Firms' exposures on COVID-19 and stock price crash risk: Evidence from China

Abstract

This study examines the impact of firms’ exposures on COVID-19 sentiment on the stock price crash risk. We show the exposure on COVID-19 sentiment related to the medical, travelling and uncertain aspects significantly increases the stock price crash risk, while the exposure on COVID-19 sentiment related to vaccines significantly decreases the risk of stock price crash. Furthermore, our findings are stronger for non-state-owned firms and firms with low information transparency. Overall, we provide timely policy implication for economic impacts of the COVID-19 on the stock market.

1. Introduction

The COVID-19 is a global pandemic considered by World Health Organization (WHO). By the end of June 2022, there are close to 570 million reported cases and over 6 million deaths, according to the Johns Hopkins University Coronavirus Resource Centre (https://coronavirus.jhu.edu/). Therefore, the sentiment in stock market can be easily affected by different news of COVID-19 (Lwin et al. 2020) and then affect the stock price dynamics as well as the traders’ behavior (John and Li 2021).

Examining the effect of investors sentiment incurred by the epidemic on stock price crash in an emerging market during the COVID-19 period motivates our research. Now the COVID-19 is not over so evaluating its impact specifically on economics can help policymakers to respond with various policies for a speedy recovery of economics.

In this study, we empirically investigate the impact of the exposure of COVID-19 sentiment on the stock price crash risk. Consistent with Lwin et al. (2020), Cevik et al. (2022), and Akram and Haider (2022), we use the news about COVID-19 to measure the sentiment. Specifically, we use the comprehensive indices constructed by Narayan et al. (2021) which capture the news about COVID-19 from different aspects.

Due to the sentiment brought by COVID-19 news, including medical, travelling, and uncertain aspects, the tolerance for other bad news of managers and investors declines, as the total tolerance of bad news is limited. In this condition, more bad news is hidden to maintain the positive attitude of firms from investors. Therefore, we believe the sentiment related to general news of COVID-19, and the detailed aspects of medical, travelling, and uncertain can lead to an increase in stock price crash risk. While vaccines can help us drive COVID-19 out, so the sentiment exposed to vaccines is opposed to the sentiment exposed to other news of COVID-19, which is consistent with the point of Cevik et al. (2022). Hence, sentiment exposed to vaccines can reduce the risk of stock price crash.

The results support our hypotheses. Specifically, we find heterogeneous impacts of the exposure of COVID-19 sentiment on the stock price crash risk. On the whole, the exposure of COVID-19 sentiment significantly increases the stock price crash risk using the aggregate COVID-19 index. And the exposure of COVID-19 sentiment related to the COVID-19, medical, travelling, and uncertain aspects significantly results in the increase of the stock price crash risk. But the exposure of COVID-19 sentiment related to vaccines significantly decreases the risk of stock price crash. We further find results above are milder in the SOEs and firms with the high information transparency.

The study contributes to the literature in the following ways. Firstly, it contributes to the ongoing discussion on the real effect on economics of COVID-19 (Cevik et al. 2022; John and Li 2021; Kong et al., 2022a; Kong et al., 2022b). Prior studies concentrate more on the whole effects of COVID-19 using the number of new confirmed cases or deaths, which ignore more detailed information brought by COVID-19. Secondly, this study is also related to the board literature on the economic consequences of sentiment. Based on the development of society, we capture the impacts of sentiment related to the macro environment (COVID-19) on stock price crash risk. Thirdly, it adds more evidence to the determinants of the stock price crash risk (Chen et al. 2001; Callen and Fang 2015; Wu et al., 2022; Wen et al. 2019). Although several studies have explored the role of investors’ sentiment on the stock price crash, the role of heterogeneous investor sentiment incurred by COVID-19 in the crash risk has received little attention. Our results provide new evidence to support the bad news hoarding theory.

2. Prior literature and hypotheses development

There is a series of academic studies using the bad news hoarding theory to explain stock crash risk. Managers tend to delay or hide negative information to avoid disclosing unfavorable information to the market (Chang et al. 2017; Graham et al. 2005). However, the stock price crash risk increases when bad news accumulates, and the crash happens when bad news reaches a limited threshold level (Callen and Fang, 2015). Wen et al. (2019) provide evidence that firms with higher retail investor attention have a lower future stock price crash risk. Xu et al. (2021) find the decrease of investors’ ability to find information online leads to the increase of the crash risk, as firms are more likely to hide adverse information.

