. No other information intermediaries(i.e., analysts, auditors, or the legal system) are cited.22 Many ofthese articles provide a clear description of press-initiated investigative reporting,but others question the accounting without discussing the reasonfor investigating the firm.23Analysts are referenced in 22 articles (29%), making them the secondmost common source. In most cases, it is difficult to determine whether sucharticles are the result of press-initiated discussion of the firm or if the analystidentified the issue and took it to the press.24 Limited information in thearticles suggests both occur.25 Only 8 of the articles refer to written analystreports (3 sell side, 5 newsletters), while 10 articles identify the analysts asworking for the buy side/shorts who would not generally provide public#p#分页标题#e#reports. The private nature of these entities indicates that the press is thelikely vehicle for bringing many of these allegations to the general public.To further understand the relationship between analysts and the press, Icompare fraud identification rates and sources for firms with sell-side coverageto those without (per I/B/E/S). Thirty-six of the 133 firms with nocoverage are identified in an article (27%), while 40 of the 130 firms withanalyst coverage are identified (31%). These amounts are statistically equivalent.Turning to the sources of coverage, analysts are cited in only 7% ofthe articles for firms without sell-side coverage, but in 43% of the articlesfor firms that have sell-side coverage. Conversely, 47% of the articles forthe noncovered firms are based on reporter analyses, compared with only28% of those for firms with coverage. It is possible that the high number ofreporter-generated articles is indicative of the actual proportion of articlesthat are press initiated and the analysts’ cites are simply an attempt to useanalysts as third-party “experts” in a press-identified article. Alternatively,the shift in percentage could indicate that the press is more likely to undertakeindependent analyses in the absences of analysts’ coverage. In eithercase, it indicates that the press is capable of identifying a large proportionof articles without analysts’ support.22 If an analyst, lawsuit, or auditor resignation was mentioned, the article was classified inthat related category regardless of the extent of additional analysis provided.23 As an example of articles that indicate press-initiated investigative reporting occurred,a reporter noticed an ad to sell specialized used computer equipment as part of liquidatinga line of business. No company was identified, but there was a local (Silicon Valley) number.The reporter called the number and asked what company he had reached. He investigatedthe company and found they had made several recent public statements regarding the strongperformance of that sector of their business with no mention of liquidation.24 I attempt to search for analyst reports that alleged accounting malfeasances prior to thepublication of a press article. However, it is difficult to obtain analyst reports over much of theperiod of the study.25 Anecdotal discussion with analysts and reporters also suggests both happen. Several analystshave indicated the press as an important ongoing source of information. Further, analystsindicated they are frequently contacted by the press regarding companies they do not followor follow “passively.”1018 G. S. MILLERThe third most frequently cited source is legal cases, which occur in 15(20%) of the articles. They range from shareholder lawsuits alleging malfeasance#p#分页标题#e#to civil cases involving wrongful discharge and criminal cases investigatingtheft. It is easy to determine causality in legal cases as the filing leadsthe reporter to investigate the company. Some articles appear to rely entirelyon case information, but most articles include additional investigativereporting such as speaking to officers or customers of the company.The final category is auditor changes, cited in 12 (16%) of the articles.Similar to the legal cases, the reporters use the change as a cue to look morecarefully at the company and then develop information for articles that goesbeyond that found in the 8-K filings with the SEC.26Next, I examine the type of publications that print articles alleging inappropriateaccounting. Given the technical nature of accounting fraud, Iexpect that publications that specialize in business reporting are more likelyto have the tools to identify fraud and a subscriber base that will find such articlesnewsworthy. Untabulated analyses find that the articles are publishedin 40 different outlets from 54 authors. The Wall Street Journal and Dow JonesNews Service have the greatest number of publications, with 14 each, followedby The San Francisco Chronicle, with 6.27 While several publications havemultiple articles, the majority have one. Only one author has more than onearticle (Herb Greenberg, who accounts for five of the San Francisco Chroniclearticles); 18 of the articles do not have an author in the byline.Classifying publication type is inherently subjective (Jamieson andCampbell ). With this caveat in mind, table 3, panel B provides aggregatedinformation regarding the types of publications that cover accountingfraud. Twenty-nine of the articles are published in national business publications(The Wall Street Journal, Business Week, etc.). Twenty-three are from“local market” publications (Chicago Sun-Times, Miami Herald, etc.). Fourteenof the articles are from electronic media (Dow Jones News Service andBloomberg). Seven are from trade publications (Boating Industry, Business Insurance,etc.). Finally, two are from nonbusiness national publications (USAToday and The New York Times).28 Consistent with my expectations, thesedescriptive data show the importance of business publications in serving as26 I include analyst following and auditor change as firm characteristics in the cross-sectionalstudy of attributes that impact the publication of an article. I planned to include legal cases,but a pilot search of 14 firms (5% of the sample) found that all but one had legal cases duringthe period of the fraud. These cases are so common that there is effectively no variance.27 These two sources are related. Dow Jones News Service has over 20 regional offices thatprepare articles that go out on the Dow Jones Wire Service. The articles are also submitted to#p#分页标题#e#the editors of the The Wall Street Journal who determine whether some version of the articleshould appear in The Wall Street Journal (Thompson, Olsen, and Dietrich ).28 Factiva provides limited coverage of The New York Times during some portions of my sampleperiod. Thus, the number of New York Times’ articles may be understated. As a robustness check,a research assistant uses LexisNexis to search 10 companies for which no article was found and10 companies with an article from a publication other than The