First differences - Ordered Probit

[Dayla Spassio]

Hi,
I have a dataset (T=2 and N=4300 (2300 in the first wave of which 2000 panels are present in the second wave)).
I'm estimating an ordered probit, where the
dependent variable y= takes values between 0 and 2
I wanted to ask you, is it possible to apply the method of first differences if the dependent variable is an ordinal variable and almost all the regressors are dummies?
do I have to consider the difference prime of all the regressors included in the model?
Thank you.

Best,

[Fei Wang]

May I ask why you'd like to apply first difference? Differences of an ordinal variable may not make any sense. For example, 1 is the difference between 1 and 0, or between 2 and 1, but they may indicate different things.

[Dayla Spassio]

It was recommended to me to eliminate fixed effects, but I had doubts. I also think like you.

[Fei Wang]

You may apply correlated random effects (CRE) estimation to the ordered probit, to account for panel-specific effects.