A broad wave of sentiment will disproportionately affect the price of stocks (Baker and Wurgler, 2006). Stambaugh et al. (2012) find sentiment is one of the reasons of anomalies. Specifically, Ding et al. (2019) divide the investor sentiment into long- and short-run components, finding the opposing impacts of long- and short-run sentiment on stock returns. Similar to our topic, Fu et al. (2021) using the traditional sentiment index, find a significant positive relationship between firm-specific investor sentiment and stock price crash risk.

Unlike prior literature, we investigate the impact of sentiment resulted from the change of COVID-19, which is a focus of society. Existing studies show that the COVID-19 epidemic can affect the capital market significantly, such as the analyst's forecasts (Zhang et al., 2022), trading and returns (Bing and Ma, 2021), and individual investor sentiment (Sun et al., 2021). More specifically, apart from the whole sentiment under the environment of COVID-19, we capture the more detailed sentiment using various fields of news about COVID-19, including the medical, travelling, uncertain and vaccines’ aspects.

News of COVID-19 pandemic, including the number of patients, the use of masks, the condition of travelling and uncertainty, can easily influence the sentiment of people. While the ability of tolerance for bad news is limited, so in this condition, despite the news of COVID-19, both managers and investors are all hoping for positive information from stock market. To ensure the accessibility of financing, managers of firms are more likely to hide negative information, which increases the risk of stock price crash. Based on the analysis above, we assume the sentiment exposed to the overall news of COVID-19 or medical, travelling, and uncertain aspects of COVID-19, can increase the stock price crash risk.

On the contrary, the news of vaccines can bring people positive sentiments (Cevik et al. 2022). Vaccines present the hope for removing viruses. According to the reports of Chinese National Health Commission (NHC), there are stronger impacts of vaccines on the population of 18-59-year-old, who are also the main workforce of firms and more active in stock market. Thus, we develop the hypothesis that exposures to vaccines of COVID-19 sentiment can reduce the stock price crash risk.

3. Data and variables

3.1. Data sources

To capture the exposure on COVID-19 sentiment, we collect the time-series dataset computed by Narayan et al. (2021) from ResearchGate (https://www.researchgate.net/). Considering the accessibility of indices relating to COVID-19, our sample includes Chinese listed firms from the first quarter of 2020 to the third quarter of 2021. We extract financial data from the China Stock Market & Accounting Research database (CSMAR). We remove firms in financial service industry and firms with missing observations of key financial variables. To avoid the influence of outliers, all continuous variables are winsorized at the upper and lower 1% levels.

3.2. Variables

Independent Variable. To measure the exposure of COVID-19, we use the indices constructed by Narayan et al. (2021) which can reflect different types of news including an aggregate COVID-19 index (LA_COVID), a COVID-19 index (LCOVID), a medical index (LMedical), a travel index (LTravel), an uncertainty index (LUncertain) and a vaccine index (LVaccine). Specifically, using the ProQuest TDM Python algorithm to count the number of times each word in different dictionaries presented different topics appear in popular newspapers. Then, adjusting proxies of different types of news about COVID-19 based on the results after running a heteroskedasticity-consistent ordinary least squares (OLS) regression of the times from each detailed topic of words on day-of-the-week dummy variables. More detailed information of calculating the proxies of COVID-19 news, including the lists of newspapers, the dictionary of words for each aspect of COVID-19 news and the equation used to adjust the proxies are shown in the paper of Narayan et al. (2021) .

Dependent Variable. Following Kim et al. (2011), we compute the negative conditional return skewness (NCSKEW) to measure the crash likelihood used in baseline regression. We use Model (2) to estimate the firm-specific daily return W_i,d which equals to the natural log of one plus the residual returned from Model (1). Then we use W_i,d to compute NCSKEW quarterly based on Model (3).

