Post on 30-May-2020
1
Sentiment-prone investors and volatility dynamics between spot and
futures markets
Pilar Corredor, Elena Ferrer and Rafael Santamaria
Public University of Navarre
October, 2012
Abstract
This paper analyses the role of investor sentiment in the contemporaneous dynamics of
spot and futures markets and in volatility spillovers between them. They are a potential
effect of high investor sentiment leading to an increase in noisy trading and a drop in
arbitrage activity due to institutional investors’ attempts to limit their risk exposure.
This reduces correlation between the spot and futures markets. Consistent with the
impact of overconfidence and self-attribution bias, both of which are stronger in noise
traders, prices take longer to adjust news. In fact, shocks on volatility have less impact
during periods of high sentiment.
Keywords: Investor Sentiment, noise traders, spot-futures correlation, volatility
spillovers
JEL: G10, G13, G14
2
Sentiment-prone investors and volatility dynamics between spot and futures
markets
1. -Introduction
The introduction of futures markets brought about a significant improvement in
the news transmission mechanism by allowing a more rapid adjustment of prices to new
information (Antoniou et al, 1998). In fact, by attracting additional traders, futures
markets can increase the possible channels of cross-market information flow (Cox,
1976). This may well determine the way information is incorporated into both futures
and spot market prices. The extent of the impact will depend upon the types of traders
active in the two markets (Antoniou et al, 1998). Noise traders, in particular, react to
information in a way that would not occur in a fully rational model because they trade
on noise as if it were information (Black, 1986). Similarly, Shiller (1984) claims that
some investors use a trend-chasing strategy based upon so-called “popular models” that
can be related to fundamentals, but involve an element of overreaction to news. In the
same vein, Shleifer and Summers (1990) show that uninformed investors are likely to
overreact to news. More recently, Kumar (2009) has shown empirically that individual
investors exhibit stronger behavioural biases when assets are more difficult to value and
when market-level uncertainty is higher. In contrast, several authors have shown that
institutional investors are sophisticated traders, emphasizing that their superior capacity
to acquire and process information gives them an advantage over other types of traders.
Their presence contributes to efficient asset pricing (see Bartov et al, 2000, Jiambalvo et
al, 2002, Collins et al, 2003 and Lewellen, 2011).
Thus, factors that might influence the investor mix, in either or both of these
markets, could also provoke changes in the dynamics of information transmission
between spot and futures markets. The literature has in fact shown that regulatory
reform or changes in the overall economic environment have had considerable impact
on these dynamics1. In this respect, investor sentiment can be a key variable. Noise
traders tend to be more active in bullish than in bearish markets (Baker and Stein, 2004)
and to have less capacity to react to news, since their overconfidence and self-attribution
biases increase in the presence of high market sentiment. Yu and Yuan (2011) also
1 See, for example, the effect of variation in the transaction costs of futures markets (Aragó et al, 2003) or the
changing nature of volatility contagion between financial markets (Saha and Chakrabati, 2011).
3
argue that sentiment-driven investors participate and trade more aggressively in high-
sentiment periods, due to their reluctance to take short positions in low-sentiment
periods. More recently, Antoniou et al (2012) also state that noise traders are less active
during pessimistic periods than optimistic ones. Indeed investors usually take short
positions during bad times but such positions are more difficult to initiate for noise
traders than long positions (which they actively take during good times). In addition,
behavioural finance has shown that the arbitrage activity of informed traders is limited
when investor sentiment is high because of noise trading risk, that is, the risk that arises
from the unpredictability of noise traders’ behaviour. In periods such as these, informed
investors will stay out of the market (Shleifer and Vishny, 2003). Sophisticated traders,
aware of the overpricing that accompanies moments of high market sentiment, may also
significantly reduce their exposure at such times, thereby increasing the role of noise
traders in price setting. This difference in trading behaviour can affect the trading
volume and investor mix in both these markets. The extent of the effect on trading
volume in the spot market is unclear, because the reduced activity of institutional
investors may be offset, wholly or in part, by a significant influx of noise traders. In the
futures market, however, the predominance of institutional investors (Kavussanos et al
2008 and Bohl et al, 2011) means that trading volume is likely to drop significantly.
Change in the investor mix, however, will be more marked in the spot market, which
will see a more significant increase in the presence of noise traders, while the futures
market will continue to be dominated by institutional investors. Finally, the reduced
activity of institutional investors in both these markets may have a significant impact on
the level of arbitrage between them.
These circumstances raise the interest in examining the impact of the level of
investor sentiment on the contemporaneous dynamics of the spot and futures markets,
and on volatility spillovers between the two. The focus of the study is to analyse the
joint dynamics of several stock indexes and their respective futures contracts,
specifically, S&P500 index for the US market, and the CAC40, the DAX30, the
IBEX35, the FTSE100, and the Eurostoxx50 for the European market.
This study makes several contributions to the literature. Firstly, this, as far as we
know, is the first attempt to analyse the impact of investor sentiment, as a latent variable
affecting trading behaviour and the investor mix, on the contemporaneous dynamics of
the spot and futures markets and on volatility spillovers between them. In more detailed
4
terms, this study attempts to investigate issues such as the significance of changes in the
contemporaneous correlation between the two markets during periods of high market
sentiment; the impact of own-market or other-market news on volatility; and the extent
of the asymmetric effect on volatility of good or bad news from either market. These
issues will be explored using bivariate GJR models to examine the time-varying
correlation between financial markets taking into account the investor sentiment level.
As well as for academics, this study holds interest for practitioners, because
knowledge and understanding of the variables influencing the degree of integration
between the two markets and the mechanisms by which news is incorporated into spot
and futures prices and transmitted across markets are important when considering
trading or hedge positions.
The rest of this article comprises four more sections. Section 2 discusses the
theoretical framework for the analysis and the formulation of the hypotheses to be
tested. Section 3 describes the data; section 4 presents the empirical model and the
results, and section 5 summarizes the main conclusions.
2. Theoretical Framework and Testable Hypotheses
If interest rates and dividend yields were non-stochastic, in a perfectly frictionless
world, price movements in the spot and futures markets would be contemporaneously
perfectly correlated and non-cross autocorrelated (Chan, 1992). Relationship between
price movements in the futures index and underlying spot markets should be
instantaneous, because they are both driven by the same market information and both
reflect the aggregate value of the underlying shares. Thus, in efficient market
conditions, it would make no difference to trade in one market or the other. Under
certain market conditions (liquidity, transaction costs, investor typology), however, one
market may assimilate new information more quickly than the other, thereby affecting
volatility spillovers.
Classical finance theory neglects the role of investor sentiment assuming investors
to be rational. Even if some investors are not rational, arbitrageurs can exploit their
irrational behaviour, thus causing prices to reflect future discount cash flows. The
behavioural finance literature suggests, however, that investor sentiment, defined as
investors’ opinions regarding future cash flows and investment risk (Chang et al, 2012),
5
affects trading decisions. The influence of investors’ future expectations may result in
mispricing that will affect pricing models.
Early US empirical studies focused on the role of sentiment in predicting stock
returns (Kothari and Shanken, 1997; Neal and Wheatley, 1998; Shiller, 1981, 2000;
Baker and Wurgler, 2000; and Brown and Cliff, 2005) and on the effect of sentiment on
small-stock premiums (Lee et al, 1991; Swaminathan, 1996; Neal and Wheatley, 1998;
Brown and Cliff, 2004; and Lemmon and Portniaguina, 2006). Research on the
sentiment-return relationship in other financial markets includes, Wang (2001) on the
futures market; Han (2008), Lemmon and Ni (2011) on the options market; Ahn et al
(2002) on the currency market; and Burghardt et al (2008) and Schmitz et al (2009) on
the warrants market. A more modest amount of research has been conducted on the
effect of sentiment on volatility (Brown (1999) or Lee et al (2002)) finding them to be
inversely related.
To the best of our knowledge, however, there is no research examining the effect
of sentiment on the interaction between the spot and futures markets. The key question
is whether it is reasonable to expect the level of investor sentiment to affect the joint
dynamics of these two markets.
A possible argument to support such an idea is variation in the mix of traders who
are active when market sentiment is high. For example, noise traders tend to trade more
when markets are bullish than when they are bearish (Baker and Stein, 2004; Yu and
Yuan, 2011; and Antoniou et al, 2012). Barber and Odean (2008) argue that individual
traders are more prone to cognitive biases and Kumar (2009) finds empirical evidence
to support this, especially in assets that are hard to value and during periods of higher
market-level uncertainty. This noise trader risk pushes asset prices away from
equilibrium (Barberis et al, 1998 or De Long et al, 1990) and makes institutional traders
less inclined to engage in arbitrage trading. They may also prefer less exposure in the
equity market in the knowledge that this kind of assets, especially those that are hard to
value or present limited arbitrage opportunities, are over-priced and will tend towards
medium- to long-term reversion (see Baker and Wurgler, 2006). Institutional trading
will not affect spot and futures markets to the same degree, however. In fact, De Long et
al (1990) report a higher percentage of this type of trader in markets dealing in complex
assets, such as the futures market. Kavussanos et al (2008) argue that the futures market
is less prone to noise trader risk, and Bohl et al (2011) find futures markets to be
6
dominated by institutional investors, who are assumed to be informed or rational. These
sophisticated traders, may reduce their arbitrage activity and their exposure in equity
markets, and thereby reduce trading volume to a greater extent in futures markets than
in spot markets. At the same time, however, the larger increase in noise trading that
occurs in spot markets will probably cause a more significant change in investor mix
than it does in futures markets.
