15

Computational Considerations in

Empirical Microeconometrics:

Selected Examples

David T. Jacho-Chávez and Pravin K. Trivedi

Abstract

The substance and style of modern microeconometrics is shaped by its role in analyses of pub-
lic policy issues. Computational considerations have proved to be an important influence on the
methodology and scope of empirical analyses that address these issues. To be convincing to a wide
readership the empirical analyses need to be based on representative data and flexible modeling
approaches. In this chapter we illustrate, through a variety of empirical examples, how modelers
handle the complexities that arise from the richness of survey data and the heterogeneity in behav-
ior of market participants. After introductory sections on data and programming languages, the
remainder of the chapter covers many leading computationally intensive econometric techniques.
These are illustrated by means of specific numerical examples. An algorithmic format is used to
describe the computational features.

15.1 Introduction 776
15.2 Preliminary 778
15.2.1 Programming languages 778
15.2.1.1 Characteristics 779
15.2.2 Foreign language interface 779
15.2.3 Parallelization 780
15.3 Computing and modeling 780
15.3.1 Data summary and visualization 781
15.3.2 Numerical optimization 781
15.3.3 Simulation-assisted estimation 784
15.3.3.1 MNP example 784
15.3.3.2 Heterogeneity example 785
15.3.4 Resampling methods 786
15.3.5 Structural models based on dynamic programming 787
15.4 Non/semiparametric methods 788
15.4.1 Nonparametric estimation 789
15.4.1.1 Example: kernel density estimation 789
15.4.1.2 Example: conditional density estimation 790
15.4.1.3 Example: additive models 792

775