2SLS with time series in R (with dynlm package?)

Question

I want to do the two-stage least squares estimation with time series. I know I can do it manually, but the standard errors then are not right.

I accidentally saw in the description of the dynlm package that it includes 2SLS estimation. However, I don't seem to understand how to do it. All that we are given in the instructions is:

To specify a set of instruments, formulas of type y ~ x1 + x2 | z1 + z2y ~ x1 + x2 | z1 + z2 can be used where z1 and z2 represent the instruments.

How do I specify which variable ( x1 or x2 ) is endogenous (and would be dependent variable in the first stage)?

Answer 1

On the right of |you want to specify all the instruments - including the exogenous variable already in the equation. Whatever is not on the right will be instrumented - assumed to be endogenous. So, if x1 is exogenous, x2 is endogenous, and you have one extra instrument z1 , then you want

y ~ x1 + x2 | z1 + x1

To see that this is right, try including just z1 on the right: then one coefficient will be not identified.