Positional Option Trading (Wiley Trading)
Page 11
Many studies (for example, Foster et al., 1984; Bernard and
Thomas, 1989; Watts, 1978) have shown that the anomalous drift
occurs over a period of months. Almost all of the drift occurs
within 9 months for small firms and within 6 months for large
firms. And a disproportionate amount of the return occurs within
5 days of the earnings announcement (13% for small companies,
18% for medium companies, and 20% for large companies).
Unfortunately, what is not clear is why this anomaly exists.
A number of behavioral explanations have been proposed. If the
effect is predominantly caused by psychological biases, that would
point to PEAD being an inefficiency rather than a risk premium.
Some research (for example, Daniel et al., 1998) suggests that
investor overconfidence causes investors to be anchored to the
pre-earnings price and incorporate the new information only
slowly.
Another possibility is that investors don't have the time and
resources to process the new information quickly. DellaVigna and
Pollet (2009) postulate that this effect is even more pronounced
on Fridays when investor attention is at its lowest. Although this
seems like a reach or a post-hoc justification of data mining, a
long-short portfolio of stocks announcing on Friday did
outperform a portfolio based on stocks that announced on other
days (9.76% over the subsequent 75 trading days versus 5.14%).
Hirshleifer et al. (2009) also reason that information processing
capacity is a reason for the anomaly. They show that a hedged
portfolio constructed from stocks that announce on days with few
other releases underperform a portfolio of stocks that announce
on busy days (2.81% quarterly compared to 5.37%).
Bartov et al. (2000) show that firms with high institutional
ownership have lower PEAD. They suggest that this is because
these supposedly more sophisticated investors are faster to absorb
the new information. Similarly, Taylor (2011) shows that the
anomaly is largest for stocks where retail traders trade against the
direction of the initial price spike. Further, he also finds a positive
correlation between PEAD size and transaction costs, which
presumably are proportionally a lower deterrent to larger
investors due to their more sophisticated execution methods and
access to a wider range of liquidity sources.
90
There are also explanations that ascribe PEAD to a risk factor. For this to be true firms with positive (negative) earnings surprises
must become riskier (less risky) after the release. It is well
accepted in classical finance that stocks with higher returns are
riskier, and if we are looking for a rational expectation explanation
we need to accept this as true. The harder thing to explain is why
stocks become riskier immediately after the earnings report is
published. Despite attempts to do this (Bernard and Thomas,
1989) no convincing supporting evidence has been discovered.
In summary, the available evidence suggests that PEAD is an
inefficiency, albeit a very persistent one.
Trading Strategy
Because implied volatility tends to be comparatively cheap after
earnings releases (although this isn't enough to profit from delta-
neutral volatility strategies; see Chung and Lewis, 2017), this is an
effect that can be traded from the long side. For stocks with
positive earnings surprises, being long a short-dated 40 delta/10
delta call spread is usually a cheap way to leverage PEAD. For
bearish positions a 50 delta/20 delta put spread achieves the same
effect while also selling one of the most expensive parts of the
implied volatility curve.
An alternative approach is to sell a covered call or put. Although
the variance premium being collected won't be significant, this
method is guaranteed to profit if any drift occurs at all.
Confidence Level Two
These strategies are not as promising as those of level three. They
may be lacking in empirical evidence or theoretical justification.
Or they could provide more limited opportunities in some way.
Nonetheless, a trader can be reasonably confident when using
these strategies.
Trading Equity Options over Earnings
Announcements
The front-expiration implied volatility will almost always rise
significantly in the weeks before a company's earnings release.
91
After the release the implied volatility will collapse. This effect has
been known since at least 1979 (Patell and Wolfson, 1979).
Buying options before the release is profitable (Chung and Lewis,
2017). And even in cases when the options do not increase in
price, the rise in implied volatility gives a very cheap long position
where the vega profits negate most of the time decay.
I conducted a test of this effect in Sinclair (2013). Using data for
73 large cap stocks from Q1 2005 to Q3 2010, a strategy of buying
front expiration, ATM straddles 10 trading days before the
earnings release had these results:
Forty-three percent of trades were profitable.
Win size/loss size was 1.70.
Results were skewed positively (skewness of 5.4) and fat-tailed
(excess kurtosis of 76.6).
Putting a $10,000 notional size bet on each trade led to the
cumulative P/L curve of Figure 5.1.
