The VIX market volatility index and U.S. stock index returns.

Sarwar, Ghulam Pub Date:12/01/2011 Name: Journal of International Business and Economics Issue Date: Dec, 2011 Source Volume: 11 Source Issue: 4 1.

INTRODUCTION

The market volatility index (VIX) of the Chicago Board Options Exchange (CBOE) has been referred to as the investor fear gauge since high levels of VIX have coincided with high degrees of market turmoil (Whaley, 2000). VIX has been a popular indicator of short-term market volatility as it represents a market consensus estimate of future (30 days) stock market volatility (Fleming, Ostdiek, and Whaley, 1995; Banerjee, Doran, and Peterson, 2007). When CBOE first introduced VIX in 1993, the index was constructed based on the volatilities implied in options on the S&P 100 index. In 2003, CBOE revamped the definition and calculation of VIX. The new VIX uses the options on the S&P 500 index and reflects a much larger range of options across strike prices (CBOE, 2009; Carr and Wu, 2006). The CBOE renamed the old VIX as VXO and calculated the new VIX back to 1990. The CBOE started trading futures on VIX in March 2004. Trading of options on VIX started in February 2006. The VIX value is computed and disseminated to the investors on a real-time basis just like the information on major stock market indexes. The real-time VIX value offers portfolio managers, hedge fund managers, and other investors an up-to-the-minute estimate of future one-month stock market volatility in order to make informed asset allocation and portfolio insurance decisions. VIX options and futures afford investors viable tools to hedge market volatility. Some investigators advocate the use of VIX as a stock market timing or options trading timing tool (Copeland and Copeland, 1999; Arak and Majid , 2006). Much of the empirical research in the extant VIX literature is based on VXO (old VIX) than on VIX (new VIX). Giot (2005) examines if high levels of VXO indicate oversold markets and finds that future returns are always positive (negative) after very high (low) levels of VXO. Copeland and Copeland (1999) examine VXO as a market timing tool to shift between the style and size portfolios, and report that large and value stocks earn high returns following high VXO values. Arak and Majid (2006) conclude that selling options when VXO levels are high provide potential profits. Dash and Moran (2005) show that VXO is negatively correlated to hedge fund returns and this correlation profile is asymmetric. Guo and Whitelaw (2006) find that market returns are positively related to implied volatilities. Blair, Poon, and Taylor (2001) demonstrate that VXO contains nearly all relevant information about the future realized volatility of index returns for both in-sample and out-of-sample forecasting. Fleming, Ostdiek, and Whaley (1995) find a large negative contemporaneous correlation between VXO changes and S&P 100 index returns, suggesting an inverse relation between VXO and stock market prices. Guo and Wohar (2006) find significant structural breaks (regime shifts) in the mean levels of VXO and VIX at three distinct periods: pre-1992, 1992-1997, and post-1997. Hong and Stein (1999) and Daniel, Hirshleifer, and Subrahmanyam (1998) find that stock market returns may behave differently depending on the state of the market. This paper contributes to the market volatility literature by focusing on VIX than on VXO. We

extend prior studies in three important ways. First, we study the statistical properties of VIX and the intertemporal relationships between VIX and stock market returns from 1992 to 2009. Our study period includes the time period since February 2006 when both futures and options on VIX were actively traded on CBOE. None of the previous studies examined relationships between VIX and stock market returns when either VIX futures or both VIX futures and options were available as viable instruments for hedging market volatility. Our study is similar in spirit to Fleming, Ostdiek, and Whaley (1995) and Whaley (2000) who examined VXO and S&P 100 returns until 1992. Second, following the finding of significant structural breaks in the mean levels of VIX by Guo and Wohar (2006), we separate our analysis of VIX and stock market returns into the 1992-1997 and 1998-2009 subperiods. In addition, we examine the relationships between VIX and stock market returns separately from March 2004 to February 2009 when VIX futures (and VIX options from February 2006) were available to hedge market volatility. Finally, we examine the intertemporal relationships between VIX and S&P 500 index, VIX and S&P 100 index, and VIX and S&P 600 small index returns. No previous study analyzed the relationship between VIX (or VXO) and small stock index returns. The rest of the paper is organized as follows. Section 2 discusses the major differences between VIX and VXO (old VIX) market volatility indexes. Section 3 describes the methods for examining the intertemporal relationships between VIX and stock market indexes and the sources of data. Section 4 discusses the results. The summary and conclusions are provided in the final section 5. 2. CBOE VIX AND VXO MARKET VOLATILITY INDEXES: The VXO (old VIX) is based on the implied volatilities of eight near-the-money S& P 100 options estimated from the Black-Scholes (1973) and Merton (1973b) option pricing models (CBOE, 2009). The cash-dividend adjusted binomial method is used to account for the American-style feature of the S&P 100 options (Fleming, Ostdiek, and Whaley, 1995). At the two nearest maturities, the CBOE selects two call and two put options for each maturity at the two strike prices that straddle the underlying spot price (Carr and Wu, 2006). The implied volatilities from these options are then interpolated to obtain one-month (22 trading days) at-themoney implied volatilities. Thus, VXO is an average of 30-day at-the-money Black-Scholes implied volatilities. Carr and Wu suggest that VXO approximates the volatility swap rate with a 30-day maturity. However, a replicating portfolio of the volatility swaps is difficult to construct. The VIX (new VIX) volatility index is based on market prices of the S&P 500 index options at the two nearest maturities. The calculation of VIX involves a much larger set of call and put options at many different strike prices and is not dependent on any particular option pricing model. Thus, VIX captures information from the whole volatility skew rather than the volatility implied by at-the-money options (CBOE, 2009). As such, VIX measures the market expectation of 30-day volatility conveyed by S&P 500 index option prices (CBOE, 2009). Since S&P 500 index is widely viewed as a proxy for the overall US stock market, the VIX represents a consensus estimate of the future short-term market volatility. Carr and Wu (2006) argue that the VIX squared approximates a 30-day variance swap rate. Because variance swap contracts on major equity indexes are actively traded in over the counter market, a replicating portfolio of the VIX squared payoffs is not difficult to construct. This ease of creating a replicating portfolio may have contributed to the CBOE decision to launch VIX futures and options in March 2004 and February 2006, respectively, after introducing VIX in September 2003.

