This study proposes an alternative measure of discretionary accruals that can be used in testing for intentional earnings management. Prior research has shown the prevalence of measurement error in all models used to estimate discretionary accruals (Healy (1985), DeAngelo (1986), Jones (1991) and modified Jones models (Dechow et al., 1995). The alternative measure proposed relies on the premise that managers use one or more components of accruals (accounts receivable, inventories, accounts payable, other working capital and depreciation) to manipulate bottom-line income in a given direction, consistent with their incentives. In other words, components of discretionary accruals are expected to be positively correlated. If they are not, this is an indication of high measurement error in the models estimating them. The alternative measure is tested in terms of its power (type II error) and specification (type I error) and compared to the traditional discretionary accruals measure. The power of the tests is measured in random samples with added accrual manipulation as well as a sample of firms targeted by the Securities and Exchange Commission for alleged fraud and a sample of firms that violated their debt covenants. The results indicate that the power of this alternative discretionary measure is higher than that of the traditional discretionary accruals measure. The specification (specificity) is tested in random samples chosen from the full sample as well as random samples chosen from extreme income and cash from operations observations and a sample in which discretionary accruals is a noisy measure of the estimated discretionary accruals. The results indicate that the specification of detecting earnings management behavior is improved by using the alternative discretionary accruals measure.