Topic：Does overnight information play an important role in predicting daytime volatility in the financial markets?

Guest：Dr. Nirodha Jayawardena

Time：10月31日（周一）下午15：30-17：00

Venue：2/F Seminar Room, Business School（商学院二楼研讨室）

Host：Prof. Chen Siyan（陈斯燕副教授）

Language: English

Abstract：

“Does overnight information play an important role in predicting daytime volatility in the financial markets?” This is an unresolved question in the literature on financial volatility. Due to the global integration of financial markets, the need for market efficiency is becoming more pronounced. More specifically, the need to account for information on the overnight or non-trading period is even more pertinent in contexts such as the Australian Stock Exchange (ASX) and the Tokyo Stock Exchange (TSE), because their geographical proximity means that they are outside the trading hours of major global markets such as the New York Stock Exchange (NYSE) and the London Stock Exchange (LSE). Thus, when a new business day dawns, much pertinent information is waiting to impact price developments.

There are five essays in my thesis. The first essay undertakes a thorough investigation of after-hour information’s impact on the volatility forecast of the ASX. I adopt a new forecasting approach by using squared overnight returns, pre-open volatility of the same assets and realised volatilities of related assets from other markets, where intraday data is still available while the ASX is closed, in order to predict stock return volatility. I use a number of different specifications of the well-specified, Heterogeneous Autoregressive (HAR) model introduced by Corsi (2009) to identify an optimal way to incorporate this additional information. The empirical analysis of the ASX constituents confirms the usefulness of using pre-open volatility of the same asset and realised volatilities of related assets from other markets when the ASX is closed, in order to forecast future volatility.

The second essay is an empirical application of the methodology developed in the first essay to the Tokyo Stock Exchange (TSE). Aside from an overnight period, the TSE also has a midday break for lunch, creating an interesting research platform. Similar to the first essay, I study volatility forecasting for the TSE using the realised volatilities of related assets traded in other markets around the world and related neighbour-market information where the intra-day data is available while the TSE is closed for night and for lunch respectively. The empirical analysis confirms the potential of using global and neighbour-market information in forecasting equity volatility on the TSE.

The third essay evaluates the impact of augmented volatility (the adjusted volatility accounting for the overnight/non-trading gap of the ASX) on the predictability of the future returns of the ASX. Apart from the well documented traditional approach of monitoring risk-return regressions using the second moment (volatility), I conjointly account for other higher-order moments such as the third and the fourth moments (skewness and kurtosis) to investigate the impact of overnight information corrected moments on predicting the future returns, using the Co-integrated Fractional Vector Autoregressive system (CFVAR). I find that the monthly compounded realised volatility-return and realised skewness-return adjusted with the fractional integration parameter show significant negative and positive relationships with the subsequent monthly returns respectively, and the addition of after-hour information improves the regression fit. Further, the multivariate setting of this study implies that there exists a co-integrating relationship between the two variance series — realised volatility and VIX, which can be related as the variance risk premium, as discussed in Bollerslev, Osterrieder, Sizova, and Tauchen (2013). The fourth essay graphically depicts the relationship between the realised higher moments (realized volatility-return, realized skewness-return) and the future return predictability of the ASX using a graphically powerful directional predictability technique named cross-quantilogram introduced by Han, Linton, Oka, and Whang (2016). The results reveal a significant directional predictability between realised higher moments and return when the quantiles in both time series are either low or high.

The fifth essay deploys four volatility-based trading strategies and evaluates the economic significance of the augmented HAR models using three profitability criteria: Sortino ratio, Leland’s alpha and performance-based switching fee, for various risk tolerance levels. One finding that is consistent in both statistical and economic loss functions is the relevance of exploiting additional information, over and above historical returns, for daily forecasting. The results reveals that the statistical significance does not always result in an economic value of the forecasts; in other words, this analysis suggests that de facto statistical metrics such as the Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) may be of little value to practitioners when making their investment decisions.

Biography of Ms. Nirodha Jayawardena

Ms. Nirodha Jayawardena recently completed her PhD in Finance at Griffith University, Australia. Her research title is “Essays on stock market volatility using high-frequency data: The role of overnight information”. She did her honors degree in Sri Lanka, University of Colombo, specializing statistics. She has also completed CIMA(UK) and CFA level 1 examinations. She has published three of my research studies in peer-reviewed journals including Economic Modelling (ABDC rank A) and has two other working papers in revised and resubmit stage in International Journal of Forecasting (ABDC rank A). She served as an assistant lecturer in University of Colombo, Sri Lanka and as a research assistant in Griffith university, Australia. Her research areas are; volatility modelling, applied financial econometrics, and empirical asset pricing.

欢迎全校师生参加！

商学院

2016-10-28