New York Times. No articles werefound.PRESS AS WATCHDOG 1019TA B L E 4http://www.ukassignment.org/Accounting_Essay/Relation between Sources of Information and Characteristics of the PublicationsPanel A: Relation between sources of information and type of publicationNational Local Electronic TradeInformation Source Business Market Business PublicationsReporter-generated information 12 8 3 4Analyst 12 4 6 0Legal cases 2 8 4 1Auditor resignation 3 5 1 2χ2 (p-value) 15.15(0.0869)Panel B: Relation between sources of information and frequency of articlePercentage Percentage ofof Recurring One One-TimeInformation Source Recurring Sample Time SampleReporter-generated information 10 55.6 17 29.8Analyst 7 38.9 15 26.3Legal cases 1 5.56 14 24.6Auditor resignation 0 0 11 13.5Total 18 100 57 100χ2 (p-value) 12.20(0.0067)Source is based on the source that is attributed as providing the primary information for or to haveinitiated the article. National business publications are publications that cover the national (or global)areas. This category includes The Wall Street Journal, Fortune, etc. Local market publications are generallyregional papers, such as The San Francisco Chronicle, Chicago-Sun Times, etc. However, for the purposes ofthis table, they also include the two national nonbusiness publications (The New York Times and USA Today).Electronic business publications consist almost entirely of the Dow Jones News Service with one articlefrom Bloomberg. Trade publications are based on covering a specific industry in depth. Examples includeBoating Industry and Business Insurance. Recurring articles are from a column that is published by a singlereporter/set of reporters on a regular basis. χ2 is based on a likelihood ratio χ2.watchdogs for accounting fraud, but they also show that many other sourcesuncover and publish articles regarding accounting fraud.Next, I examine relations between the sources and types of publications.The goal in this interaction is to examine whether the specialization ofbusiness reporters leads to more articles based on the original (or “investigative”)reporting interaction with other information intermediaries. Toreduce dimensionality, I combine national nonbusiness publications with#p#分页标题#e#local market publications, as both are general interest news outlets.Table 4, panel A shows that reporters from all publication types use awide variety of sources to identify accounting fraud. However, there arepatterns in the relative usage of sources consistent with specialization leadingto more active reporting. First, the national and electronic businesspublications have a high reliance on analysts, while local market and tradepublications rarely use that source. Second, national business publicationsrely equally on reporter analysis and analysts (12 articles from each), but donot appear to write articles in response to lawsuits and auditor resignations,suggesting they focus on original analysis rather than rebroadcasting legaland regulatory filings. The χ2 value of 0.09 indicates that these differencesare statistically significant.1020 G. S. MILLERPanel B compares authors with a recurring column with articles that appearas one-time coverage. Recurring authors may be more focused on providingindividual analysis, and thus developing the brand of their column.Additionally, they are more likely to be business specialists. Accordingly, Iexpect recurring authors to be more likely to rely on active reporter investigation.The findings in panel B support this hypothesis, with the majorityof the recurring articles being based on reporter analysis while nonrecurringarticles are more likely to rebroadcast. A χ2 test finds these levels tobe significant at the 0.007 level. Combined, the source differences acrosspublication and author type indicate that specialization allows reporters todevelop a skill set and contacts that assist in providing original information.The final investigation of the source of the news employs market returns toexamine whether the market views the articles as providing new information.The market should respond negatively to new information alleging accountingmalfeasances. The prediction is less clear for rebroadcasted information.Some studies indicate the press impacts perceptions by rebroadcasting information(Huberman and Regev , Dyck and Zingales , Dyck,Volchkova, and Zingales ). On the other hand, in an efficient market,repeating known information should not impact returns. This event-studyfaces several threats to its validity. First, the day on which the article becameavailable is difficult to determine. This threat is relatively limited for wireservices (which report real time) and newspapers (which generally are publishedin the morning, or at least available before trading ends). Magazines,however, are frequently published in advance of the “issue date.” Difficultiesin identifying the exact date the article became public introduce noiseand reduce the power of the tests. Second, some articles cite relatively recent#p#分页标题#e#events, such as lawsuits and auditor changes, that likely impact returns.If these events occur within the return window used for the event study,then results may be biased in favor of finding a market reaction. To reducethis bias, tabulated analyses rely on one-day returns, rather than the standardthree-day window. Third, my sample of 60 firms divided into multiplecategories results in relatively low-power tests, creating some problems in interpretingnonsignificant results. Finally, these tests only examine whetherthe market treats this as new information. It is possible that this informationis already in price, but that rebroadcasting is still informative to other firmstakeholders.I am able to identify clean returns data for 60 of the articles using acombination of CRSP, Datastream, and Bloomberg.29 Untabulated univariateresults find an average (median) one-day market adjusted reaction of−6.3% (−2.9%). The same measure for the three-day return centered on29 I identify 63 of the firms on these databases. However, two firms had significant eventsannounced after trading ended the day before the article and one was a penny stock tradingat $0.08 per share. If these firms are included, the results are similar except the returns forarticles based on auditor changes are significant in some specifications.PRESS AS WATCHDOG 1021article publication date is −8.2% (−4.5%).30 All of these amounts are statisticallysignificant. This evidence indicates that these articles provide newinformation to the market.31Table 5 reports regressions of day zero abnormal returns on the source,publication, and recurrence of author, consistent with the categories intables 3 and 4. This analysis examines whether differences in the sourcesused to write the article, venue of publication, or type of author impactthe informativeness of the article. All regressions suppress the interceptand include indicator variables for