Date post: | 29-Dec-2015 |
Category: |
Documents |
Upload: | fariz-febrianto |
View: | 22 times |
Download: | 0 times |
Electronic copy available at: http://ssrn.com/abstract=1974753
0
Fundamental Analysis and Option Returns
Theodore Goodman Purdue University
Monica Neamtiu
University of Arizona
and
X. Frank Zhang Yale University
January 2013
Abstract This paper investigates whether fundamental accounting information is appropriately priced in the options market. We find that fundamental accounting signals exhibit incremental predictive power with respect to future option returns above and beyond what is captured by implied and historical stock volatility, suggesting that the options market does not fully incorporate fundamental information into option prices. Transaction costs substantially reduce the overall profitability of hedge strategies that exploit the information in these fundamental accounting signals, but the strategies still earn economically and statistically significant returns for options with low transaction costs. Key words: Fundamental analysis, return, volatility, accounting signals JEL: G11, G12, G13, G14, M41 We thank Stan Markov (FARS discussant), Cathy Schrand, Mark Trombley, and workshop participants at Cornell University, Fudan University, HKUST 2012 Accounting symposium, Yale University, The University of Texas at Dallas, University of Toronto, and the 2012 FARS conference for helpful comments and suggestions. Zhang thanks the Yale School of Management for financial support.
Electronic copy available at: http://ssrn.com/abstract=1974753
1
1. INTRODUCTION
Extensive work on fundamental analysis has examined the association between
accounting data and future stock returns (e.g., Ou and Penman 1989; Bernard and Thomas 1990;
Holthausen and Larcker 1992; Sloan 1996; Abarbanell and Bushee 1998; Piotroski 2000;
Beneish et al. 2001). Typically, the motivation for this line of research is that accounting data are
informative about a firm’s expected future cash flows and that stock investors do not fully
impound this information into stock prices. This paper explores another dimension of
fundamental analysis – the extent to which market participants can use accounting information to
evaluate the volatility of a firm’s operations and whether the options market appropriately prices
this information.
Compared with prior work on fundamental analysis, our paper represents two major
innovations. First, we apply fundamental analysis to the options market rather than to the stock
market. The options and stock markets have their own distinct features and clienteles. On one
hand, the leveraged nature of option contracts attracts sophisticated investors who wish to exploit
public and private information. On the other hand, several institutional features of the options
market make it less efficient than the stock market. For example, an option contract based on a
firm’s stock typically has considerably lower trading volume than the stock itself (Battalio and
Schultz 2006; Roll et al. 2010). Options markets also have relatively high transaction costs (e.g.,
bid–ask spreads) that may impede arbitrageurs from ensuring that the option prices appropriately
reflect all available information (Fleming et al. 1996; Pool et al. 2008). Second, and more
importantly, we focus on the volatility channel rather than the expected cash flow channel, which
is the focus of prior studies on fundamental analysis.1 Volatility has a direct numerator effect in
1 Prior work on fundamental analysis typically examines whether investors under-react or over-react to fundamental information (e.g., Bernard and Thomas 1990; Sloan 1996). These papers suggest that stock returns are predictable
Electronic copy available at: http://ssrn.com/abstract=1974753
2
determining option prices, a distinct feature that differentiates the options market from the stock
market.2 Our research provides insight into the extent to which investors incorporate
fundamental volatility information from accounting signals into option prices.
In line with the finance literature on the volatility channel (e.g., Goyal and Saretto 2009),
we focus on one specific derivative contract: an at-the-money straddle. A straddle contract
includes both a call option and a put option (both of which have an exercise price close to the
prevailing stock price), resulting in a payoff that is a function of the absolute price movement in
the underlying equity security. As the payoff from a straddle is not directional with respect to the
nature of news, the fundamental information particularly relevant for a straddle differs from
information relevant for the stock market.
We draw on a broad set of signals related to a firm’s fundamental volatility. We include
four short-term signals based on the information reported in the firm’s most recent quarter
related to earnings (magnitude of surprise, incidence of loss), accruals (magnitude of accruals),
growth (magnitude of sales growth and asset growth), and DuPont measures (magnitude of
changes in profit margin and changes in asset turnover). We also calculate long-term signals
based on the standard deviation of earnings, accruals, growth, and DuPont measures over the
previous five years. Finally, we include information from the previous fiscal year on the firm’s
dividend policy. We synthesize our collection of fundamental signals into a single measure of the
expected benefits that could accrue to an investor from holding a straddle position that matures
in one month.
because investors do not fully price fundamental information into the numerator of stock prices (expected cash flows). 2 In the stock market, the link between volatility and expected stock returns is weak at best. Theoretical models suggest that volatility captures risk and thus should be positively correlated with expected stock returns. However, empirical studies find a weak negative correlation between volatility and future stock returns (Ang et al. 2006; Zhang 2006). In contrast, in the option market, the numerator or payoff from owning a straddle is directly tied to the volatility of stock prices.
3
We begin our empirical analysis by examining the hedge returns that are generated from
a trading strategy based solely on fundamental signals. A hedge portfolio with a long position in
high fundamental volatility options and a short position in low fundamental volatility options
yields an average of 16.4% per month, suggesting that the options market does not fully
incorporate fundamental information into option prices. The magnitude of the hedge return is
similar to that of Goyal and Sarreto’s (2009) historical volatility strategy. When fundamental
volatility is high, implied volatility is too low. As a result, option prices are too low and future
option returns become positive. By the same token, future option returns tend to be negative
when fundamental volatility is too low relative to implied volatility. Furthermore, we show that
the information contained in fundamental signals regarding future option returns is largely
orthogonal to that contained in Goyal and Sarreto’s (2009) historical volatility signal, with the
correlation between these two signals around 20%. Our multivariate analysis finds that
fundamental signals have incremental predictive ability for future straddle returns after
controlling for the elements of Goyal and Sarreto’s (2009) trading strategy (historical volatility
and implied volatility).
Importantly, transaction costs are high in the option market. The monthly hedge return
drops from 16.4% to -1.6% (t = -0.90) if we assume the actual bid-ask spread to be equal to the
observed closing bid-ask spread.3 We interpret these results in the framework of limits to
arbitrage (e.g., Shleifer and Vishny 1997). Accounting information does not appear to be fully
impounded into options prices, an inefficiency that exists partially because it is costly to pursue a
strategy based solely on fundamentals. While it is too costly to execute this strategy across all
firms, we find evidence of positive after-transaction-cost hedge returns for firms with low levels
3 In a similar vein, the hedge return to Goyal and Saretto’s (2009) historical volatility strategy drops from 18.8% per month to an insignificant level (RET = -1.3%, t = -0.57). Arguments can be made that the closing bid-ask spread overstates the transaction costs, which are discussed in more detail in Section 4.3.
4
of transaction costs. When partitioning the sample into three groups on the basis of a transaction
cost proxy, we find the hedge return after transaction costs to be 7.0% (t = 2.94) for large firms,
11.2% (t = 4.51) for options with low bid-ask spreads, and 12.4% (t = 5.35) for options with
large trading volume.
To provide more evidence that the information from fundamental signals complements
the information in historical volatility, we consider a strategy that combines these two
approaches. Investors have multiple signals related to the value of a straddle contract and
naturally combine these signals in their trading strategies. We find that adding fundamental
signals to Goyal and Saretto’s historical volatility trading strategy improves the hedge return
after transaction costs from -1.3% (t = -0.57) to 7.4% (t = 2.15) per month. Thus, incorporating
fundamental signals has a meaningful marginal effect on the performance of the trading strategy
shifting the after-transaction cost portfolio returns from an insignificant level (based on the
historical volatility) to a positive and significant level (based on the combined fundamental
signals and historical volatility).
Our research contributes to existing literature in multiple ways. First, our study is the first
to use fundamental analysis to predict option returns. By examining the returns from various
straddle positions, we provide direct insight into the extent to which fundamental volatility
information is priced efficiently in the options market. Our results indicate that accounting-based
fundamental signals are highly correlated with future straddle returns, suggesting that
fundamental information is not efficiently impounded into options prices. Our work also
responds to the call by Richardson, Tuna, and Wysocki (2011) to explore the role of fundamental
analysis in the pricing of credit derivatives. While we do not specifically explore credit
derivatives, we provide insight into the use of accounting information to price option derivatives
5
related to a firm’s stock price. Second, our research provides new insight into the type of
information that can be gleaned from accounting signals and used in fundamental analysis. Prior
work on the stock market has focused on the ability of accounting signals to provide information
about future cash flows. In contrast, we examine fundamental signals that predict volatility,
which is uniquely relevant to option returns and the options market, suggesting that different
mechanisms (cash flow vs. volatility) exist between stock and options markets.4
The remainder of the paper is organized as follows. The next section provides a review of
the literature on implied volatilities and fundamental analysis. Section 3 describes our variable
measurement and sample data. Section 4 presents our main results. Section 5 provides sensitivity
analyses, and section 6 concludes.
2. PRIOR LITERATURE AND HYPOTHESIS DEVELOPMENT
2.1 Option Returns
A growing body of research has examined option returns to make inferences about
expected returns and market efficiency.5 Early work on option returns focused on the returns to
option positions based on indexes (e.g., an S&P 100 index call option). For example, Coval and
Shumway (2001) provide a theoretical and empirical analysis of the expected returns associated
with option positions. They explain that owing to the leverage implicit in an option, call (put)
4 Our study also helps to better understand how the capital markets price volatility in general. If the volatility associated with fundamental signals is diversifiable, then theoretical models suggest that equity investors would not use these signals to determine expected returns. If fundamental volatility provides insight into the systematic component of volatility and thus is relevant for expected returns, empirical tests attempting to identify an association between these signals and realized stock returns would lack power owing to the small variance of expected returns. These two issues partially explain the mixed empirical evidence on the association between volatility proxies and realized returns (e.g., Fama and French 1992; Ang et al. 2006; Zhang 2006. In contrast, our research on option returns identifies a setting where signals related to either systematic or idiosyncratic volatility are highly relevant. Thus, by examining option returns we provide additional insight into whether investors efficiently use accounting information that is informative about volatility in the capital market. 5 Alternatively, several papers have explored the rationality of options investors by examining the ability of implied volatility to predict future realized volatility. For example, Christensen and Prabhala (1998) find that implied volatility outperforms historical realized volatility as a predictor of future volatility.