In Model (1) to Model (3), the subscripts i, d and q respectively represent different stocks, the time of each day and each quarter, variables R and Rmkt represent the daily return and the corresponding value-weighted market index respectively. And n is the number of market days for stock i during quarter q. The larger the value of NCSKEW, the higher likelihood of stock price crash.

$$R_{i,d}={\alpha }_{i,d}+{\beta }_{1}Rmk{t}_{i,d-2}+{\beta }_{2}Rmk{t}_{i,d-1}+{\beta }_{3}Rmk{t}_{i,d}+{\beta }_{4}Rmk{t}_{i,d+1}+{\beta }_{5}Rmk{t}_{i,d+2}+{\epsilon }_{i,d}$$ (1)

$$W_{i,d}=ln\left(1+{\epsilon }_{i,d}\right)$$ (2)

$$NCSKE{W}_{i,q}=\ln \frac{-\left[n{\left(n-1\right)}^{\frac{3}{2}}\sum {W_{i,d}}^3\right]}{\left(n-1\right)\left(n-2\right){\left(\sum {W_{i,d}}^2\right)}^{\frac{3}{2}}}$$ (3)

In robustness checks, we use the down-to-up volatility (DUVOL) to weigh the crash risk using Model (4) referring to Chen et al. (2001). We firstly calculate the average W_i,d in quarter q of firm i, recorded as Wave_i,q. Then we count the number of days when W_i,d is higher than Wave_i,q for each firm in each quarter, labelled as n_up. Also, n_down is the the number of days when W_i,d is lower than Wave_i,q for firm i during quarter q. The risk of stock price crash is higher if the value of DUVOL is larger.

$$DUVO{L}_{i,q}=\text{ln}\frac{\left(n_{up}-1\right){\sum }_{down}{W}_{i,d}}{\left(n_{down}-1\right){\sum }_{up}{W}_{i,d}}$$ (4)

Control Variables. According to the existing literature on the determinants of stock price crash (Fu et al., 2021; Kong et al., 2021a; Kong et al., 2021b; Wen et al., 2019; Xu et al., 2021), we also introduce a vector of control variables, including profitability (LROA), the book-to-market equity ratio (LBM), firm size (LSize), leverage (LLev), the volatility (Lsd) and mean value (LR) of stock prices in a quarter. LROA is the return on assets, defined as the ratio of net income to total average assets of a firm last quarter. LBM is the ratio of the book value of equity to the market value of equity of a firm last quarter. We measure firm size (LSize) using the natural logarithm of a firm's circulation market values last quarter. LLev is the ratio of total debt to total assets of a firm last quarter. Using standard deviation of the daily return of each firm last quarter to calculate the volatility (Lsd). LR is the mean value of daily return of firm i in quarter q-1.

4. Empirical results

4.1. Descriptive statistics

Table 1 shows the summary statistics of key variables and the number of observations is 26,404. All independent variables and control variables are lagged by one quarter to avoid the effect of reverse causality. The mean value of NCSKEW is -0.318, while the mean value of DUVOL is -0.255.

Table 1.

Summary statistics.

	N	Mean	SD	Min	Max
NCSKEW	26,404	-0.318	0.872	-2.303	2.339
DUVOL	26,404	-0.255	0.698	-1.833	1.625
LA_COVID	26,404	47.355	10.149	26.596	64.493
LCOVID	26,404	43.462	9.134	26.279	60.476
LMedical	26,404	44.846	9.085	26.869	60.466
LTravel	26,404	25.018	8.823	17.816	43.519
LUncertain	26,404	52.242	10.180	31.951	69.054
LVaccine	26,404	32.406	15.621	15.810	56.022
LROA	26,404	0.026	0.045	-0.148	0.184
LBM	26,404	0.648	0.267	0.103	1.247
LSize	26,404	22.243	1.133	20.084	25.619
LLev	26,404	0.408	0.202	0.053	0.884
Lsd	26,404	0.031	0.027	0.009	0.238
LR	26,404	0.002	0.007	-0.006	0.054

Notes: This table reports the summary statistics of variables in this paper.