During periods of market optimism, these changes in the investor mix, together
with less arbitrage activity due to lower participation of institutional investors, could
reduce the price correlation of these two markets within the no-arbitrage band2, by
lowering the pressure for price movements within that band. Indeed, any drop in
investor activity will, in itself, reduce the correlation between the two markets because
trading volume and correlation are directly related (see Stoll and Whaley, 1990, Chan,
1992). In the same vein, Bohl et al (2011) show that derivatives and spot markets will
correlate increasingly as institutional investors become more active. With this in mind,
we test the following hypothesis:
H1: Periods of high market sentiment reduce the correlation between spot and
futures markets.
According to the noise trading hypothesis, order flow is less informative when
investors are optimistic. Daniel et al (1998) assume that investors are overconfident
about their private information. If investors are also affected by self-attribution bias,
they will react asymmetrically to confirming versus disconfirming pieces of news and
become even more over confident after receiving confirming news. Self-attribution bias
leads investors to under react to the release of public information. The conservatism bias
hypothesis states that investors do not fully adjust their priors to the arrival of new
information (Barberis et al, 1998). During periods of high investor sentiment, these
biases will make investors in general, and noise traders in particular, less alert to
information coming from their own market, thus reducing the impact of volatility
shocks. By the same token, they will also pay less attention to information coming from
the other market. Furthermore, noise traders’ reaction to bad news that contradicts their
2This band is given by ( , ) where and ; where t
is the current date; T is the expiration date of the futures contract; S is the price of the underlying asset at time t; ρ= ln(1 + i), i being the riskless interest rate; is the time T value of dividends paid on the component stocks between
t and T. Finally, and are the present values of the sum of transaction costs involved in the arbitrage strategies.
7
prior beliefs will have less impact on price formation. This means that, during periods
of high investor sentiment, the impact of news from either market will be less
asymmetric. Thus, the hypotheses to be tested are as follows:
H2: During periods of high market sentiment, the impact of own-market news on
volatility will be weaker and less asymmetric.
H3: During periods of high market sentiment, the impact of other-market news on
volatility will be weaker and less asymmetric.
3. Database
For the implementation of the analysis, this study uses daily closing prices and
trading volume of the spot and futures markets for a period running from February 2001
to December 2011. The data are taken from the US stock market and four European
markets: namely France, Germany, the United Kingdom and Spain. The EuroStoxx50 is
also included in order to represent the Euro zone. The reason for this choice of
European markets is that the UK, France and Germany are considered, along with the
US and Japan3, as extremely prominent economies on the global stage (Chang et al,
2012). According to the World Federation of Exchanges classification for 2011, the
London SE Group is the largest European stock exchange grouping in terms of
capitalization followed by NYSE Euronext (Europe), the Deutsche Börse and the BME
Spanish Exchanges. The homogeneity of their financial development levels does not
rule out some variation in shareholder structure, corporate governance (see La Porta et
al, 1998) and cultural dimensions (see Hofstede, 2001) between the selected European
countries, however. The market sample also includes representatives of both the Anglo
Saxon and Continental financial systems. This combination of similarity and diversity
strengthens the relevance of our findings by allowing us to determine whether
institutional factors, unrelated to financial development, play a significant role in the
impact of investor sentiment on cross-market correlation and volatility spillovers.
The closing prices data, taken from the Datastream database (Thomson Financial),
refer to the S&P500 index for the US stock market and to the five key European stock
market indexes, namely, the CAC40 for France, the DAX30 for Germany, the FTSE-
100 for the UK, the IBEX35 for Spain, and the EuroStoxx50 index. The closing prices
3 Although it would have been interesting to include Japan, the necessary data were unavailable.
8
of the respective futures contracts were drawn from the Bloomberg database. The
returns for the spot index (St,i) and the futures index (Ft,,i) computed each day t for each
index i are defined as Rs,t,i=Ln(St,i/St-1,i) and Rf,t,i=Ln(Ft,i/Ft-1,i).
The trading volume data on these markets (spot and futures) are drawn from the
Datastream and the Bloomberg database, respectively. The variable used in the analysis
is abnormal trading volume4, calculated for each index as:
(1)
where is the ordinary trading volume of each index i (S&P500, CAC40, DAX30,
IBEX35, FTSE100 and EuroStoxx50) on day t, for each market m (s=spot and
f=futures).
Another of the variables considered in this analysis is investor sentiment. Previous
studies have used a variety of sentiment indicators, and there is no consensus as to the
best means of representing this unobservable variable. Indicators used in previous
research include: investor survey findings (Jansen and Nahuis, 2003; Brown and Cliff,
2005; Lemmon and Portniaguina, 2006; Schmeling, 2009; and Stambaugh, et al, 2012),
investor mood (Kamstra et al, 2003), retail investor trades (Barber et al, 2006;
Greenwood and Nagel, 2009; and Kumar and Lee, 2006), mutual fund flows (Brown et
al, 2003; Frazzini and Lamont, 2008 and Ben-Rephael et al, 2012)), the dividend
premium (Baker and Wurgler, 2004a and b), the closed-end fund discount (Zweig,
1973; Lee et al, 1991; Swaminathan, 1996; Neal and Wheatley, 1998 and Doukas and
Milonas, 2004), option implied volatility (Whaley, 2000), the number of IPOs and
average first-day IPO returns (Ritter, 2003 and Ljungqvist et al, 2006), turnover or
trading volume (Jones, 2002; Sheinkman and Xiong, 2003; and Baker and Stein, 2004),
the share of equity issues in total equity and debt issues (Baker and Wurgler, 2000),
insider trading (Seyhun, 1998) or composite sentiment indexes (Brown and Cliff, 2004;
Baker and Wurgler, 2006, 2007; Ho and Hung, 2009; Baker et al, 2012; and Chang et
al, 2012) among others.
4 The selected measure is similar to that used in papers such as Llorente et al (2002), Dennis and Strikland (2002) or
Covrig and Ng (2004). Given that our interest is in trading volume in futures markets, we use trading volume instead
of turnover.
9
For the purposes of our proposed analysis, we require a short-term measure of
sentiment. The majority of the above references describe long-term timing measures
used to test their predictive power on future stock returns. The majority, moreover, are
market-based measures whose construction requires complementary techniques that
may bias the final results. The measure selected will also need to be a frequently-
published and updated indicator, the mode of construction and date of which are known
and understood by traders.
To obtain a measure fulfilling all these requirements, we select two surveys that
directly measure the sentiment of market participants. For the U.S. market, we follow
DeBondt (1993), Fisher and Statman (2000) and Brown and Cliff (2004) whose
sentiment measure is based on the American Association of Individual Investor (AAII)
survey data. Originally started, in 1987, as a weekly survey of randomly selected AAII
members, this survey asks participants to predict the likely direction of the stock market
during the next six months and measures the percentages of individual investors that
respond “up”, “down”, and “the same”. The AAII then labels these responses as a
bullish, bearish or neutral on the stock market, respectively.
For a measure of investor sentiment in the European indexes analysed, we use
survey data from SentixEuroStoxx 505. Since this survey began in February 2001, it has
surveyed Sentix investors weekly, and currently has over 3100 registered participants,
more than 77% of whom are individual investors. Participants are asked whether they
are bullish, bearish, neutral, or have no opinion with regard to the future trend of the
EuroStoxx50 stock index over the following one- and six-month periods6.
We use the two surveys measures (Sentix and AAII) as the spread between the
percentages of bullish and bearish investors. Every week is sorted according to the level
of investor sentiment as either a bullish (above-the-median) sentiment week or a bearish
(below-the-median) sentiment week7. Both the AAII and the Sentix survey meet the
necessary criteria with respect to frequency and trader awareness and both indexes
capture market sentiment well because they are calculated from a direct survey on the
5 In the absence of any sentiment measure of this kind for the UK, we consider this a valid approximation. 6 The results shown are those obtained using the Sentix 6 month-ESX 50 Index to be consistent with the AAII. For
robustness checks, we later repeat the analysis using the Sentix 1 month-ESX 50 Index. 7 Replies to the weekly survey are accepted up to Friday of the week in question, but the results are not published
until the following Monday before trading opens. For the purposes of our study, we take the moment of
optimism/pessimism to be Friday when the survey replies are being given. Repetition of the analysis using a dummy
variable beginning the day after close of survey produced similar results. The results are available upon request.
