investor sentiment and the cross-section of japanese · pdf fileinvestor sentiment and the...
Post on 18-Feb-2018
219 Views
Preview:
TRANSCRIPT
1
Investor sentiment and the cross-section of Japanese stock returns
by
Joyce Khuu*
Robert B. Durand
Lee A. Smales
This paper examines sentiment as an augmentation to the Fama and French three-
factor model. International tests of empirical asset pricing models show that only three
pricing factors are relevant, but perhaps not sufficient, in modelling the cross-section of
Japanese stock returns. We find that that sentiment can help explain the cross-section of
Japanese stock returns and is able to remove excess returns when added to the three-factor
model. We also observe asymmetric effects of sentiment on stocks in the cross-section of
size. Small stocks and large stocks are more affected by sentiment. This paper demonstrates
that the addition of a factor capturing sentiment should be considered when modelling
Japanese stock returns.
* Corresponding author. E-mail: Joyce.Khuu@curtin.edu.au
The authors are from the School of Economics and Finance, Curtin University, Bentley,
Western Australia, Australia. We are grateful to our colleagues for their helpful input. All
errors are our own.
2
Preface
Thesis title: Japan’s Lost Decade and The Role of Sentiment
Supervisors: Prof. Robert Durand, Dr. Lee Smales
Sentiment is often referred to as the overall market mood, and sentiment as the market mood
does not necessarily correspond to fundamental or economic information. A growing body
of literature suggests sentiment impacts stock market behaviour and can predict stock returns.
These findings are contradictory to the traditional efficient market hypothesis approach in
finance and are a behavioural explanation of asset pricing anomalies. Traditional models of
asset pricing have so far been unable to explain the puzzles that are the lost decades of the
Japanese equity market. The dissertation aims to explore whether market sentiment can be
used to help explain stock returns in the Japanese equity market during the latter part of this
period, January 2003 to October 2014. This time period encompasses the “second lost decade
of Japan” which describes part of 20 years of stagnation for the Japanese economy and equity
market. This research will encompass the first section of the dissertation. Further to this
research, there is currently no consensus within the literature as to how sentiment should be
measured or captured. To fill this gap in the literature, this dissertation also explores the
potential measurement issues, relationships and sensitivities of the sentiment proxies
currently used in the literature. The structure of this dissertation is as follows:
Chapter 1: Introduction
Chapter 2: Melancholia and Japanese stock returns
Chapter 3: Investor sentiment and the cross-section of Japanese stock returns
Chapter 4: Capturing sentiment the importance of measurement and proxies
Chapter 5 Conclusions
This paper has been taken from the third chapter of my dissertation and is adapted for a three
paper publication format. It examines the role that sentiment has in empirical asset pricing
within the Fama and French framework and focuses on the cross-section of Japanese stock
returns. It extends upon previous published (forthcoming) work in the dissertation which
suggested that negative market sentiment in Japan may be related to poor equity returns over
the time period examined.
3
I. Introduction
Augmentations to the three-factor model such as momentum (Carhart, 1997),
profitability and investment (Fama and French 2015a), have achieved varied success in
explaining the cross-section of US and global stock market returns. However, these
additional factors have notably, and repeatedly, “failed” in the Japanese context (Cakici
2015; Fama and French 2015b). Unlike other developed markets international tests of
empirical asset pricing models reveal that only three pricing factors are relevant in modelling
the cross-section of Japanese stock returns (Fama and French 2012; 2015b). These three
factors are, the market premium (Rm - Rf), value premium (HML) and size premium (SMB)
which are based on seminal US studies (Fama and French 1993; 1996). However, the three
factors may not be sufficient to fully explain the cross-section of returns as unexplained
excess returns (α) seemingly persist. As standard factor augmentations have failed for Japan,
we examine whether sentiment is a priced factor and we find that sentiment is useful in
explaining stock returns. In this paper, we find that the addition of sentiment to the three
factors removes α in the majority of our sample portfolios. The effect of sentiment is greatest
for growth stocks (as expected given the extant literature) and also large stocks (which is
contrary to the literature).
A growing body of literature suggests that sentiment influences market behavior and
as a result stock prices and share markets (Baker and Wurgler 2006; Brown and Cliff 2005;
Tetlock 2007; Tetlock et al. 2008; Stambaugh et al. 2012). Sentiment may also provide a
useful addition to the Fama and French three-factor model. There are two potential reasons
for this: the first is that sentiment may have market wide effects and could influence the Fama
and French factors. Alternatively, sentiment may act as a separate additional factor.
Khuu et al. (2016) found that news sentiment can help explain the prolonged negative
average stock returns in Japan – a phenomenon which challenges the positive relationship
between risk and expected returns. They find a positive relationship between news sentiment
and stock returns, where on average the market has negative, (or near negative) sentiment
which they link to poor market returns in aggregate. They also document a relation between
sentiment and firm size that is common in the sentiment literature. Smaller stocks seem to
4
be more susceptible to “sentiment” with “limits-to-arbitrage” presenting one explanation
(Baker and Wurgler 2006; 2007). 1 Size appears to be an important characteristic when
examining the effects of sentiment, and is explicitly priced in the Fama and French empirical
framework through SMB. This common variable in size suggests that sentiment might be
associated with SMB for Japan. Khuu et al. (2016) suggests that prolonged periods of
negative sentiment can help explain poor stock market returns in Japan. As Rm-Rf represents
the market premium, this may also be influenced by sentiment.
The Fama and French (1993) three-factor model provides an empirically-based
explanation for cross-sectional patterns in stock returns that were not captured by the single
factor CAPM model of Sharpe (1964) and Lintner (1965). In addition to the market risk
premium, the three-factor model includes two other factors, SMB and HML. SMB captures a
size premium where stocks with lower market capitalization earn higher returns, over stocks
with higher market capitalization. HML captures a value premium, where higher returns are
related to stocks with high book values of assets to market values than stocks which have
low book values to market values. The excess returns equation of this model is as follows:
, , , , ,( )p t f t p p m t f t p t p t p tR R R R SMB HML (1.1)
where Rp,t is the return of the portfolio; Rf,t is the return of a risk free asset; Rm,t is the
return of a market portfolio; HMLt is the difference between a portfolio of high book to
market and low book to market stocks; SMBt is the return of a portfolio of small minus big
stocks; εp,t is the error term. αp represents the intercept or abnormal return of the expected
return, which is equal to zero if the factors capture all the variation in expected returns. In
this model the factor loadings represent a risk premia associated with sensitivity to HML and
SMB. As Japanese stock returns are highly correlated to book to market (B/M) (Chan et al.,
1991), we would expect this to be captured by HML. Though this model is often augmented
by a momentum factor (Carhart 1997), we do not employ it here given that momentum effects
1 Berger and Turtle (2012) find that “sentiment prone” stocks are young, volatile, small firms with “opaque” characteristics. Brown and Cliff (2005),
Lemmon and Portniaguina (2006), and Schmeling (2009) also note that sentiment had a greater influence on small firms, although there is conflicting evidence
as to whether the effect is greatest for stocks categorized as value or growth.
