Journal of Economic Behavior and Organization 193 (2022) 443–472 

Contents lists available at ScienceDirect 

Journal of Economic Behavior and Organization 

journal homepage: www.elsevier.com/locate/jebo 

The effects of public sentiments and feelings on stock market 

behavior: Evidence from Australia 

Aviral Kumar Tiwari a , b , c , ∗, Emmanuel Joel Aikins Abakah 

d , 
Christiana Osei Bonsu 

e , Nana Kwasi Karikari f , Shawkat Hammoudeh 

g 

a Rajagiri Business School, Kochi, India 
b South Ural State University, Lenin Prospect 76, Chelyabinsk, 454080, Russian Federation 
c CEENRG Research Fellow, Department of Land Economy, University of Cambridge, UK 
d University of Cape Coast School of Economics, Cape Coast Ghana 
e Australian Institute of Business, Australia 
f University of Ghana Business School, Accra Ghana 
g Drexel University, Philadelphia, United States 

a r t i c l e i n f o 

Article history: 

Received 13 April 2021 

Revised 10 November 2021 

Accepted 21 November 2021 

Available online 12 December 2021 

JEL Classifications: 

C14 

G10 

G12 

G14 

G40 

Keywords: 

Consumer sentiments 

Sector stock market returns 

Nonparametric causality test 

a b s t r a c t 

This paper investigates the empirical evidence of the effects of public sentiments on in- 

dustry stock returns and volatility dynamics in Australia based on the states of the market 

that relates to the conditional quantiles of public sentiments and sectoral stocks, using 

the robust nonparametric causality-in -quantile test. We adopt the monthly overall con- 

sumer sentiments index and four of its components including the sentiments for rural 

Australia and the age groups 18–24, 25–44, and 45 and above. Our nine industry stocks 

include Health Care, Consumer Discretionary, Consumer Staples, Utilities Financials, Real 

Estate, Industrials, Basic Materials and Energy, with data spanning from October 1974 to 

October 2020. The results from the nonlinear causality test show a directional and bidirec- 

tional causality between measures of consumer sentiments and returns of industry stocks. 

Interestingly, we note that the sentiments of individuals aged 45 and above cause the re- 

turns of all the nine sectors. Next, we explore the predictive power of sentiments on in- 

dustry stock returns, using the nonparametric causality-in-quantile test. We find that the 

predictability between sentiments and industry stock returns is high in the normal mar- 

ket state but drops when the consumers’ perceptions enter into the extreme bearish and 

bullish states. Additionally, the findings show a risk (volatility) transfer from sentiments 

to the industry stock returns in some cases under different market conditions. We offer 

some implications based on our findings for the stakeholders and market participants who 

develop their strategies depending on market conditions and sentiments. 

© 2021 Elsevier B.V. All rights reserved. 

 

 

1. Introduction 

The global stock market is, to a large extent, one of the world’s largest financial markets. For the past decade, factors

such as financial, economic, political, environmental, and health crises have urged financial researchers to analyze the ef- 

fects these factors on stock market returns and volatility (e.g., Balcilar et al., 2016 ; Tiwari et al., 2020 ; Le et al., 2021 among
∗ Corresponding author. 

E-mail addresses: aviral.eco@gmail.com , aviral.kr.tiwari@gmail.com (A.K. Tiwari). 

https://doi.org/10.1016/j.jebo.2021.11.026 

0167-2681/© 2021 Elsevier B.V. All rights reserved. 

https://doi.org/10.1016/j.jebo.2021.11.026
http://www.ScienceDirect.com
http://www.elsevier.com/locate/jebo
http://crossmark.crossref.org/dialog/?doi=10.1016/j.jebo.2021.11.026&domain=pdf
mailto:aviral.eco@gmail.com
mailto:aviral.kr.tiwari@gmail.com
https://doi.org/10.1016/j.jebo.2021.11.026


A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

others). Because of its critical implications for portfolio allocation, asset pricing, and policy development, stock return pre- 

dictability is of interest to both investors and policymakers ( Gil-Alana et al., 2018 ). As a result, many economic agents place

a high value on accurate stock market returns and volatility prediction. Naturally, there is a large body of works on the

predictability of stock market returns and volatility, with a general conclusion that a range of macroeconomic and financial 

variables may predict stock returns in both developed and emerging stock markets. Fama and French (1992) show that finan-

cial variables such as book-to-market, cash-flow-yields and size have a strong forecasting ability in predicting the U.S. stock 

market. Similarly, other studies have highlighted the importance of common risk variables such market risk, book-to-market 

and size in forecasting stock markets ( Fama and French, 1993 ; Gupta et al., 2014 ), while other studies have underscored the

potential predictive power of a set of macroeconomic factors including inflation, interest rates and output ( Cochrane, 1991 ;

Fame and French, 1989 ; Ferson and Harvey, 1993 ; Persaran and Timmermann, 20 0 0 ). Since the seminal study of Cootner

(1960) , the issue of identifying factors that can predict stock returns which are appropriate to conduct such a study has

remained a hotly debated topic in the finance and economic research. One general theme that emerges from this literature 

is that predicting market returns and volatility fluctuations is a difficult task. 

From the aforementioned backdrop, the thrust of our study is to employ the novel nonparametric causality-in-quantile 

test advanced by Balcilar et al. (2016) to examine the predictability of returns and the volatility of industry stocks in Aus-

tralia over the monthly period 1974:10–2020:10, based on Australia’s consumer sentiments. Theorists in the area of be- 

havioral finance argue that investors have bounded rationality and have psychological and emotional factors that influence 

their decision-making processes. As a result, rather than the statistical features of stock returns, the state of investors’ psy- 

chology plays a considerably more important role in the price formation process. As Hirshleifer (2001 , p2) suggests, " The

central task of asset pricing is to examine how expected returns are related to risk and to investor mis-valuation ". In his ex-

tensive review, he finds that over time, a broader psychological paradigm that includes full rationality as a special case will

dominate the purely rational paradigm. The optimistic and pessimistic feelings of investors’ general emotions about the mar- 

kets are considered an additional source of undiversifiable or systematic risk in the behavioral finance theories ( Kıyılar and

Akkaya, 2016 ). Concisely put, these feelings are known as sentiments, and it has been suggested that unpredictable changes 

in sentiments are likely to have an impact on stock prices ( Verma et al., 2008 ). 

One of the notable elements of the literature of behavioral finance is the concentration on both the sentiments and the

causal relationship existing between sentiments and stock returns. To this end, we contribute to the strand of the literature 

on the relation between sentiments and stock returns, by using a direct unique survey data on public sentiments in Aus- 

tralia which is published by the Melbourne Institute and Westpac. 1 Specifically, we use the aggregate Australian Consumer 

Sentiment Index (CSI). Brown and Cliff (2004) suggest that the aggregate measure of consumer sentiments may not provide 

a sufficient impact on stock returns. Additionally, market players may have different expectations, risk profiles, views, and 

information set among other things ( Mensi et al., 2016 ). Hence, we examine the influence of disaggregate measures of the

overall consumer sentiment index on sectoral stock returns comprising of sentiments for Rural Australia, and sentiments 

for ages: 18 to 24; 25 to 44, and 45 and above. Baker et al. (2003) suggested that the extent of the effect of consumers’

and investors’ propensity to speculate on stocks may vary across different industries, according to the subjectivity of their 

valuations. Hence, for our measure of industry stock returns, we consider all the nine sectors of Australia’s stock mar- 

kets: Financials, Health Care, Consumer Discretionary, Consumer Staples, Industrials, Basic Materials, Real Estate, Energy and 

Utilities. 

Our choice of the Australian stock market is because the majority of earlier studies rely on indirect measures of senti-

ments such as the close-end-fund discounts, the advance/decline ratio ( Swaminathan, 1996 ; Neal and Wheatley, 1998 ) and

share turnovers ( Deuskar, 2004 ) or the extrapolations constructed on these measurements. However, the Australian con- 

sumer sentiment index, which is a direct survey published by the Melbourne Institute and Westpac, has several unique 

features that should help us better comprehend the influence of sentiments on stock price movement. First, it is a regu-

lar survey of public opinion on existing and future economic conditions. It is one of the most widely followed and covered

monthly news stories in the media, along with other macroeconomic data. As a result, the survey should provide a thorough

picture of public opinion on the economy. Second, the composite index is made up of five survey items, conducted approx-

imately one week before the consumer sentiment is made public. It is therefore a reliable indication of how people feel

about the economy, personal finances, and spending ( Akhtar et al., 2010 ). Third, the Australian consumer sentiment index 

has both an aggregate index for the overall country and disaggregate measures for different age groups and regions, which 

provides enough opportunity to conduct an in-depth analysis of the subject matter. In addition, the Australian sentiment 

index is only provided once a month, but other markets such as the U.S. sentiment data is frequently released in stages

throughout the month, making the reliability and compatibility of the measurement problematic ( Akhtar et al., 2010 ). 

In recent years, several studies in the finance and economic literature have examined the relationship between senti- 

ments and stock behavior across different markets including the European markets ( Jansen and Nahuis, 2003 ), the U.S. mar-

kets ( Gupta et al., 2014 ), the Asian markets ( Akhtar et al., 2010 ), and the Shanghai stock exchange ( Kling and Gao, 2008 ),

using diverse estimation techniques. In this literature, studies in this area have used linear models, cointegration models, 

vector autoregressive (VAR) models among other linear models. However, due to the limitations of linear correlation models, 
1 The Australian Consumer Sentiment Index (CSI) is comparable to the CSI produced by the University of Michigan’s Survey Research Center in the United 

States. 

4 4 4 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

the multivariate GARCH models have emerged as the standard method for modeling time-varying correlations between sen- 

timents and stock returns. As a result, Patton (2006) and Garcia and Tsafack (2011) documented one important constraint of 

the multivariate GARCH approach which is the assumption that returns innovations are characterized by a symmetric mul- 

tivariate normal or Student- t distribution. Therefore, this assumption is inconsistent with the empirical evidence because 

financial return distributions differ from normal distributions in that they have heavier tails ( Embrechts et al., 2002 ). In

view of this literature, researchers in modern times have resorted to the use of new robust estimation techniques such as

copulas, among others, to examine dependence among finance and economic variables since these new methods overcome 

the drawbacks of earlier models (e.g., Abakah et al., 2021a; Abakah et al., 2021b; Mensah and Alagidede, 2017 ; Aviral et al.,

2021). 

To differentiate our paper from prior empirical studies in terms of methodology (e.g., Lee et al., 2002 ; ; Schmeling, 2009 ),

we employ Balcilar et al. (2016) novel nonparametric causality-in-quantile test to assess the sensitivity of sentiments of 

consumers in Australia to the cross sections of its industry stock returns. This causality-in-quantile test, which we use in this

paper, combines the frameworks of Nishiyama et al. (2011) ’s k-th order causality and Jeong et al. (2012) ’s quantile causality,

and can thus be considered a more general version of the former. The following are some of the innovative features of

the causality-in-quantile approach. Firstly, it is robust to misspecification errors due to its ability to detect the underlying 

dependence structure between the time series under investigation. This could be particularly crucial, given that stock prices 

are well known to have nonlinear dynamics ( Abakah et al., 2018 ). 

Secondly, through this methodology we render the ability to test not just for causality-in-the mean (the 1st moment), 

but also for causality in the tails of the variables’ joint distribution, which is especially crucial if the dependent variable

bears fat tails. Finally, we are able to look into the causality-in-variance, and hence the volatility spillovers, because there 

are times when the causality in the conditional-mean does not occur, but higher order interdependencies may emerge. 