6
options have higher (lower) expected returns than the underlying equity securities because these
derivatives have higher (lower) exposure to risk. They confirm these predictions with empirical
analysis of S&P 100 index options. In addition, they observe that straddle positions that are
insensitive to market risk (zero-beta straddles) have negative average returns, in contrast to the
prediction from existing asset-pricing models that these securities should have an expected return
equal to the risk-free rate, raising questions about the efficient pricing of option contracts.
More recently, researchers have explored the returns from options based on individual
equity securities. For example, Goyal and Saretto (2009) find that the difference between implied
and historical volatility can predict straddle option returns. They argue that implied volatility is
incorrect when it deviates too much from historical volatility, as volatility tends to be quickly
mean-reverting. As a result, straddle option returns tend to be positive when implied volatility is
below historical volatility (implied volatility is too low) and negative when implied volatility is
above historical volatility. While conventional wisdom holds that options investors are more
sophisticated, the abnormal returns to the trading strategy in Goyal and Saretto (2009) raise
questions as to how efficiently option prices incorporate publicly available information about
volatility.
Following Goyal and Saretto (2009), several concurrent papers explore the cross-section
of option returns.6 Choy (2011) provides evidence that a firm’s zero-beta straddle positions have
more negative returns when retail investors account for a greater proportion of that firm’s
trading, a finding consistent with retail investor trades resulting in option prices where implied
volatility is not a sufficient statistic for future realized volatility owing to behavioral biases.
6 Rather than examining the cross-section of individual option returns, Driessen, Maenhour, and Vilkov (2009) investigate the importance of the correlation between the assets that compose an index by comparing the return from holding an index option position with the return from holding the individual options within that index.
7
Other papers explore the determinants of put and call returns, but not straddle returns.7 We add
to this growing literature by examining whether options investors effectively incorporate
accounting-based fundamental signals into option prices.
2.2 Accounting Information, Volatility, and Option Returns
A large literature in accounting examines the extent to which investors effectively
interpret and price financial accounting information, although this literature has focused on the
predictability of future earnings and future stock returns. A number of papers have suggested that
accounting-based signals or fundamental analysis could generate abnormal returns (e.g., Bernard
and Thomas 1990; Sloan 1996; Ou and Penman 1989; Holthausen and Larcker 1992; Abarbanell
and Bushee 1998; Piotroski 2000). On the volatility side, the literature shows that a firm’s
fundamental volatility determines (but does not fully explain) stock price volatility (Shiller 1981;
Scheinkman and Xiong 2003; Paster and Veronesi 2003; Callen 2009). The correlation between
fundamental volatility and stock volatility creates the possibility for fundamental analysis to play
a role in predicting stock volatility.
While much of the literature on financial statement analysis has focused on the prediction
of future earnings and future stock returns, research also examines whether accounting measures
provide information about future uncertainty or the magnitude of future price movements. In
direct relation to our study, Beneish et al. (2001) show that fundamental signals, such as
earnings- or sales-based variables, can predict future extreme price movements, either upward or
downward, after controlling for market-based signals.
7 Boyer and Vorkink (2011) develop a measure of the ex-ante skewness associated with an option’s return and find that this measure is negatively associated with both put and call option returns. Christoffersen, Goyenko, Jacobs, and Karoui (2011) provide evidence that illiquidity in the options market is positively associated with both put and call option returns. However, illiquidity related to a firm’s stock trading is negatively related to both put and call option expected returns.
8
Several recent accounting studies have also explored the link between accounting
information and options markets with an emphasis on implied volatilities. Rogers, Van Buskirk,
and Skinner (2009) find that the implied volatility values increase after managerial forecasts,
particularly when the forecast conveys bad news. Dubinsky and Johannes (2006) find that the
implied volatility imbedded in a firm’s options tends to change when an earnings announcement
occurs, suggesting that options investors understand the potential for a material jump in price at
an earnings announcement. Barth and So (2009) explore whether accounting information is
associated with the gap between implied volatility and the subsequent realized volatility during
an earnings announcement window. They find that firms with losses or more volatile earnings
are more likely to have implied volatilities that are higher than the subsequent realized
volatilities at the earnings announcement and interpret the difference as a risk premium.8 None
of these papers examines the link between accounting signals and future option returns,
especially after controlling for market-based signals used in the finance literature.
2.3 Hypothesis Development
Building on the prior literature on accounting signals and future price volatility, this
paper examines the role of fundamental signals in predicting option returns. The financial
reporting system produces a rich set of fundamental variables that capture the uncertainty or
volatility of a firm’s operation. Historical stock volatility and implied volatility in option
contracts may not fully reflect such underlying fundamental volatility, which manifests in the
future. Similar to Goyal and Saretto (2009), who suggest that options investors under-react to
8 Pan (2002) suggests that investors price options with a risk premium to account for jumps in prices that could occur in the future (i.e., jump risk). This additional risk results in greater values for implied volatility values, which exceed the subsequent realized volatility. A premium related to uncertainty over future volatility may also partially explain the observed negative performance of a straddle position over time (Coval and Shumway 2001). We further address the risk issue in Section 4.2.
9
historical volatility (i.e., ignore the role of historical volatility in a mean-reverting process), we
posit that option implied volatility may temporarily deviate from fundamental volatility and, as a
result, fundamental signals predict option returns. This conjecture leads to the central prediction
of our paper: historical fundamental signals predict option returns.
In tests of our hypothesis, two issues are important to address, both conceptually and
empirically. First, we must show that fundamental signals convey incremental information about
future option returns beyond what is captured in historical volatility, which the finance literature
has shown to predict option returns. Historical volatility is a noisy measure of a firm’s
underlying volatility, leaving room for fundamental volatility to play a role. Second, we must
show that predictable option returns are not due to higher risk borne by options investors.
3. RESEARCH DESIGN AND SAMPLE DATA
3.1 Measurement of Individual Fundamental Signals
We explore a number of fundamental signals that are related to fundamental volatility
and the volatility of stock price movements. Section 3.1 discusses the measurement of individual
fundamental signals. A timeline for the measurement of the individual fundamental signals is
included in Panel A of Figure 1. Section 3.2 discusses the aggregation of these individual signals
into a single score.
(1) Short-term Signals – Earnings
Our first category of fundamental volatility signals is a collection of earnings signals. We
first consider the magnitude of the earnings surprise (|∆EARNq|), measured as the absolute value
of earnings changes relative to four quarters ago scaled by market value of equity (the Appendix
provides detailed definitions). As earnings represent a firm’s bottom-line performance, earnings
surprises capture the volatility of fundamental performance. The literature shows that larger
10
earnings surprises are related to larger price movements (Ball and Brown 1968; Bernard and
Thomas 1990). Following Beneish et al. (2001), we also consider whether the firm recently
incurred a loss (LOSS), which is defined as a dummy variable with the value of 1 if earnings
before extraordinary items are negative in quarters q, q-1, q-2, or q-3. Existing research indicates
that the process for valuing loss firms differs from that for valuing profit firms. For example,
losses are less informative than profits about the firm’s future prospects (Hayn 1995), implying
that loss firms have poorer information sets and greater volatility. We define the short-term
earnings signal (EARNINGS_ST) by taking the average of the decile rank of |∆EARNq|,
transformed to be on a scale of [0, 1], and the LOSS dummy.
(2) Short-term Signals – Accruals
Our second category of fundamental volatility signals is based on accruals. Prior work on
the accrual anomaly (Sloan 1996; Mashruwala et al 2006) indicates that firms with extreme
changes in working capital have higher stock price volatility. Accruals are informative about a
firm’s fundamentals. Both high and low working capital accruals are associated with extreme
stock price movements. We measure accruals (ACCRUAL_ST) as the decile rank of seasonal
differences in the primary working capital accounts (accounts receivable plus inventory less
accounts payable) scaled by total assets, transformed to be on a scale of [0, 1].
(3) Short-term signals – Growth
Our third category of fundamental volatility signals is a collection of growth signals. We
first consider sales growth, defined as the absolute value of seasonally adjusted quarterly sales
growth (|SGRq|). Analogous to earnings capturing a firm’s bottom-line performance, sales
represent the top-line performance. As a result, the magnitude of sales growth captures the
volatility of a firm’s fundamental performance. We complement sales growth with asset growth,
11
measured as the absolute value of seasonally adjusted quarterly asset growth (|AGRq|). In
contrast to sales, which are a flow variable, assets are a stock variable and represent the scale of a
firm’s operation. Naturally, assets growth captures fundamental volatility, with both large
positive and large negative asset growth signaling more volatile fundamentals. Our short-term
growth signal (GROWTH_ST) is the simple average of decile ranks of |SGRq| and |AGRq|, each
transformed to be on a scale of [0, 1].
(4) Short-term signals – DuPont analysis
DuPont analysis is a commonly used technique to evaluate asset turnover or profit margin
through which a firm generates a return on its assets. Changes in these variables could indicate
instability in the firm’s strategy that would translate into fundamental volatility. Soliman (2008)
provides evidence that changes in profit margins and asset turnover are associated with future
earnings and future returns. Thus, we expect that absolute changes in DuPont components
provide information about the absolute value of changes in fundamental value. We define short-
term DuPont measure (DUPONT_ST) as the simple average of decile ranks of the changes in
asset turnover and profit margin, each transformed to be on a scale of [0, 1].
