会计欺诈外文翻译文献(文档含中英文对照即英文原文和中文翻译)译文：会计欺诈和机构投资者查得·R·拉森介绍美国资本市场依赖财务报告系统来帮助有效分配资本。
最近的财务报告过程中故障在许多高调公司新的人员的监管机构,会计欺诈和市场参与者的兴趣。
两个重要的经验规律的文献记录极端操纵收益的决定因素和后果。
首先,股票市场反应的启示会计处理显著负面。
估计下降公告后的市场价值会计操作范围从20 - 40%( Palm rose、理查德森和 2003李,和马丁2007)。
第二,会计操作是可预测的。
文献会计操作的文档可以预测的措施准确的质量、会计性能、非金融变量声明和股市变量(如,Beneish 1999;Dechow,通用电气,拉尔森和斯隆2007)。
虽然会计操作导致重大投资损失和与公司相关的特点和性能,几乎没有证据表明存在成熟的投资者是否能够避免损失与会计欺诈。
机构投资者已成为市场的重要力量在过去的几十年。
上世纪八十年代初到九十年代末,机构投资者所有权翻倍,股市50%以上(龚帕斯和Metric时,2001)。
机构投资者在美国市场存在的上升有意义的促使文献调查他们是否执行是有利可图的交易。
文献的结果是喜忧参半。
几项研究文档积极变化之间的相互关系,这些机构投资者的资产和未来的收益和回报,这表明机构通知交易员(如，柯和2005;阿里列弗,确认我也承认泰德•克里斯坦森的指导,没有它我就不会了博士学位的挑战。
最后,我感谢我的家人的支持。
没有这篇论文的完成并不意味着几乎一样多。
介绍美国资本市场依赖财务报告系统来帮助有效分配资本。
最近的财务报告过程中故障在许多高调公司新的人员监管机构,会计欺诈和市场参与者的兴趣。
两个重要的经验规律的文献记录极端操纵收益的决定因素和后果。
首先,股票市场反应的启示,会计处理显著负面。
估计下降公告后的市场价值会计操作范围从20 - 40%(Palm rose、理查德森和朔尔茨2003;Karpo_,李,和马丁2007)。
第二,会计操作是可预测的。
文献会计操作的文档可以预测的措施,合格的质量、会计性能、非金融变量声明和股市变量(如,Beneish 1999;Dechow,通用电气,拉尔森和斯隆2007)。
虽然会计操作导致重大投资损失和与公司相关的特点和性能,几乎没有证据表明存在成熟的投资者是否能够避免损失与会计欺诈。
机构投资者已成为市场的重要力量在过去的几十年。
从八十年代初到九十年代末,机构投资者所有权翻倍,股市50%以上(。
机构投资者在美国市场存在的上升促使文献调查他们是否执行有利可图的交易。
文献的结果是喜忧参半。
几项研究文档积极变化之间的相互关系,这些机构投资者的资产和未来的收益和回报,这表明机构通知交易员(如。
、柯和Ramalingegowda 2005;柯和Petron 2004;阿里Dutch列弗,Trembles 2004)。
另一方面,一些文献表明,知情交易可能更有限,发现性能优越的共同基金很少持续(Carat 1997;布朗一个曼1995年)和交易考虑通知可能只是动量交易的结果(Bushee和古德曼,2007)。
会计欺诈和大市场的可预测性与会计欺诈相关的损失表明,它是一个理想的设定检查成熟的机构投资者。
如果机构投资者拥有优越的信息和复杂的会计信息的使用者对会计欺诈,他们应该在欺诈上市公司公开披露前欺诈。
我的主要研究问题是机构投资者预期会计欺诈的启示和剥离的股票欺诈公司公开披露之前骗子。
作为一个次要的研究问题,我检查是否机构作为有效的公司监控预防欺诈。
我使用的会计、审计和执行版本(AAER)涉及欺诈和会计操作作为一个代理第一新闻文章Factiva提及会计违规的公众披露欺诈。
我检查机构交易模式在322年企业,美国证券交易委员会(SEC)中标识执行行动从1982年到2005年有操纵会计收益。
我的分析是在两个阶段进行。
第一阶段是本文分析聚合机构公司级和欺诈检查他们的交易行为。
第二阶段是一个阶段分析,利用机构投资者之间的异构性,欺诈行为检查他们的交易行为。
公司的分析,我遵循Bushee(2001)组织机构分为三类根据自己的投资风格:瞬态和专用。
多元化的投资组合和较低的投资组合营业额机构的特点。
多样化的投资组合,投资组合交易描述瞬态高机构,和高度集中的投资组合和较低的投资组合营业额专门机构的特点。
符合文学之前,我希望发现瞬态机构最有可能发起有利交易欺诈启示的预期和机构不大可能发起有利益可得交易欺诈的预期启示我没有强烈的专门机构研究预测通常找到贸易基于即将到来的未来事件。
然而,欺诈是一个独特的设置可能导致专门机构剥离他们的位置。
如果专门机构投资公司基于他们的信心的和意愿完整性管理、检测欺诈行为会引起专门机构剥离他们的股份。
此外,由于专门机构的特点是高度的投资组合,他们可能会有更大比例的投资组合风险欺诈是否显示。
因此,他们可能会有最强的激励预测欺诈和诈骗剥离他们的股票。
文献表明,机构投资者作为公司监控(如果是这样的话,那么有可能是诈骗公司低水平的机构投资欺诈行为,因为他们之前缺乏效率监控。
因此,我的第一组测试检查机构所有权水平是否欺诈公司立即释放之前第一次欺诈收益报告不同于人口控制的公司。
在不变的分析中,我发现欺诈公司实际上有更高水平的总机构所有权,所有权和瞬态比公司所有权制度。
专门机构所有权没有显著区别样本公司。
接下来,我将进行回归分析,控制公司的特点。
我立即发现欺诈的开始之前,机构所有权欺诈公司的总体水平高于制度所有权控制公司的一个示例。
然而,我发现更高层次的机构主要是所有权的结果表明斜面更高层次的瞬态机构所有权,专门机构所有权控制后几乎是相同的形式特征。
大学和回归结果表明,机构所有权水平不作为一个伶俐的监控装置在欺诈的预防。
我的下一套测试提供相关的证据,我的主要研究问题。
我第一次检查机构所有权水平变化在欺诈公司在此期间公司提交欺诈。
季度发行前的第一次欺诈收益报告,直到季度会计欺诈的公开披露之前,我发现机构所有权欺诈公司增加了近14%,代表一家欺诈公司已发行股份的 3.9%。
2因为欺诈公司经验股价下跌约35%一旦发现欺诈,欺诈的机构增加3.9%所有权不是微不足道的。
事实上,计算表明,总机构损失322年我的样品是诈骗公司的1380亿美元。
增加3.9%机构所有权欺诈时期代表约200亿美元的损失。
之前的研究已经检查机构是否能预测即将发生的事件在短窗口(Briber,詹金斯和王,2006)。
因此,在我的下一组测试,我观察机构所有权的变化后的季度马上前,公开揭露欺诈。
我在季度立即欺诈曝光之前,所有权制度降低了大约一个半欺诈公司已发行股份的百分比。
我发现能有效的降低瞬态机构持有,而专门的机构持有的变化无关紧要的不会。
我也在季后立即找到有效减少欺诈启示。
这些结果是强劲的几个控制变量包括现在和过去的股票回报,意想不到的收益,以及分享营业额的变化。
虽然我找到证据表明瞬态机构能够预测欺诈前一个时期它的启示,这些证据必须解释的证据我之前测试。
一个半百分比下降机构所有权欺诈曝光之前,虽然显著,稍微减轻相当大的机构投资者的损失。
机构更异于三类我受聘于企业层面分析。
因此,我进行第二次分析进一步利用机构投资者之间的异质性。
我创建代理机构的信息环境和机构的激励机制,以避免负面的市场后果与会计欺诈的启示。
条件拥有欺诈公司股票欺诈开始之前,我测试是否与机构的所有权的变化相关联的代理是欺诈公司前会计欺诈的启示。
结果提供一些证据表明与最强的激励制度,以避免会计欺诈和最高的质量信息环境剥离前股票欺诈公司会计欺诈的启示。
尽管数据符合资产剥离率的增加在这些机构中,我无法确定这些关系的结果通知交易或自然的所有权水平均值回归。
这项研究应该感兴趣的机构投资者和研究者。
研究结果表明,机构投资者失去钱的重要性通过投资公司提交会计欺诈。
进一步的证据有助于我的研究文献记录复杂的机构投资者。
至少在这个特殊的背景下,大多数机构似乎没有复杂的会计信息用户;然而,我确实提供了有限的证据前立即通知本季度交易欺诈机构之间的一个子集的启示。
这些投资的标准可能会导致这些机构倾斜特征,更有可能来证明他们的投资组合的审慎投资。
例如发现高的银行向企业倾斜投资组合标准普尔股票评级。
Bushee和古德曼(2007) ,这是一个指示符变量等于一个如果一个机构的市场价值的股票投资组合的五等分顶层，否则所有机构在一个特定的季度和0。
因为大多数机构拥有更多的资源,我希望ISIZE是一个机构的代理获取和处理信息的能力。
因此,我认为,大型机构将更有可能出售公司的股票有欺诈行为。
我发现了两个额外的措施,代理机构的私人信息和激励措施,以避免会计欺诈。
选择,第一个是一个变量,措施的相对大小的股权机构风险在一个特定的公司。
打赌测量作为应声股本旗下机构j公司我在一季度t扩展机构j的总市场价值的投资组合在季度t。
我希望押注是负相关的制度变迁在欺诈公司的所有权。
与更高水平的机构选择相对比水平较低的企业风险价值选择,因此,这些机构有更大的激励来收集私人信息,避免投资公司有欺诈行为。
我最后的机构公司水平变量,这是一个指示符变量等于一个如果一个机构持有的流通股总量的百分比在公司五等分顶层的制度,公司所有权和零。
我希望块与私人信息优势,因为这些机构更有可能获得私人信息和更愿意承担私人信息采集和处理的成本。
因此,我希望阻止将负相关的机构持有的诈骗公司的变化。
Bushee和古德曼(2007)是第一个采用这两种措施指出是一个很好的衡量去激励收集公司信息。
