key: cord-1029189-biv4j7mi authors: Pham, Anh Viet; Adrian, Christofer; Garg, Mukesh; Phang, Soon-Yeow; Truong, Cameron title: State-level COVID-19 Outbreak and Stock Returns date: 2021-03-19 journal: Financ Res Lett DOI: 10.1016/j.frl.2021.102002 sha: c98a2a07f23d350c60d77e7050f9bd91619ff240 doc_id: 1029189 cord_uid: biv4j7mi We use state-level data to evaluate the connection between outbreaks of COVID-19 and stock returns over the period January-June 2020. We show that daily increases in the number of infected cases, hospitalized cases, and deaths are negatively associated with next day stock returns of firms headquartered in the same state. The relationship is weaker among states with high levels of medical resources and states that are likely to get support from the federal government. In addition, we find that the negative effect is reduced for firms that report an expectation that an outbreak will increase revenues and for firms with a strong corporate social responsibility practice. We believe our study is the first paper to assess cross-sectional stock price reactions to COVID-19 as a function of the state-level impact of the pandemic outbreak. • We use state-level data to evaluate the relation between outbreaks of COVID-19 and stock returns over the period January-June 2020. • Daily increases in the number of infected cases, hospitalized cases, and deaths are negatively associated with next day stock returns of firms headquartered in the same state. • The association is less pronounced among states with high levels of medical resources and states that are likely to get support from the federal government. • Also, the negative effect is mitigated for firms that report an expectation that an outbreak will increase revenues and for firms with a strong corporate social responsibility practice. The COVID-19 pandemic has placed a significant negative effect on the U.S. society. As of September 2020, more than 6 million people in the U.S. have been infected with the coronavirus, more than 180,000 people have died, and tens of thousands of newly infected cases are reported daily. 1 While the level of the impact of the disease on capital markets is without precedent (Baker et al., 2020; Demers et al., 2020) , the exact mechanisms by which the pandemic affects security prices have largely remained unexplored. In this study, we investigate whether cross-sectional differences in the geographical dispersion of the pandemic affects asset prices. Specifically, we study if increases at the state level in: i) COVID-19 infected cases, ii) death cases, and iii) hospitalized cases relate to future security returns of U.S. firms headquartered in the same state. Daily data on state-level infected cases, death cases, and hospitalizations are obtained from covidtracking.com over the period from January 22 to June 30, 2020. To measure statelevel COVID-19 outbreak, we construct the ratio of daily increases in infected cases over the states' population, the ratio of deaths over the total number of confirmed cases for the state, and the ratio of daily increases in the number of hospitalized cases over the total number of confirmed cases for the state. Building on the premise that core business activities of firms often occur within proximity to their headquarters (Pirinsky and Wang, 2006; Chaney et al., 2012) and asset prices exhibit a significant regional component (Coval and Moskowitz, 1999; Korniotis and Kumar, 2013), we match each firm's location with measures of the state-level pandemic outbreak using the state in which the firm is headquartered. 2 1 In this study, we focus on the U.S. experience. The scale of the COVID-19 shock is also unprecedented across the globe. As of September, 2020, the pandemic has infected more than 29.5 million people and claimed almost one million lives around the globe (WHO, 2020; available at: https://covid19.who.int/). 2 Korniotis and Kumar (2013) document that expected returns for local stocks would rise during local adverse economic conditions because local risk aversion increases while local risk sharing abilities decrease. The authors argue that the return predictability based on local factors is predominantly based on a form of investor local bias. Our empirical results show that daily increases in: i) the number of infections at the state level, ii) the number of deaths, and iii) the number of hospitalizations are negatively associated with next day stock returns of firms headquartered in the same state. This relation is of large economic significance and cannot be explained by risks, firm characteristics, or state-level macroeconomics. For example, we document that a 1% increase in the number of new COVID-19 positive cases over the total state population results in an average decrease in next day market-adjusted stock returns of 310 basis points. We find that the negative stock return impact of a COVID-19 outbreak within a state is less pronounced for firms headquartered in states with higher available capacity for treating COVID-19 patients. We conclude that the market views medical resources as important in how a state handles the outbreak. Next, we also find that the negative effect of a state-level COVID-19 outbreak on stock returns is weaker among firms headquartered in election battleground states or in states run by Republican governors. We suggest that the incumbent federal