Académique Documents
Professionnel Documents
Culture Documents
Investor Sentiment
Wei-Peng Chen, Junmao Chiu, Huimin Chung, Keng-Yu Ho
_____________________________________________________________________
ABSTRACT
This paper sets out to explore the relationship among investor sentiment, market
liquidity, and trading behavior. We classify the states of „fear and exuberance‟ as the
shorter-term sentiments of investors towards future returns, with the traditional
„bearish and bullish‟ states being identified as the medium-term sentiments. Our results
show that „bearish‟ and „fear‟ sentiments induce more sell orders along with a
reduction in market liquidity, and vice versa. As for the interaction between
shorter-term and medium-term sentiments, our findings show that when institutional and
individual sentiment is bearish and the investors‟ fear sentiment increases, investors would
decrease their net buying volume and market liquidity. In addition, the trading behavior
of individual investors is more sensitive than institutional investors in bearish state
when investors are more fearful of the future market movements. Finally, we find that
when investors feel more „fear‟ sentiment based on institutional sentiment in a bullish
state, they are likely to increase sell orders and the overall level of liquidity. We
suggest that the findings obtained from our exploration of the relationship between
investor sentiment and net buying volume could help to provide a better understanding
of the relationship between investor sentiment and market liquidity.
Keywords: Fear sentiment; Exuberance sentiment; Bearish sentiment; Bullish
sentiment; Net buying volume; Market liquidity
JEL Classification: G10; G11; G14
Wei-Pen Chen is with the Department of Finance, Shih Hsin University, Taiwan; Junmao Chiu and Huimin
Chung (the corresponding author) are both with the Graduate Institute of Finance, National Chiao Tung
University, Taiwan, and Keng-Yu Ho is with the Department of Finance, National Taiwan University,
Taiwan. Address for correspondence: Graduate Institute of Finance, National Chiao Tung University, 1001
Ta-Hsueh Road, Hsinchu 30050, Taiwan; Tel: +886-3-5712121, ext. 57075; Fax: +886-3-5733260; e-mail:
chunghui@mail.nctu.edu.tw.
1
1. INTRODUCTION
This paper sets out to examine whether the two types of investor sentiment, „bearish
versus bullish‟ and „fear versus exuberance‟, affect the trading behavior of market
investors, as well as the ways in which such sentiment can affect market liquidity.
Most of the previous studies on „bullish versus bearish‟ sentiment and market return
Investors (AAII) indicators.1 The typical forecast horizon of the newsletter for II is
from one to three months (Han, 2008), and AAII provides an indication as to where
individual investors expect the stock market to be in six months time (Brown and Cliff,
2004). Therefore, both II and AAII sentiment indicators can reveal the medium-term
With regard to the „fear versus exuberance‟ investor sentiments, although Whaley
(2000) argues that the implied volatility index (VIX) is the investor fear gauge, Low
(2004) nevertheless suggests that this index encompasses not only fear but also
exuberance sentiments. While some of the recent studies use the market-based fear
sentiment index, implied option volatility and the put-call ratio, when predicting future
returns,2 other studies have also used the investor fear sentiment indicators, such as high
VIX level and high put-call ratio to proxy bearish investor sentiment (Simon and
Wiggins, 2001). According to their features, the above two measures could be used to
1
Examples in the literature on the II sentiment index include Siegel (1992), Solt and Statman (1988),
Clarke and Statman (1988), Shefrin (1999), Fisher and Statman (2000, 2003, 2006), Lee et al. (2002),
Brown and Cliff (2004, 2005), Wang et al. (2006), Liu (2006), Schmeling (2007), Han (2008), Kurov
(2008), and Ho and Hung (2008). Studies on the uses of the AAII sentiment index include Brown (1999),
Bange (2000), Fisher and Statman (2000, 2003), Brown and Cliff (2004, 2005), Liu (2006), Wang et al.
(2006), Schmeling (2007), and Kurov (2008).
2
Examples in the literature using VIX/VXN as the fear sentiment indicators include Whaley (2000),
Simon and Wiggins (2001), Simon (2003), Low (2004), and Bauer et al. (2009). On the other hand,
Simon and Wiggins (2001), Wang et al. (2006), Dennis and Mayhew (2002), and Bauer et al. (2009) use
the put-call ratio as the fear sentiment indicator.
2
gauge the shorter-term investor expectations of the future returns of the underlying assets.3
market liquidity and investor trading behavior are both examined in this study. The
focus of the theoretical models has invariably been placed on the ways in which such
sentiment affects market liquidity (Baker and Stein, 2004; Deuskar, 2005). To the best
of our knowledge, no studies have yet been undertaken to document these issues
empirically. Our primary aim is to use these results to provide a better understanding of
the relationship that exists among investor sentiment, market liquidity, and trading
bullish‟ and „fear versus exuberance‟ sentiments have significant impacts on market
liquidity and net buying volume (buyer-initiated volume minus seller-initiated volume).
We then attempt to separate the bullish and bearish components of the institutional and
discernible among sentiment, net buying volume, and market liquidity based on bullish
and bearish sentiment. Finally, we examine market liquidity and trading behavior based
examine the relationships among „bearish versus bullish‟ sentiment, liquidity, and net
buying volume using the institutional and individual sentiment indicators; we also
determine whether there are any systematic differences in the influence of the bearish
and bullish sentiment. Our results indicate that more bullish investor sentiment leads to
an increase in both buy orders and market liquidity, thereby providing support for the
theoretical models of Baker and Stein (2004), Deuskar (2005); furthermore, we also find
that bearish and bullish sentiments have different impacts on market liquidity and
3
The weighting scheme is such that the composite quantity represents the implied volatility of an
at-the-money S&P 100 option hypothetical over a 30 calendar-day (22 trading-day) period.
3
investor trading behavior in individual stocks.
exist in which „fear versus exuberance‟ sentiment can affect investor trading behavior
and market liquidity. We find that an increase in fear sentiment amongst investors
decreases net buying volume and reduces market liquidity; the relationship between net
buying volume and the sentiment component of implied volatility could help to explain the
relationship between the implied volatility index and stock index returns reported by Gito
(2005).
Third, based upon the theoretical model of Baker and Stein (2004), we
hypothesize that the different trading cost can affect the relationships among investor
sentiment, market liquidity, and trading behavior. In contrast to Liu (2006),4 we try to
examine if there is any systematic difference between ETFs and individual stocks with
regard to the relationships among sentiment, market liquidity, and net buying volume.
The results indicate that, when investors become more bearish, there is likely to be a
greater reduction in the market liquidity of individual stocks than ETFs. Howoever, the
reduction in the net buying volume for individual stocks is less than that for ETFs. We
expectations; our results could provide a better understanding of how investor expectations
affect the investors‟ trading decisions and market liquidity. The results show that when
institutional and individual sentiment is bearish and the investors‟ fear sentiment increases,
investors would decrease their net buying volume and market liquidity.5 However, when
investors are fearful of the future returns on the underlying assets in an institutional
4
Using a monthly Amihud illiquidity measure that was computed from common stocks on the NYSE
and AMEX, as well as sentiment indicators which included institutional, individual and consumer
sentiment, Liu (2006) proved that more bullish sentiment can help the market become more liquid.
5
Our results are similar to the findings of Hammed et al. (2009).
