مدلسازی پویایی مشترک نوسانات شاخص سهام و عملکرد اوراق قرضه خنثی از ریسک
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|22483||2014||13 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Banking & Finance, Volume 38, January 2014, Pages 216–228
This study examines the joint evolution of risk-neutral stock index and bond yield volatilities by using the Chicago Board Option Exchange S&P500 volatility index (VIX) and the Bank of America Merrill Lynch Treasury Option Volatility Estimate Index (MOVE). I use bivariate regime-switching models to investigate the alternation of “high-risk” and “low-risk” markets, where the high-risk regime is characterized by higher and more volatilities with weaker cross-market linkages. Common information about economic and financial conditions appears to drive VIX and MOVE fluctuations between the two risk regimes. Two-regime specifications also distinguish between information spillover and common information effects. Ignoring regime shifts leads to spurious extreme persistence and incomplete inferences about asymmetric volatility. The findings carry important implications for asset allocation.
Volatility is one of the most important determinants of asset value for stocks and bonds, the two most important asset classes. Expectations of future market volatility and their linkages have important implications for asset pricing, portfolio management and hedging effectiveness. If volatility is directly related to the rate of information flow (Ross, 1989), volatility expectations may persist over time due to the gradual incorporation of information (Anderson and Bollerslev, 1997) or incomplete information from traders and subsequent revisions in beliefs after a structural break (Timmerman, 2001). Investors can also react differently to positive and negative information of the same magnitude. Therefore, asymmetric volatility is documented both for stocks (Black, 1976a and Campbell and Hentschel, 1992) and interest rates (Chan et al., 1992). Furthermore, the linkages between the stock and bond markets reflect common information (Ederington and Lee, 1993) and cross-market information spillover effects (Fleming et al., 1998). Typically, researchers estimate volatility from the time series of historical price changes.1 However, such volatility estimates are ex post measures and reflect only part of the impact of information arrival on perceptions of volatility. Information not only causes asset price changes but also induces revisions of investor beliefs about the future volatility of asset prices and macro-economic variables ( Stulz, 1986). Although not directly observable, implied volatility estimates derived from prices of options or other derivatives represent investor beliefs about the underlying asset price volatility ( Patell and Wolfson, 1979). Recently, implied volatility has gained more popularity in the literature and among practitioners. In contrast to ex post physical volatility measures, implied volatility is the ex ante risk-neutral expectation of future volatility and it reflects both immediate and longer-term effects of information flow. Another problem of volatility is its high persistence, a sign of structural change in variance which can be better characterized empirically by regime-switching models. 2 There is substantial time-variation and regime-dependence in the relation between stock and bond returns. Multivariate regime-switching models have become increasingly popular in investigating asset allocation between stock and bonds (Guidolin and Timmermann, 2005, Guidolin and Timmermann, 2006, Guidolin and Timmermann, 2007, Baele et al., 2010, Yang et al., 2010 and Chan et al., 2011). Also, regime-switching models have been used to study asymmetric correlations across asset returns and to draw implications for asset allocation (Ang and Bekaert, 2002a, Ang and Chen, 2002 and Guidolin and Timmermann, 2008). Ang and Timmermann (2012) conduct a good survey of regime switching and financial markets. All these studies have demonstrated that the regime-switching model is better than the single-state model at capturing the joint return distribution. However, very few studies have explored the regime shifts for joint distribution of risk-neutral stock and bond volatilities. My paper fills this gap and makes several contributions to the literature. Firstly, with implied volatility indices for the S&P500 stock index and US Treasury bond yields from 1990 to 2010, I find that bivariate two-state regime-switching models fit the data much better than a single-regime model would, thereby suggesting substantial regime-dependence in the relationship between risk-neutral stock index and bond yield volatilities. These models are particularly appealing for implied volatilities because news about business cycles and financial conditions can simultaneously alter investor expectations both in stock and bond markets, as indicated in Timmermann (2001). In particular, the two regimes in my model can be characterized as “high-risk” and “low-risk” regimes.3 During the high-risk markets, both stock and bond risk neutral volatilities are higher and more volatile. Moreover, these ex ante stock and bond volatilities have a lower correlation in the high-risk regime, which is consistent with the stock–bond return correlation pattern found in Yang et al. (2009). By contrast, the low-risk regime is associated with lower volatility expectations, lower volatility of volatilities and stronger cross-market linkages between ex ante volatilities. Secondly, allowing for regime shifts can empirically distinguish between information spillover and common information effects. I report strong evidence that macro-economic and financial variables commonly used in the literature predict the transition probability