Lecture 12: Nested Fixed Point Algorithm (NFXP)

Plan for lecture 12

We will go though Rust (1987) in detail. This is a path-breaking paper that introduces a methodology to estimate a single-agent dynamic discrete choice models by MLE. The main contribution is that Rust develop and implement the so-called Nested Fixed Point Algorithm (NFXP) to obtain ML estimates of the structural parameters. It's a nested algorithm, since it requires the researcher to repeatedly solving a fixed point problem for each evaluation of the sample criterion used in estimation (in this case the likelihood function). Another key contribution of Rust is the formulation of a set of assumptions, that makes dynamic discrete choice models tractable. Finally, he devise a "bottom-up approach", where aggregate demand goods are based on a micro aggregation of individual decision rules. In this first lecture on discrete choice models, we will focus on the illustrative empirical application: A simple model of engine replacement. Rust model will be the leading example of discrete choice model's throughout the lectures.

Preparation:

Material: 

Online lectures

Video 1: Plan for lectures in dynamic discrete choice models

https://www.youtube.com/watch?v=SbVIgzWt8So Links to an external site.

Video 2: Lecture 12 - part 1 - Solving Zurcher

https://www.youtube.com/watch?v=JfFCZhBYgGw Links to an external site.

Video 3: Lecture 12 - part 1 - Structural Estimation (NFXP) and Empirical Results.

https://www.youtube.com/watch?v=YpCptgY9vzw Links to an external site.