The impact of loss function assumptions in rational expectations tests on the efficiency of financial analysts' earnings forecasts
Su, Haochen (2008) The impact of loss function assumptions in rational expectations tests on the efficiency of financial analysts' earnings forecasts. [Dissertation (University of Nottingham only)] (Unpublished)
Prior studies document that financial analysts' earnings forecasts are inefficient with respect to various information variables. All of these studies conduct tests based on the ordinary least squares (OLS) regression, which implicitly assumes that analysts face a quadratic loss function. However, we argue that financial analysts are most likely to face a linear loss function. In this case, analysts try to minimize their mean absolute forecast errors (rather than their mean squared forecast errors). Therefore, we test the rational expectations hypothesis using the LAD (the linear loss function) as well as the OLS (the quadratic loss function) regressions, and incorporate most information variables documented in prior research. Consistent with prior studies, we find that analysts' forecasts are inefficient with OLS regressions. In contrast, we find little evidence of forecast inefficiency with LAD regressions. We conclude that the evidence of analysts' forecast inefficiency in prior research is largely attributed to the assumption of a quadratic loss function.
Actions (Archive Staff Only)