COVID-19 and Investor Sentiment Influence on the US and European Countries Sector Returns

Although some studies recently address the association between COVID-19 sentiment and returns, volatility, or stock trading volume, no one conducts an analysis to measure the impact of investor rationality or irrationality on the influence on countries and sectors’ returns.<br><br>This work creates a text media sentiment and combines its influence with the outbreak cases on the stock market sector returns of the US, Europe, and their main countries most affected by the pandemic.<br><br>This allows us to perceive the ranking impact of rationality or irrationality on country and sector stock returns. This work applies a random-effects robust panel estimation, with an M-estimator. This paper concludes that US returns are more sensitive to sentiment, and thus more prone to irrational factors than confirmed cases compared to Europe and that country factors influence the returns differently. In Italy and Spain as the most punished countries in Europe apart from the UK, present sector indexes return more reactive to verified cases, or rationality, namely, tourism, real estate, and the automobile (this last one in Italy).<br><br>The importance of this work resides in providing a new in-depth analysis of irrational behavioral metrics among countries, which allows for comparison. Moreover, it allows observing which sectors’ and which countries’ asset returns are most sensitive to rational or irrational expressions of events, allowing for arbitraging, financial planning for investors, decision-makers, and academia on an in and out of pandemic context.


INTRODUCTION
COVID-19 pandemic affected everyone worldwide. A death rate in Europe of 9.4%, 6% for the Americas and 3% for Asia, and a high contagion rate have created the economic and social chaos infecting 5 million people worldwide and 330 thousand deaths (May 19th, 2020). Although the death rate seems to be decreasing, the number of cases is going up. The US (with 1,529,000) and Europe (1,740,000) present a higher number of cases and within Europe, the United Kingdom with 248,818, Spain with 232,037, Italy with 226,699, Germany with 176,007, and France with 143,427, lead the ranking. The death rate seems to be decreasing in Europe (from February till May; 2.17%, 7.36%, 15.66%, and 14.60%, respectively) and the US (from March till May; 1.93%. 6.60%, 6.33%, respectively) (European Centre for Disease Prevention and Control, 2020).
Academia has widely proven that financial asset prices movement is also explained by sentiment, mainly in periods of irrational, unjustified panics or exaggerated optimism. Investor sentiment is related to emotions, pessimism, or optimism that can influence the investment decisions and, thus, asset prices, as documented by Benhabib, Liu, and Wang (2016), Jitmaneeroj (2017). The sentiment is documented as the deviation bias between asset price sustained by fundamentals and its current price (Zhou, 2018;Giglio & Kelly, 2017), which can be considered mispricing.
If the share price deviates from its theoretical price, then sentiment plays a determinant role to justify the difference. Stock prices can thus become irrational. Accordingly, the Eugene Fama Efficient Market Hypothesis, characterized by a market full of well-informed investors, investments properly priced, and reflecting all available information, is not sustained. That is why market efficiency for asset pricing has been put in question by behavioral finance. Investors are biased and irrational, and these behavioral features play a determining role in asset prices. When incorporated into the models' supply, investor behavior is an explanation for stock returns and volatility. Sharma and Kumar (2019) defend that researchers should present robust behavioral asset pricing models backed by enough empirical evidence worldwide, incorporating investor psychological biases in new robust behavioral asset pricing models. Also, Aggarwall (2019) sustains that financial theories' sentiment construct needs to be revisited according to the sentiments defined in psychology.
Sentiments can be defined as phantasy relationships built unconsciously in mind (Tuckett & Taffer, 2008). This can lead to emotional speculations about future price movements. Perceiving sentiments and human cognition will lead to an understanding of market asset pricing. The impact of decision-making over rational investments built upon company fundamentals or sustained on decisions made upon available information of real events, and irrationality -measured as an expectation based on feelings and not real premises -should be studied and compared in terms of sector and countries' performance.
Accordingly, this paper addresses a new approach by comparing the asset price reaction to verified events (immediate emotions) against those events' anticipation through investor sentiment (lagged and expected emotions) to observe market and sector returns biases regarding psychological behavior. This work intertwines a real psychological event as the fear and panic caused by the COVID-19 pandemic, with a text investor sentiment index that captures the anxiety embedded in different countries' and sectors' stock markets. The importance of the theme lies in the fact that by knowing which sectors and which countries are most reactive to rational or irrational expressions of events, institutional investors, companies providing financial information, political and academic professionals can anticipate market reactions and thus monitor procedures, information, and arbitraging. In academic terms, in addition to meeting the suggested academic gaps in the works of Sharma and Kumar (2019) and Aggarwall (2019), it serves to create the path for the measurement and monitoring of the causes that influence these behaviors to provide the market with an antidote to minimize the asymmetric impacts of erratic behavior and peak levels of sector and country asset exchanges. Furthermore, this work allows analyzing which country is more sensitive to irrational behaviors and compares them, which will contribute to the solution of the sentiment different bias among stock markets. Therefore, this work's research question is to find which country and sector returns response is more prone to rational or irrational behavior. This work is structured as follows: after this introduction section, the recent literature review is provided, then the methods, afterward, the results and their discussion, and finishing with the concluding remarks. Traditional finance theory defends that stock prices reflect the discounted value of expected cash flows and that arbitrageurs eradicate the impact of irrational behavior by investors, but behavioral finance suggests that waves of irrational sentiment -optimistic or pessimistic expectations -can persist and influence asset prices for significant periods of time and consequently cause crises (Zouaoui et al., 2011).

