time-varying coefficient

[jeanbaptiste]

Hi guys,

I'm just wondering if it is possible to estimate a time-varying coefficient using eviews. I'm hypothesising that the relationship between my dependant variable and one of my independant variables is not constant but rather that it varies over time.

The specification I would like to model is:

y = c(1) + B(t)*b(t-1) + c(2)*g + c(3)*y + e(t)

where B(t) is the average coefficient for the variable b(t-1) and then I want to calculate an additional time-varying coefficient B'(t) which is the deviation from the average coefficient in time t.

Thus, the actual value of the reaction coefficient for b(t-1) is B(t)+B'(t) and is different for each time period t.

Is this possible? I've tried to estimate it in a state space object but my results don't look right. Here's what I've used:

@signal y = c(1) + sv1*b(-1) + c(2)*g + c(3)*y + [var = exp(c(4))]

@state sv1 = sv1(-1) + [var=exp(c(5))]

Any thoughts or advice would be greatly appreciated!

Thanks in advance.

[trubador]

You defined the coefficent as a random walk. Here are other plausible specifications:

Stochastic coefficient:

Code: Select all

@state sv1 = c(6) + [var = exp(c(5))]

AR(1) coefficent:

Code: Select all

@state sv1 = c(6) + c(7)*sv(-1) + [var = exp(c(5))]

Please refer to the manual or search the forum for other possible specifications...