Körber, Lena
(2015)
Essays in panel data econometrics with cross-sectional dependence.
PhD thesis, London School of Economics and Political Science.
Abstract
The behavior of economic agents is characterized by interdependencies that arise from common shocks, strategic interactions or spill-over effects. Developing new econometric methodologies for inference in panel data with cross-sectional dependence is a common theme of this thesis. Another theme is econometric models that allow for heterogeneity across individual observations. Each chapter takes a different approach towards modeling and estimating panels with cross-sectional dependence and heterogeneity. In all chapters, the perspective is one where both the time series and the cross-sectional dimension are large.
The first chapter develops a methodology for semiparametric panel data models with heterogeneous nonparametric covariate effects as well as unobserved time and individual-specific effects that may depend on the covariates in an arbitrary way. To model the covariate effects parsimoniously, we impose a dimensionality reducing common component structure on them. In the theoretical part of the chapter,
we derive the asymptotic theory of the proposed procedure. In particular, we provide the convergence rates and the asymptotic distribution of our estimators. The asymptotic analysis is complemented by a Monte Carlo experiment that documents the small sample properties of our estimator.
The second chapter investigates the effects of fragmentation in equity markets on the quality of trading outcomes. It uses a unique data set that reports the location and volume of trading on the FTSE 100 and 250 companies from 2008 to 2011 at the weekly frequency. This period coincided with a great deal of turbulence in the UK equity markets which had multiple causes that need to be controlled for. To achieve this, we use the common correlated effects estimator for large heterogeneous panels
that approximates the unobserved factors with cross-sectional averages. We extend this estimator to quantile regression to analyze the whole conditional distribution of market quality. We find that both fragmentation in visible order books and dark trading that is offered outside the visible order book lower volatility. But dark trading increases the variability of volatility and trading volumes. Visible fragmentation has the opposite effect on the variability of volatility, in particular at
the upper quantiles of the conditional distribution.
The third chapter develops an estimator for heterogeneous panels with discrete outcomes in a setting where the individual units are subject to unobserved common shocks. Like the estimator in chapter 2, the proposed estimator belongs to the class of common correlated effects estimators and it assumes that the unobserved factors are contained in the span of the observed factors and the cross-sectional
averages of the regressors. The proposed estimator can be computed by estimating binary response models applied to regression that is augmented with the crosssectional averages of the individual-specific regressors. The asymptotic properties of this approach are documented as both the time series and the cross-section tend to infinity. In particular, I show that both the estimators of the individual-specific coefficients and the mean group estimator are consistent and asymptotically normal. The small-sample behavior of the mean group estimator is assessed in a Monte Carlo experiment. The methodology is applied to the question of how funding costs in corporate bond markets affect the conditional probability of issuing a corporate bond.
Actions (login required)
 |
Record administration - authorised staff only |
Download statistics
Download statistics