盈余质量的定量分析[外文翻译].doc

1、外文文献翻译译文一、外文原文原文：Quantitative analysis of earning qualityEarnings quality analysis has long been known as the qualitative investment managers best defense against low quality financial reporting. While not widely known, recent academic and proprietary research shows that earnings quality metrics are

2、 also useful in generating alpha in the context of quantitative investment and trading strategies.The Gradient Analytics Earnings Quality Model (EQM) is the first quantitative factor that objectively measures earnings quality across a broad spectrum of companies for both longer-term (3-12 month) and

3、 shorter-term (1-3 month) holding periods. The model was developed using rigorous statistical methods to ensure a robust factor that generates excess returns on a standalone basis while also capturing a unique dimension of returns not captured by other quantitative factors. The end product is a high

4、ly unique factor with exceptional returns and low correlation in relation to other commonly used factors such as those derived from estimate revisions, earnings momentum, price momentum, cash flow, corporate insider, growth and value.The U.S. accounting profession and the Securities Exchange Commiss

5、ion (SEC) have worked diligently through the years to develop the most rigorous system of accounting procedures in the world. Nevertheless, there is still a significant gap (appropriately called the expectations gap) between what investors and creditors expect and what the accounting profession can

6、deliver. The expectations gap exists in part because publicly traded companies have a great deal of discretion in choosing accounting principles and in making estimates that impact their reported financial results. Under Generally Accepted Accounting Principles (GAAP), the amount of discretion that

7、a company has in preparing financial statements is controlled by two fundamental principles of accounting: conservatism and objectivity. However, in reality these two guiding principles are often stretched to the limit or ignored.When Conservatism or Objectivity is Impaired, Earnings Quality is Comp

8、romisedWhile, in theory, a firms accounting staff should employ procedures that are objective and conservative, in practice, management has many competing motivations that drive their choice of accounting policies and influence their periodic estimates. Because of these competing motivations, many c

9、ompanies manipulate accounting numbers in order to facilitate the financial reporting goals established by management. In this regard, virtually all firms working within the bounds of GAAP use minor accounting “gimmicks” to present financial results in a particular light (i.e., overstate or understa

10、te their true profitability or financial condition). Micro Strategy, Inc., for example, (1999-2000) was using an aggressive revenue recognition policy that, while not in violation of GAAP, tended to overstate the companys true profitability. However, it wasnt until the SEC mandated a change in the w

11、ay that technology companies account for contract revenue that the market ascertained the extent of the overstatement (although access to earnings quality analysis qualitative or quantitative would have revealed the deception prior to the change in accounting rules). After the SEC mandated change, M

12、icro Strategy was twice forced to restate its earnings and its shares fell over 98% in the ensuing 12-month period.While Micro Strategys treatment of contract revenues was very aggressive, arguably it was operating in a gray area of GAAP. In more extreme cases, however, some companies go so far as t

13、o commit fraudulent acts that materially misstate their financial statements in ways that do not conform to GAAP. For example, Enron used an extremely aggressive scheme of off-balance-sheet financing in order to hide mounting losses. The end result was arguably one of the most spectacular financial

14、reporting disasters in history. Those unlucky enough to hold Enron shares during this period lost close to 100% of the value of their investment.Finally, while intentional manipulation of accounting numbers is common, earnings quality problems are not always the result of intentional acts by managem

15、ent. For example, the quality of inventories at Lucent Technologies (as reported in their first quarter 10Q filed May 2000) suggested an apparent backlog of inventory that indicated a possible slowdown on the horizon. In all likelihood, this was (at least initially) a case of earnings quality proble

16、ms resulting from unintentional acts (slow sales). Nevertheless, Lucents earnings continued to disappoint and the stock was down more than 85% in the ensuing twelve months. (Subsequent evidence suggests that there may also have been some intentional misstatements on the part of lucent management in

17、order to hide the magnitude of the sales slowdown.)How do companies manipulate earnings?Despite the efforts of the accounting profession to ensure objectivity and conservatism, it is still relatively easy to manipulate accounting numbers through either unethical (but not necessarily illegal) and/or

