One of my student is checking the effect of corporate governance on earnings management.
earnings management is measured using Discretionary accrual (calculated by Modified Jones model). the da values are not normal so I go for normal distribution. for the purpose of normalizing she use the abs value of da and then normalize it. The resultant normalized abs have some negative values.
Is it correct and if yes;
She want to substantiate this. Please suggest insights.
Negative accruals are perfectly fine from my results as well. Modified Jones model does not require Normality , as OLS does assume it but we have already weighted the variables by price. I use robust errors. Absolute value, per me, should be an incorrect use of the Modified Jones model information. My assumption for negative accruals(cause) is the more dependence on Current Liabilities and for an equi-measured explanation in Working Capital investment terms ( The TA represent Net Working Capital investments)
Dear Sir,
For the purpose of running regression that is to say regressing da as dependent variable on independent variables, isn't normality mandatory?
As per OLS, yes, but given a sample size of more than 30/100, and use of t-tests, the requirement need not be established and noise evaluated as a OLS modeled data
In the OLS methodology, da estimates anyways will include higher noise and test reliability is verified in the literature. DSS(1999), Dechow(2011)