3. RESEARCH METHODS AND DATA We first examine the statistical properties of the VIX and stock index return series and estimate the autocorrelation coefficients and cross correlations between VIX and stock market returns. Given the information on cross correlations and following Fleming, Ostdek, and Whaley (1995), we then use the following multivariate regression to study the intertemporal relationships between VIX changes and stock market returns: (1) [DELTA][V.sub.t] = [alpha] + [summation over (i=j)] [[beta].sub.s,i] [R.sub.s,t+i] + [[beta].sub.[absolute value of s]] [absolute value of [R.sub.s,t]] + [[epsilon].sub.t] where [DELTA][V.sub.t] is the change in VIX at time t, [R.sub.st+i] is the stock index return at time t+i, [absolute value of [R.sub.s,t]] is the absolute stock index return at time t, [[beta].sub.s,i] is the regression coefficient of the relation between [R.sub.s,t+i] and [beta][V.sub.t], [[beta].sub.[absolute value of s]] is the regression coefficient for [absolute value of [R.sub.s,t]], [alpha] is the regression intercept, and Et is an error term. Since Schwert (1989, 1990) and Fleming, Ostdek, and Whaley (1995) find asymmetric contemporaneous relationship between stock market returns and expected volatility, the coefficients [[beta].sub.s,i] and [[beta].sub.[absolute valeu of s]] jointly are designed to measure the asymmetric relation between [R.sub.s,t] and [DELTA][V.sub.t]. Because VIX is known as the investor fear gauge, we expect the regression coefficient [[beta].sub.s,0] for the contemporaneous relation between change in VIX and stock index returns to have a negative sign. The CAPM models of Merton (1973a), Sharpe (1964), and Lintner (1965) predict that stock prices will fall as expected volatility rises, a prediction consistent with negative [[beta].sub.s,0] value. The discounted cash flow medel of stock valuation would suggest that, assuming expected cash flows are unaffected , a rise in expected market volatility will raise the discount rate for the cash flows which will lead to lower stock prices. Black (1976) and Christie (1982) suggest an inverse relationship between stock prices (or stock returns) and changes in future volatility because a drop in stock prices increases leverage which causes expected volatility to rise. Indeed, Fleming, Ostdek, and Whaley (1995) and Banerjee, Doran, and Peterson (2007) report a strong negative association between stock market returns and changes in future volatility. Giot (2005) suggests that high levels of VIX indicate oversold stock markets which lead to a positive relationship between volatility changes and future stock market returns. Guo and Whitelaw (2006), French, Schwert, and Stambaugh (1987), and Fleming, Ostdek, and Whaley (1995) also report positive relationship between changes in market volatility and future stock returns. Guo and Wohar (2006) find that VIX is mean reverting which suggests that a strong negative contemporaneous relation between VIX changes and stock market returns may be followed by a positive relation between VIX changes and past stock market returns. Given these results, we expect the regression coefficients [[beta].sub.s,i], which capture the relation between VIX changes and lead and lagged stock market returns, to have positive signs. The coefficient [[beta].sub.[absolute value of s]] captures the effect of the size of stock return movements, regardless of the direction of returns, on the contemporaneous VIX changes. Schwert (1990) and Fleming, Ostdek, and Whaley (1995) report an asymmetric relation between market volatility