of regime switches. Thus, common information about economic and financial conditions, especially the default spread, causes regime shifts in the joint evolution of volatility expectations of stock and bond markets. There is also evidence that VIX and MOVE can predict each other, indicating a bi-directional information spillover effect. At a short (weekly) horizon, higher bond yield volatility tends to follow higher stock volatility in the previous period more significantly in the high-risk regime, suggesting increased information flow from the stock market to the Treasury bond market when investors flight to safety in the bad time. However, such stronger information spillover effect becomes insignificant in a longer (monthly) horizon, implying that flying to safety is a short-term phenomenon. Thirdly, I document additional new evidence on volatility clustering and asymmetry. Volatility expectation forms clusters in each regime, suggesting the gradual incorporation of information. Moreover, high-volatility expectation persists for 4.44 weeks whereas low-volatility expectation persists for 17.54 weeks. Ignoring regime shifts leads to the spurious appearance of extreme persistence. Also, a very significant and robust negative relation between innovations in stock returns and expected stock volatility exists and it is consistent with the asymmetric volatility literature using implied volatilities (for example, Dennis et al., 2006). A notable new finding is that the asymmetric volatility effect is much larger in the high-risk regime. This suggests that non-diversifiable stock market volatility as an asset class4 should be very appealing for stock portfolio diversification, especially in bad times. Moreover, the relation between bond yield implied volatility and the level of the long-term interest rate is regime-dependent, negative in the high-risk regime but positive in the low-risk regime. This adds to the literature on interest rate volatility that typically examines volatility of the short-term interest rate and finds mixed relationships (Trolle and Schwartz, 2009). Finally, this study features two very prominent volatility indicators, the Chicago Board Option Exchange’s S&P500 volatility index (VIX) and the Bank of America Merrill Lynch’s Treasury Option Volatility Estimate Index (MOVE). VIX is widely covered by the financial media, and is even included on the ticker of the CNBC financial news cable television network. Investors view the VIX index as reflecting both fear and the demand for portfolio insurance (Whaley, 2000 and Whaley, 2008) whereas academics find VIX an increasingly useful and interesting measure of the market’s expected future stock index volatility. MOVE is a widely-followed measure of government bond yield volatility.5 MOVE is also included by the IMF in a statistical appendix of Global Financial Stability Reports together with VIX. However, MOVE is seldom studied in the literature, either by itself or in relation to VIX.6 My study fills this gap. The rest of the paper is organized as follows. Section 2 describes the data. Section 3 discusses the regime-switching models and develops testable hypotheses. Section 4 presents the empirical results. Finally, Section 5 gives concluding remarks.
نتیجه گیری انگلیسی
Using implied volatility indices for S&P500 returns and US Treasury bond yields from 1990 to 2010, I examine the joint evolution of risk-neutral stock and bond volatilities. While there are many studies of the Chicago Board Option Exchange S&P500 volatility index (VIX), this study also features the bond market’s equivalent to VIX,17 the Bank of America Merrill Lynch Treasury Option Volatility Estimate Index (MOVE). The bivariate two-state regime-switching models developed in the paper fit the data much better than a single-regime model and capture the alternation of high-risk and low-risk markets, where the high-risk regime is characterized by higher and more volatile expected volatility. Allowing for regimes can also distinguish between the information spillover effect and the common information effect. There is strong evidence that common information about economic and financial conditions plays a significant role in driving the dynamics of regime switches and that the information spillover effect between VIX and MOVE is bi-directional and regime-dependent. Ignoring regime shifts in volatility expectations can lead to spurious extreme persistence and incomplete inferences about asymmetric volatility. Given the overwhelming existence of regimes, we can draw important implications for asset allocation. Clearly, an investor can do better by holding VIX futures, options and/or interest rate volatility instruments in his stock and/or bond portfolio when the high-risk regime prevails. If the investor ignores regimes, he would hold too much equity for too long a period of time and would not be much better off shifting to bonds when the high-risk regime hits. To hedge against the risk of switching to the high-risk regime, even though it occurs relatively rarely, the investor can predict the possibility of a regime shift by closely watching economic and financial indicators and then rebalancing his portfolio accordingly. More implications can be discussed for portfolio diversification and risk management, especially on an out-of-sample basis. Another dimension for future research is to investigate stock and bond ex ante volatilities across countries and to relate associations to international macro-economic and financial conditions. Finally, derivatives exchanges have recently expanded the range of implied volatility indices to oil, gold, exchange rates, and other stock indices, thus suggesting further room for studies of the joint evolution of implied volatility across asset classes.