If Global or Local Investor Sentiments are Prone to Developing an Impact on Stock Returns, is there an Industry Effect?

This paper investigates the heterogeneous impacts of either Global or Local Investor Sentiments on stock returns. We study 10 industry sectors through the lens of 6 (so called) emerging countries: China, Brazil, India, Mexico, Indonesia and Turkey, over the 2000 to 2014 period. Using a panel data framework, our study sheds light on a significant effect of Local Investor Sentiments on expected returns for basic materials, consumer goods, industrial, and financial industries. Moreover, our results suggest that from Global Investor Sentiments alone, one cannot predict expected stock returns in these markets.


Introduction
To consider investor sentiments (IS), as a persistent vehicle to capture market fluctuations, is a modern research field (Neal and Wheatley, 1998), but had been already considered by Keynes (1936). Statman (2000, 2003) even distinguished « Wall Street strategists from individual investors or newsletter writers ». That was followed by Brown and Cliff (2005), Chang et al. (2011), and Baker et al. (2012). Schmeling (2009) studied IS and stock returns (SR) in 18 industrialized countries and Baker and Wurgler (2006) found that IS have « larger effects on securities whose valuations are highly subjective and difficult to arbitrage ».
Since equities from different industries have different characteristics, the impacts of investor sentiments should be expected to vary among industries, as promoted by Chen, Chen, and Lee (2013). In fact, empirical and theoretical evidences (Bu and Li, 2014;Dhesi and Ausloos, 2016) recently show that investors' (irrational or not) emotions, tied to economic or other news, seem to discrepantly sway the future returns of stocks.
Yet to our knowledge there is only a small number of studies which focus on the correlation between specific industries' expected returns and investor sentiments. Beside the remarkable paper by Chen et al. (2013), only a few scholars have regarded emerging markets as the investigated objects for an IS-SR nexus research.
Our study focusses on industry type correlations with both global or local IS in a few emerging markets. Both local investor sentiments (LIS) and global investor sentiments (GIS) are collected within a panel data set which contains the 2000 to 2014 period for 6 (so called) emerging countries, and 10 industries: basic materials, consumer goods, consumer services, financials, health care, industrials, oil and gas, technology, telecommunications, and utilities.
An industry refers to a group of companies share the similar business activities, i.e., producing products, services or close substitutes. The classification schemes reflect different sources, revenues, and the stages of industry cycle, therefore respond differently to information. For each industry, the stock returns are computed from the data of FTSE global price indexes. We use each country's (thus local) consumer confidence index to reflect LIS; we measure the GIS using the State Street Investor Confidence Index (http://www.statestreet.com/ideas/investor-confidence-index.html). We include business cycle as a control variable in our regression framework returns, following McLean and Zhao's (2014) recommendation.
We find that the LIS has a heterogeneous effect on expected stock returns in different industries. In particular, the empirical results suggest that the expected returns of basic materials, consumer goods, industrials and financials industries have significant and positive correlations with LIS. This finding is consistent with most of previous conclusions. However, we find that GIS do not allow to predict expected stock returns. Moreover, our conclusive results also suggest that it is inadequate to only use LIS and GIS to perform a rational anticipation of future stock returns.
Consequently, this study contributes to the literature in three folds. First, the study focuses on the role of industry in the correlation between IS and expected SR. One of our main motivations is to find which industry is the most vulnerable to investors' emotion fluctuations and subsequently which industry is most stable during the bullish or bearish sentiment periods.
Second, we take both GIS and LIS into consideration. Third, this paper provides studies through the lens of emerging countries: China, Brazil, India, Mexico, Indonesia and Turkey.
We use the data between January 2000 and December 2014 to guarantee the volatility and practicability of the research. To the best of our knowledge, this is the first study to investigate the correlation between investor sentiments and different industries' expected stock returns with the panel data of emerging countries, on an equal footing.
The remainder of the paper is organized as follows. We propose our hypotheses in Section 2.
Section 3 outlines the methodology with our comments on relevant variables. Section 4 presents data adjustments. In Section 5, we test our hypotheses and reports the findings.
Section 6 serves as a conclusion. There are several Appendices containing « numerical and technical details ».