10
expected future state of the market. The survey results are also comparable because of
the homogeneity similarity of the question they put to the participants. The investor
sentiment data were also drawn from Datastream.
4. Methodology and Results
4.1 Preliminary results: Trading volume analysis
The argument to support the idea of possible variation in the contemporaneous
dynamics between the futures and spot markets is based, not only on the activity of
noise traders or institutional investors, but also on the change in the trader mix that
occurs during periods of high market sentiment. Firstly, as noted by Baker and Stein
(2004), Yu and Yuan (2011) and Antoniou et al (2012) there is an increase in the
number of noise traders in bullish markets. Given that the highest concentration of noise
traders is found in less complex assets, their activities will be more noticeable in spot
markets. Furthermore, seeing the market to be overpriced, institutional traders are likely
to limit their activity until prices to revert to their fundamentals. This drop in trading by
institutional investors will occur in both markets, but more significantly in the futures
markets where they dominate (Bohl et al, 2011) and where their activity will not be
offset by an increase in that of noise traders.
To obtain empirical evidence to support these arguments, we begin by testing for a
variation in the abnormal trading volume in both the spot and the futures market at
times of high investor sentiment. In addition, because the Engle’s test results reveal the
presence of ARCH effects, the variance is modelled by means of a GARCH(1,1)
specification, which takes the following form:
(2)
where follows a N(0, );
where is the abnormal daily trading volume for market m (spot or futures) and
index i. As independent variables, we include a dummy (SENT), which takes a value of
1 if investor sentiment is above the median level and 0 otherwise and 4 day-of-the-week
dummies ( which take a value of 1 if it is Monday, Tuesday,
11
Thursday or Friday, respectively and 0 otherwise. The equation is estimated using an
AR(5).
As shown in table I, Panel A, during periods of high investor sentiment, overall
trading volume in spot markets is not significantly affected. This may be because the
drop in trading by rational investors is offset by the increase in noise trading at such
times (Baker and Stein, 2004; Yu and Yuan, 2011; and Antoniou et al, 2012). These
markets do, nevertheless, see a major change in the investor mix due to the large influx
of noise traders.
The results for the futures market, shown in table I, Panel B, show a negative
effect which is clearly significant in all indexes considered8. In an optimistic market,
trading volume decreases as institutional investors, who are the principal agents in these
markets, decide to close their positions and temporarily cease trading in order to avoid
exposure to the arbitrage risk created by irrational investors, and await the subsequent
reversion of prices to fundamentals. Given the strong presence of institutional investors
in these markets, their reduced activity will have more impact on trading volume.
However, since it is not accompanied in futures markets by an increase in noise trading,
it may not have same the impact on the investor mix as it does in spot markets.
This observed difference in the trading volume and, probably, also in the investor
mix in both markets strengthens the rationale for testing their capacity to trigger
alterations in price dynamics between spot and futures markets.
4.2 Empirical model
In order to model the effects of investor sentiment on the correlation between spot
and futures index returns and between the linkage in the second moments of the two
markets, we propose a bivariate Glosten-Jagannathan-Runkle (1993) (GJR) process.
The model (henceforth, Model 1) for each index i (i = S&P500, CAC40, DAX30,
IBEX35, FTSE100 and EuroStoxx50) takes the following form9:
8 The exception is the Spanish index. Nevertheless, this negative effect is only significant at a 14% significance level. 9 Although not reported in the tables, some diagnostic tests of the residuals were performed. No indications of model
misspecification were observed. The autocorrelations and partial correlations for the squared standardized residuals
for stock index and index futures returns are all insignificantly different from zero.
12
; (3)
where is the error correction term imposing the long-term equilibrium
on index i in the two markets; ( ) is the innovation in the spot (futures) market
at day t for index i; =var( /Ωt-1,i) is the conditional variance of the spot market
and =var( /Ωt-1,i) is the conditional variance of the futures market, where Ωt,i is
the information set available at t for index i.
As shown, in the above variance equation the cross-market innovations have been
added to a GJR specification. It is interesting to note that the innovation ( ) is
used instead of ( ). The reason for this choice is the intense cross-correlation
between and which could lead to misleading estimates. The innovation
( ) is the information from the spot (futures) market which is transmitted to the
futures (spot) market and is not included in ( ). Thus, ( ) is orthogonal
to ( )10
. and have been incorporated into the and equations
respectively, to analyse the volatility spillover between the two markets on each index
i. ( ) is a dummy variable which is 1 if <0 ( <0) y 0 otherwise;
( ) is a dummy variable which is 1 if <0 ( <0) and 0 otherwise, and is the
returns correlation between the two markets. In the specification of the covariance, the
constant correlation implied in the cost-of-carry model is11assumed.
In order to test hypothesis 1, we introduce the dummy variable (SENT) into the
model to allow this correlation to change as a function of investor sentiment. As already
stated, this variable takes a value of 1 when the sentiment index is above the median
10 ( ) se calculan como los residuos de la siguiente regresión ( )=k0,i+k1,i ( )+ ( )
11 The covariance specification is similar to that used in Koutmos and Tucker (1996) which is based on the
specification in Bollerslev, (1990).
13
level and 0 otherwise. As sentiment proxies, we use the AAII for the US index and the
Sentix index for the European indexes under analysis. The coefficient of these
indexes indicates whether there is a change in the contemporaneous correlation between
the futures and spot markets. Consistent parameter estimates are obtained using the
BHHH algorithm.
Furthermore, these equations allow these innovations ( and ) to influence
the conditional volatility asymmetrically, as do their own innovations ( and ).
Thus, and measure the magnitude effect, whereas and measure the sign
effect. The intuitive interpretation of these coefficients is very similar to that of their
own innovations, but they are relative to cross-market volatility spillovers.
4.3 Impact of investor sentiment on correlation between spot and futures markets
The estimates from Model 1 are shown in table II. With respect to the means, it is
worth noting the significantly negative sign of the coefficient on the lagged return in all
spot and futures markets analysed. Meanwhile, the error correction term parameter is
significant in all of the markets. The parameter data for the conditional variance
equation show that volatility is affected by own-market shocks12 ( and ). Both the
persistence coefficients ( and ) and the asymmetry coefficients ( and ), are
positive and significant with values that fall within the usual ranges, thus confirming
that negative shocks increase volatility within a given market.
The model also captures other parameters affected by global volatility spillovers
and negative shocks. We find that both the parameters involved in global information
transfer from the other market ( and ) are, as expected, positive and significant
overall13. In the case of the parameters involved in the asymmetric impact of negative
shocks ( and ) the results are less clear because only some of them are positive and
significant14.
12 In the case of DAX30 and S&P500 these parameters are not significantly different from zero. 13 Given the availability of the sentiment indicator affecting the trend of the German market DAX30, we performed a
robustness test by repeating the analysis using this measure. Since the findings were practically the same as for the
Sentix Eurostoxx50, we decided to adopt the latter for its consistency with other European markets. The results are
available from the authors upon request. 14Note that the figures of these parameters ( , and ) are not comparable to ( , and ) because they
are obtained using instead of
14
With respect to hypothesis H1, given that the model permits the correlation to vary
as a function of market sentiment, we need to examine the parameter that is associated
with this change ( ). The results reveal that, when investor sentiment is high,
correlation decreases in all the markets analysed. This decrease is significant at the 1%
level in all cases. This finding appears to support the hypothesis that, when sentiment is
high, noise traders become more active, while institutional investors decrease their
activity. These changes do not have the same impact in both types of market, however.
The significant influx of noise traders to spot markets widens the gap in terms of
investor mix between these and futures markets, thereby reducing the contemporaneous
price correlation between the two. Moreover, the decrease in the proportion of
institutional investors in both markets reduces arbitrage activity and allows prices to
deviate further from their fundamentals. This obviously results in lower correlation
between the two markets, thus confirming Hypothesis 1.
4.4 Effects of investor sentiment on the volatility of its own market
In order to analyse the effect of sentiment on the information, we adjust Model 1
to include the dummy variable (SENT) described earlier, but now also associated to any
information coming from the market under analysis (Model 2) and to negative news
coming from its own market (Model 3). We also include the SENT variable as it affects
information coming from the other market (Model 4) and the asymmetric response of
volatility to news coming from the other market (Model 5). The unrestricted model onto
which we impose different restrictions to create the rest of the above-mentioned models
is presented below:
(4)
15
Given that the aim of this section is to analyse the effect of sentiment on
information coming from its own market, the models to be analysed are, specifically,
Model 2 and Model 3, which impose the following restrictions:
Model 2: = 0; and =0
Model 3: = 0; and =0
The estimates from Model 2 are given in table III. The variables it shares with
Model 1 behave, overall, as described earlier. Observation of the coefficients associated
to the influence of sentiment on information (α6 for the spot market and β6 for the
futures market), shows that they are in 5 of 6 cases negative and significant. The
negative sign tells us that, during periods of high investor sentiment, information
reaching the market has a lower impact on prices, consistent with over-confidence and
self-attribution among uninformed investors, and thus less impact on volatility. These
arguments are confirmed by the results for both types of markets.