each category. In the first three regressions,categories are mutually exclusive, thus the coefficient is equivalent toa mean.32 As shown in the final regression, the regression specification hasthe advantage of including controls for venue and recurrence.The first regression shows the response to articles based on source of information.The response is statistically negative for articles based on reportergeneratedanalysis or analysts. I expect reporter-generated articles, whichare based on analysis and/or sources not commonly available to the public,to have the greatest incremental information content for the market. Consistentwith this, the −13.9% reaction to these stories is statistically greaterthan that for the articles based on any of the other sources. In this specification,articles citing analysts are also significant, though the −4.5% return#p#分页标题#e#is both statistically and economically smaller in magnitude than the returnsfor reporter-generated articles.Table 6 also shows the response to articles in different press outlets. Iexpect that business-related press outlets are more likely to provide informativearticles due to both their original reporting sources and their sophisticatedaudience. Consistent with this expectation, articles in the twobusiness-focused outlets (wire services and national business publications)generate a statistically negative response, while those in the other publicationsare indistinguishable from zero. The other finding of note in thisregression is the much greater magnitude of the response to stories fromthe wire services (−12.6%) than to those from the other publication outlets.This is consistent with the timeliness of wire coverage providing usefulinformation to sophisticated market participants.The third regression compares the response to articles from a recurringcolumn with those from nonrecurring authors. Again, I expect that30 The greater magnitude for the three-day response is consistent throughout the markettests. The relative magnitudes across categories remain consistent with the one-day measure.However, many of the statistical significance levels are weaker, suggesting that the inclusion ofthe three-day period introduces noise.31 As an alternative explanation for unconditional market response, it may be that presscoverage imposes costs on a firm even if the underlying information is already known (such aspotential litigation issues, stakeholder communication issues). In that case, there would be anegative response to all articles and no differential response across categories as all categoriesinvolved press coverage.32 Untabulated tests of medians show similar patterns of returns and levels of significanceto those in the mean/regression tests.1022 G. S. MILLERTA B L E 5Regression of Day Zero Abnormal Returns on Categories of Sources, Publications, and Recurring ArticlesF -Value F -Value(p-Value) (p-Value)Compared to Compared toCoefficient Reporter Coefficient Coefficient Coefficient ReporterCategory (p-Value) Generated (p-Value) (p-Value) (p-Value) GeneratedReporter-generated −13.9% −12.6%information (0.0001) (0.0004)Analyst −4.5% 4.45 −2.2% 5.69(0.0716) (0.0197) (0.2808) (0.0103)Legal cases −1.7% 5.38 1.4% 7.25(0.3376) (0.0120) (0.3753) (0.0047)Auditor resignation 0.5% 5.28 1.9% 5.40(0.4595) (0.0127) (0.3591) (0.6876)National business −6.3%(0.0247)Local market −4.3%(0.1317)Electronic business −12.6% −11.0%(0.0011) (0.0051)Trade publications −0.4%(0.4661)Recur −5.1% −1.5%#p#分页标题#e#(0.0079) (0.3611)Nonrecur −6.7%(0.0040)Abnormal returns are the firm market returns per CRSP, Datastream, or Bloomberg on the day the article is released less the CRSP market returns on the same day. Returns data were identifiedfor 60 firms. All variables are 0/1 indicators. Source is based on the source that is attributed as providing the primary information or to have initiated the article. National business publications arepublications that cover the national (or global) areas. This category includes The Wall Street Journal, Fortune, etc. Local market publications are generally regional papers, such as The San FranciscoChronicle, Chicago-Sun Times, etc. However, for the purposes of this table, they also include the two national nonbusiness publications (The New York Times and USA Today). Electronic businesspublications consist almost entirely of the Dow Jones News Service with one article from Bloomberg. Trade publications are based on covering a specific industry in depth. Examples includeBoating Industry and Business Insurance. An article is considered regional if it is from a local publication in the same region or a national publication but provides a byline indicating the story waswritten locally. Recurring articles are from a column that is published by a single reporter/set of reporters on a regular basis. As all articles are predicted to have a negative market response,p-values are one-sided.PRESS AS WATCHDOG 1023TA B L E 6Cross-sectional Examination of Characteristics of Firms and Fraud That Lead to an Article by the PressPanel A: Means and mediansp-Value ofVariable Caught Not Caught DifferencePRESSINT Mean 51.72 24.15 0.031Median 8.23 5.11 0.062ANALYST Mean 0.33 0.22 0.033Median 0 0 0.033MV Mean 12.6 11.8 0.014Median 12.5 11.7 0.048AUDITOR Mean 0.40 0.28 0.0316Median 0 0 0.0317BIGADS Mean 0.14 0.18 0.281Median 0 0 0.286NUMINV Mean 4.3 3.2 0.004Median 4 3 0.016AMOUNT Mean 3.00 2.06 0.001Median 2.78 2.04 0.002MISLEAD Mean 0.37 0.22 0.008Median 0 0 0.005STEAL Mean 0.33 0.20 0.018Median 0 0 0.012Panel B: Logit analyses (p-values in parentheses below coefficients)Caught = αt + β1VISIBILITY + β2AUDITOR + β3BIGADS + β4NUMINV + β5AMOUNT+β6MISLEAD + β7STEAL + εVisibility VariablePRESSINT ANALYST MVINTERCEPT −2.4321 −2.3958 −3.9020(0.0001) (0.0001) (0.0002)VISIBILITY 0.0028 0.3762 0.1502(0.0481) (0.1356) (0.0424)AUDITOR 0.6959 0.6654 0.3815(0.0123) (0.0158) (0.1465)BIGAD 0.0335 −0.0011 −0.2788(0.4678) (0.4990) (0.2805)NUMINV 0.1013 0.0824 0.1462(0.0356) (0.0656) (0.0117)AMOUNT 0.1833 0.1934 0.0976(0.0050) (0.0083) (0.1763)MISLEAD 0.6302 0.6664 0.6902#p#分页标题#e#(0.0251) (0.0192) (0.0437)STEAL 0.6090 0.6272 0.6139(0.0336) (0.0299) (0.0642)Likelihood ratio 299.41 301.06 213.19(χ2) (31.02) (29.37) (23.43)Two-tailed p-value for INTERCEPT, one-tailed for other variables.This table presents analyses of differences between firms for which fraud was identified (CAUGHT = 1) and thosefor which it was not (CAUGHT = 0). Sample size for all analyses other than those including MV is 263. Seventy-five firmsare coded as CAUGHT = 1. Sample size for analyses with MV is 181. Fifty-four of those firms are coded as CAUGHT = 1.PRESSINT is measured as the total number of articles by the press over the fraud violation period divided by the monthsof the violation. Articles were identified using Factiva. ANALYST is an indicator variable coded as 1 if the firm is in the topquartile of analyst following for the sample, 0 