COVID-19, a blessing in disguise for the Tech sector: Evidence from stock price crash risk

Abstract

In this paper we document that although COVID-19 has brought uncertainties to the overall economy, the Technology (tech) sector is the systematic beneficiary of the pandemic. Using a quasi-natural setup, we find a significant notion that the Stock Price Crash Risk (SPCR) of firms within the Tech sector decreases during the COVID-19 pandemic compared to the recent past and firms belonging to other sectors. Our analyses further reveal that firms in the Tech sector with stronger external monitoring and better information environment receive an even greater advantage from the pandemic. Overall, our study suggests that the higher systemic dependency on the Tech sector during the COVID-19 outbreak results in an economic benefit for this sector.

“Nero fiddled while Rome burned!”

1. Introduction

Every corner of the world is still experiencing the devastating impact of COVID-19 in one way or another and will continue to do so. By now, most readers know that the socio-economic toll of this pandemic has already surpassed any previous unexpected events, including the two World Wars and past economic recessions (Nicola et al., 2020, Ding et al., 2021, Yeyati and Filippini, 2021). Given the uncertain nature of this pandemic, while it is obvious that many individuals, institutions, corporations, industries, and nations have experienced negative productivity shocks some groups have systemically benefited from COVID-19. Intuitively, there is always room for somebody to be advantaged even during uncertain times while many others suffer. The Corporate World is no exception. Many recent financial economic studies have discussed the impact of COVID-19 on various corporate issues (see e.g., Li et al., 2020; Ramelli and Wagner, 2020; Bae et al., 2021; Ding et al., 2021; Hasan et al., 2021, Hasan et al., 2021; Huber et al., 2021). However, if the strategic effects are strong (weak), the higher level of uncertainty could bring more (less) growth opportunities for corporations (Courtney et al., 1997, Kulatilaka and Perotti, 1998). Ample anecdotal evidence suggests (and understandably so) that the technology (hereafter tech) sector has experienced a boom rather than suffering a downturn during this period, most likely due to its competitive and strategic advantage over others.1 Particularly, the economy-wide extensive dependence on tech usage during the pandemic and the tech sector’s suitability for “work from home” may be two important catalysts behind this (Bai et al., 2021).

In this study, we investigate the issue of the tech sector’s exceptional advantage during COVID-19 by comparing the Stock Price Crash Risk (hereafter SPCR) of tech vs. non-tech firms during pre- and post-pandemic periods.2 Studying SPCR in this context is important. The stock market volatility during this pandemic has received increasing investor, academic, and policy attention (see e.g., Acharya et al., 2021; Huang and Liu, 2021; Huber et al., 2021; Ramelli and Wagner, 2020). The SPCR is a crucial accounting feature of the equity market and often serves as an indicator of the financial stability of a firm that investors, corporate managers, and policymakers alike find useful. Conceptually, SPCR originates from “managerial bad news hoarding” about a firm’s (unfavorable) financial health from the market that eventually reveals and results in a sudden but drastic drop of its stock price (Hutton et al., 2009, Kim et al., 2019, Kothari et al., 2009). Another reliable explanation of SPCR is built around the default (distress) risk of a firm indicating its incapacity to fulfill financial obligations, i.e., financial constraints (Andreou et al., 2021, He and Ren, 2022). From both perspectives, during growth time (such as for tech firms during the pandemic), managers usually have less incentive to conceal bad news from the market. In addition, the probability of default becomes less of a problem due to enhanced financial solvency (see e.g., He and Ren, 2022).3 Ding et al. (2021), among others, focusing on the COVID-19 pandemic, argue that a set of pre-pandemic corporate characteristics such as higher profitability, larger cash holdings, and a lower level of outstanding debt provide better corporate immunity to the financial markets. Moreover, tech firms are naturally more innovative, better monitored, and more transparent in the flow of information, making them less susceptible to SPCR (Hutton et al., 2009, Kim and Zhang, 2014). Altogether, we conjecture that the post-COVID-19 SPCR for tech firms would decline compared to their non-tech counterparts and pre-pandemic periods.

Our findings support this prediction. We follow the well-established SPCR literature (see Hutton et al., 2009; Kim et al., 2011a; Kothari et al., 2009) for our methodology. Focusing on the three most common SPCR measures (Negative Skewness [NCSKEW], Down to Up Volatility [DUVOL], and Probability of Crash [CRASH]), we document a significantly reduced SPCR for tech firms during the post-pandemic period (i.e., in the years 2020 and 2021), compared to the two preceding years (i.e., 2018 and 2019) and their counterparts (i.e., non-tech firms). This finding is also economically meaningful. Specifically, the magnitude of reduction in our SPCR measures, namely, NCSKEW, DUVOL, and CRASH, are comparable with 22.09%, 19.14%, and 29.02% to their respective standard deviations.