LITERATURE REVIEW
The rational part of the sentiment comes from the expectation of the behavior of future cash flows based upon the company fundamentals, or real news or events that may influence future firm financial behavior. The irrational part comes from over-optimism or pessimism, like panic or exaggerated fear that is subjective. Hirshleifer (2015) shows that investor sentiment is the fluctuating attitude to investment categories, and it may be associated with changes in assessments of expected returns or risk. This attitude might include waves of irrational enthusiasm or repugnance for certain investment characteristics and shifts in the emotional or cognitive opinion that the economic environment triggers. If sentiment brings mispricing, then sentiment measures should predict future abnormal returns.
Many authors compute investor sentiment measure using survey-based measures as the consumer sentiment index or economic sentiment indicators as proxies, or text data sustained upon media collection of words or based on data analysis.
Investor sentiment -the irrational part -is calculated as the error term of a regression between the sentiment proxy and macroeconomic variables or fundamentals to remove the "rational term". The random error term captures the in-  Verma et al. (2008) defend that irrational sentiments have a more rapid and pronounced effect on stock market returns than rational sentiments. They defend that the individual and institutional investor sentiments are driven by rational and irrational factors with different effects on stock market returns. They consider the role of economic fundamentals as determinants of stock market returns, which is the rational part of the authors' sentiment. This rational part of sentiment has a much greater effect on returns explanation than the sentiments-induced by noise or irrational. If excessive optimism drives prices above intrinsic values, periods of high sentiments should be followed by low returns as market prices revert to fundamental values. They also find that rational sentiments are incorporated in stock prices at a slower speed than the irrational sentiments. Verma et al. (2008) define as the rational part of the sentiment the company fundamentals that justify the returns and the irrational part of the error term of the regression between returns and fundamentals.

The pandemic
Zhang, Hu, and Ji (2020)  Accordingly, this work's research question is to find which country and sector returns response is more prone to rational or irrational behavior.

