Chupter 22
PANEL DATA
GARY CHAMBERLAIN*
University of Wisconsin-Madison and NBER
Contents
1. Introduction and summary
2. Specification and identification: Linear models
2.1. A production function example
2.2. Fixed effects and incidental parameters
2.3. Random effects and specification analysis
2.4. A consumer demand example
2.5. Strict exogeneity conditional on a latent variable
2.6. Lagged dependent variables
2.1. Serial correlation or partial adjustment?
2.8. Residual covariances: Heteroskedasticity and serial correlation
2.9. Measurement error
1248
1254
1254
1256
1257
1259
1262
1264
1267
1268
1269
3. Specification and identification: Nonlinear models 1270
3.1. A random effects probit model 1270
3.2. A fixed effects logit model: Conditional likelihood 1274
3.3. Serial correlation and lagged dependent variables 1278
3.4. Duration models 1282
4. Inference 1285
4.1. The estimation of linear predictors 1286
4.2. Imposing restrictions: The minimum distance estimator 1288
4.3. Simultaneous equations: A generalization of three-stage least squares 1292
4.4. Asymptotic efficiency: A comparison with the quasi-maximum likelihood e: 1294
4.5. Multivariate probit models 1296
5. Empirical applications 1299
5.1. Linear models: Union wage effects 1299
5.2. Nonlinear models: Labor force participation 1304
6. Conclusion 1311
Appendix 1311
References 1313