4
bullish state, more sell orders would occur in the market than normal in our research
institutional investors when investors are more fearful of the future market
movements.
literature review and our hypothesis development. Section 3 describes the sample
selection procedure and research methodology. Section 4 reports and analyzes the
empirical results. Finally, the conclusions drawn from this study are presented in Section
5.
affects the market trading behavior of investors and leads to a change in market liquidity.
Two different theoretical models are used to examine the relationship between
investor sentiment and market liquidity. The first model, which was based on short
sales constraints, was constructed by Baker and Stein (2004) who demonstrated that
information contained in the order flow; thus, they argue that bullish sentiment is
market liquidity based upon the argument of an association among high perceived
volatility, low current prices, high expected returns, and the nervousness of investors.
5
bullish and bearish investor sentiment, return volatility, and the prediction of future
returns.6 For example, from their empirical study, Lee et al. (2002) find a negative
relationship between a bullish shift in sentiment and market volatility. High perceived
volatility predicts lower liquidity; therefore, when investor sentiment becomes more
liquidity.
Since investors‟ arbitrage capacity is limited, they would be unwilling to bear the
excess risk. According to the „prospect‟ theories (Kahneman and Tversky, 1979) and
the „disposition effect‟, investors that can invest will tend to hold their positions or
reduce their trading activity when they are experiencing losses. The prior literature
points out that investors are more fearful of the potential loss during the down market
period than the potential gain that they enjoy during the up-market period (Benartzi and
Thaler, 1995). As a result, when they have more fear sentiment, investors will hold
their long position or reduce their trading activity, as a result of which there will be a
The evidence provided by Baker and Wurgler (2006) is consistent with the notion
that the demand for stocks by uninformed investors is driven by investor sentiment;
thus, investors buy stocks when they are bullish, thereby driving up stock prices and
6
Examples include Lee et al. (2002), Brown and Cliff (2004), and Wang et al. (2006).
6
traders to trade more when they are bullish than when they are bearish (Baker and
Stein, 2004). Given that noise traders are prone to trend chasing, one can expect
There are also a few studies which suggest that the weather may affect the mood
and sentiment of investors. It is also noted that when investors are in a generally
pessimistic mood, they will have less desire to trade. Therefore, they may tend to sell
rather than buy (Loughran and Schultz, 2004; Goetzmann and Zhu, 2005; Chang et al.,
Hypothesis 3: Where investor sentiment becomes more bullish (bearish), this will
According to Whaley (2000), when fear sentiment is high amongst investors, there
will be a rise in demand for put options to protect their long positions, which would lead
to an increase in both implied volatility and the put-call ratio. Fear sentiment provides an
indication of investors‟ short-term expectations for the future returns of the underlying
assets, with the capacity for arbitrage by investors being limited in various ways,
including short horizons, costs, and both trading and short-selling risks; given that such
investors are risk averse, they will be unwilling to bear these risks.
Therefore, when investors are more fearful of the future returns on the underlying
asset, they will be reluctant to short-sell the stock and may tend to participate by
buying fewer stocks or by closing out their existing long positions. Thus, sell orders
returns on the underlying asset, they will reduce (increase) their net
buying volume.
7
3. RESEARCH METHODOLOGY
3.1 Sample Selection
In order to compare whether the different trading cost affects the relationships among
investor sentiment, market liquidity, and trading behavior; we use data from both ETFs
and individual stocks. Because the VIX (VXN) index is a weighted average of the
implied volatility of S&P 500 (NASDAQ 100) index options, we select SPY ETFs (the
exchange-traded funds which track the S&P 500 Index) and QQQQ (the
exchange-traded funds which track the NASDAQ 100 Index), and then follow
Ascioglu et al. (2007) using the matching firm approach to find two NYSE stocks and
We identify all individual stocks listed on the NYSE and NASDAQ exchanges
between January 2001 and December 2007 from the CRSP database and search the
NYSE Trade and Quote (TAQ) database daily summary statistics file to obtain the
daily statistics, such as the daily volume, average daily price, and standard deviation of
daily returns on all of the NYSE- and NASDAQ-listed individual stocks. We then
estimate a matching statistic for each NYSE (NASDAQ) stock and SPY (QQQQ) pair,
as follows: C C12 C22 C32 , where Ci is the (DIS - DETF)/[(DIS + DETF)/2] for
statistics i; the DIS are the individual stock daily statistics; and the DETF are the ETF
(SPY, QQQQ) daily statistics. We calculate the total matching statistics for the
individual stock-ETF pairs and also exclude the individual stocks relating to Regulation
SHO.7 Those firms with matching statistics which are lower than, or close to 0, are
better matching stocks for each ETF.8 Finally, we select GE (General Electric Co.) and
PFE (Pfizer Inc.) as the matching firms for the SPY and AAPL (AAPL Inc.) and MSFT
7
Regulation SHO (Reg.SHO) provides a new regulatory framework governing the short selling of
securities in the US equity markets. Starting from 2 May 2005, roughly 1,000 US stocks (so-called Pilot
Stocks) were traded without short-sale price tests (Diether et al., 2009).
8
The results are not reported here in order to save space; however, they are available upon request.
8
(Microsoft Corporation) as the matching firms for the QQQQ.
In this study, we employ intra-day data on ETFs and individual stocks taken from
the TAQ, using the daily abstract trading and quotes data from 9:30 am to 4:00 pm. We
follow previous literature on controlling different trading mechanism and include all the
data in AMEX, NYSE, NASDAQ and NYSE Arca (Archipelago) exchanges for our
samples. The period under examination is the post-decimalization period which runs
from 29 January 2001 to 31 December 2007; this period contains the dotcom bubble
industry cycle as well as the sub-prime mortgage crisis period. We note that AAPL and
MSFT are listed on the NASDAQ, where tick-size decimalization differs from the
convention used on the NYSE; thus, the data ranges from 26 March 2001 to 31
December 2007 for AAPL, and from 9 April 2001 to 31 December 2007 for MSFT
from the sample for the year when AAPL and MSFT switched from a NASDAQ to a
NYSE listing. Finally, according to Huang and Stoll (1996) and Chung (2006), we
eliminate the quote and trade data from our research sample.
Following Brown and Cliff (2004), we collect the direct measures of bearish and
bullish sentiment from the Investor Intelligence (II) and American Association of
market newsletters each week. Following the reading of the newsletters, the market is
organization, which asks each individual investor where they expect the stock market
will be in six months, and the results are classified as bullish, bearish, or neutral.
