Hi everyone,
I have a question for those with expertise in using ARDL and ECM.
Suppose you have the following short-run equation in a panel data model:
Δ yit = αi + λ1i Δ yi,t-1+... +λpiΔ yi,t-p + θ1i xit + β0i Δ xit+ β1i Δ xi,t-1 + ... + βpi Δ x1i,t-p + γ0i Δ zit +γ1i Δ zi,t-1+ ...+γpi Δ zi.t-p+ uit
where we allow for slope heterogeneity, and the error term is cross-sectionally correlated. Here, xit , yit and zit are observable scalars.
Now, the long-run effect for individual iii of a permanent change in x on Δy is given by: θ1i / (1 - λ1i - ... - λpi )
The short-run effect for individual i of a change in z on Δy at time t should, I assume, be γ0i . I'm assuming that z does not have long-run effects on Δ y.
Is there a way to calculate a cumulative short-run effect of z on Δy?
The first idea that comes to mind is using a CS-ARDL, but I’m not sure if I can restrict z to only have short-run effects. Is there a way to test such a restriction in a panel setting?
Thanks!
I have a question for those with expertise in using ARDL and ECM.
Suppose you have the following short-run equation in a panel data model:
Δ yit = αi + λ1i Δ yi,t-1+... +λpiΔ yi,t-p + θ1i xit + β0i Δ xit+ β1i Δ xi,t-1 + ... + βpi Δ x1i,t-p + γ0i Δ zit +γ1i Δ zi,t-1+ ...+γpi Δ zi.t-p+ uit
where we allow for slope heterogeneity, and the error term is cross-sectionally correlated. Here, xit , yit and zit are observable scalars.
Now, the long-run effect for individual iii of a permanent change in x on Δy is given by: θ1i / (1 - λ1i - ... - λpi )
The short-run effect for individual i of a change in z on Δy at time t should, I assume, be γ0i . I'm assuming that z does not have long-run effects on Δ y.
Is there a way to calculate a cumulative short-run effect of z on Δy?
The first idea that comes to mind is using a CS-ARDL, but I’m not sure if I can restrict z to only have short-run effects. Is there a way to test such a restriction in a panel setting?
Thanks!
