Viewing Discretionary and Non-Discretionary Accruals Through the Univariate Lens: A Conditional Heteroscedastic Mean-Variance Approach

Abstract

Unlike no study in accounting literature, I do a firm-by-firm application of the univariate conditional heteroscedastic (UCH) mean-variance modeling technique –hereafter called ‘UCH Model’ –to propose a new way to decoupling a firm’s discretionary from non-discretionary accruals. What the UCH Model essentially does is it kills any linear dependence in accruals estimation errors to isolate firm-specific shocks to accruals reporting so that ‘model misspecification’ and ‘low test power’ concerns –known to have plagued existing accruals models –are completely eliminated. The UCH mechanism invokes an intuitive, but popular econometric time-series argument that for the same time series variable (e.g. stock return, earnings per share, accruals), each sample unit (i.e. firm) tends to have a unique behavior and should be modelled distinctively. I provide empirical evidence supporting this and conduct a series of validation tests on this new univariate measure of discretionary accruals to assess if it appeals to economic theories largely popularized in the earnings management literature. Specifically, I find that the UCH discretionary accruals measure –capturing managers’ opportunistic reporting of accruals –significantly and positively associates with the ‘smoothing’ incentive (i.e. managerial intents to report abnormally large accruals when current earnings is already above different types of earnings targets [industry profitability, prior-period earnings, and analysts’ forecasts]). Cross-sectional analyses also reveal that a number of other management intents (e.g. avoidance of debt covenant violation, large firms’ desire to minimize exposure to political costs, high operations uncertainty, etc.) affect the nature and extent to which managers report abnormally large accruals.