Simulation techniques - ganong-noel/lab_manual GitHub Wiki

Dynamic stochastic optimization

Here are a few related resources. I'm not sure which will turn out to be most relevant or useful, so I'm giving several:

solvingmicrodsops.pdf. This is Chris Carroll's explanation of the general problem. He invented a method to speed up solving these models accurately which is called the method of endogenous grid points. (Prior to this, researchers would use the same grid in every time period, which is less efficient).
- Very detailed explanation of what is happening in general. Possibly too detailed. It gets very dense and in the weeds math-wise at times. The most relevant information starts around Section 6, but prior sections still need to be read for context.
problemSet3_2013.pdf, PS3_Noel_FINAL.PDF. See question 2 in this problem set and my writeup. This was from Alp Simsek's macro field course at MIT. I am going to save what I think is the matlab code with the solution, which might be helpful in strategic_drive/lit/numerical methods/.
Alp (and Laibson, see below) cites Stokey and Lucas. Alp specifically cites Chapter 4. I saved a pdf of that textbook to strategic_drive/lit/numerical methods/, and you might also find it at the library.
David Laibson notes from first-year PhD macro. If you look at slide 13 of lecture 1, he cites three methods for solving these dynamic programing problems. I think what they are doing is related to option (3), i.e. iterating numerically, which is discussed a bit more in Lecture 2. 2010c Lecture 01 Fall 2015.pdf 2010c Lecture 02 Fall 2015.pdf

Original tweet

Answers:

The biggest thing is making sure they understand GMM first. For that, I recommend staring at a bunch of examples showing OLS and IV as special cases. SMM comes from wanting to do GMM but not being able to explicitly write the model moments. The general rule for doing econometrics when you can't/aren't smart enough to do the analytic derivation is simulation! A good single resource is @QuantEcon's SMM notes. It includes Python code, and the best way to learn econometrics is doing it.
I encoutered the same Q during my MPhil at Oxford. I learned SMM techniques by reading papers and reviewing their code. I recommend structural lifecycle papers by Attanasio, Blundell, Low, Pistaferri et al.
Chapter 10 of Ken Train's book