administration seeking re-election is more likely to provide federal support to battleground states during the 2020 election year and states led by Republican governors are more likely to be assisted by the Trump administration. Thus, these states are in a better position to fight against the COVID-19 outbreak. We also document that firms who self-report expectations of a positive effect from COVID-19 and firms with strong CSR activities experience a less adverse impact of statelevel COVID-19 outbreak on their stock returns. In our final avenue of inquiry, we conduct a battery of robustness tests with the separate inclusion of the following control variables in our baseline model: i) the proportion of state population whose age is 65 or above, ii) the stocklevel daily trading turnover, and iii) the country-level economic policy uncertainty. Consistent with our main findings, after controlling for these variables, we continue to find that daily state-level increases in infected cases, deaths, and hospitalized cases are negatively related to next day stock returns of firms headquartered in the same state. Our study offers two important contributions to the literature. First, contemporary research has uncovered several potential explanations for cross-sectional variations in stock market reactions to the COVID-19 pandemic. These include policy responses (Baker et al., 2020) , liquidity access (Acharya and Steffen, 2020) , and environmental and social ratings (Albuquerque et al., 2020; Demers et al., 2020) . Our study shows that state-level factors have considerable effects on the cross-sectional reaction of stock prices to COVID-19 outbreaks. Second, this paper contributes to the literature on the locality of firms and their stock returns. Prinsky and Wang (2006) document that stock returns of firms headquartered in the same geographic area exhibit a strong degree of co-movement. This phenomenon can be explained by the geographic segmentation of the equity market where investors show a tendency of overweighting local firms in their portfolios (Coval and Moskowitz, 1999; and Korniotis and Kumar, 2013) . Our findings suggest that there is a significant regional component of price formation in the equity market during the COVID-19 outbreak. The rest of this paper is organized as follows. Section 2 describes our sample and data. In Section 3 we present our empirical methodologies and results, while Section 4 concludes the study. We collect data on daily increases in the number of COVID-19 positive cases, death cases, and hospitalized cases from covidtracking.com. The sample period is between January 22 and June 30, 2020. We estimate three different measures of state-level COVID-19 outbreak. The first measure (Positive_Cases) is the number of new COVID-19 positive cases identified each day as a percentage of the population of a state. The second measure (Death_Cases) is the ratio of the number of COVID-19 related deaths each day over the total number of positive cases for a state. The third measure (Hospitalized_Cases) is the number of new hospitalizations for COVID-19 as a percentage of the total number of positive cases for a state. Following the existing literature (Coval and Moskowitz, 1999; 2001; Pirinsky and Wang, 2006; Korniotis and Kumar, 2013) , we use the state in which a firm has its headquarter as the location of the firm's operations. We then match measures of state-level COVID-19 outbreak with firms' locations based on the state in which the firm is headquartered. We collect firm-level financial statement data from Compustat Fundamentals Annual. Daily stock returns are obtained from the Center for Research in Security Prices (CRSP) and data on Fama and French's (2015) five factors are from Kenneth French's data library. We control for state-level personal income growth, gross domestic product (GDP) growth, and personal consumption growth based on data collected from the U.S. Bureau of Economic Analysis. Table 1 provides descriptive statistics for all variables used in this study. Overall, the main sample contains 393,097 observations for 3,562 U.S. publicly listed firms. The daily market-adjusted stock returns, which are calculated as daily stock returns minus CRSP valueweighted market returns, has a mean value of -0.035%. The daily ratios of new COVID-19 positive cases, death cases, and hospitalized cases have an average value of 0.005%, 0.128%, and 0.303%, respectively. We perform the following ordinary least squares (OLS) regression of daily marketadjusted stock returns on state-level measures of COVID-19 severity: ( 1) where and denote firm and day, respectively. Ind is the industry-fixed effects based on two-digit SIC code, Month is the month-fixed effects and is the error term. The dependent variable is market-adjusted stock returns ( ). COVID19 is a vector for our main independent variables of interest, including Positive_Cases, Death_Cases, and Hospitalized_Cases. Controls is a vector for control variables. In column (4), we include all three measures of the extent of the COVID-19 outbreak for each state into the regression specification. Similarly, the results in column (4) indicate significant negative associations between all three COVID-19 severity measures and future daily market-adjusted stock returns. The effect of COVID-19 on future stock returns is economically significant. For example, column (4) suggests that each 1% increase in the ratio of daily new COVID-19 positive cases over the total state population results in a decrease in the daily market-adjusted stock returns by 310 basis points. Overall, the results from Table 2 suggest that firms headquartered in states that experience a higher level of the COVID-19 outbreak exhibit lower future stock returns. 