R. Jared DeLisle
Department of Economics and Finance
Utah State University
James S. Doran
Implied Capital
Kevin Krieger
Department of Accounting and Finance
University of West Florida
December 2014
Communications
Author:
James S. Doran
Address:
2500 30th Street Suite 206
Boulder, CO. 80303
Tel.:
(850) 644-7868 (Office)
E-mail:
jdoran@impliedcapital.com
1. INTRODUCTION
Extreme stock market downturns are the most disconcerting periods for investors
since they are risk averse, wish to limit their exposure to volatility, and seek to avoid
negatively skewed payoffs (Kumar, 2009). Significant market downturns, such as those
experienced from October, 2007 to March, 2009, resulted in massive losses for most all
investors as equity indices retreated over 50%. The anxiety of such movements even
provoked many investors to sell large portions of equity, mutual fund, and exchange-traded
fund (ETF) holdings, thus guaranteeing large losses.1 The ability to more effectively hedge
equity investments with holdings that would reduce portfolio downside risk and without
giving up upside potential could combat a considerable amount of investor unease.
While a number of assets exist with returns that are, on average, negatively correlated
with equities, these instruments may not provide the desired hedge against market
downturns since the return correlations become positive in times of distress. For example,
during the market collapse in 2008, the value of both equities and commodities fell, though
traditionally, commodities are negative beta assets.2 Bond holdings also declined in value
during the 2008 crash, as the risk in commercial borrowing increased while liquidity fell.
Many hedge funds designed to cushion losses in equity markets experienced reversals over
the 2007-2009 period. Szado (2009) documents the increased correlation of asset returns
over the crisis period above the levels seen in the 2004-2006 period, implying that as the
need for diversification grew, the ability of many assets to hedge equity holdings shrank at
the most inopportune time. Additionally, while tail oriented products like long puts did very
Mary Pilon, Many Bought Shares High, Sold Low, Wall Street Journal, May 18th 2009.
Numerous commodity indices also retreated by more than 50% of value during the equity market decline of
2007-2009 as inflation ground to a standstill and consumption of raw materials slowed.
1
2
well in crash periods, in period of stability, such as the recent bull market from 2009-2014,
they have had significant negative performance.3 Holding volatility as an asset may make
traditional negative or counter-cyclical investments moot for hedging purposes because of
the mean-reverting nature of volatility and its negative skewed payoff.
Since the introduction of VIX, it has widely been regarded as the economys indicator
of risk in the equities market. As noted by Whaley (1993, 2000) , the VIX index is considered
the investors fear gauge index. An important byproduct of introducing VIX was the
newfound opportunity for investors to trade in futures, introduced in March 2004, and
options, introduced in February 2006, thus allowing investors to enter into contracts which
generate payoffs specifically related to volatility. Additionally, exchange-traded products
have been established, starting in 2009 with the iPath S&P 500 VIX Short-Term Futures ETN
(ticker: VXX), that offer a more direct way to access volatility as an investment. While it is
possible to invest in equity options and futures on the S&P 500 and construct a payoff that
would be related to the volatility of the index, a more direct investment in VIX might require
less management of the position and should provide less tracking error. Thus, we wish to
explore the performance of VIX portfolios to assess the implications of holding the VIX index
alongside the S&P 500. In part, this should reveal whether investing in the VIX provides an
effective hedge to long-equity positions, either through lower costs or superior returns, than
alternative hedges, such as purchasing index puts. This seems especially relevant given the
current low levels of volatility and steepness of the volatility skew. Our results show that, if
Refer to the performance on a put writing strategy by the Asset Consulting Group that noted total return for
put writing strategy is 1153% since 1986. This is also consistent with the findings of Litterman (2011) and
Ilmanen (2012).
3
the VIX were directly investable, holding VIX in a portfolio with S&P 500 yields positive and
significant alphas across all market cycles.
With more widespread access to market information, VIX has gained increased
exposure in recent years, particularly as it rose to rare levels during the 2008 market decline.
In the wake of increased attention, Whaley (2009) sought to clarify the meaning of VIX and
discuss its characteristics. He emphasizes that, like the S&P 500 index, the VIX index is not
directly investable. However, while it is quite simple to replicate the payoff of the S&P 500
by holding the 500 underlying stocks in the appropriate proportions (or more simply, via
investment in low-cost ETFs), it is difficult or nearly impossible to replicate VIX by holding
the underlying S&P 500 options. This is in part because VIX is constructed using the first two
monthly expiration call and put out-of-the money options, with weights that are squared.
Additionally, these weights change daily. Thus, even if a portfolio were able to hold the
correct proportions on a given day, which would require a significant investment in many
option contracts, the next day the proportions would change, and the rebalancing costs
would be prohibitive making the VIX index, viewed as an asset, practically untradeable.
The introduction of futures and options on VIX, and more recently exchange-traded
notes (ETNs), made it possible to invest in volatility, or at least to take a position on its future
direction. However, it is unclear whether using these products provides a payoff similar to
that of the index, or provides the hedge against increases in volatilities that most investors
desire. As Szado (2009) notes, exposure to VIX calls and puts, as well as VIX futures, does
not directly mimic holdings in the spot levels of VIX given that the mean-reverting nature of
the underlying are priced into the derivative values. Since volatility mean-reverts, investing
in the VIX index when it is low could be seen as likely to provide protection against volatility
3
increases. Giot (2005), Dennis, Mayhew, & Stivers (2006) and DeLisle, Doran, & Peterson
(2011) document the asymmetric relationship between VIX and the S&P 500 and specifically
show that VIX increases and S&P 500 declines are more strongly correlated than VIX
decreases and S&P 500 increases. Whaley (2009) documents the mean-reverting nature of
VIX and also describes its asymmetric nature such that VIX will rise more (less) dramatically
during a stock market decline (rally). Furthermore, Simon (2003) notes the tendency of
traders to overvalue (undervalue) the equity market when volatility levels are unusually low
(high). Consistent with Daigler & Rossi (2006), it would appear that investing in VIX when
it is low not only provides a hedge against declines in the S&P 500, but will not proportionally
penalize investors when the S&P 500 increases.
A number of researchers have thus considered the possibility of hedging portfolios
with VIX-mimicking assets. Dash & Moran (2005) initially considered the ability of newly
formed VIX-based products to lower portfolio risk. Emerging possibilities then developed
for such a strategy, including the use of VIX futures, VIX options and VIX-based ETNs.
Brenner, Ou, & Zhang (2006) introduce an option on a straddle designed to hedge volatility
risk. This instrument is sensitive to volatility innovations and thus useful as a hedge.
Windcliff, Forsyth, & Vetzal (2006) consider the variations in the contract designs of
volatility derivatives and discuss the difficulties of hedging the returns with such
instruments, particularly given delta and delta-gamma hedging techniques. Black (2006) and
Moran & Dash (2007) find that adding VIX futures to a passive portfolio can significantly
reduce portfolio volatility. VIXs quick movements during risky markets also improve the
skewness and kurtosis of the overall portfolios. Briere, Burgues, & Signora (2010) advocate
a sliding approach when hedging in which more (fewer) VIX futures contracts are held when
4
VIX levels are notably lower (higher) due to the mean-reverting nature of the index. Jones
(2011) and Warren (2012) also advocate using VIX futures only in a tactical manner.
While Whaley (2013) points out that trading VIX products results in poor
performance, an important remaining question is the cost-effectiveness of using VIX and VIXstyle products as a hedging or speculative strategy. If VIX were tradable, investing in the S&P
500 and VIX, perhaps when VIX is relatively low, might provide investors with a portfolio
that will increase in value when the S&P 500 increases and which will be somewhat
protected when the S&P 500 falls. Along with looking at VIX as a complementary investment
to the S&P 500, we will look at certain levels of VIX to assess when it might be best to invest
in volatility.