Hence, the II and AAII are the sentiment indicators of the medium-term expectations
9
of the institutional and individual stock markets. In the present study, we follow Wang
et al. (2006) to adopt the ratio of the bearish and bullish percentages as our measures
of investor sentiment; when they are higher (lower), market investors demonstrate
We use the sentiment components of the implied volatility indices, VIX and VXN, and
the put-call ratio to measure investor fear versus exuberance sentiment. The VIX (VXN)
index, computed from the Black-Scholes model, represents the implied volatility of S&P
According to the prior studies, several studies on S&P 100 index options find that
implied volatility contains historical volatility and has the capability of forecasting future
volatility (Day and Lewis, 1992; Christensen and Prabhala, 1998). Theoretically, implied
proposed by Shefrin (2002). Several empirical studies show that sentiment has impacts
on implied volatility (Vlad, 2004; Deuskar, 2006); Fleming (1998) also argues that
implied volatility may be an effective measure of market sentiment, noting that it contains
prediction of volatility.
volatility component which could affect market liquidity (Domowitz et al., 2001). We
thus decompose implied volatility into volatility and sentiment components. The
regression approach is often used within the prior literature to orthogonalize the
variables (Banerjee et al., 2007); for example, Baker and Wurgler (2006)
using the residuals as pure sentiment proxies. In the present study, we decompose
10
implied volatility into its volatility and sentiment components by regressing the
implied volatility index on one-year historical volatility, and we refer the residuals to
the sentiment components of implied volatility. These pure sentiment proxies are
returns on the underlying assets. Hence, we define the VIX and VXN sentiment
calculated through the use of options. It is equal to the total trading volume of the put
option contracts divided by the total trading volume of the call option contracts (Simon
and Wiggins, 2001). The PCR is also viewed as an indicator of investor fear. Whilst
the data on the VIX and VXN sentiment components and the put-call ratio are all
daily-based, the AAII and II sentiment indicators are all weekly-based. In order to
resolve this data frequency problem, we adopt the method whereby each trading day of
We use the percentage spread as the illiquidity proxy. The formula for the percentage
spread is (Askt – Bidt)/[(Askt + Bidt)/2], where Askt and Bidt are the respective intraday
ask and bid prices at time t. We then calculate the average of all the percentage spreads
in one day as the liquidity variable and use the algorithm proposed by Lee and Ready
(1991) to distinguish between all transactions that are buyer and seller initiated. The
algorithm classifies a trade as a buyer (seller) initiated trade if the traded price is higher
(lower) than the mid-point of the bid and ask price. We assign a value of +1 (–1) where it
is clear that each transaction is a buyer (seller) initiated trade, multiply the signals by
trade volume, and sum up the multipliers that occur each day. As a result, we can
ascertain whether each trade arises from buying or selling pressure, which provides the
11
net buying volume variable. Finally, we standardize the net buying volume variable by
dividing net buying volume by trading value. Therefore, the percentage standardized net
buying volume is the net buying volume variable used in our analysis.
4. EMPIRICAL RESULTS
4.1 Basic Statistics
As we can see from Table 1, the summarized statistics show that LogV (V is the daily
trading volume) is found to be larger for ETFs than for individual stocks, and the
median spread and net buying volume in SPY are both smaller than those of GE and
PFE. In addition, both individual (AAII) and institutional (II) investors are more
bullish in our sample period, especially institutional investors. Finally, we show from
the correlation statistics of the sample variables that the VIX/VXN sentiment
indicators used in our study have more sentiment components and lower volatility
components than the VIX/VXN indices.9 Thus, we argue that in the examination of
the relationship among the degree of investor uncertainty (or fear sentiment), liquidity,
and net buying volume, our implied volatility sentiment components are better proxies
We first examine the relationship between „bearish versus bullish‟ sentiment and
market liquidity using the Institutional and Individual variables and investigate the
following regression model in order to control for those factors which may be of
9
The results are not reported here in order to save space; however, they are available upon request.
12
where SPt is the daily average of the percentage spread for ETF/individual stock on
day t, Vt is the daily trading volume, Rett is the daily return, the DtWK are the day of the
week dummies, and Institutional (Individual) is the II (AAII) measure, which is the
ratio of the bearish percentage to the bullish percentage on day t. The results in Panel A
of Table 2 show that the relationships between Institutional, II, and spread are
significantly positive in all of our samples with the exception of MSFT, whereas in
terms of the relationships between Individual, AAII, and spread, only MSFT is
significantly positive.
These results provide support both for our hypothesis and the theoretical models
of Baker and Stein (2004) and Deuskar (2005), demonstrating that with more bullish
investor sentiment, investors would be more inclined to trade in the market, thereby
improving market liquidity. The results also provide support for Lee et al. (2002) by
implying that bullish sentiment is associated with higher market liquidity, lower
volatility, and low expected returns. Moreover, we find that institutional sentiment has
sentiment and market liquidity using the VIX/VXN and PCR variables and investigate
the following regression model in order to control for those factors which may be of
where the VIX/VXN variables are the sentiment components of the VIX/VXN indices
on day t, and PCR is the put-call ratio, which is the total trading volume of the put
option contracts divided by the total trading volume of all call option contracts on day t.
The remaining control variables are the same as in Equation (1). As we can see from
indicator is significant in SPY and PFE, and that the VXN indicator is significant in QQQQ
and AAPL. PCR is significant in GE, PFE, QQQQ, AAPL, and MSFT. As a result, an
increase in PCR and the VIX/VXN components leads to increase market spread.
Since the capacity for arbitrage by investors is extremely limited, they will tend to
be risk averse; thus, when investors are more fearful of the future returns on the
underlying assets, they will demonstrate their reluctance to engage in the short-selling
of stocks. Such investors are therefore likely to either hold their current long positions
or reduce their overall trading activity, with a resultant reduction in market liquidity.
These results provide support for our second hypothesis. Finally, by using our
aggregate sentiment indicators, we find that the impact on the liquidity of individual
stocks listed on the NYSE is likely to be more sensitive than the corresponding impact
on ETFs.
Secondly, we examine whether investor sentiment could affect net buying volume.
By following Chordia et al. (2002) to control for the factors that may be of importance
versus bullish‟ sentiment and net buying volume, using the following regression
model:
5 5 5
OIBNUM t ∑ 1i OIBNUM t - j ∑ 2i Max(0, Ret tj ) ∑ 3i Min (0, Ret tj )
j 1 j 1 j 1
, (3)
4
4 LogVt ∑ 5i DtWK 6 Institutio nal t 7 Individual t t
i 1
where OIBNUMt is the standardization percentage net buying volume variable, which is
the number of buyer-initiated trades less the number of seller-initiated trades divided by
10
The remaining control variables are the same as those in Equation (1).
14
In Panel A of Table 3, the relationship between individual sentiment and net buying
significantly negative for SPY. During a period of bullish investor sentiment, investors
have greater optimism and more desire to trade, with noise traders demonstrating a
penchant for trend chasing; thus, we expect to find a greater prevalence of positive
feedback trading, with investors demonstrating a general tendency to buy rather than sell.
Our results indicate that the research samples listed on the NYSE support our Hypothesis
3.
We then examine the relationship between „fear versus exuberance‟ sentiment and
where the VIX/VXN variables and PCR are the same as in Equations (2). As shown in
Panel B of Table 3, the market sentiment indicators have a strong negative correlation
with contemporaneous net buying volume. We find that the VIX is significant in both
SPY and GE, and that the VXN is significant in QQQQ, whilst PCR is significant in all
of the ETFs and individual stocks with the exception of AAPL. Furthermore, our
sentiment indicators have more significant effects on net buying volume in the ETFs and
individual stocks listed on the NYSE than those listed on the NASDAQ. Finally, we find
that both the Institutional and VIX/VXN sentiment indicators have more significant
impacts on the net buying volume of ETFs than individual stocks. Since all of our
sentiment indicators are aggregate measures, the aggregate sentiment indicators may
have more direct effects on the ETFs tracking the S&P 500 and NASDAQ 100 indices
than on individual stocks. Therefore, the sentiment indicators have larger impact on the
15
net buying volume for ETFs than for individuals stocks. In addition, it is noted that the
market depth of ETFs could be deeper than individual stocks. Such phenomenon further
leads to the smaller effect of the sentiment indicators on market spread for ETFs,
These results provide support for Hypothesis 4 to the effect that if investors have
uncertainty with regard to the future returns on the underlying assets (i.e., where the
VIX/VXN variables and PCR are higher), this will reduce the buy orders in the market.
