I'm using ARDL model to regress the permanent income hypothesis. Regarding the Engle-Granger and bounds test, I conclude that the consumption and income series are cointegrated. However, there is a significant structural break detected by Gregory-Hansen structural break test.
I placed a dummy variable for that year to make ARDL significant but with or without the break dummy, ARDL results are inconsistent, only break dummy and consumption cointegrated in the long run.
Did I make model-misspecification?