Order of estimation on PS5

[olivianatan]

@rickecon -
Maybe I'm missing something, but I'm not sure why parts A and B of Q 1 aren't switched. We need to specify alpha to get our z_t series, so it makes sense to me to first estimate (alpha, beta), then estimate (rho, mu), since our results there vary by what we use for alpha - What am I missing?
Olivia

[bobaekang]

@rickecon In addition to the question posed by @olivianatan, I am also wondering if we should assume the value for z_0 to be something to get a vector for z_{t-1}, just as we did in PS3. Is it different this time? Or should we just use a (99, 1) vector for each variable? Maybe I'm just not understanding how to approach this problem...

[rickecon]

@olivianatan and @bobaekang . You're right. Here is the correct way to estimate the Brock and Mirman (1972) model by GMM.

1. Guess a values for alpha, beta, rho, and mu.
2. Given alpha and the data r_t and k_t, you can use the first order condition of the firm for r_t to back out the series for z_t.
3. Once you have z_t, you can compute the four moments listed in the problem set.
4. Your GMM estimates for alpha, beta, rho, and mu are the ones that minimize the weighted sum of squared errors from those four moment conditions.

I will update the problem set question after class.

[rickecon]

I just pushed the changes to Exercise 1 in Problem Set 5.