and stock market returns, implying that an increase in market volatility from a negative stock market returns is expected to be much larger than the decrease in market volatility from a similar positive stock market returns. Hence, we expect that the joint coefficient ([[beta].sub.s,0] [[beta].sub.[absolute value of s]]), which represents the effect of negative returns on VIX, will be significantly larger in magnitude than the joint coefficient ([[beta].sub.s,0] + [[beta].sub.[absolute value of s]]), which reflects the effect of positive returns on VIX. Guo and Wohar ( 2006) find that mean levels of VIX shifts over time. Fleming, Ostdek, and Whaley (1995) and Carr and Wu (2006) show significant autocorrelations in stock market returns and VIX changes. To account for possible heteroskedasticity and autocorrelation problems in estimated parameters and standard errors, we estimate equation (1) using Hansen's (1982) method of moments estimator. The data for this study for the two market volatility indexes (VIX and VXO) and three stock market indexes (S&P 100, S&P 500, and S&P 600 small) came from the CBOE web site. The data cover the time period from January 2, 1992 to February 27, 2009 and have a total of 4324 daily observations. (1) The daily data represent the closing values of VIX, VXO, and stock market indexes. Guo and Wohar (2006) find that VIX has three distinct regimes: pre-1992, 19921997, and post-1997. To conform to these regimes (structural shifts) of VIX, we divide our data into 1992-1997 and 1998-2009 subperiods. In addition, we propose a third subperiod from March 26, 2004 to February 27, 2009 to capture the time period when VIX futures and VIX options (since February 24, 2006) were available to hedge stock market volatility. It is possible that the trading opportunities in VIX futures and options might affect the behavior of VIX. The subperiod 2004-2009 also includes the year 2008 when stock market experienced a severe contraction and VIX reached its highest value since 1990. 4. RESULTS Table 1 presents the summary statistics of daily VIX, VXO, and S&P 100, 500, and 600 returns for the entire sample period 1992-2009 and two subperiods 1992-1997 and 1998-2009. As expected, VIX has slightly lower mean and standard deviation and narrower range than VXO during 1992-2009. VIX ranges from a minimum of 9.31% to a maximum of 80.86%, its highest value on November 2, 2008. (2) Also, as expected, the mean and standard deviation of S&P 600 returns are larger than those of S&P 100 and 500 returns. The summary statistics are more revealing when examined by the two subperiods. The mean VIX rose by 42% from 15.54 during 1992-1997 to 22.11 during 1998-2009, while the standard deviation of VIX (volatility of volatility) jumped by 128% from 4.07 to 9.31 during the same subperiods. The range of VIX values was 70.97 percentage points during 1998-2009, compared to a VIX range of only 28.89 percentage points during 1992-1997. These VIX summary statistics corroborate with the findings of Guo and Wohar (2006) that VIX experienced a significant structural shift (regime change) from the 1992-1997 period to the post-1997 period. Like VIX, the mean and standard deviation of stock index returns are also substantially different between 1992-1997 and 1998-2009. The mean daily returns for both S&P 100 and 500 indexes were negative 0.009% during 1998-2009, a period often referred to as the lost decade. Not only the mean daily returns were positive for the S&P 100, 500, and 600 indexes during 1992-1997, but also the standard deviations of these stock index returns were 80% lower during this period relative to the period 1998-2009.