原文：Accounting Fraud and Institutional InvestorsBy Chad R. LarsonI also acknowledge the mentorship of Ted Christensen without which I would have never taken on the challenge of a doctorate. Lastly, I am grateful for the support of my family. Without them the completion of this dissertation would not mean nearly as much. Introduction U.S. capital markets rely on financial reporting systems to help effectively allocate capital. The recent breakdowns in the financial reporting process at many high profile companies have renewed researchers', and market participants' interest in accounting fraud. Two important empirical regularities emerge from the body of literature documenting the determinants and consequences of extreme earnings manipulations. First, stock market reactions to the revelation of accounting manipulations are significantly negative. Estimated declines in market value following the public announcement of accounting manipulations range from 20 to 40 percent (Palmore, Richardson, and Scholz 2003; Karpo_, Lee, and Martin 2007). Second, accounting manipulations are predictable. A body of literature documents that accounting manipulations can be predicted with measures of accurate quality, accounting performance, non-financial statement variables, and stock market variables (e.g., Beneish 1999; Dechow, Ge, Larson, and Sloan 2007). Although accounting manipulations result in significant investor losses and are associated with firm characteristics and performance, little evidence exists on whether sophisticated investors are able to avoid losses associated with accounting fraud..Institutional investors have become a significant market force over the last several decades. From the early 1980s to the late 1990s, institutional investors doubled their ownership in the equity markets to over 50 percent (Gompers and Metrick, 2001). The rising presence of institutional investors in the U.S. markets has spurred a signi_cant body of literature investigating whether they execute pro_table trades. The results of the literature are mixed. Several studies document positive associations between changes in institutional investors' holdings and future earnings and returns, suggesting that institutions are informed traders (e.g., Ke and Ramalingegowda 2005; Ke andPetroni 2004; Ali, Durtschi, Lev, and Thrombley 2004). On the other hand, some literature suggests that informed trading might be more limited, finding that superior mutual fund performance is rarely persistent (Carhart 1997; Brown an Goetzmann 1995) and trading patterns previously considered informed might simply be the result of momentum trading (Bushee and Goodman, 2007).The predictability of accounting fraud and the large market losses associated with accounting fraud suggest that it is an ideal setting to examine the sophistication of institutional investors. If institutional investors possess superior information and are sophisticated users of accounting information with respect to accounting fraud, they should sell shares in fraud firms prior to public revelations of fraud. My primary research question is whether institutional investors anticipate accounting fraud revelations and divest shares in fraud firms prior to the public revelation of frauds. As a secondary research question, I examine whether institutions act as effective firm monitors in the prevention of fraud. I use Accounting, Auditing, and Enforcement Releases (AAER) involving accounting manipulations as a proxy for fraud and the first press article in Factiva mentioning an accounting irregularity as the public revelation of fraud.1 I examine institutional trading patterns in 322 firms that the Securities and Exchange Commission (SEC) identified in enforcement actions from 1982 through 2005 as having manipulated their accounting earnings. My analysis is conducted in two stages. The first stage is a firm-level analysis that aggregates institutions at the firm-level and examines their trading behavior in fraud _rms. The second stage is an institution-level analysis that exploits the heterogeneity among institutional investors and examines their trading behavior in fraud _rms. For my firm-level analysis, I follow Bushee (2001) by grouping institutions into three categories based on their investment styles: quasi-indexer, transient, and