Arash Ghorbani; Mohammad Reza Abbaszadeh
Abstract
Using a sample including 2642 observations of 2003-2016 annual data of firms listed in Tehran Security Exchange, this study investigates the implications of correlation between non-discretionary accruals and partitioning variables when testing the positive accounting theories hypotheses. In earnings ...
Read More
Using a sample including 2642 observations of 2003-2016 annual data of firms listed in Tehran Security Exchange, this study investigates the implications of correlation between non-discretionary accruals and partitioning variables when testing the positive accounting theories hypotheses. In earnings management detection tests, it is common for researchers to use variables which partition their sample into two groups, for which differences in motivation for income manipulation are predicted. Since earnings management stimuli are assumed to be correlated with variables like firm performance, leverage or size, the use of these proxies are popular in empirical tests of positive accounting theories hypotheses. The correlation between non-discretionary accruals and the partitioning variable implies that part of the variation of non-discretionary accruals is generated by the partitioning variable. If accruals model does not control for the correlation, this part will be added to the discretionary accruals. In this study, we provide evidence that, when correlation between non-discretionary accruals and the above-mentioned partitioning variables remains uncontrolled, accrual-based models tend to generate measurement error in the estimate of discretionary accruals that significantly affects the sign and the magnitude of correlation between discretionary accruals and the partitioning variables. The findings of a Monte Carlo simulation also indicate that the Jones model relatively generates less type I error when it is adjusted to control for the relation between non-discretionary accruals and firm performance.