This study further uses a nonlinear causality test to examine the consumer sentiment-stock return relationship. The 

inclusion of nonlinear causality tests seeks to overcome the bottlenecks that characterize the linear causality tests. The in- 

ability of linear causality to detect nonlinear relationships present in the economic and financial variables is catered for by 

nonlinear causalities employed in this study. The presence of nonlinearity could result in a multiplicative and disproportion- 

ate impacts on financial variables ( Abakah et al., 2018 ; Kyrtsou and Labys, 2006 ), which for this study is stock returns, due

to minute changes in other variables such as consumer sentiments. 

A critical question that we have yet to address is: What are the intuitive and theoretical reasons that can lead one to

think that consumer sentiments predict sectoral stock returns and volatility? According to the rational asset pricing the- 

ory, there is a positive relationship between expected returns and risk following the assertion of Merton (1980) . However,

behavioral theorists in finance added the concept of noise trader sentiment predictive power and its effects prevalent in 

financial markets for strategic and prudent allocation of financial assets ( De Long et al., 1990 a; Fisher and Statman, 20 0 0 ).

The theoretical idea provides a pragmatic proof that the consumer sentiments can impact the prices of financial assets un- 

der two assumptions: (i) the predominant role of sentiment (noise) traders in the movement of assets is crucial, and (ii)

the limitation of arbitrage in terms of transaction costs is prevailing. Some empirical findings 2 argue from the conventional 

wisdom, and thus makes a contribution to the subject by providing strong evidence that investor sentiments affect stock 

returns. 

At the most basic level, one could consider sentiments to be either optimistic or pessimistic regarding the economy. Pos- 

itive expectations are seen to boost stock prices, while negative expectations are thought to depress them. When sentiments 

are strong and persistent, they can lead to irrational exuberance, as seen in the late-90 s high-tech financial bubble known

as the Dot com, or the panic scenario that occurred during the October 1987 crisis. Such occurrences do not indicate any

abrupt changes in the fundamentals, but rather significant adjustments in consumer or investor sentiments ( Siegel, 1992 ). 

At this extreme level, investor sentiments may be linked to the propensity to speculate. Furthermore, given that financial 

asset returns are a function of the state of the economy, which is susceptible to fluctuations caused by sentiments, uncer-

tainties, and other factors, would suggest an indirect channel via which sentiments can affect stock returns and volatility. 

Benigno and Ricci (2011) and Martín and Urrea (2007) formalize these channels using new Keynesian general equilibrium 

frameworks. 

Our contribution to the literature on the relationship between the public sentiment and the returns of stocks is five- 

fold. First, to the best of our knowledge, this work premieres the use of the causality-in-quantile approach to investigate 

the predictability of industry stock returns and volatility simultaneously using public sentiment in Australia. Thus, we give 

new empirical evidence on the impact of public sentiment on stock returns based on the states that pertain to the mar-

ket. This corresponds to the conditional quantiles of the sentiments and stock markets, based on Bacilar et al. (2016) ’s

novel nonparametric test. The causality-in-quantile methodology is unique because it is unaffected by outlier observations. 

It can also typify Granger causality throughout the entire distribution. Because financial time series are characterized by 

fat tails due to the occurrence of either individual or simultaneous co-jumps in financial and commodity markets, which 

constitute the basis for our stock sectors, this method takes care of regime changes and jumps in the data ( Dungey and

Hvozdyk, 2012 ; Chevallier and Lelpo, 2014 ). 
2 ( Baker and Wurgler, 2006 ; Kumar and Lee, 2006 ; Tetlock, 2007 ; Edmans et al., 2007 ; Da et al., 2011 ; Stambaugh et al., 2012 ; Siganos et al., 2014 ; 

Da et al., 2015 ; Huang et al., 2015 ). 

445 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Second, because the rejection of causality-in-mean does not rule out the possibility of causation in higher moments, 

the paper looks into causality-in-variance between sentiments and Australian sectoral stocks. Third, determining the inten- 

sity and direction of causation under varied market circumstances would be beneficial to stakeholders who develop their 

strategies depending on conditions of the market. Fourth, since consumers may differ in terms of expectations, beliefs, and 

risk profiles, we use a component of the overall aggregate measure of consumer sentiments, including sentiments for rural 

Australia and different age groups. We are the first to examine the impact of sentiments of various age groups and rural

Australia on industry stock returns and volatility dynamics in different market conditions. This is a noble contribution to 

the literature. Fifth, we examine the cross-sectional effect of sentiments on sectoral stock returns. Evidently, while several 

studies have assessed the effect of sentiments or the emotional states of asset traders on stock prices (see Akhtar et al.,

2011 ; Baker and Wurgler, 2006 ; Breaban and Noussair, 2018 ; Chu et al., 2016 ; Hu et al., 2015 ), the decomposition of the

aggregate stocks in relation to sentiments is largely unexplored. In particular, the paper examines the distributional pre- 

dictability between changes in consumer sentiments and nine stock sectors, mindful that sectors with less predictable cash 

flows are more likely to face swings in sentiment than the traditional and matured sectors ( Baker and Wurgler, 2006 ). 

Our findings from the nonlinear causality test reveal a directional causality between Overall CSI and Health Care, and 

Financial sectors. We further note that CSI Rural Australia granger causes stock returns of Energy, Utilities and Basic Ma- 

terials sectors. Additionally, sentiments of consumers aged between 18 and 24 cause Health Care, Financial, Real Estate, 

Industrials, Energy and Utilities sectors. The findings show a bidirectional relationship between sentiments of the individ- 

uals aged between 25 and 44 and Real Estate, Industrials Basic Materials and Utilities sectors. Finally, we document di- 

rectional causality between CSI for the age group 45 and above and the nine sectors. The result from the nonparametric

causality-in-quantile reveals the causality-in-return is asymmetric and unidirectional running from consumer sentiments to 

sectoral stock returns in some cases, implying that the predictability of the sectoral stock returns is high in the normal

market but drops when the consumers’ perceptions shift to the extreme bearish and bullish states. Under certain mar- 

ket situations, the causality-in-variance shows a directional and bidirectional causality between sentiments and industry 

returns. 

The remainder of the paper is structured as follows. The second section briefly reviews the existing literature. The third 

section presents the methodological framework adopted in the study. We then analyze the findings and conclude the paper 

in the fourth and fifth sections, respectively. 

2. Literature review 

Asymmetric information in the financial markets influences investor behaviors ( Shleifer and Vishny, 1990 ). The imper- 

fection of the hypothesis for efficient markets, which states that prices of stocks do not reflect all available information in

the market could be attributed to (irrational) investor behaviors that are seldom quantifiable ( Hsu et al., 2005 ). Shleifer and

Summers (1990) also allude to the existence of rational and irrational traders that are very likely to influence equity prices.

Thus, an investor sensitivity nexus to stock returns has attracted research concerns. 

Several studies in the finance and economic literature have examined the relationship between sentiments and stock 

behavior. For example, Jansen and Nahuis (2003) find evidence of a positive correlation between investor sentiments and 

stock returns of 11 European stock markets. Similar results were obtained by Gupta et al. (2014) in the U.S. stock mar-

ket returns.. However, Brown and Cliff (2004) find no predictability power of stock returns emanating from both investor 

and institutional sentiments. Other subsequent studies indicate that consumer sentiments contrarily predict future stock 

returns ( Bathia and Bredin, 2013 ; Schmeling, 2009 ), whereas by using the Shanghai stock exchange movement, Kling and

Gao (2008) detect a reverse causality from stock returns to institutional investor sentiments in the short run. However, this 

relation did not exist for the long-run, and further negative returns were more sensitive and impactful than positive returns. 

Li (2019) finds an existence of a contagious effect from investor sentiments to stock returns and volatilities, thereby affirm- 

ing the sentiment-stock price relationship. Otoo (1999) also demonstrates for the U.S. stock market that changes in equity 

values and changes in consumer sentiments are concurrently correlated. Moreover, rises in equity prices enhance consumer 

confidence with a (short) lag, but that the reverse does not hold. Measuring sentiments based on the COVID-19 pandemic, 

several studies show that investor sentiments impact stock market returns ( Haroon and Rizvi, 2020 ; Lyócsa et al., 2020 ;

Salisu and Vo, 2020 ). 

The sensitivity of stock returns to the sentiments of investors remains fundamentally a central issue in finance and eco- 

nomic research since the existing literature could be recounted as not exhaustive. The inconclusive findings and the diver- 

gence in the results continue to attract research in this area. Prior studies, spew mixed findings which could be attributed

to the proxies used for investor sentiments ( Kothari and Shanken, 1997 ; Baker and Stein 2004 ; Baker and Wurgler, 2006 ;

Kurov, 2010 ; Krokida et al., 2020 ), the empirical methods adopted and the dynamics of stock markets to asset, institutional

and economic heterogeneities in jurisdictions of the concentration of the studies. 

Rahman and Shamsuddin (2019) studied the role of investor sentiments in explaining the price-earnings ratios of stocks 

of the G7 countries. Their findings indicate that the price-to-earnings ratio increased with the improvements in the senti- 

ments of investors in the region. In an earlier study, Baker and Stein (2004) had underscored that irrational investors traded

more in a bullish market when these investors felt cheerful, and also traded more in the bearish market when a gloomy

feeling was dominant. Again, Chen et al. (2019) contended that investor sentiments have an important effect on the degree

of market efficiency and equity mispricing. The literature from Yu and Yuan (2011) also supported earlier assertions that 
446 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

show that market participation increases when there is an increase in sentiment-driven market participants. Those authors’ 

study shows evidence of a strong and positive equity return-volatility trade-off when investor sentiments are low and vice 

versa. 

Akhtar et al. (2011) also analyzed the effect of consumer sentiment announcements on equity markets. Their study in- 

dicated that the announcements of consumer sentiments had a significant negative effect on the Australian stock markets. 

The Akhtar et al. (2011) study endorses previous and later literature that finds an inverse relationship between investor 

sentiments and stock returns, thus implying that high investor sentiments were related to low future equity returns and 

vice versa (see Brown and Cliff, 2005 ; Baker and Wurgler, 2007 ; Baker et al., 2012 ; Bathia and Bredin, 2013 ; Siganos et al.,

2014 ). According to Ritter (2003) and Ljungqvist (2006) , the underpricing of IPOs was associated with investor sentiments

which were described as enthusiastic. Schmeling (2009) also looked at how consumer confidence (used as a proxy for the 

sentiment of individual investors) impacted expected stock returns in 18 developed countries. The findings of Schmeling re- 

vealed a negative forecast of aggregate stock market returns on average by investor sentiments across the countries studied. 

Higher levels of investor sentiments (consumer confidence) led to lower future stock returns and vice versa. This relation- 

ship, according to Schmeling, held true for the returns of value stocks, growth stocks, small stocks, and different forecasting 

timeframes. 

