한국 재무분석가 이익예측치의 속성 분석：정보성․합리성․대칭성

The purpose of this study is to empirically test three properties of financial analysts’ earnings forecasts, informativeness․rationality․symmetry. Prior studies suggest that Korean financial analysts’ earnings forecasts are more accurate than time-series models’ earnings forecasts. Also, they find that financial analysts’ earnings forecasts have optimistic bias. However, there have been few studies on the rationality and symmetry of financial analysts’ earnings forecasts. In order to empirically investigate the informativeness, rationality, and symmetry of financial analysts’ earnings forecasts, this study collects 14,428 weekly earnings forecasts that were released by 20 Korean Securities Companies from 2003 to 2005. The earnings forecasts is the average of financial analysts’ earnings forecasts for firm i in week t. The research method and empirical results are as follows. First, weekly market-adjusted stock returns are regressed on contemporaneous weekly earnings forecasts revisions after controlling for prior stock returns. The earnings forecast revisions are concentrated on 0. Especially, 48.5 percent of all revisions are 0. The median of revisions is 0, but the mean is -0.0024, implying that analysts’ earnings forecasts are likely to be revised downward during the sample period. The empirical results on the first regression model report that the coefficient of earnings forecast revision is statistically significant. That is, there appear to be informative of financial analysts’ earnings forecasts about contemporary stock prices. However, financial analysts’ earnings forecasts do not predict future stock price returns. Second, current earnings forecast revisions are regressed on prior stock returns in order to test whether all available information is reflected on contemporaneous earnings forecasts. If financial analysts’ earnings fore- casts reflect all value-relevant information, there is no significant association between current earnings forecast revisions and prior stock returns. However, the empirical results on the second regression model report that current earnings forecast revisions are significantly associated with prior stock returns, imply that all available information is not reflected on contemporaneous earnings forecasts. Third, current earnings forecast revisions are regressed on positive stock returns and negative stock returns, individually, in order to test whether financial analysts symmetrically react to good news and bad news. The empirical tests on positive stock returns show that the coefficient of stock returns in week t-1 is significantly positive. However, coefficients of stock returns in week t-2, t-3, and t-4 are not statistically significant. These results imply that positive information about firm value is reflected on financial analysts’ earnings forecasts after about one week. On the other hand, the empirical tests on negative stock returns show that coefficients of stock returns in week t-1 and t-2 are significantly positive. However, coefficients of stock returns in week t-3 and t-4 are not statistically significant. These results imply that negative information about firm value is reflected on financial analysts’ earnings forecasts after about three or four week. Therefore, financial analysts react to positive information differently from negative information due to financial analysts’ reward system. Fourth, current earnings forecast revisions are regressed on positive stock returns and negative stock returns for 40 big firms and 40 small firms, individually. The empirical results on positive stock returns for 40 big firms report that coefficients of all prior stock returns are not statistically significant, implying that positive information about firm value of 40 big firms is reflected on financial analysts’ earnings forecasts without delay. The empirical results on negative stock returns for 40 big firms report that the coefficient of stock returns in week t-1 is not statistically significant, but coefficients of stock returns in week t-1, t-2, and t-3 are significantly positive, implying that negative information about firm value of 40 big firms is reflected on financial analysts’ earnings forecasts after about two or four week. On the other hand, the empirical results on both positive and negative stock returns for 40 small firms show that coefficients of all prior stock returns are not statistically significant.