Several studies have characterized the relation between discretionary accruals and earnings before-taxes to test for the existence of earnings smoothing behaviors. In this paper, we argue that the characteristic response of accruals to earnings is not linear, as the literature has shown. Instead, it is likely to be driven by non-linear patterns since both the incentives to manipulate earnings and the practical way to do so depend, in part, on the relative size of earnings. Using a sample of 9,442 US banks in the period 1999-2008, this paper shows that bank managers tend to use provisions as a smoothing devise when earnings are substantial ("cookie-jar" strategies), engage in earningsdecreasing strategies when losses are relatively large ("big-bath" accounting) and, most of the time, use provisions as an earnings-increasing tool. Hence, it is shown that nonlinear specifications are more informative with regard to the different strategies employed to manipulate earnings. Abstract Several studies have characterized the relation between discretionary accruals and earnings before-taxes to test for the existence of earnings smoothing behaviors. In this paper, we argue that the characteristic response of accruals to earnings is not linear, as the literature has shown. Instead, it is likely to be driven by non-linear patterns since both the incentives to manipulate earnings and the practical way to do so depend, in part, on the relative size of earnings. Using a sample of 9,442 US banks in the period 1999-2008, this paper shows that bank managers tend to use provisions as a smoothing devise when earnings are substantial ("cookie-jar" strategies), engage in earningsdecreasing strategies when losses are relatively large ("big-bath" accounting) and, most of the time, use provisions as an earnings-increasing tool. Hence, it is shown that nonlinear specifications are more informative with regard to the different strategies employed to manipulate earnings.
Marina Balboa