| Modified Jones Model |
The so-called “modified Jones model” is a variation of the original Jones model (see Jones, 1991) proposed by Dechow et al. (1995) to separate nondiscretionary and discretionary accruals, which are used for earnings management purposes.
Like in the original Jones model, total accruals are regressed on a set of independent variables that are supposed to drive the extent of nondiscretionary accruals in the reporting period, thus letting the error term capture the unobservable extent of discretionary accruals. The only modification compared with the original model is that the change in revenues is adjusted for the change in receivables.
This adjustment is only made in the event period (where earnings management is supposed), and the original model is fitted in the other periods. The reasoning behind this adjustment is that, contrary to the assumption in the original model, managers indeed have discret ion over recognizing revenues, particularly when it comes to sales on credit.
  |
Hence, changes in sales on credit are more likely to be manipulated and therefore drive rather discretionary than nondiscretionary accruals. However, the modification implies that all sales on credit in the event period are connected to the earnings management activities.
This is not a more convincing assumption than supposing that revenue recognition is not a subject to earnings management at all. It thus seems likely that the modified Jones model will overstate discretionary accruals (i.e., earnings management) when sales and receivables increase.
However, Dechow et al. (1995) provide evidence for the modified model, exhibiting more power in detecting earnings management than the original model. Like the original model, the modified model also was criticized for overestimating the level of discretionary accruals within periods of extreme financial performance.
Consequently, Kothari et al. (2005) empirically find that discretionary accruals estimations based on the original or the modified Jones model can be enhanced by performance matching. In the literature, various other modifications of the original Jones model have been discussed. However, distributional tests and rankings have more often been used than accruals models in recent research.
