Volatility contagion across commodity, equity, foreign exchange and ...

4 downloads 79 Views 141KB Size Report
across commodity, equity, foreign exchange and Treasury bond markets ... prices of CBOE exchange-traded fund (ETF) options used to calculate EVZ, GVZ.
Volatility contagion across commodity, equity, foreign exchange and Treasury bond markets

R. López Department of Economic Analysis and Finance, University of Castilla-La Mancha, Albacete, 02071, Spain E-mail: [email protected]

Over the last years, the Chicago Board of Options Exchange (CBOE) has launched a set of implied volatility indices based on new asset classes following the success of equity-based volatility indices. Using some of the newly created volatility indices, this study shows that evidence of implied volatility transmission across commodity, equity, foreign exchange and Treasury bond markets cannot be accounted for by news announcements on economic fundamentals, suggesting volatility contagion. The findings are robust over the recent financial crisis period and the post crisis period.

Acknowledgements Raquel acknowledges the financial support provided by Ministerio de Educación y Ciencia grant ECO2011-28134.

I. Introduction

Implied volatility indices measure the market expectations of volatility of asset prices over a fixed horizon (i.e., they are forward-looking measures of volatility). The term implied refers to the fact that volatility is derived from or implied by the market prices of options on the asset whose volatility is being estimated (see Demeterfi et al., 1999 and references therein). Several studies document the existence of implied volatility spillovers across international equity markets using equity-based volatility indices (see e.g., Nikkinen and Sahlström, 2004; Äijö, 2008; Konstantinidi et al., 2008; Jiang et al., 2012; Siriopoulos and Fassas, 2012). Jiang et al. (2012) further show that volatility spillovers across international equity markets cannot be explained by the prevailing economic conditions as reflected by news releases on macroeconomic fundamentals, indicating the presence of volatility contagion (see e.g., Connolly and Wang, 2003; Bekaert et al., 2005; Boyson et al., 2010). The present study extends the previous finding by providing evidence of volatility contagion across US equity and non-equity markets. To that end, it uses data from five volatility indices disseminated by the Chicago Board of Options Exchange (CBOE) - the foreign exchange volatility index EVZ, the commodity-based volatility indices GVZ and OVX, the equity volatility index VIX and the recently introduced Treasury bond volatility index VXTYN. 1 The results are found to be consistent across the recent financial crisis period and the post crisis period.

1

The CBOE and CME Group launched VXTYN in May 2013.

II. Data

The data set includes the daily closing prices of five volatility indices – EVZ, GVZ, OVX, VIX and VXTYN – and scheduled news announcements on macroeconomic fundamentals. Due to data availability, the sample encompasses the date range 1 June 2008 to 31 March 2013 so as to study the five indices over a common time period. Data for the volatility indices were obtained from the CBOE website. They track the market expectations of volatility of the US dollar/euro exchange rate (EVZ), gold prices (GVZ), crude oil prices (OVX), the S&P 500 (VIX) and 10-year Treasury note futures prices (VXTYN) over a constant 30-day horizon. 2 Closing prices of CBOE exchange-traded fund (ETF) options used to calculate EVZ, GVZ and OVX are measured at 4:00 pm ET, whereas CBOE S&P 500 options and CME Treasury futures options involved in computing VIX and VXTYN cease trading at 4:15 pm ET and 5:00 pm ET, respectively. 3 Hence, the chance that changes in one volatility index contain far more (or less) information than changes in another index is low. This is a key factor in successfully investigating the effect of scheduled news announcements on implied volatility spillovers. We collect data on the dates, release times, actual released figures and survey forecasts from Bloomberg for a set of 14 scheduled news announcement items which comprehensively characterize the US macro economy (see e.g., Ederington and Lee, 1996; Fleming and Remolona, 1997; Balduzzi et al., 2001; Jiang et al., 2012). Table 1 provides a summary of the announcements. We can see that all of

2

Additional information on the volatility indices can be retrieved from http://www.cboe.com/micro/volatility/introduction.aspx 3 Contract specifications of options traded on the CBOE and the CME Group are provided on the websites of the exchanges.

the news releases occur before options involved in the calculation of the volatility indices cease trading. Survey forecasts regarding the values of economic variables conducted by Bloomberg on the Friday of the week before they will be released are employed to estimate the surprise element of the scheduled news announcements.