We then explore the benefits of investing in VIX futures and ETNs and study whether
these products mimic the payoff to the VIX index. Since these are volatility products that are
tradable, if the payoffs do not replicate the underlying index, it may imply that a tradable
asset on volatility does not provide investors with appropriate protection, especially at the
prices required for these products. Our results show that this is the case for most VIX-based
products. These findings are contrary to those of Chen, Chung, & Ho (2011), who determine
VIX futures do enhance the performance of a portfolio of equities, and are consistent with
the warnings of Whaley (2013) against using VIX-related exchange-traded products. Next,
we examine the benefit of using VIX call options for portfolio insurance, relative to S&P 500
put options, to assess whether the payoffs to VIX options provide similar downside
protection. We find that VIX options provide a reasonable hedge, as well as capture the
positive skewness of VIX. However, while VIX options maybe cheaper, on a relative basis,
compared to S&P 500 puts, the long-term expenses are still cost-prohibitive long-term.
5
Finally, we construct a low-cost portfolio that attempts to capture the increases in VIX by
exploring which S&P 500 options drive the changes in VIX. We demonstrate that, by
deconstructing VIX into the individual option components, it is possible to form a portfolio
of liquid S&P 500 options that captures the payoff to the VIX index which eludes the ETN,
future, and option contracts on VIX. This synthetic VIX position improves the hedging
prospects of otherwise passive, long-equity investors especially when VIX calls are added to
the portfolio.
The rest of the article is as follows: Section 2 presents the data and its sources, Section
3 presents the analyses of the performance of VIX and existing VIX-like assets as a hedge
against market declines. Section 4 investigates which parts of VIX make it a favorable hedge
and evaluates a low-cost portfolio constructed to mimic VIX returns. Section 5 concludes the
paper.
http://www.cboe.com/micro/VIX/historical.aspx
which mimics the performance of the S&P 500, also from CRSP. VXX ETN returns are
collected from Bloomberg.
Futures positions are used as one method for creating a VIX based portfolio. VIX
futures began trading on the CBOE futures exchange in March of 2004. The daily prices of
VIX and S&P 500 futures contracts are collected from Bloomberg, beginning with the arrival
of VIX futures. However, VIX futures with two-month expiration dates were not continuously
available until the end of 2005. Thus, we limit our analyses of the VIX futures strategy to the
time period spanning 2006-2013. A second method for assembling a tradable VIX-based
portfolio is constructed with the use of VIX options. These options began trading on the CBOE
in February, 2006, and their daily prices are also collected from Bloomberg through 2013.
Additionally, data is collected for S&P 500 index options, SPX, beginning in 1996 and
ending in 2013. SPX calls and puts are used to create the synthetic VIX position. While our
focus is on the hedging ability that VIX futures, options, and the synthetic position possess,
the hedging abilities of S&P 500 put positions are also considered for comparison purposes.
The Fama & French (1993) MKT, SMB, and HML factors and the Carhart (1997) UMD factor
are provided on Kenneth French's data library website.5 Additionally, SPX options are used
to create ATM straddles, which are typically employed as plays on volatility, and whose
returns are used as an option-related factor in evaluating the results of various strategies.
Using the returns from this strategy as an additional factor in a risk model can capture the
volatility premium that is present in many hedge fund and option based strategies.
TABLE I presents the summary statistics for S&P 500 and VIX returns. Panel A shows
that S&P 500 returns over the entire sample period average 70 basis points per month, while
VIX returns average 130 basis points per month. However, the standard deviation of the VIX
monthly returns is 18.83%, which is so volatile that the average VIX returns are statistically
indistinguishable from zero. Additionally, the median return is negative, highlighting the
positively skewed distribution and the mean reverting nature of volatility.
The correlation between S&P 500 and VIX returns is -0.65, demonstrating a strong
negative relation as should be expected. Panel B limits the sample to months in which S&P
500 returns are positive. Limiting the sample in this manner yields average S&P 500 and VIX
returns of 3.22% and -6.52% per month, respectively. The correlation between S&P 500 and
VIX returns during these months is -0.282. Panel C, when the sample is limited to months
where S&P 500 returns are negative, shows that the average S&P 500 and VIX returns are 3.61% and 14.66% per month, respectively. The return correlation for these months is 0.531. When separating the returns into up and down periods, mean returns for both the
S&P 500 and VIX are significantly different from zero, highlighting the importance of
differing market conditions. More interestingly, the absolute value of the mean S&P 500
returns is similar for up and down market conditions, while the positive mean VIX returns
are twice as large during periods of negative S&P 500 movement as the negative mean VIX
returns in periods of positive S&P 500 returns. This result highlights the asymmetric relation
between S&P 500 returns and VIX returns and may be indicative of the potential for using
volatility as a hedging instrument against S&P 500 losses.
Panel D breaks the sample into months where VIX and the S&P 500 have different
monthly signed returns, and month where the signed returns are the same. Almost 75% of
the sample has months when the VIX and S&P 500 have different signed returns, with 46%
of those when the S&P 500 is up and VIX is down. Just over 17% of the sample has months
when both VIX and the S&P are up, which is not unreasonable since large positive
movements can result is increases in short-term volatility.6 The 8% of observations when
volatility falls while the S&P 500 falls is unusual, but reflects that volatility and returns are
not always negatively correlated.
In order to construct a proper hedge ratio for the inclusion of VIX in a portfolio
consisting of the S&P 500, we first need to estimate how VIX and the S&P 500 move together.
Considering the potential for a non-linear relation between VIX and S&P 500, we estimate
the following regression over the entire sample period:
, = + &500, + &500, 2 +
(1)
where , is the return of the VIX index on day t, &500, is the return of the S&P 500 on
day t, is a constant, and is the residual on day t. Since it is well-documented that the
relation between market returns and volatility is asymmetric, we split the sample into
This can happen because call options can be bid resulting in a steepening of the volatility curve resulting in a
volatility smile.
6
positive and negative S&P 500 returns and re-estimate the regression. TABLE II presents the
estimations' results.
When the full sample is used, the relation between VIX and S&P 500 is highly
significant, as a 1 % change in the S&P results in a 2.7% change in the VIX index. However,
when separating the sample into positive and negative S&P 500 returns, the relation is
clearly not symmetric. When the S&P falls is statistically significant at least at the 5% level,
which is not the case when the S&P rises. Using the coefficient estimates, we then calculate
the weight of VIX required in a portfolio long both VIX and S&P 500 to yield a portfolio return
of zero when the S&P 500 experiences a 5% loss. On average, over the entire time period, the
portfolio would require a weight of VIX of 22.5% (i.e. a loss of 5% in the S&P 500 would be
completely offset by an increase the VIX portion of the portfolio so that the return is 0%). We
use a 5% loss threshold since it is common for long mangers and quantitative strategies to
enact stop losses or rebalance at or around this level (Khandani & Lo, 2011).
The entire time series of data is not available to an investor at any particular point in
time, however. Therefore, each month we estimate equation (1) at the end of month t using
only the data from months t-60 to t-1. We then recalculate the VIX weight at the end of month
t that would yield a portfolio return of 0% if the S&P 500 were to fall by 5% over the next
month.
ATM straddle returns have been used by Coval & Shumway (2001) and Broadie, Chernov, & Johannes
(2009) to help explain option, or non-normal returns. We follow a methodology to construct the factor similar
to Broadie, Chernov, & Johannes (2009) who use a ratio of 1:1 long puts and calls that holds the next month
expiration and rolls over each contract at the end of the month to the following expiration. For example, at the
end of January 2008, a call and a put are identified that are closest to ATM and expire in March 2008. These
two options are purchased at the ask price at the end of January. These options are then sold at the bid price
at the end of February and new ATM options are purchased with April expiration dates.
