Earnings management research design - An evaluation of aggregate and specific accruals models

Sammanfattning: This study aims to evaluate whether accruals models are useful to detect earnings management in listed Swedish companies. We analyse the ability of aggregate and specific accruals models to detect manipulation, to provide unbiased estimates of discretionary behaviour and to identify known cases of accruals fraud. Our results show that both types of accruals models have limited usefulness in detecting manipulation at levels associated with fraudulent accounting, which is partly due to noise in the estimation procedure. While specific accruals models outperform aggregate accruals models, we argue that required detection rates for these models are higher if manipulation is spread across items. These findings contribute to recent studies criticising the ability of accruals models to detect earnings management by showing that specific accruals models may not be a preferable option to aggregate accruals models.