The data on the effect of sentiment on the asymmetric impact on volatility of own-
market bad news are given in table IV. The coefficients on the variable used to capture
the effect of sentiment on volatility asymmetry (α8 and β8) are clearly significant. In
fact, all six indexes analysed show a significant decrease in volatility asymmetry in the
presence of negative shocks. Once again, we observe this pattern in both types of
markets.
This set of results confirms hypothesis H2 and suggests that when investor
sentiment is high, news plays a somewhat less important role in price setting driven by
the biases of noise traders, whose percentage presence at such times is higher than when
investor sentiment is low. As expected, this less prominent role of information is
particularly noticeable in the asymmetric effect on volatility, probably as a consequence
of noise traders’ failure to react to bad news that contradicts their prior beliefs.
4.5 Effects of investor sentiment on volatility spillovers
The next step is to test the effect of sentiment on volatility spillovers. For this,
starting from the unrestricted model described in the previous section, we devise two
16
new models, Model 4 which analyses the impact on information coming from the other
market and Model 5 which examines the asymmetric effect of that information on
volatility. In more specific terms, the said models impose the following restrictions on
the general model:
Model 4: = 0; and =0
Model 5: = 0; and =0
The estimates from Model 4 are shown in table V. Coefficients α7 and β7 capture
the impact of sentiment on information coming from the other market. It should be
noted that the cross-market shocks considered, including those affected by sentiment,
are orthogonal to the information originating in their own markets. As can be seen,
when sentiment is high, we find a generalized decrease in cross-market volatility
spillovers. In fact, both coefficients in all 6 indexes analysed are highly statistically
significant. These results are consistent with those obtained for the effect on own-
market volatility. It is also important to note that news can originate not only from the
release of exogenous information, but also from that of endogenous information
conveyed through trading. If, during periods of high sentiment, there is a drop in
trading, there will also be a drop in trading news and, presumably, in the amount of
trading news being transmitted to the other market.
Table VI gives the coefficient estimates obtained from the estimation of Model 5.
Coefficients α9 and β9 associated with the effect of sentiment on the transmission of
negative shocks in the spot and futures markets, respectively, are nearly all negatively
signed, although significant only in the case of futures on FTSE100. This means that, in
this case, the level of investor sentiment does not affect the asymmetric reaction of
volatility to negative shocks coming from the other market.
These results only partially confirm H3, since although we observe, as predicted,
that, during periods of high investor sentiment, volatility in one market is less affected
by news coming from the other, the decrease in volatility asymmetry following bad
news from the other market lacks statistical significance.
17
Finally, the findings vary very little across the cases analysed, allowing us to
conclude that they are robust to possible country-specific institutional or cultural
factors, at least, that is, in the developed market context in which this paper is situated.
4.6 Robustness
Our purpose in this next section is to analyse the robustness of the results reported
above, by examining two issues: a) their sensitivity to the selected dummy for
extremely bullish sentiment and b) their sensitivity to the time horizon for the sentiment
measure15.
The first test is to adjust the sentiment dummy in order to check the robustness of
the results to its mode of construction. This variable was initially defined to identify a
period in which market sentiment had risen above the median level. In this new
analysis, the variable is adjusted to capture periods of more extreme levels of sentiment.
Taking the top 25% to be high sentiment periods, the variable takes a value of 1 in these
periods and 0 otherwise. Table VII summarizes the coefficient estimates for this
analysis. The results show that cross-market correlation drops significantly during
periods of high investor sentiment, thus confirming H1. They also reveal that volatility
is less affected by news from either market. At the same time, volatility asymmetry
during such periods is found to be less affected by own-market news, while the effect of
other-market news remains unchanged. This confirms H2 and partially confirms H3.
This consistency with the results of the initial analysis confirms their robustness to the
construction of the sentiment variables.
Finally, as already stated, the Sentix survey issues two EuroStoxx forecasts: a one-
month forecast and a six-month forecast. The results given in the tables shown so far are
based on the six-month forecast. However, since the AAII issues only a six-month
forecast, we repeated the analysis using the Sentix EuroStoxx one-month forecast. The
resulting coefficient estimates, given in table VIII, are similar to those reported above,
in that high market sentiment triggers a significant decrease in correlation, the reaction
of volatility to own-market news (models 2 and 3) and volatility spillovers (models 4
and 5). This clear drop in correlation allows us to confirm H1. The results for H2 and
H3, however, differ slightly from those reported in the earlier analyses. Although the
impact of news on volatility decreases, as predicted in both these hypotheses, there is a
15 The robustness test will be available only for the Sentix measure, since AAII does not consider horizons of less
than 6 months.
18
difference in the asymmetric impact of bad news on volatility. While there is barely any
significant change in the effect of “own-market” bad news, a large number of the
markets analysed show a significant reduction in the impact of “other-market” negative
news. This enables confirmation of H3 and partial confirmation of H2.
Overall, the results obtained, both in terms of correlation and the information
effect show no major variations attributable to the choice of time horizon for estimation
of the sentiment variable or to its mode of construction, and can therefore be considered
highly robust. The only difference worth noting is that which can be observed in the
asymmetric impact of news on volatility. When we use the six-month sentiment index,
asymmetric volatility decreases only as a reaction to shocks in its own market, whereas,
when we use the one-month sentiment index, it is found to decrease in response to news
from the other market. These findings confirm the impact of investor sentiment on
volatility asymmetry, although the type of information that produces the effect appears
to depend on the time horizon.
5. Conclusions
This study establishes a link between the published research on volatility
dynamics and investor sentiment. Through its potential influence on investor behaviour,
high sentiment can have a significant impact on volatility dynamics. Noise traders, in
particular, will show an increased presence in the market, while sophisticated investors,
faced with higher arbitraging risk driven by the irrational behaviour of noise traders, and
conscious of over-pricing, will reduce their activity until prices revert to their
fundamental values. Due to the characteristic differences between spot and futures
markets, these changes affect their trading volume and investor mix in different ways
and may therefore significantly alter the contemporaneous dynamics between them. To
explore this issue, we analyse spot and futures markets on stock market indexes in
different countries: the S&P500 for the US, and a representative set of European
indexes (CAC40, DAX30, FTSE100, IBEX35 and Eurostoxx50).
Consistent with expectations, during periods of high investor sentiment in all of the
countries considered, trading volume drops notably in the futures markets due to the
significant reduction in the activity of institutional investors. The effect in spot markets
is not significant because the reduction in the activity of institutional investors is offset
by an increase in the participation of noise traders. This variation in the investor mix
19
can have a major impact on the joint volatility dynamics between the two markets. In
fact, the results show that the level of cross-market correlation decreases significantly in
all the countries analysed. This is due not only to the imbalance created by the activity
of noise traders themselves but also to institutional investors slackening their arbitrage
activity, unless prices deviate considerably from the no-arbitrage bands. Consistent with
the impact of overconfidence and self-attribution bias, which is stronger in individual
investors and during periods of higher market-level uncertainty, prices take longer to
adjust news. In fact, shocks on volatility in either market have significantly less impact
during periods of high sentiment. To a lesser degree, the same can be said of the
asymmetric impact of negative shocks on volatility, although it is worth noting that the
results are sensitive to the time horizon employed in the estimation of investor
sentiment. This issue, while exceeding the scope of the present paper, might be an
interesting avenue of future research.
Finally, the results obtained are very similar across all the markets analysed,
suggesting that cultural and institutional frameworks do not play a crucial role in this
issue, or at least not in the developed market context in which this paper is situated.
These findings reveal that the joint dynamics of the spot and futures markets is
strongly influenced by the diversity and mix of investors at any given moment and also
by variables affecting trading behaviour, one being investor sentiment. The latter’s
usefulness in describing cross-market conditional correlation and the reaction of stock
prices to news justifies examination of its role in the dynamics of these two markets.
Acknowledgements: This paper has received financial support from the Spanish
Ministry of Science and Innovation (ECO2009-12819) and Spanish Ministry of
Economy and Competitiveness (ECO2012-35946-C02-01)
References
Ahn, C., Lee, E-B., and Suh, E-H., (2002): “Noise trading in the Korean foreign
exchange market: some questionnaire evidence”, Bank of Korea Economic Papers, 5,
pp. 133-155.
Antoniou, C., Doukas, J. A., and Subrahmanyam, A., (2012): “Sentiment and the
CAPM”, Available at SSRN: http://ssrn.com/abstract=2101724.
