We measure the commonality of analysts' information using the BKLS correlation measure ([rho]), which is a measure of the across-analyst correlation in forecast errors and is thus a direct measure of the construct examined by Fischer and Verrecchia (1998).
BKLS further show that under certain assumptions (which we review in Section V), one can use this measure of commonality in analysts' forecast errors to draw inferences about the information individual analysts possess at the time they report their forecasts.
The BKLS measures allow us to use observable attributes of analysts' forecasts to derive empirical estimates of two key (unobservable) dimensions of an information structure among multiple analysts: (1) the precision of their information (i.e., the inverse of uncertainty), and (2) the commonality among different analysts' beliefs (i.e., their consensus), measured as the across-analyst correlation in forecast errors.
Selection of Individual Forecasts Used to Calculate BKLS Measures
This procedure ensures that when comparing the BKLS measures of analysts' information before and after earnings announcements, we base our analysis on the same set of individual analysts.
We substitute ex post realized dispersion (D) and squared error in these analysts' mean forecast (SE) for the expected dispersion (D) and squared error in the mean forecast (SE) used in the BKLS model.
Substituting ex post realizations for expected values introduces measurement error into our measures of expected dispersion and squared error in the mean forecast and, in turn, into our estimates of BKLS measures.
We compute estimates of [rho], h, and s--[rho], h, and s--by substituting the number of our selected analysts forecasts (N), together with the ex post realized dispersion (D) and squared error in the mean forecast (SE) for these forecasts, both scaled by the absolute actual earnings per share, into the BKLS equations shown here in Equations (3), (4), and (5).
Table 1 shows the median and mean levels of BKLS correlation ([rho]) and the BKLS precision measures (h and s) for the [Q2.sub.B] forecast window.
(6) The Appendix provides a detailed discussion of the BKLS framework and implications for our predictions.
In addition, the BKLS framework would consider any common information disclosed privately to all analysts to be effectively public information.
(8) In the BKLS framework, error in the mean forecast is defined as the squared difference between the mean analyst forecast and actual earnings.