(5) Long-term Signals
In addition to using the most recent realization to calculate short-term signals, we also
consider long-term signals on the basis of a long firm-specific time series. For each
corresponding short-term signal, we calculate fundamental volatility using the standard deviation
of that signal over a five-year window, with a minimum of ten quarterly observations.
(6) Recent Dividend Policy
A large body of research (e.g., Guay and Harford, 2000) has examined the signaling role
of dividends. Specifically, a firm’s decision to pay a dividend signals a commitment to maintain
12
that dividend, implying a level of stability in the firm’s operations. Thus, managers can use a
dividend to signal lower fundamental volatility. We calculate a dividend-based measure (DIV) as
dividends in prior fiscal year multiplied by minus one, scaled by prior year’s average assets.
3.2 Fundamental Signals and Future Stock Price Movements
We aggregate individual fundamental signals into a single score that reflects the
information they convey about the magnitude of stock price movements relevant to pricing a
straddle position. As our analysis explores the returns an investor can earn from buying a
straddle contract and holding it to maturity one month later, we examine the extent to which our
signals are related to the absolute value of future monthly returns. We focus on absolute return
during a month, because this return is equivalent to the value that could be realized from an at-
the-money straddle the investor purchased at the beginning of the month.9
We begin by calculating the average absolute value of monthly returns in the three
months following the month during which a firm announces its earnings (|RETq+1|). To be
consistent with our fundamental signals, we transform |RETq+1| into a scaled decile rank where,
to avoid a look-ahead bias, the rank calculation is based on the return distribution in the previous
calendar quarter. Then, we estimate the historical association between our fundamental signals
and the subsequent price movements using the following equation:
�𝑅𝐸𝑇𝑞+1� = 𝜃0 + 𝜃1𝐸𝐴𝑅𝑁𝐼𝑁𝐺𝑆_𝑆𝑇 + 𝜃2𝐴𝐶𝐶𝑅𝑈𝐴𝐿_𝑆𝑇 + 𝜃3𝐺𝑅𝑂𝑊𝑇𝐻_𝑆𝑇 +
𝜃4𝐷𝑈𝑃𝑂𝑁𝑇_𝑆𝑇 + 𝜃5𝐸𝐴𝑅𝑁𝐼𝑁𝐺𝑆_𝐿𝑇 + 𝜃6𝐴𝐶𝐶𝑅𝑈𝐴𝐿_𝐿𝑇 +
𝜃7𝐺𝑅𝑂𝑊𝑇𝐻_𝐿𝑇 + 𝜃8𝐷𝑈𝑃𝑂𝑁𝑇_𝐿𝑇 + 𝜃9𝐷𝐼𝑉 + 𝑒𝑞+1 (1)
As noted above, the independent variables in Equation (1) are from three groups: short-term
variables measured using data from the earnings announcement at time q (e.g., EARNINGS_ST), 9 Importantly, the value that could be realized at the end of the month is highly correlated with, but distinct from, the return on the straddle, as it does not incorporate the initial purchase price.
13
long-term variables measured using quarterly data from the most recent five years (e.g.,
EARNINGS_LT), and dividend data from the previous year (DIV). The dependent variable is
measured over the three months following the month when earnings for quarter q are released.
Using the following process, we calculate rolling estimates of Equation (1) on the basis
of data available when each firm reports its earnings. For each calendar quarter, we estimate
Equation (1) using the five years of historical data that are available at the beginning of that
calendar quarter. We then take coefficients for Equation (1) estimated from historical data and
apply them to the current period’s fundamental signals to obtain a predicted value of subsequent
price movement (E[|RETq+1|]). For example, a firm reporting earnings during February 1999
would be assigned to the first calendar quarter of 1999. For this quarter, we create a sample to
estimate Equation (1) using data available before January 1, 1999, containing all firm-quarters
that reported earnings between the third quarter of 1993 and the third quarter of 1998. This date
range ensures that three months of returns following the earnings announcement (needed to
calculate the dependent variable in Equation (1)) are also observable before January 1, 1999. The
coefficients from this historical model would be applied to the signals available at the earnings
announcement during February 1999, which would then be used to predict straddle returns in
March, April, and May 1999. Thus, rather than equally weighting our signals, we use the
coefficients as weights to aggregate the signals to vary based on the historical regression
coefficients.
We begin our analysis of fundamental signals by estimating Equation (1) with all non-
financial firms (SIC codes not in 6000-6999) that have sufficient Compustat data to calculate the
fundamental signals and sufficient CRSP monthly return data to estimate the average absolute
value of monthly returns in the three months following the earnings announcement. We apply the
14
following sample selection criteria. We drop firm quarters with earnings report dates that are
more than 90 days after fiscal quarter-end, firm quarters with extremely low quarterly closing
prices (less than $1), and observations where the deflators (sales, total assets, and market value
of equity) are less than $1 million. We also require that each firm have non-missing market value
of equity at the end of quarter q. As we examine option returns that occur between January 1996
and December 2011, we estimate 65 versions of Equation (1) covering rolling windows from the
fourth calendar quarter of 1995 through the fourth quarter of 2011.
Panel A of Table 1 presents the average correlation matrix across the 65 samples to
provide insight into the univariate associations among subsequent absolute returns (|RETq+1|) and
fundamental signals. As expected, each fundamental signal is positively correlated with |RETq+1|.
The pair-wise correlations between individual fundamental signals are also positive. Panel B of
Table 1 presents the distribution of coefficients from Equation (1) across the 65 samples. To
indicate the explanatory power from each group of signals, columns 1, 2, and 3 present models
with short-term, long-term, and dividend signals, respectively. Column 4 presents the full model
used in our main analysis. Consistent with the positive pair-wise associations between each
fundamental signal and |RETq+1| in Panel A, the average coefficients on all of the fundamental
signals are positive. In addition, variation in coefficients occurs across signals, suggesting that
equally weighting the signals may not be optimal.10 We emphasize that our research purpose is
not to test whether these coefficients in Equation (1) are positive. Instead, these coefficients are
the first step (aggregating our fundamental signals into a single variable) in our analysis of
option returns.
10 When calculating E[|RETq+1|], we weight fundamental signals by their respective coefficients estimated from historical data. As Panel B of Table 1 shows, the coefficients vary across fundamental signals.
15
Panel C of Table 1 reports the association between the out-of-sample E[|RETq+1|] and
realized |RETq+1|. When we partition firms on the basis of E[|RETq+1|] into low, medium, and
high groups, we find that realized |RETq+1| increases monotonically from 0.336 in the low group
to 0.654 in the high group, with the spread between low and high groups statistically and
economically significant. This finding suggests that fundamental signals have strong predictive
power with respect to subsequent absolute price movements. Our subsequent tests explore the
extent to which investors price these fundamental signals in the options market.
3.3 Option Sample and Descriptive Statistics
While the analysis of fundamental volatility in Table 1 is based on all firms with
sufficient Compustat and CRSP data, we restrict the remainder of our analysis to the subset of
those firms with sufficient Optionmetrics data to calculate straddle option returns (SRETt+1). We
calculate SRETt+1 following Goyal and Seratto (2009). Specifically, we consider options that
mature in the next month, and select the contracts that are close to at-the-money, with moneyness
between 0.975 and 1.025.11 To form a straddle, for each stock and for each month in the sample,
we select the call and the put contracts with the same striking price that are close to at-the-month
and expire the next month. After next-month expiration, we repeat the procedure and select a
new pair of call and put contracts. As the straddle has both call and put contracts, the payoff to
the straddle is determined purely by the deviation of stock price a month later from its exercise
price. Whether the stock price goes up or down is irrelevant, a concept in line with the volatility
channel that we emphasize in the paper.
Matching our data on fundamental signals and historical volatility to Optionmetrics
results in a sample of 89,805 firm-months composed of 53,449 firm-quarters (3,521 unique 11 Our results are also robust to relaxing this constraint (including straddles that are either in-the-money or out-of-the-money).
16
firms). A comparison of the number of firm-months with the number of firm-quarters reveals
that we are typically able to use the information from a given firm-quarter for between one and
two straddle positions (89,805/53,449 = 1.68 monthly straddle contracts per firm-quarter) that
are originated during the three months following the earnings announcement month.12
Ex-ante, options investors may partially price the volatility information captured in
fundamental signals. Therefore, E[|RETt+1|] should be positively associated with implied
volatility embedded in option contracts.13 To identify cases where the information from
fundamental signals does not appear to be fully impounded into option prices, we adjust
E[|RETt+1|] relative to the level of implied volatility by using the following model:
E[|𝑅𝐸𝑇𝑡+1|] = 𝛾0 + 𝛾1𝐼𝑉𝑂𝐿𝑡 + 𝑒𝑡 (2)
where IVOLt is the natural log of implied volatility from Optionmetrics in month t. We label the
residual from Equation (2) as DIFF_FUND, which is our main variable of interest in the
subsequent analysis to predict future option returns (Panel B of Figure 1 presents a timeline). A
positive DIFF_FUND means fundamental volatility is high relative to implied volatility,
suggesting that both implied volatility and option prices are too low. Similarly, a negative
DIFF_FUND means implied volatility is too high and options are too expensive.
Panel A of Table 2 presents descriptive statistics for the monthly regression estimates of
Equation (2). As expected, fundamental volatility proxied by E[|RETt+1|] is positively correlated
with implied volatilities, suggesting that fundamental volatility is partially priced in the options
12 While we match the information available during an earnings announcement month to the subsequent three months, this match does not imply that all earnings announcements are matched to three straddle returns. The absence of a straddle return is largely attributable to the constraint that we require the straddle position to be roughly at-the-money. 13 We switch the subscript from q+1 to t+1 as we start to predict straddle returns on a monthly basis. E(RETq+1) in the fundamental analysis corresponds to E(RETt+1), E(RETt+2), and E(RETt+3) in the option return analysis.
