Long Run VAR alpha and beta estimation
Posted: Wed Mar 05, 2014 6:26 am
Hi,
I am using a VAR with 2 variables and 4 lags. I am combining the coefficients of these variables to get an overall alpha (a) and beta (B) value for in the form Y = (a) + (B)X
In order to get the long run alpha, I am using the following formula:
a = C/[1-a(1)-a(2)-a(3)-a(4)], where the alphas are the coefficients for lagged Y.
and the following formula for the beta calculation:
B = [B(1)+B(2)+B(3)+B(4)]/[1-a(1)-a(2)-a(3)-a(4)], where Bs are coefficients for lagged X and a are coefficients for lagged Y.
Am I correct in saying that these provide the LONG RUN alpha and beta values, or are these equations incorrect?
Also, and most importantly, what equation would I use to find the significance of these headline alpha and beta values, are they derived from the standard errors/t-stats?
Attached is an example of the above, with the headline alpha and beta coefficients derived from the above formulas.
Many thanks in advance,
Adam
I am using a VAR with 2 variables and 4 lags. I am combining the coefficients of these variables to get an overall alpha (a) and beta (B) value for in the form Y = (a) + (B)X
In order to get the long run alpha, I am using the following formula:
a = C/[1-a(1)-a(2)-a(3)-a(4)], where the alphas are the coefficients for lagged Y.
and the following formula for the beta calculation:
B = [B(1)+B(2)+B(3)+B(4)]/[1-a(1)-a(2)-a(3)-a(4)], where Bs are coefficients for lagged X and a are coefficients for lagged Y.
Am I correct in saying that these provide the LONG RUN alpha and beta values, or are these equations incorrect?
Also, and most importantly, what equation would I use to find the significance of these headline alpha and beta values, are they derived from the standard errors/t-stats?
Attached is an example of the above, with the headline alpha and beta coefficients derived from the above formulas.
Many thanks in advance,
Adam