18、fraudulent means. The list presented below provides a high level overview of how Management can manipulate accounting numbers.1. Recording fictitious transactions or amounts2. Recording transactions incorrectly3. Recording transactions early4. Recording transactions late5. Misstating percentages or

19、amounts involved in a transaction6. Misstating the amounts of assets or liabilities7. Changing accounting methods or estimates for no substantive reason8. Using related party transactions to alter reported profitsAcademic Research on Earnings Quality and Future ReturnsIn addition to the anecdotal ev

20、idence provided by qualitative earnings quality services (i.e., those that use subjective evaluations of financial statements to render an earnings quality grade), academic research also supports the notion that quantitative models of earnings quality can be used to earn excess returns. The followin

21、g brief review of the academic literature highlights some of the most important factors that form the basis for Gradients approach to quantitatively modeling earnings quality and forecasting related excess returns.The very first studies to investigate issues related to earnings quality were conducte

22、d by G. Peter Wilson of Harvard University (1986, 1987) using an event study methodology. Wilsons key conclusions are that operating cash flows and total accruals (i.e., changes in current accruals plus non current accruals) are differentially valued and that both are value relevant. That is, the ma

23、rket appears to react to the disclosure of detailed cash flow and accrual data (value relevance) and that cash flows are more highly valued than accruals (differential valuation). Wilsons basic findings are also supported by a number of studies that use an association methodology7, including Rayburn

24、 (1986), Bowen, Burgstahler and Daley (1987), Chariton and Katz (1990), Levant and Zeroing (1990), Vickers (1993), Ali (1994), Pfeiffer et al. (1998), and Vickers, Vickers and Betties (2000).The fact that the market values a dollar of cash flow more than a dollar of current or non current accruals i

25、mplies that higher levels of accruals are indicative of lower quality of earnings. In other words, the degree to which a company must rely on accruals to boost net income results in lower quality earnings. Nevertheless, it is possible that the market “sees through” the deception and appropriately va

26、lues companies based on some notion of baseline or sustainable earnings. However, the first studies to investigate this issue (Sloan, 1996 and Swanson and Vickers, 1997) find that, contrary to the efficient markets hypothesis, disaggregating earnings into cash flow and accrual components is useful i

27、n identifying securities that are likely to outperform (or under- perform) in the future. Thus, the results of these studies imply that security prices do not fully reflect the information contained in the cash flow and accrual components of earnings.Following in the path of Sloan (1996) and Swanson

28、 and Vickers (1997), academic researchers are currently focusing on the development of simple empirical models that objectively assess earnings quality in order to predict future return performance. (See, for example, Sloan, Solomon and Tuna, 2001; Chan, Chan, Legatees and Lakonishok, 2001; and Penm

29、an and Zang, 2001.) Table 1 below summarizes the results of recent academic working papers that focus on the predictive ability of simple earnings quality models. As shown in the table, these studies find that companies with relatively high (low) levels of accruals tend to under-perform (outperform)

30、 for periods of 12-36 months after the disclosure of detailed financial data. Specifically, the return spread between stocks with the highest level of accruals (lowest earnings quality) and the lowest level of accruals (highest earnings quality) ranges from 8.8% to 21.7%, depending on the approach u

31、sed by the authors in forming portfolios. The implication is that measures of earnings quality can be used in forming profitable investing and trading strategies.Gradient Analytics Earnings Quality ModelThe latest academic research demonstrates that the market does not fully impound information abou

32、t earnings quality at the time that detailed financial statement data are released. That is, a statistically-based approach to analyzing earnings quality can yield profitable investment and trading strategies. Thus, the next step was to develop a robust model that is designed to optimize the excess

33、returns that can be realized from an earnings quality strategy. The Gradient Analytics Earnings Quality Model (EQM) has been developed to achieve this goal.The Gradient EQM is the first quantitative factor that measures earnings quality across a broad spectrum of companies. The EQM provides two week

34、ly scores ranging from 1 (strong sell) to 8 (strong buy) for each of the top 5000 companies ranked by market capitalization. The Long-Term Score provides a 1-8 ranking based on a stocks expected future performance over a 3-12 month holding period. The Short-Term Score ranks each stock based on its e