Table 2 presents the mean and autocorrelation structure of changes (innovations) in market volatility index and stock index returns to examine possible linkages between changes in VIX (investor fear gauge) and U.S. stock market returns. The mean daily change in VIX rose from 0.004% during 1992-97 to 0.007% during 1998-2009, a daily increase of 75%. The exact opposite change occurred to the mean daily returns of S&P 100 and 500 indexes which both fell from 0.056% during 1992-97 to -0.009% during 1998-2009, a daily drop of 116%. For the entire period 1992-2009, the autocorrelations of VIX changes and stock market returns, except for S&P 600, at the first two lags are negative and statistically significant at the 5% level. However, the autocorrelation structure of VIX changes and stock market returns differ substantially between the subperiods 1992-97 and 1998-2009. While autocorrelations of S&P 100 and 500 returns at the first two lags are statistically significant during 1998-2009, none of these return autocorrelations are significant during 1992-97. Similarly, the autocorrelations of VIX changes are significant at the first two lags during 1992-97, but only the VIX autocorrelation at lag one is significant during 1998-2009. The autocorrelations of S&P 600 returns also differ significantly between 1992-97 and 1998-2009. Table 3 presents cross correlations between the current VIX change and past, current, and future stock market returns. The contemporaneous correlations between changes in VIX and S&P 100, 500, and 600 returns were negative and significant for the entire period 1992-2009 and subperiods 1992-1997 and 1998-2009. The contemporaneous correlations between VIX changes and S&P 100 and 500 returns were larger during 1998-2009 than during 1992-1997 (-0.81 versus -0.61). These negative contemporaneous correlations are consistent with the results of Black (1976), Christie (1992), and Fleming, Ostdiek, and Whaley (1995), and they indicate that stock prices and returns fall (rise) when expected market volatility rises (falls). However, correlations between current VIX changes and stock index returns at lag and lead 2 reverse signs to positive and drop to modest levels, but they are still statistically significant at the 5% level. These positive correlations between VIX changes and lagged and lead stock market returns are consistent with the mean reverting feature of VIX and with the results of Banerjee, Doran, and Peterson (2007), French, Schwert, and Stambaugh (1987), and Fleming, Ostdiek, and Whaley (1995). A significant positive correlation also exists between current VIX change and one-period lagged and lead S&P 100 and 500 returns during 1998-2009. But this correlation at lag and lead one was not statistically significant during 1992-97. Thus, similar to autocorrelations of VIX changes and stock market returns, the non-contemporaneous cross correlations of VIX changes and stock market returns are quite different between the subperiods 1992-97 and 1998-2009. These results corroborate with the findings of Guo and Wohar (2006) that VIX experienced a significant structural shift from 1992-97 to post-1997 period. The results of multivariate regressions that examine the intertemporal relationshsips between expected stock market volatility and stock market returns are presented in Table 4.3 The regression results show a strong negative contemporaneous relationship between the stock market returns and changes in VIX. This negative contemporaneous association holds for all three stock market indexes (S&P 100, 500, and 600) and for the entire sample period and three subperiods. These results for the contemporaneous effect are consistent with the results of Fleming, Ostdiek, and Whaley (1995) and Whaley (2000) who examined the relationship between S&P 100 index and expected market volatility index VXN until 1992. Our study

demonstrates that the strong negative association between stock market returns and expected market volatility extends to the small stock market returns as reflected in the S&P 600 small stock index. For the entire period 1992-2009, one percent change in the S&P 500 return is accompanied on average by a contemporaneous 0.99 percent change in VIX in the opposite direction. The size of the contemporaneous coefficient and regression [R.sup.2] indicate that the negative relationship is stronger between VIX and S&P 500 return than between VIX and S&P 100 (or S&P 600) return regardless of the time period of the analysis. These results are consistent with the cross correlation analysis for the three indexes in Table 3. The coefficient of the contemporaneous absolute stock market return in Table 4 is positive and significant for the three stock market indexes over the entire period 1992-2009 and subperiod 1992-97. The significant [[beta].sub.[absolute value of s]] suggests that the size of the stock market move, regardless of its direction, has a significant positive relationship with the contemporaneous expected change in stock market volatility. The coefficient [[beta].sub.[absolute value of s]] plus (minus) [[beta].sub.s,0] yields the effects of positive (negative) changes in market returns on VIX. The magnitude of positive return coefficient, [[beta].sub.s+], is much smaller than that of negative return coefficient, [[beta].sub.S-], during 1992-2009 and 1992-1997, indicating an asymmetric relationship between stock market returns and VIX. This asymmetric relationship is much stronger during 1992-97 than during 1992-2009. However, surprisingly, the asymmetric relationship between stock market returns and VIX disappears during 1998-2009 and 2004-2009 as the absolute return coefficient [[beta].sub.[absolute value of s]] is not significantly different from zero during these periods, except for S&P 600 index during 1998-2009. Interestingly, as shown in Table 1, the subperiod 1998-2009 experienced 48% higher mean and 128% higher volatility for VIX than the subperiod 1992-97. The results of Tables 4 and 1 jointly suggest that the asymmetric relationship between market returns and VIX depends on the mean and volatility regime of VIX. The asymmetric VIX-returns relation gets stronger when VIX is lower and less volatile, but this relation disappears when VIX is higher and more volatile. The positive return coefficient [[beta].sub.s+] (or [[beta].sub.[absolute value of s]] plus [[beta].sub.s,0]) is -0.89 for the period 1992-2009, which suggests that one percent increase in the S&P 500 return is accompanied by a 0.89 percent drop in VIX. However, the negative return coefficient [[beta].sub.S-] (or [[beta].sub.[absolute value of s]] minus [[beta].sub.s,0]) is -1.09, which indicates that a drop of one percent in the S&P 500 return is associated with an increase of 1.09 percent in VIX. The difference in the effects of rises and falls in stock market returns on VIX is much larger for the subperiod 1992-1997 which experienced low mean and volatility for VIX. During this subperiod, one percent increase in the S&P 500 return was associated with a 0.43% drop in VIX, while one percent drop in the S&P 500 return was accompanied by an increase of 1.11% in VIX. A difference of at least two times in magnitudes of [[beta].sub.s+] and [[beta].sub.S-] estimates was also observed for the S&P 100 and 600 returns, and it clearly indicates asymmetric relation between stock market returns and VIX. The expected market volatility index (VIX) responds much more aggressively to negative changes in stock market return that it responds to positive changes in returns of similar size. This result is consistent with the popular notion that VIX is more of a gauge of investor fear than investor positive sentiment. In Table 4, the regression coefficients at lags one and two and at lead 2 are positive and