17
market. However, the average monthly R2 is 46.3%, indicating that cases exist in which
E[|RETt+1|] and IVOLt differ substantially.
Panel B of Table 2 presents descriptive statistics for firm-months with both fundamental
signal and straddle return data. The mean (median) SRETt+1 is -0.013 (-0.193), a result consistent
with the prior literature and suggesting that straddle contracts on average lose money. Panel C of
Table 2 reports the correlation matrix of the main variables of interest in this study. We observe
that DIFF_FUND has a positive association with straddle option returns (SRETt+1), consistent
with the main empirical prediction of the paper. As expected, IVOL and HVOL exhibit a strong
positive correlation, suggesting that while implied volatility may not efficiently capture all of the
information in historical volatility, these two signals overlap considerably. The difference
between HVOL and IVOL, DIFF_HVOL, is positively correlated with SRETt+1, a result
consistent with Goyal and Saretto (2009). Finally, DIFF_FUND and DIFF_HVOL are only
moderately correlated, with Pearson and Spearman correlations of 0.226 and 0.197, respectively.
Such low correlations suggest that fundamental signals are likely to capture incremental
information about future option returns above and beyond Goyal and Saretto’s historical
volatility.
4. MAIN RESULTS
4.1 Do Fundamental Signals Predict Future Option Returns?
In this section, we use both portfolio and regression approaches to formally investigate
whether accounting information predicts future option returns. Under the portfolio approach, for
each month we sort straddle options into ten deciles based on DIFF_FUND. Panel A of Table 3
shows a positive association between DIFF_FUND and decile option returns. Straddle option
returns increase fairly monotonically from -9.8% in D1 to 6.5% in D10, resulting in a D10-D1
18
hedge return of 16.4% per month (t = 10.11). For comparison purposes, we replicate Goyal and
Sarreto’s (2009) historical volatility strategy in our sample. Column 2 of Panel A shows that
option returns increase from -11.4% in D1 to 7.4% in D10, resulting in a D10-D1 hedge return of
18.8% (t = 8.93) per month.14 Overall, we find that the hedge return from the fundamental
strategy is comparable to that from the historical volatility strategy (a slightly lower hedge return
and a higher t-statistic for the fundamental strategy).
While DIFF_FUND is orthogonal to IVOL by definition (Equation 2), Panel C of Table 2
shows a positive correlation between DIFF_FUND and DIFF_HVOL (Pearson=0.226,
Spearman=0.197), suggesting some overlap in the information contained in fundamental signals
and historical volatility. However, the correlation is sufficiently low that that there could be
useful information about volatility in fundamental signals that is not in historical volatility. To
assess the degree of information overlap between DIFF_FUND and DIFF_HVOL with respect
to future straddle returns , we employ a multivariate regression framework that includes
DIFF_HVOL as a control variable. Specifically, we consider the following regression model:
𝑆𝑅𝐸𝑇𝑡+1 = 𝛼0 + 𝛼1𝐷𝐼𝐹𝐹_𝐹𝑈𝑁𝐷𝑡 + 𝛼2𝐷𝐼𝐹𝐹_𝐻𝑉𝑂𝐿𝑡 + 𝑒𝑡+1 (3)
where the explanatory variables are transformed into decile rankings converted into a scale of [0,
1].
Panel B of Table 3 presents the regression results. Column 1 (Column 2) contains a
univariate model where only DIFF_FUND (DIFF_HVOL) is included as an independent
variable. We find significant coefficients on DIFF_FUND and DIFF_HVOL, a result that is
similar to the portfolio results in Panel A. In column 3, we include both measures in the model
14 Our replicated hedge return of 18.8% is slightly lower than the 22.7% documented in Goyal and Saretto (2009), a result that is largely expected as we limit our sample to firms with fundamental signals. Goyal and Saretto show that hedge returns are higher for smaller firms and firms with higher transaction costs, which are more likely to be excluded from our sample.
19
and find that both coefficients are highly significant, suggesting that fundamental signals and
historical volatility convey much orthogonal information with respect to future straddle returns.
By comparing Column 3 with Columns 1 and 2, we find that the coefficients on DIFF_FUND
and DIFF_HVOL drop by 25.6% (=(0.133-0.099)/0.133) and 17.4% (=(0.144-0.119)/0.144),
respectively. Roughly speaking, of the total information in DIFF_FUND about future straddle
returns, only a quarter is common to DIFF_HVOL. Therefore, although DIFF_FUND and
DIFF_HVOL are positively related to each other, much of the information contained in
DIFF_FUND regarding future option returns is separate from that contained in DIFF_HVOL. In
sum, both the portfolio and the regression results show that fundamental signals can predict
future straddle option returns.
4.2 Is a Fundamental-Based Option Strategy Risky?
To assess whether our fundamental-based option strategy is risky, we conduct two sets of
analysis. First, we examine the co-movement between hedge portfolio returns and common risk
factors. The finance literature suggests that theoretical asset pricing models, such as CAPM, also
apply to options (e.g., Coval and Shumway 2001). The intuition is that common state variables,
such as consumption, capture all kinds of risks in the capital markets, including the options
market. Therefore, the expected return on an option should relate to the option’s sensitivity to
common risk factors. Empirically, prior studies use common return factors to capture systematic
risks (e.g., Goyal and Saretto 2009; Choy 2011; Boyer and Vorkink 2011). Following prior
work, we use the four-factor model below to examine whether our fundamental trading strategy
exposes investors to systematic risks.15
15 We also include the return to a straddle based on an S&P500 index option as another risk factor in Equation (4) and our inferences related to the portfolio alpha do not change. In addition, to further investigate whether our hedge
20
𝑆𝑅𝐸𝑇𝐻𝐸𝐷𝐺𝐸 = 𝛼 + 𝛽1𝑀𝐾𝑇_𝑅𝐹 + 𝛽2𝑆𝑀𝐵 + 𝛽3𝐻𝑀𝐿 + 𝛽4𝑈𝑀𝐷 + 𝑒 (4)
where SRETHEDGE is the D10-D1 hedge portfolio return based on DIFF_FUND. MKT_RF, SMB,
HML, and MOM are the market, size, book-to-market, and momentum factors, respectively. The
four-factor data are from Kenneth French’s website. The intercept (a) provides an estimate of the
monthly abnormal returns earned by the DIFF_FUND trading strategy after controlling for these
four factors. As our straddle contracts typically span two months (e.g., a straddle initiated on
May 19 that expires on June 21) and do not all start on the same date, we present results for two
versions of Equation (4) using the factors for the initiation and the expiration months.
Panel A of Table 4 presents the analysis of the hedge return from the DIFF_FUND
trading strategy. In all cases, abnormal returns (intercepts) and their associated t-statistics from
the four-factor model are virtually the same as the raw returns reported in Panel A of Table 3. In
addition, none of the factor coefficients is statistically significant regardless of whether the
factors for the initiation or the expiration months are considered. In sum, the results suggest that
the hedge portfolio returns from our option strategies are not correlated with four common
factors.
Our second set of analyses examines the time series properties of our fundamental
analysis portfolio. An alternative risk-based perspective suggests that portfolios with higher
average returns must have inconsistent performance. Investors may have positive returns over a
long window, but the portfolio strategy exposes the investor to large negative outcomes. This
concern particularly arises for options-based trading strategies, where selling a straddle exposes
the seller to potentially extreme negative outcomes. To address this concern, we examine the
returns are capturing a volatility factor, we test whether our hedge returns have significant correlations with the VIX index. We find that these correlations are not significantly different from zero.
21
performance of our fundamental analysis portfolio over time, noting the overall frequency of
negative returns and the performance by year.
Panel B of Table 4 provides further details on the distribution of monthly returns for the
D10-D1 hedge return. The fundamental strategy delivers positive returns for 74% of months
during the sample window. The hedge returns are less than -20% for less than 5% of months in
the sample period. Finally, Figure 2 plots the average monthly hedge returns by year. The hedge
return is positive every year in the 16-year sample period.
In summary, we find fairly consistent strong positive returns for the fundamental-based
option strategy. Our evidence is unlikely to reflect an appropriate reward for a risky investment
strategy caused by the long positions being more risky than the short positions in the strategy.
4. 3 The Impact of Transaction Costs
In this section, we examine the impact of transaction costs on trading strategy
performance. Specifically, we consider the impact of the bid-ask spread, which is crucial in
interpreting empirical results in option studies.16 The bid-ask spread is typically much higher in
the option market than in the stock market. The main results in previous sections are based on the
assumption that options are traded at the mid-point of bid and ask prices. It is possible that
investors cannot trade options at that price in every circumstance. Many finance studies (e.g.,
Mayhew 2002; De Fontnouvelle et al. 2003) show that the effective spreads for options are large
in absolute terms but are typically much smaller than the quoted spreads, with the effective to
16 We do not consider the price impact of trade because options are derivatives of stocks and thus option trading may not necessarily move stock price in the stock market. However, investment capacity is limited in arbitrage activities depending on option trading volume. We also ignore margin requirements for two reasons. First, an examination of margin requirements does not go much beyond Goyal and Sarreto (2009), so the incremental contribution to the literature is small. We expect similar margin requirements to those of Goyal and Sarreto, as these two strategies produce similar t-statistics Second, the effect of margin requirements on trading profitability is likely to be slight if investors can borrow money at a rate close to the risk-free rate. The average risk-free rate is 0.215% per month in our sample period, compared to a hedge return of 16.4%. We partially address the margin requirement concern by considering a long-only strategy in call options (section 5.4), where there is no need for margin requirements.