7
11
See Carr & Wu (2009) for details on synthesizing and pricing variance swaps.
12
the VIX index when it sinks below a certain level may hedge against future decreases in the
S&P 500 without the expense of constantly maintaining the hedge. Thus, along with the basic
or no threshold case, we present results for portfolios which purchase VIX or VIX-like
assets only when the VIX is below the thresholds of 12.51, 20.2, 27.89, and 35.35 at the
beginning of a one-month period.
The results in panel A of TABLE III show that VIX theoretically serves as an effective
hedge to holdings in the S&P 500. Using no threshold, a portfolio that holds 20% VIX results
in a monthly alpha of 83 basis points, and is significant at the 1% level. This alpha remains
significant controlling for CAPM, Fama-French, and option based factors, and is in excess of
50 basis points a month. In addition, CAPM, 4-factor and 5-factor betas decline substantially
from 1 when portfolios are hedged with theoretical VIX. The specific investing threshold
alphas increase when all thresholds are at or above the mean, with a maximum monthly
alpha of 109 basis points, for the portfolio that only invests in VIX when it is at or below one
standard deviation above the VIX historical mean. As might be expected, market betas
decline across higher thresholds for VIX levels, which highlights the dampening effect of
adding VIX to a portfolio. Funds are placed entirely in the S&P 500 on a more frequent basis
when a lower VIX threshold is invoked, and thus while hedging costs may be removed,
portfolio betas rise with lower thresholds.
Panels B and C break the periods into two sub-periods, 2006-2013, and 2009-2013.
This is done to capture the high volatility and subsequent low volatility period over the past
13
seven years, and isolate the low volatility period that has been experienced over the prior
four years. Both sub-periods show that adding VIX, especially when below the one standard
deviation threshold, results in a significant alpha even after controlling for typical factors.
The fact that some of the alphas are significant in the low volatility period is surprising since
the average level of volatility over the past four years has been significantly lower than the
long-term historical mean. This suggests that holding VIX, even when volatility is low, has
added benefits, especially for small periods of volatility shocks, such as the flash crash in May
2010 and the sovereign wealth crisis in 2011.
FIGURE 1 demonstrates the performance of portfolios which embark on the various
investment strategies hedging directly in theoretical VIX based on the ratios of TABLE II and
the thresholds given.
[Insert FIGURE 1]
Assuming an initial investment of one million dollars, portfolios which are allowed to
invest in theoretical VIX and the S&P 500 from the 1996-2013 period have final values that
are two to three times greater than investment in the S&P 500 index alone. This performance
discrepancy is even more pronounced for the 2006-2009 sub-period, which coincide with
the availability of VIX futures and VIX calls and periods of higher volatility. Unfortunately, it
is not possible to directly invest in VIX, and as such, we question whether the performance
of VIX based products will be analogous to the impressive returns offered by the theoretical
VIX hedge. It is entirely possible that the pricing and day-to-day management of these assets
may erase any theoretical improvements for investors.
14
As a VIX futures contract approaches expiration, basis risk increases and the divergence between frontmonth and next-month contracts increases. Additionally, there is no way to hedge the difference between last
price and settlement price, which can cause huge variation in returns to the strategy.
9
15
findings is Zhang, Shu, & Brenner (2010). Holding VIX futures does impart a reduction to the
market beta, in fact resulting in a market beta insignificantly different from zero in the no
threshold case.
Portfolios which hold VIX futures substantially underperform those which hold the
theoretical VIX by a difference of 149 basis point per month using the four-factor model for
the 2006-2013 period, and more importantly, underperform the S&P 500 index itself.
Attempts to utilize the cost-savings thresholds result in even poorer portfolio performance,
as the VIX futures underperform by 183 basis points. This is, in part, driven by the fact that
VIX futures would not be purchased in the Fall of 2008, at a time when VIX levels were
historically high, but during which even higher levels of VIX were later realized. So, while VIX
futures can dampen periods of high volatility, the overall costs are prohibitive.
[Insert FIGURE 2]
FIGURE 2 highlights the expense of the VIX futures hedge using the one standard
deviation threshold starting in 2009. As can be seen, during this recent period, where VIX
started in the 40s and then retreated to the low teens, the VIX futures hedge has significantly
underperformed, resulting in an overall negative performance over the period compared to
the equity only portfolio and the theoretical VIX-hedged portfolio.
16
Due to the shorter time series of VIX calls relative to the series of the VIX index, we compute fictional
options in 2005 using historical volatility of the VIX index. These fictional options are used in the averaging
scheme so that all options available in 2006 can be used in the portfolio analysis.
10
17
essentially one-to-one. Using this methodology to construct the size of the hedge results in
allocating about 2% of funds, on an annualized basis, to the options, which is similar to to the
allocation typically used to hedge equity exposure.11 In Panels B and C of TABLE III, we see
the performance of S&P 500 portfolios which are hedged with VIX calls and S&P puts,
respectively. This includes, for comparison purposes, consideration of the option hedge at
the given hedging thresholds. Using this 2% threshold makes sense for the options, versus
the 20% number for the VXX and VIX futures, because the options are designed as a tail hedge
and have higher built-in leverage while the other products are assets that do not have any
associated theta component.
We find lower market betas for the portfolios which utilize S&P puts across most of
thresholds, however, the betas of VIX-call hedged portfolios are almost zero in the no
threshold case. This suggests a significant and highly skewed response when volatility
exceeds 36. Furthermore, VIX-call hedges appear to be more cost effective as the mean
returns and alphas to such hedges are superior to the S&P put hedges, particularly in the
base, no threshold case. While the CAPM, 4-factor, and 5-factor alphas of portfolios hedged
with S&P 500 puts are significantly below zero at the 5% confidence level, in the no threshold
case, no such significance presents for the VIX-call hedged portfolios.
We see the superiority of the VIX-call hedged portfolio in FIGURE 2. While the S&P
500 put hedge, VIX call and VIX futures all underperform the S&P 500, even in the presence
of a substantial market crash during a period of overall low volatility, the VIX-call hedge
strategy has the best performance. The relative inexpensiveness of the VIX calls can be seen
early in the period as the portfolio hedged with VIX calls does not impart nearly the costs of
11
Refer to the paper An introduction to Tail Risk Parity by Alliance Bernstien for the cost of heding
18
the S&P put hedge. Hedging simple, passive positions like long-S&P 500 with VIX calls, rather
than S&P 500 puts, appears to be the best potential hedging strategy. It is the ability of VIX
calls to capture positive skewness that suggests they are a superior alternative with low
relative cost. What remains unclear is whether the difference in the payoffs in the two
hedging strategies is due to underpricing of VIX calls or overpricing of S&P 500 puts. There
is recent evidence that suggests skewness is priced in equity returns (Boyer, Mitton, &
Vorkink, 2010) and index option prices (Doran & Krieger, 2010). The portfolio result
suggests that the options on volatility have not incorporated corresponding positive
skewness or the increase in skewness in VIX relative to the decrease in skewness in the
index. As such, it appears that VIX calls are cheap only because of the differences in
underlying return distributions of volatility and equity instruments, not necessarily because
they are mispriced.
12
The weights of VXX in the portfolios are identical to the weights of theoretical VIX.
19
FIGURE 2 further highlights the poor performance of a portfolio that uses VXX to
hedge volatility. The return performance is similar to that of the VIX futures hedging strategy,
but costs slightly more. This can be tied back to fees and the costs of rolling over futures
contracts each month. In summary, it appears that while VIX itself provides an excellent way
to hedging market downside without significant expense, the market products that are
available to invest in are just not viable as long-term hedging solutions.