significant over the entire period 1992-2009 for the S&P 100, 500, and 600 returns. (3) Similarly, both lag 1 and lead 2 coefficients were positive and significant at the 5% level for S&P 100 and 500 returns during 1992-1997. However, only lead 2 coefficient was significant and positive during 1992-97 for S&P 100 and 500 returns. But all of these positive and significant non-contemporaneous relationships between stock market returns and VIX were much smaller in magnitude (average coefficient 0.069) than the negative and significant contemporaneous relations between stock market returns and VIX (average coefficient 0.93). Nonetheless, the positive non-contemporaneous relation between stock market returns and VIX is consistent with the results of French, Schwert, and Stambaugh (1987), Fleming, Ostdiek, and Whaley (1995), and Whaley (2000). This positive association between stock market returns and VIX also lends support to the mean reversion feature of VIX as documented by Guo and Wohar (2006). The regression results in Table 4 are more revealing when separated by the three subperiods 1992-97, 1998-2009, and 2004-2009. The negative contemporaneous relationship between VIX and stock market returns grows much stronger in magnitude and statistical significance for all three stock market indexes as we shift analysis from the subperiod 1992-1997 to 1998-2009 and then to 2004-2009. The subperiod 2004-2009 reflects the time period when VIX futures and options were available to investors for hedging stock market return volatility. For the S&P 500 returns, the coefficient of contemporaneous relation between stock market returns and VIX increases in magnitude from -0.77 during 1992-97 to -1.02 during 1998-2009 and then to -1.17 during 2004-09. Similarly, the contemporaneous return-VIX coefficient for the S&P 100 returns is -0.72, -0.99, and -1.20 during 1992-97, 1998-2009, and 2004-09, respectively. A difference in contemporaneous return-VIX coefficients of similar magnitude over the three subperiods was found for S&P 600 returns. These results strongly suggest that the negative relationship between stock market returns and VIX has changed substantially over time, and they lend support to Guo and Wohar (2006) who reported a structural shift in VIX from 1992-97 to post-1997 period. One important observation for the 2004-09 period is that the non-contemporaneous relationships between VIX and S&P 100 and 500 returns at lags and leads one and two do not exist. (4) In addition, the relationship between VIX and absolute changes in stock market returns (i.e., [[beta].sub.[absolute value of s]] estimate) is no longer significant during this period. The only significant VIX-returns relation that remained during 2004-2009 is the negative contemporaneous relationship between stock market returns and VIX. During the latest five year period, one percentage point change in S&P 500 and 100 returns triggers a 1.17 and 1.20 percentage point change in VIX, respectively, in the opposite direction. The domination of negative contemporaneous relationship between stock market returns and VIX and the absence of non-contemporaneous effects between VIX and returns during 2004-2009 may be related to the excessively high mean and volatility of VIX and the existence of VIX futures and options as viable tools for hedging market volatility during this period. 5. SUMMARY AND CONCLUSIONS The CBOE market volatility index (VIX) is often referred to as the investor fear gauge. Most prior studies examine the time series behavior of VXO (old VIX) and its relation with different equity portfolio returns. This study examines the univariate time-series properties of VIX and the