dedicated. Diversified portfolios and low portfolio turnover characterize quasi-indexer institutions. Diversified portfolios and high portfolio turnover characterize transient institutions, and highly concentrated portfolios and low portfolio turnover characterize dedicated institutions. Consistent with prior literature, I expect to find that transient institutions are the most likely to initiate profitable trades in anticipation of a fraud revelation and quasi-indexer institutions are unlikely to initiate profitable trades in anticipation of a fraudrevelation (e.g., Ke and Ramalingegowda 2005; Hribar, Jenkins, and Wang 2006). I make no strong predictions for dedicated institutions as research typically finds that they do not trade based on impending future events. However, fraud is a unique setting that may lead dedicated institutions to divest their positions. If dedicated institutions invest in firms based on their confidence in the vision and integrity of management, detecting a fraud might lead dedicated institutions to divest their shares. In addition, since dedicated institutions are characterized by highly-concentrated portfolios, they are likely to have a larger percentage of their portfolios at risk if fraud is revealed. Therefore, they are likely to have the strongest incentives to anticipate fraud and divest their shares in fraud _rms. A body of literature suggests that institutional investors act as firm monitors (e.g.Chung, Firth, and Kim 2002; Chen, Harford, and Li 2007). If this is the case, then it is possible that fraud firms have low levels of institutional investment prior to committing fraud because they lack efficiency monitoring. Therefore, my first set of tests examines whether institutional ownership levels in fraud firms immediately prior to the release of a first fraudulent earnings report differ from a population of control firms. In unvaried analysis, I find that fraud firms actually have higher levels of total institutional ownership, quasi-indexer ownership, and transient institutional ownership than non-fraud firms. Dedicated institutional ownership is not significantly different from the non-fraud sample of firms. Next, I conduct regression analysis that controls for firm characteristics. I find that immediately prior to the beginning of a fraud, fraud firms' total level of institutional ownership is higher than institutional ownership for a sample of control firms. However, I find that the higher level of institutional ownership is primarily the result of a sign cant higher level of transient institutional ownership, while quasi-indexer and dedicated institutional ownership is nearly identical after controlling form characteristics. The university and regression results suggest that the level of institutional ownership does not act as a sapient monitoring device in the Prevention of fraud.My next sets of tests provide evidence relating to my primary research question. I first examine changes in institutional ownership levels in fraud firms over the period firms commit fraud. From the quarter prior to the issuance of a first fraudulent earnings report until the quarter prior to thepublic revelation of an accounting fraud, I find that institutional ownership in fraud firms increases by almost 14 percent, representing 3.9 percent of a fraud firm's outstanding stock.2 Because fraud firms experience stock price declines of approximately 35 percent once the fraud is revealed, the 3.9 percent increase in institutional ownership over the fraud period is not trivial. In fact, calculations suggest that total institutional losses for the 322 fraud firms in my sample are in order of $138 billion. The 3.9 percent increase in institutional ownership over the fraud period represents approximately $20 billion of those losses. Prior research has examined whether institutions can predict impending events over short windows (Hribar, Jenkins, and Wang, 2006). Accordingly, in my next set of tests, I observe changes in institutional ownership in the quarters immediately prior to and following the public revelation of fraud. I am that in the quarter immediately prior to