# Capital in VAR model

Dear All:
Greeting to all old members.
I have a simple Cash in the Advance model in a closed economy and after log linearizing, it became a VAR model of seven variables. There is one K_{t+1} and K_t and H_{t+1} and H_t, which is labor hours, in the var model, I have two problems here.

1. It seems there is not exclusive physical capital data on Fred and other organizations. If no capital stock data for production function, then how would I proceed with the Var model?.
2. I have been thinking so hard about the labor hours. I assume it is between 0 and 1, I also collected employment data from Fred, how will I normalize the employment data as between 0 and 1?.
If anyone could help, I would really appreciate. I am new to this form and Dynare.

Dear muhtar,

1. There is a direct mapping between investment and capital: I_t=K_t-(1-\delta)K_{t-1}, so what is usually used is data on investment.
2. You can use hours worked.

Dear Cmarch:
So, I will extract gross fixed capital formation from Fred as an investment and use it directly as capital or by setting an initial value for K_0, then extract K_t?.
So, how will normalize the hours worked to between 0 and 1?.
Thanks.

@muhtar You need to elaborate what your are doing. Usually, DSGE models have a solution in the form of a restricted VARMA process in the observables. Why did you get a (seemingly unrestricted) VAR process after linearization, but before solving the model?

Also, it is unusual to estimate a model with the capital stock as an observable. Using investment is more common.

In K_t = I_t + (1-\delta)K_{t-1}, DYNARE sets K_{t-1} = K_{ss} (steady state capital) or K_{t-1} = 0. More likely steady state, but I wanna to be sure, meaning K_{t-1} is unique in DYNARE can’t be chosen outside the model? Or we can? Thanks.

What do you mean? How the initial period K_0 is set?

Yes, can one choose K_0 = K_{ss} or say K_0 = 0 or some other value in dynare.

Predetermined values can be set with histval for many purposes. See the Dynare manual for details. In estimation, the smoother can be used to infer to initial values.

1 Like