Use this URL to cite or link to this record in EThOS:
Title: The structure of cross-sectional dependence in analysts forecasts of earnings per share: evidence and implications
Author: El-Galfy, Ahmed Mohamed Mohamed
ISNI:       0000 0001 3442 7529
Awarding Body: University of Manchester : University of Manchester
Current Institution: University of Manchester
Date of Award: 2003
Availability of Full Text:
Access from EThOS:
A number of papers have provided empirical evidence suggesting that analysts' earnings forecasts do not conform to Muthian rationality. Analysts' earnings forecasts have been found, in both early research and more recent studies, to be irrational, i.e., biased, inefficient or both (see for instance: Brown et al., 1985 and Easterwood and Nutt, 1999). De Bondt and Thaler (1990), in particular, documented generalised overreaction in the analysts' forecasts, showing that the forecasts, tracked by I/B/E/S, are too optimistic and too extreme to be considered rational. Conversely, a notable study developed by Keane and Runkle (1998) claimed that analysts' forecasts of corporate profits may be rational after all if we take into account two complications: (1) the cross-sectional correlation in contemporaneous forecast errors across analysts and firms, and (2) discretionary asset write-downs, which affect earnings but are intentionally ignored by analysts when they make earnings forecasts. They proposed an estimation technique for testing the rationality of analysts' forecasts based on the GMM of Hansen (1982). This thesis challenges the finding that cross-sectional correlation "explains" the irrationality of analysts' earnings forecasts. The thesis further investigates Keane and Runkle's (1998) claim. The rationality of analysts' earnings forecasts per share is tested in the current study using two data-set samples and employing two complementary empirical methods. Firstly, the quality of "annual" analysts' earnings forecasts for a "balanced panel" of US firms is investigated. In particular, I test for forecast accuracy by using two alternative forecast errors metrics, that are: the "mean absolute percent forecast errors" and the "mean percent forecast errors". Then, I evaluate the superiority of analysts' earnings forecasts by comparing the forecasts of my balanced panel of firms with those generated by a variety of alternative benchmark models, using Theil's if metric which evaluates the adequacy of analysts' forecasts based on the mean-square error of their predictions. Next, the unbiasedness and efficiency are tested by using OLS regression of the actual change of EPS on the forecasted change of EPS. The study is extended in this part to investigate the degree to which the analysts vary in their predictive performance and the impact of some inappropriate forecasts on the adequacy of analysts' forecasts as a whole. Secondly, GMM correction to the variance-covariance matrix is employed to test the rationality of "quarterly" analysts' earnings forecasts per share, allowing for cross-sectional dependence, as well as controlling for the extreme special items. I use in this part the framework advanced by Keane and Runkle (1998), and also introduce some extensions to their framework to try to improve the understanding of this issue.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available