| Summary: | Prior studies using ordinary least squares (OLS) regression find that financial analysts do not efficiently utilize public available information when making forecasts. The OLS-based tests implicitly assume that analysts face a quadratic loss function which aim to minimize their mean squared forecast errors. However, Gu and Wu (2003), and Basu and Markov (2004) argue that analysts face a linear loss function and thus minimize their mean absolute forecast errors. Therefore, rationality test should be examined by least absolute deviation (LAD) regression. As suggested by Basu and Markov (2004), we examine the rationality of analysts' forecasts with different information variables by using OLS and LAD regressions. Consistent with the prior studies, the OLS results show that analysts do not efficiently use the information when making forecasts under a quadratic loss function, but we find very little evidence of forecasts' inefficiency under a linear loss function.
|
|---|