I applied OLS regression to predict total accruals by using modified Jones model.
the model is as follows
TA=alpha1*1/At-1+alpha2*(delta_Revenue - delta_Receivables)/At-1+alpha3*(PPE)/At-1.
the results shows that
alpha2 is negative and insignificant and alpha3 also negative but significant
R-squared value is 0.18 with significant F-statistic.
with this regression results, Can I derive Non-Discretionay accruals(i.e., predicted total accruals) and Discretionary accruals(i.e., residual from the regression).
I am also enclosing regression output
thanks in advance
You may want to read the following paper to help you in this regard:
Is modified Jones model effective in detecting earnings management? Evidence from a developing economy
MA Islam, R Ali, Z Ahmad - International Journal of Economics and Finance, 2011
You just need to predict the residuals, then you have the discretionary accruals (or abnormal accruals). After that you can use in absolute value as well. I forgot to mention a point, check if you need to winsorize at 1% on each tail variables accruals, d.rev and ppe. You can also keep the constant.
In addition, maybe Faisal Nawaz could be helpful to you regarding the financial data analysis as he is the expert to my knowledge.
Hi there,
I am using the same model for earnings management.But my coefficients are high and signs are reversed .The coefficeint value comes out to be negative for 1 /lagged assets and also for Total accruals coefficeint sign is negative. I then standardized all variables but the issue persists.
I am using panel data and results show significant p values with R square value of 0.96 .
Am i using wrong regression or is it ok to have reversed signs for MJM model?
P.s: I have to use the Discretionary accruals as a IV in another model.
Please suggest me something ?