FIGURE 5.1 Results of the long straddle strategy.
A far more comprehensive study was completed by Gao et al.
(2017). They bought delta-neutral straddles three days before the
earnings announcement. In the period 1996 through 2013, selling
a day before the release gave a return of 1.9%, selling on the
release day returned 2.60%, and holding for a day after earnings
had a return of 1.98%. (All of these returns were calculated using
92
mid-market prices.) These time periods are not entirely
independent. Some companies on the earnings date report before
the open and some after the close. So the second group includes
both pre- and post-report companies. Notably, holding the
straddle for a full day after earnings had a lower return than
selling on the release date. This indicates that the return on the
straddle is negative after earnings are announced.
The authors don't present a separate study for the post-earnings
period. But using my smaller sample, I directly tested a short
volatility strategy. As close as possible before earnings, sell the
front straddle and hope that the collapse in implied volatility
compensates for the movement of the underlying.
The results are as follows:
Sixty-four percent of trades were profitable.
Win size/loss size was 0.69.
Results were skewed negatively (skewness of negative 5.4) and
fat-tailed (excess kurtosis of 12.2).
FIGURE 5.2 Results of the short straddle strategy.
Putting a $10,000 notional size bet on each trade led to the P/L
curve of Figure 5.2.
93
There haven't been many published studies of this effect and the sample size of my study is very small, so it i
s dangerous to draw
any conclusions about the profitability of subcategories of stocks.
However, it seems that there is a small tendency for this strategy
to fare poorly with high PE stocks. Studies have shown that such
growth and value stocks do react to earnings differently and this
effect might be related (He et al., 2010). However, the difference is
so weak it can probably be ignored.
There was no significant relationship was between profitability of
the option strategies and industry or company size. However,
there is a link between the dispersion of analyst estimates (defined
as high estimate–low estimate)/average estimate) and
profitability. Trading the effect on stocks with high dispersion is
more profitable. High dispersion literally means that the analysts
disagree about the unreleased earnings, a situation of uncertainty.
The profitability of option trades with edge, either from the long
or short side, is generally tied to uncertainty.
Other signifiers of uncertainty are with higher historical
volatilities, larger historical earnings surprises, and more volatile
earnings surprises. The strategies also perform better in these
cases. Gao et al. (2017) also found that the strategy was more
profitable in stocks that were covered by fewer analysts.
There are likely two reasons for this. High uncertainty leads to a
greater increase in implied volatility, so we will be selling the
options at a good price. But we also benefit from the uncertainty in
the actual move of the stock price that happens due to the release.
Stocks move when news comes out. News is something contrary to
expectations. Uncertainty implies no coherent expectations. There
can be no “surprise” if no one agrees on what to expect.
The fact that pre-earnings long straddles are profitable makes it
unlikely that this effect is due to a risk premium. Gao et al. (2017)
postulate that one reason for the underestimation of uncertainty is
the difficulty of extracting a signal from sparse, noisy past
information. This seems plausible.
Separating the effect into the pre-earnings implied volatility
increase and the post-earnings implied volatility collapse and
stock price adjustment makes finding possible explanations easier.
The pre-earnings situation seems to be driven by similar factors to
PEAD. That is, investors can process the information only from
previous earnings results slowly and inefficiently. This is
94
consistent with the link to analyst coverage and estimate
dispersion. However, unlike the case of PEAD, there have been no
further studies to test this idea. The profitability of the post-
earnings straddle sales could be due to ambiguity aversion. But,
again, this has not been independently verified or studied.
The short side of this strategy hasn't done well recently (2016–
2018). I think this is probably a temporary, variance-driven period
of poor performance. Usually, selling options before situations of
uncertainty is a good trade. I still think that is the case here. As
any casino knows, sometimes the customers have to win.
Otherwise they won't come back.
Trading Strategy
On the long side, buy the shortest-dated straddles that span the
earnings date. Or, if the smile is relatively flat, strangles are a
higher risk/higher reward play.
On the short side, sell the shortest-dated straddles that span the
earnings date. After the earnings release, either buy the straddle
back (generally preferable if the trade is a winner), or, if the stock
has a big move, hedge the losing leg as 100 delta. Over-hedging
one side of the trade usually works due to PEAD, and the short-
dated options not giving the stock enough time to reverse. Stocks
in which the earnings number is outside of the most extreme
estimates are most likely to not reverse.