To address potential concerns with sample selection bias, covariate imbalance, and model misspecification, we further verify our baseline results using the sample created from the propensity score matching (PSM) technique. A point to be noted here is that our methodology has been designed as a quasi-natural experiment (i.e., a Difference-in-Differences set-up). From an econometric viewpoint, this design is superior to the traditional ones, given the outcome is less likely to be endogenously determined (see e.g., Doidge and Dyck, 2015; Glaeser and Guay, 2017; Gropp et al., 2019; Yang et al., 2021). Moreover, our test for parallel trend assumptions assures no preceding (i.e., pre-pandemic) trending difference in SPCR reduction for tech firms.

The external monitoring and information environment are two vital dimensions of the SPCR literature (Kim et al., 2016, Bauer et al., 2021). Therefore, we also investigate the cross-sectional implications of these two factors in our context and document consistent results with our predictions derived from the relevant literature (e.g., Kim et al., 2019; Ma et al., 2020). Specifically, our study finds that the post-pandemic reduction in SPCR for tech firms becomes even more evident with a higher level of external monitoring and a better information environment.4

Although the scope of our study is limited, we have been able to explore several important avenues of literature. First, our paper joins a big crowd of emerging literature that focuses exclusively on the abrupt impact of the COVID-19 pandemic on various financial-economic aspects. While most of the studies in this line of inquiry focus on the downside stories of the COVID-19 pandemic, we have been able to document a rare opportunity for tech firms that brings a certain amount of tension to the literature. Second, our study fits with the recently prevalent trend of quasi-natural experiment-based studies (e.g., Gropp et al., 2019; Xu et al., 2021; Yang et al., 2021) that are arguably less susceptible to endogeneity concerns. Third, our paper adds to the stream of SPCR literature (e.g., Chang et al., 2017; Chowdhury et al., 2020; Hutton et al., 2009; Kothari et al., 2009; Kim et al., 2011a and 2011b; 2014; 2016; 2019; 2020; Xu et al., 2021). Fourth, our study leaves its mark on established corporate governance issues, such as external monitoring and information opacity. Finally, our findings shed light on the superior financial performance of firms in the tech sector during the difficult period of the worldwide pandemic.

The remainder of this manuscript is designed as follows: Section 2 discusses the extant literature and develops the main and supporting hypotheses based on the current research gap for our empirical exploration. Section 3 describes the sample construction and methodology used in this study. Section 4 reports the empirical findings and justifications. Section 5 conveys a practitioner’s perspective on the context of this study. Section 6 concludes our study.

