| The purpose of this study was to examine the quarterly time-series properties of selected financial-statement line items and subtotals. Once time-series models for these items were determined, another purpose of the study was to determine whether deviations from predicted future values of these items was consistent with the behavior of the capital markets. Both ARIMA and vector autoregressive techniques were employed in order to determine time-series models for six financial statement subcomponents (accounts receivable, gross margin, inventory, current liabilities, depreciation expense, and selling and general administrative expense). Once such models were developed, firms were ranked based upon the size of the deviation from the one-quarter-ahead prediction, and several event studies were undertaken to determine whether the deviation introduced value-irrelevant noise into the earnings response coefficient. The more sophisticated expectation models (ARIMA and VAR) outperformed the more naive models (random walk and seasonal random walk) for three of the six subcomponents. However, the sophistication of the expectation model did not appear to be relevant to the capital markets. If a deviation from the expectation model for a given subcomponent was value-relevant, it generally was value-relevant no matter which expectation model was used. |