intertemporal relationships between VIX and returns of the S&P 100, 500, and 600 indexes for the period 1992-2009. Because previous studies show evidence of structural shifts in VIX in the pre- and post-1997 periods, we conduct our analysis separately for 1992-1997 and 1998-2009 subperiods. In addition, we perform the VIX-returns analysis for the subperiod 2004-2009 during which VIX futures and options were available for hedging stock market volatility. We examine whether the relation between stock market returns and VIX has significantly changed over time. Our results from cross correlation analysis indicate a strong negative contemporaneous relationship between VIX and stock market returns. This relationship is much stronger in 19982009 than in 1992-97 and is followed and preceded by positive non-contemporaneous relations between VIX and stock market returns. The regression analyses suggest a strong negative contemporaneous relation between daily changes (innovations) in VIX and S&P 100, 500, and 600 returns during the three subperiods. This negative contemporaneous relationship grows much stronger in magnitude and statistical significance as we shift analysis from 1992-1997 to 1998-2009 and then to 2004-2009. The results suggest that the strength of contemporaneous VIX-returns relationship depends on the mean and volatility regime of VIX. The strongest concurrent response of VIX to S&P 100 and 500 returns was -1.20 during the latest five-year period when VIX was both high and highly volatile. Our results also indicate a strong asymmetric relation between daily stock market returns and innovations in VIX during 1992-2009 and 1992-97. The VIX responds much more aggressively to negative changes in stock market returns that it responds to positive changes in returns of similar size, suggesting that VIX is more of a gauge of investor fear than investor positive sentiment. However, unlike the contemporaneous VIX-returns relation which gets stronger when VIX is high and more volatile, the asymmetric relationship between stock market returns and changes in VIX is much weaker or nonexistent when VIX is large and more volatile. In fact, during 2004-2009, the contemporaneous negative relationship between stock market returns and changes in VIX was the most dominating and the only significant relationship among the contemporaneous, asymmetric, and non-contemporaneous relationships between stock market returns and VIX. REFERENCES: Arak, M. and Mijid, N., "The VIX and VXN Volatility Measures: Fear Gauges or Forecasts", Derivatives Use, Trading & Regulation, Vol. 12, 2006,14-27. Banerjee, P. S., Doran, J. S., and Peterson, D. R., "Implied Volatility and Future Portfolio Returns", Journal of Banking and Finance, 31, 2007, 3183-3199. Black, F., "Studies of Stock Price Volatility Changes", Proceedings of the 1976 Meetings of American Statistical Association, Business and Economics Section, 1976, 177-181. Black, F. and Scholes, M., "The Pricing of Options and Corporate Liabilities", Journal of Political Economy, 81, 1973, 637-654. Blair, B. J., Poon, S-H., Taylor, S. J., "Forecasting S&P 100 Volatility: The Incremental

Information Content of Implied Volatilities and High-Frequency Index Returns", Journal of Econometrics, 105, 2001, 5-17. Carr, P. and Wu., L., "A Tale of Two Indices", The Journal of Derivatives, 13, 2006, 13-29. Chicago Board Options Exchange, 2009, http://www.cboe.com. Christie, A. A., "The Stochastic Behavior of Common Stock Variances: Value, Leverage, and Interest Rate Effects", Journal of Financial Economics, 10, 1982, 407-432. Copeland, M. and Copeland, T., "Market Timing: Style and Size Rotation Using the VIX", Financial Analysts Journal, 55, 1999, 73-81. Dash, S. and Moran, M. T., "VIX as a Companion of Hedge Fund Portfolios", The Journal of Alternative Investments, 8, 2005, 75-82. Daniel, K., Hirshleifer, D., and Subrahmanyam, A., "Investor Psycology and Investor Security Market Under- and Over-reaction", Journal of Finance, 53, 1998, 189-209. Fleming, J., Ostdiek, B., Whaley, R.E., "Predicting Stock Market Volatility: A New Measure", Journal of Futures Markets, 15, 1995, 265-302. French, K. R., Schwert, G.W., and Stambaugh, R. F., "Expected Stock Returns and Volatility", Journal of Financial Economics, 19, 1987, 3-29. Giot, P., "Relationships between Implied Volatility Indexes and Stock Index Returns", Journal of Portfolio Management, 26, 2005, 12-17. Ghosh, P. R., "VIX's Actions Trick Market Traditions", Wall Street Journal, July 17, 2009. Guo, H., and Whitelaw, R., "Uncovering the Risk-Neutral Relationship in the Stock Market", Journal of Finance, 61, 2006, 1433-1463. Guo, W. and Wohar, M. E., "Identifying Regime Changes in Market Volatility", The Journal of Financial Research, 29, 2006, 79-93. Hansen, L. P., "Large Sample Properties of Generalized Method of Moment Estimators", Econometrica, 50, 1982,1029-1054. Hong, H., and Stein, J., "A Unified Theory of Underreaction, Momemtum Trading, and Overeaction in Asset Markets", Journal of Finance, 54, 1999, 2143-2184. Merton, R. C., "An Intertemporal Capital Asset Pricing Model", Econometrica, 41, 1973a, 867888. Merton, R. C., "Theory of Rational Option Pricing", Bell Journal of Economics and Management Science, 4, 1973b, 141-183.