otherwise. Analyst following is measured at the last month of the violationperiod and is from I/B/E/S. MV is the log of market value of the firm in billions. Market value is measured on the last dayof the violation period and is obtained from CRSP. AUDITOR is an indicator variable coded as 1 if the auditor changedduring the period of the fraud or in the following year. Otherwise, coding is 0. BIGAD is an indicator variable coded as1 if the firm is in an industry that was in Advertising Age’s top 15 advertisers for 1985, 1990, 1995, and 2000. Otherwise,the coding is 0. NUMINV is the number of persons involved in the fraud as cited in the AAER. AMOUNT is the log ofthe sum of the dollar amount of the violations documented in the AAER. MISLEAD is an indicator variable coded as 1if the violations on the AAER included a materially misleading public statement or report, 0 otherwise. STEAL is an indicatorvariable coded as 1 if the AAER indicates management misappropriated funds as a portion of the fraud, 0 otherwise.1024 G. S. MILLERrecurring authors will be more likely to provide original analysis with stronginformation content. While the findings in table 4 suggest the authors ofrecurring articles are more likely to provide original analysis, the marketresponse is statistically equivalent across these two categories.The final column reexamines the response to the information source categorieswith controls for wire service and recurring articles. The response toarticles relying on analysts is lower in magnitude and no longer statisticallydifferent from zero, all other results are consistent with those previouslydiscussed. This provides further support for the contention that press reportingis most useful when based on analysis not available through otherinformation intermediaries rather than rebroadcasting of information. It isimportant to note that these tests are based on one day returns and thus aredesigned to exclude the period in which the nonpress analysis was initially#p#分页标题#e#provided (i.e., the day of the analyst report, auditor resignation, or lawsuit).To the degree that the test design successfully excludes these events, ittests only the importance of rebroadcasting—it does not examine whetherinformation from these sources had an initial impact on price.As an additional examination of information content, I collect daily tradingdata from the Trade and Quote (TAQ) database. TAQ data do not beginuntil 1993 and are also more restricted in coverage than CRSP data, resultingin data for only 37 firms. For each of these 37 firms, I aggregate the dataover the day of the article to calculate number of trades, volume, averagetrade size, and volatility. I calculate abnormal values by using the same datafor the period seven days prior to the article. Many of the variables (suchas number of trades) are highly correlated with size, thus, I use percentagechange in those variables. In an untabulated analysis, I find that article dayshave a statistically higher mean and median standard deviation in volatility.Further, trade volume increases by 375% while average number of trades increasesby 429%. Although there is an increase in trades of all sizes, there isa statistically significant shift upwards in the proportion of small trades (lessthan 1,000 shares). Combined, these data suggest that the articles includeinformation to which the market responds and appears to be most informativeto small traders (who are likely individuals). While cross-sectional testsgenerally confirm the findings regarding sources in table 5 (i.e., reportergeneratedinformation and auditors’ resignations are most informative),they are not robust to specification checks and thus are not presented.33Overall, the market return evidence indicates press articles are informativeto market participants. In particular, press articles based on reporteranalysis seem to create significant changes in the market’s assessment of afirm.33 For the returns and TAQ data, I randomly omit 10% of the sample and reperform theregressions 100 times each. Results for the returns data are generally consistent with those intable 6. The aggregate TAQ data results reported in the text are robust, but the cross-sectionalexaminations by categories are unstable.PRESS AS WATCHDOG 1025To further assess the economic importance of these actions, I examinedescriptive data regarding the eventual announcement of an issue by managementor the SEC. I find that press articles lead the “hard” announcementby a mean of 294 days. However, the median of 99 is more informative, as themean is skewed by the small number of firms that withhold any announcement.34 Medians for articles based on reporter analysis, analysts, and lawsuitsare relatively similar at 112, 92.5, and 96 days, respectively. However, the median#p#分页标题#e#for auditor changes is only 16 days. While I cannot provide a conclusiveanswer for this difference, it is likely that auditor changes attract the attentionof the SEC and prompt outsiders to begin questioning management,thus forcing an admission of an investigation. As robustness checks, I recalculateall previous tests excluding articles issued within 7 days or greater than365 days of the management announcement. Results remain qualitativelyand statistically similar. Finally, to provide a benchmark for the magnitudeof the market response to the articles, I examine the market response to themanagement/SEC announcement of the issue. I amable to identify returnsfor 50 firms with a mean response of −16.2%.354.2 FIRM AND FRAUD CHARACTERISTICS THAT INFLUENCE COVERAGEI next turn to investigating cross-sectional factors associated with whethera given firm-observation results in a published article. As discussed in section2.3, I expect that the press will consider costs and benefits to providingcoverage that vary with characteristics of both the firm visibility and fraudcommitted.I expect that reporters/publishers will maximize the benefits and minimizethe cost of articles by focusing on firms that have high visibility. Imeasure visibility with three variables. First, press intensity (PRESSINT) capturesthe amount of attention received in general by the press. Recall thatPRESSINT is calculated by identifying all articles written over the period ofthe violation and then deflating by the length of the violation in months.36Second, I examine analyst following (ANALYST) measured in the last monthof the violation period. Consistent with prior research, I address the skewnessof analyst following by using a dichotomous coding of 1 if the firmhas analyst following in the top quartile of the distribution of my sample34 An examination of these firms indicates the majority were companies that were usedexplicitly as a vehicle for fraud rather than legitimate companies that became involved in frauds.For example, several were expost identified as being controlled by serial market manipulators.35 I identify returns for 118 firms that do not receive an article. The response of −18.8% isstatistically equivalent to that of the firms that receive an article. Given evidence in the crosssectionalreturns tests that information leaks to the market occur through other intermediariesand findings in the subsequent determinants of article tests that both firms and frauds aresystematically different across the groups, I cannot provide an ex ante expectation of differenceson these dates.36 As robustness checks, all tests are run using press coverage in the quarter prior to thefraud and in the two weeks prior to the AAER announcement. All results are similar.