20
Antoniou, A. Holmes, P., and Priestley R., (1998): “The effects of stock index futures
trading on stock index volatility: an analysis of the asymmetric response of volatility to
news”, Journal of Futures Markets, 18, pp. 151-166.
Aragó, V., Corredor, P., and Santamaría, R., (2003): “Transaction costs, arbitrage, and
volatility spillover: a note”, International Review of Economics and Finance, 12, pp.
399-415.
Baker, M., and Stein, J., (2004): “Market liquidity as a sentiment indicator”, Journal of
Financial Markets, 7, pp. 271-299.
Baker, M., and Wurgler, J., (2000): “The equity share in new issues and aggregate stock
returns”, Journal of Finance, 55, pp. 2219-2257.
Baker, M., and Wurgler, J., (2004a): “A catering theory of dividends”, Journal of
Finance, 59, pp. 1125-1165.
Baker, M., and Wurgler, J., (2004b): “Appearing and disappearing dividends: the link to
catering incentives”, Journal of Financial Economics, 73, pp. 271-288.
Baker, M., and Wurgler, J., (2006): “Investor sentiment and the cross-section of stock
returns”, Journal of Finance, 61, pp. 1645-1680.
Baker, M., and Wurgler, J., (2007): “Investor sentiment in the stock market”, Journal of
Economic Perspectives, 21, pp. 129-151.
Baker, M., Wurgler, J., and Yu, Y., (2012): “Global, local and contagious investor
sentiment”, Journal of Financial Economics, 104, pp. 272-287.
Barber, B.M., y Odean, T., (2008): “All that glitters: the effect of attention and news on
the buying behaviour of individual and institutional investors”, Review of Financial
Studies, 21, pp. 785-818.
Barber, B. M., Odean, T., and Zhu, N., (2006): “Systematic noise”, Journal of Financial
Markets, 12, pp. 547-569.
Barberis N., Sheifer A. and Vishny R. (1998): “A model of investor sentiment” Journal
of Financial Economics, 49, pp 307-343.
Bartov, E., Radhakrishnan, S. and Krinsky, I., (2000): “Investor sophistication and
patterns in stock returns after earnings announcements”, Accounting Review, 75, pp. 43-
63.
Black, F., (1986): “Noise”, Journal of Finance, 41, pp. 529-543.
Ben-Rephael, A., Kandel, S., and Wohl, A., (2012): “Measuring investor sentiment with
mutual fund flows”, Journal of Financial Economics, 104, pp. 363-382.
21
Bohl, M.T., Salm, C. A., y Schuppli, M., (2011): “Price discovery and investor structure
in stock index futures”, Journal of Futures Markets, 31, pp. 282-306.
Bollerslev, T., (1990): “Modelling the coherence in short-run nominal exchange rates:
A multivariate generalized ARCH approach”, Review of Economic and Statistics, 72,
pp. 498–505.
Brown, G.W., (1999): “Volatility, sentiment and noise traders”, Financial Analysts
Journal, 55, pp. 82-90.
Brown, G. W., and Cliff, M. T., (2004): “Investor sentiment and the near-term stock
market”, Journal of Empirical Finance, 11, pp. 1-27.
Brown, G. W., and Cliff, M. T., (2005): “Investor sentiment and asset valuation”,
Journal of Business, 78, pp. 405-440.
Brown, S., Goetzmann, W. N., Hiraki, T., Shiraishi, N., and Watanabe, M., (2003):
“Investor sentiment in Japanese and U.S. daily mutual fund flows”, NBER Working
Paper 9470, http://www.nber.org/papers/w9470.
Burghardt, M., Czink, M., and Riordan, R., (2008): “Retail investor sentiment and the
stock market”, Available at SSRN: http://ssrn.com/abstract=1100038.
Chan, K., (1992): “A further analysis of the lead-lag relationship between the cash
market and stock index futures market”, Review of Financial Studies, 5, pp. 123-152.
Chang, Y. Y., Faff, R., and Hwang, C-Y., (2012): “Local and global sentiment effects,
and the role of legal, information and trading”, Available at SSRN:
http://ssrn.com/abstract=1800550 or http://dx.doi.org/10.2139/ssrn.1800550.
Collins, D.W., Gong, G. and Hribar, P., (2003): “Investor sophistication and the
mispricing of accruals”, Review of Accounting Studies, 8, pp. 251-276.
Covrig, V. and Ng, L., (2004): “Volume autocorrelation, information and investor
trading”, Journal of Banking and Finance, 28, pp. 2155-2174.
Cox, C.C. (1976): “Futures trading and market information”, Journal of Political
Economy, 84, pp. 1215-1237.
Daniel, K., Hisrsleifer, D. and Subramanyan, A., (1998): “Investor psychology and
security markets under and-overreactions”, Journal of Finance, 53, 1839-1885.
De Bondt, W.F.M., (1993): “Betting on trends: intuitive forecasts of financial risk and
return”, International Journal of Forecasting, 9, pp. 355-371.
De Long, J.B., Shleifer, A., Summers, L.H., and Waldmann, R.J., (1990): “Noise trader
risk in financial markets”, Journal of Political Economy, 98, pp. 703-738.
22
Dennis, P. and Strikland, D., (2002): “Who blinks in volatilite markets, individuals or
institutions?”, Journal of Finance, 57, pp.1923-1949.
Doukas, J. A., and Milonas, N. T., (2004): “Investor sentiment and the closed-end fund
puzzle: out-of-sample evidence”, European Financial Management, 10, pp. 235-266.
Fisher, K.L., and Statman, M., (2000): “Investor sentiment and stock returns” Financial
Analysts Journal, 56, pp.16-23
Frazzini, A., and Lamont, O., (2008): “Dumb money: mutual fund flows and the cross-
section of stock returns”, Journal of Financial Economics, 88, pp. 299-322.
Glosten, L.R., Jagannathan, R., and Runkle, D.E., (1993): “On the relation between the
expected value and the volatility of the nominal excess return on stocks”, Journal of
Finance, 48, pp. 1779-1801.
Greenwood, R., and Nagel, S., (2009): “Inexperienced investors and bubbles”, Journal
of Financial Economics, 93, pp. 239-258.
Han, B., (2008): “Investor sentiment and option prices”, Review of Financial Studies,
21, pp. 387-414.
Ho, C. and Hung, C.-H., (2009): “Investor sentiment as conditioning information in
asset pricing”, Journal of Banking and Finance, 33, pp. 892-903.
Hofstede, G., (2001): “Culture’s Consequences: Comparing Values, Behaviors,
Institutions, and Organizations across Nations”, Sage Publication, Beverly Hills.
Jansen, W. J., and Nahuis, N. J., (2003): “The stock market and consumer confidence:
European evidence”, Economic Letters, 79, pp. 89-98.
Jiambalvo, J., Rajgopal, R. and Venkatachalam, M., (2002): “Institutional ownership
and the extent to whick stock prices reflect future earnings”, Contemporary Accounting
Research, 19, pp. 117-145.
Jones, C. M., (2002): “A century of stock market liquidity and trading costs”, Available
at SSRN: http://ssrn.com/abstract=313681.
Kamstra, M. J., Kramer, L. A., and Levi, M. D., (2003): “Winter blues: a SAD stock
market cycle”, American Economic Review, 93, pp. 1257-1263.
Kavussanos, M.G., Visvikis, I.D., y Alexakis, P.D., (2008): “The lead-lag relationship
between cash and stock index futures in a new market”, European Financial
Management, 14, pp. 1007-1025.
Kothari, S. P., and Shanken, J., (1997): “Book-to-market, dividend yield, and expected
market returns: A time-series analysis”, Journal of Financial Economics, 44, pp. 169-
203.
23
Koutmos, G., and Tucker, M., (1996): “Temporal relationships and dynamic
interactions between spot and futures stock markets”, Journal of Futures Markets, 16,
pp. 55-69.
Kumar, A., and Lee, C., (2006): “Retail investor sentiment and return comovement”,
Journal of Finance, 61, pp. 2451-2486.
Kumar, A., (2009): “Hard-to-value stocks, behavioural biases, and informed trading”,
Journal of Financial and Quantitative Analysis, 44, pp. 1375-1401.
La Porta, F., López-de-Silanes, F., Shleifer, A., and Vishny, R.W., (1998): “Law and
Finance”, Journal of Political Economy, 106, pp. 1113-1155.
Lee, W.Y., Jiang, C.X., and Indro, D.C., (2002): “Spot market volatility, excess returns
and the role of investor sentiment”, Journal of Banking and Finance, 26, pp. 2277-
2299.
Lee, C., Shleifer, A., and Thaler, R. H., (1991): “Investor sentiment and the closed-end
fund puzzle”, Journal of Finance, 46, pp. 75-109.
Lemmon, M., and Ni, S. X., (2011): “The effects of investor sentiment on speculative
trading and prices of stock and index options”, AFA 2011 Denver Meetings Paper.