20
Before deriving an empirical specification to test the relationship between S&P 500
options and VIX changes, we briefly summarize the methodology used to calculate the VIX
Index. The specification for VIX is given by:13
2 =
2 30
= 100{1 12 [
2 1
( ) [ 1]
30 1
] + 2 22 [
2 1
]}
365
30
(2)
(3)
where is the strike price of the options that are currently closest to ATM or OTM, is the
difference in strike prices, ()is the midpoint price of the option at strike , 1 is the
expiration of the front month option, 2 is the expiration of the next-month option, N is the
minutes to expiration and represents the ATM forward price of the index. Using all options
available that are closest to ATM and OTM, equation (1) constructs a 2 for both the frontmonth and next-month options, while equation (2) weights the 2 to create a 30-day
measure of implied volatility.
Although both equations are straightforward to calculate and recreate the current
level of VIX, the non-linear nature makes replicating the payoff to the VIX index difficult to
accomplish. The weight on each option is a function of a root-weighted time variable, which
changes daily, making rebalancing costs prohibitive while creating indivisible option units.
However, it is possible to unwind the formula into smaller components, allowing for a test of
21
the effect of each option, or the changes in the prices and moneyness of each option, on the
changes in VIX.
Taking the natural log of equation (2) removes the nonlinear term such that equation
(2) become linear and allows for a simple regression to assess the impact of the individual
options on the change in VIX. Since changes are of interest, testing the difference
in VIX each day. The key explanatory variables to include are whether the option was a call
or a put, the time to expiration of the option, the price change, the strike price or moneyness
change, and an interaction term between these variables. It is necessary to incorporate an
interaction term, since the first term in equation (1) is multiplicative in these variables.
The following specifications are run, controlling for time variation:
ln (
ln (
(4)
(5)
where ,,,t is the change in price of (call/put) option with a strike of and maturity
at time t. represents the change in moneyness of the option, where put moneyness is
equal to K/S and call moneyeness is equal to S/K. Thus, more positive changes bring both
definitions of moneyness closer to ATM. is the time to maturity of the option, and is the
interaction of all three variables. is a dummy variable equal to one if the option is a call
and zero if the option is a put. Equation (3) interacts the dummy variables with change in
price and moneyness variables to capture differences between call and put options. Equation
22
(4) examines the effect of price and moneyness changes of calls and puts separately on VIX
changes. Each specification clusters on the date to avoid overstating the t-stats. The results
are shown in TABLE V.
[---Insert TABLE V---]
Seven estimations are conducted with the first three using equation (3) and the final
four using equation (4). The first utilizes the full sample while the second (third) considers
the positive (negative) VIX changes only. The fourth and fifth estimations use only puts and
segment the sample based on whether VIX is above or below the historical mean of 20.8. The
last two estimations use only calls, again segmenting the sample based on VIX levels above
and below 20.8.
The results of the first estimation reveal that increases in option prices are positively
related to VIX changes. Specifically, if the option becomes one dollar more expensive, then
VIX will increase by 0.25% on average. However, if the option is a call, for each dollar
increase, the change in VIX is only 0.11% as the coefficient of the call dummy is -0.14. This
suggests a greater impact from put price changes than call price changes. Options that are
closer to ATM have a greater effect on VIX values, as a 0.01 increase in moneyness results in
an increase of 3.32% in VIX. Again, this is only for puts, because as call options approach ATM
status, the effect on VIX is a fall of -4.01%. This is consistent with negative correlation
between VIX and the S&P 500 since puts (calls) become more expensive as the S&P 500 falls
(increases). There is no significant relationship with the expiration of the option, although
23
the negative coefficient on DM is consistent with a negative option theta. The interaction
term is positive, as expected, but the effect on the change in VIX is small.14
The results of the second and third estimation are similar in direction for all the
coefficients, but the interpretation of the coefficients reveals interesting results. The
coefficient on is 0.141 larger and statistically different than the coefficient when VIX
is increasing versus decreasing, suggesting an asymmetric response to put price changes on
VIX changes. Also the discrepancy between put and call price changes on VIX changes is
larger when VIX is increasing. When VIX decreases, is insignificant, again
highlighting the importance of put price changes for VIX increases, which are highly
negatively correlated with negative S&P returns. The results of the final four estimations
further emphasize these findings. Put price changes have a greater effect on VIX changes at
all levels of VIX, relative to calls, but the effect is especially strong for VIX levels less than the
historical mean.
These results suggest that ATM put options are most responsible for driving changes
in VIX, and this appears particularly true when VIX is below its historical mean. Given these
findings, and because many investors are interested in a portfolio that hedges against
downturns in the market, holding a portfolio that is long in ATM S&P 500 puts appears to be
a natural fit. However, it has been well documented that puts are expensive.15 Thus, it is
necessary to not only go long puts, but sell either OTM puts or corresponding calls to offset
the cost. Given the results in TABLE V, we opt to form a portfolio that will buy ATM puts and
A one dollar change times a 0.01 change in moneyness for a one-month option, multiplied by the 48.47
coefficient on I, results in a 0.02% change in VIX.
15 For example, Jackwerth (2000), At-Sahalia, Wang, & Yared (2001), Coval & Shumway (2001), Bakshi &
Kapadia (2003), Bondarenko (2003), Bollen & Whaley (2004), and Liu, Pan, & Wang (2005) generally find that
the historical costs of puts, particularly OTM and ATM puts, are too expensive to be justified.
14
24
OTM calls, and sell ATM calls and OTM puts. This will allow the portfolio to benefit from
volatility shocks by having a put spread, but will have a relative low cost from selling a call
spread. This call spread not only funds the volatility protection but combined with the net
long equity positon can have a payoff structure similar to a covered call. This means that the
portfolio is susceptible to underperformance if there are big upward moves in the market.
Since one of the hurdles in replicating VIX is the difficulty in daily rebalancing costs
of the options, the replicating portfolio will enter into the put and call spreads investments
at the beginning of the month, using both the front maturity options only. The investment in
the options will only occur when the VIX index is below the plus one standard deviation level
since earlier results have shown the negative correlation and asymmetric response between
the VIX index and the S&P 500. The portfolio will hold the equivalent number of options
corresponding to 100 shares in the index, while accounting for the margin required by selling
options. As such, 20% of the notional value of the portfolio to invest in volatility will be used
equally to pay for the premium of the put spread minus the call spread and the margin
required for the short call spread. This somewhat penalizes the portfolio since the short call
positions are covered by the long equity investment. However, since the 80% long portfolio
has to be maintained, unless it is at 100% at higher levels of volatility, a call position that
finished in the money and is exercised will still require a selling of the underlying to cover
the expense, even on cash settlement. Following the exercise the portfolio must then be
rebalanced to achieve the 80%/20% ratio. To avoid this added complication, initially setting
aside cash for margin purposes reduces the added step from call option exercise. This fairly
simple structure, or synthetic VIX, has a correlation of 0.45 to the VIX index and has a similar
skewness that captures the asymmetric payoff to volatility
25
capture large declines in the S&P 500 and becomes moot as a hedge. The returns to the
portfolio that uses actual VIX holdings are significantly higher, which is unsurprising since
trading in the actual VIX portfolio does not account for transaction costs, bid-ask spread, or
portfolio turnover.
There is a specific downside to the synthetic VIX portfolio approach and ignoring VIX
levels above 36. Such a position cannot capture the extreme positive skewness in the VIX
index and is subject to crash risk. Hedging skewness in VIX, which corresponds to large
increases in the index and typically coincides with sharp declines in the S&P 500, is very
costly and thus not worth implementing consistently over a long period. For example, from
September through November of 2008, when the S&P 500 fell over 400 points and VIX
increased from a level of 20 to the high 80s, the synthetic strategy would have provided
protection given by the put spread only, while 80% of the portfolio would have suffered large
losses. This also would have occurred in March and September/October of 2001, July 2002,
May 2010, and August 2011 when VIX spiked and the S&P fell sharply. In total, the hedge
portfolio would not be able to cover significant market losses of greater than 5% in 26 out of
the 211 months in the sample. This doesnt mean the portfolio underperformed, but it would
not be able to capture the full extent of the market losses.
