The RATS Software Forum

Dear Tom:

In general time series analysis, If they has unit roots, we should test cointegration between them, then decide to estimate VECM or VAR. In panel time series analysis, most panel analysis focous on individul and period, fixed and random effect, even dynamic model. If panel analysis also focous on its time series attribution, why only give panel unit roots and panel cointegration test,neglate panel VECM and panle VAR. If these model need consider, how they code?

thanks
Hardmann

On the first page of the this section of the forum, about 1/2 the questions are about some form of estimation of cointegrating vectors or error correction models (PANELDOLS, PANELFM, Pooled Mean Group, etc.). You don't see many full panel VAR's or panel VECM's because it is often unreasonable to assume that the short-run dynamics are common across individuals. As a result, those short-run dynamics are treated as "nuisance" parameters to the estimation of the (presumed fixed across individuals) cointegrating vector.