An Empirical Evaluation of Accounting Income Numbers, 1968, Ball & Brown, University of Chicago and University of Western Australia, Journal of Accounting Research

By relating income numbers to stock prices, this paper examines the information content of annual reports.

It starts with two OLS equations and gets their residuals as new information contained in income changes and causing unexpected stock returns. The first equation regresses change in firm j's income on the average change in all other firms' income. The second equation regresses the monthly price relative of firm j minus one on the market's monthly rate of return minus one. The monthly price relative of firm j equals dividend plus closing price divided by opening price. The market's monthly rate of return uses Fisher's computed index.

Table 1: Coefficients of correlation between firm and market income changes

Table 2: First-order autocorrelation of income regression residuals

The paper then computes the Abnormal Performance Index (API) and plots it for three sets of portfolios in Figure 1. The first set is for the positive income forecast errors for the three variables ( net income and EPS for the regression model and EPS for the naive model); the second set is for the negative income forecast errors for the three variables; the third set is a single portfolio consisting of all firms and years in the sample.

The paper then computes the Total Information (0.731), Net Information (0.165) and Income Information (0.081 for variable 1, 0.083 for variable 2 and 0.077 for variable 3) and concludes that 25% information persists, half of which is contained in income reports. It further concludes that only 10% to 15% of the information contained in income reports has not been anticipated by the month of their release and that income reports make up only 20% of all information on the market in the month of their release.