It is not clear whether having a hedged portfolio that can result in positive return in
these extreme situations is worthwhile. The average loss in these 26 months for the S&P 500
is -8% while the synthetic VIX portfolios loss is -3%. That is a significant reduction in the
drawdown without the added expense that can mute normal returns. These tail risks, if
hedged consistently, would cost more than the total gains from the extreme event. This
leaves the portfolio manager with a difficult choice, namely, whether to buy expensive
27
insurance or expose the portfolio to significant jump risk. Since the replicating portfolio
captures the asymmetric relationship between volatility and returns, it can capture most of
the downside risk investors wish to hedge. The fact that the portfolio does not capture the
extreme negative skewness is a tradeoff to hedging most of the downside risk. However, a
potential alternative to hedging skewness is reverting to VIX calls.
Adjusting the portfolio to incorporate VIX calls as a hedge when VIX is above 28,
instead of the synthetic-VIX position, captures the extreme negative skewness without the
cost of S&P 500 puts. Using the calls at these levels is significantly preferable because the
cost of the calls does not dampen the potential return from volatility mean reversion. FIGURE
3 highlights this ability of the VIX calls to hedge the market decline in late 2008 as a result of
the positive skewness in the VIX index. This results in annualized return of 9.3% and a
volatility of 12.1%. As shown in Table VI, The alphas are significant for each regression
specification, and the annualized five factor model is 11.04% and is significant at the 5%
level. Additionally, the correlation between the combined synthetic-VIX and VIX call
portfolio and the VIX index is 0.58, and it is the use of the VIX calls to capture the positive
skewness in VIX that can really enhance returns. Thus, it appears that a portfolio that
combined the synthetic strategy with some VIX calls may provide the ideal hedging
combination to the long-equity portfolio.
5. CONCLUSION
Our results illustrate the feasibility and effectiveness of attempting to hold volatility
as an asset class in order to avoid market shortfalls by hedging downside risk. Since the
current asymmetric relationship between the VIX index and the S&P 500 generates the
28
strongest correlations when the market is falling, holding the VIX is a natural candidate for
hedging market risk. In fact, our results show that if VIX were investable, a portfolio
comprised of VIX and the S&P 500 would provide returns and risk levels that far outpace the
traditional buy-and-hold portfolio.
Of course, those initial results are hypothetical, and more so, not perfectly replicable
via investable VIX products because such assets price anticipated VIX mean reversion and
the mispricing cannot be arbitraged. The portfolio of holding VIX futures or the VXX ETN in
combination with the S&P 500 significantly underperforms. VIX calls, however, appear to be
a less expensive and more effective hedge of significant market downturns, outperforming a
more traditional protective strategy that utilizes S&P 500 puts. However, the expense of the
VIX calls is still a long-term hindrance on equity returns.
Understanding what is responsible for the changes in VIX is critical to assessing what
drives the asymmetric relationship between market prices and volatility. The largest
component responsible for positive VIX changes are front-month ATM puts. With this
understanding, a portfolio can be formed that holds a combination of long and short S&P 500
option contracts that best captures increases in VIX while keeping expenses low. This
synthetic VIX portfolio is quite liquid, and performs extremely well, capturing the increases
in VIX without proportionally penalizing the portfolio when the market increases. While this
portfolio does require the use of margin, it neutralizes the downside market risk investors
wish to remove while allowing for the upside gain. The aspect it cannot capture is the
extreme positive skewness in volatility. However, by holding VIX calls when appropriate,
investors may be further protected against significant shocks in the market in a much more
cost effective manner.
29
References
At-Sahalia, Y., Wang, Y., & Yared, F. (2001). Do option markets correctly price the probabilities of
movement of the underlying asset? Journal of Econometrics, 102(1), 67-110.
Bakshi, G., & Kapadia, N. (2003). Delta-hedged gains and the negative market volatility risk
premium. Review of Financial Studies, 16, 527-566.
Black, F., & Scholes, M. (1973). The Pricing of Options and Corporate Liabilities. Journal of Political
Economy, 81(3), 637-654.
Black, K. (2006). Improving Hedge Fund Risk Exposures by Hedging Equity Market Volatility, or
How the VIX Ate My Kurtosis. Journal of Trading, Spring, 6-15.
Bollen, N. P. B., & Whaley, R. E. (2004). Does Net Buying Pressure Affect the Shape of Implied
Volatility Functions? Journal of Finance, 59(2), 711-753.
Bondarenko, O. (2003). Why are Put Options So Expensive. Working paper.
Boyer, B., Mitton, T., & Vorkink, K. (2010). Expected Idiosyncratic Skewness. Review of Financial
Studies, 23(1), 169-202.
Brenner, M., Ou, E. Y., & Zhang, J. E. (2006). Hedging volatility risk. Journal of Banking & Finance,
30(3), 811-821.
Briere, M., Burgues, A., & Signora, O. (2010). Volatility exposure for strategic asset allocation.
Journal of Portfolio Management, 36(3), 105-116.
Broadie, M., Chernov, M., & Johannes, M. (2009). Understanding Index Option Returns. Review of
Financial Studies, 22(11), 4493-4529.
Carhart, M. M. (1997). On Persistence in Mutual Fund Performance. Journal of Finance, 52(1), 57-82.
Carr, P., & Wu, L. (2009). Variance Risk Premiums. Review of Financial Studies, 22(3), 1311-1341.
Chen, H.-C., Chung, S.-L., & Ho, K.-Y. (2011). The diversification effects of volatility-related assets.
Journal of Banking & Finance, 35(5), 1179-1189.
Coval, J. D., & Shumway, T. (2001). Expected Option Returns. Journal of Finance, 56(3), 983-1009.
Daigler, R. T., & Rossi, L. (2006). A Portfolio of Stocks and Volatility. Journal of Investing, Summer,
99-106.
Dash, S., & Moran, M. T. (2005). VIX as a Companion for Hedge Fund Portfolios. Journal of
Alternative Investments (Winter), 75-80.
DeLisle, R. J., Doran, J. S., & Peterson, D. R. (2011). Asymmetric pricing of implied systematic
volatility in the cross-section of expected returns. Journal of Futures Markets, 31(1), 34-54.
Dennis, P., Mayhew, S., & Stivers, C. (2006). Stock returns, implied volatility innovations, and the
asymmetric volatility phenomenon. Journal of Financial and Quantitative Analysis, 41, 381406.
Doran, J. S., & Krieger, K. (2010). Implications for Asset Returns in the Implied Volatility Skew.
Financial Analysts Journal, 66(1), 65-76.
Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal
of Financial Economics, 33, 3-56.
Giot, P. (2005). Relationships between implied volatility indexes and stock index returns. Journal of
Portfolio Management, 31(3), 92-100.
Ilmanen, A. (2012). Do Financial Markets Reward Buying or Selling Insurance and Lottery Tickets? .
Financial Analysts Journal, 68(5), 26-37.
Jackwerth, J. (2000). Recovering risk aversion from option prices and realized returns. Review of
Financial Studies, 13(2), 433-451.
Jones, T. L. (2011). A Look at the Use of VIX Futures in Investment Portfolios: Buy-and-Hold versus
Tactical Allocations. Journal of Trading, 6(2), 22-29.
Khandani, A. E., & Lo, A. W. (2011). What happened to the quants in August 2007? Evidence from
factors and transactions data. Journal of Financial Markets, 14(1), 1-46.