5
are commonly regarded as absent in Japan (Fama and French 2012). Recent evidence also
suggests that the new profitability and investment factors (Fama and French 2015) add little
to the three-factor model when applied to Japan (Cakici 2015; Fama and French 2015b)2.
Both rational and behavioral explanations have been offered for the pattern of Japanese stock
returns.
Sentiment is not directly observable, but is often associated with the market "mood" or
"feeling". While sentiment itself is not observable, its effects maybe which requires a proxy
(Chan et al. 2016). There is no commonly defined sentiment proxy and there are three
common approaches. One sentiment proxy includes Baker and Wurgler’s (2006; 2007)
macroeconomic based measure3 which captures market sentiment through the use of
macroeconomic and market variables. Papers which employ this metric include Baker and
Wurgler (2006; 2007), Tsuji (2006), Yu and Yuan (2011), Baker et al. (2012), Chung et al.
(2012) and Stambaugh et al. (2012). However, there is debate as to whether these proxies are
effective (Chen et al. 1993; Lemmon and Portniaguina 2006).
The second approach employs periodic survey based indices (Akhtar et al. 2011;
Antoniou et al. 2013; Brown and Cliff 2005; Lemmon and Portniaguina 2006). Examples
include the Conference Board Consumer Index (CBCI) and Michigan Consumer Sentiment
Index (MCSI), that poll market or household opinions on a regular basis.
The third approach, which we employ, is the use of text-based sentiment measures
which is becoming increasingly prevalent in the literature (Allen et al. 2015; Dzielinski 2011;
García 2013; Groß-Klußmann and Hautsch 2011; Smales 2014; Tetlock 2007; Tetlock et al.
2 Different explanations have been put forward for the pattern of Japanese stock returns. Daniel et al. (2001) argued that a characteristics based
model rather than a risk factor based model is more suitable for Japan. Chiao and Hueng (2005) finds evidence for overreaction in Japan (Chang et al, 1995;
Gunaratne and Yonesawa, 1997), which are independent of the characteristics and risk factor hypotheses. However, we do not focus on these explanations
here.
3 Baker and Wurgler (2006) report that the first measure of sentiment explains 49% of the sample variance of the set of candidate sentiment proxies
and that the second measure explains 51% of the cumulative variance of the orthogonalized proxies that they use. They utilize only the first principal component
and do not report the eigenvalues of their principal components analysis. Our attempts to replicate the PCA analysis finds that the second principal component
could also be considered.
6
2008; Uhl 2014). We utilize Thomson Reuters News Analytics (TRNA) as a text-based proxy
of sentiment over our sample period. One advantage of using a sentiment proxy based on
text-based news is that it captures dynamic changes in sentiment; news is released and
updated frequently, eliciting changes in sentiment and influencing investor behavior. Tetlock
(2007) finds that media pessimism predicts lower stock returns on the Dow Jones Industrial
Average (DJIA), and this suggests the existence of a psychological link between news and
market prices. García (2013) analyzes the text of a Wall Street Journal (WSJ) news column
and finds that the predictive power of news sentiment is concentrated in recessions and notes
evidence of an irrational reaction to market news on days of pessimism and optimism. Uhl
(2014) reports that a text based sentiment measure performs better in forecasting returns than
in predicting macroeconomic factors. Dzielinski (2011) compared returns on positive and
negative news days, using the TRNA dataset, and found that US stock returns have above
(below) average returns on positive (negative) days. Aman (2013) identifies a potential
relationship between active media coverage (newspaper articles) and extreme market
volatility (crashes) in Japan. He finds that investors have extreme and large reactions to
increased intensity of news coverage.
The remainder of this paper is structured as follows: section II describes the data and
methodology utilized in this paper, section III presents our results and section IV concludes.
II. Data
Our study utilizes daily data for common stocks that are listed on the Tokyo Stock
Exchange (TSE) from January 2003 to July 2014.4 We choose daily data as it is more likely
to capture the dynamic relationship of sentiment on stock prices which would otherwise be
lost by using monthly data.
4 Our sample period is limited by the availability of the TRNA data provided by SIRCA used to construct our sentiment time
series.
7
We compute our sentiment measure, Psent, using TRNA. TRNA is a contextual text-
based sentiment measure which captures the effect of news on stock markets. TRNA uses
neural network and machine learning to categorize sentiment associated with news stories as
“positive” (1), “negative” (-1) and “neutral” (0).5 Each news item is accompanied by a GMT
date and time stamp as well as a Reuters Instrument Code (RIC) that identifies the stock that
the news item is related to. We aggregate all daily news items for Japan during trading hours.
Any news articles that are released after close of trading are assigned to the following trading
day since that is when the news will be able to impact prices and returns. Prior to constructing
our sentiment measure, we filter the news items in the TRNA data set using several
information fields:
1. Sentiment probability: The measure of sentiment for a news article is
categorized as positive (1), neutral (0), or negative (-1). TRNA also assigns a probability that
the news item is correctly signed as positive, neutral or negative. For example, if there is an
80% probability that a news item is positive, the news item would be signed as positive (+1),
with 80% probability. From this we can construct a probability weighted sentiment score as
+0.8 (i.e. +1 x 80%).6 We utilize probability weighted scores in this study.
2. Relevance: A rating between 0 and 1 that indicates how relevant the news
item is to a specific firm. A score of 1 (0) means the news item is highly relevant (irrelevant).
We filter for news articles with a relevance score above 0.8 to ensure that the sentiment
measure we construct is relevant7 to stock prices and returns (Groß-Klußmann and Hautsch
2011; and Smales 2014) whilst filtering out noise. This filter does not necessarily mean that
5 Studies that have utilized this data set include Hendershott et al. (2015) and Smales (2014).
6 The TRNA sentiment scores provide measures of positivity (+1) and negativity (-1) of any news signal, as well as the magnitude (probability).