Gupta and Banerjee (2019) examined the impact of sentiments from foreign news on the financial performance of domes- 

tic firms in the energy sector of the Organization of Petroleum Exporting Countries (OPEC). Their study finds a dominance 

of the negative effect of news on the performances. However, that study reports that in eras of low consumer confidence,

positive OPEC news negatively influenced the U.S. energy stock returns. Kim and Kim (2014) , using internet messages to

measure investor sentiment, find no significant relation between investor sentiments and stock returns. Using Facebook’s 

Gross National Happiness Index, Siganos et al. (2014) investigate the relationship between daily sentiments and trading 

behavior in 20 international stock markets. They discover a positive contemporaneous relationship between emotions and 

stock returns. Huynh et al. (2021) used the 17 largest economies’ six behavioral indicators (media coverage, false news, fear, 

sentiment, media hype, and infodemic) during the COVID-19 pandemic to construct connectedness among these indicators 

and the stock markets. Their study shows that total and net connectedness indicate that China, Germany, the UK and the

U.S. were the epicenters of the sentiment shocks that were transferred to other countries. 

Using a Granger causality test developed by Shi et al. (2018) , Ça ̆glı et al. (2018) analyzed the causal relationship between

stock returns and investor sentiments in Borsa Istanbul and found no causality but alluded to the possibility of not using

the non-linearity causality. On the other hand, Wang et al. (2006) found that most of measures of investor sentiments

were caused by stock returns and volatility rather than by sentiments, therefore causing stock return and volatile behavior. 

Their study again concludes that all sentiment variables have an extremely limited forecasting power once the returns were 

included as a forecasting variable. 

The long-term and short-term variation effects of investor sentiments were accounted for by Ni et al. (2015) as their

study documents a mean reverting behavior of stock returns, assuming predictability with a large and positive effect in the 

short-run and a contrary (small and negative) effect in the long-run. Alouiet al., (2021) examine the relevance of investor

sentiments to the Islamic stock-bond interplay in the time-frequency domain, using a multiple and partial wavelet coher- 

ence as well as bivariate and multivariate nonlinear causality tests. They conclude that connectedness exists in both the 

long-run and short-run horizons. Using Consumer Confidence Index and Investor Intelligence survey in the United States, 

Lemmon and Portniaguina (2006) and Qiu and Welch (2004) indicate that investor sentiments were useful in predicting the 

equity returns of stocks with less than large capitalizations and successive equity returns. Chen et al. (2019) document an

investigation of the impact of investor sentiments on the probability of offering seasoned stocks and stock price performance 

of these stocks. Their study finds a positive impact on the probability of offering seasoned equities and shows the short-run

price falls during periods of high investor sentiments. 

Extending the arguments beyond stock returns, French and Li (2017) examined the impact of investor sentiments on 

Thailand’s equity market returns and foreign equity flows. They find a strong negative effect of global risk aversion on for-

eign investment flows in Thailand, whereas sentiments positively impact foreign equity flows and is a significant variable 

for forecasting returns in the markets in Thailand. Using investor sentiments created from the first principal component 

of the consumer confidence index, advance/decline ratio, and volatility premium, Vuong and Suzuki (2020) suggested that 

the market sentiments can be a valid predictor of stock returns in the short-term horizons. Additionally, by decompos- 

ing the total sentiment in each market into regional and local indices, Vuong and Suzuki (2020) find that the market-

level results are driven mostly by the local sentiments. Some studies ( Berger and Turtle, 2012 ; Hirshleifer et al., 2006 )

consider the sensitivity of stock returns to the market sentiments. Yang and Hu (2021) show that sensitivity of stock re-

turns in China which changes with respect to individual stock sentiments. By supporting previous studies, their research 

shows that the effect of the beta of individual stock sentiment on stock returns is positive and significant in different stock

markets. 

The abovementioned literature can be recounted as non-exhaustive in the investor sentiment- stock returns controversy. 

The current study further investigates the impacts of sentiments considering sectoral stocks in Australia. Interestingly, it 

aims to fill this gap in the literature by examining the predictability of the sentiments on the sectoral stocks by making

use of the nonlinear causality-in-quantile method proposed by Balcilar et al. (2016) . This paper differs from the existing

literature, as it examines the age-varying effect of the consumer sentiments on returns, using the causality-in-quantile and 

nonlinear causality tests, which predicts a near accurate relationship of series. 
447 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3. Empirical methodology: Causality- in- quantiles test 

We investigate the causality-in-quantiles between the components of consumer sentiments (( y t ) and aggregate sectoral 

stock returns ( x t ) in Australia, using the unique nonlinear causality-in-quantile approach proposed by Balcilar et al. (2016) .

Following Jeong et al. (2012) , 3 we test that x t does not cause y t in the θ-quantile with regards to the lag-vector of

{ y t−1 , . . . , y t−p , x t−1 , . . . , x t−p } if 
Q θ ( y t | y t−1 , . . . , y t−p , x t−1 , . . . , x t−p ) = Q θ ( y t | y t−1 , . . . , y t−p ) (1) 

However, x t presumably causes y t in the θ-th quantile with regards to { y t−1 , . . . , y t−p , x t−1 , . . . , x t−p } if 
Q θ ( y t | y t−1 , . . . , y t−p , x t−1 , . . . , x t−p ) � = Q θ ( y t | y t−1 , . . . , y t−p ) (2) 

where Q θ ( y t |·) is the θ-th quantile of y t . The conditional quantiles of y t , Q θ ( y t |·) depend on t, and the quantiles are limited

to a range of zero to one, i.e., 0 < θ < 1 . 

Let us express the vectors Y t−1 ≡ ( y t−1 , . . . , y t−p ) , X t−1 ≡ ( x t−1 , . . . , x t−p ) , and Z t = ( X t , Y t ) . The functions F y t | Z t−1 
( y t | Z t−1 )

and F y t | Y t−1 
( y t | Y t−1 ) are the conditional distribution functions of y t conditioned on vectors Z t−1 and Y t−1 , respectively. The 

conditional distribution F y t | Z t−1 
( y t | Z t−1 ) is presumed to be completely continuous in y t for nearly all Z t−1 . By defining 

Q θ ( Z t−1 ) ≡ Q θ ( y t | Z t−1 ) and Q θ ( Y t−1 ) ≡ Q θ ( y t | Y t−1 ) , we can see that 

F y t | Z t−1 
{ Q θ ( Z t−1 ) | Z t−1 } = θ, 

which is true with a probability of one. As a result, based on Eqs. (1) and (2) , the causality-in-quantile hypothesis can be

expressed as 

H 0 : P 
{

F y t | Z t−1 
{ Q θ ( Y t−1 ) | Z t−1 } = θ

}
= 1 (3) 

H 1 : P 
{

F y t | Z t−1 
{ Q θ ( Y t−1 ) | Z t−1 } = θ

}
< 1 (4) 

A causality in a fat tail is different from that at the center of the distribution ( Lee and Cheng, 2007 ) . 4 

4. Data and descriptive statistics 

To investigate the distributional predictability between the Consumer Sentiments Index (CSI) and the sectoral stock re- 

turns of Australia in order to ascertain the perceptions of economic agents on stock returns, we obtain monthly indices of

Australia’s aggregate Consumer Sentiments index and components of the aggregate index which includes CSI Rural Australia, 

CSI-Aged 18–24; CSI – Aged 25 to 44, and CSI–Aged 45 and above. For our measure of sectoral stock returns, we consider the

following nine sectors: Health Care, Financials, Real Estate, Consumer Discretionary, Consumer Staples, Utilities, Industrials, 

Basic Materials and Energy. Our data obtained from Datastream spans the period from October 1974 to October 2020. 

Table 1 reports the statistical properties of the monthly returns for the nine sectoral stock price indices and our five

sentiment measures of CSI: CSI- Overall, CSI Rural Australia, CSI Aged 18–24; CSI – Aged 25 to 44, and CSI – Aged 45 and

above. Comparing the mean returns of the nine sectoral stocks, we find that the daily mean returns of all the stock sectors

are positive, with the Consumer Discretionary sector recording the highest mean return, which is equal to 0.921, followed 

by the Health Care sector (0.901). The least mean return which is equal to 0.601 was recorded by the financial sector. We

note that the mean of all the sentiment indices is positive with the exception of CSI Aged 18–24, which recorded a negative

mean of −0.001. Concerning the standard deviation of the sectoral stocks and the consumer sentiment indices, the results 

illustrates a considerable number of fluctuations in all series under examination, even though the extent of volatility and 

variability is low for all the series. 

On skewness, we see a negative skewness for all the series except that of CSI Rural Australia, which is mostly affected

by conditions in urban Australia. Negative (positive) skewness implies the tendency of higher negative (positive) returns 

without matching the tendency of positive (negative) returns. For kurtosis, all the series examined in this study recorded 

a kurtosis which exceeds the threshold 3 which is for the normal distribution. This surmises that the return series for the

period have flatter tails, compared to what one could expect from a normally distributed series. We use the Jarque-Bera (JB)

test to check for the normality assumption in the series. We reject the null hypothesis of normality for all the series under

investigation at the 1% level. We use Dickey and Fuller’s (1979) ADF test, Philips and Parron’s PP test, the KPSS stationarity

test and the ZA tests to test for the stationarity/nonstationarity of all the variables examined. Our results show the series

are integrated of order one i.e., they are log first difference stationary. 

Fig. 1 shows the graphical outlook of the overall distribution of the data together as well as the pairwise correlations

between sectoral stock returns and consumer sentiment indices. Once again, Fig. 1 confirms that the data used in this

investigation is normally distributed. We focus our analysis here on the correlation between the nine sectors and our overall 

measure of consumer sentiments. The highest correlation between the Overall CSI for Australia and any of the sectoral stock 
3 The exposition in this section closely follows Nishiyama et al. (2011) and Jeong et al. (2012) . 
4 We have made use of the bandwidth h = 0.05 and the kernel type used both for ( •) K and ( •) L is Gaussian. 

448 



A
.K

.
 T

iw
a

ri,
 E

.J.A
.
 A

b
a

k
a

h
,
 C

.O
.
 B

o
n

su
 et

 a
l.
 

Jo
u

rn
a

l
 o

f
 E

co
n

o
m

ic
 B

eh
a

v
io

r
 a

n
d
 O

rg
a

n
iza

tio
n
 19

3
 (2

0
2

2
)
 4

4
3

–
4

7
2
 

Table 1 

Summary statistics. 

Mean Median Max Min Std.Dev. Skewness Kurtosis Jarque-Bera ADF PP KPSS ZA Obs. 

