Doctor of Philosophy
Dr. John Knight
This dissertation contributes to the theory and the applications of copulas to problems in economics, econometrics and finance. The second chapter proposes a new measure of macroeconomic news which is termed the "Macroeconomic News Index". Using the copula approach, new findings about the relationship between macroeconomic news and the stock markets are revealed. The third chapter aims to improve the existing non-parametric copula-based tests for stochastic independence. It provides an extension to the test statistic of Kojadinovic and Holmes (2009), which is obtained through the introduction of a weighted functional norm. The addition of the weights creates a channel through which the power properties of the test can be manipulated. Certain choices of the weights are shown to give the statistic a significant power advantage. The third chapter provides additional results which enable the application of the test to regression model residuals. The test is used to probe for the presence of conditional heteroscedasticity, and is shown to have a power advantage over the test of White (1980). The fourth chapter provides a serial extension to the statistic, and further extends the results of Quessy (2010), which permits the application of the statistic to the testing for the goodness of fit of serial copulas. An upper bound for the independence test statistic is derived in Chapter 4, and a standardized version of the statistic is proposed, which can serve as an omnibus measure of vectorial serial dependence. A computational formula for the new copula-based dependence measure is provided.
Medovikov, Ivan, "News, Copulas and Independence" (2013). Electronic Thesis and Dissertation Repository. 1176.