TRNA analysis provides an analysis of the sentiment likely opined from the perspective of the author of the News item for consistency, not how the market
perceives the news item. 7 Not all news items referring to a firm may be directly relevant to it. For example, a discussion about Firm A may also mention Firm B in passing.
TRNA provides information on relevancy to ensure that the sentiment being distilled from a news article is not mistakenly associated with firms which are not
necessarily the focus of the article. Relevant articles do not necessarily translate to fundamental information, as the relevancy field does not provide information
on the content of news articles.
8
the news contains fundamental information, as this field does not distinguish between the
content or topic of the news articles.
3. Novelty: This measures how unique a particular news item is when compared
to previous similar news items within a defined period. We filter for content that is considered
“novel”, i.e. news items that are not similar to previous articles or “stale news”.
We construct our sentiment proxy using the sentiment classification (positive,
negative or neutral) attached to a news item and multiply this by the TRNA assigned
probability that the classification is correct. This provides a probability weighted sentiment
score Psent:
(1) ( ) ( 1) ( )
1;1positive negative
positive negative netural
P sentiment P sentimentPsent
nsentiment nsentiment nsentiment
(1.2)
where Psent is the sentiment of the market; P is the probability of classification; and
nsentiment is the number of sentiment news items with corresponding positive, negative or
neutral scores. As neutral news items have (0), or zero sentiment classification they do not
affect the numerator the equation. However, they weigh the denominator of this measure
towards neutral sentiment as the number of neutral news items increases.
Stock market and accounting data are taken from Thomson Reuters Datastream and
Bloomberg. The risk free rate Rf used in this study is the 30 day Gensaki repo rate which is
one of the most liquid proxies for the Japanese risk free rate and is commonly used in the
literature (Daniel et al., 2001) The market return Rm is the average return of the TOPIX. We
exclude stocks which do not have 24 months of returns before portfolio formation dates, as
well as stocks with negative book equity. Unlike firms in the United States, firms in Japan
tend to have fiscal years ending March 31. As a result we follow Daniel et al. (2001) and
Chiao and Hueng (2005), in the timing of all our portfolio formations, rather than the
traditional June to December formation periods. Return portfolios are formed on the first
trading day of October in year t and held until the last trading day of September t + 1. For
return portfolios, book equity (BE) of a firm is that which runs April t - 1 to March year t.
Book to market, is BE divided by market equity (ME) on the last trading day of March year
t and Size, is taken as the market equity of a firm on the last trading day of September year
9
t. The 6-month lag between portfolio formation and fiscal year end is commonly used to
ensure that accounting information is publically available and has been disseminated.
To construct our Japanese specific Fama and French Factors, we follow Fama and
French (1993) to construct size (SMB) and book to market (HML) factors. We construct six
(2x3) size and book to market return portfolios from the intersection of two ME and three
B/M independent sorts. Stocks are first sorted into two portfolios by median market
capitalization at the end of March year t. We then independently sort stocks into three
portfolios by book to market using a split of 30:40:30 percentiles. We define the bottom 30th
percentile as low, the middle 40th percentile as medium and the top 30th percentile high.
These portfolios are rebalanced every year. The SMB factor is constructed as the average
return on the three small portfolios minus the average return of the three big portfolios. The
HML factor is constructed as the average return on the two high HML portfolios, minus the
average return of the two low HML portfolios. Table 1 presents the number of stocks in the
six (2x3) size and book to market return portfolios formed from the intersection of two ME
and three B/M independent sorts
Table 1 Average Number of Stocks in Each Portfolio 2x3 B/M and Size
B/M Low Med High
Small 280 434 549
Big 478 578 210
We also form twenty-five (5x5) size and book to market return portfolios from the
intersection of stocks sorted into quintiles by size and book to market. Stocks in our sample
are first sorted into market equity quintiles from small to large and then again independently
sorted into book to market quintiles from low to high. The value weighted daily returns are
calculated September year t to October t+1.8 Table 2 displays the number of stocks sorted in
to the (5x5) portfolios. Table 3 below presents the average excess returns and statistics for
8 As a check, we first download CRSP and Compustat data to replicate a subsample of Fama and French’s daily factors before making the required
adjustments to our programming code for Japanese data.
10
the 5x5 size and book to market portfolios. These results demonstrate the puzzle of Japan’s
stock market as of recent times. The average excess return for the majority of 5x5 portfolios
are close to zero, however despite this there is significantly large variation in returns.
Table 2 Average Number of Stocks in Each Portfolio 5x5 B/M and Size
B/M Low 2 3 4 High
Small 76 67 73 98 190
2 77 71 88 117 154
3 85 86 107 127 100
4 109 114 130 106 47
Big 159 168 109 58 15
This persistent relationship of positive risk yet zero return contradicts Merton’s
proposition of positive risk and positive expected return. Khuu et al. (2016) report that
sentiment has a role in explaining this phemonema.
There are two potential channels through which sentiment could influence asset
prices: The first, is by acting through or influencing the factors themselves (which are
negatively correlated and statistically significant). Alternatively, sentiment could present as
a separate additional factor in itself. Therefore, to remove the influence of market wide
sentiment we consider orthogonalizing our Japanese factors to Psent. This allows us to
separate the effect of sentiment from the factors. To obtain the orthogonalized factors we
follow Durand et al. (2016) and regress each Fama and French factor against Psent, utilizing
the residuals as orthogonalized factors in the following analysis. These factors are denoted
by orthog in Table 4. Table 4 presents summary statistics of our constructed factors. Panel A
of Table 4 presents the summary statistics for the constructed factors, and sentiment measure.
Panel B presents correlations of the factors. Panel B of Table 4 indicates that there are
correlations between the factors and sentiment. Sentiment is positively and statistically
significantly correlated with the market premium (0.2688), and negatively related to SMB (-
0.1215) and HML (-0.0837). The reported correlations may indicate that positive (negative)
sentiment is related to positive (negative) premiums.
11
Table 3 Average Daily Excess Returns for Twenty-five Portfolios Formed on Size and B/M
B/M Low 2 3 4 High
Excess Returns
Small -0.0001 0.0000 0.0001 0.0001 0.0002
2 0.0000 0.0004 0.0002 0.0003 0.0003
3 0.0000 0.0002 0.0002 0.0003 0.0004
4 0.0002 0.0003 0.0003 0.0003 0.0004
Big 0.0001 0.0002 0.0003 0.0004 0.0005
Std. Dev.