30
Kumar, A. (2009). Who Gambles in the Stock Market? Journal of Finance, 64(4), 1889-1933.
Litterman, R. (2011). Who Should Hedge Tail Risk? Financial Analysts Journal, 67(3), 6-11.
Liu, J., Pan, J., & Wang, T. (2005). An Equilibrium Model of Rare-Event Premia and Its Implication for
Option Smirks. Review of Financial Studies, 18(1), 131-164.
Moran, M. T., & Dash, S. (2007). VIX Futures and Options, Pricing and Using Volatility Products to
Manage Downside Risk and Improve Efficiency in Equity Portfolios. Journal of Trading, 2(3),
96-105.
Shu, J., & Zhang, J. E. (2012). Causality in the VIX futures market. Journal of Futures Markets, 32(1),
24-46.
Simon, D. (2003). The Nasdaq Volatility Index During and After the Bubble. Journal of Derivatives,
11(2), 9-24.
Szado, E. (2009). VIX Futures and options: A Case Study of Portfolio Diversification During the 2008
Financial Crisis. Journal of Alternative Investments, 12(2), 66-85.
Warren, G. J. (2012). Can Investing in Volatility Help Meet Your Portfolio Objectives? Journal of
Portfolio Management, 38(2), 82-98.
Whaley, R. (1993). Derivatives on Market Volatility: Hedging Tools Long Overdue. Journal of
Derivatives, 1(1), 71-84.
Whaley, R. (2000). The investor fear gauge. Journal of Portfolio Management, 26, 12-17.
Whaley, R. (2009). Understanding the VIX. Journal of Portfolio Management, 3, 98-105.
Whaley, R. (2013). Trading Volatility: At What Cost? Journal of Portfolio Management, 40(1), 95-108.
Windcliff, H., Forsyth, P. A., & Vetzal, K. R. (2006). Pricing methods and hedging strategies for
volatility derivatives. Journal of Banking & Finance, 30(2), 409-431.
Zhang, J. E., Shu, J., & Brenner, M. (2010). The new market for volatility trading. Journal of Futures
Markets, 30(9), 809-833.
31
TABLE I:
Correlations and Summary Statistics
TABLE I, below, demonstrates the correlation of monthly S&P 500 returns with the returns that would be realized from simultaneous investment in the
VIX index. Panel A presents results for the full sample. Panel B presents results for those months that experience an increase in the S&P 500 level while
Panel C presents results for months that experience a decrease in the S&P 500 level. Panel D shows the percentage of outcomes where the S&P 500 and
VIX had both positive, negative, and different signed returns in a given month.
Std Dev
4.27%
18.33%
Skewness
-0.63
1.17
Kurtosis
1.22
2.64
Correlation
S&P 500
VIX Index
Skewness
0.98
0.43
Kurtosis
0.53
0.41
Correlation
S&P 500
VIX Index
Skewness
-1.60
1.17
Kurtosis
3.18
2.11
Correlation
S&P 500
VIX Index
Median
1.11%
-1.35%
VIX - Return
17.07%
45.99%
28.92%
8.01%
32
-0.651
-0.283
-0.531
TABLE II:
Relation Between VIX and S&P Returns
TABLE II, below, presents the results from estimating the regression using the entire sample from 1990 to
2009:
, = + &500, + &500, 2 +
where , is the return of the VIX index on day t, &500, is the return of the S&P 500 on day t, is a constant,
and is the residual on day t. The column "Portfolio Weight of VIX to Hedge 5% S&P Drop" presents the
necessary proportions of VIX in a portfolio with the S&P 500 that yield a portfolio return of zero when the S&P
500 returns -5%, given the estimates from the regression. Robust test statistics are given in parentheses. *, **
and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.
Panel A: VIX
Regressions
Full sample
Months S&P 500
returns>0
Months S&P 500
returns<0
S&P
Returns
-2.70***
(-11.80)
Squared S&P
Returns
7.25**
(2.17)
Constant
0.02*
(1.86)
-1.61
(-1.33)
1.72
(0.13)
-0.02
(-0.74)
3.09**
(-2.53)
1.18
(0.13)
0.03
(1.17)
0.209
33
TABLE III:
Theoretical VIX Index and S&P 500 Portfolio Returns
TABLE III, below, provides return and abnormal return measures for portfolios that hold the S&P 500 and hedge using the VIX index. Panel A gives monthly results from
1996 to 2013 for threshold strategies which invest the prescribed dynamic hedge portfolio weight in the VIX and the remainder in the S&P 500 when VIX levels are below
the numerical level presented and entirely in the S&P 500 otherwise. Panel B presents analogous results for positions that invest from 2006-2013. Panel C shows results
for positions that invest from 2009-2013. Mean monthly returns are presented, as are the CAPM alpha and beta which result from the regression of monthly portfolio
returns on the value-weighted monthly market return, and the Carhart (1997) 4-factor alpha and market beta which result from the regression of monthly returns on the
three Fama and French (1993) factors (MKT, SMB, and HML) and Carharts (1997) momentum factor (UMD), which are obtained from Ken Frenchs website. The 5-Factor
model consists of the returns on an ATM or OTM straddle in addition to the factors in the Carhart model. Test statistics are given in parentheses. *, ** and *** denote
statistical significance at the 10%, 5% and 1% levels, respectively.
CAPM
MKT Beta
CAPM
Alpha
4-Factor
MKT Beta
4-Factor
Alpha
5-Factor
(ATMS) MKT
Beta
5-Factor
(ATMS) Alpha
5-Factor
(OTMS) MKT
Beta
5-Factor
(OTMS) Alpha
0.83***
0.15***
0.55**
0.14**
0.61***
0.15***
1.39***
0.21***
1.72***
(3.80)
(2.65)
(2.39)
(2.32)
(2.69)
(2.67)
(5.03)
(4.12)
(5.56)
Mean
S&P500 Return
0.70***
(2.76)
No Threshold
12.51 (-1sd) Threshold
20.2 (Mean)Threshold
27.89 (+1sd)Threshold
35.58 (+2sd)Threshold
0.74**
0.97***
0.14
0.97***
0.15
0.97***
0.16
0.97***
0.25
(2.38)
(69.94)
(1.53)
(52.37)
(1.55)
(52.99)
(1.33)
(63.35)
(1.48)
1.04***
0.79***
0.51**
0.78***
0.54**
0.79***
0.78**
0.82***
1.09***
(3.33)
(17.00)
(2.49)
(15.71)
(2.53)
(15.34)
(2.52)
(17.22)
(2.94)
1.09***
0.41***
0.71***
0.39***
0.78***
0.40***
1.51***
0.47***
1.98***
(4.02)
(4.99)
(2.78)
(4.49)
(2.92)
(4.64)
(4.14)
(5.88)
(5.16)
0.95***
0.30***
0.62**
0.28***
0.68***
0.29***
1.39***
0.35***
1.82***
(3.77)
(3.62)
(2.48)
(3.13)
(2.59)
(3.22)
(3.80)
(4.25)
(4.77)
34
CAPM MKT
Beta
CAPM
Alpha
4-Factor MKT
Beta
4-Factor
Alpha
5-Factor MKT
Beta
5-Factor
Alpha
5-Factor MKT
Beta
5-Factor
Alpha
0.87**
0.02
0.74*
-0.09
0.81**
-0.02
1.82***
0.08
2.32***
(2.42)
(0.17)
(1.95)
(-0.79)
(2.26)
(-0.21)
(3.64)