III. Methodology

We investigate the presence of implied volatility spillovers between commodity, equity, foreign exchange and Treasury bond volatility indices and the impact of news announcements on volatility transmission following the approach of Jiang et al. (2012). Evidence of implied volatility spillovers is examined using the following vector autoregressive VAR(p) model: p

DLVOLt = α + ∑ ϕl DLVOLt −l + ε t (1) l =1

where DLVOLt = ln(VOLt / VOLt −1 ) is a (5 x 1) vector of daily log changes in EVZ, GVZ, OVX, VIX and VXTYN, α is a 5 x 1 vector of intercepts, ϕl is a 5 x 5 matrix with ϕijl being the coefficient of the l-day lagged jth volatility index used to explain changes in the ith volatility index, ε t is a 5 x 1 vector of residuals, and p denotes the lag length of the system. Then, we conduct the Wald test under the null hypothesis that ϕijl = 0 for all i ≠ j , so that rejection of the null hypothesis would indicate that there is a transmission of implied volatility across commodity, equity, foreign exchange and Treasury bond markets.

The next step consists of analyzing whether scheduled news releases have a role in explaining volatility spillovers. To this end, we construct the following standardized measure of surprise (see e.g., Balduzzi et al., 2001; Beber and Brandt, 2006, 2009):

S kt =

Akt − Fkt

σk

(2)

where A kt is the released value for announcement k on day t, F kt denotes the corresponding Bloomberg forecast, and σ k is the standard deviation of the surprise component (i.e., A kt - F kt ) of the announcement k for the whole sample. 4 This variable equals zero on days where there is no announcement. Then, we estimate an augmented VAR(p) model of the form: p

DLVOLt= α + ∑ ϕl DLVOLt −l + Φ St + ε t (3) l =1

14

where Φ is a 5 x 1 vector of coefficients, St = ∑ S kt is a 1 x 1 vector which k =1

stands for the aggregate absolute surprise component of all considered announcements, and all other variables and parameters are defined as in Equation 1.

IV. Results

We estimate a VAR(3) model based on Equation 1. The appropriate lag length is determined using the Akaike information criteria along with the Lagrange Multiplier (LM) test for VAR residual serial correlation. The results are reported in Table 2. 4

The standardization of the surprise variable helps in comparing the effect of announcements that differ in the units of measurement.

The outcomes of the Wald test reveal that implied volatility spills over across commodity, equity, foreign exchange and Treasury bond markets. Thus, this finding extends the evidence of volatility transmission between US markets reported in the works by Badshah et al. (2013) and Liu et al. (2013), which do not include the recently created fixed-income volatility index VXTYN. Consistent with the results provided in the above mentioned works, we find that changes in VIX lead changes in the rest of the volatility indices (included VXTYN) and that changes in the US dollar/euro exchange rate volatility index (EVZ) have a significant impact on the commodity-based volatility indices (GVZ and OVX). In addition, this study shows that implied volatility is transmitted from the Treasury market (VXTYN) to the equity market (VIX), whereas lagged changes in the gold volatility index (GVZ) and VIX significantly affect changes in VXTYN. Thus, results derived from the inclusion of VXTYN can be applied to improve forecasts of changes in market expectations of volatility in the equity and Treasury markets. They are of particular importance to the design of investment strategies involving volatility derivatives, which allow traders to hedge and speculate against future levels of volatility. 5 In Table 3, we present the estimation results from the VAR(3) model after accounting for the surprise component of scheduled news announcements expressed in Equation 3. We can see that the coefficients of the announcement variable are not statistically significant in any case, whereas the explanatory power of lagged changes in other volatility indices remains. This finding indicates that implied volatility spillovers are not driven by public information about economic fundamentals, and this supports the notion that volatility contagion is

5

Currently, the CBOE lists options and futures in a number of volatility indices.

