Abstract ABSTRACT: This paper analyzes the manner in which financial analysts' forecasts of annual earnings are formed. A sample of over 6,000 earnings forecasts made over a period of 11 years is examined. The results from the time-series tests show that analysts' forecasts of annual earnings are rational in the sense that they fully utilize the information contained in the past history of earnings and their own forecasts. The findings also suggest that the adaptive expectations model adequately describes the process of the formation of earnings forecasts. The adaptation coefficient, however, is not constant across companies and over time. The results imply that the use of mean forecasts and cross-sectional (rather than time-series) tests in previous studies has reduced the tests' power and efficiency, but still led to valid conclusions.
Dan Givoly (Mon,) studied this question.