Our results may therefore help to explain the relationship between the implied
volatility index and stock index returns found by Gito (2005), who reported a positive
Gito (2005) surmised that “if, as thought by some market practitioners, large or
very large implied volatility levels do indeed indicate oversold markets, then the
forward-looking returns for long positions in the underlying stock index triggered by
these large implied volatility levels should be attractive”. We suggest that this could be
caused by investors who are fearful of the future returns on the underlying assets, in
which case both the VIX/VXN components and PCR become higher, thereby resulting
4.3 Bearish and Bullish Sentiment, Liquidity, and Net Buying Volume
When the „bearish versus bullish‟ sentiment indicators (II and AAII) are higher than 1, it
means that for most investors, their expectations with regard to the future returns of the
underlying assets will be more bearish than bullish. In other words, aggregate sentiment
in the market is bearish. This provides us with an opportunity to compare the way in
which market liquidity and trading behavior may be affected by bullish and bearish
sentiment. First, we identify the relationship among bearish sentiment, bullish sentiment,
and market liquidity by investigating the following regression model in order to control
16
for the factors that may be of importance in determining the spread:
4
SPt 1SPt -1 2 LogVt 3 Ret t ∑ 4i DtWK 5 Institutio nalt D Inst
i 1
, (5a)
6 Institutio nalt (1 - D Inst ) t
4
SPt 1 SPt -1 2 LogVt 3 Ret t ∑ 4i DtWK 5 Individal t D Ind
i 1
, (5b)
6 Individualt (1 - D Ind
) t
Inst
where the dummy variable D is equal to 1 on days when the II indicator is equal to
Ind
or greater than 1, and D is equal to 1 on days when the AAII indicator is equal to or
Inst Ind
greater than 1. The variables Institutional × D and Individual × D therefore
indicate that institutional and individual investors expect future market movements to
The empirical results are presented in Table 4, from which we can see that the
Inst
relationship between Institutional × D sentiment and market liquidity is significantly
positive in all of the samples with the exception of MSFT. In addition, the significantly
Ind
positive relationships are discernible between Individual × D sentiment and market
Inst
liquidity, and are significantly positive for both SPY and MSFT. Institutional × (1– D )
sentiment is significantly positive for SPY, GE, PFE, and QQQQ. We then examine
whether there are any systematic differences in the influence of the bearish and bullish
sentiment. We find that there are systematic differences between Institutional bearish
and bullish sentiment in GE, PFE, and AAPL, whereas in the relationship between
Individual bearish and bullish sentiment, only GE is significant. These results show that
there are systematic differences between bearish and bullish sentiment in individual
All of these results provide support for our Hypothesis 1; furthermore, we can see
that the institutional investor sentiment indicator (II) affects market liquidity more
significantly than the individual investor sentiment indicator (AAII), particularly for
17
Inst
Institutional × D . Most of our research results indicate that sentiment has a more
significant impact on the market liquidity of individual stocks than of ETFs, since the
individual stocks are affected by short-sales constraints, with arbitrage in such stocks
tending to be particularly risky and costly. When mispricing is at its highest level,
volatility in individual stocks may be higher, in which case investors may provide more
funds for arbitrage activities or short selling; this would tend to result in a greater
reduction in the market liquidity of individual stocks than ETFs (Baker and Stein, 2004).
As a result, from our comparison between ETFs and individual stocks, we find that the
latter have more sensitive relationships between investor sentiment and market
liquidity.
and net buying volume. We investigate the following regression models according to
5 5 5
OIBNUM t ∑ 1i OIBNUM t - j ∑ 2i Max(0, Ret tj ) ∑ 3i Min (0, Ret tj ) 4 LogVt
j 1 j 1 j 1 , (6b)
4
Inst Ind
where the dummy variables Institutional × D and Individual × D are the same as
in only SPY and PFE. In addition, we also find significantly negative relationships
18
among bearish sentiment, bullish sentiment, and net buying volume for ETFs (SPY,
QQQQ), while the results for in our research on individual stocks are only a negative but
insignificant. These results provide support for our Hypothesis 3. In addition, most of
our findings suggest that individual bearish sentiment has a more significant impact on
net buying volume than individual bullish sentiment. Our results are similar to those of
Institutional bearish and bullish sentiment in PFE and MSFT, whereas in terms of the
relationship between Individual bearish and bullish sentiment, only PFE is significant.
Overall, we find that there are systematic differences between bearish and bullish
We also find that the relationship between institutional sentiment and net buying
consistent with the results of Equation (3). Since financial intermediaries and hedge
funds often trade and hedge their activities in the ETF markets, the net buying volume
stocks.
investors. When investors envisage a rise in the stock market in six months‟ time, the
bullish percentage will be higher than the bearish percentage; however, unexpected
negative news could raise investor fear sentiment, since investors are more concerned
with the potential for a short-term loss, and they are unwilling to bear the short-term
11
They examine the good news and bad news in the up and down markets. Their results indicate that
investors could be more inclined to engage in herding behavior in the down market.
19
excess risk in the future. Investors will thus tend to close out their long positions,12
In the previous literature, when using the moving average indicator, Kavajecz and
Odders-White (2004) found that when the long-run moving average is higher than the
short-run moving average, the extreme selling pressure which triggers a negative
moving average signal causes the quotes to walk down the book; thus, there will be an
on market liquidity and net buying volume when investors are more fearful of the
future returns on the underlying assets in either the bullish or bearish states.
Investor) Expectations
We now examine whether investors could close out their existing long positions or
hold their positions to affect market liquidity when investors are more fearful of the
future returns on the underlying assets in institutional bullish and bearish states. The
the future returns on the underlying assets in an institutional bearish state, and vice
versa.
12
See Pring (1995) and Whaley (2000).
20
when investors are more fearful of the future returns on the underlying assets in the
Inst
institutional bullish and bearish states. Those results for the VIX/VXN × D indicator,
which are positive in all of the samples, are also significant in SPY, two NYSE stocks,
Inst
and AAPL. In addition, PCR × D has a positive impact on market liquidity in
QQQQ and AAPL. These results indicate that when institutional investors already
expect the stock market to be lower in the future, if investors further exhibit fear
(exuberance) in regard to the future returns in the stock market, they would reduce
(increase) their trading within the market in the shorter-term period. The results are
similar to Hammed et al. (2009) who found that when the stock price lasts for up to
two weeks in large down markets, it tends to be higher in periods of high liquidity
commonality.
Inst
However, we have also found that the relationship between PCR × (1– D ) and
market spread is significantly negative for GE and PFE, and there is no significantly
Inst
positive relationship based on VIX/VXN × (1– D ). This result is interesting and
indicates that when institutional investors have bullish expectations of future market
movements, if investors are fearful of the future returns of the stock market, the market
liquidity would increase. Such findings are similar to those of Kavajecz and
Odders-White (2004).