4 In this section, we examine whether the market reaction to the outbreak of COVID-19 varies across states with different levels of medical treatment capacity. Naturally, we would expect the impact of the COVID-19 outbreak to be less pronounced in states with a greater medical treatment capacity. For this analysis, we utilize the ratio of the number of hospital beds in a state over the state population (Hospital_Beds) as a proxy for the state's medical treatment capacity to handle new COVID-19 cases. A higher Hospital_Beds indicates that the state's medical facilities have higher available resources. We augment model (1) with interaction terms between each of the COVID-19 outbreak measures and Hospital_Beds. The results reported in Table 3 Panel A show positive and significant coefficients on all three interaction variables (Positive_Cases*Hospital_Beds, Death_Cases*Hospital_Beds, and Hospitalized_Cases*Hospital_Beds) at the 5% significant level or lower. These findings suggest that the effect of severe COVID-19 cases on stock returns is less pronounced among firms headquartered in states with more hospital beds per capita. In addition, we employ the ratio of the number of physicians in a state over the state population (Physicians) as another proxy for the state's medical resources. Higher Physicians suggests that the state has a greater medical treatment capacity. We incorporate model (1) with interaction terms between each of the COVID-19 severity measures and Physicians. As shown across all three columns in Table 3 Panel B, firms headquartered in states facing a greater level of COVID-19 severity experience a significant decrease in their stock returns. However, the coefficients of interaction terms between each of the COVID-19 severity measures and Physicians are all positive and statistically significant. These results indicate that the negative relation between our COVID-19 severity measures and stock returns is weaker among firms headquartered in states with a higher number of physicians per capita. Taken together, results from Table 3 suggest that the negative impact of the state-level COVID-19 outbreak on future stock returns is less pronounced among firms headquartered in states with a greater medical treatment capacity to handle COVID-19 patients. Prior literature documents that due to political goals, election battleground states tend to receive more federal support, especially during periods of disasters or election years (Hudak, 2011; Chen, 2013; Kriner and Reeves, 2015) . Since the COVID-19 pandemic struck in an election year, the incumbent federal administration seeking re-election may seek to enhance their chances by providing more federal support to the battleground states to fight against the spread of COVID-19. Therefore, the adverse impact of the COVID-19 outbreak is likely to be somewhat mitigated in battleground states due to potential greater federal support. To investigate this conjecture, we modify our baseline model (1) to include interaction terms between each of the state-level COVID-19 outbreak measures and the indicator variable for battleground states (Battleground_States). Based on the 2016 presidential election results, we define a battleground state as one where the absolute difference in voting results between Donald Trump and Hillary Clinton was equal to or less than 5%, and zero otherwise. (2) where the significant negative relation between Death_Cases and stock returns is less pronounced among firms from battleground states. Collectively, results from Similarly, column (2) indicates that the negative relation between Death_Cases and stock returns is also weaker among firms from states governed by Republican politicians. Overall, these findings show that stock returns of firms from states led by Republican governors experience a lesser COVID-19 impact, possibly due to greater federal support to these states. In this section, we examine whether the negative association between COVID-19 severity and stock returns varies with certain firm characteristics. In particular, we investigate whether the effect of COVID-19 on stock returns varies across firms with different sentiments on the potential impact of COVID-19 on their performance. Intuitively, the COVID-19 pandemic may have detrimental effects on the performance of firms in some industry sectors (e.g., hospitality, transport, oil, and gas) while creating considerable growth opportunities for others (e.g., pharmaceuticals, e-commerce, online entertainment). We We conduct a battery of tests to ascertain the robustness of our results. First, we consider the possibility that a state is more vulnerable to the negative effect of COVID-19 due to the greater proportion of its older population. 6 Therefore, we include the proportion of the state population whose age is 65 or above (Age_65) in model (1). Results from Table 6 Panel A show that our main findings remain unchanged after controlling for the proportion of the population in each state that is more vulnerable to COVID-19. 