Small 0.0160 0.0125 0.0112 0.0105 0.0095
2 0.0144 0.0131 0.0106 0.0104 0.0105
3 0.0155 0.0124 0.0113 0.0114 0.0119
4 0.0139 0.0129 0.0125 0.0128 0.0142
Big 0.0134 0.0137 0.0136 0.0145 0.0161
(Continued on next page)
12
B/M Low 2 3 4 High
Min
Small -0.2041 -0.1619 -0.1631 -0.1593 -0.1504
2 -0.1842 -0.1625 -0.1360 -0.1508 -0.1459
3 -0.1573 -0.1336 -0.1379 -0.1450 -0.1206
4 -0.1260 -0.1266 -0.1216 -0.1326 -0.1302
Big -0.0912 -0.0998 -0.1173 -0.1177 -0.1075
Max
Small 0.1007 0.0925 0.1109 0.1074 0.1014
2 0.1210 0.1306 0.1101 0.1131 0.1171
3 0.1263 0.1330 0.1109 0.1023 0.1006
4 0.1191 0.1261 0.1144 0.1142 0.1024
Big 0.1070 0.1320 0.1148 0.1341 0.1178
13
Table 4 Summary Statistics for Constructed Factors
Panel A: Summary Statistics Factors Orthogonolized Factors
Psent SMB HML (Rm-Rf) SMBOrthog HMLOrthog (Rm-Rf) Orthog
N 3,000 3,000 3,000 3,000 3,000 3,000 3,000
Mean -0.0481 -0.0001 0.0003 0.0002 0.0000 0.0000 0.0000
Sd 0.1398 0.0072 0.0138 0.0042 0.0072 0.0042 0.0132
Min -0.5016 -0.0565 -0.0952 -0.0235 -0.0588 -0.0245 -0.0948
Max 0.3658 0.0435 0.1370 0.0339 0.0422 0.0327 0.1370
Panel B: Correlation Matrix
Psent
SMB -0.1215***
HML -0.0837*** -0.1403***
(Rm-Rf) 0.2688*** -0.5483*** -0.1817***
SMBOrthog 0.0000 0.9911*** -0.1527*** -0.5193***
HMLOrthog 0.0000 -0.1519*** 0.9962*** -0.1597*** -0.1532***
(Rm-Rf)Orthog 0.0000 -0.5345*** -0.1652*** 0.9630*** -0.5393*** -0.1659***
PsentOrthog 0.9623*** 0.0000 0.0000 0.0000 0.1178*** 0.0808*** -0.2686***
14
Khuu et al. (2016) link positive sentiment to positive market premiums. Periods of high
(low) sentiment has also been associated with future reversals in the size premium (Baker and
Wurgler 2006) with exuberance and over confidence (Yu and Yuan 2011) providing one
explanation. The positive correlation with the market premium suggests that sentiment has market
wide effects.
To assess if sentiment is potentially useful we examine 6 different model specifications
centered around the three-factor Fama and French model. These models are run with and without
the orthogonalized factors. The first model is the standard three-factor model (excess returns):
, , , , ,( )p t f t p p m t f t p t p t p tR R R R SMB HML (1.3)
where Rp,t is the return of the portfolio on day t; Rf,t is the return of the risk free asset on day t; Rm,t
is the market return on day t; SMB is the size factor on day t; HML is the book to market factor on
day t; εp,t is the error term for the portfolio on day t; and αp is the intercept term of the portfolio p.
The second specification includes the addition of Psent, and we run this for both un-orthogonal
and orthogonalized factors:
, , , , ,( )p t f t p p m t f t p t p t p t p tR R R R SMB HML Psent (1.4)
where Rp,t is the return of the portfolio on day t; Rf,t is the return of the risk free asset on day t; Rm,t
is the market return on day t; SMB is the size factor on day t; HML is the book to market factor on
day t; Psent is sentiment on day t; εp,t is the error term for the portfolio on day t; and αp is the
intercept term of the portfolio p.
The final specification includes dummy variables for January, March and July and day of the week
effects. 9 We run this regression for the 2x3 and 5x5 portfolio sorts.10,11 The final specification is
as follows:
9 July is the month when the majority of financial statements and accounts are finalized in Japan.
10 We run all regressions with robust standard errors.
11 We report only the 5x5 portfolio sorts for the sake of brevity.
15
, , , ,
,
( )p t f t p p m t f t p t p t p t
p t p t p t p t p t
R R R R SMB HML Psent
j Jan March l July d Day
(1.5)
where Rp,t is the return of the portfolio on day t; Rf,t is the return of the risk free asset on day t; Rm,t
is the market return on day t; SMB is the size factor on day t; HML is the book to market factor on
day t; Psent is sentiment on day t; εp,t is the error term for the portfolio on day t; and αp is the
intercept term of the portfolio p. Jan is a dummy variable for January, March is a dummy variable
for March, July is a dummy variable for July, and Day are day of the week dummies.
III. Results
Table 5 presents statistics commonly utilized to assess different model specifications
(Fama and French 2012). We examine the average absolute alpha for each model specification and
utilize a zero intercept rule as a selection criterion (Merton 1973). Panels A and B of Table 5
presents results for a Gibbons Ross Shanken test (GRS) test of finite sample.
Panel A focuses on tests for the 5x5 portfolio sorts whilst Panel B presents results for 2x3
portfolio sorts. In each case, we test the null hypothesis that the alphas for each specification of
our model are jointly equal to zero. As in Fama and French (2012), SR(α) is equal to 1/ 21
where α is the column vector of the 25 regression intercepts produced by a model when applied to
25 global or local Size and B/M portfolios, and Σ is the covariance matrix of regression residuals.
“Lower is better” for this statistic that is interpreted by Fama and French as a “Sharpe ratio for the
intercepts (unexplained average returns) of a model” (p.466).