a fraud revelation, institutional ownership decreases by approximately one and a half percent of a fraud firm's outstanding stock. I find significant decreases for transient institutional holdings, while changes in quasi-indexer and dedicated institutional holdings are insignificant. I also find significant decreases in the quarter immediately following the fraud revelation.These results are robust to several control variables including current and past stock returns, unexpected earnings, and changes in share turnover. Although I find come evidence that transient institutions are able to anticipate fraud one period prior to its revelation, this evidence must be interpreted in light of evidence from my previous tests. The one and a half percent decrease in institutional ownership prior to fraud revelations, though statistically significant, only slightly mitigates substantial losses for institutional investors.Institutions are more heterogeneous than the three categories I employ in my firm-level analysis. Therefore, I conduct a second analysis at the institution-level that further exploits the heterogeneity among institutional investors. I create proxies for institutions' information environments and institutions' incentives to avoid the negative market consequences associated with the revelation of accounting fraud. Conditional on owning fraud firm shares prior to a fraud beginning, I test whether the proxies are associated with institutions' ownership changes in fraudfirms prior to the revelation of an accounting fraud. The results provide some evidence that institutions with the strongest incentives to avoid accounting fraud and with the highest quality information environments divest shares in fraud firms prior to the revelation of accounting fraud. Although the data are consistent with an increased rate of divestitures among these institutions, I am unable to establish whether these relations are a result of informed trading or natural mean reversion in ownership levels.This study should be of interest to both institutional investors and researchers. The results suggest that institutional investors lose significant amounts of moneyby investing in firms that commit accounting fraud. My study contributes further evidence to the body of literature documenting the sophistication level of institutional investors. At least for this particular context, most institutions do not appear to be sophisticated users of accounting information; however, I do provide limited evidence of informed trading in the quarter immediately prior to fraud revelations among a subset of institutions.The remainder of my dissertation proceeds as follows. Chapter 2 examines prior literature and outlines my empirical predictions. Chapter 3 outlines my research design. Chapter 4 describes my sample selection process and provides descriptive statistics. Chapter 5 details my tests and presents results and chapter 6 concludes. Prior Literature and Empirical Predictions My dissertation builds on two streams of prior literature. The first stream of literature examines the determinants and consequences of accounting manipulations. The second stream of literature examines the trading behavior of institutional investors.Accounting Manipulations Prior research has identified characteristics of firms that manipulate their financial statements. Dechow, Ge, Larson, and Sloan (2007) investigate a comprehensive sample of all 895 firms subject to Accounting, Auditing, and Enforcement Releases (AAER) from 1982 through July 2005. They examine the use of several financial statement variables, o_-balance sheet and non-financial variables, and market-related variables to predict accounting manipulations. They had that firms accused by the SEC of manipulating their _financial performance tend to have had strong performance prior to manipulations. They also and thatmanipulations appear to be motivated by managers' desire to obfuscate deteriorating financial performance. During manipulation years, they find that cash profit margins and return on assets are declining while accruals are increasing. They also find that firms manipulating