Initial condition problem in RE probit

[Maya Lani]

Dear all,

I am struggling a bit in understanding what is the initial condition problem truly and really is.

Also, how can you adjust the RE probit to remedy for the initial condition problem.

Please note, I do not have data, my questions are theoretical (for now).

[Carlo Lazzaro]

Maya:
googling with the following string -initial condition problem probit- gives back many promising entries.

[Maya Lani]

Dear Carlo,

I have done that already. But every paper I came across state that the initial condition is a problem for dynamic models and that you need to specify the initial value.

However, I don't understand why that is a problem...

[Stephen Jenkins]

In a "dynamic random effects probit model", the presence of the lagged dependent variable means that there a correlation induced between the error term and one of the explanatory variables = bad news for the properties of the standard probit estimator. Hence the various estimators proposed to account for this issue. (NB as per Forum FAQ, you should use full literature references, rather than refer to "every paper".)