4.2. Baseline results

To identify the impact of the exposures on COVID-19 sentiment on the stock price crash risk, we use Model (5) as follows:

$$Cras{h}_{i,q}={\alpha }_{i,q}+{\beta }_{1}Sentimen{t}_{i,q-1}+{\beta }_{2}Control{s}_{i,q-1}+{\beta }_{3}FirmFixedEffec{t}_{i,q}+{\epsilon }_{i,q}$$ (5)

where Crash represents the stock price crash risk, we use NCSKEW to measure it. Sentiment represents the sentiment affected by different news of COVID-19. Controls represent a series of control variables. To further alleviate omitted variable bias, we include the firm fixed effects.

The results of baseline regression are shown in Table 2 . The coefficient of LUncertain is 0.007, and coefficients of LA_COVID, LCOVID, LMedical, LTravel are 0.006, which are all significant, indicating the sentiment affected by total COVID-19 news, COVID-19 news, medical news, travelling news, uncertain news significantly increases the stock price crash risk. While, LVaccine is significantly negative in Column (6) of Table 2, illustrating the sentiment affected by vaccines decreases the stock price crash risk. Therefore, the results of the baseline regression indicate that the negative sentiment exposed to the overall news of COVID-19 can increase the stock price crash risk while positive sentiment exposures to vaccines of COVID-19 can reduce the stock price crash risk.

Table 2.

COVID-19 and stock price crash risk.

	(1)	(2)	(3)	(4)	(5)	(6)
	NCSKEW	NCSKEW	NCSKEW	NCSKEW	NCSKEW	NCSKEW
LA_COVID	0.006***
	(11.39)
LCOVID		0.006***
		(11.40)
LMedical			0.006***
			(10.19)
LTravel				0.006***
				(8.31)
LUncertain					0.007***
					(13.45)
LVaccine						-0.003***
						(-6.84)
LROA	-0.740***	-0.777***	-0.779***	-0.668***	-0.731***	-1.130***
	(-4.36)	(-4.59)	(-4.60)	(-3.85)	(-4.32)	(-6.80)
LBM	-1.674***	-1.643***	-1.677***	-1.520***	-1.641***	-1.404***
	(-12.32)	(-12.15)	(-12.31)	(-11.19)	(-12.16)	(-10.14)
LSize	0.154***	0.168***	0.154***	0.255***	0.163***	0.299***
	(4.47)	(4.92)	(4.46)	(7.34)	(4.79)	(8.07)
Llev	-0.192	-0.185	-0.201	-0.073	-0.175	-0.097
	(-1.55)	(-1.49)	(-1.61)	(-0.58)	(-1.42)	(-0.79)
Lsd	-0.632	-0.476	-0.481	-1.482***	-0.793*	-0.541
	(-1.52)	(-1.15)	(-1.16)	(-3.48)	(-1.91)	(-1.30)
LR	6.810***	6.216***	6.498***	8.941***	6.891***	6.113***
	(5.16)	(4.70)	(4.92)	(6.73)	(5.23)	(4.56)
Constant	-2.816***	-3.165***	-2.808***	-5.060***	-3.132***	-5.897***
	(-3.37)	(-3.81)	(-3.35)	(-5.96)	(-3.78)	(-6.58)
Observations	26,404	26,404	26,404	26,404	26,404	26,404
R-squared	0.201	0.201	0.201	0.200	0.203	0.199
Firm FE	YES	YES	YES	YES	YES	YES

Notes: The t statistics reported in parentheses are based on all standard errors clustered at the firm level. *, **, and *** indicate two-tailed significance at the 10%, 5%, and 1% levels, respectively.

4.3. Robustness checks

To further verify the impacts of firms’ exposures on COVID-19 on the risk of crash, we use the alternative proxy of crash risk (DUVOL) to repeat our baseline regression. The results of running Model (5) replacing NCSKEW with DUVOL as the dependent variable are provided in Table 3 . Consistent with baseline regression, LA_COVID, LCOVID, LMedical, LTravel and LUncertain are positive at 1% significant level and the value of corresponding coefficient is similar with the baseline result. The coefficient of LVaccine is -0.002 and significant at 1% level, which is consistent with the result using NCSKEW to measure Crash.

Table 3.

Robust tests.

