본 연구는 2001년부터 2008년까지의 표본을 대상으로 재무분석가가 자신의 고유한 인적 특성으로 인해 분석 대상 기업과 무관하게 이익예측에 있어서 지속적이고 체계적인 예측편의(Foreccast bias)를 나타내게 되는지를 살펴보았다. 이와 더불어 체계적 예측편의를 나타내는 재무분석가의 예측정확성과 체계적 예측편의를 나타내지 않는 재무분석가의 예측정확성을 비교하였다. 체계적 예측편의를 나타내는 재무분석가의 예측정확성은 그렇지 않은 재무분석가의 예측정확성보다 낮을 것으로 예상된다. 왜냐하면 높은 수준의 전문적 지식을 바탕으로 정교한 판단을 내리는 재무분석가라면 분석대상 기업별로 과거 영업실적, 특정산업에서 해당 기업이 차지하는 위치 및 전반적인 경기 동향 등을 바탕으로 예측활동을 하게 될 것이고, 결과적으로 나타나는 예측편의는 양(+) 또는 음(-)의 값을 나타낼 것이지 분석대상 기업과 무관하게 체계적으로 양(+)의 값이나, 음(-)의 값을 나타내지는 않을 것이기 때문이다. 실증분석 결과에 의하면 일부 재무분석가의 경우 분석대상 기업과 무관하게 체계적으로 낙관적 혹은 비관적 예측편의를 나타내고 있었다. 그리고 체계적 예측편의를 나타내는 재무분석가의 예측정확성이 그렇지 않은 재무분석가의 예측정확성보다 낮은 것으로 나타났다. 또한, 체계적으로 낙관적 예측편의를 나타내는 재무분석가의 예측정확성이 가장 낮은 것으로 나타났는데, 이는 낮은 분석능력이 재무분석가가 낙관적 예측편의를 나타내는 데에 대한 원인이 될 수 있음을 보인 Beyer (2008)의 이론적 분석결과와 일치한다. 대부분의 실증연구는 재무분석가가 예측에 있어 편의를 나타내게 되는 이유를 소속증권사 및 재무분석가의 재무적 유인에서 찾으려 하였다. 본 연구는 이와 같은 외적 요인 뿐 만 아니라 재무분석가의 인적 특성 즉 내적 요인도 재무분석가가 예측 편의를 나타내는 데에 대한 원인이 될 수 있음을 보였다는 데에 공헌 점이 있다. 또한, 재무분석가 더미변수를 이용하여 기업 수준이 아닌 개별 재무분석가 수준에서 예측편의를 측정할 수 있음을 보였다는 점에서도 공헌 점이 있다.
Are analysts perfect? Analysts are known to have behavior bias, even though they have considerable expertise in forecast activity(Stotz and Nitzsch 2005; Hilary and Menzly 2006). Then, it is conceivable that some portion of analysts`` forecast bias is attributable to imperfection of analysts. However, prior literature focuses on monetary incentives such as incentives to generate trading commissions to explain analysts`` forecast bias. This study attempts to address this gap. Specifically, this study examines whether analyst human characteristics affect forecast bias. Then, I investigate whether an analyst that shows a tendency toward systematic bias in forecast issues less accurate forecast. Prior literature finds that analysts`` forecasts are biased and the bias is related to analysts`` incentives to appease managers to obtain investment banking business, to maintain access to firm managers who are a primary source of information flow, to generate trading commissions and so on (Lin and McNichols 1998; Michaely and Womack 1999; Dechow, Hutton, and Sloan 2000; Lin, McNichols, and O`Brien 2005; Cowen et al. 2006; Jacob et al. 2008; Lee et al. 2005). The common characteristic of these literature is that they focus on monetary incentives that lead analysts to show forecast bias. However, I expect that not only monetary incentives but also imperfection of analysts could be the source of forecast bias. Beyer (2008) theoretically demonstrates that analysts who have little information or low ability in forecasts tend to forecast overoptimistically. This is because analysts with low-ability can make use of the tendency of managers for meeting or beating analysts`` forecasts. When analysts forecast earnings optimistically, managers endeavor to increase earnings to meet or beat analysts`` forecasts and this managerial behavior results in the decrease of the difference between analyst forecasts and reported earnings. Analysts that recognize managers`` tendency for meeting or beating analysts`` forecasts and have little sources for accurate forecasts find it optimal to forecast optimistically rather than pessimistically. This implies that analysts`` low ability could be a source of forecast bias. Behavior economics finds that man has behavior bias and sometimes shows irrationality. They argue that man`s irrationality could explain considerable portion of man`s behavior and this individual irrationality persists. I argue that analyst as an imperfect man also can show forecast bias persistently. In total, analysts`` low-ability and irrationality can be sources of analysts`` forecast bias. This leads to my assertion that analyst human characteristics affect forecast bias. To test my conjecture, I estimate forecast-bias specific analyst fixed effects by regressing forecast bias on analyst indicator variables with other controls (Ge et al. 2009; DeJong and Ling 2009; Demerjian et al. 2010; Dyreng et al. 2010). This regression model is designed to identify each analyst`s heterogeneity in forecast bias, controlling for firm`s time-varing and time-invariant characteristics and year effects. When forecast bias (dependent variable) is defined as reported earnings minus analyst forecst, the analyst who has significant negative (porsitive) coefficient on his indicator variable could be considered to issue optimistic (pessimistic) forecasts persistently. I call these analysts`` behaviors as "systematic forecast bias", because they show forecast bias in the same direction irrespective of firms that they analyze. After estimating the coefficients of analyst indicator variables, I do F-test to jointly test the significance of the coefficients. If the coefficients of analyst indicator variables are significantly different from zero, this implies that analyst human characteristics affect forecast bias. And if analysts`` low-ability and irrationality are sources of analysts`` forecast bias, I can argue that analysts who show systematic forecast bias have low accuracy in forecast. This leads to my hypothesis. To test my hypothesis, I classify analysts into two goups depending on whether they have systematic forecast bias or not. If an analyst has significant coefficient on his/her indicator variable, he/she is classified into systematic bias group. If systematic bias group has less accurate forecast accuracy, my hypothesis is supported. The results show that analysts`` indicator variables are jointly significant in the regression model, which implies that analyst human characteristics affect forecast bias. And I find that the group with systematic bias has less accurate forecast accuracy than the group without systematic bias. Specifically, evidence shows that the analysts who show optimistic bias systematically have the least accurate forecast accuracy. This paper contributes to the literature by extending the prior research on the relation between analysts and forecast bias by providing evidence that analysts`` human characteristics can affect forecast bias. Prior literature on the relation between analysts and forecast bias focuses mainly on analysts`` monetary incentives, In contrast, This paper focuses on analysts`` individual characteristics and provides evidence that not only monetary incentives but also analysts`` human characteristics affect forecast bias, This paper also contributes to the literature by introducing new methodology for estimating forecast bias attributable to analysts controlling for firm characteristics. When forecast bias is defined as reported earnings minus forecast earnings, the forecast bias is influenced by both firm and analyst. So, the forecast bias is not solely attributable to analyst. However, the coefficient on an analyst indicator variable in this paper`s regression model is solely attributable to the analyst because it is estimated controlling for firm`s time-varing and time-invariant characteristics.