Hypotheses development
There are numerous studies which focus on the correlation between stock returns and investor sentiments; details of those previous studies are displayed in Table 1. However, since the existing studies of investor sentiments and stock returns are too numerous to be all mentioned, we only list articles which are representative and frequently cited in the literature. According to Table 1, we can conclude that investor sentiments are proved to be influential to expected stock returns in many countries. We therefore formulate our first hypothesis as follows: Hypothesis 1. LIS has a positive impact on industry stock returns in emerging countries.
Next, let it be observed that there is a scarcity of evidence on the impact of global sentiments.
As noted earlier, it is important to investigate global sentiments along with local sentiments. Chang et al. (2011:36) confirmed the impact of GIS across international markets, with the claim that the LIS effect is "just an empirical manifestation of the global effect''.
Therefore, our second hypothesis reads as follows: Hypothesis 2. GIS has a positive impact on industry stock returns in emerging countries, with the effect that more accessible markets will be associated with a stronger GIS.
Notice that Baker et al. (2012) found that both global and local market sentiments have a predictive power on industry returns, global sentiment playing a more significant role. Since only a few researchers perform empirical studies on the relationship(s) between sentiments and stock returns, distinguishing different industries, the research question (RQ) tied to H1 and H2 immediately follows:

RQ:
If GIS or LIS are prone to developing an impact on SR, is there an industry effect ? -at least in emerging countries and for a considered time interval.

Methodology and Variables
We employ a classical linear panel fixed-effect regression model (Simpson & Ramchander, 2002;Brown & Cliff, 2006;Baker & Wurgler, 2006;Tsuji, 2006;Schmeling, 2009;Chen et al., 2013) which will be separately conducted for the different industries in order to make some comparison: 2, 3, …, N ; t=1, …,T; i and t refer to the country number and time, respectively; N is the maximum number of countries so examined. Notice that the error should be considered as taken at the date of the forecast, t+1, not the day before.
Let for China, Brazil, India, Mexico, Indonesia and Turkey, the country number to be 1, 2, 3, 4, 5, and 6, respectively. SR represents the expected stock returns for a specific industry, assumption on ε i t+1 indicates that the data is expected to be independent, identically distributed. This assumption allows one to model the joint probability of the data as the product of the probability of individual data points. The parameter is a time independent term which allows the possibility of country-specific fixed effects.

Industry Stock Returns
The dependent variable is the stock returns of different industry sectors. As company's demand and supply within a given industry are affected by environmental forces and economic performances in much the same way (Bain, 1959;Porter, 1980), their stock returns are correlated. We follow the industry classification benchmark of the Financial Times Security Exchange (henceforth FTSE) group, which divides all traded equities into ten Firstly, it only employs some specific portfolios to represent the whole market. Secondly, the FSTE price index is not available for several countries' specific industry sectors; in our case, for example, there is no FSTE price index for Mexico technology industry.

The Business Cycles
We control the business cycle for macro-economic impacts on stock returns, using the unemployment rate in each country to capture each country's long-term macro-economic situation. It can be argued that the unemployment rate could only partly reflect a country's long-term macro-economic situation, because it is merely one particular indicator of a country's particular economic condition. Aware of such arguments and limitations, nevertheless, we use the unemployment rate as indicator for measuring the macro-economic status for the practical reason of data availability. One thing needs to be noticed: since the increase of unemployment rate reflects the deterioration of one country's economy, a high unemployment rate therefore denotes negative macro-economic effects. Yet, an increase in unemployment rate might be considered as « good news » by some investors (Boyd et al., 2005).

Data and Preliminary Numerical Results
The data used in this study are obtained from DataStream from January 2001  The data that support the findings of this study are available from Datastream. Restrictions apply to the availability of these data, which were used under license, at the University of Leicester Library, for this study. Data are available from the authors with the permission of University of Leicester and Datastream, as obtained from https://www2.le.ac.uk/library/find/databases/d/datastream use a widely accepted fraud detection method in order to investigate the problematic aspects of the data set. We then adjusted the data set by deleting all the abnormally repetitious values pointed out when applying Benford's laws (Ausloos et al., 2015). Table 2 reports the adjusted data set's descriptive statistics. Notice (1)