METHOD
Although most studies address the relation amongst COVID-19 sentiment and returns, volatility, or stock trading volume, no one conducts an analysis towards measuring the effect of the rationality of investors on the influence on countries' and sectors' returns. The rational reaction can be measured through real events, where irrationality can be portrayed by sentiment behavior when expecting upcoming events (see Verma et al., 2008, about rational and irrational behavior). The authors' method is divided into three phases: 1) the building of the index sustained on Google Trends; 2) the orthogonalization trough ordinary least squares against macro variables; 3) the data analysis using panel data robust analysis with random effects. This work based the sentiment index on the method proposed by Gao, Ren, and Zhang (2018) that applied households' Google search behavior to construct sentiment indices for different markets. They show that their sentiment measure is a contrarian predictor of country-level market returns. Also, the work of Schatteman and Waymire (2017) contributes to the choice of words for our sentiment. This paper relies on the Merriam-Webster dictionary and finds negative words classified in the group of "econ". This work proceeds with the search of the words on the news from Reuters, CNBC, Bloomberg, and Wall Street Journal websites as from April 2004 till April 16 th , 2020.
Then the combined terms in Google Trends "market crash + depression + recession + "short selling" + panic + default + bankruptcy + losses" are searched to come up with the sentiment index in the subgroup "companies and industries" from 2015 till April 19 th , 2020. This form of words combination code allowed us to check simultaneously whichever of those words were searched specific words were translated into the country language using Google Translate. Accordingly, this work produces a sentiment index with a range value between 0 and 100, meaning that a value of 100 is a very low sentiment (fear, panic, pessimism), implicating an increase in those search words, and 0 otherwise.
Considering that macroeconomic factors may influence this sentiment proxy and thus be biased, to capture real irrational and rational behavior related to the sentiment, this paper follows Baker and Wurgler's (2006) procedure to orthogonalize the index against three macroeconomic variables (Brent, sovereign 10-year yield, and Bitcoin) trough ordinary least squares. Brent and sovereign yield allow us to withdraw systematic risk and bitcoin the speculative risk. Afterward, the residuals are used as true sentiment measure: Then with COVID-19 cases and the sector and global country index returns, this paper carried on a panel data regression analysis with robust (heteroscedasticity and serial correlation, robust standard errors) with random effects (verified after the performance of the Hausman test) for European countries (UK, France, Italy, Spain, Germany) and individual robust standard errors OLS regressions for individual European countries and the US.
Robust regressions use not only an M-estimator (Huber followed by bisquare) but also include a first step that removes high-leverage outliers (based on Cook's D) (Maronna, 2006;Huber, 1973). The standard least-squares method tries to minimize the error, which is unstable if there are outliers present in the data. Outlying data give such a strong effect in the minimization that the parameters thus estimated are distorted. The M-estimator reduces outliers' effect by replacing the squared residuals by another function of the residuals, yielding more optimal estimators (Zhang, 1997). The sentiment measure and cases are standardized with zero mean and one standard deviation for more legible results: where R is the index or the sector ETF return, S is the standardized sentiment obtained in expression (1) and is C the first lagged difference of coronavirus cases, with , β θ and , ϑ the coefficients, i is the country and t is the time, with , , it µ , , it ε mean the between entity and within entity error, respectively. Table 2 presents the detailed variable descriptive statistics.

0.01
Standard errors in parentheses + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001 Standard errors in parentheses + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001 Note: Table 3 reports the influence of standardized sentiment index and standardized first lagged difference of COVID-19 cases on global and sector return indexes or ETFs. For the US globally (dowjones) has more sensitivity to sentiment (0.1% alpha) and in tourism sector and real estate (0.1% alpha). Real cases affect also real estate with an α = 0,1%. For Europe as a whole, markets are more prone to rationality as it can observe that the market as a whole and all sectors are influenced by rational events (α = 0.1% with the exception of telecom sector. Standard errors in parentheses + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001 Standard errors in parentheses + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001 Note: Table 4 reports the influence of standardized sentiment index and standardized first lagged difference of COVID-19 cases on global and sector return indexes or ETFs. Spain's tourism returns are affected by real cases, and Italy is more rationally affected globally in tourism, real estate, and auto industries. Standard errors in parentheses + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001 Standard errors in parentheses + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001 Note: Table 5 reports the influence of standardized sentiment index and standardized first lagged difference of COVID-19 cases on global and sector return indexes or ETFs. Germany's tourism returns are more sensitive to rationality (alpha of 1%), and the UK's telecom sector is more sensitive to rationality.   Standard errors in parentheses + p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001 Note: Table 6 reports the influence of standardized sentiment index and standardized first lagged difference of COVID-19 cases on global and sector return indexes or ETFs. France tourism returns are more influenced by rationality at alpha of 5%. Sentiment data duly standardized with a mean 0 and 1 variance for comparison purposes Figure 2. Sentiment for the US and European countries