2. Literature review and hypotheses development

The concept of Stock Price Crash Risk (SPCR) is central to the temporal managerial bad news hoarding behavior. It mainly arises from the possibility of unanticipated poor firm performance (and thus poor valuation) if such bad news gets revealed to the market (i.e., to the shareholders and investors). In modern corporate settings, managerial incentives and job security are closely linked to firm performance (Bamber et al., 2010, Farrell and Whidbee, 2003, Graham et al., 2005, Roychowdhury et al., 2019). It is well established in the corporate governance literature (and understandably so) that managers have more information about firms than anyone else (see e.g., Diamond, 1985; Dye, 1985). Some insider information, if revealed, could persuade shareholders and investors to withdraw their investments from a company, resulting in upward pressure on financial constraints (and thus downward pressure on firm performance). In such a situation, managers are likely to feel uncomfortable transparently propagating bad news to the market, considering they could suffer a decline in incentives, and a possible threat to their job security (Graham et al., 2005, Roychowdhury et al., 2019). However, hoarding bad news from the market is very challenging for managers, particularly today with advances in information systems. Moreover, this conservative practice could produce worse results for firms if the market suddenly and unexpectedly learns about bad news that managers have wanted to conceal for a long time (Hutton et al., 2009, Kothari et al., 2009). Market participants usually conclude that there may be additional bad news that they are not yet aware of, and may accordingly over-react by trying to sell their stock immediately (Conrad et al., 2002, Defond and Zhang, 2014). When too many shareholders simultaneously react with such panic, the stock price drops drastically, experiencing a stock price crash (Kim et al., 2011a and 2011b; Chen et al., 2017). The financial market has experienced such events in the recent past, in the Enron scandal and the WorldCom bubble, for instance (Unerman and O’Dwyer, 2004). Therefore, a corporation’s susceptibility to1 such risk (i.e., SPCR) is closely monitored by other stakeholders, policymakers, the media, and academicians. From a policy perspective, firm managers have undertaken numerous initiatives to prevent SPCR; a remarkable example is the moderation of accounting standards by the Sarbanes-Oxley Act – SOX (Zhang, 2007). To date, the financial accounting literature provides numerous factors affecting SPCR. Broadly speaking, this literature looks into factors such as financial reporting and corporate disclosure (Chen et al., 2017, Francis et al., 2016, Hutton et al., 2009, Kim et al., 2011a), accounting standards (DeFond et al., 2015), corporate social responsibility (Kim et al., 2014), earnings manipulation (Graham et al., 2005), conservative accounting policies (Kim and Zhang, 2016), managerial characteristics and incentives (Kim et al., 2011b, Andreou et al., 2017), internal and external corporate governance mechanisms (Andreou et al., 2016, Callen and Fang, 2013, Chen et al., 2017, Kothari et al., 2009, Xu et al., 2014;), political connections (Lee and Wang, 2017), individual norms (Callen and Fang, 2015), institutional ownership (An and Zhang, 2013), external labor market incentives (Chowdhury et al., 2020), brand capital (Hasan et al., 2021, Hasan et al., 2021) and so on. Habib et al. (2018) provide an excellent review on the empirical SPCR literature and indicate a research gap in sector-specific and event-specific SPCR. Studies on sector- and event-specific SPCR are almost non-existing, Huang and Liu (2021) being one exception: focusing on the outbreak of COVID-19, they document a decrease in SPCR for Chinese energy firms.

Our study attempts to fill this research gap by extending Huang and Liu’s (2021) to the tech-sector. For several reasons, the nexus between COVID-19 and the tech sector provides an excellent research setup to study SPCR. First, the capital structure of the firms in the tech sector mainly relies more on the stock market (i.e., shareholders’ equity) for external financing than on debt since tech firms are essentially growth firms and require much research and development (Guiso, 1998, Brown et al., 2009 and 2013). As the name implies, SPCR mainly captures stock-market risk. Besides, many of the natural characteristics of tech firms such as the higher growth opportunities (Brown et al., 2009), innovativeness (Bao et al., 2012, Guiso, 1998), information transparency (Granados and Gupta, 2013; Hodge et al., 2004; Zhu, 2004), and less dependence on liquidity, relatively speaking (Guiso, 1998, Levitas and McFadyen, 2009) provide direct channels to SPCR (see e.g., An et al., 2018; Ben‐Nasr et al., 2021; Chang et al., 2017; Hong et al., 2017; Zaman et al., 2021) that suggest a reduced SPCR for tech firms. Second, one aspect of COVID-19 caused tech firms to experience a boom for obvious reasons, i.e., the increased overall dependence on higher technology during the early COVID-19 isolation months and a tech firm’s suitability to “work from home” – ensuring that their operations remained largely uninterrupted (Bae et al., 2021). Notably, the greater financial and/or operational solvency of tech firms during the pandemic took two essential forms. First, a skyrocketing demand for consumer tech products related to entertainment, electronic devices, virtual education, communication, and telemedicine, while maintaining social distance. Second, to facilitate the ability to work from home for tech and other sectors, required an enormous structural change, mainly with the help of tech facilities such as cloud computing platforms and extended cyber securities.5 Moreover, it is no secret that tech firms rely substantially more on intangible assets compared to their non-tech counterparts (Kwon and Yin, 2006, Skinner, 2008). Thus, while non-tech firms were primarily affected by the slow-down in supply chains process during the pandemic, this was less of an issue for many tech firms. Overall, COVID-19 provided tech firms with greater financial solvency (i.e., they had fewer financial constraints) – another remarkable inhibitor for the SPCR phenomena (Andreou et al., 2021, He and Ren, 2022). Intuitively, when a firm is financially solvent, its managers have less motivation to show “bad news hoarding behavior” given their higher incentives and job security, a result of the financial solvency of the firm) – suggesting a reduction in SPCR. Collectively, we can formulate the following main hypothesis for this study:

H1: “The stock price crash risk (SPCR) of Tech firms during post-COVID-19 period goes down compared to the pre-COVID19 period and the non-Tech firms.”