Available at SSRN: http://ssrn.com/abstract=1572427.
Lemmon, M., and Portniaguina, E., (2006): “Consumer confidence and asset prices:
some empirical evidence”, Review of Financial Studies, 19, pp.1499-1529.
Lewellen, J., (2011): “Institutional investors and the limits of arbitrage”, Journal of
Financial Economics, 102, pp. 62-80.
Ljungqvist, A., Nanda, V., and Singh, R., (2006): “Hot markets, investor sentiment, and
IPO pricing”, Journal of Business, 79, pp. 1667-1703.
Llorente, G. Michaely, R., Saar, G. and Wang, J., (2002): “Dynamic volume-return
relation of individual stocks”, Review of Finance Studies, 15, pp. 1005-1047.
Neal, R., and Wheatley, S. M., (1998): “Do measures of investor sentiment predict
returns?”, Journal of Financial and Quantitative Analysis, 33, pp. 523-547.
Ritter, J., (2003): “Investment banking and securities issuance”. In Handbook of the
Economics of Finance, ed. Constantinides, George, Milton Harris, and Rene Stulz,
chap. 5. Elsevier Science.
Saha, S. and Chakrabati, G., (2011): “Financial crisis and financial market volatility
spill-over” International Journal of Applied Economics and Finance, 5, pp.185-199.
Schmeling, M., (2009): “Investor sentiment and stock returns: some international
evidence”, Journal of Empirical Finance, 16, pp. 394-408.
24
Schmitz, P., Glaser, M., and Weber, M., (2009): “Individual investor sentiment and
stock returns – What do we learn from warrant traders?”, Available at SSRN:
http://ssrn.com/abstract=923526.
Scheinkman, J., and Xiong, W., (2003): “Overconfidence and speculative bubbles”,
Journal of Political Economy, 111, pp. 1183-1219.
Seyhun, H. N., (1998): “Investment Intelligence from Insider Trading”, Cambridge,
MA: MIT Press.
Shiller, R. J., (1981): “Do stock prices move too much to be justified by subsequent
changes in dividends?”, American Economic Review, 71, pp. 421-436.
Shiller, R. J., (1984): “Stock prices and social dynamics” Brookings Papers on
Economic Activity, pp. 457-498.
Shiller, R. J., (2000): “Irrational Exuberance”, Princeton UP, Princeton.
Shleifer, A. and Summers, L.H., (1990): “The noise trader approach to finance” Journal
of Economic Perspectives, 4, pp. 19-33.
Shleifer, A. and Vishny R. (2003) Stock market driven acquisitions, Journal of
Financial Economics, 70, pp. 295-311.
Stambaugh, R.F., Yu, J., and Yuan, Y., (2012): “The short of it: investor sentiment and
anomalies”, Journal of Financial Economics, 104, pp. 288-302.
Stoll, H. R. and Whaley, R.E., (1990): “The dynamics of stock index and stock index
futures returns” Journal of Financial and Quantitative Analysis, 25, pp. 441-468.
Swaminathan, B., (1996): “Time-varying expected small firm returns and closed-end
fund discounts”, Review of Financial Studies, 9, pp. 845-887.
Wang, C., (2001): “Investor sentiment and return predictability in agricultural futures
markets”, Journal of Futures Markets, 21, pp. 929-952.
Whaley, R. E., (2000): “The investor fear gauge”, Journal of Portfolio Management, 26,
pp. 12-17.
Yu, J., and Yuan, Y., (2011): “Investor sentiment and the mean-variance relation”,
Journal of Financial Economics, 100, pp. 367-381.
Zweig, M. E., (1973): “An investor expectations stock price predictive model using
closed-end fund premiums”, Journal of Finance, 28, 1, pp. 67-87.
25
Table I. Effect of sentiment on abnormal volume in the spot and futures markets. 2001-2011
Panel A: Spot Market
CAC40 DAX30 EUROSTOXX50 IBEX35 FTSE100 S&P500
α 0.097*** 0.056*** 0.066*** 0.064*** 0.057*** 0.055***
β -0.012 0.017 -0.001 0.002 -0.003 -0.004
γ1 -0.286*** -0.279*** -0.236*** -0.213*** -0.239*** -0.142***
γ2 -0.012 -0.026 -0.015 -0.019 0.027** -0.007
γ3 -0.009 -0.014 -0.011 -0.036** -0.025** -0.026***
γ4 -0.050*** 0.020 -0.020 -0.022 -0.070*** -0.078***
α0 0.026*** 0.007*** 0.036*** 0.005*** 0.015*** 0.010***
α1 0.249*** 0.196*** 0.172*** 0.048*** 0.204*** 0.256***
α2 0.354*** 0.749*** 0.123 0.869*** 0.477*** 0.380***
Panel B: Futures Market
CAC40 DAX30 EUROSTOXX50 IBEX35 FTSE100 S&P500
α 0.155*** 0.099*** 0.133*** 0.098*** 0.026 0.044***
β -0.031* -0.019* -0.029** -0.017 -0.029*** -0.025**
γ1 -0.178*** -0.203*** -0.271*** -0.099*** -0.141*** -0.161***
γ2 0.079** 0.019 0.041 0.105*** 0.056** 0.066***
γ3 -0.033 -0.012 -0.043 -0.133*** -0.007 -0.053***
γ4 -0.366*** -0.093*** -0.229*** -0.349*** -0.099*** -0.087***
α0 0.074*** 0.098*** 0.042*** 0.061* 0.017*** 0.003***
α1 0.084*** 0.423*** 0.062*** 0.042** 0.358*** 0.056***
α2 0.522*** 0.081 0.600*** 0.517** 0.601*** 0.914***
The sentiment effect (coefficient β) on abnormal trading volume in the spot market (Panel A) and the futures market (Panel B). AV is the abnormal volume of index i and market m (spot or futures). SENT is the dummy variable that
takes a value of 1 if sentiment is above the median level and 0 otherwise. DM, DT, DTh y DF are dummy variables that
take a value of 1 on Mondays, Tuesdays, Thursdays and Fridays, respectively. The estimation includes an AR(5) process.***, ** and * indicate 1%, 5%, and 10% levels of significance, respectively.
26
Table II. Impact of investor sentiment on the correlation between spot market and futures market (Model 1).
2001-2011
CAC40 DAX30 EUROSTOXX50 FTSE100 IBEX35 S&P500
A0 -0.227 0.018 0.720** -0.129 0.130 1.459***
A1 -0.183*** -0.164*** -0.214*** -0.220*** -0.137*** -0.190***
A2 0.224*** 0.006 -0.102** 0.024 -0.077 -0.165***
A3 0.999*** 0.999*** 0.991*** 0.994*** 0.999*** 0.987***
B0 -0.506* -0.420** -0.119 -0.848** -0.126 -0.551
B1 -0.186*** -0.151*** -0.231*** -0.207*** -0.134*** -0.179***
B2 0.468*** 0.366*** 0.019 0.155*** 0.150** 0.061
α0 0.125*** 0.026*** 0.049*** 0.029*** 0.067*** 0.015***
α1 0.019*** 0.006 0.012* 0.016** 0.037*** 0.002
α2 0.782*** 0.879*** 0.839*** 0.891*** 0.831*** 0.898***
α3 0.071** 0.122*** 0.156*** 0.098*** 0.144*** 0.135***
α4 1.297*** 0.679*** 0.550*** 0.398*** 0.807*** 0.262***
α5 0.134*** 0.654*** 0.211** -0.060 -0.055 0.332***
β0 0.131*** 0.027*** 0.050*** 0.027*** 0.071*** 0.016***
β1 0.028*** 0.002 0.017** 0.019*** 0.038*** -0.006
β2 0.772*** 0.879*** 0.828*** 0.890*** 0.823*** 0.907***
β3 0.193*** 0.118*** 0.106*** 0.002 -0.084*** 0.144***
β4 1.624*** 0.945*** 0.960*** 0.326*** 1.107*** 0.335***
β5 -0.137 0.010 0.044* 0.091*** 0.238*** -0.116**
γ0 0.989*** 0.983*** 0.969*** 0.980*** 0.990*** 0.981***
γ1 -0.003*** -0.009*** -0.004*** -0.004*** -0.004*** -0.007***
Model 1
where is the error correction term imposing the long-term equilibrium on index i in the two
markets; ( ) is the innovation in the spot (futures) market at day t for index i; =var( /Ωt-1,i) is the
conditional variance of the spot market and =var( /Ωt-1,i) is the conditional variance of the futures market,
where Ωt,i is the information set available at t for index i. The innovation ( ) is the information from the spot
(futures) market which is transmitted to the futures (spot) market and is not included in ( ). The dummy
variable SENT has a value of 1 if sentiment is above the median level and 0 otherwise. We use the Sentix 6 month-
ESX 50 Index as the sentiment proxy for the European indices and AAII for the US index. The dummy
variable ( ) is equal to 1 if ( ) <0. The dummy variable ( ) is equal to 1 if ( ) <0.