DISCUSSION
An increase in negative sentiment implying fear, pessimism, and panic, is negatively associated with returns in the US global index and in tourism and in real estate, the most impacted sectors with the pandemic (Table 3). Irrationality prevails and conditions market returns. Europe also presents the same reaction, although at a lower level than in the US. In Europe, the automobile industry also was affected by investor sentiment. An increase in COVID-19 cases, although relevant in the US for the global, tourism, and real estate indexes, has a more pronounced effect on Europe in all sectors (telecommunications return has a lower reaction). Accordingly, rationality reaction prevails. The US reacts more in anticipation of the sentiment, a measure that captures actual and future concern and fear, rather than with the effective COVID cases. On the other hand, Europe seems to react more to the real scenario and landscape and lower to sentiment, to the anticipation of any scenario. The adjusted R 2 confirms the predictive power of sentiment and COVID-19 cases of returns, mainly on the more damaged sectors of the economy: hotel, leisure, and real estate. Spain and Italy are the European countries that suffered the most, and it is confirmed that the tourism sector had the most impacted returns (Table 4). Sentiment and cases both have equally predicted power on returns for Tourism. In Italy, real estate and automobile indexes returns (car makers are important in Italy) react more to the real cases than on sentiment or mood. Germany and France index returns are more prone to confirmed cases in the Tourism sector than on sentiment (Tables 5 and 6).
According to Figure 1, the sentiment index is accurate and a manifestation of irrationality, considering that it also follows the main events (whether in the US or Europe) that have affected the stock markets. For instance, the main events, such as the ones described in the note to Figure 1, justify the model's accuracy in Europe and the USA.
The contagion and spillover effect among European countries and the US are confirmed in this work. Moreover, the UK and the USA have a stronger connection to the irrational behavior of investors. This spillover effect is capitalized on the COVID-19 period, where it can be seen the panic and fear of investors more underlined (see Table 7). Figure 2 allows concluding that among countries, the UK and the US are more prone to irrational sentiment as it can also be testified by UK sector returns that respond more to irrationality than on real COVID-19 cases (telecommunications sector). Also, Italy and France have strong irrational behavior peaks, but when confronting sector returns over the two -rational and irrationality, rationality appears as the main driver of returns on some sectors. The sentiment index captures the nervous investors before market constraints, and it seems that mood alterations are often seen in European markets than in the US. Despite the mood features, rationality is more present in European sectors than on US sector returns (see Table 3-6).

CONCLUSION
Coronavirus has brought panic, negativity, and contagious effects among global economies. Although this may have caused huge daily and accumulated losses on stock indexes also created opportunities for recoveries and huge gains. The sentiment index proved to be more effective in predicting returns than the real COVID-19 cases that conditioned the market. This encompasses the idea that irrational feelings than rational ones mainly condition investor behavior. The research question is duly answered considering that the US reacted in anticipation when compared with Europe as a whole or even before country-specific effects, and so subject to more irrationality behavior despite the possibility of being more future assertive. Tourism (travel and leisure) and real estate sectors are the more responsive ones to investors' irrational behavior in the US, while tourism, real estate, and automobile are more affected by rationality in Europe. This may prove that even before market contagion and spillovers, the US stock markets react more to the anticipation of bad news and worst scenarios than Europe that reacts more to real pandemic verified confirmed cases. The excessive financial news providers with corporations such as the CNBC, Thomson Reuters, Bloomberg, Wall Street Journal, Moody's, S&P GMI, MSCI,