Pieter T. Elgers, May H. Lo, Reductions in Analysts' Annual Earnings Forecast Errors Using Information in Prior Earnings and Security Returns, Journal of Accounting Research, Vol. 32, No. 2 (Autumn, 1994), pp. 290-303

[This study evaluates the relation between security returns and funds-based earnings components. We document that proxies for market expectations of the components that are based on measures of historical serial-and cross-dependencies are substantially more accurate than random-walk proxies. Moreover, we detect significantly higher valuations of the operating cash flow component of earnings, relative to current accruals, when market expectations are represented using the dependency-based predictions. Such differential valuation is not detectable for random-walk representations. Contrary to results in Ali (1994), we find incremental information in unexpected cash flows over the whole spectrum (moderate and extreme) of unexpected cash flow realizations.]

This paper documents that the weighting of analysts' annual earnings forecasts implicit in security prices is lower than the historical relation between financial analysts' forecasts and realized earnings. Short positions in securities in the bottom decile and long positions in the top decile of the crosssectional distribution of analysts' early-in-the-year earnings forecasts generate significant hedge-portfolio returns in the year after portfolio formation. This delayed price response is more pronounced for firms with relatively low analyst coverage, consistent with the premise that low financial analyst coverage is associated with a variety of factors that impede the information efficiency of the security market. The hedge-portfolio returns concentrate in the months of subsequent quarterly earnings announcements, suggesting that the delayed security price adjustments reflect the market's failure to incorporate information in analysts' forecasts about future earnings, rather than deficiencies in our conditional expectations of security returns.

This study evaluates the relation between security returns and funds-based earnings components. We document that proxies for market expectations of the components that are based on measures of historical serial- and cross-dependencies are substantially more accurate than random-walk proxies. Moreover, we detect significantly higher valuations of the operating cash flow component of earnings, relative to current accruals, when market expectations are represented using the dependency-based predictions. Such differential valuation is not detectable for random-walk representations. Contrary to results in Ali (1994), we find incremental information in unexpected cash flows over the whole spectrum (moderate and extreme) of unexpected cash flow realizations.