We investigate the time series properties of semi-annual earnings and compare the out-of-sample predictive ability of the seasonal random walk model and the various models identified using the Box-Jenkins method. Using 195 manufacturing firms during t...
We investigate the time series properties of semi-annual earnings and compare the out-of-sample predictive ability of the seasonal random walk model and the various models identified using the Box-Jenkins method. Using 195 manufacturing firms during the first half of 1987 to the first half of 1997 (21 observations), we document that the autocorreation coefficients of semi-annual earnings display some seasonality. We also provide evidence that the simple seasonal random walk process is not statistically inferior to other more complex models based on the time-series properties in out-of-sample predictive ability.