	(1)	(2)	(3)	(4)	(5)	(6)
	DUVOL	DUVOL	DUVOL	DUVOL	DUVOL	DUVOL
LA_COVID	0.006***
	(14.82)
LCOVID		0.006***
		(14.77)
LMedical			0.006***
			(13.28)
LTravel				0.005***
				(10.19)
LUncertain					0.007***
					(17.61)
LVaccine						-0.002***
						(-7.66)
LROA	-0.454***	-0.497***	-0.494***	-0.415***	-0.445***	-0.840***
	(-3.32)	(-3.65)	(-3.61)	(-2.99)	(-3.27)	(-6.22)
LBM	-1.631***	-1.598***	-1.634***	-1.479***	-1.596***	-1.386***
	(-14.40)	(-14.19)	(-14.39)	(-12.98)	(-14.23)	(-11.88)
LSize	0.167***	0.182***	0.167***	0.267***	0.177***	0.301***
	(5.64)	(6.20)	(5.62)	(8.84)	(6.04)	(9.41)
Llev	-0.071	-0.063	-0.079	0.042	-0.053	0.012
	(-0.70)	(-0.62)	(-0.78)	(0.41)	(-0.52)	(0.11)
Lsd	0.656**	0.818**	0.813**	-0.124	0.488	0.764**
	(1.98)	(2.47)	(2.45)	(-0.36)	(1.48)	(2.29)
LR	1.898*	1.290	1.572	3.912***	1.983*	1.325
	(1.77)	(1.20)	(1.46)	(3.60)	(1.85)	(1.22)
Constant	-3.175***	-3.546***	-3.163***	-5.369***	-3.505***	-5.987***
	(-4.43)	(-4.97)	(-4.40)	(-7.30)	(-4.94)	(-7.74)
Observations	26,404	26,404	26,404	26,404	26,404	26,404
R-squared	0.237	0.237	0.236	0.234	0.240	0.232
Firm FE	YES	YES	YES	YES	YES	YES

Notes: The t statistics reported in parentheses are based on all standard errors clustered at the firm level. *, **, and *** indicate two-tailed significance at the 10%, 5%, and 1% levels, respectively.

To sum up, the sentiment exposure to COVID-19, medical, travelling, and uncertain news or aggregate news of COVID-19 can significantly increase the stock price crash risk, while sentiment exposed to news of vaccines leads to the decrease in crash risk. Thus, the results in robustness checks again support our hypotheses.

4.4. Cross-sectional analysis

4.4.1. Effects of state-owned enterprises (SOEs)

SOEs generally have close connections with local governments and take some social functions in China. Therefore, SOEs can obtain long-term loans with low credit worthiness requirements (Lin et al. 2020). Accordingly, SOEs may face less financial constraints than private firms during the COVID-19 epidemic. In this subsection, we thus explore the moderation effect of SOEs on our main findings. Specifically, we divide the sample into SOEs and private firms, then regress again based on Model (5).

Column (1) to (6) of Table 4 shows the variable of Sentiment isn't significant except LTravel in the sample of SOEs. But for private firms in Column (7) to (12) of Table 4, the variable of Sentiment is significant positive. These results show the impact of exposures on COVID-19 sentiment on the stock price crash risk is milder in SOEs.

Table 4.

The effects of SOEs.