Health. Care 0.901 1.028 17.519 −44.726 5.077 −1.367 ∗ 14.493 ∗ 3215.9 ∗ −7.29022 −524.784 0.084436 −8.152 ∗ 553 

Financials 0.601 1.089 26.101 −40.032 5.499 −1.008 ∗ 11.489 ∗ 1754.1 ∗ −8.12039 −536.455 0.401358 −7.575 ∗ 553 

Real. Estate 0.623 0.864 27.435 −47.517 5.672 −1.752 ∗ 17.337 ∗ 5018.8 ∗ −7.47508 −551.809 0.500582 −6.564 ∗ 553 

Consumer. disk. 0.921 1.319 28.091 −61.623 7.036 −1.399 ∗ 15.663 ∗ 3875.2 ∗ −7.01359 −494.566 0.547212 −6.541 ∗ 553 

Consumer. Staples 0.538 0.911 22.918 −58.068 7.295 −1.357 ∗ 12.605 ∗ 2295.3 ∗ −7.24875 −618.75 0.292138 −7.110 ∗ 553 

Industrials 0.609 1.287 18.954 −38.457 5.57 −0.969 ∗ 8.364 ∗ 749.3 ∗ −7.39916 −496.16 0.378097 −7.461 ∗ 553 

Basic. Materials 0.681 1.108 19.378 −45.291 6.25 −0.876 ∗ 8.633 ∗ 801.9 ∗ −7.72626 −525.624 0.102599 −7.309 ∗ 553 

Energy 0.612 0.689 23.498 −44.038 7.137 −0.703 ∗ 6.690 ∗ 359.3 ∗ −7.7247 −508.861 0.288453 −6.999 ∗ 553 

Utilities 0.684 0.659 32.788 −49.644 6.977 −0.24823 10.554 ∗ 1320.5 ∗ −7.66389 −512.329 0.268958 −7.049 ∗ 553 

CSI.OVERALL 0.027 0.087 16.551 −19.512 5.218 −0.331 ∗ 4.251 ∗ 46.1 ∗ −9.27422 −556.953 0.015123 −8.565 ∗ 553 

CSI.RURAL 0.042 0 27.793 −20.289 5.99 0.094709 4.310 ∗ 40.4 ∗ −8.85577 −597.787 0.017798 −8.586 ∗ 553 

CSI.AGED18TO24 −0.001 0.018 62.421 −48.221 8.179 0.277 ∗ 11.411 ∗ 1637.3 ∗ −10.1585 −590.013 0.015157 −8.829 ∗ 553 

CSI.AGED25TO44 0.038 0.078 19.841 −25.44 6.396 −0.350 ∗ 4.394 ∗ 56.1 ∗ −9.23014 −623.293 0.015074 −8.705 ∗ 553 

CSI.AGED45andOVER 0.035 −0.01 19.268 −19.468 5.828 −0.06685 3.696 ∗ 11.6 ∗ −9.10355 −558.813 0.015638 −8.887 ∗ 553 

Notes . The table reports the summary statistics for the indices examined in this study. Std.Dev denotes the standard deviation. JB denotes the Jarque-Bera test for normality. 
∗ denotes significance at the 1% level. 

4
4

9
 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Fig. 1. Causality-in-quantile between Components of Consumer Sentiments Index Australia and Health Care Sector market returns. Notes : The figure plots 

the estimates of the nonparametric causality test statistics (y-axis) corresponding to various quantiles (x-axis). The dark horizontal yellow and gray lines 

represent the 10% and 5% critical values, respectively. 

450 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

sectors is seen for the case of the Overall CSI and Real Estate sector which is equivalent to 0.22. Focusing on the components

of Overall CSI and the sectors under examination, we observe the highest significant correlation between CSI Rural Australia 

and the various sectors is seen for the case of CSI Rural Australia and the Basic Materials sector (0.14). For the remaining

sentiment indices, we find a significant correlation between the CSI Aged 18–24 and Industrials (0.12); the CSI Aged 25–44 

and Real Estate (0.22) and the CSI Aged 44 and above and Financials (0.18) and Basic Materials (0.18). 

The low pairwise correlation displayed in Table 1.1 between the various sectors and the sentiment indices is not sur- 

prising since the sectoral stock returns are impacted by several factors within the broader economy. However, we argue 

that this lays a solid foundation to do further tests, using robust methodologies to examine the extent of the distributional

predictability among the series. 

5. Discussion of the empirical results 

5.1. Linear and nonlinear causality tests 

The tests for causality provide a mechanism of testing for the dependence between the measures of sentiment and the 

returns of stocks under consideration in Australia. Table 2 describes the tests of the linear and nonlinear causality tests be-

tween the aggregate/overall CSI Australia and the nine sectoral stocks and vice versa. Schmeling (2009) reports that there is a

two-way causality in which the sentiments depend on the lagged returns, and the returns depend on the lagged sentiments. 

Focusing on the causality between the aggregate CSI measure and the sector returns, we find no statistical dependency for 

many of the relationship. However, three directional dependencies are identified. First, from the linear causality tests, we 

find that changes in the overall consumer sentiments influence the health care sector returns, financial and industrials sec- 

tors. Second, most importantly the result from the nonlinear tests shows that the stock returns of the health care sector and

financial sector are influenced by the changes in the overall sentiments of consumers in Australia, but the reverse causality 

is not true. 

We find that most of the sectoral stock returns are not dependent on the overall consumer sentiments, which sug- 

gests that the aggregate CSI is not an important measure driving returns. This initial finding agrees with the findings of

Salmes (2017) who established that the consumer sentiments in Australia have a direct or an undeviating effect on the re-

turns of certain sectors including the financials and industrials sectors where the valuations of firms in these sectors are 

highly subjective and not easy to arbitrage. 

Next, we focus on the causal relationship between the disaggregate components of Australia’s overall CSI and the returns 

of sectoral stocks. Table 3 reports the results for the case of the causality between the Rural Australia Consumer Sentiments

and the sectoral stocks returns. We observe from the linear causality test that the sentiments of consumers in the rural

Australia cause returns of the health care, financial, real estate and industrial sectors. For the non-linear causality tests, we 

observe that the CSI Rural Australia granger causes the Energy, Utilities and Basic Materials sectors, in addition to the Health

Care, Financial, and Real Estate sectors, a result that is similar to what we obtained from the linear causality tests. 

On the reverse causality test, we find a bidirectional causality only for the case of the CSI Rural Australia and the In-

dustrials sector under the test of linear causality, and the case of the CSI Rural Australia and the Real Estate and Industrials

sectors, using the nonlinear tests. Thus, changes in the returns of the industrial sector significantly influence the sentiments 

of the consumers across rural Australia. This result is not surprising for some varying reasons. For example, the spillover

effects of activities of the firms in the industrial sector, which include automobile, clothing, food services and mining firms, 

can extend themselves to the rural settings. This is because the consumers in the rural settings also depend on products

from the industrial sector. 

Table 4 depicts the linear and non-linear causalities for the relationship between the Consumer Sentiments for the age 

group 18–24 and the nine sectors. Interestingly, we find that variations in sentiments for this age group do not affect the

returns of any of the nine sectors examined in this study. This outcome can be attributed to the fact that this age group

does not have much money to invest in stocks. However, for the case of the reverse causality, we observe that the returns of

the following sectors: health care, financial, real estate, industrials, energy and utilities impact the perceptions and feelings 

of consumers in Australia within the age bracket of 18–24 years for both linear and non-linear causality tests. This result is

interesting since the behavioral theorists suggest that this age of the people in this bracket is the time of the lifespan when

very little is normative. It is a period of frequent changes and explorations that covers many aspects of the life of this group

from home, family, work, school, resources, and roles in society ( Johnson and Naka, 2014 ). Hence, it is not surprising that

variations in returns of several sectors affect the sentiments of this age group. This result is a unique contribution to the

literature since this paper is the first to examine sentiments of diverse age groups and stock prices. 

Table 5 presents the causality test for the case of the sentiments for the age group 25–44 and the nine sectors. We find

a bidirectional relationship between the sentiments for this age group and the real estate, industrials, basic materials, and 

utilities sectors. It is surprising that we find no dependency between the sentiments of this age group and the variations

in stock prices for sectors such as the financial, energy and utilities. This is perhaps because this age group constitutes the

adult consumers who are active in the labor force with varied needs spanning from health to energy to financing, among

other sectors. Hence, it is expected that changes in sectors such as the financial, energy and utilities will affect the overall

perceptions and sentiments of individuals within this age group. Our findings are not in agreement with the conclusion of 

Salmes (2017) . 
451 



A
.K

.
 T

iw
a

ri,
 E

.J.A
.
 A

b
a

k
a

h
,
 C

.O
.
 B

o
n

su
 et

 a
l.
 

Jo
u

rn
a

l
 o

f
 E

co
n

o
m

ic
 B

eh
a

v
io

r
 a

n
d
 O

rg
a

n
iza

tio
n
 19

3
 (2

0
2

2
)
 4

4
3

–
4

7
2
 

Table 1.1 

Pairwise correlations. 

Health 

Care 

Financials Real 

Estate 

Consumer 

Disk 

Consumer 

Staples 

Industrials Basic 

Materials 

Energy Utilities CSI 

Australia 

CSI Rural 

Australia 

CSI AGED 

18 to 24 

CSI AGED 

25 to 44 

CSI AGED 45 

and Over 

Health Care 1 

Financials 0.742 1 

Real Estate 0.587 0.903 1 

Consumer Disk 0.733 0.968 0.910 1 

Consumer Staples 0.538 0.675 0.711 0.731 1 

Industrials 0.766 0.869 0.875 0.873 0.887 1 

Basic Materials 0.704 0.865 0.701 0.845 0.601 0.732 1 

Energy 0.680 0.904 0.722 0.874 0.643 0.759 0.969 1 

Utilities 0.739 0.968 0.843 0.954 0.696 0.843 0.916 0.942 1 

CSI Australia −0.073 0.166 0.188 0.179 −0.019 0.005 0.086 0.086 0.147 1 

CSI Rural Australia −0.069 0.138 0.140 0.144 −0.087 −0.048 0.073 0.062 0.119 0.961 1 

CSI AGED 18 to 24 0.090 0.357 0.368 0.378 0.205 0.223 0.266 0.280 0.341 0.841 0.769 1 

CSI AGED 25 to 44 −0.018 0.153 0.173 0.176 0.002 0.025 0.078 0.072 0.144 0.975 0.934 0.797 1 

CSI AGED 45 and Over −0.054 0.228 0.219 0.223 −0.019 0.027 0.156 0.163 0.203 0.966 0.942 0.768 0.905 1 

4
5

2
 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Table 2 

Linear and nonlinear causality: Overall CSI Australia and sectoral stocks returns. 

H0: non causality Linear Non-Linear 

H1: causality Causality Test Causality Tests 

Sectoral Stocks Linear Granger General 

Taylor-based 

Semi-additive 

Taylor-based 

P-General Taylor 

based 

ANN-based 

H0: Overall CSI does not cause sectoral stock returns 

Health Care 2.0696 [0.0677] ∗∗∗ 1.8841 [0.0525] ∗∗∗ 0.8903 [0.5338] 2.0696 [0.0677] ∗∗∗ 0.9671 [0.4667] 

Financials 2.5431 [0.0274] ∗∗ 1.1359 [0.3287] 1.1056 [0.3572] 2.5431 [0.0274] ∗∗ 1.5363 [0.1326] 

Real Estate 1.5867 [0.1620] 4.1866 [0.0000] ∗ 1.7314 [0.0797] ∗∗∗ 1.5867 [0.1620] 0.9044 [0.5212] 

Consumer Disk 0.7121 [0.6145] 1.3179 [0.2399] 0.6366 [0.7659] 0.7121 [0.6145] 0.6177 [0.7822] 

Consumer Staples 1.4502 [0.2046] 2.3696 [0.0290] ∗∗ 0.8346 [0.5845] 1.4502 [0.2046] 0.8333 [0.5857] 

Industrials 4.4663 [0.0005] ∗ 2.2794 [0.0033] ∗ 2.0204 [0.0357] ∗∗ 4.4663 [0.0005] ∗ 2.2302 [0.0193] ∗∗

Basic Materials 1.7702 [0.1171] 1.3414 [0.1855] 1.0631 [0.3890] 1.7702 [0.1171] 1.0065 [0.4338] 

Energy 0.7080 [0.6176] 0.9690 [0.4778] 0.7412 [0.6711] 0.7080 [0.6176] 0.4281 [0.9200] 

Utilities 1.3064 [0.2596] 4.4594 [0.0000] ∗ 0.5658 [0.8252] 1.3064 [0.2596] 0.7820 [0.6332] 