#p#分页标题#e#1026 G. S. MILLER(Bradshaw, Bushee, and Miller ).37 The third variable is the log ofmarket value (MV ) of the firm measured at the end of the period of fraud.Consistent with prior research (Feroz, Park, and Pastena ), I find thatmany firms cannot be identified on CRSP (82, or 31% of the sample, almostevenly distributed between the caught and not caught subsamples). Obviously,missing firms are excluded in the MV analyses.38 However, they areincluded in all other analyses. Untabulated analyses find that the three variablesmeant to capture information environment are highly correlated. Eventhe lowest correlation, ANALYST and PRESSINT, has a Pearson (Spearman)correlation of 0.64 (0.72). The Cronbach’s alpha for the three variables is0.83, which exceeds Nunnally’s  suggested value of 0.70 for a reliableindex. Finally, factor analysis of the three variables finds only one factor.When multiple factors are forced, the first factor loads with an eigenvalue of1.72; all forced factors have negative eigenvalues, implying that they reducethe ability to explain variation in comovement. Given this high correlation,I do not combine these variables in my logit regressions. Rather, I run thelogit three times, once with each information environment variable.39Auditors are an important information intermediary and changes in auditormay signal an accounting issue to the press. I identify changes in auditorsboth during and one year after the fraud using Compustat. For firms missingthe Compustat data items or with auditor identified as “other,” I perform asearch of Factiva for any indication of auditor changes.I examine whether the press is less likely to publish articles for firms ina high advertising industry by including an indicator variable if the pressis in a high advertising industry. I use industry, rather than firm-specific,advertising spending, as publishers must consider future, not just current,advertising spending.40 My dependent variable is coded based on whether37 This variable is calculated using I/B/E/S data. If a firm cannot be identified on I/B/E/S,analyst following is assumed to be zero. Results are similar if such firms are discarded. Results arealso similar if the variable is left in its raw continuous form, logged, measured at the beginningof the violation period, and as an average over the violation period38 As robustness checks, all tests are also run using market value at the beginning of thefraud period and average market value over the fraud period. All results are similar. I also usemultiple imputations to generate observations for missing market data (Yuan ). Resultsare similar.39 A logit using the factor score in place of the information variables finds a positive andsignificant coefficient. The findings on all other variables remain the same as those presented#p#分页标题#e#in the tables. I do not use the factor score in the primary analysis due to the large number oflost observations caused by missing market values.40 Consistent with the CRSP coverage, many of these firms are not covered on Compustat. Areview of several of these firms indicates that they tend to file abbreviated financial statements(small business 10-Q) and thus do not provide the financial statements used by Compustat tocreate their data. Several firms that are covered by CRSP are not covered on Compustat, and viceversa, suggesting that the lack of coverage is database specific, rather than an inability to identifycusips and permnos. As a robustness test, I reperform these tests using firm-specific spendingas available on Compustat. Due to the frequency that this variable is missing, I perform thetest using only advertising during the period of the fraud and using advertising for the closestperiod included on CRSP. The variable is insignificant in all variations.PRESS AS WATCHDOG 1027any member of the press identifies accounting fraud, so I do not attemptto isolate spending with a given publisher. This variable is created from theannual advertising rankings provided by Advertising Age.Aspects of the accounting fraud may influence whether the press detectsthe fraud and writes an article. First, I expect that frauds with more peopleinvolved are more likely to result in a leak, thus decreasing the costof following the fraud. I measure the number of people involved from theAAER (NUMINV ). Second, severe frauds are more likely to be of interestto the public, and thus to offer a higher benefit to members of the press.I measure severity by the magnitude of the total dollars included in theAAER (AMOUNT).41 Third, I expect frauds that involve material publicmisrepresentations and misleading statements (through a press release orfinancial statements) (MISLEAD) to be more likely to attract press and public.This both decreases costs and increases the benefit of publishing anarticle. Fourth, I expect that fraud that involves some form of managementmisappropriation of funds will be viewed as more intriguing. A companyis coded as having such misappropriation (STEAL) if the AAER includedcensure for insider trading, illicit payments, undisclosed compensation, orpure theft.Table 6, panel A provides univariate and panel B provides logit analysesof the difference in these variables between firms caught and those notcaught by the press. The results indicate that visibility variables are relatedto the likelihood that the press will publish an article regarding accountingfraud. Caught firms have a significantly higher mean and median level ofPRESSINT, ANALYST, and MV . Findings are similar in the logit analysis,except for the ANALYST variable, which is no longer statistically significant.