یک روش جدید برای پیش بینی خطاهای پیش بینی تحلیلگر: آیا سرمایه گذاران پیش بینی تحلیلگر را دچار خطا می کنند؟
|کد مقاله||سال انتشار||تعداد صفحات مقاله انگلیسی||ترجمه فارسی|
|13080||2013||26 صفحه PDF||سفارش دهید|
نسخه انگلیسی مقاله همین الان قابل دانلود است.
هزینه ترجمه مقاله بر اساس تعداد کلمات مقاله انگلیسی محاسبه می شود.
این مقاله تقریباً شامل 20760 کلمه می باشد.
هزینه ترجمه مقاله توسط مترجمان با تجربه، طبق جدول زیر محاسبه می شود:
- تولید محتوا با مقالات ISI برای سایت یا وبلاگ شما
- تولید محتوا با مقالات ISI برای کتاب شما
- تولید محتوا با مقالات ISI برای نشریه یا رسانه شما
پیشنهاد می کنیم کیفیت محتوای سایت خود را با استفاده از منابع علمی، افزایش دهید.
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Financial Economics, Volume 108, Issue 3, June 2013, Pages 615–640
I provide evidence that investors overweight analyst forecasts by demonstrating that prices do not fully reflect predictable components of analyst errors, which conflicts with conclusions in prior research. I highlight estimation bias in traditional approaches and develop a new approach that reduces this bias. I estimate characteristic forecasts that map current firm characteristics into forecasts of future earnings. Contrasting characteristic and analyst forecasts predicts analyst forecast errors and revisions. I find abnormal returns to strategies that sort firms by predicted forecast errors, consistent with investors overweighting analyst forecasts and predictable biases in analyst forecasts influencing the information content of prices.
Estimating a firm's future profitability is an essential part of valuation analysis. Analysts can facilitate the valuation process by translating a mixture of public and private information into forecasts of future earnings. However, a substantial literature spanning finance, economics, and accounting raises concerns about the use of these forecasts for investment decisions, commonly citing a significant incentive misalignment between analysts and those of the end users of the earnings forecasts.1 The collective evidence from this literature suggests that reliance on analyst forecasts can produce biased estimates of firm value. Recognition of this problem has motivated researchers to develop techniques to identify the predictable component of analyst forecast errors. The development of these techniques also reflects a desire to better understand what information is reflected in price. To the extent that investors overweight analyst forecasts, a firm's share price is unlikely to fully reflect the earnings news associated with predictable analyst forecast errors.2 Thus, if overweighting is systematic, the identification of predictable forecast errors is potentially useful in disciplining prices. The goal of this paper is to determine whether and to what extent investors systematically overweight analysts' earnings forecasts. Motivated by a similar goal, Hughes, Liu, and Su (2008) conclude that investors do not overweight analyst forecasts. They find that a strategy of sorting firms by predicted forecast errors fails to generate abnormal returns and attribute this finding to market efficiency with respect to the predictable component of analyst errors. I argue that their findings are unlikely to result from market efficiency and are instead a reflection of their methodology. The traditional approach to predicting forecast errors, used by Hughes, Liu, and Su (2008) among others, involves regressing realized forecast errors on lagged, publicly observable firm characteristics. The resulting estimated coefficients are applied to current characteristics to create a fitted prediction of future forecast errors. I show that the traditional approach can introduce biases into predicted forecast errors. Biases emerge whenever the observable firm characteristics used to predict forecast errors are correlated with unobservable inputs to analyst forecasts such as analysts' incentive misalignment or private information. Predicted forecast errors can be consistently above or below the realized forecast error depending on the sign and magnitude of these correlations. Moreover, biases in predicted forecast errors can vary across firms, limiting their ability to meaningfully sort stocks in the cross section. Because tests of overweighting rely on sorting firms by predicted errors, assessing whether investors overweight analyst forecasts is difficult without first making progress on a methodological front. In this paper, I develop and implement a new approach to predicting analyst forecast errors that circumvents many of the problems hampering the traditional approach. This new approach also involves the use of historically estimated relations but shifts the focus toward the prediction of future earnings and away from regression-based fitting of past forecast errors. I show that this approach is less sensitive to estimation bias and offers significant predictive power for realized forecast errors and future returns. The methodology highlighted in this paper is referred to as the characteristic approach to predicting analyst forecast errors. This term reflects the fact that I contrast analysts' earnings forecasts with characteristic forecasts of earnings, in which both forecasts are measured several months before firms' annual earnings announcements. I construct characteristic forecasts by fitting current earnings to the firm characteristics used by Fama and French (2000) in the prediction of future profitability: lagged earnings, book values, accruals, asset growth, dividends, and price. I estimate pooled cross-sectional regressions to capture large sample relations between earnings and lagged firm characteristics. I apply historically estimated coefficients to firms' most recent characteristics to create ex ante forecasts of future earnings. I first show that characteristic forecasts are an unbiased predictor of realized earnings and contrast these forecasts with those issued by sell-side analysts. When contrasting characteristic and analyst forecasts, several interesting patterns emerge. First, firms with characteristic forecasts exceeding consensus analyst forecasts tend to have realized earnings that exceed the consensus, and vice versa. Second, when discrepancies exist, analysts subsequently revise their forecasts in the direction of characteristic forecasts leading up to earnings announcements. Third, analysts are more likely to raise investment recommendations for a given firm when characteristic forecasts exceed the consensus analyst forecast, and vice versa. These results suggest that analysts are slow to incorporate the information embedded in characteristic forecasts when forecasting future firm performance and that overreliance on analyst forecasts could result in substantial valuation errors. Given the potential for valuation errors when relying on analyst forecasts, I conduct a series of tests to examine whether investors overweight analyst forecasts. Using a simple two-period framework, I establish how researchers can precisely test for efficient market weights by relating future returns with differences between characteristic and analyst forecasts. To implement this test, I develop a new metric that I refer to as characteristic forecast optimism, defined as the ex ante characteristic forecast minus the prevailing