H0: Sectoral stock returns do not cause Overall CSI-Australia 

Health Care 0.7306 [0.6007] 0.2084 [0.9589] 0.0793 [0.9886] 0.7306 [0.6007] 0.9716 [0.4628] 

Financials 0.5190 [0.7620] 0.1551 [0.9785] 0.1496 [0.9801] 0.5190 [0.7620] 0.2444 [0.9877] 

Real Estate 1.1940 [0.3107] 0.3089 [0.9076] 0.3686 [0.8701] 1.1940 [0.3107] 0.9379 [0.4918] 

Consumer Disk 1.6186 [0.1532] 0.1667 [0.9747] 0.1704 [0.9734] 1.6186 [0.1532] 0.5792 [0.8144] 

Consumer Staples 1.7711 [0.1170] 0.3141 [0.9045] 0.3364 [0.8909] 1.7711 [0.1170] 0.9621 [0.4709] 

Industrials 0.9770 [0.4311] 0.3960 [0.9046] 0.2531 [0.9382] 0.9770 [0.4311] 0.5682 [0.8233] 

Basic Materials 0.2764 [0.9261] 0.2414 [0.9439] 0.2736 [0.9275] 0.2764 [0.9261] 0.4740 [0.8921] 

Energy 0.9834 [0.4272] 0.2893 [0.9188] 0.3098 [0.9071] 0.9834 [0.4272] 0.6844 [0.7232] 

Utilities 1.5970 [0.1591] 0.2223 [0.9529] 0.2602 [0.9346] 1.5970 [0.1591] 1.3734 [0.1977] 

Notes . This table presents the linear and nonlinear granger causality tests between CSI and sectoral stock returns. The figures in the curry brackets are the 

p -values. ∗ , ∗∗ and ∗∗∗ denote significance at the 1%. 5% and 10% levels. 

Table 3 

Linear and nonlinear causality: CSI- Rural Australia and sectoral stocks returns. 

H0: non causality Linear Non-Linear 

H1: causality Causality Test Causality Tests 

Sectoral Stocks Linear Granger General 

Taylor-based 

Semi-additive 

Taylor-based 

P-General Taylor 

based 

ANN-based 

H0: CSI-Rural Australia does not cause sectoral stock returns 

Health Care 3.0788 [0.0095] ∗ 2.2340 [0.0078] ∗ 1.3170 [0.2255] 3.0788 [0.0095] ∗ 1.5309 [0.1344] 

Financials 3.0384 [0.0103] ∗∗ 1.3212 [0.1797] 1.3463 [0.2107] 3.0384 [0.0103] ∗ 1.8784 [0.0533] ∗∗

Real Estate 2.2054 [0.0524] ∗∗∗ 2.9399 [0.0004] ∗ 1.6783 [0.0919] ∗∗∗ 2.2054 [0.0524] ∗∗∗ 1.2288 [0.2752] 

Consumer Disk 0.8134 [0.5403] 1.5732 [0.1207] 0.6566 [0.7482] 0.8134 [0.5403] 0.5312 [0.8519] 

Consumer Staples 1.2906 [0.2664] 1.5363 [0.1237] 0.7596 [0.6540] 1.2906 [0.2664] 0.5377 [0.8471] 

Industrials 5.3026 [0.0001] ∗ 2.5443 [0.0003] ∗ 2.0762 [0.0304] ∗∗ 5.3026 [0.0001] ∗ 2.4000 [0.0116] ∗∗

Basic Materials 1.8416 [0.1030] 1.6905 [0.0416] ∗∗ 1.3360 [0.2158] 1.8416 [0.1030] 0.8133 [0.6042] 

Energy 0.9528 [0.4463] 1.6202 [0.0603] ∗∗ 0.7486 [0.6643] 0.9528 [0.4463] 0.5491 [0.8384] 

Utilities 1.5323 [0.1779] 3.2125 [0.0000] ∗ 0.5962 [0.8004] 1.5323 [0.1779] 0.8940 [0.5305] 

H0: Sectoral stock returns do not cause CSI-Rural Australia 

Health Care 0.7676 [0.5734] 0.6620 [0.6525] 0.6969 [0.5944] 0.7676 [0.5734] 0.9418 [0.4883] 

Financials 1.7092 [0.1306] 0.5394 [0.7464] 0.6071 [0.6946] 0.5394 [0.7464] 0.8482 [0.5720] 

Real Estate 2.2342 [0.0497] ∗∗ 0.9657 [0.4384] 1.1301 [0.3435] 0.9657 [0.4384] 1.8248 [0.0619] ∗∗∗

Consumer Disk 1.6905 [0.1350] 0.6897 [0.6314] 0.7109 [0.6155] 1.6905 [0.1350] 0.8772 [0.5456] 

Consumer Staples 2.2566 [0.0476] ∗∗ 1.1241 [0.3467] 1.1591 [0.3285] 2.2566 [0.0476] ∗∗ 1.5564 [0.1261] 

Industrials 2.7724 [0.0175] ∗∗ 1.4649 [0.1585] 1.2531 [0.2833] 2.7724 [0.0175] ∗∗ 1.6823 [0.0909] ∗∗∗

Basic Materials 0.7208 [0.6080] 0.7944 [0.5541] 0.8825 [0.4926] 0.7208 [0.6080] 0.5096 [0.8677] 

Energy 1.5859 [0.1622] 0.5982 [0.7319] 0.7458 [0.5896] 1.5859 [0.1622] 0.6794 [0.7278] 

Utilities 3.0176 [0.0107] ∗∗ 0.6763 [0.6416] 0.7005 [0.6233] 3.0176 [0.0107] ∗∗ 1.6864 [0.0899] ∗∗∗

Notes . This table presents linear and nonlinear granger causality tests between CSI and sectoral stock returns. The figures in the curry brackets are the 

p-value s. ∗ , ∗∗ and ∗∗∗ denote significance at the 1%. 5% and 10% levels. 

 

 

 

 

Table 6 denotes both the linearity and nonlinearity tests of causality for the sentiments for the age group 45 years and

above and the nine sectors. We observe some interesting results in this table, demonstrating that perceptions and sentiments 

of elderly consumers granger cause the returns of all nine sectors under examination. However, the reverse is not true. This

is an insightful finding but is not surprising because the individuals in this age group arguably are more concerned about

wealth. Otoo (1999) concludes that stock returns can increase consumer sentiments since the stock returns bring wealth 

which boosts consumer confidence in the economy. In addition, Otoo (1999) asserts that changes in wealth including the 

wealth from the stock markets are likely to affect sentiments less than changes in income because changes in wealth affect

consumption much less than changes in income. 
453 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Table 4 

Linear and Nonlinear causality: CSI – AGED 18 TO 24 and Sectoral Stocks Returns. 

H0: non causality Linear Non-Linear 

H1: causality Causality Test Causality Tests 

Sectoral Stocks Linear Granger General 

Taylor-based 

Semi-additive 

Taylor-based 

P-General Taylor 

based 

ANN-based 

H0: CSI –Aged 18 to 24 does not cause sectoral stock returns 

Health Care 0.7271 [0.6033] 0.7116 [0.5843] 0.7105 [0.5851] 0.7271 [0.6033] 1.1763 [0.3083] 

Financials 0.8179 [0.5372] 0.3598 [0.8372] 0.3596 [0.8373] 0.8179 [0.5372] 0.9967 [0.4419] 

Real Estate 0.8962 [0.4832] 0.0937 [0.9844] 0.0914 [0.9851] 0.8962 [0.4832] 0.8354 [0.5837] 

Consumer Disk 1.1195 [0.3488] 0.3248 [0.8614] 0.3140 [0.8686] 1.1195 [0.3488] 0.8654 [0.5563] 

Consumer Staples 1.2600 [0.2798] 0.0285 [0.9984] 0.0290 [0.9984] 1.2600 [0.2798] 0.7606 [0.6531] 

Industrials 1.0201 [0.4049] 0.2543 [0.9070] 0.2535 [0.9075] 1.0201 [0.4049] 0.9112 [0.5152] 

Basic Materials 0.7162 [0.6115] 1.1505 [0.3322] 1.1291 [0.3421] 0.7162 [0.6115] 0.8229 [0.5952] 

Energy 0.7966 [0.5524] 0.4942 [0.7400] 0.4875 [0.7449] 0.7966 [0.5524] 0.6592 [0.7459] 

Utilities 0.6683 [0.6477] 0.4452 [0.7759] 0.4381 [0.7811] 0.6683 [0.6477] 0.7013 [0.7078] 

H0: Sectoral stock returns do not cause CSI – Aged 18 to 24 

Health Care 2.0765 [0.0668] ∗∗∗ 1.7441 [0.0970] ∗∗∗ 0.1503 [0.9628] 2.0765 [0.0668] ∗∗∗ 0.8017 [0.6149] 

Financials 4.0015 [0.0014] ∗ 7.0641 [0.0000] ∗ 0.0832 [0.9948] 4.0015 [0.0014] ∗ 0.5124 [0.8658] 

Real Estate 2.3430 [0.0404] ∗∗ 3.2360 [0.0014] ∗ 0.1080 [0.9905] 2.3430 [0.0404] ∗∗ 0.3637 [0.9518] 

Consumer Disk 1.2514 [0.2837] 0.1507 [0.9798] 0.1434 [0.9819] 1.2514 [0.2837] 0.5421 [0.8438] 

Consumer Staples 1.0572 [0.3832] 0.2640 [0.9326] 0.2741 [0.9272] 1.0572 [0.3832] 0.9197 [0.5077] 

Industrials 2.5380 [0.0277] ∗∗ 5.7848 [0.0000] ∗ 0.1779 [0.9708] 2.5380 [0.0277] ∗∗ 0.5850 [0.8096] 

Basic Materials 2.7183 [0.0194] ∗∗ 7.0490 [0.0000] ∗ 0.2326 [0.9481] 2.7183 [0.0194] ∗∗ 1.5429 [0.1304] 

Energy 0.4928 [0.7817] 3.7805 [0.0000] ∗ 0.0804 [0.9952] 0.4928 [0.7817] 0.9114 [0.5150] 

Utilities 2.0040 [0.0765] ∗∗∗ 1.8066 [0.0739] ∗∗∗ 0.1708 [0.9733] 2.0040 [0.0765] ∗∗∗ 1.2719 [0.2499] 

Notes . This table presents linear and nonlinear granger causality test between CSI and sectoral stock returns. The figures in the curry brackets are the 

p-value s. ∗ , ∗∗ and ∗∗∗ denote significance at the 1%. 5% and 10% levels. 

Table 5 

Linear and Nonlinear causality: CSI – AGED 25 TO 44 and Sectoral Stocks Returns. 