present between the commodity, equity, foreign exchange and Treasury bond markets. This result adds to the literature on implied volatility contagion. While Jiang et al. (2012) document volatility contagion across international equity markets using US and European volatility indices, the present study shows that there is also volatility contagion across equity- and non-equity-based volatility indices. Finally, we analyze whether evidence of volatility contagion is robust to the recent financial crisis period and the post crisis period. To this end, the augmented VAR(3) model is estimated from 1 June 2008 to 30 June 2009 (crisis period) and from 1 July 2009 to 31 March 2013 (post crisis period). 6 Overall, the results for these two periods are consistent with our main findings, indicating that the presence of volatility contagion is not driven by the crisis. 7

V. Conclusions

The present article provides evidence that implied volatility spills over across US commodity, equity, foreign exchange and Treasury bond markets using data from five CBOE implied volatility indices – EVZ, GVZ, OVX, VIX and VXTYN. Based on the just recently introduced Treasury bond volatility index (VXTYN), we show for the first time that implied volatility is transmitted from the equity market to the Treasury bond market and vice versa. The key contribution of this study to existing literature is the finding that economic news announcements do not account for observed volatility spillovers, suggesting volatility contagion across US equity and non-equity markets. Previous studies have reported the 6

According to the National Bureau of Economic Research (NBER), June 2009 is the end date of the recessionary period. See also Liu et al. (2013). 7 Detailed results can be obtained from the authors upon request.

presence of volatility contagion across international equity markets using exclusively equity-based volatility indices. We further find that evidence of volatility contagion is robust over the recent financial crisis period and the post crisis period. The findings of the present study create some strands for future research, namely, investigation of the effect of the surprise component of individual news announcement items on implied volatility transmission (see, e.g., Balduzzi et al. (2001), Beber and Brandt (2006, 2009)) and whether the announcement effect varies with the sign of the surprise (i.e., good and bad surprises) in the news announcement (see e.g., Beber and Brandt (2006)).

References

Äijö, J. (2008) Implied volatility term structure linkages between VDAX, VSMI and VSTOXX volatility indices, Global Finance Journal, 18, 290-302. Badshah, I. U., Frijns, B.and Tourani-Rad, A. (2013) Contemporaneous SpillOver Among Equity, Gold, and Exchange Rate Implied Volatility Indices, Journal of Futures Markets, 33, 555-572. Balduzzi, P., Elton, E. J.and Green, T. C. (2001) Economic news and bond prices: Evidence from the U.S. Treasury market, Journal of Financial and Quantitative Analysis, 36, 523-543. Beber, A.and Brandt, M. W. (2006) The Effect of Macroeconomic News on Beliefs and Preferences: Evidence from the Options Market, Journal of Monetary Economics, 53, 1997-2039.

Beber, A.and Brandt, M. W. (2009) Resolving Macroeconomic Uncertainty in Stock and Bond Markets, Review of Finance, 13, 1-45. Bekaert, G., Harvey, C. R.and Ng, A. (2005) Market integration and contagion, Journal of Business, 78, 39-69. Boyson, N. M., Stahel, C. W.and Stulz, R. M. (2010) Hedge Fund Contagion and Liquidity Shocks, The Journal of Finance, 65, 1789-1816. Connolly, R. A.and Wang, F. A. (2003) International equity market comovements: Economic fundamentals or contagion?, Pacific-Basin Finance Journal, 11, 23-43. Demeterfi, K., Derman, E., Kamal, M.and Zou, J. (1999) More than you ever wanted to know about volatility swaps, Goldman Sachs Quantitative Strategies Research Notes. Ederington, L. H.and Lee, J. H. (1996) The creation and resolution of market uncertainty: The impact of information releases on implied volatility, Journal of Financial and Quantitative Analysis, 31, 513-539. Fleming, M.and Remolona, E. (1997) What moves the bond market?, FRBNY Economic Policy Review, 3, 31-50. Jiang, G. J., Konstantinidi, E.and Skiadopoulos, G. (2012) Volatility spillovers and the effect of news announcements, Journal of Banking and Finance, 36, 2260-2273. Konstantinidi, E., Skiadopoulos, G.and Tzagkaraki, E. (2008) Can the evolution of implied volatility be forecasted? Evidence from European and US implied volatility indices, Journal of Banking and Finance, 32, 2401-2411.

