Nonparametric transition based-tests for Markov processes in financial econometrics

点击翻译按钮获取中文摘要

Nonparametric model identification and specification play a critical rolein asset pricing and risk management. The objective of the investigators isto study a number of important model specification problems usingnonparametrically estimated transition densities and transitiondistributions. The proposed method integrates a number of seeminglyunrelated model identification problems into a common paradigm, by framingthem as various restrictions on the transition density or distribution ofthe underlying process, sampled at the observation frequency. This willprovide a rich framework for analyzing other similar problems. Byestimating the transition density and distribution under null andalternative hypotheses, appropriate test statistics will be proposed and investigated. Theproperties of the tests includeboth the asymptotic null distributions and power functions and will alsobe analyzed via carefully designed simulations. The relative merits ofvarious methods will be critically examined. Several testing problemsarising frequently in the empirical finance will be carefullyinvestigated.The investigators intend to develop new techniques and theories that arisefrom frontiers of financial econometrics. The investigators will developmodels and cutting-edge technologies for model specification problems ineconomics and financial econometrics. Common characteristics of theseproblems are that the dynamics of stocks and interest rates are hard tospecify and hence contemporary nonparametric techniques are particularlypowerful for these problems. The proposed techniques and results will havestrong impact on assest pricing, portfolio allocations, risk managementsand other problems in empirical finance. The project will integrateresearch and education by working closely with both undergraduate andgraduate students, creating datasets and developing publicly availablecomputer code. Postdoctoral fellows and underrepresented groups will betrained as a part of our research investigation. The results will bedisseminated broadly through presentations at seminars, conferences,professional association meetings and via internet.