| The identification of estimation samples is important for estimating discretionary accruals using an accrual model. This study proposes an alternative identification of estimation samples using a firm's life cycle (i.e., life cycle-based estimation samples). Analyses using U.S. and international data show that when detecting accrual manipulation, life cycle-based estimation samples outperform industry-based or size-based estimation samples in sample retention, specification, and detection power. Improved detection power by life cycle-based estimation samples is also evident in the AAERs sample. Lastly, I reexamine Dechow, Richardson, and Tuna (DRT) (2003) and Teoh, Wong, and Rao (TWR) (1998) applying life cycle-based estimation samples. I find that life cycle-based estimation samples change the inferences from DRT by improving test power and mitigate misspecification in TWR. Collectively, the current study provides empirical evidence that supports the use of life cycle-based estimation samples over other existing estimation samples. |