The Overnight Effect
The index variance premium is realized overnight. Muravyev and
Ni (2018) wrote a paper that studies this. Although it is very well
known that returns to index options are negative (about −0.7% a
day in terms of actual premium decay for the S&P), it turns out
that all of this comes from overnight decay. Specifically, delta-
hedged option returns are −1% overnight, and intraday returns are
positive at 0.3% per day. The anomaly holds for options of all
maturities and moneyness and also equity options.
This can be partially explained by the fact that overnight returns
are less volatile than intraday returns (since 2000 the S&P has
had an annualized volatility of 10.4% overnight and 15.9% during
the day). The effect is also consistent with the variance premium
being a mispriced risk premium. The risk is much higher during
95
the untraded (or illiquid) hours. We get paid for taking the risk of having a bet that we can't get out of or hedge if it goes against us.
Trading Strategy
Being short options overnight and flattening during the day is
impractical due to transaction costs. Instead the best way to profit
from this is to sell very short-term options that include as many
overnight periods as possible. For example, instead of selling daily
options on the morning of expiration, sell them the night before.
FOMC and Volatility
The most important news for an equity is the earnings release. The
equivalent for the broad market is the monetary policy decision of
the Federal Reserve Open Market Committee (FOMC). And, just
as company earnings releases lead to predictable pre-event price
drift and post-event volatility collapse, analogous phenomena
occur in the broad market around the FOMC announcements.
In “When No News Is Good News—The Decrease in Investor Fear
after FOMC Announcements” Fernandez-Perez et al. (2017) show
that the VIX and VIX-Futures drop significantly after the
announcements of the meeting.
From their abstract, “We find that the VIX and the VIX futures
start to decline immediately after the FOMC announcement, and
this decline persists for about 45 minutes after the announcement.
The VIX declines by about 3% on announcement days, whereas
the nearest term VIX futures contract declines about 1.4% around
the announcement.”
Similar results have been obtained by others. Nikkinen and
Sahlström (2004) showed that the VIX index declined after FOMC
announcements, CPI and PPI releases, and the Non-farm Payroll
report. Chen and Clements (2007) and Vähämaa and Äijö (2011)
replicated the results for the VIX. Gospodinov and Jamali (2012)
found both implied and realized volatility dropped after the
release and any initial market reaction. Fuss et al. (2011) showed
that German equity implied volatility decreased after the release of
GDP, CPI, and PPI data, and Shaikh and Padhi (2013) confirmed
the effect in the Indian market.
96
The stock earning's related volatility collapse has to be played with options. Implied volatility will reliably collapse following the news
releas
e, but this will generally be partially offset by an increase in
realized volatility. Trading the FOMC-related implied volatility
collapse with VIX futures avoids this problem. We can trade an
implied effect with an implied contract and have no realized
volatility exposure.
Why does implied volatility collapse? Ederington and Lee (1996)
argued that the release of prescheduled news can be viewed as the
resolution of uncertainty, and an announcement reduces the
degree of future uncertainty, and hence implied volatility, left in
the market.
This effect can be combined with a directional anomaly. Equity
market returns have been about 30 times larger on announcement
days than on normal days. Surprisingly most of these excess
returns occur before the announcement (Lucca and Moench,
2011). Stocks tend to rally strongly in the day and a half before the
announcement.
It is difficult to think of a convincing risk-based reason for these
returns. Although investors holding stocks over the
announcement period are exposed to jump risk, the excess return
occurs mainly before the announcement. So, it is possible to sell
the stocks before the news and earn the return without taking this
risk. Further, realized volatility is lower than usual before the
release. Market risk is actually lower than normal in the period
when the excess returns are generated.
It seems more likely that this is a market-wide version of the pre-
earnings drift seen in individual equities. As in that case, investors
buy stocks because of an attention-grabbing event (see Barber and
Odean, 2008; Lamont and Frazzini, 2007).
Trading Strategy
Buy S&P 500 futures two days before the FOMC release. Sell these
before the announcement and at the same time sell VIX futures.
Cover the short VIX futures 45 minutes after the announcement.
The Weekend Effect
97
Equity options decay more than expected over weekends. The
variance premium is probably the most important effect in option
trading. On average options are overpriced. But the premium
doesn't accrue consistently: some periods are better than others.
In particular, options lose more value over weekends than over