AbstractsBusiness Management & Administration

Topics in applied financial econometrics

by Νικόλαος Βαφειάδης




Institution: University of Ioannina; Πανεπιστήμιο Ιωαννίνων
Department:
Year: 2012
Keywords: Αποτίμηση περιουσιακών στοιχείων; Φίλτρα Kalman; Κλασματική συνολοκλήρωση; Αναλογία διακυμάνσεων; Υποδείγματα μεταβλητότητας; Κλασματική μνήμη σε υποδείγματα μεταβλητότητας; Υποδείγματα μεταβλητότητας που ενσωματώνουν φίλτρα; Ανταλλαγή μεταβλητότητας ως μέτρου του κινδύνου με αποδόσεις; Asset pricing models; KALMAN FILTER; Fractional cointegration; Variance ratio; Volatility models; Fractional memory in volatility models; Volatility models incorporating filters; Volatility return trade off
Record ID: 1153515
Full text PDF: http://hdl.handle.net/10442/hedi/26798


Abstract

The present dissertation is oriented towards the empirical application of certain models and econometric techniques drawn from recent developments in the financial econometric literature. The aims of this project are targeted a) in testing the proposed financial models to financial data sets, b) in enriching and strengthen the analysis by inducing new aspects into the proposed methodologies, and finally c) in producing inferences and comparing the outcomes with other results existing in many related empirical applications. Each chapter in the present dissertation corresponds a different section of applied econometrics and therefore three empirical projects are carried out. Those are : a) the estimation and test of the joint conditional CAPM model introduced by Morelli (2011), b) the detection of fractional cointegrating relations using the variance ratio approach introduced by Nielsen (2010), and finally c) a comparative analysis of different volatility models, aiming a) the comparison of their volatility forecasting potentials under various forecasting horizons, and b) the detection of possible statistically significant volatility - return relations. Specifically : Chapter 1 follows the approach of Morelli (2011) and carries over the estimation and test of the joint conditional CAPM model. The analysis uses the monthly returns of the 25 Fama-French portfolios in the period from July 1926 to June 2008 to evolve in two phases. The first part of the analysis through the application of four different methodologies estimates corresponding versions of the time varying beta coefficients series, while the second based on those latter estimates tests the statistical significance of the beta - return relation, especially when the last is conditioned upon the sign of excess market returns. Note that the above methodologies correspond a) the volatility approach, where conditional covariances and variances that define the notion of conditional beta are modeled as ARCH, GARCH, FIGARCH and FIEGARCH processes, b) the recursive OLS approach, and c) the Kalman filter analysis, where two different assumptions have been applied on the definition of the state equation. Those are a) the random walk approach and the b) AR(1) alternative. In spite of differences existing in all four versions of the estimated time varying beta coefficients series, results in all four procedures reject the conditional and the joint conditional CAPM versions, while results appear robust either when the analysis examine the full sample case or two equal sub-samples. The key feature of chapters 2 and 3 evolves around the idea of long memory that is detected both in cointegrating relations and volatility return series. From the initial work of Granger (1981) to nowadays there has been an increasing amount of evidence supporting the presence of long memory in different financial and macroeconomic series, with the list including data over exchange rates, interest rates, indexes of production, consumption, unemployment, estimated series on volatility and many…