and options give protection against these jumps in a way that a dynamic
hedging scheme cannot, making them more attractive to buyers (hedgers or
speculators). The non-redundancy of options can be seen as both a cause and
consequence of the variance premium.
Trading Restrictions
Many retail traders have restrictions placed on them by their brokers. It is
common for them to be forbidden from selling naked options. This means that
an entire class of speculators can only be long volatility, thus driving the
variance premium higher.
Market-Maker Inventory Risk
Contrary to popular belief, bookies do not try to completely balance their risk.
They often speculate on a team to win. Similarly, option market-makers usually
have speculative views on the markets and use the income from market-making
operations to smooth the losses from these positions. There are far fewer
market-makers now than there were 10 years ago. Increasing automation means
fewer people are needed. On the floor, a trader could cover two or three stocks
but now it is routine for a single trader to trade hundreds of stocks. But the
overall profits to the community are still enormous. Spreads have narrowed, but
this has been compensated for by increasing volumes. As long as a market-
maker can stay in business, she will eventually be successful.
The best trading opportunities for liquidity providers are in times of turmoil.
Spreads widen. Volume soars. Customers are not as price sensitive as usual. So,
it is imperative that market-makers can trade aggressively in these situations.
And this can only happen if they aren't stuck with a loss from the move. Ideally, they will have a nice profit and can trade freely and with no restrictions. This means that they are almost always long teeny options. A standard risk-73
management heuristic for market-makers is always to be net long options. They
are aware that options are overpriced, but they need them as insurance. Not
inventory insurance: business insurance.
Path Dependency of Returns
Grosshans and Zeisberger (2018) show that investors also care about exactly how their returns are realized. They performed surveys that asked people to
imagine that they had six stocks. Three made 10% and three lost 10%. But each
of the three had different paths: up-down, a linear path, or down-up. These are
shown stylistically in Figures 4.9 and 4.10.
FIGURE 4.9 The three different positive return paths.
FIGURE 4.10 The three different negative return paths.
For both winners and losers, the participants were happiest when prices first
declined and then rose. People were even slightly disappointed when stocks rose
and then fell but were still winners! People are happiest when they feel that they have recovered from adversity, snatching victory from the jaws of defeat.
74
Although this is only survey data it gives another plausible reason for the
variance premium.
Think about a $100 stock and the $100 strike call and put. Assuming no interest
rates and a volatility of 30%, both the 1-month call and put will each be worth
$3.43. One trader sells the put, and another buys the call. Consider what
happens if the stock jumps from $100 to $106.86 right before expiration. Each
trader makes $3.43 at expiry, but the P/L evolves slightly differently over time for the buyer and seller. These return paths are shown in Figures 4.11 and 4.12.
The largest PL difference is $5.09 on the day before the jump.
The long option maintains the ability to “snatch victory from the jaws of defeat.”
People prefer this, so options will tend to be overpriced, hence the variance
premium.
This is really a gamma effect. The short put steadily collects theta, but the long call wins because of gamma. Redemption comes from the possibility of extreme
price moves due to high gamma. This explains why the variance premium is
greatest for short-dated options and also why it tends to be higher when
volatility is low.
FIGURE 4.11 The P/L for a short put, with a stock jump at expiration.
FIGURE 4.12 The P/L for a long call, with a stock jump at expiration, 75
The Problem of the Peso Problem
It is possible that options are actually not overpriced. Perhaps volatility and the skew are fairly priced, and the apparent variance premium is due to the fact that the events option buyers are insuring against haven't occurred in our sample
period but will at some point in the future.
It is hard to see how an argument based on “just you wait” can ever be refuted.
We certainly know that in the history we have, implied volatility has been
overpriced. It is possible that someday an event will occur that is so large that all option-selling profits will be eradicated. It also seems unlikely. As with all trading decisions, we can either assume history will be an accurate guide to the future or it won't. “This time is different” is an appealing idea because it means we don't have to do any studying of how things have behaved in the past, but it
is rarely true.
Conclusion
Implied volatility tends to be an upward-biased estimator of the future realized volatility. This is the most important empirical fact to a volatility trader. The effect applies to most underlying situations and has existed for as long as we
have data to look at. There are a large number of economic, distributional, and
psychological reasons for the variance premium. Although it could be
diminished by the emergence of more institutions trying to capture it, I doubt
that it will ever disappear completely.
Summary
Historically it has been profitable to be short options. There are persuasive
arguments that suggest that this will continue.