22
quoted spread ratio to be less than 0.50. However, Battalio et al. (2004) find the effective to
quoted spread ratio to be between 0.8 and 1 for a small sample of large stocks. To examine the
impact of transaction costs, we recalculate option returns under the assumption that the effective
spread is equal to the quoted spread (the effective-to-quoted spread ratio of 1): investors always
buy options at the ask price and write options at the bid price. We view this assumption as
relatively conservative because it produces the lowest after-transaction cost returns relative to
alternative effective spread assumptions used by prior literature.17 As tick-by-tick option
transaction data are not available, we follow the prior literature and use the closing bid-ask
spread to proxy for the average bid-ask spread.
We repeat the main analysis in Table 3 using option returns after transaction costs and
report empirical results in Panel A of Table 5. We find that the D10-D1 hedge return from the
fundamental strategy drops from 16.4% (t = 10.11) before transaction costs to an insignificant
level of -1.6% (t = -0.90) after transaction costs. In comparison, the hedge return from historical
volatility strategy drops from 18.8% (t = 8.93) to -1.3% (t = -0.57).18 The inability of the returns
for the fundamental or historical volatility strategy in isolation to exceed transaction costs
provides evidence consistent with a “limits to arbitrage” framework (Shleifer and Vishny 1997).
The returns tests suggest that option prices do not fully capture fundamental or historical
volatility signals. However, transaction costs are sufficiently large to limit an arbitrageur’s ability
to profit from option mispricing in isolation, allowing the mispricing to exist.
17 For example, Goyal and Saretto (2009) also consider effective spread as 50% or 75% of quoted spread. If we use these alternatives, our trading strategy would generate higher after-transaction-cost returns than those reported in the paper. For example, for the overall sample, the D10-D1 hedge returns would increase from -1.6% to 3.89% (t = 2.40) and 1.47% (t = 0.90) if we use effective spread as 50% and 75% of quoted spread, respectively. 18 Similarly, Goyal and Saretto (2009) note that, in their sample, the hedge return drops from 22.7% (t = 10.41) to an insignificant level of 3.9% (t = 1.84) when they assume that the effective spread is equal to the quoted spread.
23
Our second approach to address the transaction cost issue is to examine cross-sectional
variations in trading strategy performance based on the proxies for transaction costs and
liquidity. Specifically, we consider three proxies: firm size, the bid-ask spread, and option
trading volume. For each proxy, we partition our sample into three groups each month and then
implement our fundamental trading strategy in each resulting group. Finally, we calculate the
D10-D1 hedge returns both before and after transaction costs. The results are reported in Panels
B, C, and D of Table 5.
We find that hedge portfolio returns before transaction costs do not vary much with the
proxies for liquidity and transaction costs. For example, pre-transaction-cost returns are 17.2%
and 16.3% per month for small and large firms, respectively. The hedge returns before
transaction costs are also similar between options with small bid-ask spreads (17.7%) and
options with large bid-ask spreads (17.9%). When using option volume as the proxy for liquidity,
we find that pre-transaction-cost returns are higher for more liquid options (21%) than for less
liquid options (15%). As transaction costs vary monotonically with these proxies, the hedge
returns after transaction costs increase monotonically from options with high transaction costs to
options with low transaction costs. The hedge returns after transaction costs are reliably positive
and significant for options with low transaction costs: 7.0% (t = 2.94) for large firms, 11.2% (t =
4.51) for options with low bid-ask spreads, and 12.4% (t = 5.35) for options with large trading
volume.
In sum, we find that transaction costs significantly affect the performance of fundamental
trading strategy in isolation. For the overall sample, the D10-D1 hedge returns drop from16.4%
before transaction costs to an insignificant level after transaction costs. However, the D10-D1
24
hedge returns are still highly positive for options with low transaction costs, such as large firms,
options with low bid-ask spread, and options with large trading volume.
4.4 Combing fundamental and historical volatility signals
The preceding analysis shows that investors can enhance portfolio performance by
focusing on options with low transaction costs. In this section, we consider how investors could
combine fundamental signals and historical volatility together in their trading strategies. As the
previous sections show that fundamental signals contain information about future straddle
returns that is incremental to what is captured in historical volatility, we expect higher hedge
returns by combining historical volatility with fundamental signals.
We adopt the following trading strategy. We begin with the bottom and top deciles (D1
and D10) of DIFF_HVOL, which are the short and long positions in Goyal and Saretto (2009).
Next, we further sort the observations in the D1 and D10 into four quartiles, each based on
DIFF_FUND. To calculate the hedge return from using both signals, we identify cases where the
fundamental and historical volatility signals agree. We pick options in D1 with the lowest
fundamental score as our short position and options in D10 with the highest fundamental score as
our long position. Panel A of Table 6 provides an analysis of the returns to combining these two
types of signals. Consistent with our expectations, incorporating fundamental signals into the
historical volatility strategy increases the hedge return from 18.8% in Panel A of Table 3 to
30.1% in Panel A of Table 6. After transaction costs, the hedge returns are still significantly
positive (return = 7.4%, t = 2.15). In contrast, when the two sets of signals disagree, the hedge
portfolio with a long position in D10(DIFF_HVOL) and Q1(DIFF_FUND) and a short position
in D1(DIFF_HVOL) and Q4(DIFF_FUND) yields much lower returns (before-transaction-cost
return = 9.64%, after-transaction-cost return = -8.40%). In sum, historical volatility and
25
fundamental signals are complements and the two signals can be combined into a single strategy.
The incremental return from incorporating fundamental signals allows the hedge return to stay
positive and significant even in the presence of transaction costs.
5 ADDITIONAL ANALYSES AND SENSITIVITY CHECKS
5.1 The Time-Series Pattern of Implied Volatility around the Portfolio Formation Date
We have motivated our analysis from the fundamental volatility perspective. That is, our
fundamental signals capture fundamental volatility, and implied volatility may deviate from the
true underlying volatility. In portfolios sorted by our fundamental score, implied volatility is too
low relative to fundamental volatility for D10 options and too high for D1 options. If this story
holds, a natural prediction is that implied volatility should increase for D10 options and decline
for D1 options after the portfolio formation date, given that over time, implied volatility should
converge to the true underlying volatility.
In Figure 3, we plot the time-series pattern of implied volatility of D1 and D10 options.
We consider a range of 12 months before and 12 months after the portfolio formation date (t =
0). The results are striking. For D1 options, implied volatility is higher at time zero than in
previous months. After time zero, implied volatility decreases to the level of the previous
months. In contrast, for D10 options implied volatility is much lower at time zero than in
adjacent months, resulting in a clean “V” shape of implied volatility. Overall, the results are
consistent with the story that implied volatility temporally deviates from fundamental volatility,
resulting in predictable option returns.
5.2 The Analysis of Individual Fundamental Signals
26
In the main analysis, we aggregate individual fundamental signals into a single score
when examining future option returns. In this section, we examine individual signals to see
which signal contributes more to the hedge returns. We run the following expanded regression
models:
𝑆𝑅𝐸𝑇𝑡+1 = 𝛼0 + ∑𝑎𝑘𝑆𝐼𝐺𝑁𝐴𝐿𝑘 + 𝛼𝑘+1𝐻𝑉𝑂𝐿𝑡 + 𝛼𝑘+2𝐼𝑉𝑂𝐿𝑡 + 𝑒𝑡+1 (5)
where SIGNALK are individual fundamental signals, HVOL is historical volatility, and IVOL is
implied volatility. All explanatory variables are transformed into decile rankings converted into a
scale of [0, 1].
Table 7 presents estimates of Equation (5). The coefficients on the individual
fundamental signals reveal that EARNINGS_ST, ACCRUAL_ST, EARNINGS_LT, GROWTH_LT,
and DIV have the greatest positive association with future straddle returns. In addition, when
estimating Equation (5), we also test the joint hypothesis that none of the signals is informative.
The sum of the individual signals can also be interpreted as the hedge portfolio returns based on
individual fundamental signals (Abarbanell and Bushee 1998). Table 7 shows that the sum of
these coefficients is positive and highly significant in both regression models.
5.3 A Simplified Strategy to Mitigate Data Mining Concerns
Beneish et al. (2001) identify a number of fundamental signals to predict extreme price
movements. To avoid the risk of data mining, we repeat our analysis estimating an alternative
version of Equation (1) based only on eight fundamental signals identified in Beneish et al.
(2001): sales growth, sales growth decline indicator, changes in earnings, loss indicator, R&D
intensity, accruals, capital expenditures, and changes in gross margins. When we repeat our
analysis with these signals, we continue to observe positive and significant raw returns (D10-D1
hedge return = 9.75%, t = 5.23). The hedge return drops to -8.82% (t = -4.25) after transaction
27
costs. Although arbitrageurs cannot profit from this simple trading strategy in isolation, the
results still suggest that option prices are not right with respect to fundamental volatility captured
in these accounting signals.
5.4 Analysis of options that are also exposed to equity risk
All of our preceding analysis has focused on straddle contracts, which do not depend on
directional equity returns. However, as stock prices tend to go up because of a positive risk-free
rate and risk premium, longing put options and shorting call options tend to lose money,
especially after transaction costs. Therefore, an alternative trading strategy is to long call options
with high fundamental scores and to short put options with low fundamental scores. This
alternative trading strategy yields an impressive hedge return of 44.3% per month before
transaction costs and 26.9% per month after transaction costs.19 After adjusting for the four
factors included in Equation (4), the abnormal returns remain highly significant after transaction
costs, with a value of 25.15% (t = 2.70) and 17.73% (t = 2.13) using factor data measured in
months t and t+1, respectively. As shorts are undesirable to some investors, we also examine a
long-only portfolio with call options (D10) that still yields an average monthly return of 11.6% (t
= 2.35) after transaction costs, which corresponds to abnormal returns of 10.18% (t = 2.14) and
8.32% (t = 1.87) after controlling for the four comment factors measured in month t and t+1,
respectively. One caveat applies to the approach of using individual options. Unlike straddle
contracts, these trading strategies do not hedge against market movements, resulting in more
volatile portfolio returns over time. In addition, portfolio returns are a result of both volatility
mispricing and market movements.