Schwert, G. W., "Why Does Stock Market Volatility Change Over Time?", Journal of Finance, 44, 1989,1115-1154 Schwert, G. W., "Stock Volatility and Crash of '87'", Review of Financial Studies, 3, 1990, 77102. Whaley, R. E., "The Investor Fear Gauge", Journal of Portfolio Management, 26, 2000, 12-17. Ghulam Sarwar, California State University, San Bernardino, California, USA ENDNOTES: (1.) This was the latest data on VIX and three stock market indexes in the master file of CBOE archives at the time this paper was written. (2.) A VIX value of more than 30% is considered to be high and outside of the normal range (Whaley, 2000). Practitioners often think that a VIX below 30 means that stock market is relatively stable, while a VIX above 30 reflects a sense of panic or capitulation (WSJ, 2009). (3.) Coefficients of stock index returns beyond lag and lead 2 were not statistically significant at the 5% level in nearly all of the cases. Similarly, coefficients of lead and lag absolute returns were predominantly statistically insignificant. Hence, these coefficients are not included in the regression analysis presented in Table 4. (4.) We also conducted the regression analysis for the subperiod February 24, 2006 to February 27, 2009 when VIX options were traded in addition to VIX futures which started trading on March 26, 2004. The regression analyses for this subperiod (not reported here) are virtually identical to those shown in Table 4 for March 26, 2004 to February 27, 2009 (i.e., 2004-2009).
TABLE 1. SUMMARY STATISTICS OF DAILY VOLATILITY AND INDEX RETURNS Variable Mean (%) Maximum (%) Std Dev (%) Minimum 1992-2009 (N = 4324) VIX 80.86 VXO S&P 500 Return S&P 100 Return S&P 600 Small Ret. 87.24 0.013 0.014 0.024 20.67 1.17 1.19 1.26 9.43 -9.46 -9.18 -11.62 1992-1997 (N = 1518) VIX 15.54 4.07 9.31 9.04 10.95 10.65 8.11 19.80 8.47 9.31 (%)

38.20 VXO S&P 500 Return S&P 100 Return S&P 600 Small Ret. 39.96 0.055 0.057 0.061 15.91 0.72 0.76 0.71 4.54 -7.11 -7.09 -6.33 1998-2009 (N = 2806) VIX 80.86 VXO S&P 500 Return S&P 100 Return S&P 600 Small Ret. 87.24 -0.009 -0.009 0.004 23.25 1.35 1.37 1.47 10.35 -9.46 -9.18 -11.62 9.04 10.95 10.65 8.11 22.11 9.31 9.89 9.04 4.98 5.60 3.36

Notes: The daily data reflect the closing values of the variables and came from the Chicago Board Options Exchange (CBOE). The choice of subperiods 1992-1997 and 1998-2009 was based on Guo and Wohar (2006) who showed significant structural changes in VIX and VXO between the two subperiods. VIX and VXO are the CBOE volatility indexes based on the S&P 500 index options and S&P 100 index options, respectively. S&P 600 Small Ret is the daily return on the S&P 600 small index. TABLE 2. STATISTICAL PROPERTIES OF VOLATILITY INDEX CHANGES AND STOCK INDEX RETURNS Variable Mean (%) [rho] (1) [rho] (2) 1992-2009 VIX S&P S&P S&P Change 500 Return 100 Return 600 Small Ret. 0.006 0.013 0.014 0.024 -0.098 * -0.063 * -0.072 * 0.011 -0.111 -0.069 -0.067 -0.036 * * * * 0.004 0.029 0.025 0.046 * [rho] (3)

1992-1997 VIX S&P S&P S&P Change 500 Return 100 Return 600 Small Ret. 0.004 0.055 0.057 0.061 -0.048 * 0.012 -0.012 0.234 * -0.040 -0.034 -0.032 -0.018 1998-2009 VIX S&P S&P S&P Change 500 Return 100 Return 600 Small Ret. 0.007 -0.009 -0.009 0.004 -0.105 * -0.075 * -0.083 * -0.017 -0.122 -0.075 -0.074 -0.039 * * * * 0.011 0.040 0.037 0.051 * -0.044 -0.050 -0.051 0.006

Notes: The autocorrelations [rho] (i) are shown for the first three lags. An asterisk '*' indicates that autocorrelations are significant at the 5% level. VIX is the CBOE market volatility index based on the S&P 500 index options.