H0: non causality Linear Non-Linear 

H1: causality Causality Test Causality Tests 

Sectoral Stocks Linear Granger General 

Taylor-based 

Semi-additive 

Taylor-based 

P-General Taylor 

based 

ANN-based 

H0: CSI –Aged 25 to 44 does not cause sectoral stock returns 

Health Care 1.4797 [0.1946] 1.4226 [0.1847] 1.0214 [0.4189] 1.4797 [0.1946] 0.7968 [0.6194] 

Financials 1.5046 [0.1865] 1.2585 [0.2518] 1.3653 [0.2097] 1.5046 [0.1865] 0.9351 [0.4941] 

Real Estate 0.8927 [0.4856] 4.1548 [0.0000] 2.0939 [0.0352] 0.8927 [0.4856] 0.8588 [0.5623] 

Consumer Disk 0.3028 [0.9112] 1.1448 [0.3359] 0.6950 [0.6961] 0.3028 [0.9112] 0.4737 [0.8923] 

Consumer Staples 0.7554 [0.5823] 2.7673 [0.0179] 0.5459 [0.8218] 0.7554 [0.5823] 0.6986 [0.7103] 

Industrials 2.6363 [0.0228] 2.4704 [0.0030] 1.8983 [0.0586] 2.6363 [0.0228] 1.7742 [0.0710] 

Basic Materials 2.7654 [0.0177] 1.6872 [0.0736] 1.8308 [0.0696] 2.7654 [0.0177] 1.7920 [0.0677] 

Energy 0.9193 [0.4679] 0.9375 [0.5039] 0.5885 [0.7875] 0.9193 [0.4679] 1.0333 [0.4122] 

Utilities 0.4423 [0.8189] 2.9947 [0.0028] 0.4367 [0.8989] 0.4423 [0.8189] 0.2457 [0.9874] 

H0: Sectoral stock returns do not cause CSI – Aged 25 to 44 

Health Care 1.3561 [0.2393] 0.1127 [0.9895] 0.0796 [0.9886] 1.3561 [0.2393] 0.9727 [0.4620] 

Financials 0.7642 [0.5759] 0.2460 [0.9417] 0.1208 [0.9877] 0.7642 [0.5759] 0.2938 [0.9764] 

Real Estate 1.8356 [0.1041] 0.2342 [0.9474] 0.2108 [0.9579] 1.8356 [0.1041] 1.0871 [0.3709] 

Consumer Disk 1.5204 [0.1816] 0.0640 [0.9972] 0.0662 [0.9970] 1.5204 [0.1816] 0.6231 [0.7775] 

Consumer Staples 1.3672 [0.2350] 0.2688 [0.9301] 0.3226 [0.8994] 1.3672 [0.2350] 0.9740 [0.4608] 

Industrials 1.2856 [0.2685] 0.7420 [0.6364] 0.0894 [0.9939] 1.2856 [0.2685] 0.6429 [0.7603] 

Basic Materials 0.4089 [0.8427] 0.0378 [0.9992] 0.0463 [0.9987] 0.4089 [0.8427] 0.5582 [0.8312] 

Energy 1.1798 [0.3178] 0.2588 [0.9353] 0.1523 [0.9793] 1.1798 [0.3178] 1.0290 [0.4157] 

Utilities 1.7857 [0.1139] 0.1134 [0.9894] 0.1768 [0.9712] 1.7857 [0.1139] 1.2381 [0.2696] 

Notes . This table presents linear and nonlinear granger causality test between CSI and sectoral stock returns. Figures in curry brackets are p-value s. ∗ , ∗∗

and ∗∗∗ denote significance at the 1%. 5% and 10% levels. 

 

 

5.2. Causality- in-quantile tests 

Post to the establishment of the unstable nature of the causality parameters of the linear causality, as well as the exis-

tence of the non-linearity causality tests in the relationship between the Australian consumer sentiments and sectoral stock 

returns, we extend our analysis with the application of the unique causality-in-quantile methodology of Balcilar et al. (2016) ,

deemed as robust to the aforesaid economic issues under examination. 
454 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Table 6 

Linear and Nonlinear causality: CSI – AGED 45 and above Sectoral Stocks Returns. 

H0: non causality Linear Non-Linear 

H1: causality Causality Test Causality Tests 

Sectoral Stocks Linear Granger General 

Taylor-based 

Semi-additive 

Taylor-based 

P-General Taylor 

based 

ANN-based 

H0: CSI –Aged 45 and above does not cause sectoral stock returns 

Health Care 1.3176 [0.2549] 1.8523 [0.0500] 1.1965 [0.2952] 1.3176 [0.2549] 0.7900 [0.6258] 

Financials 2.7738 [0.0174] 1.4985 [0.1017] 2.2486 [0.0183] 2.7738 [0.0174] 2.1729 [0.0229] 

Real Estate 2.6706 [0.0213] 5.0062 [0.0000] 3.6436 [0.0002] 2.6706 [0.0213] 1.6113 [0.1095] 

Consumer Disk 1.0789 [0.3710] 0.9014 [0.5151] 1.3385 [0.2146] 1.0789 [0.3710] 0.7856 [0.6299] 

Consumer Staples 2.3831 [0.0374] 2.3729 [0.0288] 1.0816 [0.3749] 2.3831 [0.0374] 1.4590 [0.1608] 

Industrials 5.9822 [0.0000] 3.1884 [0.0000] 3.8198 [0.0001] 5.9822 [0.0000] 3.2952 [0.0007] 

Basic Materials 1.3368 [0.2471] 1.3684 [0.1649] 0.9691 [0.4650] 1.3368 [0.2471] 0.7883 [0.6273] 

Energy 1.0994 [0.3596] 2.0883 [0.0097] 0.7439 [0.6686] 1.0994 [0.3596] 1.2671 [0.2527] 

Utilities 2.4757 [0.0312] 4.8017 [0.0000] 2.2522 [0.0181] 2.4757 [0.0312] 1.6965 [0.0875] 

H0: Sectoral stock returns do not cause CSI – Aged 45 and above 

Health Care 0.4807 [0.7907] 0.1005 [0.9920] 0.0566 [0.9940] 0.4807 [0.7907] 0.6966 [0.7121] 

Financials 1.1447 [0.3355] 0.1940 [0.9648] 0.1915 [0.9657] 0.1940 [0.9648] 0.6643 [0.7414] 

Real Estate 1.3257 [0.2516] 0.3912 [0.8549] 0.4509 [0.8126] 1.3257 [0.2516] 1.5311 [0.1344] 

Consumer Disk 1.4322 [0.2109] 0.0892 [0.9939] 0.0896 [0.9939] 1.4322 [0.2109] 0.5776 [0.8157] 

Consumer Staples 1.8633 [0.0990] 0.1112 [0.9899] 0.1334 [0.9846] 1.8633 [0.0990] 0.9649 [0.4686] 

Industrials 0.5210 [0.7605] 0.5704 [0.8023] 0.2629 [0.9332] 0.5704 [0.8023] 0.4655 [0.8976] 

Basic Materials 0.1625 [0.9761] 0.1183 [0.9883] 0.1328 [0.9848] 0.1625 [0.9761] 0.3594 [0.9536] 

Energy 0.7939 [0.5544] 0.2604 [0.9344] 0.2676 [0.9307] 0.2604 [0.9344] 0.4947 [0.8782] 

Utilities 0.8817 [0.4930] 0.2626 [0.9333] 0.3465 [0.8845] 0.8817 [0.4930] 1.5009 [0.1450] 

Notes . This table presents linear and nonlinear granger causality test between CSI and sectoral stock returns. Figures in the curry brackets are the p-value s. 
∗ , ∗∗ and ∗∗∗ denote significance at the 1%. 5% and 10% level. 

 

 

 

 

 

 

 

 

 

 

 

 

The causality-in-mean (denoted in blue) and the causality-in-variance (denoted in green) are predicted in a time-varying 

context. This is conditioned on the states of the markets represented by the corresponding quantiles found in the distribu- 

tion of the dependent variable. Grounded on the distribution quantiles, the markets have been classified as normal (aver- 

age market), bullish (good market) and bearish (bad market). These classifications correspond to the 50th (average), 90th 

(higher) and 10th (lower) quantiles, respectively. The quantiles are presented on the x-axis and the nonparametric causal- 

ity test statistics are shown on the y-axis. The dark horizontal yellow line represents the 10% critical values, and the gray

lines are representative of the 5% critical values. The center of the distribution focuses on the normal markets, whereas the

lower and higher extreme quantiles show the bearish and bullish markets as indicated. The nature of the causality curve 

is indicative of the strength of the causality and the degree of the predictableness that exist between the sentiments and

the sectoral stock returns across the quantiles which correspond to particular conditions of the market, be they bear, bull 

and normal market conditions. The finding replicates three varieties of asymmetry in causality. This is an affirmation of the 

choice of the methods, causality- in-quantile, over other existing and prior linear and non-linear models that are employed 

in the estimation of the average conditional causality. The hump-shaped curvature of the causality is indicative of a strong 

conditional causality at the average. Again, this shows that the strength of the causality fades or loses strength towards the

extreme quantiles considered far away from the mean or the median. The inverted hump -shaped curve suggests evidence 

of causality with an altered degree at the quantiles distant from the average. The left tail (which shows a hump towards the

right) and the right tail (showing a hump towards the left) are indicative of an asymmetric causality, which the causalities

of linearity and non-linearity do not capture. 

5.2.1. Relationship between consumer sentiments and the health care sector stock returns 

The results of the causality-in-quantile between the aggregate/overall and disaggregate components of the consumer 

sentiments index in Australia and the health care sector’s market returns are presented in Fig. 1 . These results show a clear

difference between the quantile causality in the mean and the variance analyses. The causality-in-mean is bidirectional with 

its presence seen only in the normal market conditions. This suggests a presence of a sturdy predictability around only the

normal market condition. Issue of no causation in the bearish and bullish market conditions implies a failure of consumer 

sentiments to lead the stock returns of the health care sector. Hence, for policy decisions in terms of investments and also

risk management, the quantile causality bears or provides more information in different market circumstances. Thus, this 

finding upholds that the sentiments are proficient predictors of the health care sector stock price as applied in the asset

pricing model around the normal market condition. 

The causality-in-variance is also bidirectional around the conditions pertaining to a normal market. The no causation in 

the bear market and bull market suggests the inability of the sentiments in playing their economic roles as a transmission

of risk and an effective predictor of the health care sector’s returns in the bad and good conditions of the market. The

causality-in-variance running from aggregate sentiments to the health care sector returns is indicative of a transfer of risk 

from sentiments to the health care sector market in the normal market condition. This suggests that sentiments lead the 
455 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

health care sector in the issue of the efficiency in information and the discovery of price in the condition of the market

deemed as normal but not in the bearish and bullish market conditions. The reverse is also true, running from the health

care sector to the overall sentiments in the normal market condition. 

Concerning the causation between the sentiments of rural Australia and the sectoral returns, we observe that the bidirec- 

tional results spilling over from the causality-in-mean and the causality-in -variance are similar to the findings we observed 

concerning the overall CSI Australia and the health care sector returns. Thus, in the normal market condition, the senti- 

ments of consumers dwelling in the rural Australia impact the price levels of the health care sector but with no evidence

of causation around the good and bad market conditions. Thus, the causality is not asymmetric, and that the potency of

this causality is strong only in the normal condition of the market, something the linear and nonlinear causality tests fail to

account for. 

The causation is bidirectional both the in mean and the in variance running from the sentiments of age group 18 to

24 to the healthcare sector. There are no causality- in variance in the lower and upper quantiles from the sentiments of

consumers within the age group of 18–24 to the health care sector indicates that sentiments of this age group is inefficient

in information in a bad market and a good market. Regarding the causality between the sentiments for the age group 25

to 44, we observe a similar outcome just like in the case of the relationship between the sentiments of age group 18–24

and the health care sector returns. However, for the causation between the sentiments of age group 45 and above, we find

a bidirectional causation for the causality- in- mean running from the bad market to the normal market, while for the

causality-in-variance, we observe a unidirectional relationship from the bearish market to the normal market conditions for 

sentiments to the health care sector. 

