On the identification and estimation of panel data dynamic discrete choice models

IZA Logo

IZA Seminar

Place: Schaumburg-Lippe-Str. 9, 53113 Bonn

Date: 24.02.2009, 12:15 - 13:30


Presentation by 

Bo E. Honore (Princeton University)


While it is generally difficult to deal with individual-specific effects in dynamic nonlinear panel data models, some progress has been made in specific models. The new results presented here will generalize known results for simpler models to logit models with second order state dependence and state-specific parameters. These results are for example relevant if one wants to model the transitions between employment and unemployment. In that case, one would not want to assume that the parameters that govern the transition from employment to unemployment are the same as the parameters that govern the transition from unemployment to unemployment. Results concerning bivariate logit models will also be presented. These are relevant if one wants to jointly model two discrete variables such as health status and labor force participation.

For more information, please contact seminar@iza.org