Many economic phenomena require limited variable models for an appropriate treatment. In addition, panel data models allow the inclusion of unobserved individual-specific effects. These models are combined in this thesis. Distributional assumptions in the limited dependent variable models are important for consistency of the estimators. Parametric models require strong distributional assumptions. The hypothesis of correct specification of these distributional assumptions is often rejected. Therefore these assumptions are relaxed, resulting in semiparametric models. The empirical evidence shows that the results based on less restrictive semiparametric models often differ in terms of both parameter estimates and economically interesting issues like Engel curves and equivalence scales. Testing the specification of the semiparametric models sometimes leads to the conclusion that an even more general model is required.
|Qualification||Doctor of Philosophy|
|Award date||28 Nov 1997|
|Place of Publication||Tilburg|
|Publication status||Published - 1997|