Overall, the findings align with the asset pricing theory that the sentiment is a systemic risk and an efficient predictor

of the stock returns in this case which is the health care sector. The observed causality seen in both the mean and variance

in the normal market condition supports the adoption and the usage of the non-parametric causality in quantile models 

instead of the linear and nonlinear granger causality tests. 

5.2.2. Relationship between consumer sentiments and the financial sector stock returns 

Fig. 2 presents the quantile causality between sentiments and the financial sector returns. We document some interesting 

findings. For the case of the overall CSI and financial sector returns, we note the causality-in-mean is bidirectional which 

occurs only in the bearish to normal markets, but with no causation in the market conditions seen as bullish. The causality-

in-variance is also bidirectional and left tailed, and the shaped (humped towards the right) causality-in-variance curvature 

indicates that the causality is not symmetric, implying that there exists a near accurate predictability around the normal 

conditions of the market and that the strength dwindles as we move towards the bullish market condition. The implication 

drawn is that the efficiency of information about the aggregate consumer sentiments in leading the financial sector markets 

falls in the extreme bullish condition. 

Focusing on the causation between the rural Australia sentiments and the financial sector stock returns, the findings 

show a one-directional causality running from CSI rural Australia to the financial sector around the bearish market to the 

normal market conditions. However, for the reverse causality, we find no evidence of causation exists under all the market 

conditions for both the causality-in-mean and in-variance. This connotes that variations in the financial sector’s stock prices 

do not impact the sentiments of consumers in rural Australia. This result is surprising since spillover effects from the finan-

cial sector impact the overall broader economy ( Otoo, 1999 ). Another possible reason for this results is that economic agents

in rural Australia may not be exposed to too much media coverage of the traumatic financial events/news in the financial

sector and as result their feelings are not impacted by what pertains to the financial sector. We make this claim because

Collimore et al. (2008) suggest that the exposure to media reportage on sensitive matters generates increasing levels of 

anxiety. Arguably, residents in rural Australia are likely not to be active users of Twitter and other social media platforms to

access adequate information concerning financial markets. 

As evidenced in Fig. 2 , a unidirectional causality-in-mean is seen from the financial sector returns to the sentiments of

individuals within the age group of 18–24, which runs from the normal market condition to the extreme bullish market 

situation. In the case of the causality-in-variance, there exists a significant causation from sentiments to the financial sector 

returns during the normal to bullish market conditions. However, for the reverse causality-in-variance, there exists a causa- 

tion at the lower quantiles. Thus, we find a transfer of risk from the financial sector to consumers within the aged bracket

of 18–24 years. From the CSI aged 25–44 to the financial sector stock returns, the causality-in-mean reveals a unidirectional 

causality during the normal market condition from the financial sector to the sentiments of individuals aged between 25 

and 44 years. 

In the case of the causality-in-variance, there exists a causality from sentiments to the financial sector. However, the 

reverse is untrue. Unlike the results obtained from the linear and nonlinear results reported earlier, there is no causation in

the causality-in-mean running from CSI aged 45 years and above, and the causality-in-variance from the financial sector to 

the sentiments of the aged group 45 years and above. However, we document a one-way directional causality-in- variance 

running from the financial sector to the sentiments of this aged group. This suggests that there is a transfer of risk from the

financial sector to the individuals within the age bracket of 45 years and above. This finding differs from what we recorded

by using the linear and nonlinear causality models, which supports the superiority of the quantile-in-causality test. 

Consistent with the existing literature, an increase in individual sentiments impacts the financial industry returns and 

volatility ( Brown and Cliff, 2004 ; Verma et al., 2008 ; Calafiore, 2010 ). Overall, in the financial sector we note that the impacts
456 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

 

 

 

of sentiments of consumers on this sector and vice versa differing in different quantiles. Thus, investors can rely on our

findings in their portfolio strategy formation since we observe strong predictability under the normal market condition and 

to some extent during the bearish market condition. 

5.2.3. Relationship between consumer sentiments and real estate sector returns 

The causality dynamics between sentiments and real estate sector returns are reported in Fig. 3 . For the causation run-

ning from the overall sentiment index and the real estate sector, the causality-in-mean is unidirectional and runs from 

the normal to the bullish market conditions. The stock price variations in the real estate sector are intense; hence, it is

not surprising that we find no causation during the bearish market conditions, which also holds for the reverse causality. 

One would have expected the reverse causality running from the real estate price variation to influence the sentiments of 

consumers. However, that is not the case for the Australian market. 

Regarding the causality-in-variance analysis, we find evidence of a bidirectional causation during the normal market 

conditions. This is in alignment with the existing studies which argue that risk transfer from the real estate sector impacts

household consumption. This further shows that the effects of price pressure (noise traders’ pressure on the stock prices 
Fig. 2. Causality-in-quantile between Components of Consumer Sentiments Index Australia and Financial Sector market returns. Notes : The figure plots 

the estimates of the nonparametric causality test statistics (y-axis) corresponding to various quantiles (x-axis). The dark horizontal yellow and gray lines 

represent the 10% and 5% critical values, respectively. 

457 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Fig. 2. Continued 

 

 

 

 

 

 

 

 

reduces the expected return) don’t dominate the markets and that noise traders don’t benefit from the market during the 

extreme market conditions where sentiments are high (Haritha & Rashid, 2020). 

Focusing on the causation between disaggregate measures of sentiments and the real estate sector, we find a bidirectional 

causation for the causality-in-mean as well as the causality-in-variance running from the rural Australia sentiment to the 

real estate sector during the bearish to normal and bullish market conditions. However, the extent of predictability declines 

during the good and bad markets conditions. Next, we find that there exists a unidirectional causation in the causality-in- 

mean for the case of the real estate stock returns to the sentiments of individuals within the age group of 18–24 years,

running from the normal to bullish market conditions. However, there is no causation under all market conditions for the 

causality running from the sentiments of aged group 18–24 years and the real estate sector returns. 

For the causality-in-variance from the sentiments of aged group 18–24 years to the real estate sector, the results show 

the sentiments lead the returns during the normal market conditions. However, once the noise traders start making profits, 

their expectations on the return and risk will increase. Thus, it may not create a reverse causality because of the information

inefficiency which is exactly what we have recorded since the reverse causality is not true for the causality-in-variance 

analysis. In the case of causation between the sentiments from age group 25–44 and the real estate sector returns, we

underscore the existence of a unidirectional causality-in-variance under all market conditions and a directional causality-in- 

mean during the normal market conditions. The absence of causality- in-variance in the bear market implies an inability 

of the sentiments in serving its economic role as a fair (transfer of risk) and efficient predictor (efficiency in informational

needs) of the real estate stock prices in an unfavorable market condition. We obtain similar findings for the aged group 45

and above and the real estate sector. Similarly, to what we obtained using the linear and nonlinear causality tests, this age

group seems to have a dominant impact on the several sectoral stocks since we again have not had a significant causation-

in-variance under all market conditions. In brief, the real estate sector is impacted significantly by noise traders under the 

normal market conditions with a significant reduction in the transfer of risk from the real estate sector to the sentiments

in the lower and upper quantiles. 

5.2.4. Relationship between consumer sentiments and the sector for consumer discretion 

The sector for consumer discretion involves businesses that trade nonessential products and services which consumers 

may avoid short of any major consequences to their well-being. The sector comprises numerous companies operating in a 

number of industries which include retail, restaurants, media, consumer durables and apparel. We present the results of 

the causality-in-quantile between the various components of the consumer sentiments index in Australia and the consumer 
458 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

 

 

 

discretionary returns in Fig. 4 . First, we find a causality-in-mean only exists around the extreme market conditions from 

the overall CSI to consumer discretionary sector, while for the causality-in-variance, the causality exists only in the normal 

market condition. Thus, during the extreme market conditions, consumers react to the prevailing factors in the sector. The 

demand for goods based on for consumer discretion is typically much more responsive to demand factors in comparison to 

the demand for consumer staple goods. The elasticity of demand implies that the demand for such goods can plummet very

quickly in response to falls in consumers’ incomes or rises in prices of consumer discretionary goods. For instance, during 

the COVID 19 pandemic lockdown, the decline in income of consumers due to the loss of jobs among other factors affected

the returns of firms within the sector. Hence, it is not surprising that we find evidence of a strong predictability only in the

normal market condition. 

For the relationship between the components of the aggregate sentiment index and the consumer discretionary sector, 

we note a unidirectional causality both in the mean and the variance from the CSI rural Australia to the consumer dis-

cretionary sector under all market conditions. There exist no causality-in-mean and in-variance for the sentiments from 

the age groups 18–24 and 25–44 to the consumer discretionary sector during the extreme market conditions, which is not 

surprising. Interestingly, the causality-in-mean and in-variance from the sentiments of the age group 45 and above to the 
Fig. 3. Causality-in-quantile between Components of Consumer Sentiments Index Australia and Real Estate Sector market returns. Notes : The figure plots 

the estimates of the nonparametric causality test statistics (y-axis) corresponding to various quantiles (x-axis). The dark horizontal yellow and gray lines 

represent the 10% and 5% critical values, respectively. 

459 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Fig. 3. Continued 

 

 

 

 

 

 

 

consumer discretionary sector is bidirectional for the causality -in-mean with asymmetric features. This signals the existence 

or presence of a robust predictability around the normal market settings and the strength falls as we approach the extreme

quantiles, that is, the 90th and 10th quantiles, which correspond to the bullish markets and bearish markets, respectively. 

5.2.5. Relationship between consumer sentiments and the consumer staples sector 

Consumer staples are deemed as essential products that consist of typical products such as foods & beverage, household 

goods, and hygiene commodities. Nevertheless, this category also includes items such as alcohol and tobacco. The staple 

nature of these goods implies that people are unable or unwilling to cut out of their budgets, regardless of their financial

situations. From Fig. 5 , we observe a bidirectional causation from the overall sentiments to the consumer staples sector 

both in the mean and in the variance from the bearish to the normal market conditions. Thus, during the bearish market

condition, sentiments of consumers increase until the market conditions stabilize. However, when the conditions are good 

in the market, we find no transfer of risk from the causality -in-variance analysis. For the causation between disaggregate 

measures of sentiments and consumer staples both in mean and in variance under the various market conditions, we ob- 

serve the causation varies depending on the prevailing market condition. It is noteworthy to mention that there exists a 

significant bidirectional causality under the bearish and normal market conditions for the various categories. 

5.2.6. Relationship between consumer sentiments and the industrial sector 

We show the results of the causality-in-quantile between the aggregate/overall and the disaggregate measures of the 

consumer sentiments index in Australia and the industrial sector market returns in Fig. 6 . This sector is one of the key

sectors that form the economy, and economists are of the opinion that this sector creates wealth unlike the service sector.

We argue that the relationship between sentiments and the stock returns in the industrial sector will exist under all market 

conditions, following the significant role this sector plays in the overall economy. For the causation between the overall 

sentiments and the industrial sector, the causality- in-mean is bidirectional and runs from the bearish market to the normal 

market conditions, although the predictability is strong around normal market conditions. There is no causation during 

bullish market conditions which suggests that consumers in Australia only react when conditions are bad in the industrial 

sector. This is not surprising because in cases where there is a shortage in a particular product that is essential, one would

expect consumer sentiments to increase. 