financial reporting are more likely to issue debt and equity.Ranking firms based on the predicted likelihood of accounting manipulations from a logistic model, they classify almost 50 percent of manipulation firms in the top 20 percent of their manipulation index and 65 percent of manipulation firms in the top 40 percent of their index. Beneish (1999) creates a fraud prediction model based on a sample of 74 firms that manipulated earnings and a sample of 2,332 matched firms. Estimating probity models of accounting manipulations as a function of eight accounting based variables (indexed day's sales in receivables, gross margin, asset quality, sales growth, depreciation, sales, general and administrative expenses, leverage, and accruals to total assets) he is able to correctly classify approximately 50 to 75 percent of fraud firms, while incorrectly classifying only 10 to 20 percent of matched firms. Several other studies document relations between earnings manipulation firms and firm characteristics. Two other notable studies include Dechow, Sloan, and Sweeney (1996) and Brazel, Jones, and Zimbelman (2006). Dechow et al. (1996) examine a sample of 92 firms with an AAER from 1982 to 1992. They document that AAERs are associated with external financing needs and poor corporate governance. They also show that AAER firms experience signifies cant increasesin their cost of capital after the revelations of accounting manipulations. Brazel etffal. (2006) also and that several non-financial measures can be useful in predicting accounting manipulations.Although the number of Type I errors in fraud prediction models is relatively high, the relative cost of Type I to Type II errors for institutional investors is likely extremely low. Several studies have estimated investment losses when accounting manipulations are revealed. The latest large sample evidence suggests that the cost of Type II errors average approximately 40 percent of an institution's investment in a fraud firm (Karpo, Lee, and Martin, 2007). On the other hand, the cost of a Type I error is extremely low in a market with many substitute assets as investors can simply choose not to hold firms with a high probability of fraud. Investors may alsobe privy to private information regarding firm performance and accounting manipulations. To the extent that investors possess private information and choose to use other qualitative information, they may be able to significantly reduce the number of Type I and II errors incurred when attempting to identify accounting frauds. The high number of Type I errors associated with using earnings manipulation prediction models might also suggest that investors would be willing to live with the negative returns associated with fraud _firms if the negative returns are balanced out with suficiently positive returns from non-fraud _firms with strong signals of fraud. In a concurrent working paper, Beneish and Nichols (2007), show that this is not the case. Their results reveal that _firms with a high probability of manipulated earn have lower future earnings and returns. They also show that a trading strategy based on the probability of earnings manipulation yields an abnormal hedge return of 13.9 percent.Through additional tests they conclude that the returns, which are concentrated on the short side, are not a result of asymmetric arbitrage costs, but rather a result of asymmetric errors in market expectations.Beneish and Nichols (2007) do not provide direct evidence on _firms that actually manipulate earn, rather they examine portfolios of firms with a high probability of manipulation. They find that institutional investors increase their holdings in firms with a high probability of manipulation. My study focuses on the actual incidence of fraud. I am able to provide more detailed and direct evidence on the trading behavior of institutions in actual fraud _firms before, during, and after the period in which firms commit fraud and the frauds become public. 