Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

A Factor Analytical Approach to Dynamic Panel Data Models

Norkute, Milda LU (2016)
Abstract
This thesis deals with the development and application of new estimation approaches based on factor analysis for estimation and inference in dynamic panel data models with fixed-effects.

A new factor analytical method (FA) for the estimation of fixed-effects dynamic panel data models is proposed in Bai ("Fixed-Effects Dynamic Panel Models, A Factor Analytical Method". Econometrica 81, 285-314, 2013), which has the unique and very useful property that it is not subject to the incidental parameter problem and is therefore asymptotically bias-free. In Chapter I we complement Bai's theoretical study by providing Monte Carlo evidence of the good small-sample performance of the FA method.

One of the attractive features of the... (More)
This thesis deals with the development and application of new estimation approaches based on factor analysis for estimation and inference in dynamic panel data models with fixed-effects.

A new factor analytical method (FA) for the estimation of fixed-effects dynamic panel data models is proposed in Bai ("Fixed-Effects Dynamic Panel Models, A Factor Analytical Method". Econometrica 81, 285-314, 2013), which has the unique and very useful property that it is not subject to the incidental parameter problem and is therefore asymptotically bias-free. In Chapter I we complement Bai's theoretical study by providing Monte Carlo evidence of the good small-sample performance of the FA method.

One of the attractive features of the FA method is that it does not require explicit detrending, a practice that is known to cause problems of bias and low power in (near) unit root panels, especially in models that include a linear trend. While certainly appealing, the FA approach, as originally proposed, is restricted to fixed-effects models without a unit root. In Chapter II we investigate the properties of this estimator when it is used to estimate (near) unit root panels with a possible trend.

In Chapter III we propose an extension of the FA method to cover dynamic panel data models with interactive fixed-effects and a moving average error structure. The limiting distribution of the modified FA method is derived and it is shown that this estimator is asymptotically normal and unbiased. The performance of the FA method is evaluated by conducting a small-scale Monte Carlo simulation exercise.

Most empirical evidence suggests that the efficient futures market hypothesis, stating that spot and futures prices cointegrate with a unit slope on futures prices, does not hold. In Chapter IV we test this hypothesis using the FA method and a large panel data set including 17 commodities from March 1991 to August 2012. The empirical results suggest that this hypothesis cannot be rejected.

In Chapter V we investigate if the New Keynesian Phillips curve (NKPC) is supported empirically in the Euro Area using a data set that is disaggregated by both country and sector. The NKPC are estimated using the Factor-GMM estimator, which is superior to GMM in terms of efficiency. We test the NKPC by combining the individual results. Our results provide no empirical evidence in favour of the NKPC. (Less)
Please use this url to cite or link to this publication:
author
supervisor
opponent
  • Professor Urbain, Jean-Pierre, Maastricht University, Netherlands
organization
publishing date
type
Thesis
publication status
published
subject
keywords
Dynamic panel data models, Factor analytical method, Bias, Incidental trends, MA errors, Factor-GMM
pages
279 pages
defense location
EC3:211, Holger Crafoord Centre, Tycho Brahes väg 1, Lund
defense date
2016-05-24 13:00:00
ISBN
978-91-7623-831-8
978-91-7623-830-1
language
English
LU publication?
yes
id
a9fe4bbe-d339-45f9-a91d-658a28063f54
date added to LUP
2016-04-26 12:17:02
date last changed
2019-09-17 14:03:36
@phdthesis{a9fe4bbe-d339-45f9-a91d-658a28063f54,
  abstract     = {{This thesis deals with the development and application of new estimation approaches based on factor analysis for estimation and inference in dynamic panel data models with fixed-effects. <br/><br/>A new factor analytical method (FA) for the estimation of fixed-effects dynamic panel data models is proposed in Bai ("Fixed-Effects Dynamic Panel Models, A Factor Analytical Method". Econometrica 81, 285-314, 2013), which has the unique and very useful property that it is not subject to the incidental parameter problem and is therefore asymptotically bias-free. In Chapter I we complement Bai's theoretical study by providing Monte Carlo evidence of the good small-sample performance of the FA method.<br/><br/>One of the attractive features of the FA method is that it does not require explicit detrending, a practice that is known to cause problems of bias and low power in (near) unit root panels, especially in models that include a linear trend. While certainly appealing, the FA approach, as originally proposed, is restricted to fixed-effects models without a unit root. In Chapter II we investigate the properties of this estimator when it is used to estimate (near) unit root panels with a possible trend. <br/><br/>In Chapter III we propose an extension of the FA method to cover dynamic panel data models with interactive fixed-effects and a moving average error structure. The limiting distribution of the modified FA method is derived and it is shown that this estimator is asymptotically normal and unbiased. The performance of the FA method is evaluated by conducting a small-scale Monte Carlo simulation exercise.<br/><br/>Most empirical evidence suggests that the efficient futures market hypothesis, stating that spot and futures prices cointegrate with a unit slope on futures prices, does not hold. In Chapter IV we test this hypothesis using the FA method and a large panel data set including 17 commodities from March 1991 to August 2012. The empirical results suggest that this hypothesis cannot be rejected. <br/><br/>In Chapter V we investigate if the New Keynesian Phillips curve (NKPC) is supported empirically in the Euro Area using a data set that is disaggregated by both country and sector. The NKPC are estimated using the Factor-GMM estimator, which is superior to GMM in terms of efficiency. We test the NKPC by combining the individual results. Our results provide no empirical evidence in favour of the NKPC.}},
  author       = {{Norkute, Milda}},
  isbn         = {{978-91-7623-831-8}},
  keywords     = {{Dynamic panel data models; Factor analytical method; Bias; Incidental trends; MA errors; Factor-GMM}},
  language     = {{eng}},
  school       = {{Lund University}},
  title        = {{A Factor Analytical Approach to Dynamic Panel Data Models}},
  url          = {{https://lup.lub.lu.se/search/files/7508461/Thesis_Milda_Norkute.pdf}},
  year         = {{2016}},
}