Regarding the causality in the variance analysis, we observe a unidirectional causation from the industrial stock returns 

to consumers sentiments under all market conditions. Interestingly, this result is in alignment with the notion that the 

industrial sector creates wealth; hence, the variations in the stock prices of this sector affect the feelings of consumers and
460 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Fig. 4. Causality-in-quantile between Components of Consumer Sentiments Index Australia and Consumer Disk. Sector market returns. Notes. The figure 

plots the estimates of the nonparametric causality test statistics (y-axis) corresponding to various quantiles (x-axis). The dark horizontal yellow and gray 

lines represent the 10% and 5% critical values, respectively. 

461 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Fig. 4. Continued 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

noise traders. From the causality-in-mean analysis, sentiments of consumers dwelling in the rural Australia are also seen to 

be a strong predictor of returns in the industrial sector all through the bearish and normal market conditions, which is not

surprising following the significance of this sector in the broader economy. 

Comparing the causation between the sentiments of individuals within the age groups of 18–24; 25–44, and 45 and 

above and the industrial stock returns, the result is interesting. For the age 18–24, the causality-in-mean is unidirectional 

only during market conditions seen as normal, while the reverse does not hold. Interestingly, for the adult group aged 

between 24 and 44, the relationship is bidirectional from the causality-in-mean analysis which is not news since the con- 

sumers in this bracket are more concerned about making wealth, and as a result, they closely follow this sector in the

bearish and normal market conditions. We document similar findings for the consumers within the age group of 45 and 

above. In addition, for the three age groups under examination, the causality-in-variance analysis shows that sentiments fail 

to act in its economic role as an unbiased and an efficient estimator of the industrial sector stock returns under the bearish

and bullish market conditions. 

5.2.7. Relationship between consumer sentiments and the basic materials sector 

The basic materials sector comprises companies involved in the discovery, extraction, and the processing of raw materials. 

The basic materials market is based on the products that make use of such materials. The sub-sectors found within the mar-

ket for the basic materials include chemicals, metals and mining, and forestry. This causality-in-quality result reported in 

Fig. 7 for this sector shows that there exists no causality between the overall sentiments of consumers in Australia and this

sector from the causality-in-mean point of view. This result is not surprising since the sector is not regarded as a significant

sector in the broader financial economy. We find the same results hold for the causation between the disaggregate measures 

of sentiments and the basic material under all market conditions. 

For the causality-in-variance, we find some level of a bidirectional causation from the normal to bull market conditions 

running from the overall sentiments and CSI Aged 18–24 to the basic material sector. One can argue that a variation in the

stock prices or activities of firms in this sector is not of great concern to consumers since the firms in this sector, such as

BHP among others, are very strong firms and can bounce back to the normal times even after intense fluctuations. 

5.2.8. Relationship between consumer sentiments and the energy sector 

The causality dynamics between the sentiments and the energy sector returns are reported in Fig. 8 . For the causation

running between the overall sentiment index and the energy sector, one finds that the causality-in-mean is bidirectional and 

emanates from the bearish to normal to bullish market conditions, although the strength of predictability declines during 

the extreme market conditions that are in the 10th and 90th quantiles. The causality-in-variance analysis is unidirectional 

running from the energy sector to the overall sentiments under all markets conditions, except at the extreme quantiles. 

The energy sector is a key contributor to the growth in the broader economy. As a result, it is not surprising that volatility

spillovers from the sector predict the sentiments of consumers. Focusing on the causation between the disaggregate mea- 

sures of sentiments and the energy sector, we find a unidirectional causality for both the causality-in-mean and in-variance 

running from the stock price fluctuations in the energy sector to the rural Australia sentiment levels during the normal 

market condition. 

Next, we find there exists no causation in the causality-in-mean and for the mean for the case of energy stock returns

to the sentiments of individuals within the age group of 18–24. We however find the opposite results for the causation

between age groups 25–44 and 45 and above and the energy sector. This is to say that the adult consumers in Australia are

concerned about the variations in the activities of the firms in the energy sector, which impacts their feelings and emotions.

The perceptions and sentiments of the young adults of consumers of ages18–24 seemed not to be impacted by the energy
462 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Fig. 5. Causality-in-quantile between Components of Consumer Sentiments Index Australia and Consumer Staples Sector market returns. Notes : The figure 

plots the estimates of the nonparametric causality test statistics (y-axis) corresponding to various quantiles (x-axis). The dark horizontal yellow and gray 

lines represent the 10% and 5% critical values, respectively. 

463 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Fig. 6. Causality-in-quantile between Components of Consumer Sentiments Index Australia and Industrials Sector market returns. 

Notes. The figure plots the estimates of the nonparametric causality test statistics (y-axis) corresponding to various quantiles (x-axis). The dark horizontal 

yellow and gray lines represent the 10% and 5% critical values, respectively. 

464 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

sector under all market conditions. Thus, for policy decision- making regarding investment and the management of risk, the 

quantile causality bears more information during varied conditions of the market since it is evident that the sentiments of 

the adult workforce which includes noise traders in Australia are influenced by the energy sector’s activities. 

5.2.9. Relationship between consumer sentiments and the utility sector 

The utilities sector represents a class of companies that provide basic services and essentials, such as water, electricity, 

sewage services, dams, and natural gas. Investors typically treat the sector for utilities as long-term holdings and use them 

to generate steady income flows for their portfolios. The results of the causality-in-quantile between the consumer senti- 

ments in Australia and the utilities stock returns are presented in Fig. 9 . Regarding the overall CSI and the stock returns

in the utilities sector, we find the causality-in-mean and in-variance are bidirectional across the normal states of the mar- 

kets and the extreme states. The hump-shaped causality- in -mean curvature indicates that the causality is not symmetric. 

This implies the presence of a strong predictive power around the conditions of a normal market. However, the predictive 
Fig. 7. Causality-in-quantile between Components of Consumer Sentiments Index Australia and Basic Materials Sector market returns. 

Notes. The figure plots the estimates of the nonparametric causality test statistics (y-axis) corresponding to various quantiles (x-axis). The dark horizontal 

yellow and gray lines represent the 10% and 5% critical values, respectively. 

465 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Fig. 7. Continued 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

strength declines in the approach towards the extreme quantiles, i.e. the 90th and 10th quantiles, respectively, correspond- 

ing to the bull and bear markets. This result demonstrates that the role of information efficiency of consumers in the lead

of the utilities market dwindles in the extreme bearish and bullish market conditions. 

For the aggregate measures of sentiments and the utilities sector, we document several findings. For example, we find 

evidence of a causation during the normal market conditions for most cases. Fluctuations in the utilities sector in all the

market states don’t affect the sentiments of the consumers in rural Australia and those within the age bracket of 25–44

and 45 above. No causation also exists between the sentiments of consumers within the age brackets of 18–24 and utilities

sector. However, we find a causation running from sentiments of consumers who are within the age of 24–44 and 45 years

and above and the utilities sector during the normal markets condition. Since individuals hold the defensive utility stocks 

for long term purposes, they fail to react even in the extreme market conditions. Our finding that a shock to the Australian

consumer sentiments causes substantial or significant changes in returns and volatility of the Australian utility stocks is 

inconsistent with the findings of Sayim et al. (2013) with regards the effects of sentiments on the US utility stocks. 

Overall, considering the sectoral level, this study finds that the energy, financials, industrials, consumer stables sectors 

are vastly influenced by movements in the sentiments. These sectors can be seen as more idiosyncratic in the valuation. The

empirical evidence supports the theory of behavioral finance that sentiments play a vital role in stock pricing. Consumer 

sentiments seem to lack a predictive power of the subsequent aggregate and most of the sector returns across the bearish

and bullish markets states. This result is indicative that any trading strategy set to take advantage of public sentiments 

needs to be exercised with caution, as the market tends to be informationally efficient during such market conditions. 

5.3. Discussion 

From the results presented so far, we observe that differences exist in the various age groups’ sentiments and industry 

stock returns, which is consistent with the findings of Johnson and Naka (2014) in the case of the U.S. markets. The re-

sults suggest that younger consumers tend to be more optimistic than older generations, which is in agreement with the 

life cycle investment hypothesis. A possible explanation from the behavioral standpoint would be that younger consumers 

have more years of their life to participate in the labor force and earn money, subsequently are more hopeful in the cur-

rent and future savings and consumption, while older consumers have less years to participate in the labor force. Ando and

Modigliani (1963) and Johnson and Naka (2014) find that younger consumers have higher propensity to consume, compared 

to middle and older consumers. In addition, older consumers with children may reduce consumption as their children en- 

ter adulthood. With respect to how sentiments of different aged groups predict stock volatility, we observe findings that 

are similar to what we obtained in the case of the mean returns. We establish that sentiments of different age groups

impact industry stocks, which is similar to the findings of Akhtar et al. (2011) , Baker and Wurgler (2006) , Lemmnon and

Portniaguina (2006) among others who established that sentiments affect firms of various sizes and industry differently. 

One of the interesting findings we observed has to with the predictive power of sentiments of older consumers on stock

returns and volatility. Johnson and Naka (2014) note that older individuals are less likely to have a significant stock market

exposure because of their near-term focus on retirement and inclination to possibly hold a larger portion of fixed income 

securities. However, that is not the case in this paper. For example, we find a causation running from the sentiments of con-

sumers who are 45 years and above and the utilities sector during the normal market condition. A possible factor that may

have caused the change could be the older generation’s exposure to social media. We make this claim because according 

to Siganos et al. (2014) , Facebook is no longer the exclusive domain of younger people. Kramer and Chung (2011) find that

the average age of Facebook users is about 31 years with more than a quarter of Facebook users being older than 45 years.

Since the media coverage escalates the level of gloomy attitudes towards investment decisions according to Brigida and 

Pratt (2017) , we hold the view that the older generations in Australia are well informed about market conditions than be-

fore. 
466 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Fig. 8. Causality-in-quantile between Components of Consumer Sentiments Index Australia and Energy Sector market returns. Notes. The figure plots the 

estimates of the nonparametric causality test statistics (y-axis) corresponding to various quantiles (x-axis). The dark horizontal yellow and gray lines 

represent the 10% and 5% critical values, respectively. 

467 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

Fig. 9. Causality-in-quantile between Components of Consumer Sentiments Index Australia and Utilities Sector market returns. Notes : The figure plots 

the estimates of the nonparametric causality test statistics (y-axis) corresponding to various quantiles (x-axis). The dark horizontal yellow and gray lines 

represent the 10% and 5% critical values, respectively. 

468 



A.K. Tiwari, E.J.A. Abakah, C.O. Bonsu et al. Journal of Economic Behavior and Organization 193 (2022) 443–472 

 

 

 

 

 

 

 

 

 

 

 

 

Furthermore, a strand of studies finds a direct relationship between false news and financial markets. Fake news about 

markets is becoming an issue to many economists. Clarke et al. (2020) find that the equity price reaction to false news

is discounted when compared to the authentic news articles. Hence, we argue that consumers’ feelings and sentiments in 

Australia could be impacted by fake news that flows through the media. H owever, its effect will be dependent on the age of

consumers since younger consumers are likely to ignore fake news than older generation. This is because younger generation 

are more concerned about increasing wealth and as a result may have a strong desire to take more risk compared to older

generations. 

From the psychological standpoint, news of traumatic events significantly increases consumer’s levels of fear and anxiety. 

Thus, fear and panic are seen as an indirect channel through which media coverage impacts investors’ feelings since fear and

panic are associated with perceived risk among