Floor trader sayings: “If in doubt, hands out” and “Whenever an option
trades good market-makers will know if they want to buy or sell at that
price, and if they are unsure they should sell.”
76
CHAPTER 5
Finding Trades with Positive
Expected Value
In Volatility Trading I wrote, “There is no attempt here to give a
list of trading rules. Sorry, but markets constantly evolve, and
rules rapidly become obsolete. What do not become obsolete are
general principles. These are what I attempt to provide. This
approach isn't as easy to digest as a list of magic rules, but I do not
claim markets are easy to beat, either.”
This is still true.
But I am going to give a list of edges. Edges aren't rules. And the
difference is important. A rule is a defined way to monetize an
edge, whereas the edge is an observed phenomenon that could
potentially be profited from in many ways. Edges can persist.
Aside: Crowding
There is a common perception that crowded trades will have
diminishing performance. Whether this is always true is unclear,
and it was addressed in the context of factor investing by Baltas
(2019).
First, it is unclear what exactly is meant by crowded. Sometimes it
is taken to mean the size of a particular subindustry (e.g., high-
frequency trading firms), who are then exposed to the effects of a
broad unwinding (such as the “quant meltdown” of August 2007
or the volatility ETN problems of February 2018). It can also
relate to m
arket capacity. Generally, the idea encapsulates a
number of situations in which the unintentional coordinated
actions of traders create feedback effects.
Exactly how these feedback events unfold depends on the
dynamics of the specific strategy. Baltas classifies strategies into
either convergence or divergence strategies. Convergence
strategies, such as value investing, have a natural target price. This
creates a stabilizing effect. Crowding will help the realization of
profits in convergence strategies.
77
Divergence strategies, such as trend-following, have no anchoring price. The more these assets go up, the more are bought. This
creates a destabilizing feedback effect. In the short term this will
help profits, but it will eventually lead to a bubble and a
subsequent crash.
Baltas postulated that crowdedness could be quantified by looking
for increases in co-movements of assets beyond their beta (an idea
first stated by Cahan and Luo, 2013). For example, if all value
stocks started more strongly moving together, this would be an
indicator of a crowding increase in this factor.
Baltas tested his convergence/divergence idea by using global
equities data from September 2004 to May 2018, the constituents
of the S&P GSCI Commodity Index from January 1999 to May
2018, and 26 currency pairs from January 2000 to May 2018. For
the stocks, he studied the value, size, momentum, quality (return
on assets), and low beta factors. For commodities he looked at
momentum. And for currencies he examined momentum and
value (defined by purchasing power parity). His thesis about how
crowding affects various strategies was broadly confirmed. In the
presence of a natural target, crowding stabilized markets. If there
was no natural target value, crowding was a destabilizing factor.
An important caveat to these results is that the presence of
leverage can be overwhelming regardless of the specific trade
dynamics. For example, LTCM employed an ostensibly mean-
reverting strategy, trading the spread between off-the-run and on-
the-run treasuries. They employed huge leverage, and so did their
prime brokers who copied the trade. When the spread widened,
traders were forced to deleverage. This pushed the spread further
against them creating a destabilizing feedback loop. What should
have been a stable trade was turned into an unstable one by excess
leverage.
An example of products being caught in a destabilizing feedback
loop were the VIX ETN's in February 2018.
On Friday, February 2, the S&P 500 dropped 2.1% and the VIX
rallied 28.5% from 13.47 to 17.31. This was a large move: at the
time it was the 34th largest in history. But Monday, February 5,
was truly exceptional. The S&P 500 dropped 4.2%, but the VIX
rallied 115.6%. This was the largest VIX move ever and nearly
twice the previous record of 64% (and in that case the move was
only from 11.15 to 18.31). As the VIX was only implying a daily
78
move of about 1% the S&P 500 return was a genuinely large event.
However, the response of the VIX was well out of line with an
underling move of that size. A linear regression model linking the
S&P 500 returns to the VIX returns predicted that a 4.2% drop in
the equity index would correspond to a 12.8% move in the VIX.
The actual VIX increase was nine times the expected amount.
On the next day the VIX had its largest range ever, with a low of
22.42 and a high of 50.3 (the 70th highest value recorded).
I've seen nonpublic results for volatility fund returns in these few
days ranging from up 25% to down 95%. And some of the losers
were large funds. One operation with assets close to a billion
dollars lost 89%. The median was a loss of about 30%. Some of
these volatility funds lost money by being short index options but
the option space losses appear to have been relatively benign
compared to those experienced in the VIX ETNs.