***, ** and *indicate 1%, 5%, and 10% levels of significance, respectively.
27
Table III. Effect of investor sentiment on spot (futures) volatility (Model 2). 2001-2011
CAC40 DAX30 EUROSTOXX50 FTSE100 IBEX35 S&P500
A0 -0.222 0.043 0.771** -0.078 0.105 1.239**
A1 -0.184*** -0.163*** -0.212*** -0.224*** -0.140*** -0.191***
A2 0.219*** -0.010 -0.106*** 0.017 -0.045 -0.141***
A3 0.999*** 0.998*** 0.991*** 0.994*** 0.999*** 0.988***
B0 -0.503* -0.393** -0.081 -0.754** -0.167 -0.731*
B1 -0.187*** -0.150*** -0.230*** -0.210*** -0.139*** -0.179***
B2 0.462*** 0.348*** 0.014 0.144*** 0.177** 0.084*
α0 0.133*** 0.030*** 0.055*** 0.034*** 0.062*** 0.013***
α1 0.033*** 0.016*** 0.015** 0.025*** 0.059*** 0.001
α2 0.772*** 0.872*** 0.839*** 0.880*** 0.836*** 0.900***
α3 0.065** 0.119*** 0.158*** 0.010*** 0.125*** 0.135***
α4 1.344*** 0.738*** 0.536*** 0.459*** 0.780*** 0.281***
α5 0.135*** 0.689*** 0.224** 0.031 -0.032 0.317***
α6 -0.016** -0.016*** -0.021*** -0.017*** -0.027*** 0.011
β0 0.138*** 0.031*** 0.054*** 0.033*** 0.064*** 0.015***
β1 0.041*** 0.012* 0.022*** 0.030*** 0.059*** -0.004
β2 0.762*** 0.872*** 0.827*** 0.878*** 0.826*** 0.908***
β3 0.189*** 0.116*** 0.112*** 0.005 -0.091*** 0.134***
β4 1.712*** 1.017*** 0.952*** 0.426*** 1.128*** 0.362***
β5 -0.175 0.009 0.038 0.087*** 0.228*** -0.145**
β6 -0.014* -0.018*** -0.019*** -0.021*** -0.025*** -0.006
γ0 0.990*** 0.983*** 0.969*** 0.981*** 0.989*** 0.981***
γ1 -0.003*** -0.010*** -0.006*** -0.005*** -0.003*** -0.007***
Model 2: = = = 0 and = = = 0
where is the error correction term imposing the long-term equilibrium on index i in the two
markets; ( ) is the innovation in the spot (futures) market at day t for index i; =var( /Ωt-1,i) is the
conditional variance of the spot market and =var( /Ωt-1,i) is the conditional variance of the futures market,
where Ωt-1,i is the information set available at t-1 for index i. The innovation ( ) is the information from the
spot (futures) market which is transmitted to the futures (spot) market and is not included in ( ). The dummy
variable SENT has a value of 1 if sentiment is above the median level and 0 otherwise. We use the Sentix 6 month-
ESX 50 Index as the sentiment proxy for the European indices and AAII for the US index. The dummy
variable ( ) is equal to 1 if ( ) <0. The dummy variable ( ) is equal to 1 if ( ) <0.
***, ** and *indicate 1%, 5%, and 10% levels of significance, respectively.
28
Table IV. Effect of investor sentiment on asymmetries in spot (futures) volatility (Model 3). 2001-2011
CAC40 DAX30 EUROSTOXX50 FTSE100 IBEX35 S&P500
A0 -0.198 0.102 0.757** -0.110 0.166 1.375***
A1 -0.183*** -0.162*** -0.211*** -0.221*** -0.140*** -0.191***
A2 0.189** -0.050 -0.103** 0.023 -0.088 -0.153***
A3 0.999*** 0.998*** 0.991*** 0.994*** 0.999*** 0.987***
B0 -0.490* -0.370** -0.099 -0.815** -0.116 -0.647
B1 -0.187*** -0.148*** -0.229*** -0.208*** -0.139*** -0.180***
B2 0.428*** 0.321*** 0.016 0.152*** 0.132* 0.071
α0 0.142*** 0.030*** 0.055*** 0.034*** 0.071*** 0.014***
α1 0.025*** 0.010* 0.015** 0.017*** 0.043*** 0.005
α2 0.767*** 0.872*** 0.829*** 0.876*** 0.824*** 0.889***
α3 0.085*** 0.133*** 0.168*** 0.120*** 0.185*** 0.132***
α4 1.412*** 0.746*** 0.605*** 0.452*** 0.917*** 0.298***
α5 0.135*** 0.712*** 0.230** 0.010 -0.037 0.226**
α8 -0.047*** -0.040*** -0.023** -0.044*** -0.097*** -0.036*
β0 0.147*** 0.031*** 0.057*** 0.033*** 0.073*** 0.016***
β1 0.035*** 0.004 0.020** 0.021*** 0.044*** -0.005
β2 0.756*** 0.871*** 0.816*** 0.875*** 0.815*** 0.908***
β3 0.208*** 0.135*** 0.124*** 0.032* -0.037 0.141***
β4 1.794*** 1.041*** 1.064*** 0.400*** 1.283*** 0.240***
β5 -0.173 0.007 0.040 0.085*** 0.231*** 0.018
β8 -0.043*** -0.044*** -0.027* -0.050*** -0.090*** -0.045**
γ0 0.990*** 0.984*** 0.969*** 0.980*** 0.990*** 0.981***
γ1 -0.003*** -0.010*** -0.005*** -0.006*** -0.004*** -0.007***
Model 3: = = = 0 and = = = 0
where is the error correction term imposing the long-term equilibrium on index i in the two markets;
( ) is the innovation in the spot (futures) market at day t for index i; =var( /Ωt-1,i) is the conditional
variance of the spot market and =var( /Ωt-1,i) is the conditional variance of the futures market, where Ωt-1,i is the
information set available at t-1 for index i. is the conditional covariance between spot and futures markets. The
innovation ( ) is the information from the spot (futures) market which is transmitted to the futures (spot) market
and is not included in ( ). The dummy variable SENT has a value of 1 if sentiment is above the median level and
0 otherwise. We use the Sentix 6 month-ESX 50 Index as the sentiment proxy for the European indices and AAII for the
US index. The dummy variable ( ) is equal to 1 if ( ) <0. The dummy variable ( ) is equal to 1 if
( ) <0. ***, ** and *indicate 1%, 5%, and 10% levels of significance, respectively.
29
Table V. Effect of investor sentiment on volatility spillovers (Model 4). 2001-2011
CAC40 DAX30 EUROSTOXX50 FTSE100 IBEX35 S&P500
A0 -0.211 0.067 0.799** -0.127 0.108 1.612***
A1 -0.184*** -0.162*** -0.212*** -0.219*** -0.137*** -0.195***
A2 0.213*** 0.029 -0.107*** 0.024 -0.051 -0.180***
A3 0.999*** 0.998*** 0.991*** 0.993*** 0.999*** 0.987***
B0 -0.486 -0.412** -0.069 -0.891** -0.171 -0.451
B1 -0.188*** -0.148*** -0.229*** -0.206*** -0.136*** -0.183***
B2 0.454*** 0.336*** 0.012 0.157*** 0.173** 0.051
α0 0.128*** 0.031*** 0.054*** 0.027*** 0.062*** 0.020***
α1 0.019*** 0.008 0.014** 0.011** 0.038*** 0.008
α2 0.779*** 0.868*** 0.830*** 0.895*** 0.832*** 0.889***
α3 0.075*** 0.121*** 0.156*** 0.097*** 0.143*** 0.168***
α4 1.521*** 1.212*** 0.952*** 0.552*** 1.221*** 0.307***
α5 0.130*** 0.792*** 0.197** -0.031 -0.034 0.367***
α7 -0.339* -0.718*** -0.579*** -0.319*** -0.619*** -0.094***
β0 0.133*** 0.032*** 0.057*** 0.027*** 0.066*** 0.020***
β1 0.028*** 0.001 0.019** 0.016*** 0.041*** -0.005
β2 0.769*** 0.869*** 0.816*** 0.892*** 0.822*** 0.902***
β3 0.194*** 0.140*** 0.122*** -0.007 -0.087*** 0.174***
β4 1.809*** 1.478*** 1.570*** 0.550*** 1.616*** 0.390***
β5 -0.091 -0.010 0.027 0.097*** 0.241*** -0.164**
β7 -0.291* -0.662*** -0.889*** -0.408*** -0.694*** -0.074***
γ0 0.990*** 0.984*** 0.972*** 0.980*** 0.989*** 0.983***
γ1 -0.003*** -0.012*** -0.012*** -0.007*** -0.003*** -0.009***
Model 4: = = = 0 and = = = 0
where is the error correction term imposing the long-term equilibrium on index i in the two
markets; ( ) is the innovation in the spot (futures) market at day t for index i; =var( /Ωt-1,i) is the
conditional variance of the spot market and =var( /Ωt-1,i) is the conditional variance of the futures market,
where Ωt-1,i is the information set available at t-1 for index i. is the conditional covariance between spot and
futures markets. The innovation ( ) is the information from the spot (futures) market which is transmitted to the
futures (spot) market and is not included in ( ). The dummy variable SENT has a value of 1 if sentiment is
above the median level and 0 otherwise. We use the Sentix 6 month-ESX 50 Index as the sentiment proxy for the
European indices and AAII for the US index. The dummy variable ( ) is equal to 1 if ( ) <0. The
dummy variable ( ) is equal to 1 if ( ) <0. ***, ** and *indicate 1%, 5%, and 10% levels of
significance, respectively.