19 In comparison, the historical volatility strategy that longs call options in D10 and shorts put options in D1 yields a hedge return of 38.5% before transaction costs and 18.9% per month after transaction costs.
28
6. CONCLUSION
In this paper, we examine the extent to which accounting information is useful in
evaluating the volatility of a firm’s operations and whether this information is appropriately
priced by the options market. We find evidence that information about a firm’s fundamental
volatility is not fully priced in option contracts. Hedge portfolios with long and short straddle
contract positions based on accounting signals earn statistically and economically significant
returns before transaction costs. However, the high level of transaction costs in the options
market limits the profitability of fundamental trading strategy in isolation. Transaction costs
render the D10-D1 hedge returns to an insignificant level for the overall sample, but the hedge
returns remain highly positive for options with low transaction costs. We also show that the
fundamental-based strategy is additive to the historical volatility-based strategy. A strategy that
combines both historical volatility and fundamental signals earns returns in excess of transaction
costs. Investors may also increase hedge portfolio returns by focusing on individual call and put
options rather than straddle contacts.
Overall, our evidence provides insight into a new dimension of fundamental analysis—
using accounting signals to evaluate fundamental volatility and examining whether such
information is priced in the options market. Our evidence complements prior fundamental
analysis research on equity returns, which focused on fundamental signals to predict future
operating performance and stock returns.
29
REFERENCES
Abarbanell, J. and B. Bushee. 1998. Abnormal returns to a fundamental analysis strategy. The Accounting Review 73, 19-45.
Ang, A., R. Hodrick, Y. Xing, and X. Zhang. 2006. The cross-section of volatility and expected returns. Journal of Finance 51, 259-299.
Ball, R. and P. Brown. 1968. An empirical evaluation of accounting income numbers? Journal of Accounting Research 6, 159-178.
Barth, M. and E. So. 2009. Earnings announcements equity volatility and risk premia: evidence from equity returns and option prices. Stanford University working paper.
Battolio, R., B. Hatch, and R. Jennings. 2004. Toward a national market system for U.S. exchange-listed equity options. Journal of Finance 59, 933-962.
Battalio R. and P. Schultz, 2006, Options and the bubble, Journal of Finance, 59 (5), 2017-2102.
Beneish, M., Lee, C. and R. Tarpley. 2001. Contextual fundamental analysis through the prediction of extreme returns. Review of Accounting Studies 6, 165-189.
Bernard, V. and J. Thomas. 1990. Evidence that stock prices do not fully reflect the implications of current earnings for future earnings. Journal of Accounting and Economics 13, 305-340.
Boyer, B. and K. Vorkink. 2011. Stock option lotteries. Brigham Young University working paper.
Callen, J., 2009. Shocks to stocks: a theoretical foundation for the information content of earnings. Contemporary Accounting Research 26, 135-166.
Choy, S., 2011. Retail clientele and option returns. University of Toronto working paper.
Christensen, B. and N. Prabhala. 1998. The relation between implied and realized volatility. Journal of Financial Economics 50, 125-150.
Coval, J. and T. Shumway. 2001. Expected option returns. Journal of Finance 56, 983-1009.
Cristoffersen, P., Goyenko, R., Jacobs, K. and M. Karoui. 2011. Illiquidity premia in the equity option market. McGill University working paper.
De Fontnouvelle, P., P. Fisher, and R. Harris. 2003. The behavior of bid-ask spreads and volume in options markets during the competition for listings in 1999. Journal of Finance 58, 2437-2463.
Driessen, J., Maenhout, P. and G. Vilkov. 2009. The price of correlation risk: evidence from equity options. Journal of Finance 64, 1377-1406.
Dubinsky, A. and M. Johannes. 2006. Fundamental uncertainty, earnings announcements, and equity options. Working paper.
30
Fama, E. and K. French. 1992. The cross-section of expected stock returns. Journal of Finance 47, 427-466.
Fleming, J., B. Ostdiek, and R. Whaley. 1996. Trading costs and the relative rates of price discovery in stock, futures, and option markets. Journal of Futures Markets 16, 353–387.
Goyal, A. and A. Saretto. 2009. Cross-section of option returns and volatility. Journal of Financial Economics 94, 310-326.
Guay, W. and J. Harford. 2000. The cash-flow permanence and information content of dividend increases versus repurchases. Journal of Financial Economics 57, 385-415
Hayn, C. 1995. The information content of losses. Journal of Accounting and Economics 15, 226-244.
Holthausen, R. and D. Larcker. 1992. The prediction of stock returns using financial statement data. Journal of Accounting and Economics 15, 317-411.
Mashruwala, C., S. Rajgopal, and T. Shevlin. 2006. Why is the accrual anomaly not arbitraged away? The role of idiosyncratic risk and transaction costs. Journal of Accounting and Economics 42, 3-33.
Maydew, S. 2002. Competition, market structure, and bid-ask spreads in stock option markets. Journal of Finance 57, 931-958.
Ou, J. and S. Penman. 1989. Financial statement analysis and the prediction of stock returns. Journal of Accounting and Economics 11, 295-329.
Pan, J. 2002. The jump-risk premia implicit in option prices: evidence from an integrated time-series study. Journal of Financial Economics 63, 3-50.
Pastor, L. and P. Veronesi. 2003. Stock valuation and learning about profitability. Journal of Finance 58, 1749–1789.
Piotroski, J. 2000. Value investing: The use of historical financial statement information to separate winners from losers. Journal of Accounting Research 38.
Pool, V. K., Stoll, H. R., and R. E. Whaley. 2008. Failure to Exercise Call Options: An Anomaly and a Trading Game. Journal of Futures Markets, 11(1), 1-35.
Richardson, S., Tuna, I. and P. Wysocki. 2010. Accounting anomalies and fundamental analysis: a review of recent research advances. Journal of Accounting and Economics 50, 410-454.
Rogers, J., Van Buskirk, A., Skinner, D., 2009. Earnings guidance and market uncertainty. Journal of Accounting and Economics 48, 90-109.
Roll, R., E. Schwartz, and A. Subrahmanyam.2010. O/S: The Relative Trading Activity in Options and Stock, Journal of Financial Economics, 96, 1-17.
31
Scheinkman, J. and W. Xiong. 2003. Overconfidence and speculative bubbles. Journal of Political Economy 111, 1183-1219.
Shiller, R., 1981. Do stock prices move too much to be justified by subsequent changes in dividends? American Economic Review 71, 421-436.
Shleifer, A and R. Vishny. 1997. The limits of arbitrage. Journal of Finance 52, 25-55.
Soliman, M. 2008. The use of DuPont analysis by market participants. The Accounting Review 83, 823-853.
Sloan, R., 1996, Do stock prices fully reflect the information in accruals and cash flows about future earnings? The Accounting Review 71, 289-315.
Zhang, X. F. 2006. Information uncertainty and stock returns. Journal of Finance 51, 105-137.
32
APPENDIX – VARIABLE DEFINITIONS
Variable Definition
EARNINGS_ST = Short-term earnings signal, measured as the average of the decile rank of |∆EARNq|, transformed to be on [0, 1] scale, and the LOSS dummy. |∆EARNq| is the absolute value of ∆EARNq, which is measured as (IBQq – IBQq-4)/MVEq-4, where IBQq is income before extraordinary items in quarter q and MVEq-4 is the market value of equity at the end of quarter q-4. LOSS is an indicator with the value of 1 if IBQ is negative during quarters q, q-1, q-2, or q-3, and 0 otherwise.
EARNINGS_LT = Long-term earnings signal, measured as the decile rank of standard deviation of ∆EARNq over quarters q through q-20, where the decile rank is transformed to be on a scale of [0, 1] .
ACCRUAL_ST = Short-term accrual signal, measured as the decile rank of |∆WCq|, transformed to be on a scale of [0, 1]. |∆WCq| is the absolute value of ∆WCq, where ∆WCq is seasonally adjusted change in working capital divided by beginning total assets. ∆WCq =( (RECTQq + INVTQq - APTQq) - (RECTQq-4 + INVTQq-4 - APTQq-4) )/ATQq-4, where RECTQ is accounts receivable, INVTQ is inventory, APTQ is accounts payable, and ATQ is average total assets.
ACCRUAL_LT = Long-term accrual signal, measured as the decile rank of standard deviation of ∆WCq over quarters q through q-20, where the decile rank is transformed to be on a scale of [0, 1].
GROWTH_ST = Short-term growth signal, measured as the average of decile ranks of |SGRq| and |AGRq|, transformed to be a scale of [0, 1]. |SGRq| is the absolute value of SGRq, where SGRq is seasonally adjusted sales growth during quarter q, measured as (SALEQq - SALEQq-4)/SALEQq-4). |AGRq| is the absolute value of AGRq, where AGRq is seasonally adjusted asset growth during quarter q, measured as (ATQq - ATQq-4)/ATQq-4.
GROWTH_LT = Long-term growth signal, measured as the average of decile ranks of standard deviations of SGRq and AGRq over quarters q through q-20, where the decile ranks are transformed to be on [0, 1] scale.
DUPONT_ST = Short-term DuPont signal, measured as the average of decile ranks of (|∆ATOq| and |∆PMq|, transformed to be a scale of [0, 1]. |∆ATOq| is the absolute value of (ATOq - ATOq-4), where ATOq (=SALEQq /ATQq-1) is sales from quarter q divided by total assets from quarter q-1. |∆PMq| is the
33
absolute value of (PMq - PMq-4), where PMq (=IBQq / SALEQq) is income from quarter q divided by sales from quarter q.