	Results based on SOEs							Results based on private firms
	(1)	(2)	(3)	(4)	(5)	(6)		(7)	(8)	(9)	(10)	(11)	(12)
LA_COVID	0.001							0.008***
	(1.12)							(13.20)
LCOVID		0.001							0.009***
		(0.60)							(13.54)
LMedical			0.001							0.009***
			(0.63)							(12.14)
LTravel				0.004***							0.007***
				(2.99)							(7.83)
LUncertain					0.001							0.010***
					(1.56)							(15.18)
LVaccine						-0.000							-0.004***
						(-0.07)							(-7.90)
LROA	-1.793***	-1.834***	-1.830***	-1.568***	-1.770***	-1.874***		-0.522***	-0.570***	-0.562***	-0.487**	-0.532***	-1.069***
	(-4.30)	(-4.40)	(-4.40)	(-3.73)	(-4.26)	(-4.62)		(-2.62)	(-2.88)	(-2.83)	(-2.38)	(-2.67)	(-5.50)
LBM	-1.796***	-1.785***	-1.788***	-1.759***	-1.797***	-1.775***		-1.649***	-1.603***	-1.661***	-1.433***	-1.588***	-1.230***
	(-7.54)	(-7.49)	(-7.50)	(-7.39)	(-7.56)	(-7.28)		(-9.46)	(-9.28)	(-9.49)	(-8.26)	(-9.24)	(-7.04)
LSize	0.333***	0.338***	0.336***	0.382***	0.334***	0.342***		0.099**	0.118***	0.097**	0.236***	0.114***	0.321***
	(4.44)	(4.51)	(4.46)	(4.92)	(4.46)	(4.26)		(2.35)	(2.85)	(2.29)	(5.61)	(2.77)	(7.14)
LLev	0.710***	0.713***	0.711***	0.813***	0.713***	0.716***		-0.407***	-0.396**	-0.419***	-0.273*	-0.382**	-0.265*
	(2.70)	(2.70)	(2.70)	(3.05)	(2.71)	(2.70)		(-2.63)	(-2.58)	(-2.71)	(-1.78)	(-2.49)	(-1.74)
Lsd	1.539	1.588	1.590	0.337	1.466	1.598		-1.172**	-0.968**	-0.990**	-1.948***	-1.351***	-1.091**
	(1.51)	(1.56)	(1.57)	(0.32)	(1.44)	(1.57)		(-2.42)	(-2.00)	(-2.04)	(-3.91)	(-2.79)	(-2.25)
LR	7.828***	7.822***	7.828***	10.256***	7.833***	7.931***		6.505***	5.658***	6.110***	8.464***	6.557***	5.455***
	(2.75)	(2.74)	(2.75)	(3.67)	(2.76)	(2.75)		(4.10)	(3.57)	(3.85)	(5.27)	(4.14)	(3.38)
Constant	-6.908***	-7.010***	-6.958***	-8.118***	-6.960***	-7.089***		-1.699*	-2.175**	-1.642	-4.653***	-2.188**	-6.345***
	(-3.80)	(-3.86)	(-3.82)	(-4.30)	(-3.84)	(-3.65)		(-1.68)	(-2.18)	(-1.62)	(-4.56)	(-2.20)	(-5.91)
Observations	8,025	8,025	8,025	8,025	8,025	8,025		16,135	16,135	16,135	16,135	16,135	16,135
R-squared	0.193	0.193	0.193	0.194	0.193	0.193		0.214	0.215	0.213	0.209	0.217	0.210
Firm FE	YES	YES	YES	YES	YES	YES		YES	YES	YES	YES	YES	YES

Notes: The t statistics reported in parentheses are based on all standard errors clustered at the firm level. *, **, and *** indicate two-tailed significance at the 10%, 5%, and 1% levels, respectively.

4.4.2. Information transparency

Bad-news-hoarding activities finally lead to the stock price crash (Chang et al. 2017; Cevik et al. 2022; Ding et al. 2019). We set the high transparency sample with the rating of A or B and low transparency sample with the rating of C or D according to the information transparency rating issued by Stock Exchange. While less transparency provides firms with more chances to hide bad news. So we predict that the impacts of firms’ exposures on COVID-19 sentiment on stock price crash risk are stronger for firms with low transparency which are more likely to hide bad news. Then, we follow Model (5) to identify heterogeneous results.

The results are presented in Table 5 . As for the variable measuring sentiment, the coefficient of low information transparency sample in Column (7) to (12) of Table 5 is more than twice as large as the coefficient of high information transparency sample in Column (1) to (6) of Table 5. These results support our hypothesis, that is, the impact of the sentiment on the stock price crash risk is larger in the low information transparency sample.

Table 5.

Heterogeneous results by information transparency.