16
Table 5 Model Performance Statistics
Panel A 5x5 Portfolio Sorts Statistics Model GRS |α| R2 S(α) SR(α) (1) Japan three - factor model 1.90*** 0.0001 0.8855 7.56 x 10-5 0.3594 (2) Japan orthogonalized three - factor model 12.96*** 0.0002 0.8311 9.54 x 10-5 0.5803 (3) Japan three - factor model with psent 1.32 0.0001 0.8856 8.03 x 10-5 0.3382 (4) Japan orthogonalized three - factor model with psent 4083.12*** 0.0012 0.8850 8.01 x 10-5 2.5144 (5) Japan three-factor model with psent and control variables 1.10 0.0001 0.8856 1.78 x 10-4 0.4823 (6) Japan three-factor model with psent and control variables with orthogonalized factors
5.25x1014*** 0.0014 0.8856 1.78 x 10-4 2251.6660
(7) Japan three-factor model with Psent orthogonal to three factors 1.91*** 0.0001 0.8856 7.56 x 10-5 0.1296 Panel B 2x3 Portfolio Sorts Statistics Model GRS |α| R2 S(α) SR(α) (8) Japan three - factor model 4.25*** 0.0001 0.9676 4.01 x 10-5 0.0926 (9) Japan orthogonalized three - factor model 4090.63*** 0.0002 0.9070 6.99 x 10-5 2.8645 (10) Japan three - factor model with psent 2.66** 0.0001 0.9677 9.14 x 10-5 0.0779 (11) Japan orthogonalized three - factor model with psent 3635.64*** 0.0013 0.9671 4.27 x 10-5 2.8561 (12) Japan three-factor model with psent and control variables 0.67 0.0001 0.9677 9.45 x 10-5 0.0869 (13) Japan three-factor model with psent and control variables with orthogonalized factors
230.41*** 0.0014 0.9677 9.44 x 10-5 1.6087
(14) Japan three-factor model with Psent orthogonal to three factors 4.26*** 0.0001 0.9677 4.01 x 10-5 0.0927
17
Based on Table 5, two models satisfy the factor criteria, that is they do not reject the null
hypothesis. These are (3), (5) and (12) which are models augmented with Psent.12 In addition, the
specification with the lowest values of SR(α) occurs in the three-factor model augmented with
Psent which suggests that the inclusion of Psent removes the excess returns left unexplained by
the other three factors. We present results for model (3), as model (5) and (12) adds little to the
discussion.13 Table 6 presents results for model (3). The reported results are based on 5x5 double
sorted portfolios. The results in Table 6 show that only two α’s remain in the 25 portfolios. The
market premium Rm - Rf is positively and statistically significant for all the portfolios, whilst SMB
is statistically significant except in the largest growth portfolio. HML is also significant in all but
two portfolios and depicts a monotonically increasing coefficient, from strongly negative to
strongly positive. These results indicate that the three factors are useful in explaining Japanese
stock returns.
Psent however, is only significant for five portfolios, with the majority of them occurring
in the smallest stocks and the largest growth stocks. The main contribution from sentiment is in
the removal of α when Psent is included in models (3) and (5) with the other three factors. The
identified pattern for Psent lends evidence to the most common patterns observed for sentiment
effects in the literature. There is an asymmetric effect which is explained in the sentiment literature
by Size. Small stocks and the largest stocks are those which tend to be most affected by sentiment,
as opposed to the stocks in the “middle”. Small stocks tend to be more “sentiment prone” as they
are more likely to be harder to value due to opaque characteristics. We also find that large growth
stocks are affected. Large growth stocks with low B/M may have characteristics, such as
intangibles, which make them more easily influenced by sentiment. Another potential reason is the
reaction to news. Luo et al. (2015) argues that institutional investors will react more to news in
larger stocks than in smaller stocks since their holdings are concentrated in larger stocks; we
observe evidence for this here. One further explanation is that the amount of news per firm, or
news coverage, is most concentrated in the largest stocks.
12 The same models also have the highest average adjusted R-squared values. 13 The results for models (5) and (12) are generally the same and various controls do not add to discussion materially.
18
Table 6 Time-Series Regressions of the 25 Size and Book to Market Sorted Portfolios Japan 3 Factor Model with Psent
B/M Low 2 3 4 High Low 2 3 4 High
α t(α)
Small -0.0002 -0.0002 -0.0001 -0.0001 -0.0000 -1.584 -1.367 -1.041 -0.965 -0.744
2 -0.0003*** 0.0002 -0.0001 -0.0000 -0.0001 -3.100 1.525 -0.999 -0.673 -1.185
3 -0.0002 -0.0001 -0.0001 -0.0001 0.0000 -1.559 -1.863 -1.707 -1.243 0.1290
4 -0.0001 -0.0000 -0.0001 -0.0002** -0.0001 -1.043 -0.621 -1.765 -2.259 -1.018
Big -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -1.100 -1.74 -1.752 -1.078 -0.539
β t(β)
Small 1.1416*** 0.9476*** 0.8959*** 0.8638*** 0.7979*** 41.372 55.241 50.886 53.324 59.652
2 1.1342*** 1.0025*** 0.9063*** 0.9179*** 0.9390*** 64.282 72.264 103.849 82.678 127.002
3 1.1672*** 1.0417*** 0.9713*** 0.9922*** 1.0344*** 61.349 104.848 88.036 70.384 53.377
4 1.0775*** 1.0388*** 1.0148*** 1.0357*** 1.1441*** 62.469 95.736 83.776 79.909 49.357
Big 0.9195*** 0.9497*** 0.9720*** 1.0614*** 1.1029*** 88.515 168.856 67.896 92.627 52.858
δ t(δ)
Small 1.6070*** 1.2831*** 1.1712*** 1.1168*** 1.0559*** 39.394 47.858 36.207 37.752 40.66
2 1.4667*** 1.2310*** 1.0303*** 0.9925*** 1.0203*** 50.995 43.247 61.63 48.386 72.43
3 1.2402*** 0.9367*** 0.8164*** 0.8447*** 0.8885*** 37.472 46.105 43.523 40.533 28.728
4 0.8197*** 0.5760*** 0.5219*** 0.5100*** 0.6512*** 27.059 26.912 24.588 23.310 19.579
Big -0.0127 -0.1084*** -0.0112 0.1027*** 0.1500*** -0.8430 -10.827 -0.472 4.4170 4.399
(Continued next page)
19
B/M Low 2 3 4 High Low 2 3 4 High
γ t(γ)
Small -0.2396*** 0.1894*** 0.3099*** 0.3599*** 0.5012*** -4.772 5.527 10.074 13.273 20.733
2 -0.1921*** -0.0682 0.3697*** 0.4579*** 0.7175*** -5.357 -1.326 18.614 26.041 49.265
3 -0.0807 0.1447*** 0.4120*** 0.5924*** 0.7768*** -1.863 6.281 19.682 28.26 30.358
4 -0.1152*** 0.2411*** 0.4134*** 0.6497*** 0.8615*** -2.939 8.579 16.147 24.962 25.279
Big -0.3798*** -0.0377*** 0.3134*** 0.5755*** 0.8626*** -21.547 -2.815 14.251 23.699 20.083
κ t(κ)
Small 0.0010 0.0019** 0.0013** 0.0013** 0.0019*** 0.824 2.154 1.990 2.177 3.935
2 -0.0014 -0.0010 0.0009 0.0005 -0.0000 -1.725 -1.027 1.866 1.310 -0.023
3 -0.0007 -0.0008 0.0002 0.0007 0.0006 -0.600 -1.476 0.517 1.398 1.005
4 -0.0012 -0.0002 -0.0007 -0.0006 0.0001 -1.473 -0.289 -1.173 -1.078 0.087
Big 0.0013*** 0.0003 -0.0000 -0.0009 -0.0006 3.055 0.958 -0.037 -1.594 -0.578
R-squared
Small 0.755 0.768 0.821 0.856 0.876
2 0.868 0.795 0.906 0.942 0.953
3 0.758 0.910 0.926 0.935 0.928
4 0.837 0.903 0.911 0.914 0.879
Big 0.961 0.977 0.944 0.921 0.785
This Table reports regression results over the period January 2003 – June 2014. This regression uses the daily three factors constructed for Japan and sentiment Psent. In March of year t, stocks are ranked by book to market, and independently sorted by median market capitalization at the end of September of year t. We form 25 stock portfolios based on quintiles for book to market and size. Value-weighted daily returns on these portfolios are from October to September in year t+1. The Model is specified as: , , ,( )p t f t p p m f t p t p t p t p tR R R R SMB HML psent
