Ecm can also be applied to panel data, similar to Time series analysis. The basic idea is that, when u have 2 or more variables with I(1), the linear dependence of these variables will give a lower level of integration and the long run convergence will be achieved through the error correction term. The use of VECM is based on the relationship between the dependent variable and independent variable. If the RHS variable is assumed to be endogenous, the VECM will be appropriate, but if the RHS is exogenous, then Panel ARDL will be appropriate

Thank you, Hamid Muili, for your response.

As Hamid Muili has said you can apply error correction models in

panel data. However, you need to be very careful. Using standard panel

estimation you are assuming that

If you

When you are doing econometrics you must use a model and estimation procedure that match your economic theory. A method that imposes theoretically (economically) invalid restrictions on your model will lead to invalid results.

To answer your second question I would need to know a lot about X and

Y, their panel and time dimensions, and the economic theory that connects

X and Y. If you are making some form of causal conclusion you must use

an economic model setting out the relevant causal links. Otherwise your

estimates will almost certainly be spurious.

Panel ARDL will do two things

The theory of the bounds test starts with a k+1 VECM with a single error correction term. It establishes when one equation may be estimated using single equation methods. This requires (along with various regularity conditions) that the error correction term enter only the single equation to be estimated. If you need to examine the stability of the system or other dynamic properties, you must also estimate the remaining k equations.

ECM for panel data accounts for short-term dynamics and long-term equilibrium. Techniques like PMG, MG, or DFE estimators in Stata/EViews are used depending on homogeneity assumptions.