#p#分页标题#e#Overall, results are consistent with the press publishing articles based on thebenefits of attracting a large reader base and finding it easier to investigatefirms that have a readily available body of information.Univariate and logit results indicate that a firm is more likely to be coveredin an article if there is an auditor change.42 However, if I exclude the firmsfor which auditor was coded as the source of the article, this variable is nolonger significant. All other variables remain unchanged. This suggests thatthe press accurately reports when they rely on auditor changes, providingfurther support for the source classifications discussed previously in thepaper.http://www.ukassignment.org/Accounting_Essay/The univariate and logit results provide no evidence that press coverageis correlated to current or future advertising revenue. This may be due tonoise in my industry-level data and/or the fact that many high-advertising41 Unfortunately, the AAER are not always clear on magnitude. Further, the types of fraudvary—making aggregation somewhat problematic. I expect these aggregation issues will addnoise, but have no reason to anticipate bias.42 The logit results using MV as the visibility variable have lower statistical significance forAUDITOR and AMOUNT. Given the large number of observations missing data, I recalculatethese tests using multiple imputation to generate MV . AUDITOR becomes significant, butAMOUNT remains insignificant.1028 G. S. MILLERindustries may not advertise in the business press (e.g., consumer productcompanies may not view the readers of the business press as their target market).However, findings from concurrent research suggest that the cause maybe more complex. Reuter and Zitzewitz  study whether advertisingspending by mutual fund companies impacts the likelihood that the fundwill be recommended in a financial publication. They find no evidence ofbias for The Wall Street Journal or The New York Times, but evidence consistentwith bias for Kiplinger’s, Money, and Smart Money. Interestingly, many ofthe caught articles in my sample are found in The Wall Street Journal andits related Dow Jones Newswire (14 articles each), and The New York Timesalso provides one article. There are no articles in the three magazines withevidence consistent with bias in the Reuter and Zitzewitz  study. Advertisingspending may impact some publications, with the remaining outletscontinuing to serve a monitoring function.43Table 6 provides strong evidence that fraud characteristics are associatedwith the likelihood of an article. Frauds that involve a large number of individualsare statistically more likely to result in an article in all specifications.Similarly, the dollar magnitude of the fraud is a significant predictor in all#p#分页标题#e#tests other than the logit that includes market value of equity. Given thehigh correlation between these variables, the lower significance in that regressionis not surprising.44 Frauds that involve public misleading statementsare more likely to result in an article being written. Frauds that involve misappropriationof funds are also more likely to result in an article, consistentwith the press finding deviant stories more newsworthy.As a robustness test, each statistical analysis is reperformed randomlyexcluding approximately 5% of the sample. Results are consistent with thosepresented in the paper.The results that both firm and fraud characteristics are important suggestthat there may be an interaction between these variables. Ex ante, itis difficult to predict whether it is complementary (larger firms and moreinteresting fraud are required) or a substitute (either large firm or interestingfraud). Accordingly, I recalculate all tests including an indicator variable43 This conjecture is also consistent with a Neiman Foundation survey of members of thepress. The journalist surveyed voted The New York Times and Washington Post as doing the bestjob of watchdog journalism, followed by The Wall Street Journal. The votes for other outlets werenot given. See the Neiman Web site for details of this survey.44 I use the total magnitude of the fraud, rather than a deflated number, since the press ismost likely to focus on the undeflated number. If I deflate by assets (or market value of equitywhen assets are not available), the variable is generally not significant. Other variables are notimpacted. I also replace the magnitude of the fraud with the market response to the officialannouncement of an issue. Given the large number of missing observations, I perform thiscalculation using available data and also with multiple imputations to generate data for firmswith missing returns (Yuan ). I find that articles are more likely if there was a morenegative market response. Results for the other variables are similar to those reported, exceptunder a few specifications results for analysts become significant while those for auditors andtheft weaken.PRESS AS WATCHDOG 1029for high (low) visibility, which is coded as one if the firm is in the top (bottom)quartile of the visibility measure (PRESSINT, ANALYST, MV ). Theindicator variable is interacted with each of the fraud variables and run separatelyfor high/low visibility. In general, the (untabulated) results do notfind support for interactions. The only consistent exception is low visibilityinteracted with publicly misleading statements. The negative coefficient onthat variable indicates that publicly misleading statements are less likely toimpact coverage for less visible firms.Finally, I also perform regressions comparing articles that are based on#p#分页标题#e#reporter-generated analysis with those that rebroadcast from other informationintermediaries. I find no statistical difference. However, given the smallsize of this sample, power is likely an issue.Overall, the evidence in this section indicates that the press publishesarticles regarding accounting fraud consistent with a cost/benefit maximization.This suggests that the press may be a more effective watchdog forsome types of firms and frauds than for others.5. ConclusionAcademic research on the press’s role in commerce has been limited(Zingales ). The evidence that does exist provides a conflictingportrait of the press. Some studies suggest that the press provides littleuseful analysis, instead focusing on entertainment value and sensationalism(Jensen , Core, Guay, and Larcker , Dyck and Zingales[2002b]). Other studies find that, at the macro level, the press is an importantfactor in impacting economic development and corporate governance(Djankov et al. , Dyck and Zingales [2002a], Dyck, Volchkova, andZingales ).In this paper, I study press coverage of a group of firms that have beeninvolved in accounting malfeasances. This firm-level examination of pressinvolvement in a highly technical area of analysis provides an opportunityto greatly enhance our understanding of whether and how the press contributesto financial information flows. Specifically, I examine whether thepress is involved in the early public identification of accounting malfeasances,whether that identification is based on original analysis or rebroadcasting,the information content, and whether the press is systematicallybiased regarding the type of firms it covers.Using a sample of firms sanctioned by the SEC for accounting malfeasances,I find that the press is involved in early public identification in 29%of the cases. Many articles rebroadcast allegations made by other informationintermediaries (analysts, auditors, and lawsuits), but there is also asubstantial number of articles based on press