We further investigate whether the causal trading behavior leading to the negative
21
5 5 5
OIBNUM t ∑ 1i OIBNUM t - j ∑ 2i Max(0, Ret tj ) ∑ 3i Min (0, Ret tj ) 4 LogVt
j 1 j 1 j 1 , (8b)
4
Inst Inst
where the VIX/VXN × D and PCR × D are the same as in Equations (7a) and (7b).
Inst
As shown in Models 3 and 4 of Table 6, the relationship between VIX/VXN × D and
net buying volume is significantly negative for SPY. The relationship between net
Inst
buying volume and PCR × D has no negative significance for SPY, GE, and MSFT.
Inst
We also find that the relationship between PCR × (1– D ) and net buying volume is
significantly negative in all the ETFs and individual stocks with the exception of
AAPL. Our results find that the net buying volume decrease when institutional
investors expect that the stock market will be higher in the future with unexpected
information leading to an increase in investor fear sentiment. Since investors have very
limited capacity for arbitrage, they are also likely to be extremely risk averse, and
short selling in individual stocks could be more costly under such circumstance,
investors would be unwilling to bear the short-term excess risk and thus choose to sell
stocks.
Specifically, for GE and PFE, we find that investors would be more likely to close
out their existing long positions during a bullish period in the market than in a bearish
Inst
period, and PCR ×(1– D ) has an impact on net buying volume that is less negatively
significant than the Table 3 coefficient. In addition, the relationship between PCR ×(1–
Inst
D ) and market spread is significantly negative in Models 1 and 2 of Table 6. Since
institutional investors expect the stock market to be higher in the future, this will lead
to an increase in net buying volume; if investors exhibit fear in regard to the future
returns in the stock market, this could lead to an increase in the sell orders, and thereby
causing order imbalance and market spread to reduce. These results help to provide an
understanding of the results reported in Models 1 and 2 of Table 6, and why the
22
Inst
relationship between PCR × (1– D ) and market spread is significantly negative in
GE and PFE.
Investor) Expectations
Finally, we examine whether or not investors, when they are more fearful of the
future returns on the underlying assets in their individual bullish and bearish states,
could close out their existing long positions to affect market liquidity. The empirical
the future returns on the underlying assets in an individual bearish state and vice versa.
Models 1 and 2 of Table 7 show that all of the relationships between VIX/VXN ×
Ind
D and the spread are significantly positive for SPY, PFE, and QQQQ, whereas
Ind
VIX/VXN × (1– D ) is barely significantly positive for SPY and AAPL. Regarding the
PCR sentiment indicators in the bullish and bearish states, we find that a significantly
Ind
positive relationship exists between PCR × D and market spread except for AAPL.
Ind
We also find a significantly positive relationship between PCR × (1-D ) and the
market spread for all the ETFs and individual stocks with the exception of SPY and
GE. In our ETFs results, we find that market spread becomes larger for individual
investors in the bearish state than in bullish state when unexpected information leads to
an increase in investor fear sentiment. We then examine the type of trading behavior
which gives rise to these results by examining how the interaction between
23
shorter-term and individual investor expectations affects the net buying volume. The
OIBNUM t = α + ∑β1iOIBNUM t - j + ∑β2i Max(0, Ret tj ) + ∑β3i Min (0, Ret tj ) + β4 LogVt +
j =1 j =1 j =1 , (10a)
4
5 5 5
OIBNUM t ∑ 1i OIBNUM t - j ∑ 2i Max(0, Ret tj ) ∑ 3i Min (0, Ret tj ) 4 LogVt
j 1 j 1 j 1 , (10b)
4
Ind Ind
where the VIX/VXN × D and PCR × D are the same as in Equations (9a) and (9b).
indicators are all negatively significant in all the ETFs and individual stocks.
significance in this relationship for the samples listed on the NASDAQ. In addition,
Ind
the relationships between net buying volume and the PCR × D indicators are all
negatively significant except for AAPL. These results are similar with the findings in
Table 6. Our results also indicate that individual investors being fearful of the future
returns on the stock market would potentially lead to a decrease in net buying volume
that is more negatively significant than institutional investors and the Table 3
coefficient, thereby causing order imbalance and market spread to increase. Since
investors (Schmeling, 2007), and individual investors exhibit myopic loss aversion13
13
Benartzi and Thaler (1995) combined two behavioral concepts, loss aversion and mental accounting,
to provide a theoretical foundation for the observed equity premium puzzle. They find that investors
24
(Bernartzi and Thaler, 1995; Gneezy, Kapteyn and Potters, 2003), when investors are
more fearful of the future market movements, the trading behavior of individual
5. CONCLUSIONS
This study distinguishes between the „bearish versus bullish‟ and „fear versus
exuberance‟ sentiments and examines how these two types of sentiment affect market
liquidity and net buying volume. We find that that when investor sentiment becomes
more bullish, there is a reduction in market spread and an increase in buy orders. This
result provides support for the findings of Baker and Stein (2004) and Deuskar (2005).
Our results further show that when investors are more fearful of the future returns of the
underlying assets, investors will tend to hold on to their long positions or reduce their
trading activity. In other words, the investors decrease their buy orders in the market and
thereby leading to a reduction in market liquidity. Our results could help to explain the
We also find that there are systematic differences between bearish and bullish
sentiment in individual stocks for market liquidity and net buying volume.
Furthermore, the institutional investor sentiment indicator (II) affects market liquidity
more significantly than the individual investor sentiment indicator (AAII). These results
are similar to the Schmeling (2007) hypothesis that the sentiment of individuals has a
different effect than the sentiment of institutions. Regarding the comparison between
individual stocks and ETFs, when investors have more bearish sentiment, the reduction
in market liquidity could be greater for individual stocks than for ETFs; thus, any
reduction in net buying volume for individual stocks will be less than that for ETFs.
Such findings are more significant in the case of institutional investors, since all of our
seem to be more concerned with the potential for a short term loss than in planning for the relevant time
horizon and accepting periodic short-term losses.
25
sentiment indicators reflect the aggregate sentiment of market participants. In addition,
the previous results may also due to that individual stocks involve more trading costs
than ETFs do, and their market depth is also less than that of ETFs.
In terms of interaction sentiment, we find that when investors expect the stock
market to go down in the future with some unexpected information increasing investor
reduction in market liquidity. Our results are similar to those of Hammed et al. (2009)
in that the stock price remains stable for up to two weeks in large down markets and is
higher in periods of high liquidity commonality. Secondly, we also discover that when
investors are fearful of the future returns on the underlying assets in an institutional
bullish state, there will be a more significant reduction in net buying volume than
normal, causing an increase in the market liquidity of individual stocks. These findings
would appear to indicate that investors will tend to close out their long positions under
such circumstances. In addition, our results are consistent with those of Kavajecz and
Odders-White (2004) in that when the long-run moving average is higher than the
short-run moving average; selling pressure would increase in the market, which will
also increase sell-side market depth. Finally, in terms of the interaction between
shorter-term and individual investor expectations, we find that, for individual investors,
the reduction in market liquidity in the bearish state is more sensitive than that in
bullish state for the ETFs when unexpected information leads to an improvement in
more sensitive than institutional investors when investors are more fearful of the future
market movements. Overall, the findings obtained from our exploration of the
relationship between investor sentiment and net buying volume help to provide a better
26
REFERENCES
Ascioglu, A., M. Aydogdu, R.K. Chou and L.P. Kugele (2007), „An Analysis of
Intraday Patterns in ETF and Common Stock Spreads‟, Working Paper, Bryant
University.
