How to treat a dynamic ordered logit with a lagged dependent variable?

[Carsten Preuss]

Dear community,

I am having an ordered categorical dependent variable and a lagged version of the dependent variable as a covariate. I know that in least-squares models estimates can be inconsistent and biased upwards due to the inclusion of a lagged dependent variable, even if the regression errors are uncorrelated.

For least-squared models, there are several estimators that deal with this issue, for example GMM (xtbond2) or system GMM estimators, by first-differencing the dynamic equation and using the lagged covariates as instruments.

However, it seems to me that information about how to treat dynamic equations in a multivariate setting, such as ordered logit, is somewhat scarce. Given the different nature of categorical variables, or multivariate models, I was wondering whether there exists any consistent estimator for such dynamic models, or how to deal with this issue in general?

Thanks for your hints!

Carsten

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a helpful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output [what you posted is extremely hard to read], and sample data using dataex.

I don't know a canned estimator that does this. In such cases, you often have to move to a general estimator like GSEM. The extended regression routines in Stata 15 might also work.