analysis. The press analysisarticles provide new information to the capital markets, while articles thatrebroadcast information from other intermediaries do not. Consistent witha dual role of the press, I find that publications and authors that are likelyto specialize in business information are also more likely to provide articles1030 G. S. MILLERthat rely on reporter-based analysis, while more general press outlets almostexclusively rebroadcast from other information intermediaries.An analysis of the types of firms and frauds that receive press coveragefinds systematic biases in press coverage. These biases are consistent withthe press trading off the costs of identifying frauds with the benefits ofproviding interesting articles that will attract a wide range of readers. On#p#分页标题#e#the firm level, more visible firms and those with auditor changes are morelikely to receive an article. Both of these are consistent with a reduced costof reporting, and more visible firms are also likely to attract a larger audience.Consistent with this cost/benefit model, there are also several characteristicsat the fraud level that impact the likelihood of an article. I findthat the press is more likely to write an article when the fraud involves alarge number of people who may provide information to the press, consistentwith a higher potential for leaked information and thus lower costof analysis. Further, the press is more likely to cover large frauds, as theegregious nature of these frauds is likely to capture reader attention. Thepress is also more likely to cover firms when there is a misleading publicstatement. Such statements may attract the press’s attention initially, thusreducing search costs. This high ex ante visibility also suggests a subsequentarticle will have a large prospective reader base. As further evidence ofthe press’s desire to cover frauds the public finds compelling, the pressis more likely to write an article if managerial misappropriation of funds hasoccurred.This paper provides evidence regarding the role of the press as a watchdogfor accounting fraud. In addition to contributing to our understandingof how accounting fraud is brought to the public’s attention, this settingallows an examination of how the press undertakes reporting that involvesanalyzing and understanding highly technical information. This setting providesa strong research design for establishing a greater understanding ofthe press’s informational role in general. However, several caveats apply.First, cross-sectional analysis of press coverage at the firm level is inherentlysubject to a large number of judgments in coding data. Prior researchhas either focused primarily on country-level data or used clinical studiesexamining only one firm in one unique situation. While this has avoidedmany of the judgment issues, it also makes it difficult to develop an understandingof press activities. Second, my sample relies on firms beingidentified as having inappropriate accounting by the SEC and having anAAER written regarding that accounting. These are likely extreme cases ofmalfeasances, and thus readers should be careful in attempting to generalizethese results to the population as a whole. Third, the SEC likely investigatesfirms that have been identified in the press as engaging in some form ofaccounting malfeasances. In that case, the proportion of firms “caught” bythe press may be overstated. That is, many other frauds may have occurredand remain undetected by either the press or the SEC. Again, this suggests#p#分页标题#e#readers should use caution when trying to generalize to the population offirms.PRESS AS WATCHDOG 1031A P P E N D I X 1Definition of Variables Presented in Order They Appear in TablesPanel A: Early identification of fraudVariable Name Definition/SourceCAUGHT Indicator variable coded as one if the press writes an articlealleging accounting malfeasances prior to a company or SECannouncement regarding an issue. Articles and managementdisclosures were identified through a Factiva search.Reporter Generated Information Indicator variable coded as one if the article indicates the argumentfor an accounting issue is based on reporter analysis rather thanreports from information intermediaries. This includes analysisof SEC documents, financial statements, firm disclosures, andtips from customers, industry insiders, anonymous sources, etc.Coded based on review of the article retrieved from Factiva.Analyst Indicator variable coded as one if the article indicates theargument for an accounting issue is based on rebroadcastinganalysts’ allegations. Coded based on review of the articleretrieved from Factiva.Legal cases Indicator variable coded as one if the article indicates theargument for an accounting issue is based on rebroadcastinglawsuits or public criminal investigation. Coded based on reviewof the article retrieved from Factiva.Auditor Resignation Indicator variable coded as one if the article indicates theargument for an accounting issue is based on rebroadcastingauditor resignations. Coded based on review of the articleretrieved from Factiva.National Business Indicator variable coded as one if the article was published in aperiodical focused on business with nationalcoverage/readership. Examples include The Wall Street Journaland Fortune. Coded based on review of the article retrieved fromFactiva.Local Market Indicator variable coded as one if the article was published in aperiodical that covers general news with predominately localcoverage/readership. Examples include The San FranciscoChronicle and Chicago Sun-Times. Coded based on review of thearticle retrieved from Factiva.Electronic Business Indicator variable coded as one if the article was published by anelectronic wire service focused on business. Examples includeDow Jones Newswire and Bloomberg. Coded based on review ofthe article retrieved from Factiva.Trade Publications Indicator variable coded as one if the article was published in aperiodical that covers specialized industry news. Examplesinclude Boating Industry and Business Insurance. Coded based onreview of the article retrieved from Factiva.National Non-business Indicator variable coded as one if the article was published in a#p#分页标题#e#periodical that covers general news with predominately nationalcoverage/readership. Examples include USA Today and The NewYork Times. Coded based on review of the article retrieved fromFactiva.Recur Indicator variable coded as one if the article is from a column thatis published by a single reporter/set of reporters on a regularbasis. Coded based on review of the article retrieved from Factiva.Non-recur Indicator variable coded as one if the article is not from a columnthat is published by a single reporter/set of reporters on aregular basis. Coded based on review of the article retrievedfrom Factiva.1032 G. S. MILLERA P P E N D I X 1—ContinuedPanel B: Firm and fraud characteristicsVariable Name