2.2 Institutional Investors From 1980 to 1996, institutional investors doubled their share of the market and now control over half of the U.S. equity market (Gompers and Metrick, 2001). The increased importance and perceived sophistication of institutional investors has spawned a large body of literature.One branch of the literature examines whether institutional investors act as monitors and influence managements' decisions. The evidence suggests that the level of institutional ownership and the composition of a firm's institutional ownership base matters when determining whether institutional owners are likely to act as effective monitors. Bushee (1998) finds that managers are less likely to cut research and development expenses when facing an earnings shortfall ifinstitutional ownership is high. But he also finds that large proportions of ownership by institutions that trade based on momentum and have high portfolio turnover increase the likelihood that a firm will cut research and development to meet an earnings shortfall. Chung, Firth, and Kim (2002) find that large institutional shareholdings in a firm reduce the likelihood of earnings management using accruals. Chen, Harford, and Li (2007), using acquisition decisions to reveal monitoring, find that institutions with long-term investments specialize in monitoring while other institutions do not monitor. Bushee (2001) finds that high levels of short-term investors are associated with an over-weighting of near-term expected earnings and under-weighting oflong-term expected earnings. In light of this combined evidence, my first prediction is that fraud firms, prior to the issuance of their first fraudulent earnings report, are likely to have low levels of institutional ownership. I also expect that fraud firms will have higher levels of short-term, transient, institutional ownership and lower levels of long-term, dedicated, institutional ownership. Much of the accounting research on institutional investors' trading behavior suggests that institutional investors are sophisticated users of accounting information. For example, previous literature has documented that the higher the level of institutional ownership, the smaller the market reaction surrounding earnings announcements (El-Gazzar, 1998). Balsam, Bartov, and Marquardt (2002) find that the valuation implications of large discretionary accruals are incorporated into stock prices more quickly for _rms with large institutional investor bases. The presence of institutional investors is also positively associated with the extent that prices lead earnings (Jiambalvo, Raj Gopal, and Venkatachalam, 2002). Studies have also shown that institutional investors exploit accounting based stock price anomalies such as the post-earnings announcement drift (Ke and Ramalingegowda,2005) and the accruals anomaly (Collins, Gong, and Hribar, 2003). Lev and Nissim (2006) also show that the accruals anomaly is exploited by some institutional investors, but the magnitude of this accruals-related trading is rather small. They show that the continued persistence of the accruals anomaly is not explained by a lack of understanding among institutions, but rather an institutional distaste for extreme-accruals firms that are typically small, unprofitable, and risky. Ke and Ramalingegowda (2005) find thatinstitutions also possess information that allows them to avoid negative stock price shocks associated with a break in a string of consecutive earnings increases.Although much of the literature on institutional investors suggests that they are sophisticated users of financial information, this literature stands in contrast to evidence that questions whether institutions profit from informational advantages. For example, much of the literature on mutual fund performance suggests that superior performance is not persistent (e.g., Brown and Goetzmann 1995). Additionally, O'Brien and Bhushan (1990) _find that institutions are attracted to firms with more analyst following. Similarly, Bushee and Noe (2000) _find that institutions are attracted to firms with high-quality disclosure regimes. Therefore, if public and private information are substitutes, institutions should have fewer opportunities to benefit from informational advantages. If