30
Table VI. Effect of investor sentiment on asymmetries in spot (futures) volatility spillovers (Model 5). 2001-
2011
CAC40 DAX30 EUROSTOXX50 FTSE100 IBEX35 S&P500
A0 -0.187 0.013 0.730** -0.191 0.084 1.228**
A1 -0.186*** -0.164*** -0.214*** -0.222*** -0.139*** -0.191***
A2 0.196*** 0.011 -0.103** 0.036 -0.030 -0.139**
A3 0.999*** 0.998*** 0.991*** 0.994*** 0.999*** 0.987***
B0 -0.450 -0.420** -0.111 -0.912** -0.184 -0.770*
B1 -0.189*** -0.151*** -0.231*** -0.208*** -0.138*** -0.180***
B2 0.431*** 0.374*** 0.018 0.170*** 0.193*** 0.086*
α0 0.115*** 0.026*** 0.052*** 0.023*** 0.057*** 0.014***
α1 0.013** 0.007 0.013** 0.009 0.037*** 0.005
α2 0.796*** 0.879*** 0.833*** 0.908*** 0.841*** 0.897***
α3 0.082*** 0.121*** 0.159*** 0.095*** 0.138*** 0.132***
α4 1.135*** 0.688*** 0.585*** 0.297*** 0.740*** 0.276***
α5 0.120*** 0.676*** 0.240** -0.046 0.000 0.360***
α9 -0.007 -0.002 -0.034 0.054 -0.075 -0.087
β0 0.118*** 0.027*** 0.053*** 0.023*** 0.059*** 0.015***
β1 0.022*** 0.002 0.018** 0.012** 0.040*** -0.005
β2 0.789*** 0.879*** 0.822*** 0.906*** 0.831*** 0.908***
β3 0.190*** 0.119*** 0.110*** -0.002 -0.075*** 0.141***
β4 1.285*** 0.968*** 1.030*** 0.250*** 1.079*** 0.326***
β5 0.079 0.009 0.045* 0.093*** 0.220*** -0.121*
β9 0.012 -0.005 -0.006 -0.009** 0.006 0.027
γ0 0.990*** 0.983*** 0.969*** 0.979*** 0.989*** 0.982***
γ1 -0.004*** -0.009*** -0.004*** -0.005*** -0.002*** -0.007***
Model 5: = = = 0 and = = = 0
where is the error correction term imposing the long-term equilibrium on index i in the two
markets; ( ) is the innovation in the spot (futures) market at day t for index i; =var( /Ωt-1,i) is the
conditional variance of the spot market and =var( /Ωt-1,i) is the conditional variance of the futures market,
where Ωt-1,i is the information set available at t-1 for index i. is the conditional covariance between spot and
futures markets. The innovation ( ) is the information from the spot (futures) market which is transmitted to
the futures (spot) market and is not included in ( ). The dummy variable SENT has a value of 1 if sentiment
is above the median level and 0 otherwise. We use the Sentix 6 month-ESX 50 Index as the sentiment proxy for the
European indices and AAII for the US index. The dummy variable ( ) is equal to 1 if ( ) <0. The
dummy variable ( ) is equal to 1 if ( ) <0. ***, ** and *indicate 1%, 5%, and 10% levels of
significance, respectively.
31
Table VII. Effect of extremely bullish sentiment on correlation between markets and on volatility
spillovers, Six-month Sentix index and AAII. 2001-2011
CAC40 DAX30 EUROSTOXX50 FTSE100 IBEX35 S&P500
Coeff. Model 2
α6 -0.085*** -0.038*** -0.045*** -0.049*** -0.036*** -0.012**
β6 -0.095*** -0.040*** -0.047*** -0.058*** -0.031*** -0.011**
γ1 -0.003*** -0.013*** -0.013*** -0.011*** -0.001** -0.004***
Model 3
α8 0.028 -0.066*** -0.036*** -0.058*** -0.033** -0.039***
β8 0.029 -0.074*** -0.046*** -0.073*** -0.026* -0.029***
γ1 -0.002*** -0.012*** -0.009*** -0.007*** -0.001 -0.004***
Model 4
α7 0.014 -0.606*** -0.528*** -0.342*** -0.653*** -0.144
β7 -0.055 -0.636*** -0.920*** -0.376*** -0.868*** -0.140
γ1 -0.002*** -0.015*** -0.017*** -0.008*** -0.001** -0.004***
Model 5
α9 -0.003 0.021 -0.041 -0.169* -0.025 -0.027
β9 -0.179 -0.0111 -0.016 -0.025*** 0.008 0.002
γ1 -0.002*** -0.010*** -0.008*** -0.007*** 0.000 -0.003***
Unrestricted Model:
Model 2: = = = 0 and = = = 0
Model 3: = = = 0 and = = = 0
Model 4: = = = 0 and = = = 0
Model 5: = = = 0 and = = = 0
where is the error correction term imposing the long-term equilibrium on index i in the two
markets; ( ) is the innovation in the spot (futures) market at day t for index i; =var( /Ωt-1,i) is the
conditional variance of the spot market and =var( /Ωt-1,i) is the conditional variance of the futures market,
where Ωt-1,i is the information set available at t-1 for index i. is the conditional covariance between spot and
futures markets. The innovation ( ) is the information from the spot (futures) market which is transmitted to the
futures (spot) market and is not included in ( ). The dummy variable SENT has a value of 1 for sentiment
scores within the top 25% and 0 otherwise. We use the Sentix 6 month-ESX 50 Index as the sentiment proxy for the
European indices and AAII for the US index. The dummy variable ( ) is equal to 1 if ( ) <0. The
dummy variable ( ) is equal to 1 if ( ) <0. ***, ** and *indicate 1%, 5%, and 10% levels of
significance, respectively.
32
Table VIII. Effect of sentiment on correlation between markets and on volatility spillovers, One-month Sentix
index. 2001-2011
CAC40 DAX30 EUROSTOXX50 FTSE100 IBEX35
Coeff. Model 2
α6 0.051*** -0.007 -0.039*** -0.009* -0.010**
β6 0.054*** -0.018*** -0.054*** -0.011** -0.007*
γ1 -0.005*** -0.007*** -0.013*** -0.012*** -0.003***
Model 3
α8 -0.009 -0.005 -0.076*** 0.011 0.010
β8 -0.005 -0.015 -0.093*** 0.009 0.006
γ1 -0.005*** -0.007*** -0.012*** -0.011*** -0.004***
Model 4
α7 -0.005 -0.647*** -0.090 -0.362*** -1.268***
β7 -0.066 -0.758*** -0.273** -0.364*** -1.388***
γ1 -0.005*** -0.009*** -0.012*** -0.013*** -0.007***
Model 5
α9 -0.006 0.303** 0.084 -0.256*** -0.218**
β9 -0.314*** -0.004 -0.018 -0.011** -0.007
γ1 -0.005*** -0.007*** -0.010*** -0.011*** -0.005***
Unrestricted Model:
Model 2: = = = 0 and = = = 0
Model 3: = = = 0 and = = = 0
Model 4: = = = 0 and = = = 0
Model 5: = = = 0 and = = = 0
where is the error correction term imposing the long-term equilibrium on index i in the two
markets; ( ) is the innovation in the spot (futures) market at day t for index i; =var( /Ωt-1,i) is the
conditional variance of the spot market and =var( /Ωt-1,i) is the conditional variance of the futures
market, where Ωt-1,i is the information set available at t-1 for index i. is the conditional covariance between
spot and futures markets. The innovation ( ) is the information from the spot (futures) market which is
transmitted to the futures (spot) market and is not included in ( ). The dummy variable SENT has a
value of 1 if sentiment is above the median level and 0 otherwise. We use the Sentix 1 month-ESX 50 Index as
the sentiment proxy for the European indices. The dummy variable ( ) is equal to 1 if ( ) <0.
The dummy variable ( ) is equal to 1 if ( ) <0. ***, ** and *indicate 1%, 5%, and 10% levels of
significance, respectively.