DUPONT_LT = Long-term DuPont signal, measured as the average of decile ranks of standard deviations of ∆ATOq and ∆PMq over quarters q through q-20, where the decile ranks are transformed to be a scale of [0, 1].
DIV = Negative one multiplied by divided in prior fiscal year divided by average assets in prior fiscal year.
|RETq+1| = Natural log of the average absolute monthly return over the three months following the month when the firm’s earnings announcement occurs.
E[|RETt+1|] = Predicted absolute return in month t, |RETt+1|, based on coefficients from Equation (1) estimated over the prior five years using data available prior to the firm’s earnings announcement.
IVOL = Natural log of implied volatility, which is downloaded from Optionmetrics.
HVOL = Natural log of historical volatility, where historical volatility is estimated using daily returns over the year preceding the month when the trading strategy is initiated. The standard deviation of daily returns is transformed into an annual measure by multiplying by the square root of 252.
DIFF_FUND = Residual value from regressing E[|RET|] on IVOL (Equation (2)), where the regression is estimated monthly and IVOL is implied volatility.
DIFF_HVOL = HVOL – IVOL
SRETt+1 = Monthly return to a straddle position initiated in firm i at time t.
ASK_BID = The bid–ask spread for traded options, measured as the difference between the highest closing bid and the lowest closing ask scaled by the mid-point of these two.
VOLUME = Dollar volume of traded options, measured as option trade volume multiplied by the midpoint of the highest closing bid and the lowest closing ask prices.
MVE = The market value of equity from the end of the most recent fiscal quarter.
34
Figure 1 – Timeline
Panel A: Measurement of fundamental signals available as of the earnings announcement date (RDQ)
Calendar Quarter q
Calendar Quarters q-19 through q
Measurement of short-term fundamental signals (e.g., EARNINGS_ST) based on information released on the RDQ date
Measurement of long-term fundamental signals (e.g., EARNINGS_LT) based on information released on and before the RDQ date
Months t+1, t+2 and t+3 Month t
The RDQ date
Measurement of SRET in months t+1 through t+3
Panel B: Timeline of matching fundamental signals to straddle returns
Fundamental data E[|RETq+1|] available at the RDQ date
35
Figure 2 Average monthly returns to fundamental-based hedge portfolios over time
This figure provides the average monthly return to the hedge portfolio over time. For each month, we sort straddle options into ten deciles based on our fundamental score (DIFF_FUND). Hedge returns are returns to the hedge portfolio with a long position in the top DIFF_FUND decile and a short position in the bottom DIFF_FUND decile. Each bar represents the average monthly hedge return in a given year.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
Year-by-year average monthly hedge returns
36
Figure 3 The time series pattern of implied volatility around the portfolio formation date
This figure provides the average implied volatility for the top and bottom DIFF_FUND deciles from 12 months before the portfolio formation date to 12 months after.
0.36
0.37
0.38
0.39
0.4
0.41
0.42
0.43
0.44
0.45
0.46
0.47
-12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
The time-series pattern of IVOL for the bottom DIFF_FUND decile
0.36
0.38
0.4
0.42
0.44
0.46
0.48
0.5
0.52
-12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
The time-series pattern of IVOL for the top DIFF_FUND decile
37
Table 1 The association between fundamental signals and future stock return volatility
Panel A: Correlation matrix (Pearson upper triangle, Spearman lower triangle)
Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
(1) |RETq+1| 1.00 0.28 0.18 0.21 0.28 0.29 0.28 0.31 0.37 0.34
(2) EARNINGS_ST 0.28 1.00 0.10 0.10 0.54 0.53 0.22 0.32 0.44 0.28
(3) ACCRUAL_ST 0.18 0.11 1.00 0.41 0.24 0.15 0.50 0.22 0.27 0.21
(4) GROWTH_ST 0.20 0.11 0.40 1.00 0.38 0.12 0.27 0.42 0.33 0.22
(5) DUPONT_ST 0.28 0.56 0.24 0.38 1.00 0.39 0.33 0.43 0.56 0.27
(6) EARNINGS_ST 0.29 0.54 0.15 0.12 0.39 1.00 0.32 0.38 0.65 0.36
(7) ACCRUAL_ST 0.28 0.23 0.50 0.27 0.32 0.32 1.00 0.54 0.56 0.35
(8) GROWTH_ST 0.31 0.32 0.22 0.41 0.43 0.38 0.54 1.00 0.75 0.39
(9) DUPONT_ST 0.37 0.45 0.27 0.33 0.55 0.65 0.56 0.75 1.00 0.44
(10) DIV 0.34 0.29 0.21 0.22 0.27 0.37 0.36 0.39 0.44 1.00
Panel B: Average estimated parameters from 65 rolling models predicting |RET| |RETq+1| = θ0 + θ1EARNINGS_ST + θ2ACCRUAL_ST + θ3GROWTH_ST + θ4DUPONT_ST
+ θ5EARNINGS_LT + θ6ACCRUAL_LT + θ6GROWTH_LT + θ8DUPONT_LT + θ9DIV + error
(1)
Variable 1 2 3 4 Intercept 0.227 0.235 0.329 0.148
EARNINGS_ST 0.182 0.096
ACCRUAL_ST 0.095 0.032
GROWTH_ST 0.122 0.067
DUPONT_ST 0.160 0.058
EARNINGS_LT 0.117 0.049
ACCRUAL_LT 0.106 0.061
GROWTH_LT 0.089 0.035
DUPONT_LT 0.221 0.129
DIV 0.264 0.142
Average R2 0.127 0.153 0.115 0.199
38
Panel C: Out-of-sample analysis association between future realized |RETq+1| and E[|RETq+1|] Groups based on E[|RETq+1|]
Low
N=57,368 Medium
N=57,404 High
N=57,391 High vs. Low
t- statistic Average scaled decile rank of realized |RETq+1| 0.338 0.518 0.654 188.57 Table 1 provides descriptive statistics on the variables used to estimate the association between accounting signals and the magnitude of future stock price movements (Equation (1)), the estimated coefficients, and the out-of-sample properties of the predicted values. |RETq+1| is natural log of the average absolute month returns over three months following the month of earnings announcement. E[|RETq+1|] is predicted |RETq+1| based on Equation (1). EARNINGS_ST is short-term earnings signal. ACCRUAL_ST is short-term accrual signal. GROWTH_ST is short-term growth signal. DUPONT_ST is short-term DuPont measure. EARNINGS_LT is long-term earnings signal. ACCRUAL_LT is long-term accrual signal. GROWTH_LT is long-term growth signal. DUPONT_LT is long-term DuPont signal. DIV is the dividend signal. Please see the Appendix for detailed variable definitions. Equation (1) is estimated using 65 rolling window samples composed of firms with sufficient Compustat and CRSP data to calculate the dependent and independent variables in Equation (1). The top and bottom 1% of all variables in Panel A are winsorized in each estimation sample. Panel A presents the univariate correlations between the variables used to estimate Equation (1). Panel B presents the coefficient estimates of Equation (1) that are averaged across the 65 samples that are used to estimate Equation (1). Panel C presents evidence on the association between out-of-sample predicted values and realized return movements.
39
Table 2 Descriptive statistics
Panel A: Models to estimate abnormal levels of E[|RETt+1|] E[|RETt+1|] = γ0 + γ1IVOLt + error (2)
Independent variable Coefficient
(t-stat)
Intercept 0.148 (56.93)
IVOLt 0.881
(70.94) R2 0.463 Panel B: Descriptive statistics Variable Mean STDEV Q1 Median Q3 SRETt+1 -0.013 0.825 -0.622 -0.193 0.387 IVOLt -0.962 0.441 -1.257 -0.969 -0.673 HVOLt -0.943 0.458 -1.262 -0.958 -0.637 DIFF_HVOL 0.023 0.280 -0.147 0.010 0.167 DIFF_FUND 0 0.100 -0.071 -0.003 0.067 Panel C: Correlation matrix (Pearson upper triangle, Spearman lower triangle) Variable (1) (2) (3) (4) (5)
(1) SRETt+1 1.000 -0.012 -0.007 0.008 0.052 (2) IVOLt -0.016 1.000 0.806 -0.259 0.013 (3) HVOLt -0.009 0.814 1.000 0.363 0.152 (4) DIFF_HVOL 0.015 -0.198 0.352 1.000 0.226 (5) DIFF_FUND 0.043 0.072 0.182 0.197 1.000 Table 2 presents descriptive statistics for our main sample. SRETt+1 is straddle option returns. IVOL is implied volatility. HVOL is historical volatility. DIFF_HVOL is the difference between HVOL and IVOL. E[|RETt+1|] is predicted absolute value of monthly price movement over three months following the month of earnings announcement. DIFF_FUND is the residual from Equation (2). Panel A presents average coefficient and average t-statistic from the monthly regressions to estimate Equation (2). Panel B presents univariate descriptive statistics. Panel C presents correlation matrix for the main variables used in this study. The top and bottom 1% of all variables (except SRETt+1) are winsorized each month. Please see the Appendix for detailed variable definitions. The sample includes 89,805 firm-month observations from January 1996 to December 2011.