(0.93)
(4.02)
Mean
S&P500 Return
0.40
(0.83)
No Threshold
12.51 (-1sd) Threshold
20.2 (Mean)Threshold
27.89 (+1sd)Threshold
35.58 (+2sd)Threshold
0.68
0.96***
0.18
0.93***
0.19
0.94***
0.29
0.98***
0.56
(1.41)
(33.33)
(0.98)
(20.25)
(1.03)
(22.26)
(1.07)
(35.96)
(1.35)
1.01**
0.72***
0.61*
0.67***
0.63*
0.71***
1.05*
0.81***
1.81**
(2.09)
(8.17)
(1.67)
(6.47)
(1.69)
(7.13)
(1.71)
(8.59)
(2.58)
1.03**
0.38**
0.76
0.25
0.81*
0.31*
1.64**
0.45***
2.53***
(2.24)
0.96**
(2.44)
(1.65)
(1.40)
(1.85)
(1.89)
(2.20)
(3.04)
(3.29)
0.29*
0.73
0.13
0.79*
0.19
1.62**
0.31**
2.40***
(2.17)
(1.75)
(1.59)
(0.72)
(1.84)
(1.11)
(2.18)
(2.00)
(3.18)
Mean
CAPM Alpha
4-Factor Alpha
5-Factor
MKT Beta
5-Factor
Alpha
5-Factor
MKT Beta
5-Factor
Alpha
1.47**
(2.55)
1.05**
(2.63)
0.08
(0.69)
0.86*
(1.82)
-0.07
(-0.59)
1.06**
(2.39)
-0.09
(-0.80)
2.01***
(3.08)
-0.03
(-0.32)
2.47***
(3.48)
1.67***
(2.86)
1.53***
(2.92)
1.43***
(2.91)
0.76***
(8.07)
0.45***
(3.19)
0.30*
(1.99)
0.52
(1.11)
0.77
(1.37)
0.86
(1.53)
0.61***
(4.99)
0.32*
(1.89)
0.11
(0.62)
0.65
(1.36)
0.92
(1.66)
1.02**
(2.04)
0.63***
(5.16)
0.32*
(1.94)
0.09
(0.58)
1.09
(1.14)
1.68*
(1.81)
2.02**
(2.61)
0.67***
(6.44)
0.37**
(2.60)
0.16
(1.05)
1.82*
(1.76)
2.51**
(2.47)
2.75***
(3.32)
35
TABLE IV:
VIX Products and S&P 500 Portfolio Returns
TABLE IV, below, provides return and abnormal return measures for portfolios that hold the S&P 500 and hedge using the VIX futures, VIX calls, SPX Puts, and the VXX
ETN. Panel A gives monthly results from 1996 to 2013 for threshold strategies which invest the prescribed dynamic hedge portfolio weight in VIX futures and the remainder
in the S&P 500 when VIX levels are below the numerical level presented and entirely in the S&P 500 otherwise. Panel B presents analogous results for positions that invest
partially in VIX calls as hedging instruments, from 2006 to 2013.Panel C presents analogous results for positions that invest partially in S&P 500 puts as hedging
instruments. Panel D presents analogous results for positions that invest partially in the VXX ETN as a hedging instrument from 2009-2013. Mean monthly returns are
presented, as are the CAPM alpha and beta which result from the regression of monthly portfolio returns on the value-weighted monthly market return, and the Carhart
(1997) 4-factor alpha and market beta which result from the regression of monthly returns on the three Fama and French (1993) factors (MKT, SMB, and HML) and
Carharts (1997) momentum factor (UMD), which are obtained from Ken Frenchs website. The 5-Factor model consists of the returns on an ATM straddle in addition to
the factors in the Carhart model. Test statistics are given in parentheses. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.
Panel A: VIX Futures Strategy,
2006-2013
No Threshold
12.51 (-1sd) Threshold
20.2 (Mean)Threshold
27.89 (+1sd)Threshold
35.58 (+2sd)Threshold
Mean
CAPM Alpha
4-Factor Alpha
-0.59*
(-1.80)
0.41
(0.87)
-0.36
(-0.74)
-0.71
(-1.62)
-0.72*
(-1.70)
0.06
(0.60)
0.97***
(55.51)
0.76***
(12.06)
0.49***
(4.45)
0.41***
(3.31)
-0.73**
(-2.13)
-0.09
(-0.80)
-0.78**
(-2.36)
-1.02***
(-2.66)
-0.99**
(-2.53)
0.01
(0.11)
0.97***
(40.99)
0.71***
(9.41)
0.41***
(3.46)
0.30**
(2.44)
-0.68**
(1.99)
-0.08
(-0.62)
-0.78**
(-2.30)
-0.98**
(-2.56)
-0.94**
(-2.49)
Mean
CAPM Alpha
4-Factor Alpha
0.72
(0.78)
0.26
(0.53)
0.00
(0.00)
-0.19
(-0.42)
-0.21
(-0.49)
0.01
(0.01)
1.00***
(221.80)
0.95***
(48.36)
0.86***
(21.15)
0.84***
(19.30)
0.60
(0.55)
-0.12***
(-2.85)
-0.37***
(-3.65)
-0.53***
(-3.97)
-0.55***
(-4.03)
-0.05
(-0.07)
1.00***
(168.64)
0.94***
(48.42)
0.84***
(16.93)
0.81***
(16.29)
0.60
(0.55)
-0.12***
(-2.83)
-0.37***
(-3.63)
-0.53***
(-3.90)
-0.55***
(-4.05)
36
5-Factor
MKT Beta
0.06
(0.73)
0.96***
(40.39)
0.74***
(9.85)
0.44***
(3.82)
0.33***
(2.74)
5-Factor
Alpha
0.10
(0.23)
-0.09
(-0.47)
-0.46
(-1.02)
-0.52
(-0.97)
-0.47
(-0.87)
5-Factor
MKT Beta
0.14
(1.62)
0.99***
(58.63)
0.83***
(11.30)
0.54***
(4.72)
0.42***
(3.52)
5-Factor
Alpha
0.43
(0.86)
0.16
(0.87)
0.25
(0.55)
0.16
(0.29)
0.12
(0.22)
5-Factor
MKT Beta
0.08
(0.14)
1.00***
(151.12)
0.94***
(48.42)
0.85***
(19.53)
0.82***
(18.69)
5-Factor
Alpha
2.47
(1.00)
-0.13**
(-2.45)
-0.37**
(-2.56)
-0.35
(-1.59)
-0.35
(-1.55)
5-Factor
MKT Beta
0.14
(0.24)
1.01***
(172.55)
0.96***
(48.46)
0.89***
(23.51)
0.86***
(22.25)
5-Factor
Alpha
2.21
(0.95)
-0.06*
(-1.68)
-0.17
(-0.91)
-0.14
(-0.50)
-0.16
(-0.57)
Mean
CAPM Alpha
4-Factor Alpha
-0.31
(-1.39)
0.47
(1.50)
0.05
(0.17)
-0.19
(-0.73)
-0.27
(-1.13)
0.67***
(22.63)
0.99***
(342.61)
0.92***
(48.67)
0.76***
(18.97)
0.72***
(18.70)
-0.75***
(-10.37)
-0.08***
(-2.82)
-0.48***
(-6.65)
-0.66***
(-7.02)
-0.73***
(-8.42)
0.68***
(24.33)
0.99***
(289.45)
0.91***
(50.98)
0.75***
(20.34)
0.72***
(20.13)
-0.76***
(-10.60)
-0.07***
(-2.68)
-0.47***
(-6.77)
-0.65***
(-6.88)
-0.74***
(-8.33)
5-Factor
Alpha
-0.27***
(-3.68)
-0.07**
(-2.10)
-0.32***
(-2.75)
-0.21
(-1.50)
-0.28**
(2.20)
5-Factor
MKT Beta
0.72***
(59.05)
1.00***
(304.37)
0.93***
(54.89)
0.79***
(28.16)
0.76***
(28.55)
5-Factor
Alpha
-0.01
(-1.27)
-0.02
(-0.58)
-0.14
(-0.97)
0.01
(0.08)
-0.08
(-0.60)
-0.55
(-1.18)
5-Factor
MKT Beta
-0.02
(-0.16)
5-Factor
Alpha
0.36
(0.52)
5-Factor
MKT Beta
0.04
(0.40)
5-Factor
Alpha
0.96
(1.32)
-0.77**
(-2.25)
-0.81
(-1.47)
-0.62
(-1.23)
0.69***
(6.35)
0.35*
(2.00)
0.14
(0.82)
-0.48
(-0.66)
-0.08
(-0.09)
0.31
(0.40)
0.72***
(7.84)
0.40**
(2.59)
0.20
(1.31)
0.37
(0.53)
0.81
(0.80)
1.13
(1.34)
Mean
CAPM Alpha
4-Factor Alpha
-0.47
(-1.27)
N/A
0.11
(0.82)
-0.64
(-1.31)
0.01
(0.05)
0.26
(0.44)
-0.29
(-0.55)
-0.32
(-0.65)
0.79***
(10.92)
0.47**
(3.16)
0.33**
(2.03)
-0.79**
(-2.41)
-0.91*
(-1.68)
-0.75
(-1.37)
0.68***
(6.22)
0.36**
(2.03)
0.16
(0.90)
37
5-Factor
MKT Beta
0.68***
(47.37)
0.99***
(286.79)
0.92***
(50.43)
0.76***
(23.98)
0.73***
(23.86)
TABLE V:
VIX Factor Regression
TABLE V, below, shows the results from the following regressions:
ln (
) = + ,,,t + ,,,t + ,,,t + ,,,t + ,,,t + ,,,t +
1
and
ln (
Where ,,,t is the change in price of (call/put) option with a strike of and maturity at time t.