Baker, M. and J.C. Stein (2004), „Market Liquidity as a Sentiment Indicator‟, Journal
Baker, M. and J. Wurgler (2006), „Investor Sentiment and the Cross-section of Stock
Banerjee, P.S., J.S. Doran and D.R. Peterson (2007), „Implied Volatility and Future
Bange, M. (2000), „Do the Portfolios of Small Investors Reflect Positive Feedback
Brown, G.W., and M.T. Cliff (2004), „Investor Sentiment and the Near-term Stock
Chang, S.C., S.S. Chen, R.K. Chou and Y.H. Lin (2008), „Weather and Intraday
Patterns in Stock Returns and Trading Activity‟, Journal of Banking and Finance,
32: 1754-1766.
Chordia, T., R. Roll and A. Subrahmanyam (2002), „Order Imbalance, Liquidity and
27
Christensen, B.J. and N.R. Prabhala (1998), „The Relationship between Implied and
Evidence from the ADR Market‟, Journal of Banking and Finance, 30: 1485-1505.
Clarke, R.G. and M. Statman (1998), „Bullish or Bearish?‟, Financial Analysts Journal,
54(3):63-72.
Day, T.E. and C.M. Lewis (1992), „Stock Market Volatility and the Information
De Long, J.B., A. Shleifer, L.H. Summers and R.J. Waldmann (1990), „Noise Trader
Deuskar, P. (2005), „Aggregate Liquidity and the Role of Sentiment Traders‟, Working
Diether, K.B., K.H. Lee and I. Werner (2009), „It‟s Show Time! Short-sales Price-test
Domowitz, I., J. Glen and A. Madhavan (2001), „Liquidity, Volatility and Equity Trading
Costs across Countries Over Time‟, Journal of International Finance, 4(2): 221-255.
Fisher, K.L. and M. Statman (2000), „Investor Sentiment and Stock Returns‟,
Fisher, K.L. and M. Statman (2003), „Consumer Confidence and Stock Returns‟,
Fisher, K.L. and M. Statman (2006), „Market Timing in Regressions and Reality‟,
Gito, P. (2005), „Relationships between Implied Volatility Index and Stock Index
Gleason K.C., I. Mathur and M.A. Peterson (2004), „Analysis of Intraday Herding
Behavior among the Sector ETFs‟, Journal of Empirical Finance, 11: 681-694.
Gneezy, U., A. Kapteyn and J. Potters (2003), „Evaluation Periods and Asset Prices in
Goetzmann, W.N. and N. Zhu, (2005), „Rain or Shine: Where is the Weather Effect?‟,
Han, B. (2008), „Investor Sentiment and Option Prices‟, Review of Financial Studies,
21: 387-414.
Hammed, A., W. Kang and S. Viswanathan (2009) „Stock Market Declines and
Huang, R.D. and H.R. Stoll (1996), „Dealer versus Auction Markets: A Paired
Kavajecz, K.A. and E.R. Odders-White (2004), „Technical Analysis and Liquidity
29
Lee, C.M.C. and M.J. Ready (1991), „Inferring Trade Direction from Intraday Data‟,
Lee, W.Y., C.X. Jiang and D.C. Indro (2002), „Stock Market Volatility, Excess Returns
and the Role of Investor Sentiment‟, Journal of Banking and Finance, 26:
2277-2299.
Liu, S. (2006), „Investor Sentiment and Stock Market Liquidity‟, Working Paper,
University of Texas.
Loughran, T. and P. Schultz (2004), „Weather, Stock Returns and the Impact of
343-364.
Low, C. (2004), „The Fear and Exuberance from Implied Volatility of S&P 100 Index
Pring, M.J. (1995), Investment Psychology Explained: Classic Strategies to Beat the
Shefrin, H. (2002), „On Kernels and Sentiment‟, Working Paper, Santa Clara
University.
Shleifer, A. and L.H. Summers (1990), „The Noise Trader Approach to Finance‟,
Siegel, J.J. (1992), „Equity Risk Premia, Corporate Profit Forecasts, and Investor
Sentiment around the Stock Crash of October 1987‟, Journal of Business, 65:
557-570.
30
Simon, D.P. and R.A. Wiggins (2001), „S&P Futures Returns and Contrary Sentiment
Simon, D.P. (2003), „The Nasdaq Volatility Index During and After the Bubble‟,
Solt, M. E. and M. Statman (1988), „How Useful is the Sentiment Index?‟, Financial
Vlad, D.G. (2004), „Investors‟ Beliefs and their Implications on Asset Pricing, Excess
University.
Wang, Y.H., A. Keswani and S.J. Taylor (2006), „The Relationships between Sentiment,
Whaley, R.E. (2000), „The Investor Fear Gauge‟, Journal of Portfolio Management, 26:
12-17.
31
Table 1a Descriptive Statistics of the Dependent and Control Variables
Ticker Symbols
Variables
SPY GE PFE QQQQ AAPL MSFT
Ret
Mean 0.0000 –0.0001 –0.0004 –0.0001 –0.0003 0.0001
Median 0.0006 –0.0004 –0.0004 0.0010 –0.0005 0.0000
S.D. 0.0107 0.0167 0.0161 0.0183 0.0163 0.0174
LogV
Mean 17.6084 16.9462 16.8724 18.3113 16.9092 17.9718
Median 17.6638 16.8973 16.8547 18.3118 16.9138 17.9587
S.D. 0.7894 0.3942 0.5106 0.3633 0.5344 0.3600
SPREAD
Mean 0.0466 0.1676 0.2324 0.0793 0.0683 0.0386
Median 0.0233 0.1187 0.1187 0.0444 0.0491 0.0389
S.D. 0.0483 0.1569 0.1997 0.0757 0.0483 0.0087
OIBNUM
Mean 0.5787 4.9946 0.3036 0.3485 0.6433 -0.3049
Median 0.2875 3.8210 2.0585 0.2730 0.4961 -0.1353
S.D. 5.7139 9.5468 13.7885 3.1918 2.2121 1.9433
Note: The descriptive statistics are provided for the dependent and control variables, comprising of Ret, LogV,
SPREAD and OIBNUM. The data covers the period from 29 January 2001 to 31 December 2007 for GE,
PFE, SPY and QQQQ, from 26 March 2001 to 31 December 2007 for AAPL, and from 9 April 2001 to 31
December 2007 for MSFT.
Implied Volatility
VIX 19.0337 17.3400 7.0941
VXN 28.4717 21.3700 14.7387
Market Sentiment
VIXSENT –0.0030 –0.9102 4.4678
VXNSENT 0.5682 –0.7545 5.1415
Put-Call Ratio 0.6789 0.6600 0.1396
Direct Sentiment
Individual 0.7820 0.6874 0.4481
Institutional 0.5729 0.5399 0.2011
Note: The descriptive statistics are provided for the sentiment-related variables, comprising of VIX, VXN,
VIXSENT, VXNSENT and Put-Call Ratio. The data covers the period from 29 January 2001 to 31
December 2007 for GE, PFE, SPY and QQQQ, from 26 March 2001 to 31 December 2007 for AAPL, and
from 9 April 2001 to 31 December 2007 for MSFT.