Liu, M.-L., Ji, Q.and Fan, Y. (2013) How does oil market uncertainty interact with other markets? An empirical analysis of implied volatility index, Energy, 55, 860-868. Nikkinen, J.and Sahlström, P. (2004) International transmission of uncertainty implicit in stock index option prices, Global Finance Journal, 15, 1-15. Siriopoulos, C.and Fassas, A. (2012) An investor sentiment barometer — Greek Implied Volatility Index (GRIV), Global Finance Journal, 23, 77-93.

Table 1. Summary of scheduled news announcements. Entries report the time at which each announcement is usually released, the source of the report, the frequency of release and the total number of observations in our sample. The sample spans June 01, 2008 to March 31, 2013 Announcements and timing 8:30 am ET announcements CPI (Consumer price index) DGO (Durable goods orders) GDP (Gross domestic product) IJC (Initial jobless claims) NFP (Non-farm payrolls) PPI (Producer price index) RS (Retail sales less autos) 9:15 am ET announcements CU (Capacity utilization) IP (Industrial production) 10:00 am ET announcements CCI (Consumer confidence index) LI (Index of leading indicators) NHS (New home sales) PMI (Purchasing Manager Index) 2:15 pm ET announcements FOMC (Federal Open Market Committee rate decision)

Source of report

Frequency

Obs.

Bureau of Labor Statistics US Census Bureau Bureau of Economic Analysis Department of Labor Bureau of Labor Statistics Bureau of Labor Statistics US Census Bureau

Monthly Monthly Monthlya Weekly Monthly Monthly Monthly

58 58 58 252 58 58 58

Federal Reserve Federal Reserve

Monthly Monthly

58 58

Conference Board Conference Board US Census Bureau Institute for Supply Management

Monthly Monthly Monthly Monthly

58 58 58 58

Federal Reserve

The FOMC holds 40 eight meetings per year a GDP is a quarterly statistic, but advance, preliminary and final estimates for the quarter are released in successive months.

Table 2. Estimation results from the VAR(3) model for the period 1 June 2008 to 31 March 2013. t-statistics are provided in parentheses. One and two asterisks denote rejection of the null hypothesis at the 5% and 1% significance levels, respectively. α DLEVZ t-1 DLEVZ t-2 DLEVZ t-3 DLGVZ t-1 DLGVZ t-2 DLGVZ t-3 DLOVX t-1 DLOVX t-2 DLOVX t-3 DLVIX t-1 DLVIX t-2 DLVIX t-3 DLVXTYN t-1 DVXTYN t-2 DVXTYN t-3 Adj. R2 Wald test (H 0 : φ ijl = 0 for i ≠ j)

DLEVZ 0.000 (0.149) -0.174** (-4.988) -0.084* (-2.378) -0.100** (-2.885) 0.027 (0.863) 0.006 (0.201) 0.019 (0.644) 0.039 (1.024) 0.002 (0.075) -0.001 (-0.032) 0.071** (2.629) 0.037 (1.347) -0.018 (-0.699) 0.031 (0.950) -0.011 (-0.333) 0.009 (0.290) 0.029 131.622**

DLGVZ -0.001 (-1.150) 0.080* (2.015) -0.038 (-0.951) -0.056 (-1.415) -0.072* (2.007) -0.240** (-6.794) -0.008 (-0.241) 0.026 (0.594) 0.063 (1.425) 0.042 (1.046) 0.010 (0.340) 0.072* (2.308) 0.026 (0.844) -0.006 (-0.180) 0.072 (1.901) -0.033 (-0.885) 0.047

DLOVX -0.003* (-2.177) 0.093** (2.857) -0.052 (-1.556) 0.011 (0.367) -0.049 (-1.681) -0.030 (-1.057) 0.014 (0.510) -0.220** (-6.131) 0.047 (1.311) -0.089** (-2.686) 0.085** (3.367) 0.012 (0.487) 0.005 (0.227) 0.049 (1.581) 0.020 (0.647) 0.016 (0.526) 0.068