represents the change is moneyness of the option, where put moneyness is equal to S/K and call moneyness is
equal to K/S. Thus, more positive changes bring both definitions of moneyness closer to ATM. is the time
to maturity of the option, and is the interaction of all three variables. is a dummy variable equal to one if
the option is a call and zero if the option is a put. The first regression interacts the dummy variables with change
in price and moneyness variables to capture differences between call and put options. The second regression
examines the effect of price and moneyness changes of calls and puts separately on VIX changes. Each
specification clusters the standard errors by date to avoid overstating the t-statistics. Test statistics are given
in parentheses. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.
VIX
VIX >0
VIX <0
Puts Only
VIX>20.8
VIX
0.248***
0.339***
0.198***
0.203***
(7.36)
(4.91)
(7.98)
(6.67)
(9.18)
(3.81)
(4.70)
KS
331.8***
210.6***
172.5***
301.1***
448.5***
-364.4***
-636.3***
(32.68)
(13.11)
(16.07)
(27.59)
(21.61)
(33.74)
(21.62)
-1.031
-1.292
0.55
-0.922
-4.73**
-0.197
-3.983**
(0.91)
(3.12)
(0.17)
(2.62)
DM
Puts Only
VIX<20.8
VIX
Calls Only
VIX>20.8
VIX
Calls Only
VIX<20.8
VIX
0.818***
0.112***
0.346***
(1.35)
(1.25)
(0.76)
-0.140***
-0.262***
-0.043
(3.88)
(2.80)
(1.06)
KS*CD
-732.0***
-510.7***
-407.1***
(38.11)
(17.27)
(19.58)
48.47***
15.45
19.69***
38.08***
456.8***
44.17***
413.8***
(4.76)
-0.69
(4.80)
(4.61)
(4.42)
(2.64)
(3.38)
0.260***
2.96***
-2.62***
0.427***
0.533***
0.306**
0.381**
P*CD
Constant
OBS
R-squared
(2.85)
(24.20)
(31.39)
(3.63)
(3.94)
(2.33)
(2.37)
33681
15813
17718
9437
8177
9358
6682
0.57
0.39
0.35
0.61
0.55
0.65
0.57
38
TABLE VI:
Synthetic VIX and S&P 500 Portfolio Returns
TABLE VI, below, provides return and abnormal return measures for portfolios that hold the S&P 500 and hedge using a synthetic VIX portfolios
constructed from a VIX factor regression. The table presents monthly results from 1996 to 2013 for a threshold strategy which invests the prescribed
dynamic hedge portfolio weight in synthetic VIX and the remainder in the S&P 500 when VIX levels are below +1sd threshold and entirely in the S&P
500 otherwise. Mean monthly returns are presented, as are the CAPM alpha and beta which result from the regression of monthly portfolio returns on
the value-weighted monthly market return, and the Carhart (1997) 4-factor alpha and market beta which result from the regression of monthly returns
on the three Fama and French (1993) factors (MKT, SMB, and HML) and Carharts (1997) momentum factor (UMD), which are obtained from Ken
Frenchs website. The 5-Factor model consists of the returns on an ATM straddle in addition to the factors in the Carhart model. Test statistics are
given in parentheses. *, ** and *** denote statistical significance at the 10%, 5% and 1% levels, respectively.
Mean
CAPM Alpha
4-Factor Alpha
5-Factor
MKT
Beta
5Factor
Alpha
5-Factor
MKT
Beta
5Factor
Alpha
0.65***
(2.62)
0.46***
(6.85)
0.28
(1.34)
0.42***
(6.27)
0.28
(1.29)
0.42***
(6.28)
0.23
(0.90)
0.43***
(6.31)
0.48*
(1.83)
0.78***
0.29**
0.46*
0.24*
0.52*
0.24*
0.72*
0.27**
0.92**
(3.23)
(2.49)
(1.88)
(1.88)
(1.88)
(1.95)
(1.71)
(2.23)
(2.21)
39
VIX (1+SD)
40
S&P 500
1-Jul-12
1-Apr-13
1-Oct-11
1-Jan-11
1-Apr-10
1-Jul-09
1-Oct-08
1-Jan-08
1-Apr-07
1-Jul-06
1-Jan-05
1-Oct-05
1-Jul-03
VIX No Threshold
1-Apr-04
1-Jan-02
1-Oct-02
1-Jul-00
1-Apr-01
1-Jan-99
1-Oct-99
1-Apr-98
1-Jul-97
1-Oct-96
1-Jan-96
$-
$2,500,000
$2,000,000
$1,500,000
$1,000,000
$500,000
VIX (1+SD)
VXX
41
VIX Calls
S&P500 Puts
SPX
1-Jul-13
1-May-13
1-Mar-13
1-Jan-13
1-Nov-12
1-Jul-12
1-Sep-12
1-Mar-12
1-May-12
1-Jan-12
1-Sep-11
1-Nov-11
1-Jul-11
1-May-11
1-Jan-11
VIX Futures
1-Mar-11
1-Nov-10
1-Jul-10
1-Sep-10
1-Mar-10
1-May-10
1-Jan-10
1-Nov-09
1-Sep-09
1-Jul-09
1-May-09
1-Jan-09
1-Mar-09
$-
$9,000,000
$8,000,000
$7,000,000
$6,000,000
$5,000,000
$4,000,000
$3,000,000
$2,000,000
$1,000,000
VIX (1+SD)
Synthetic
SPX
42
Synthetic+ VIX_Calls
1-May-13
1-Sep-12
1-Jan-12
1-May-11
1-Sep-10
1-Jan-10
1-Sep-08
1-May-09
1-Jan-08
1-May-07
1-Jan-06
1-Sep-06
1-May-05
1-Sep-04
1-Jan-04
1-May-03
1-Sep-02
1-Jan-02
1-Sep-00
1-May-01
1-Jan-00
1-May-99
1-Jan-98
1-Sep-98
1-May-97
1-Sep-96
1-Jan-96
$-