32
Table 2 Investor Sentiment and Market Liquidity
NYSE NASDAQ
Variables SPY-VIX GE-VIX PFE-VIX QQQQ-VXN AAPL-VXN MSFT-VXN
Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value
Panel A: Bearish versus Bullish
Institutional 0.0054 3.014*** 0.0611 5.214*** 0.0572 4.277*** 0.0085 2.720*** 0.0456 2.051** –0.0006 –0.819
Individual 0.0011 1.326 –0.0065 –1.291 –0.0063 –1.067 –0.0001 –0.131 0.0044 0.461 0.0004 1.972**
SPt – 1 0.9245 94.582*** 0.8171 59.219*** 0.8193 58.400*** 0.9421 117.706*** 0.9093 82.889*** 0.9191 98.536***
LogV –0.0029 –5.054*** 0.0038 –0.735 –0.0253 –4.677*** –0.0037 –2.305** -0.0044 –4.567*** 0.0000 0.158
Ret –0.0934 –3.397*** –0.6774 –5.625*** –0.7278 –5.005*** –0.1481 –5.169*** –0.0749 –2.840*** –0.0038 –6.225***
C 0.0509 5.004*** –0.0628 –0.708 0.4383 4.723*** 0.0068 2.300** 0.0801 4.672*** 0.0003 6.062***
Observations 1740 1740 1740 1740 1449 1439
2
R 0.9370 0.7215 0.7642 0.9181 0.8893 0.852
Note: The dependent variable is the daily percentage spread on trading day t, which is regressed on lagged market liquidity, LogV, the return on day t, day-of-the-week dummies, direct
sentiment (Individual and Institutional), and market sentiment (VIX, VXN and Put-Call Ratio). ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.
33
Table 3 Investor Sentiment and Net Buying Volume
NYSE NASDAQ
Variables SPY-VIX GE-VIX PFE-VIX QQQQ-VXN AAPL-VXN MSFT-VXN
Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value
Panel A: Bearish versus Bullish
Institutional –2.883 –3.172*** 2.036 1.199 2.168 1.197 –0.370 –0.770 0.299 0.877 0.225 0.766
Individual –0.010 –0.003 –1.807 –3.212*** –1.328 –2.116** –0.228 –1.133 0.005 0.034 –0.101 –0.801
LogV –0.059 –0.319 –1.429 –2.377** –1.888 –3.370*** –0.016 –0.073 –0.022 –0.191 0.259 1.795*
C 2.851 0.858 26.271 2.576*** 31.155 3.272*** 1.311 0.322 0.349 0.178 –5.025 –1.954*
Observations 1740 1740 1740 1740 1449 1439
2
R 0.025 0.138 0.478 0.046 0.049 0.135
Panel B: Fear versus Exuberance
VIX/VXNSENT –0.271 –6.304*** –0.192 –3.307*** –0.008 –0.136 –0.044 –2.384** 0.002 0.157 0.013 1.158
Put-Call Ratio –2.729 –2.608*** –8.088 –5.012*** –5.482 –3.488*** –1.294 –2.276** –0.709 –1.471 –0.932 –3.145***
LogV 0.424 2.222** –1.268 –2.045** –1.613 –2.863*** 0.189 0.787 0.047 0.398 0.227 1.551
C –6.215 –1.729* 27.484 2.553** 31.491 3.402*** –2.115 –0.474 0.279 0.144 –3.534 –1.331
Observations 1740 1740 1740 1740 1449 1439
2
R 0.046 0.154 0.481 0.051 0.049 0.141
Note: The dependent variable is the daily percentage standardized net buying volume on trading day t, which is regressed on lagged percentage standardized net buying volume and past
positive and negative elements of the market return on day t, LogV, day-of-the-week dummies, direct sentiment (Individual and Institutional), and market sentiment (VIX, VXN and
Put-Call Ratio). The individual parts of the dependent variable results are not reported here in order to space; however, they are available upon request. ***, **, and * indicate
significance at the 1%, 5%, and 10% levels, respectively.
34
Table 4 Bearish and Bullish Sentiment and Market Liquidity
NYSE NASDAQ
Variables SPY-VIX GE-VIX PFE-VIX QQQQ-VXN AAPL-VXN MSFT-VXN
Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value
Panel A: Institutional
Inst
Institutional × D 0.0057 3.104*** 0.0376 5.089*** 0.0599 4.305*** 0.0001 2.856*** 0.0061 2.581*** –0.0007 –1.221
Inst
Institutional × (1 – D ) 0.0068 3.535*** 0.0048 3.675*** 0.0400 2.735*** 0.0007 2.208** 0.0020 0.745 –0.0008 –1.161
SPt – 1 0.9260 95.220*** 0.8168 59.133*** 0.8176 57.877*** 0.9421 118.239*** 0.9060 81.354*** 0.8482 61.876***
LogV –0.0027 –4.830*** 0.0024 0.257 –0.0276 –5.188*** 0.0003 –2.424** –0.0048 –4.814*** –0.0003 –1.108
Ret –0.0977 –3.572*** –0.6444 –5.511*** –0.7268 –4.989*** –0.0015 –5.176*** –0.0768 –2.912*** –0.0430 –6.321***
C 0.0463 4.719*** –0.0166 –0.204 0.4820 5.205*** 0.0694 2.421** 0.0879 4.952*** 0.0132 2.143**
Observations 1740 1740 1740 1740 1449 1439
2
R 0.9371 0.7193 0.7642 0.9181 0.8894 0.7405
F-test 0.4487 4.9408*** 2.0371* 0.2784 2.8118** 0.008
Panel B: Individual
Ind
Individual × D 0.0017 2.067** –0.0023 –0.429 0.0008 0.142 0.0001 0.815 –0.0007 –0.626 0.0005 1.686*
Ind
Individual × (1 – D ) 0.0011 0.681 –0.0172 –1.425 –0.0065 –0.521 0.0005 0.116 –0.0009 –0.396 0.0009 1.486
SPt – 1 0.9368 105.275*** 0.8408 64.842*** 0.8358 61.802*** 0.9508 130.005*** 0.9132 84.317*** 0.8498 63.010***
LogV –0.0026 –4.485*** 0.0061 1.162 –0.0265 –4.908*** –0.0003 –2.116** –0.0045 –4.564*** –0.0005 –1.609
Ret –0.0848 –3.089*** –0.6228 –5.151*** –0.6856 –4.672*** –0.0014 –4.946*** –0.0712 –2.703*** –0.0432 –6.418***
C 0.0474 4.650*** –0.0701 –0.784 0.4844 5.288*** 0.0653 2.223** 0.0842 4.935*** 0.0148 2.465**
Observations 1740 1740 1740 1740 1449 1439
2
R 0.9367 0.7179 0.7619 0.9177 0.8890 0.7407
F-test 0.3427 4.4062*** 0.7692 0.1931 0.0147 0.9226
Note: The dependent variable is the daily percentage spread on trading day t, which is regressed on lagged market liquidity, LogV, the return on day t, and day of the week dummies. The dummy
variable DInd is equal to 1 on days when the AAII indicator is equal to or greater than 1, and DInst is equal to 1 on days when the II indicator is equal to or greater than 1. Therefore,
Individual × DInd and Institutional × DInst indicate that individual and institutional investors expect future market moments to be bearish rather than bullish, and vice versa. Parts of the
estimated results are not reported here in order to save space; however, they are available upon request. The F-test examines whether the Institutional × DInst is equal to Institutional × (1–
DInst) in Panel A, and whether the Individual × DInd is equal to Individual × (1– DInd) in Panel B. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.