7 5 Albuquerque et al. (2020) and Ding et al. (2020) show that the impact of the COVID-19 pandemic on stock returns is weaker among firms with better CSR performance. However, Demers et al. (2020) Second, given that the variation in the return patterns could be possibly driven by the daily trading volume, we control for the volume of trading in the baseline model and re-run our regression. Following Chiah and Zhong (2020), we measure the trading volume using the stock-level daily trading turnover (Turnover), computed as the ratio of daily trading volume divided by the number of shares outstanding as of that day. As reported in Table 6 Covidtracking.com The natural logarithm of market capitalization Compustat Leverage The ratio of total debts over assets Compustat Cash The ratio of cash balances over assets Compustat MTB The ratio of market value of equity to the book value of equity Firm-level total environmental, social and corporate governance (ESG) score The proportion of the state population whose age is 65 or above The ratio of daily trading volume divided by the number of shares outstanding as of that day CRSP EPU Baker et al. (2016) economic policy uncertainty policyuncertainty.com (1). The main independent variable is stock-level market-adjusted stock returns (AR). The main independent variables of interests are state-level COVID-19 severity measures including Positive_Cases, Death_Cases, and Hospitalized_Cases. We include a constant term, industry-fixed effects, and month-fixed effects. t-statistics computed standard errors clustered at the industry and month-level (Petersen, 2009 ) are reported in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% levels, respectively. Variable definitions and data sources are presented in Appendix A. with the ratio of the number of hospital beds in a state over the state population (Hospital_Beds). In Panel B, we consider interaction terms between each of the COVID-19 severity measures and the ratio of the number of physicians in a state over the state population (Physicians). We include a constant term, industry-fixed effects, and monthfixed effects. t-statistics computed standard errors clustered at the industry and monthlevel (Petersen, 2009) are reported in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% levels, respectively. Variable definitions and data sources are presented in Appendix A. Adjusted R-squared 0.067 0.067 0.067 This table presents regression results of stock-level market-adjusted stock returns (AR) on interaction terms between COVID-19 severity measures and political factors. In Panel A, we interact each of the COVID-19 severity measures with a dummy variable for the 2016 presidential election battleground states (Battleground_States). In Panel B, we consider interaction terms between each of the COVID-19 severity measures and a dummy variable for states headed by Republican governors (Rep_Gov). We include a constant term, industry-fixed effects, and month-fixed effects. t-statistics computed standard errors clustered at the industry and month-level (Petersen, 2009) are reported in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% levels, respectively. Variable definitions and data sources are presented in Appendix A. Hassan et al.'s (2020) firm-level measure of COVID-19 net sentiment (COVID_Sentiment). In Panel B, we consider interaction terms between each of the COVID-19 severity measures and firm-level corporate social responsibility performance (ESG_Score). We include a constant term, industry-fixed effects, and month-fixed effects. t-statistics computed standard errors clustered at the industry and month-level (Petersen, 2009) are reported in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% levels, respectively. Variable definitions and data sources are presented in Appendix A. The risk of being a fallen angel and the corporate dash for cash in the midst of COVID Resiliency of Environmental and Social Stocks: An Analysis of the Exogenous COVID-19 Market Crash Measuring economic policy uncertainty The Unprecedented Stock Market Reaction to COVID-19 Race, Ethnicity, and Age Trends in Persons Who Died from COVID-19 -United States The collateral channel: How real estate shocks affect corporate investment Voter partisanship and the effect of distributive spending on political participation Trading from home: The impact of COVID-19 on trading volume around the world Home bias at home: Local equity preference in domestic portfolios The geography of investment: Informed trading and asset prices ESG didn't immunize stocks against the COVID-19 market crash Corporate immunity to the COVID-19 pandemic A five-factor asset pricing model Trump attacks second Democratic governor over coronavirus criticism Firm-level exposure to epidemic In Panel A, we control for the proportion of the state population whose age is 65 or above (Age_65). In Panel B, we control for stock-level daily trading turnover (Turnover). In Panel C, we replace state-level GDP growth with economic policy uncertainty (EPU). We include a constant term, industry-fixed effects, and monthfixed effects. t-statistics computed standard errors clustered at the industry and month-level (Petersen, 2009) are reported in parentheses. *, ** and *** denote significance at the 10%, 5% and 1% levels, respectively. Variable definitions and data sources are presented in Appendix A.