TABLE 3. CROSS CORRELATIONS BETWEEN VOLATILITY INDEX CHANGES AND STOCK INDEX RETURNS Correlation Pair SPX (-2), VIXC (0) SPX (-1), VIXC (0) SPX (0), VIXC (0) SPX (+1), VIXC (0) SPX (+2), VIXC (0) OEX (-2), VIXC (0) OEX (-1), VIXC (0) OEX (0), VIXC (0) OEX (+1), VIXC (0) OEX (+2), VIXC (0) SP600 (-2), VIXC (0) SP600 (-1), VIXC (0) SP600 (0), VIXC (0) SP600 (+1), VIXC (0) SP600 (+2), VIXC (0) 1992-2009 0.089 * 0.083 * -0.801 * 0.051 0.092 * 0.088 * 0.083 * -0.78 * 0.056 0.090 * 0.081 * 0.044 -0.71 * 0.016 0.065 * 1992-1997 0.096 * 0.010 -0.62 * -0.031 -0.014 0.087 * 0.004 -0.61 * -0.012 -0.013 0.109 * 0.045 -0.56 * -0.15 * 0.022 1998-2009 0.088 0.094 -0.82 0.063 0.108 0.088 0.095 -0.81 0.067 0.106 0.077 0.045 -0.73 0.039 0.071 * * * * * * * * * * * * *

Notes: An asterisk '*' indicates that correlations are significant at the 5% level. VIXC (0) is the change in CBOE volatility index VIX at lag 0. SPX (i), OEX (i), and SP600 (i) are the returns on the S&P 500 index, S&P 100 index, and S&P 600 small index at lag i or lead i, respectively. TABLE 4. INTERTEMPORAL RELATIONSHIP BETWEEN VIX CHANGES AND STOCK INDEX RETURNS Intercept Adj [[beta].sub. [absolute value of s]] [[beta].sub. s,-2] [[beta].sub. s,-1]

Period

[R.sup.2]

Panel A. S&P 500 Index -0.058 0.097 (2.63) -0.13 0.34 (5.78) -0.066 0.072 (1.71) -0.023 0.015 (0.27) 0.64 (2.27) 0.43 (4.64) 0.69 (1.96) 0.74 (0.54) 0.056 (2.37) 0.111 (3.45) 0.045 (1.65) -0.019 (0.39) 0.053 (2.53) 0.051 (1.62) 0.048 (2.09) 0.029 (0.76)

1992-2009

1992-1997

1998-2009

2004-2009

Panel B. S&P 100 Index -0.067

1992-2009

0.010 (2.81) -0.13 0.33 (5.68) -0.074 0.079 (1.85) -0.038 0.029 (0.44)

0.63 (2.51) 0.41 (4.54) 0.67 (2.11) 0.73 (0.86)

0.058 (2.44) 0.102 (3.31) 0.047 (1.74) -0.023 (0.45)

0.045 (2.22) 0.023 (0.79) 0.043 (1.93) 0.026 (0.67)

1992-1997

1998-2009

2004-2009

Panel C. S&P 600 Small Index -0.061 0.094 (2.65) -0.099 0.24 (3.29) -0.087 0.093 (2.23) -0.058 0.057 (0.87) [[beta].sub. s,0] Panel A. S&P 500 Index 1992-2009 1992-1997 1998-2009 2004-2009 -0.99 (32.35) -0.77 (20.77) -1.02 (31.37) -1.17 (22.20) Panel B. S&P 100 Index 1992-2009 -0.95 (30.28) 0.001 (0.04) 0.047 (2.19) 0.002 (0.10) -0.045 (1.73) 0.005 (0.20) 0.042 (0.92) 0.048 (2.13) -0.038 (1.47) 0.058 (2.30) 0.079 (1.67) 0.52 (2.16) 0.36 (2.70) 0.55 (2.20) 0.60 (0.95) [[beta].sub. s,+1] 0.073 (3.02) 0.093 (2.55) 0.066 (2.45) 0.040 (0.90) [[beta].sub. s,+2] 0.067 (2.36) 0.23 (6.43) 0.044 (1.42) 0.022 (0.47)

1992-2009

1992-1997

1998-2009

2004-2009 Period

1992-1997 1998-2009 2004-2009

-0.72 (20.75) -0.99 (20.48) -1.20 (21.28) Panel C. S&P 600 Small Index

-0.042 (1.57) 0.003 (0.15) 0.041 (0.89)

-0.035 (1.38) 0.057 (2.34) 0.075 (1.52)

1992-2009 1992-1997 1998-2009 2004-2009

-0.82 (28.26) -0.72 (13.72) -0.84 (26.90) -0.91 (18.31)

0.024 (0.93) -0.027 (0.81) 0.027 (0.96) 0.067 (1.41)

0.045 (2.02) 0.012 (0.43) 0.050 (2.01) 0.086 (2.08)

Notes: Parameters [[beta].sub.s,i] indicate the effect of contemporaneous and non-contemporaneous stock index returns on daily changes in VIX. The parameter [[beta].sub.[absolute value of s]] captures the effect of absolute stock index returns on VIX changes. Absolute t-statistics are given in parentheses, and Adj [R.sup.2] is the adjusted coefficient of determination. The regression was estimated using Hansen's (1982) method of moments.