I have a balanced panel data (N= 190, T=5) on income and personal characteristics of the householder.
I would like to estimate the coefficients and variances of a temporary and of a permanent income shock.
I fitted an OLS regression model as follows: $log(income)=sex+age+age^2+study+ public administration + type of house + country$
I then took the residuals and regressed them on a ARMA (1,1) model. The coefficient of AR(1) should be the permanent part and should be around 1, the MA(1) should be the temporary part. I also found that the AR(1) coefficient is only 0.2 instead of about 1. Moreover, AUTOARIMA fits them with an ARMA(2,2). How can I read the coefficients of the ARMA (2,2) to find out the temporary and permanent shock on the income?
I would like also to find the variance of the income (which is the residuals of the ARMA regression) but is it right even if the model is not correct?