35
Table 5 Bearish and Bullish Sentiment and Net Buying Volume
NYSE NASDAQ
Variables SPY-VIX GE-VIX PFE-VIX QQQQ-VXN AAPL-VXN MSFT-VXN
Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value Coeff. t-Value
Panel A: Institutional
Institutional × DInd –2.839 –2.921*** –0.351 –0.291 1.978 1.318 –0.471 –0.979 0.292 0.848 0.015 0.005
Institutional × (1– DInd) –2.919 –3.142*** –1.339 –1.141 –0.206 –0.134 –0.736 –1.427 0.320 0.847 0.403 1.130
LogV –0.061 –0.350 –1.689 –2.825*** –2.193 –3.932*** –0.086 –0.399 –0.019 –0.175 0.278 1.917*
C 2.902 0.910 30.669 3.005*** 36.657 3.831*** 2.586 0.654 0.302 0.154 –5.541 –2.128**
Observations 1740 1740 1740 1740 1449 1439
2
R 0.025 0.134 0.477 0.045 0.049 0.136
F-test 0.009 0.363 2.222* 0.355 0.007 2.154*
Panel B: Individual
Individual × DInd –0.598 –1.685* –1.495 –2.576*** –1.667 –2.594*** –0.314 –1.725* –0.023 –0.145 –0.148 –1.112
Individual × (1– DInd) –0.885 –1.652* –1.513 –1.324 –3.461 –2.687*** –0.302 –0.754 –0.133 –0.421 –0.382 –1.472
LogV 0.057 0.305 –1.451 –2.402** –1.938 –3.514*** –0.008 –0.037 –0.011 –0.092 0.253 1.753*
C 0.185 0.057 27.246 2.672*** 33.990 3.682*** 1.071 0.264 0.381 0.205 –4.675 –1.815*
Observations 1740 1740 1740 1740 1449 1439
2
R 0.020 0.137 0.479 0.046 0.052 0.136
F-test 0.342 0.001 4.314*** 0.002 0.287 1.914
Note: The dependent variable is the daily percentage standardized net buying volume on trading day t, which is regressed on lagged percentage standardized net buying volume and past
positive and negative elements of the market return on day t, LogV, and day of the week dummies. The dummy variable DInd is equal to 1 on days when the AAII indicator is equal to or
greater than 1, and DInst is equal to 1 on days when the II indicator is equal to or greater than 1. Therefore, Individual × DInd and Institutional × DInst indicate that individual and
institutional investors expect future market moments to be bearish rather than bullish, and vice versa. Parts of the estimated results are not reported here in order to save space; however,
they are available upon request. The F-test examines whether the Institutional × DInst is equal to Institutional × (1– DInst) in Panel A, and whether the Individual × DInd is equal to
Individual × (1– DInd) in Panel B. ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively
36
Table 6 Interaction between Shorter-term and Medium-term (Institutional Investor) Expectations
2. GE-VIX
Put-Call Ratio –0.0143 –1.127 –0.213 –5.378***
VIX/VXN 0.0014 2.518** –0.130 –2.042**
VIX/VXN × DInst 0.0004 4.234*** –0.135 –1.095
VIX/VXN × (1 – D ) Inst –0.0000 –1.403 –0.003 –0.605
Put-Call Ratio × D Inst
0.0147 0.728 –6.518 –3.057***
Put-Call Ratio × (1 – D Inst
) –0.0326 –2.477** –6.544 –4.670***
2
R 0.733 0.722 0.151 0.153
3. PFE-VIX
Put-Call Ratio 0.0621 3.489*** –6.944 –3.869***
VIX/VXN 0.0026 3.737*** –0.038 –0.575
Inst
VIX/VXN × D 0.0042 3.518*** 0.067 0.522
VIX/VXN × (1 – D ) Inst –0.0001 –1.366 –0.003 –0.550
Put-Call Ratio × D Inst –0.0320 –1.359 –3.082 –1.268
Put-Call Ratio × (1 – D Inst
) –0.0705 –4.594*** –5.413 –3.443***
2
R 0.766 0.767 0.481 0.481
Note: The dependent variable is the daily percentage spread on trading day t in Models 1 and 2; the independent variables include lagged
market liquidity, LogV, the return on day t, day-of-the-week dummies, the put-call ratio, and VIX/VXN. Dependent variables in
Models 3 and 4 are the daily percentage standardized net buying volumes on trading day t, They are regressed on lagged percentage
standardized net buying volume and past positive and negative components of the market return on day t, LogV, day-of-the-week
dummies, and the put-call ratio. The dummy variable DInst is equal to one on days when the II indicator is greater than or equal to
one., Therefore, a higher (lower) VIX/VXN × DInst, PCR × DInst, VIX/VXN × DInst, or PCR × DInst indicates more (less) fear amongst
investors regarding the future returns on the underlying asset in a bearish state. Parts of the estimated results are not reported here in
order to save space; however, they are available upon request. ***, **, and * indicate significance at the 1%, 5%, and 10% levels,
respectively.
37
Table 6 (Contd.)
2. AAPL-VXN
Put-Call Ratio 0.0083 2.417** –0.621 –1.354
VIX/VXN 0.0002 2.308** –0.002 –0.146
Inst
VIX/VXN × D 0.0004 2.460** 0.016 0.587
VIX/VXN × (1 – D ) Inst –0.0001 –0.451 –0.001 –1.353
Put-Call Ratio × D Inst
0.0081 1.875* –0.126 –0.219
Put-Call Ratio × (1 – D Inst
) 0.0031 0.906 –0.028 –0.661
2
R 0.890 0.890 0.054 0.049
3. MSFT-VXN
Put-Call Ratio 0.0024 3.379*** –1.288 –1.611
VIX/VXN 0.0001 0.445 0.017 1.442
VIX/VXN × DInst –0.0001 –0.789 5.349 0.114
VIX/VXN × (1 – D ) Inst –0.0004 –1.079 –1.049 0.539
Put-Call Ratio × D Inst
0.0013 1.256 –1.456 –3.369***
Put-Call Ratio × (1 – D Inst
) 0.0020 2.850** –0.936 –3.238***
2
R 0.7423 0.7419 0.137 0.143
38
Table 7 Interaction between Shorter-term and Medium-term (Individual Investor) Expectations
Note: The dependent variable is the daily percentage spread on trading day t in Models 1 and 2; the independent variables include
lagged market liquidity, LogV, the return on day t, day-of-the-week dummies, the put-call ratio, and VIX/VXN. Dependent
variables in Models 3 and 4 are the daily percentage standardized net buying volumes on trading day t, They are regressed on
lagged percentage standardized net buying volume and past positive and negative components of the market return on day t,
LogV, day-of-the-week dummies, and the put-call ratio. The dummy variable DInd is equal to one on days when the AAII
indicator is greater than or equal to one. Therefore, when VIX/VXN × DInd, PCR × DInd, VIX/VXN × DInd, or PCR × DInd is
higher, it means that investors are more fearful regarding the future return on the underlying assets in the bearish state, and
vice versa. Parts of the estimated results are not reported here in order to save space; however, they are available upon request.
***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.
39
Table 7 (Contd.)
40