	Results based on firms with high information transparency							Results based on firms with low information transparency
	(1)	(2)	(3)	(4)	(5)	(6)		(7)	(8)	(9)	(10)	(11)	(12)
LA_COVID	0.004***							0.010***
	(6.29)							(7.78)
LCOVID		0.003***							0.011***
		(5.49)							(7.88)
LMedical			0.004***							0.010***
			(5.74)							(7.35)
LTravel				0.002***							0.008***
				(2.79)							(4.37)
LUncertain					0.004***							0.011***
					(6.90)							(8.66)
LVaccine						0.001							-0.003**
						(1.46)							(-2.34)
LROA	-1.341***	-1.406***	-1.381***	-1.434***	-1.337***	-1.619***		-1.248***	-1.233***	-1.222***	-0.927**	-1.301***	-0.808**
	(-6.04)	(-6.36)	(-6.24)	(-6.18)	(-6.04)	(-7.66)		(-3.37)	(-3.34)	(-3.30)	(-2.47)	(-3.52)	(-2.11)
LBM	-1.638***	-1.611***	-1.637***	-1.555***	-1.620***	-1.639***		-1.309***	-1.308***	-1.282***	-1.315***	-1.357***	-1.166***
	(-9.38)	(-9.24)	(-9.37)	(-8.86)	(-9.31)	(-8.99)		(-3.39)	(-3.39)	(-3.32)	(-3.28)	(-3.53)	(-2.95)
LSize	0.237***	0.251***	0.238***	0.302***	0.243***	0.253***		0.336***	0.338***	0.341***	0.388***	0.322***	0.402***
	(5.22)	(5.55)	(5.22)	(6.60)	(5.39)	(5.11)		(2.98)	(3.00)	(3.02)	(3.41)	(2.86)	(3.57)
LLev	0.220	0.223	0.216	0.279	0.223	0.217		0.312	0.315	0.305	0.385	0.321	0.346
	(1.27)	(1.29)	(1.25)	(1.59)	(1.29)	(1.25)		(1.04)	(1.05)	(1.02)	(1.32)	(1.07)	(1.19)
Lsd	-0.319	-0.170	-0.122	-1.119	-0.550	-0.091		-0.588	-0.169	0.003	-2.626	-1.246	1.288
	(-0.46)	(-0.25)	(-0.18)	(-1.52)	(-0.80)	(-0.13)		(-0.37)	(-0.11)	(0.00)	(-1.45)	(-0.78)	(0.82)
LR	8.868***	8.612***	8.601***	10.524***	9.117***	9.419***		6.702	6.068	5.933	10.925**	7.276*	6.113
	(5.20)	(5.03)	(5.03)	(6.14)	(5.35)	(5.44)		(1.61)	(1.46)	(1.42)	(2.57)	(1.75)	(1.45)
Constant	-4.750***	-5.069***	-4.766***	-6.154***	-4.931***	-4.959***		-7.424***	-7.486***	-7.563***	-8.290***	-7.177***	-8.496***
	(-4.31)	(-4.61)	(-4.31)	(-5.50)	(-4.50)	(-4.14)		(-2.73)	(-2.75)	(-2.78)	(-3.03)	(-2.65)	(-3.12)
Observations	18,698	18,698	18,698	18,698	18,698	18,698		3,394	3,394	3,394	3,394	3,394	3,394
R-squared	0.216	0.215	0.215	0.214	0.216	0.214		0.288	0.288	0.286	0.278	0.292	0.274
Firm FE	YES	YES	YES	YES	YES	YES		YES	YES	YES	YES	YES	YES

Notes: The t statistics reported in parentheses are based on all standard errors clustered at the firm level. *, **, and *** indicate two-tailed significance at the 10%, 5%, and 1% levels, respectively.

5. Conclusion

Using the evidence from the COVID-19 pandemic, we find the sentiment affected by medical, travelling, and uncertain news about COVID-19 significantly increases the stock price crash risk. On the contrary, the exposure of COVID-19 sentiment related to vaccines significantly decreases the risk of the stock price crash. We further find the results above are milder in the SOEs and firms with the high information transparency.

Our findings not only enrich the existing studies on the economic consequences of the COVID-19 epidemic and the influence factors of stock price crash risk but also provide some implications for managers. Our cross-sectional analysis shows that the positive effect of sentiment affected by COVID-19 on stock price crash can be mitigated by information transparency to some extent. Therefore, the managers could allocate more resources to improve their information quality to avoid crash risks.

Author statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

• Data will be made available on request.