consensus forecast, in which higher values correspond to firms whose characteristics signal future earnings that exceed analyst projections. I find consistent abnormal returns to a strategy that buys firms in the highest quintile of characteristic forecast optimism and sells firms in the lowest quintile, consistent with investors systematically overweighting analyst forecasts and underweighting characteristic forecasts. This simple, unconditional quintile strategy generates average returns of 5.8% per year in out-of-sample tests. Strategy returns significantly increase through contextual analysis and display a number of intuitive relations with firm characteristics. In conditional tests, returns increase to 9.4% per year among firms whose stock price is highly sensitive to earnings news. Similarly, characteristic forecast optimism is a stronger predictor of returns among small firms, firms with historically disappointing earnings, and firms with low financial transparency. These results are consistent with investors being more likely to overweight analyst forecasts among firms with poor information environments or when investors are uncertain about the mapping between current and future earnings. An alternative explanation for these findings is that return predictability manifests in response to priced risk correlated with characteristic forecast optimism. To mitigate risk-based explanations, I demonstrate that return predictability is robust to Fama-French risk-adjustments and standard risk controls in cross-sectional tests. The ability of characteristic forecast optimism to predict returns is distinct from post-earnings announcement drift, momentum, the accrual anomaly, relative value strategies, and investors' reliance on analysts' long-term growth forecasts. I also find that characteristic forecast optimism predicts subsequent earnings announcement returns, consistent with forecast discrepancies signaling earnings information that is not reflected in prices in a timely fashion. Taken together, the magnitude and consistency of return prediction is striking in light of prior research concluding that investors efficiently weight analyst forecasts. The central implication of these findings is that investors fail to fully undo predictable biases in analyst forecasts and, as a result, distortions in analyst forecasts can influence the information content of prices. These findings suggest that regulators should be concerned not only with how distortions in analyst forecasts differentially impact the welfare of subsets of investors (e.g., retail versus institutional) but also how they impact the efficient allocation of capital. Two additional tests compare the characteristic approach to the traditional regression-based fittings of past forecast errors. First, I fit past forecast errors to the same firm characteristics used when constructing characteristic forecasts and demonstrate that the characteristic approach significantly outperforms the traditional approach in predicting analyst forecast errors, forecast revisions, and future returns. Second, I compare the predictive power of characteristic forecast optimism with two existing forecast error prediction models and again find evidence favoring the use of the characteristic approach. The rest of the paper is organized as follows. Section 2 provides motivation for predicting analyst forecast errors and discusses traditional approaches. Section 3 discusses the related literature. 4 and 5 discuss the empirical tests, findings, and robustness checks. Section 6 concludes.
نتیجه گیری انگلیسی
This paper provides evidence that investors systematically overweight analyst forecasts by demonstrating that prices do not fully reflect the predictable component of analyst forecast errors in a timely fashion. The central implication of these findings is that investors fail to fully undo predictable biases in analyst forecasts and, as a result, distortions in analyst forecasts can influence the information content of prices. Evidence that investors overweight analyst forecasts conflicts with conclusions in prior research relying on traditional approaches to predicting analyst forecast errors. Traditional approaches are subject to correlated omitted variable bias whenever the variables used to predict forecast errors are correlated with unobservable inputs to analyst forecasts. I develop and implement a new approach that mitigates this bias by contrasting characteristic forecasts of earnings with those issued by analysts. I estimate characteristic forecasts using large sample relations to map current firm characteristics into forecasts of future earnings and demonstrate that evaluating analyst forecasts relative to characteristic forecasts offers significant predictive power for analyst errors and future returns. I find that firms with characteristic forecasts exceeding the consensus analyst forecast tend to have realized earnings that exceed the consensus, and vice versa. Similarly, analysts subsequently revise their earnings forecasts and investment recommendations in the direction of characteristic forecasts leading up to earnings announcements. This evidence suggest that analysts are slow to incorporate the information embedded in characteristic forecasts and that overreliance on analyst forecasts likely results in valuation errors. I find that stock prices behave as if investors overweight analyst forecasts and underweight characteristic forecasts relative to the optimal Bayesian weights. Specifically, I find consistent abnormal returns to a strategy that buys firms with characteristic forecasts above analyst forecasts and sells firms with characteristic forecasts below analyst forecasts. Strategy returns significantly increase through contextual analysis and display a number of intuitive relations with firm characteristics. For example, returns are increasing in the sensitivity of firms' stock price to earnings news and the uncertainty between current and future earnings. The magnitude and consistency of return prediction is striking in light of prior research concluding that investors efficiently weight analyst forecasts. Taken together, the findings of this paper have implications for practitioners, regulators, and researchers. First, for practitioners, the findings support using characteristic forecasts as a means of evaluating analysts and identifying potential mispricing. Similarly, characteristic and analyst forecasts offer incremental predictive power for future earnings, which supports the use of both forecasts when valuing firms. Second, the evidence that investors systematically overweight analyst forecasts suggests that market regulators motivated by the efficient allocation of capital should pursue measures to improve analyst forecasts, such as the development of additional mechanisms reducing incentive misalignment between analysts and investors. Finally, for researchers, I propose a simple test for the efficient weighting of multiple earnings forecasts by relating forecast differences with future returns. Understanding how investors weight these forecasts can yield superior measures of the market's expectations of earnings and, thus, potentially improve estimates of earnings surprises and implied cost of capital that require these expectations as inputs.