Dokument: Nonlinear Panel Data Models with High-Dimensional Fixed Effects
| Titel: | Nonlinear Panel Data Models with High-Dimensional Fixed Effects | |||||||
| URL für Lesezeichen: | https://docserv.uni-duesseldorf.de/servlets/DocumentServlet?id=52065 | |||||||
| URN (NBN): | urn:nbn:de:hbz:061-20200123-100436-8 | |||||||
| Kollektion: | Dissertationen | |||||||
| Sprache: | Englisch | |||||||
| Dokumententyp: | Wissenschaftliche Abschlussarbeiten » Dissertation | |||||||
| Medientyp: | Text | |||||||
| Autor: | Stammann, Amrei [Autor] | |||||||
| Dateien: |
| |||||||
| Beitragende: | Prof. Dr. Heiß, Florian [Gutachter] Prof. Dr. Stiebale, Joel [Gutachter] | |||||||
| Dewey Dezimal-Klassifikation: | 300 Sozialwissenschaften, Soziologie » 330 Wirtschaft | |||||||
| Beschreibung: | The careful handling of unobserved heterogeneity – such as individual or time specific effects – is an essential concern in econometrics for causal analyses. If these unobserved effects are related to any explanatory variable, neglecting them causes an omitted variables bias. A great advantage of panel data over pure cross-sections is that they allow to fully control for these unobserved effects and thus offer new possibilities to researchers that go beyond proxy variable and instrumental variable approaches. Another benefit is the possibility to study different sources of persistence (see chapter 1.2 in Baltagi 2013 and Hsiao 2014 for a comprehensive list of further advantages). For example, Roberts and Tybout (1997) and Bernard and Jensen (2004) employed dynamic discrete choice models to disentangle the drivers behind firms’ exporting persistence, such as sunk costs and unobserved plant heterogeneity. Within panel data estimators, fixed effects estimators are very popular, because unlike random-effects estimators, they do not impose any distributional assumption on the unobserved heterogeneity. A flourishing part of the theoretical econometric literature is in particular concerned with nonlinear fixed effects models. However, their application confronts practitioners with several problems.
This thesis is intended to draw empirical researchers’ attention to nonlinear fixed effects estimators and to facilitate their applicability. For this purpose, the strand of econometric literature on bias corrections is linked to computational advances. This makes it possible to estimate nonlinear models even with many observations and high-dimensional fixed effects, which is more and more required due to the increasing magnitude of panel data sets. This thesis also contributes to the literature on bias corrections itself by providing further insights on finite sample properties of various corrections and proposing novel corrections for special two- and three-way fixed effects models required in international trade. Further, I offer the corresponding software routines, bife and alpaca, to make the methods presented in this thesis ready to use. | |||||||
| Lizenz: |  Urheberrechtsschutz | |||||||
| Fachbereich / Einrichtung: | Wirtschaftswissenschaftliche Fakultät » Statistik und Ökonometrie | |||||||
| Dokument erstellt am: | 23.01.2020 | |||||||
| Dateien geändert am: | 23.01.2020 | |||||||
| Promotionsantrag am: | 24.09.2019 | |||||||
| Datum der Promotion: | 10.12.2019 |
