Usage of dummy variables in URCM and causality

[lcfreitas]

Dear friends,
I have two questions that maybe are simple but that are consuming lot of time to solve. I’m trying to run a unrestricted error correction model (UECM) of gasoline consumption that consists of a dependent variable (first difference log gasoline consumption) and explanatory variables (lagged first difference of log gasoline price, lagged first difference of log gasoline consumption, lagged first difference of log income, lagged level variables for dependent and explanatory variables AND a dummy for technology shock). This technology shock refers to the introduction of a specific policy and happened in the middle of a 40 years series.

My questions are related to the dummy.

First, when i test for cointegration how should i treat the dummy? Literature teaches that the cointegration exam (Pesaran Bounds test with Wald F-stat) is done based on the coefficients of the lagged first difference of explanatory variables, in other words without the lagged level variables. In this case should i also ignore the coefficients of the dummy in my cointegration test?

Second question is related to how to manage the dummy variable in a granger causality test. Does it go to the estimation of causality as any other variable? (ps. Should it be in log form? Do I have to lag it?)

thank you in advance