I am doing cointegration analysis (all my macroeconomic series are I(1)). In addition, I want to include in all my cointegration analysis a dummy variable indicating the years where there were dictatorships (0 is democracy, 1 is dictatorship). There were two dictatorships, so the dummy looks like 0,0,0...0,1,1....,1,0,0...0. By including this dummy, my aim is to determine the effect generated these dictatorships. For me makes sense to include it as exogenous, but I am not sure.
In most of the statistical softwares and books it is indicated that the dummies can be included if they are seasonal, or indicating significant events (like a crisis). An exception is when there are breaks in the series and in this case should be followed Johansen et al (2000). But this case of Johansen et al (2000) seems not to be my case.
I have read several journal articles and I have never seen the inclusion of this kind of dummy.
I have read the Juselius (2006) book in the page 139
"We have shown that the asymptotic distributions depend on whether there is a constant and/or a trend in the VAR model and whether they are unrestricted or not. However, other deterministic components, such as intervention dummies, are also likely to influence the shape of the distributions. In particular, care should be taken when a deterministic component generates trending behaviour in the levels of the data. A typical example is an unrestricted shift dummy (...,0,0,0,1,1,1,...) which cumulates to a broken linear trend in the data. A detailed discussion of this case can be found in Johansen, Mosconi, and Nielsen (2000) and Doornik, Hendry, and Nielsen (1998). However, even if the shift dummy is restricted to lie in the cointegration relations, the asymptotic distributions will be affected (Nielsen 2004a)".
Right after she added
"Unrestricted dummy variables which do not cumulate to (broken linear) trends, for example transitory and permanent innovational dummies, are not likely to have an effect on the asymptotic distributions, but can nevertheless influence the finite-sample distributions."
From the explanation in the first paragraph of Juselius, I understand that I cannot add the dummy for the policy as exogenous (neither as endogenous), but the second paragraph, confuses me.
Please, can I add this dummy? If yes, should I include it as exogenous?