*** denotes significance at the 1% level, ** 5% level.
20
As discussed previously, we stated that market wide sentiment can potentially influence
stock returns via two channels. The first channel is indirectly, via through the three factors,
whilst the other is directly as an independent factor. The results presented in Table 5 supports
the first argument that sentiment works indirectly, that is primarily through the factors.
However, it is worthwhile discussing the orthogonalization of the three factors to Psent and the
subsequent results for Psent as a direct individual factor. Models (6) and (13) which
orthogonalize the three factors to Psent do not pass the GRS tests. However, orthogonalization
to Psent would theoretically remove the effects of sentiment on the three factors which
potentially indicates how Psent may work independently as a factor. Table 7 presents the results
of model (6). The estimated results indicate that the effect of sentiment is greatest for the largest
sized portfolios. There also appears to be a growth effect, with the coefficient for Psent
increasing for growth stocks and decreasing for value stocks. This pattern does not readily
emerge in Table 6 and model (3) where the 3 other factors are not orthogonalized to Psent.
Psent in model (6) seems to affect large stocks, more so than smaller stocks. This is a result
that is contradictory to results in Table 6, where smaller stocks and in particular smaller growth
stocks should be the most influenced by sentiment as they tend to have “opaque”
characteristics,14 yet large stocks have the largest loadings on Psent once the other factors are
orthogonalized. Again, this may be due to the fact that institutional investors react to news
(sentiment) in larger stocks than smaller stocks given their holdings in larger stocks (Luo et al.
2015).
As indicated by the failed GRS tests there are statistically significant alpha terms that
persist, indicating that this model does not explain all the cross-sectional variation in stock
returns. Psent is also significant for all portfolios which suggests that sentiment does play a
role in explaining excess portfolio returns. The day of the week dummies are insignificant
except for a few portfolios which show a significant coefficient on the Monday dummy and we
do not present those coefficients for sake of brevity.
14 Characteristics such as high information asymmetry, low liquidity and high transaction costs.
21
Table 7 Time-Series Regressions of the 25 Size and Book to Market Sorted Portfolios using Orthogonalized Factors
B/M Low 2 3 4 High Low 2 3 4 High α t(α) Small 0.0013*** 0.0011*** 0.0012*** 0.0010*** 0.0011*** 4.039 4.154 5.810 5.946 7.504 2 0.0007*** 0.0016*** 0.0011*** 0.0011*** 0.0012*** 3.521 6.075 8.977 10.030 11.901 3 0.0014*** 0.0012*** 0.0013*** 0.0012*** 0.0013*** 4.587 8.104 9.784 9.212 9.626 4 0.0018*** 0.0017*** 0.0015*** 0.0014*** 0.0017*** 7.723 9.315 9.096 8.147 8.225 Big 0.0015*** 0.0016*** 0.0017*** 0.0018*** 0.0023*** 13.538 16.755 12.608 10.820 7.197 β t(β) Small 1.1424*** 0.9475*** 0.8962*** 0.8640*** 0.7972*** 41.610 55.044 50.644 52.970 58.871 2 1.1346*** 1.0023*** 0.9066*** 0.9179*** 0.9384*** 63.586 72.548 103.324 81.542 126.315 3 1.1663*** 1.0417*** 0.9713*** 0.9923*** 1.0339*** 61.678 104.264 87.398 69.558 53.206 4 1.0769*** 1.0389*** 1.0144*** 1.0354*** 1.1441*** 62.396 95.329 83.683 79.130 48.983 Big 0.9190*** 0.9494*** 0.9722*** 1.0619*** 1.1036*** 88.141 168.689 67.564 93.172 52.306 δ t(δ) Small 1.6105*** 1.2822*** 1.1713*** 1.1171*** 1.0543*** 39.731 47.679 35.883 37.331 39.964 2 1.4688*** 1.2302*** 1.0317*** 0.9925*** 1.0190*** 50.155 43.228 61.056 47.444 71.698 3 1.2393*** 0.9376*** 0.8168*** 0.8452*** 0.8877*** 37.808 45.978 43.043 39.908 28.529 4 0.8205*** 0.5774*** 0.5218*** 0.5099*** 0.6523*** 26.937 26.906 24.449 22.977 19.395 Big -0.0132 -0.1093*** -0.0108 0.1045*** 0.1523*** -0.869 -10.887 -0.453 4.514 4.430
(Continued next page)
22
B/M Low 2 3 4 High Low 2 3 4 High γ t(γ) Small -0.2314*** 0.1894*** 0.3109*** 0.3616*** 0.4989*** -4.598 5.496 10.010 13.105 20.215 2 -0.1891*** -0.0669 0.3710*** 0.4577*** 0.7155*** -5.258 -1.294 18.834 25.615 49.218 3 -0.0852** 0.1446*** 0.4119*** 0.5921*** 0.7736*** -1.968 6.263 19.526 27.811 30.236 4 -0.1181*** 0.2409*** 0.4110*** 0.6474*** 0.8596*** -2.998 8.538 15.928 24.559 25.049 Big -0.3824*** -0.0387*** 0.3151*** 0.5793*** 0.8647*** -21.590 -2.875 14.006 23.832 20.003 κ t(κ) Small 0.0218*** 0.0184*** 0.0168*** 0.0162*** 0.0151*** 19.322 23.331 25.172 27.558 29.941 2 0.0197*** 0.0180*** 0.0174*** 0.0173*** 0.0166*** 27.301 20.602 37.680 45.016 51.168 3 0.0225*** 0.0205*** 0.0197*** 0.0201*** 0.0203*** 20.624 41.617 45.320 47.450 41.203 4 0.0225*** 0.0231*** 0.0218*** 0.0219*** 0.0240*** 30.218 41.766 42.476 42.567 34.040 Big 0.0267*** 0.0262*** 0.0250*** 0.0250*** 0.0255*** 70.677 89.780 52.929 45.778 24.460 j t(j) Small 0.0007 0.0016*** 0.0010*** 0.0009*** 0.0011*** 1.374 4.556 3.379 3.702 5.175 2 0.0009*** 0.0009*** 0.0004** 0.0007*** 0.0010*** 2.737 2.676 1.970 4.328 6.528 3 0.0012 0.0003 0.0004** 0.0003** 0.0005*** 1.388 1.338 1.891 1.718 2.998 4 -0.0002 -0.0004 -0.0002 -0.0001 -0.0002 -0.651 -1.606 -0.817 -0.463 -0.556 Big -0.0012*** -0.0007*** -0.0007*** -0.0006** -0.0008** -5.702 -5.289 -3.254 -2.270 -1.933
