Security Analysts’ Earnings Forecasts: Distributions Normality and a Comparative Analysis of Fitted Distribution Types in the Development of a Surrogate Consensus

The employment of IBES (Institutional Brokers’ Estimate System) analysts’ earnings forecast consensus in capital markets research literature presupposes normality in per period, per firm distributions of analysts’ earnings estimates on the basis that: (i) the central limit theorem holds true because analysts’ earnings estimates are independently and identically distributed (iid); and (ii) the unweighted consensus mean and/or median are the best estimator(s) of a normal distribution. However, monthly distributions of IBES analysts’ earnings forecasts for all Australian stocks from 11 months through to actual reported earnings between 1988 and 2002 were found to be significantly non-normal, with this result common across four different deflator types; (i) firm’s share price at 11 months prior to actual reported earnings; (ii) firm’s share price at each period; (iii) actual reported earnings; and (iv) average of forecast plus actual. Furthermore, respective distribution skewness and kurtosis were significantly positive in corroboration with evidence of distribution non-normality. These findings are consistent with the principal hypothesis of this thesis, which propounds the contrarian notion that distributions of IBES analysts’ earnings estimates are non-normal because IBES analysts’ earnings forecasts are neither independent nor identically distributed in practice for various reasons: post earnings announcement drift effects; serial correlations of IBES individual analysts forecast revisions; and analyst herding behaviour.Further investigation of the same distributions found that they adhere variously to the Extreme Value distribution, the Uniform distribution and other unspecified non-normal distribution types. This information was utilised to construct a best estimate of distributions, the surrogate consensus, which was found to be more accurate than the IBES consensus for the actual forecast error deflator type. It was benchmarked against four other surrogate consensus generation techniques: weighted consensus based on prior year end’s lead analyst, analysts’ prior accuracy, earnings forecast age and analyst forecast frequency, with none of these alternatives offering improvements over the IBES consensus.