institutional investors possess superior private information or information processing abilities, I expect to find support for my second prediction that institutional investors divest shares in firms that are committing accounting fraud. A lack of evidence that institutions divest shares in fraud firms prior to public revelations of fraud would suggest that either investors are unable to use private information to anticipate public announcements of fraud or the cost of anticipating the public announcements of fraud are too great relative to the benefits. Institutional investors exhibit heterogeneity in their investment styles. Prior literature has shown that the likelihood of informed trading varies with institutional investors' characteristics (e.g., Hribar and Jenkins 2004; Ke and Ramalingegowda2005). Much of the prior literature has relied on a methodology proposed by Bushee (1998). In this methodology, institutions are first classified into one of three investment strategies (quasi-indexer, transient, and dedicated institutional investors) based on portfolio turnover and stake sizes. The institutions are then aggregated at the firm level. The body of evidence that uses this methodology typically finds that profitable trading in anticipation of future events is only identity able for the transient investor category. Therefore, I expect any evidence that institutional investors predict accounting fraud will be concentrated among transient institutional investors. Because dedicated institutional owners have the largest portion of their portfolios at stake when a fraud is revealed, I also anticipate the possibility thatthey may divest shares in anticipation of fraud revelations.In a recent paper, Bushee and Goodman (2007) exploit the heterogeneity among institutional investors and the positions they hold by conducting an analysis that includes not onlyinstitution-level variables such as portfolio size and trading strategy but also institution firm-level characteristics such as the size of a position in a particular firm and the size of the position in a firm relative to an institution's portfolio size. They find that private information trading are most pronounced when large positions are taken by investment advisers in small firms. In the spirit of Bushee and Goodman (2007), I conduct an institution-level analysis that exploits the heterogeneity among institutional investors that are not captured by the threemtypes of trading strategies employed in my firm-level analysis. Using proxies for the quality of an institution's information environment and the incentives an institution has to avoid fraud _firms, I expect to_find that institutions with the strongest incentives to avoid accounting fraud and higher-quality information environments are more likely to divest shares in fraud firms prior to the revelation of accounting fraud. I define and discuss in the institution-level analysis section the total institutional ownership representing only a 4.7 percent decrease in total institutional holdings. The decrease in transient institutional ownership in the quarter immediately prior to the fraud revelations represents only a 15.1 percent decrease in their institutional holdings. Thus, although institutions mitigate losses by divesting fraud firms prior to fraud revelations, overall they still lose a considerable amount of their investments. 5.2 Institution-level Analysis Institutions exhibit significant heterogeneity beyond the three investment styles I employ in my first analysis; therefore, I conduct a second analysis at the institution level. In this analysis, I test whether institution-level proxies for incentives to avoid accounting fraud and for private information are negatively associated with changes in institutions' ownership of fraud _rms. In this section, private information refers to both an institution's ability to gather private information and an institution's ability to process both private and public information. I employee two sets of variables in my tests. The first set is measured at the institution level and the second set is measured at theinstitution-firm level.。