Realized Volatility, Jumps and the Interface between Financial Markets and the Real Economy

点击翻译按钮获取中文摘要

The recent availability of high-frequency intraday asset prices and real-time economic announcement data for a host of different financial markets and instruments has spurred a large and rapidly growing literature concerned with the statistical and empirical analysis of this new rich source of data. This project aims to further expand on our ability to extract useful information about important economic phenomena from such data through the development of new and general econometric procedures and modeling paradigms, coupled with specific empirical applications. In particular, building on the investigators' earlier work, they seek to obtain: (i) new robust non-parametric procedures for disentangling the day-today price variation into continuous and discontinuous components, and corresponding procedures for modeling, forecasting and pricing continuous and jump risks; (ii) a better understanding of the type of events, or news, that induces large price movements, or jumps, in financial asset prices and their relation to the macro economy; (iii) new and improved realized variation measures for better accommodating market microstructure complications and other frictions in the high-frequency data; (iv) a better understanding of the empirical linkages between economic fundamentals and asset markets, as illuminated by the simultaneous high-frequency response of multiple cross-country markets to specific macroeconomic news announcements; (v) new and improved procedures for measuring and modeling time-varying correlations and beta factor loadings and a better understanding of the macroeconomic determinants behind the apparent temporal dependencies.Broader Impacts: It is by now widely accepted that financial market volatility is predictable, and that this predictability has profound practical implications for asset pricing, risk management, monitoring, and oversight. In theory, the use of more frequently sampled data should result in better volatility measurements and more accurate forecasts. However, actual high-frequency financial data are beset by a host of complications relative to the stylized parametric models employed in the existing (G)ARCH and stochastic volatility literature and only very recently have some of the gains, expected to be harnessed from the use of finer sampled intraday data, started to materialize empirically. The realized volatility measures and forecasting models developed in the invesetigators' prior NSF-sponsored research have been at the forefront of these developments. The present proposal seeks to expand on these ideas in several important directions, including new robust multivariate procedures for measuring and forecasting realized correlations and factor loadings, along with the development and use of non-parametric measures for assessing the contribution to the overall price variation attributable to jumps, or discontinuities, in turn allowing for separate modeling, pricing and hedging of the continuous and discontinuous part of the price process. Importantly, the investigators also seek to obtain a better understanding of the type of economic news that induces large price movements in financial markets, both across different markets and internationally, and as such hope to shed new light on the fundamental linkages between asset markets and the real economy across business cycles. The general results of the proposed activities should therefore be of relevance to applied macroeconomists, time series econometricians, applied statisticians, financial researchers, regulators, and practitioners alike.