Local investor sentiments correlations to the stock returns
The regression analysis results for the ten industries are reported in Table 4 Additionally, except for oil and gas (OIL), the other five industries' expected returns are insignificantly connected with both GIS and LIS, i.e. investor sentiments have no obvious influence on those expected industry returns. Oil and gas companies are exposed to systematic asset price risks, (Hamilton, 1983;Burbidge and Harrison, 1984;Sadorsky, 1999; Gupta and Banerjee, 2019) both oil prices and gasoline prices (Bacon, 1991;Deltas, 2008), and their investment behavior respond to the dynamic political and market environments One should stress that, for emerging countries, LIS are more influential than GIS. The interpretation goes as follows: because in the emerging stock markets, local investors could trade shares more freely than foreign investors, the expected industry returns could only demonstrate fluctuations induced by LIS. Interestingly, technology (TECH) is negatively (but not significantly) related to local sentiments; Wurgler (2006, 2007) provided some possible explanations which we consider to be realistic, for the phenomenon. They stated that those speculative stocks are difficult to price and thus hard to arbitrage; therefore, higher optimism periods of the stock market are expected to produce lower future returns of the stocks.
However, GIS cannot predict most of the industries' expected returns, while numerous existing studies, e.g. Chen et al. (2013), suggest that global sentiments could significantly affect the expected return of several specific industries. This discrepancy between findings might be generated by the differences between research objects. In our research, we take emerging markets as the investigated object, rather than developed markets. Those emerging stock markets completely differ from the well-developed markets. On one hand, the governments play significant roles in the stock markets of emerging countries; thus the stock markets' degrees of freedom are decreased (Harvey, 1995;Claessen, Dasgupta, and Glen, 1995). Take the Chinese stock market for example: the Chinese government has issued many regulations to limit capital investment from foreign country 'for security reasons'.
Additionally, external investors are restricted to buy Chinese equities; therefore, changes of GIS could not bring tremendous impacts on emerging countries' expected stock returns. On the other hand, due to the limitations of investment knowledge, local stock traders might not respond notably or quickly to some foreign financial information; they are more sensitive to domestic news. Hence, the expected returns would not reflect the changes in the GIS. The foreign investor, however, may face formal or informal barriers when trading (Serra, 2000) because they might be more informative and efficient (Vo, 2017, Syamala and Wadhwa, 2019) but exposed to more risk of stock price crash (Vo, 2018 We suggest that a more complete anticipation of stock returns, for these markets, it seems to us, needs to bring other factors, such as the market status, company financial statements and risk-free rates, or even IPO news (Kim and Byun, 2010), into consideration. Moreover, other investor sentiments indexes might be invented (Bu and Li, 2014) and thereafter tested. It would be also interesting to examine another type of causality, whether the stock market fluctuations, in such (emerging) countries and for this set of different industry sectors, influence the investor sentiments (Chen et al., 2003) rather than the contrary, as assumed here.
Most likely the time correlations will have different spans; the investor, local or global, sentiment reaction might also have to be measured through other means, but the matter is another interesting open door.  Whether sentiment has the cross-sectional effects on stock returns or not.

Conclusions
Investor sentiment is more likely to influence smaller, young, less stable, less profitable and nondividend-paying stocks.

2006
Tsuji Japan Regression Monthly data

1994-2001
Investigates investor sentiment's predictability for expected stock return and its properties in Japan.
Investor sentiment has predict power for short-term expected stock return, while there are no "naïve extrapolation" and "salience effect" in Japan.

Canbas and Kandir
Turkey VAR model Monthly data

1997-2005
Analyzing the correlation between investor sentiment and stock return.
Investor sentiment might not have the ability to anticipate stock return.

2009
Schmeling Developed countries. data 1985-2004 Using the international evidence to prove the connection between investor sentiment and stock return.

Monthly
Investor sentiment could predict the stock return and those countries whose culture encourages investment behavior are more vulnerable to the influence of sentiment. In bear markets, investor sentiment and monetary policy decision have a strong correlation.

1963-2004
The correlation between investor sentiment and the mean-variance tradeoff.
In the low-sentiment periods, market's expected extra return positively link with its conditional variance, while in the highsentiment periods, there is no connection between them. Investigating the influence of investor sentiment on a certain industry's stock return.
In the current period, investor sentiment has a positive relationship with return of an industry, while they are negatively correlated in one lag period.

2014
Oprea and Brad

Romania
Regression Investigating the influence of investor sentiment on stock prices.
Investor sentiment has a positive relationship with returns from small stocks but not from large stocks.