(Continued next page)
23
B/M Low 2 3 4 High Low 2 3 4 High t( )
Small -0.0009 -0.0002 -0.0002 -0.0001 0.0002 -1.638 -0.380 -0.555 -0.373 0.844 2 0.0001 -0.0004 -0.0001 0.0000 0.0002 0.243 -0.897 -0.634 -0.028 1.163 3 0.0008** 0.0001 0.0000 0.0001 0.0006** 1.871 0.401 0.166 0.341 2.140 4 0.0004 0.0002 0.0004 0.0003 0.0005 1.159 0.840 1.403 1.261 1.456 Big 0.0004** 0.0000 -0.0002 -0.0002 0.0000 2.050 0.236 -0.894 -0.767 -0.054 l t(l) Small -0.0003 -0.0008** -0.0006** -0.0006** -0.0006*** -0.453 -1.822 -1.856 -2.090 -2.813 2 -0.0009*** -0.0009** -0.0007** -0.0008*** -0.0009*** -2.667 -2.265 -2.497 -4.447 -5.893 3 -0.0013*** -0.0011*** -0.0008*** -0.0007*** -0.0007*** -2.878 -4.462 -3.585 -4.134 -3.570 4 -0.0010*** -0.0006* -0.0007*** -0.0009*** -0.0009*** -2.782 -2.573 -2.782 -3.929 -2.960 Big -0.0005*** -0.0004*** -0.0008*** -0.0010*** -0.0009** -2.783 -3.050 -4.091 -4.023 -1.759 R-squared Small 0.756 0.768 0.821 0.857 0.877 2 0.869 0.795 0.906 0.943 0.953 3 0.760 0.910 0.926 0.935 0.928 4 0.838 0.903 0.911 0.914 0.879 Big 0.961 0.977 0.944 0.922 0.785
This Table reports regression results over the period January 2003 – June 2014. This regression uses the daily three factors constructed for Japan and orthogonalized to our measure of daily news sentiment Psent. In March of year t, stocks are ranked by book to market, and independently sorted by median market capitalization at the end of September of year t. We form 25 stock portfolios based on quintiles for book to market and size. Value-weighted daily returns on these portfolios are from October to September in year t+1.
The Model is specified as: , , , , ,( )p t f t p p m t f t p t p t p t p t p t p t p t p tR R R R SMB HML psent j Jan March l July d Day
Where dpDay is a day of the week dummy. *** denotes significance at the 1% level, ** 5% level.
24
As a robustness check, we do a Wald test to jointly test whether the three coefficients
and Psent are different to zero. This aim is to test whether Psent adds to the three-factor model
by testing the null hypothesis that the factors (Psent in addition to the factors Rm-Rf, SMB and
HML) are jointly equal to zero. Table 8 presents these results.
Table 8 Wald tests for usefulness of factors.
Panel A: All Factors
χ 2 5120.16***
Panel B: All Factorsorthog
χ 2 5128.72***
This Table presents a Wald test (χ 2) of the null hypothesis that the coefficients reported in Table 6 and Table 7
for all the factors, are jointly equal to zero. Panel A presents results for the three factors and Psent. Panel B
presents results for the 3 factors orthogonalized to Psent.
*** denotes significance at the 1% level, ** denotes significance at the 5% level.
Panel A presents results for the test applied to the coefficients presented in Table 6
whilst panel B presents results for the coefficients presented in Table 7. Both of the tests reject
the null hypothesis which indicates that the coefficients are not jointly equal to zero. As both
tests reject the null hypothesis this presents further evidence that Psent should be considered
in addition to the three Fama and French factors for Japanese asset pricing.
It appears that sentiment has asymmetric effects, either through influencing the Fama
and French factors directly, and when used as an individual factor. However, the effect appears
to be relatively small relative to the three factors and tends to work in the extremes through
size. Future research could extend on this by examining multiple sentiment measures to test
the sensitivity of this to other metrics and by extending to other markets.
IV. Conclusion
The results presented in this paper suggests that there is a role for sentiment in the cross-
section of Japanese stock returns. Using the Fama and French three-factor model we find that
sentiment appears to primarily influence returns indirectly, via the Fama and French three
25
factors. The effects of sentiment are relatively small yet appear to add value to our
understanding of returns when included as a factor. We document asymmetric effects of
sentiment where small growth stocks and large stocks appear to be most affected. These results
suggest that the addition of a sentiment factor should be considered in asset pricing model and
in particular for Japan.
26
References
Akhtar, S., Faff, R., Oliver, B., Subrahmanyam, A., 2011. The power of bad: The negativity bias in
Australian consumer sentiment announcements on stock returns. Journal of Banking &
Finance 35, 1239-1249.
Allen, D. E., McAleer, M. J., & Singh, A. K. 2015. Chapter 19 - Machine News and Volatility: The
Dow Jones Industrial Average and the TRNA Real-Time High-Frequency Sentiment Series.
In G. N. Gregoriou Ed., The Handbook of High Frequency Trading pp. 327-344. San Diego:
Academic Press.
Aman, H. 2013. An analysis of the impact of media coverage on stock price crashes and jumps:
Evidence from Japan. Pacific-Basin Finance Journal, 24, 22-38.
