ویژگی های واقعی یا ساختگی حافظه طولانی مدت نوسانات: شواهد تجربی از بازار نوظهور
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|13761||2013||6 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Economic Modelling, Volume 30, January 2013, Pages 67–72
We examine whether real or spurious long memory characteristics of volatility are present in stock market data. We empirically distinguish between true and spurious long memory characteristics by analysing different types and measurements of volatility, utilising different sampling frequencies and evaluating different financial markets. Because it is well known that long memory characteristics observed in data can be generated by either non-stationary structural breaks or slow regime-switching models, we additionally assess how the results of the analyses change during crisis periods by considering the effects of the US subprime mortgage crunch. The results support the presence of long memory characteristics that vary for diverse types and measurements of volatility, different financial markets, and distinct sampling periods, such as the pre-crisis and crisis periods. This result suggests that empirical investigations must be particularly careful in addressing long memory issues.
Long memory issues arise in many different fields, including hydrology, Internet traffic, economics and finance (Ohanissian et al., 2007). The behaviour and volatility of prices have long been greatly important to financial economists. Understanding the dynamics of stock market volatility poses an extremely puzzling challenge. In this paper, we aim to examine this difficult puzzle by investigating whether real or spurious long memory characteristics of volatility are present in emerging stock market data. The long memory issue is intriguing because of its importance for capital market theories. In particular, the presence of long memory in stock market volatility elucidates the higher-order correlation structure of a financial time series and supports the possibility of predicting the behaviour of these series in a market setting. The analysis of long memory in stock market volatility is important for practitioners, investors, academicians, financial institutions, and policy makers because its presence can have significant implications for risk management, portfolio selection and trading strategies (Kumar, 2012). From the literature, we know that long memory characteristics may differ in contexts involving various types of volatility measurements (such as squared returns, absolute returns, the log of squared returns, realised volatility (RV) and different types of RV), different types of financial markets (such as spot and future markets), and the presence of one or more structural breaks (Ding et al., 1993 and Leccadito and Urga, 2009).1 Thus, we distinguish between true and spurious long memory characteristics by analysing different types of volatility measurements and different types of financial markets. Moreover, we also examine how long memory characteristics vary during crisis periods. On the whole, this paper aims to make the following contributions to the existing literature by testing for the presence of long memory characteristics of volatility in Turkey. First, this study tests long memory characteristics by considering level shifts; in particular, it uses the recent Modified GPH test developed by Smith (2005) rather than methods such as the Hurst R/S, modified R/S and GPH tests, which have demonstrable shortcomings. Second, this investigation examines volatility measures that are based on low- and high-frequency data to compare the effects of different sampling frequencies and different types of volatility measurements on long memory characteristics. Finally, we test how the long memory characteristics of volatility change during crisis periods by considering the 2007–2009 US subprime mortgage meltdown. The results of this study demonstrate that the existence of long memory characteristics differs according to the types and measurements of volatility, the different sub-samples, and the different financial markets that are examined. This paper is organised as follows: Section 2 provides a literature review; Section 3 explains the data and research methods of the study; Section 4 emphasises the limitations of the research; Section 5 presents the empirical evidence of this investigation; and Section 6 consists of the summary and conclusion of the paper.
نتیجه گیری انگلیسی
In the literature, it is commonly believed that volatility is characterised by long memory processes. Given these findings, this paper aims to investigate whether real or spurious long memory characteristics of volatility are present in stock market data. We empirically distinguish between true and spurious long memory characteristics by analysing different types and measurements of volatility, utilising different sampling frequencies and examining different financial markets. In addition, we test how the study results change during the US subprime mortgage crisis. The results with respect to the presence of long memory characteristics vary depending on the volatility measurements, the different financial markets, and the different sampling periods that are examined. This variation suggests that empirical work needs to be particularly careful in addressing long memory issues. The findings of the paper have important implications for practitioners, investors, academicians, financial institutions and policy makers who are interested in volatility in financial markets because the long memory characteristics of volatility are crucial to the selection of the methodology used to forecast volatility. There appears to be many useful issues for future research. Thus far, we compare whether the long memory is true or spurious for volatility utilising low- and high-frequency data based volatility measure. However, we use only realised volatility as a proxy for high frequency based volatility measure. We leave the other measures such as kernel based estimator (Hansen and Lunde, 2006), two time scale estimator (Zhang et al., 2005) or range‐based estimator (Christensen et al., 2010) for future researches for extensive robustness checks.