Definition/SourcePRESSINT A measure of the level of press coverage during the period offraud. Calculated as total number of press articles during theperiod of the fraud divided by number of months in the fraud.Number of press articles was retrieved from a Factiva search.ANALYST Indicator variable coded as one if the firm is in the top quartile ofanalysts following for the sample. Analyst coverage based onI/B/E/S.MV The log of market value measured in billions. Market value ismeasured on the last day of the violation period and is obtainedfrom CRSP.http://www.ukassignment.org/Accounting_Essay/AUDITOR Indicator variable coded as one if the auditor changes during theperiod of the fraud or within one year following the fraud.Auditor changes are identified using Compustat. A Factiva searchwas also performed if auditor status was unclear on Compustat.BIGAD Indicator variable coded as one if the firm is in an industry that wasin Advertising Age’s top 15 advertisers for 1985, 1990, 1995, and2000.NUMINV The number of people indicated as involved in the fraud by theSEC. Calculated by summing participants across all SEC AAERrelated to the fraud.AMOUNT The log of the sum of the dollar amount of violations. Calculatedby summing amounts in all SEC AAER related to the fraud.MISLEAD An indicator variable coded as one if the fraud included amaterially misleading public statement or report. Calculated byreviewing all SEC AAER related to the fraud.STEAL An indicator variable coded as one if managementmisappropriated funds through stock trades, bonus plans, oroutright theft during the fraud. Calculated by reviewing all SEC.REFERENCESBEHR, P. “Looking for Answers in the Enron Story.” Neiman Reports 56 (2002): 8–9.BENEISH, M. D. “Incentives and Penalties Related to Earnings Overstatements that ViolateGAAP.” The Accounting Review 74 (1999a): 425–57.BENEISH,M.D. “The Detection of Earnings Manipulation.” Financial Analysts Journal 55 (1999b):#p#分页标题#e#24–36.BONNER, S.; Z. PALMROSE; AND S. YOUNG. “Fraud Type and Auditor Litigation: An Analysis ofSEC Accounting and Auditing Enforcement Releases.” Accounting Review 73 (1998): 503–32.BRADSHAW, M.; B. BUSHEE; AND G.MILLER. “Accounting Choice, Home Bias, and US Investmentin Non-US Firms.” Journal of Accounting Research 42 (2004): 795–841.BUSHMAN, R.; J. PIOTROSKI; AND A. SMITH. “What Determines Corporate Transparency?” Workingpaper, University of North Carolina–Chapel Hill, 2003.CORE, J.; W. GUAY; AND D. LARCKER. “The Power of the Pen and Executive Compensation.”Working paper, University of Pennsylvania, 2005.DEANGELO, H.; L. DEANGELO; AND S. GILSON. “The Collapse of First Executive Corporation:Junk Bonds, Adverse Publicity, and the ‘Run on the Bank’ Phenomenon.” Journal of FinancialEconomics 36 (1994): 288–336.DEANGELO, H.; L. DEANGELO; AND S. GILSON. “Perceptions and the Politics of Finance: JunkBonds and the Regulatory Seizure of First Capital Life.” Journal of Financial Economics 44(1996): 475–511.DECHOW, P.; R. SLOAN; AND A. SWEENEY. “Causes and Consequences of Earnings Manipulation:An Analysis of Firms Subject to Enforcement Actions by the SEC.” Contemporary AccountingResearch 13 (1996): 1–36.PRESS AS WATCHDOG 1033DEFOND, M., AND B. SMITH. “Discussion of the Financial and Market Effects of the SEC’s Accountingand Auditing Enforcement Releases.” Journal of Accounting Research 29 (Supplement1991): 143–8.DJANKOV, S.; C.MCLIESH; T. NENOVA; AND A. SHLEIFER. “Who Owns the Media?”Working paper,Harvard University, 2002.DYCK, A.; N. VOLCHKOVA; AND L. ZINGALES. “The Corporate Governance Role of the Media:Evidence from Russia.” Working paper, University of Toronto, 2005.DYCK, A., AND L. ZINGALES. “The Corporate Governance Role of the Media,” in The Right to Tell—The Role of Mass Media in Economic Development. Washington, DC: The World Bank Institute,2002a: 107–37.DYCK, A., AND L. ZINGALES. “The Bubble and the Media,” in Corporate Governance and CapitalFlows in a Global Economy. New York, NY: Oxford University Press, 2002b: 83–102.DYCK, A., AND L. ZINGALES. “Media and Asset Prices.”Working paper, Harvard University, 2003.FEROZ, E.; K. PARK; AND V. PASTENA. “The Financial and Market Effects of the SEC’s Accountingand Auditing Enforcement Releases.” Journal of Accounting Research 29 (Supplement 1991):107–42.FROST, C.; E. GORDON; AND A. HAYES. “Stock Exchange Disclosure and Market Liquidity: AnAnalysis of 50 International Exchanges.” Working paper, Dartmouth College, 2001.#p#分页标题#e#HAW, I.; B. HU; L. HWANG; AND W. WU. “Ultimate Ownership, Income Management, and Legaland Extra-Legal Institutions.” Working paper, Texas Christian University, 2003.HERMAN, E. “The Media and Market in the United States,” in The Right to Tell – The Role of MassMedia in Economic Development. Washington, DC: The World Bank Institute, 2002: 61–81.HUBERMAN, G., AND T. REGEV. “Contagious Speculation and a Cure for Cancer: A Noneventthat Made Stock Prices Soar.” Journal of Finance 56 (2001): 387–96.ISLAM, R. “Into the Looking Glass: What the Media Tell and Why – An Overview,” in The Rightto Tell – The Role of Mass Media in Economic Development. Washington, DC: The World Bank,2002: 1–23.JAMIESON, K., AND K.CAMPBELL. The Interplay of Influence, Fifth edition. Belmont, CA:Wadsworth,2001.JENSEN, M. “Toward a Theory of the Press,” in Economics and Social Institutions, edited by K.Brunner. Boston, MA: Martinus Nijhoff, 1979.KELLER, J. “How to Keep Watchdogs Watching.” Neiman Reports 52 (1998): 53–8.MADRICK, J. “A Good Story Isn’t Always the Right One to Tell.” Neiman Reports 56 (2002): 6–7.MCNAIR, J. “Corporations Work Hard to Prevent Reporting.” Neiman Reports 56 (2002): 12–4.Mullainathan, S., AND A. Shleifer. “Media Bias.” Working paper, Harvard University, 2002.NUNNALLY, J. Psychometric Theory. New York: McGraw-Hill Companies, 1978.REUTER, J., AND E. ZITZEWITZ. “Do Ads Influence Editors? Advertising and Content Biases inthe Financial Media.” Working paper, University of Oregon, 2003.SERRIN, J., AND W. SERRIN. Muckraking! The Journalism that Changed America. New York, NY: TheNew Press, 2002.STEIGER, P. “Not Every Journalist ‘Missed’ the Enron Story.” Neiman Reports 56 (2002): 10–2.STROMBERG, D. “Mass Media Competition, Political Competition, and Public Policy.” Workingpaper, Institute for International Economic Studies, 2002.THOMPSON, R.; C. OLSEN; AND J. DIETRICH. “Attributes of News About Firms: An Analysis ofFirm-Specific News Reported in the Wall Street Journal Index.” Journal of Accounting Research25 (1987): 245–73.WASIK, B. “Dismal Beat – The March of Personal Finance Journalism.” Harpers (March 1, 2003):81–7.WATTS, R., AND J. ZIMMERMAN. Positive Accounting Theory. Englewood Cliffs, NJ: Prentice Hall,1986.YUAN, Y. Multiple Imputation for Missing Data: Concepts and New Development. Rockville, MD:SAS Institute Inc., 2005. Accessed online at multipleimputation.pdf.ZINGALES, L. “In Search of New Foundations.” The Journal of Finance 55 (2000): 1623–53.#p#分页标题#e#