Antoniou, C., Doukas, J.A., Subrahmanyam, A., 2013. Cognitive Dissonance, Sentiment, and
Momentum. Journal of Financial and Quantitative Analysis, 48, 245-275.
Baker, M., & Wurgler, J. 2006. Investor Sentiment and the Cross-Section of Stock Returns. Journal of
Finance, 614, 1645-1680.
Baker, M., & Wurgler, J. 2007. Investor Sentiment in the Stock Market. Journal of Economic
Perspectives, 212, 129-151.
Baker, M., Wurgler, J., Yuan, Y., 2012. Global, local, and contagious investor sentiment. Journal of
Financial Economics, 104, 272-287.
Berger, D., & Turtle, H. J. 2012. Cross-sectional performance and investor sentiment in a multiple
risk factor model. Journal of Banking & Finance, 364, 1107-1121.
Brown, G. W., & Cliff, M. T. 2005. Investor Sentiment and Asset Valuation. Journal of Business,
782, 405-440.
Cakici, N., 2015. The five-factor Fama-French model: International evidence. Working Paper.
Retrieved from SSRN: ssrn.com/abstract=2601662
Carhart, M. M. 1997. On Persistence in Mutual Fund Performance. Journal of Finance, 521, 57-82.
Chan, F., Durand, B. R., Khuu, J., & Smales, A. L., 2016. The Validity of Investment Sentiment
Proxies. International Review of Finance. Forthcoming.
Chan, L. K. C., Hamao, Y., & Lakonishok, J. 1991. Fundamentals and Stock Returns in Japan.
Journal of Finance, 465, 1739-1764.
Chang, R. P., McLeavey, D. W., & Rhee, S. G. 1995. Short- Term Abnormal Returns of the
Contrarian Strategy In the Japanese Stock Market. Journal of Business Finance &
Accounting, 227, 1035-1048.
27
Chen, N.f., Kan, R., Miller, M.H., 1993. Are the Discounts on Closed‐End Funds a Sentiment Index?
Journal of Finance, 48, 795-800.
Chiao, C., & Hueng, C. J. 2005. Overreaction effects independent of risk and characteristics: evidence
from the Japanese stock market. Japan and the World Economy, 174, 431-455.
Chung, S.-L., Hung, C.-H., Yeh, C.-Y., 2012. When does investor sentiment predict stock returns?
Journal of Empirical Finance, 19, 217-240.
Daniel, K., Titman, S., & Wei, K. C. J. 2001). Explaining the Cross-Section of Stock Returns in
Japan: Factors or Characteristics? The Journal of Finance, 562, 743-766.
Durand, R. B., Limkriangkrai, M., & Chai, D. 2016. The Australian asset-pricing debate. Accounting
& Finance, 562, 393-421.
Dzielinski, M. 2011. News sensitivity and the cross-section of stock returns. NCCR Finrisk Working
Paper no. 719.
Fama, E. F., & French, K. R. 1993. Common risk factors in the returns on stocks and bonds. Journal
of Financial Economics, 331, 3-56.
Fama, E. F., & French, K. R. 1996. Multifactor Explanations of Asset Pricing Anomalies. The Journal
of Finance, 511, 55-84.
Fama, E. F., & French, K. R. 2012. Size, value, and momentum in international stock returns. Journal
of Financial Economics, 1053, 457-472.
Fama, E. F., & French, K. R. 2015a. A five-factor asset pricing model. Journal of Financial
Economics, 1161, 1-22.
Fama, E. F., & French, K. R. 2015b. International Tests of a Five-Factor Asset Pricing Model.
Working Paper. Retrieved from SSRN: http://ssrn.com/abstract=2622782
García, D. 2013. Sentiment during Recessions. Journal of Finance, 683, 1267-1300.
Groß-Klußmann, A., & Hautsch, N. 2011. When machines read the news: Using automated text
analytics to quantify high frequency news-implied market reactions. Journal of Empirical
Finance, 182, 321-340.
Gunaratne, P. S. M., & Yonesawa, Y. 1997. Return reversals in the Tokyo Stock Exchange: A test of
stock market overreaction. Japan and the World Economy, 93, 363-384.
Khuu, J., Durand, R. B., & Smales, L. A. 2016. Melancholia and Japanese stock returns – 2003 to
2012. Pacific-Basin Finance Journal. Forthcoming.
Lemmon, M., & Portniaguina, E. 2006. Consumer Confidence and Asset Prices: Some Empirical
Evidence. Review of Financial Studies, 194, 1499-1529.
28
Lintner, J. 1965. The Valuation of Risk Assets and the Selection of Risky Investments in Stock
Portfolios and Capital Budgets. The Review of Economics and Statistics, 471, 13-37.
Lemmon, M., Portniaguina, E., 2006. Consumer Confidence and Asset Prices: Some Empirical
Evidence. Review of Financial Studies, 19, 1499-1529.
Luo, Yangqiulu and Wu, Guojun and Xu, Yexiao, Idiosyncratic Risk Matters to Large Stocks!
eWorking Paper. Available at SSRN: http://ssrn.com/abstract=2716357 or
http://dx.doi.org/10.2139/ssrn.2716357
Schmeling, M. 2009. Investor sentiment and stock returns: Some international evidence. Journal of
Empirical Finance, 163, 394-408.
Sharpe , W. F. 1964. Capital asset prices: A theory of market equilibrium under conditions of risk.
The Journal of Finance, 193, 425-442.
Smales, L. A. 2014. News sentiment in the gold futures market. Journal of Banking & Finance, 490,
275-286.
Stambaugh, R. F., Yu, J., & Yuan, Y. 2012. The short of it: Investor sentiment and anomalies. Journal
of Financial Economics, 1042, 288-302.
Tetlock, P. C. 2007. Giving Content to Investor Sentiment: The Role of Media in the Stock Market.
Journal of Finance, 623, 1139-1168.
Tetlock, P. C., Saar-Tsechansky, M., & Macskassy, S. 2008. More Than Words: Quantifying
Language to Measure Firms' Fundamentals. The Journal of Finance, 633, 1437-1467.
Tsuji, C., 2006. Does investors' sentiment predict stock price changes? With analyses of naive
extrapolation and the salience hypothesis in Japan. Applied Financial Economics Letters, 2,
353-359.
Uhl, M. W. 2014. Reuters Sentiment and Stock Returns. Journal of Behavioral Finance, 154, 287-
298.
Yu, J., Yuan, Y., 2011. Investor sentiment and the mean–variance relation. Journal of Financial
Economics, 100, 367-381.
top related