35、xpected future performance over a 1-3 month holding period.In contrast to competing, commercially-available earnings quality services, the output from the EQM is derived objectively not subjectively through statistical analysis of accrual and cash flow components of earnings. The model was developed

36、 using rigorous statistical methods to ensure a robust factor that generates excess returns on a standalone basis while also capturing a unique dimension of returns not captured by other quantitative factors. More specifically, the model was constructed using a multiple regression approach (includin

37、g repressors from academic research and our own theoretically sound proprietary earnings quality constructs) estimated in pooled time series, cross section for 13 sector categories.8 Each separate sector model incorporates the most important dimensions of earnings quality for that segment of the mar

38、ket.9 When considered together, these dimensions or “sub-factors” provide a means of reliably ranking firms monotonically according to both their expected mean and median excess returns. The end product is a highly unique factor with exceptional returns and low correlation in relation to other commo

39、nly used factors.All Gradient models are developed using a disciplined scientific approach. Our approach can be characterized as follows:Variable Specification We begin by carefully specifying each variable to ensure proper measurement and scaling. When more than one specification is defensible, we

40、choose the simplest specification on the theory that simplicity will yield more generalizable results.Modeling Techniques Each model is estimated using relatively simple linear and nonlinear regression techniques. Again, we believe that simplicity is the key to generalizability.Sensitivity Analyses

41、All models are subjected to sensitivity analyses to determine whether or not our results are impacted by outliers, changes in regimes, alternative variable specifications and modeling techniques, and so on.Proper Use of In-Sample and Out-of-Sample Periods Each model is estimated using data from a st

42、rict in-sample period. The model is then tested (for generalizabity, stability, and so on) in an out-of-sample period.Control for Potential Threats to Internal and External Validity Our research efforts are designs to control for common threats to internal and external validity in financial engineer

43、ing studies (such as survivorship bias, hindsight bias, selection bias, and so on).ConclusionEarnings quality analysis is often regarded as the qualitative investment managers best defense against low quality financial reporting. The latest academic research also demonstrates that the market does no

44、t fully impound information about earnings quality at the time that detailed financial statement data are released. That is, a statistically-based approach to analyzing earnings quality can yield profitable investment and trading strategies. The Gradient Earnings Quality Model (EQM) has been develop

45、ed to achieve this goal.The EQM is the first quantitative factor that objectively measures earnings quality for the purpose of forecasting future returns. The model was developed using rigorous statistical methods to ensure a robust factor that generates excess returns on a standalone basis while al

46、so capturing a unique dimension of returns not captured by other quantitative factors. The end product is a highly unique factor with exceptional returns and low correlation with other commonly used factors.The Earnings Quality Model has been extensively back-tested across a variety of stock univers

47、es and time periods in order to ensure optimal, generalizable results. The results presented in this white paper provide extremely strong evidence on the usefulness of the EQM. As shown in the results section of this document, the model produces highly significant excess returns, performs extremely

48、well both in- and out-of-sample, and has a low correlation with other commonly used quantitative factors. And, in contrast to competing, commercially-available earnings quality services, the output from the EQM is derived objectively not subjectively through statistical analysis of accrual and cash

49、flow components of earnings.Source: gradient analytics, 2005. “Quantitative analysis of earning quality”.EB/OL website. March, pp.23-25.二、翻译文章译文：盈余质量的定量分析盈余质量分析长时间被认为是定性投资经理针对低质量财务报告的最佳防卫。虽然不广为人知，但近期的学术和专有的研究表明，盈余质量指标也对量化投资和贸易战略中用到的阿尔法的产生有帮助。盈余质量的梯度分析模型是第一个量化的因素，客观地计量在广泛的公司为期（3-12月）盈利的质量和短期（13月）期间举行。该模型是开发、利用严谨的统计方法，以确保一个强大的因素，在一个独立的基础上产生的超额利益，同时也捕获不受其他因素的捕获量的回报的独特维度。最终产品是一种特殊的回报，是相对于其他常用的估计，如