DLVIX -0.003* (-2.177) 0.039 (0.776) 0.028 (0.555) -0.015 (-0.304) -0.052 (-1.130) -0.017 (-0.381) -0.004 (-0.094) 0.004 (0.085) 0.172** (3.043) -0.020 (-0.400) -0.149** (-3.759) -0.142** (-3.540) -0.090* (-2.287) 0.120* (2.510) 0.009 (0.196) 0.009 (0.189) 0.030

DLVXTYN -0.000 (-0.625) -0.031 (-0.916) -0.013 (-0.380) -0.002 (-0.069) -0.072* (-2.330) -0.034 (1.114) -0.018 (-0.617) -0.019 (-0.513) 0.004 (0.115) -0.030 (-0.861) 0.067* (2.490) 0.020 (0.750) 0.001 (0.052) -0.011 (-0.361) -0.088** (-2.704) -0.027 (-0.838) 0.006

Table 3. Estimation results from the VAR(3) model after accounting for the effect of news announcements for the period 1 June 2008 to 31 March 2013. t-statistics are provided in parentheses. One and two asterisks denote rejection of the null hypothesis at the 5% and 1% significance levels, respectively. α DLEVZ t-1 DLEVZ t-2 DLEVZ t-3 DLGVZ t-1 DLGVZ t-2 DLGVZ t-3 DLOVX t-1 DLOVX t-2 DLOVX t-3 DLVIX t-1 DLVIX t-2 DLVIX t-3 DLVXTYN t-1 DVXTYN t-2 DVXTYN t-3

St Adj. R2 Wald test (H 0 : φ ijl = 0 for i ≠ j)

DLEVZ 0.001 (0.874) -0.180** (-5.143) -0.075* (-2.118) -0.093** (-2.676) 0.022 (0.711) 0.000 (0.025) 0.016 (0.536) 0.034 (0.892) 0.007 (0.190) -0.003 (-0.102) 0.079** (2.904) 0.038 (1.395) -0.019 (-0.728) 0.031 (0.946) -0.005 (-0.177) 0.009 (0.290) -0.001 (-1.300) 0.031 161.274**

DLGVZ -0.000 (-0.242) 0.081* (2.022) -0.037 (-0.902) -0.051 (-1.291) -0.075* (2.075) -0.246** (-6.905) -0.009 (-0.279) 0.020 (0.466) 0.062 (1.391) 0.040 (0.995) 0.016 (0.518) 0.075* (2.383) 0.024 (0.805) -0.006 (-0.182) 0.072 (1.896) -0.029 (-0.770) -0002 (-1.412) 0.048

DLOVX -0.003* (-2.179) 0.092** (2.809) -0.051 (-1.527) 0.010 (0.329) -0.049 (-1.661) -0.028 (-0.990) 0.014 (0.512) -0.218** (-6.049) 0.049 (1.349) -0.088** (-2.665) 0.084** (3.274) 0.011 (0.445) 0.006 (0.247) 0.049 (1.580) 0.021 (0.676) 0.014 (0.453) 0.000 (0.643) 0.067

DLVIX -0.004 (-0.176) 0.044 (0.860) 0.024 (0.475) -0.014 (-0.283) -0.051 (-1.126) -0.020 (-0.441) -0.003 (-0.079) 0.001 (0.024) 0.167** (2.950) -0.021 (-0.416) -0.147** (-3.684) -0.140** (-3.479) -0.091* (-2.311) 0.121* (2.511) 0.005 (0.121) 0.014 (0.297) -0.001 (-0.766) 0.030

DLVXTYN -0.001 (-0.621) -0.031 (-0.913) -0.013 (-0.384) -0.002 (-0.083) -0.072* (-2.314) -0.034 (1.125) -0.018 (-0.611) -0.018 (-0.495) 0.004 (0.117) -0.030 (-0.853) 0.066* (2.442) 0.020 (0.739) 0.001 (0.057) -0.011 (-0.361) -0.088** (-2.695) -0.028 (-0.846) 0.000 (0.169) 0.005