The first volatility ETN was VXX, which was listed in 2009. It was
designed to match the returns of a (hypothetical) 30-day VIX
future. In 2010, XIV was launched with the stated aim of
delivering the returns of a short position in the 30-day VIX future.
These, and the other similar products that followed, proved
immensely popular. First, they gave investors who were unable to
trade futures a way to speculate on implied volatility. Second, their
returns were relatively predictable. This is due to the contango
effect in the VIX futures. To maintain a 30-day notional exposure,
the manager of VXX must sell front month futures and buy second
month futures. As about 80% of the time the second future trades
at a premium to the first, this means that the VXX rebalancing
process usually must sell low and buy high. Conversely XIV
benefits from this effect. As a result, since its launch until the end
of 2017, VXX had decayed from a split-adjusted price of 107090 to
27.92. XIV had increased from 9.56 to 134.44.
Short volatility products had become more popular in 2016 and
2017 because the realized volatility of the S&P 500 index was very
low and the contango decay was very high. Open interest
increased enormously and resulted in crowding in these products
prior to the crash in February 2018. In VXX alone, short interest
increased nearly 1300% from the end of 2013 to the end of 2017.
It is obvious that a 100% rally in the VIX 30-day future would
drive XIV to zero; however, conditions for termination (named
79
acceleration in the prospectus) didn't even require this. From the prospectus,
“…an Acceleration Event includes any event that adversely
affects our ability to hedge or our rights in connection with the
ETNs, including, but not limited to, if the Intraday Indicative
Value is equal to or less than 20% of the prior day's Closing
Indicative Value.”
In this eventuality,
“… you will receive a cash payment in an amount (the
‘Accelerated Redemption Amount’) equal to the Closing
Indicative Value on the Accelerated Valuation Date.”
It is important to stress that XIV is an ETN, not an ETF. An ETF
owns the stocks, bonds, or commodities that make up the
portfolio, whereas the ETN is merely a note that pays the return
on the portfolio. Whether and how the issuers hedge their
obligations is up to them. This means that there is no direct way to
create an arbitrage between the ETN and its fair value. This led to
severe dislocation between the fair value of XIV and its price on
Monday afternoon. By the close, the one-month future had risen
by 45%, yet XIV had only dropped by 15%. Directly after the stock
market close the VIX futures spiked to an increase of 100% on the
day, which triggered the acceleration event in XIV.
This after-hours jump was not due to any nefarious manipulation.
Around the close, ETNs rebalance their exposures to the VIX. So,
on this day, short VIX ETNs needed to buy futures to reduce their
exposure and lo
ng VIX ETNs needed to buy VIX to increase
exposure. This severe buying pressure created a large imbalance
and drove the price higher. Even if the actual product issuers were
hedged with swaps, the counterparties to those agreements would
have needed to hedge.
Lessons
It is never good to be so leveraged that you are forced to exit a
trade.
ETNs are more dangerous than ETFs.
It is dangerous to be in a strategy or product that has to make
trades and has no discretion in doing so.
80
It is even more dangerous when other people know what you
need to do and can push the market against you.
“The day you say you have to do something, you're screwed.
Because you are going to make a bad deal.”
—Billy Beane, general manager of the Oakland As from
Moneyball by Michael Lewis
Trading Strategies
I assign each edge a rating on a scale of one to three with three
being the best. This is a subjective score based on the amount of
empirical and theoretical support, the longevity of the effect, and
the volatility of the associated trade results. I also suggest ways to
implement each effect. These are also subjective.
Confidence Level Three
These strategies are based on effects that have been well
documented over either many markets or time periods. They also
have convincing theoretical bases. These can form the majority of
a trader's strategy portfolio.
Implied Volatility Term Structure as a Predictor It is
well known that the term structure of commodity futures is a
predictor of future returns (see, for example, Erb and Harvey,
2006; Gorton and Rouwenhorst, 2006; Gorton et al., 2013). When
a commodity has a term structure in contango (long-dated futures
more expensive than short-dated futures), it is profitable to short
the futures. And when futures are in backwardation (long-dated
futures cheaper than short-dated futures), it is profitable to buy
futures. Essentially, the cash price is a better predictor of futures
prices than the futures are. The way convergence happens at
expiration is that the futures move toward the cash price. This is
Positional Option Trading (Wiley Trading) Page 9