Grant: Discovery Grants Program - Individual
Université de Montréal
$16,000.00
Bootstrap inference for panel data models wirh serial and cross sectional dependence of unknown form
FY 2009-10
Summary
The first part of this program considers bootstrap methods for panel data regression models estimated with fixed effects. In particular, I develop a new bootstrap method for panel regression models that is robust to serial and cross sectional dependence of unknown form in the error term. I call it the panel block bootstrap method because it resamples blocks of the vector that collects the cross section units for each time period. It is the bootstrap analogue of the Driskoll and Kraay (1998) covariance estimator for spatial and serial dependent panel data models. Preliminary simulations show that the panel block bootstrap has good finite sample properties and dominates the Driskoll and Kraay (1998) robust standard errors. I prove the first order asymptotic validity of the panel block bootstrap method under conditions similar to those used by Driskoll and Kraay (1998). Then I investigate to what extent the panel block bootstrap can outperform the asymptotic theory of Driskoll and Kraay (1998).
The second part of this research program considers bootstrap methods of inference on large factor models. Factor models are a useful dimension reduction device because they can summarize the information contained in a large number of variables. The estimated factors are often used subsequently in a regression or a forecasting model.
The main goal is to develop bootstrap methods for regression models that rely on estimated factors. As Bai and Ng (2006) show, two sources of estimation uncertainty arise: one is due to the estimation of the parameters in the regression model and the other is due to estimation of the factors. I intend to propose a bootstrap method that captures both forms of estimation uncertainty.
Program
Discovery Grants Program - Individual
Recipient
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
341322-2009
Name nom
Gonçalves, Silvia
Department département
Sciences économiques
Organization id
63
Institution établissement
Université de Montréal
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2009
Competition year année de concours
2009
Award amount
16000
Program id
RGPIN
Program nane en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
14
Committee name en
Statistical Sciences
Committee name fr
Statistique
Area of application code
1212
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Social sciences
Area of application fr
Sciences sociales
Research subject code
3002
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Time series analysis
Research subject fr
Analyse des séries chronologiques
Partie
1
Nb partie
5
Application title
Bootstrap inference for panel data models wirh serial and cross sectional dependence of unknown form
Keyword
NA
Application summary
The first part of this program considers bootstrap methods for panel data regression models estimated with fixed effects. In particular, I develop a new bootstrap method for panel regression models that is robust to serial and cross sectional dependence of unknown form in the error term. I call it the panel block bootstrap method because it resamples blocks of the vector that collects the cross section units for each time period. It is the bootstrap analogue of the Driskoll and Kraay (1998) covariance estimator for spatial and serial dependent panel data models. Preliminary simulations show that the panel block bootstrap has good finite sample properties and dominates the Driskoll and Kraay (1998) robust standard errors. I prove the first order asymptotic validity of the panel block bootstrap method under conditions similar to those used by Driskoll and Kraay (1998). Then I investigate to what extent the panel block bootstrap can outperform the asymptotic theory of Driskoll and Kraay (1998).
The second part of this research program considers bootstrap methods of inference on large factor models. Factor models are a useful dimension reduction device because they can summarize the information contained in a large number of variables. The estimated factors are often used subsequently in a regression or a forecasting model.
The main goal is to develop bootstrap methods for regression models that rely on estimated factors. As Bai and Ng (2006) show, two sources of estimation uncertainty arise: one is due to the estimation of the parameters in the regression model and the other is due to estimation of the factors. I intend to propose a bootstrap method that captures both forms of estimation uncertainty.