Table 2.

Estimation of the role of non-critical business in stock prices during the COVID-19 pandemic.

Dependent variable:	$log\left(p_{i,g,s,c,t}\right)$
Time window:	May$\left|19\text{-Oct}\right|$20 ($\pm$ 8 months)
Sample:	Full	Full	Full	Ex-U.S.	Ex-Dev.
Specification:	(1)	(2)	(3)	(4)	(5)
COVID-19${}_t$ × Non-critical${}_g$	−0.0652${}^{\ast \ast }$	−0.0782${}^{\ast \ast }$	−0.0681${}^{\ast \ast }$	−0.0760${}^{\ast \ast }$	−0.1530${}^{\ast \ast \ast }$
	(0.032)	(0.031)	(0.032)	(0.033)	(0.043)
Ex-ante controls
Firm	Yes	Yes	No	Yes	Yes
Country	Yes	Yes	No	Yes	Yes
Fixed effects
Firm	Yes	Yes	No	Yes	Yes
Time × Subsector	Yes	No	Yes	Yes	Yes
Time × Subsector × Country	No	Yes	No	No	No
Error clustering
Industrial group	Yes	Yes	Yes	Yes	Yes
Observations	1,007,391	1,007,391	1,007,391	840.532	254.331
$R$2	0.990	0.992	0.104	0.990	0.985

Note: This table reports coefficient estimates for the specification in Eq. (1). The dependent variable $log\left(p_{i,g,s,c,t}\right)$ represents the logarithm of the week’s closing price of the stock of firm $i$. We include ex-ante firm-specific controls (fixed with the latest observed values in 2019): Market capitalization, Total assets, EBITDA, and Price-to-book. We also include ex-ante controls for country (fixed with 2019 values): (GDP per capita, Unemployment, Inflation, Country openness and Population). All ex-ante controls are interacted with time to avoid collinearity with fixed effects. $Non\text{-}critica{l}_g$ identifies whether the industrial group $g$ of firm $i$ is a non-essential economic activity. $COVID\text{-}19_t$ is equal to 1 if $t\ge$ 2020-01-31 and 0 otherwise. We cluster errors at the industrial group level, which coincides with the level of variation of the variable $Non\text{-}critica{l}_g$. We report results for a time window of eight months centered on the WHO’s PHEIC declaration to cover the first wave of the pandemic. In Spec. (1), we add firm fixed effects and dynamic subsector fixed effects. In Spec. (2), we introduce dynamic fixed effects for both subsector and country. In Spec. (3), we remove all controls and firm fixed effects to decrease the saturation of the model. To address potential problems with our identification strategy, in Spec. (4) we perform the estimation without stocks of firms traded on the U.S. market. Similarly, in Spec. (5), we perform the estimation without stocks of firms traded in developed markets. Statistical significance: ${}^{\ast }$ $p$-value $<0.10$; ${}^{\ast \ast }$ $p$-value $<0.05$; ${}^{\ast \ast \ast }$ $p$-value $<0.01$.