40
Table 3 Fundamental score and future option returns
Panel A: Portfolio analysis
1 2
Deciles
Portfolios based on fundamental signals
(DIFF_FUND)
Replication of Goyal and Saretto (2009)
in our sample (DIFF_HVOL)
D1 -0.098 -0.114 D2 -0.041 -0.055 D3 -0.058 -0.048 D4 -0.028 -0.020 D5 -0.023 -0.009 D6 -0.007 -0.013 D7 -0.009 -0.012 D8 0.020 0.014 D9 0.032 0.034 D10 0.065 0.074 HEDGE (D10-D1)
0.164 (10.11)
0.188 (8.93)
Panel B: Fama-MacBeth regressions of future straddle option returns
Independent variables 1 2 3
Intercept -0.081 (-4.46)
-0.087 (-5.29)
-0.124 (-7.28)
DIFF_FUND 0.133 (11.18) 0.099
(8.36)
DIFF_HVOL 0.144 (9.10)
0.119 (7.45)
R2 0.004 0.007 0.010
41
Table 3 presents evidence on the performance of straddle option returns (SRETt+1) based on fundamental signals. In Panel A, each month, we sort straddle options into ten deciles based on the fundamental score DIFF_FUND. More positive (negative) values of DIFF_FUND indicate observations where the expected price movement based on fundamental signals is larger (smaller) than would be predicted based on historical or implied volatility. Column A presents portfolio results across ten DIFF_FUND deciles. For comparison to prior work, Column B replicates Goyal and Saretto’s results using our sample data. IVOL is implied volatility. HVOL is historical volatility. DIFF_HVOL is the difference between HVOL and IVOL. DIFF_FUND is the residual fundamental score after controlling for IVOL (Equation (2)). HEDGE is the hedge portfolio with a long position in the top decile and a short position in the bottom decile. Panel B presents multivariate regression analysis of the association between fundamental signals and straddle returns. Both DIFF_FUND and DIFF_HVOL are transformed into decile ranks converted into the [0, 1] scale. The t-statistics are based on Fama-MacBeth regressions. All variables are defined in the Appendix. The sample includes 89,805 firm-month observations from January 1996 to December 2011.
42
Table 4 Evaluation of monthly returns from fundamental analysis portfolios
Panel A: Returns to fundamental analysis portfolios and risk factors
Risk data In month t
Risk data In month t+1
Intercept 0.162
(9.87) 0.165
(10.08)
MKT_RF 0.296
(0.81) -0.498
(-1.36)
SMB -0.647 (-1.37)
-0.202 (-0.43)
HML -0.760 (-1.52)
-0.820 (-1.65)
UMD 0.296
(0.99) 0.245
(0.82)
R2 0.026 0.033
Panel B: Distribution of the D10-D1 hedge returns based on DIFF_FUND
Percent of months with returns less than 0% 0.260
Percent of months with returns less than -10% 0.094
Percent of months with returns less than -20% 0.047
Percent of months with returns less than -30% 0.031
Percent of months with returns less than -40% 0.015
Percent of months with returns less than -50% 0.005 Table 4 presents an examination of whether a fundamental-based trading strategy exposes option investors to systematic risks. Panel A report regression of D10-D1 hedge portfolio returns (SRETHEDGE) on common risk factors, where the intercept (or alpha) measures abnormal performance.
SRETHEDGE= α + β1MKT_RF + β2SMB + β3HML + β4UMD + error, where the monthly risk factors (MKT_RF, SMB, HML, and UMD) are obtained from Kenneth French’s website. T-statistics are in parenthesis. Panel B examines the frequency and magnitude of negative returns to this hedge portfolio strategy. DIFF_FUND is the residual fundamental score after controlling for IVOL (Equation (2)). The D10-D1 hedge return is the return to a hedge portfolio with a long position in the top DIFF_FUND decile and a short position in the bottom DIFF_FUND decile. All variables are defined in the Appendix. The sample includes 89,805 firm-month observations from January 1996 to December 2011.
43
Table 5 The analysis of transaction costs
Panel A: Hedge portfolio (D10-D1) return before vs. after transaction costs
Before transaction costs After transaction costs
Deciles DIFF_FUND DIFF_HVOL DIFF_FUND DIFF_HVOL
D1 -0.098 -0.114 -0.003 -0.004
D10 0.065 0.074 -0.019 -0.017
HEDGE (D10-D1)
0.164 (10.11)
0.188 (8.93) -0.016
(-0.90) -0.013
(-0.57)
Panel B: Hedge portfolio return based on DIFF_FUND by firm size
Before transaction costs After transaction costs
Deciles Bottom size
tercile Medium
size tercile Top size tercile Bottom
size tercile Medium
size tercile Top size tercile
D1 -0.107 -0.090 -0.096 0.087 -0.010 -0.049
D10 0.065 0.043 0.067 -0.048 -0.028 0.021
HEDGE (D10-D1)
0.172 (6.41)
0.133 (5.37)
0.163 (6.82) -0.136
(-3.64) -0.019
(-0.77) 0.070
(2.94)
Panel C: Hedge portfolio return based on DIFF_FUND by the bid-ask spread
Before transaction costs After transaction costs
Deciles
Bottom spread tercile
Medium spread tercile
Top spread tercile
Bottom spread tercile
Medium spread tercile
Top spread tercile
D1 -0.096 -0.086 -0.121 -0.065 -0.025 0.098
D10 0.081 0.036 0.058 0.047 -0.027 -0.079
HEDGE (D10-D1)
0.177 (7.01)
0.122 (5.07)
0.179 (7.32) 0.112
(4.51) -0.002
(-0.08) -0.178
(-5.14)
44
Panel D: Hedge portfolio return based on DIFF_FUND by option trading volume
Before transaction costs After transaction costs
Deciles
Bottom volume tercile
Medium volume tercile
Top volume tercile
Bottom volume tercile
Medium volume tercile
Top volume tercile
D1 -0.110 -0.076 -0.119 0.081 0.004 -0.079
D10 0.040 0.061 0.092 -0.080 -0.015 0.046
HEDGE (D10-D1)
0.150 (5.86)
0.137 (5.52)
0.210 (8.80) -0.161
(-4.61) -0.019
(-0.75) 0.124
(5.35)
Table 6 presents evidence on the impact of transaction costs on hedge portfolio returns. Transaction costs are incorporated into the straddle returns by assuming that long positions are acquired by purchasing the straddle at the ask price and short positions are acquired by selling straddle contracts at the bid price (effective spread = quoted spread). Panel A provides evidence on the performance of trading strategies based on fundamental signals (DIFF_FUND) and Goyal and Saretto’s historical volatility (DIFF_HVOL) before vs. after transaction costs. Panels B, C, and D report portfolio results when we partition the sample into three terciles on the basis of firm size, the bid-ask spread, or option trading volume. Firm size is the market value of equity from the end of prior fiscal quarter. The bid-ask spread is the difference between the highest closing bid and the lowest closing ask scaled by the mid-point of these two. Option trading volume is the option trade volume multiplied by the bid-ask midpoint. The sample includes 89,805 firm-month observations from January 1996 to December 2011.
45
Table 6 Combining fundamental signals and historical volatility into a single trading strategy
Panel A: Combine fundamental score with historical volatility signals DIFF_FUND DIFF_FUND
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
Before transaction costs After transaction costs
D1 (DIFF_HVOL) -0.185 -0.110 -0.093 -0.072 -0.061 -0.004 0.017 0.028
D10 (DIFF_HVOL) 0.024 0.098 0.057 0.115 -0.056 0.007 -0.031 0.013
Before transaction costs After transaction costs
Short position – the bottom DIFF_HVOL decile with the lowest 25% DIFF_FUND -0.185 -0.061
Long position – The top DIFF_HVOL decile with the highest 25% DIFF_FUND 0.115 0.013
HEDGE (Long-Short)
0.301 (9.14)
0.074 (2.15)
Table 6 presents evidence on two approaches to enhance trading strategy performance. In Panel A, we combine fundamental score (DIFF_FUND) with Goyal and Saretto’s historical volatility signal (DIFF_HVOL). Each month, we first sort straddle options into ten DIFF_HVOL deciles. Then, for the bottom and top DIFF_HVOL deciles, we further sort options into four quartiles based on DIFF_FUND. The short position includes straddle option contracts in the bottom DIFF_HVOL decile with the lowest 25% DIFF_FUND. The long position includes straddle option contracts in the top DIFF_HVOL decile with the highest 25% DIFF_FUND. Transaction costs are incorporated into the option returns by assuming that long positions are acquired at the ask price and short positions are acquired at the bid price (effective spread = quoted spread). The sample includes 89,805 firm-month observations from January 1996 to December 2011.
46
Table 7 Regressions of straddle returns on individual fundamental signals
1 2
Intercept -0.084 (-3.54)
-0.080 (-3.31)
EARNINGS_ST 0.030 (2.24)
0.039 (2.97)
ACCRUAL_ST 0.045 (3.72)
0.046 (3.97)
GROWTH_ST 0.005 (0.29)
0.018 (1.12)
DUPONT_ST 0.029 (1.61)
0.032 (1.76)
EARNINGS_LT 0.048 (2.53)
0.050 (2.57)
ACCRUAL_LT 0.000 (0.00)
0.010 (0.62)
GROWTH_LT 0.038 (1.76)
0.038 (1.74)
DUPONT_LT -0.032 (-1.01)
-0.021 (-0.68)
DIV 0.000 (0.02)
0.020 (1.78)
HVOL 0.254 (8.53)
IVOL -0.326 (-10.10)
R2 0.017 0.027
Sum of coefficients on fundamental signals 0.162
(6.20) 0.232
(9.36) Table 7 presents multivariate regression analysis of the association between individual fundamental signals and straddle returns. The dependent variable is straddle option returns (SRETt+1). EARNINGS_ST is short-term earnings signal. ACCRUAL_ST is short-term accrual signal. GROWTH_ST is short-term growth signal. DUPONT_ST is short-term DuPont measure. EARNINGS_LT is long-term earnings signal. ACCRUAL_LT is long-term accrual signal. GROWTH_LT is long-term growth signal. DUPONT_LT is long-term DuPont signal. DIV is the dividend signal. HVOL is historical volatility. IVOL is implied volatility. Please see the Appendix for detailed variable definitions. All independent variables are transformed into decile ranks converted into the [0, 1] scale. The t-statistics in this table are based on Fama-MacBeth regressions. The sample includes 89,805 firm-month observations from January 1996 to December 2011.