Empirical Modeling of Social Science Theory: Advanced Topics


  • Robert J. Franzese, University of Michigan

This is a course in the specification, estimation, interpretation, and presentation of empirical models appropriate for the context-conditionality, the over-time and cross-unit (inter-) dependence, and the ubiquitous endogeneity that characterizes modern, sophisticated social-science theory. As theory advances, the implications for empirical outcomes tend to grow correspondingly richer. In modern social-science theories, for instance, the effects of most causal factors are not constant but rather vary depending on many features of the contexts in which those factors occur. Also, the outcomes in some spatial units at sometimes depend, according to our modern theoretical & substantive understandings, on those in other units and earlier times. And, we now understand better, just about everything in social reality causes & is caused by (i.e., is endogenous to) just about everything else. This course explores how best to specify empirical models that reflect these complex substantive-theoretical understandings, and then how to estimate, interpret, & present the results of such models.

Participants are encouraged to bring their own ideas, data, and projects as applications on which to try some of the course content.

We will try to incorporate class time for instruction & practice aimed toward the furthering of these projects, and models and methods to be covered may vary depending on student interests and projects and on the availability of exemplary extant research building closely from theory through empirical-model specification, estimation, interpretation, and presentation. For certain, though, students will learn how to specify, estimate, interpret, and present empirical models that reflect the (1) interaction effects and context conditionality, (2) the temporal, spatial, and spatiotemporal dynamics and interdependence, and (3) the ubiquitous joint endogeneity of modern theories of domestic, comparative, and international political and social science. The specific empirical models and methods covered will span three broad areas:

  1. interaction & heterogeneous-parameter models, such as multiplicative-interactive models, nonlinear least-squares for nonlinear models with additive-separable moments, maximum likelihood for non-separable nonlinear models, & multilevel (a.k.a. hierarchical, random-effect/coefficient, …) models;
  2. temporal, spatial, & spatiotemporal dynamic models, such as time-serial, event-history, panel/TSCS, transition/switching, and spatial-econometric models, possibly including context-conditional dynamics;
  3. strategies for causal-parameter and causal-response estimation given ubiquitous endogeneity, some example specific models and methods for which may include systems-of-equations, instrumental-variables, vector-autoregression, discontinuity, matching, and difference-in-difference designs.
Our intent is also to cover all 3 areas for both continuous and limited &/or qualitative dependent-variable cases.

PrerequisitesThe course has no prerequisites. All methods to be employed will be thoroughly, albeit (very) quickly, introduced en route. A prior course in Linear Regression and in Qualitative-Data Analysis/Maximum-Likelihood Estimation would be helpful and is recommended, however. (Representative background would be a course in linear regression and beginning qualitative-dependent-variable analysis at the level of, for example, Gujarati, Basic Econometrics or Wooldridge, Introductory Econometrics.)

EITM certification is available for graded course completion.

Fees: Consult the fee structure.

Tags: models, EITM, empiricism

Course Sections

Section 1

Location: ICPSR -- Ann Arbor, MI

Date(s): July 23 - August 17

Time: 10:00 AM - 12:00 PM


  • Robert J. Franzese, University of Michigan