The results in Section 2.1 imply that the null hypothesis that the density of the distribution of earnings changes does not change at zero should be rejected. We next turn to the problem of estimating the frequency of earnings management to avoid earnings decreases, i.e. calculating the difference between observed fre quencies of earnings changes and frequencies which would have been expected in the absence of earnings management. Note that the model of expectations used to test the null hypothesis of no earnings management (where the expected number of observations in an interval was defined as the average of the observed numbers in the two adjacent intervals) is not appropriate for estimating the frequency of earnings management because we now have evidence that the null hypothesis does not hold. Also, while tests of significance focused on the change in density at zero (in order to minimize the assumptions required for the test), estimates of the frequency of earnings management should allow for the fact that earnings management is not necessarily confined to just the two intervals adjacent to zero.
With these considerations in mind, we adopt the following model for the purpose of estimating the frequency of earnings management. We assume that in the absence of earnings management, the distribution of earnings changes would be approximately symmetric and that the right half of the empirical distribution is largely unaffected by earnings management to avoid earnings decreases. Using this model, the observed frequencies from intervals in the right half of the empirical distribution serve as measures of the expected frequencies in the corresponding interval in the left half of the distribution. Operationally, we assume that in the absence of earnings management, the distribution of earnings changes in Fig. 1 would be symmetric around 0.01 and that managed values of earnings changes do not fall to the right of 0.01.
The results in Section 2.1 imply that the null hypothesis that the density of the distribution of earnings changes does not change at zero should be rejected. We next turn to the problem of estimating the frequency of earnings management to avoid earnings decreases, i.e. calculating the difference between observed fre quencies of earnings changes and frequencies which would have been expected in the absence of earnings management. Note that the model of expectations used to test the null hypothesis of no earnings management (where the expected number of observations in an interval was defined as the average of the observed numbers in the two adjacent intervals) is not appropriate for estimating the frequency of earnings management because we now have evidence that the null hypothesis does not hold. Also, while tests of significance focused on the change in density at zero (in order to minimize the assumptions required for the test), estimates of the frequency of earnings management should allow for the fact that earnings management is not necessarily confined to just the two intervals adjacent to zero.
With these considerations in mind, we adopt the following model for the purpose of estimating the frequency of earnings management. We assume that in the absence of earnings management, the distribution of earnings changes would be approximately symmetric and that the right half of the empirical distribution is largely unaffected by earnings management to avoid earnings decreases. Using this model, the observed frequencies from intervals in the right half of the empirical distribution serve as measures of the expected frequencies in the corresponding interval in the left half of the distribution. Operationally, we assume that in the absence of earnings management, the distribution of earnings changes in Fig. 1 would be symmetric around 0.01 and that managed values of earnings changes do not fall to the right of 0.01.
การแปล กรุณารอสักครู่..