Moreover, the discussion above makes clear that the SPCR game takes the managerial tendency to hoard bad news as the key player, and the temporary stockpiling of bad news as the playground. The extant literature on SPCR thus often underpins the arguments and anchors the findings, relying primarily on crucial factors such as how closely the corporate managers are monitored and how transparent a firm’s information environment is (see e.g., Al Mamun et al., 2020; Hu et al., 2020; Kim et al., 2016, 2019; Kothari et al., 2009). Specifically, if a firm’s external monitoring mechanism is robust, the managers cannot easily attempt to conceal firm-specific bad news from the market because of their enhanced accountability to the board and the more intense monitoring. Similarly, if the corporate information environment is transparent enough, it becomes more challenging for managers to hide bad news from the market. Most importantly, tech firms tend to choose equity more than debt in their capital structure (see e.g., Brown et al., 2009, Brown et al., 2013). There is an established notion in the literature that firms with a lower (higher) information asymmetry choose to rely more on equity (debt) financing (see e.g., Bharath et al., 2009; Petacchi, 2015). Particularly during the pandemic, as tech firms experienced success in the equity market (Gompers et al., 2022), it seems plausible to conclude that tech firms have maintained a good information environment to attract investors in the equity market (see e.g., Cui et al., 2021). Moreover, the same line of arguments is possible in the case of external monitoring (in general and with particular focus on the pandemic) within the context of the comparison between tech and non-tech firms surrounding their respective capital structures and common characteristics (see e.g., Slovin et al., 1990; Wang et al., 2022). Accordingly, we can formulate two supporting hypotheses for this study as follows:

H1-a: “If the stock price crash risk (SPCR) of tech firms during the post-COVID-19 period goes down compared to the pre-COVID19 period and non-tech firms, it would be greater for firms with a stronger external monitoring mechanism.”

H1-b: “If the stock price crash risk (SPCR) of tech firms during the post-COVID-19 period goes down compared to the pre-COVID19 period and the non-tech firms, it would be greater for firms with a more transparent information environment.”

3. Data and methodology

Our sample period covers 2018–2021. For our study, the years 2018 and 2019 (2020 and 2021) are considered as the pre- (post-) COVID periods. We collect financial and accounting data from Compustat, and stock price and return data from CRSP. We exclude firms from the utility (SIC 4900 – 4999) and financial (SIC 6000 – 6999) sectors as they operate in highly regulated environments (e.g., Kim et al., 2014). We ensure that Compustat and CRSP data are available to construct all our control variables. All continuous variables are winsorized at 1st and 99th percentiles. Our final sample consists of 7115 firm-year observations.

Our main dependent variable, a proxy for SPCR, is the negative coefficient of skewness of firm-specific daily returns (NCSKEW). We construct this measure in accordance with extant literature such as Chen et al. (2001), Hong et al. (2017), and Kim et al. (2011a, 2011b). We first estimate the residuals of from the following augmented market model:

$$r_{i,t}=a_0+{{\beta }_{1i}r}_{\mathit{mt}-2}+{{\beta }_{2i}r}_{\mathit{mt}-1}+{{\beta }_{3i}r}_{\mathit{mt}}+{{\beta }_{4i}r}_{\mathit{mt}+1}+{{\beta }_{5i}r}_{\mathit{mt}+2}+{\epsilon }_{i,t}$$ (a)

where i denotes the firm, and w indicates the week, $r_{m,w}$ means the weekly value-weighted market return. Model (a) includes the forward and lag market returns and industry returns to capture non-synchronous trading (Dimson, 1979). The firm's specific weekly return is determined as $W_{i,w}=\log (1+{\epsilon }_{i,w})$, where ${\epsilon }_{i,w}$ is the estimated residual of the above model. We estimate our SPCR proxies based on this firm-specific weekly return ($W$). Specifically, NCSKEW captures any asymmetry in return distribution skewed to the left or right. It is calculated as the third moment of each week’s returns in the year that is normalized by the standard deviation raised to the third power. Empirically, for each firm $i$ in year $t$, NCSKEW is formulated as:

$${\mathit{NCSKEW}}_{i,t}=\frac{-[{n\left(n-1\right)}^{\frac{3}{2}}\Sigma W_{i,t}^3]}{[{\left(n-1\right)\left(n-2\right)}^{\frac{3}{2}}{(\Sigma W_{i,t}^2)}^{\frac{3}{2}}]}$$ (b)

where $W$ indicates residual weekly returns. A higher NCSKEW value indicates higher SPCR. We use two alternate proxies for crash risk. First, the down-up volatility of firm-specific daily returns (DUVOL). To calculate DUVOL, we first differentiate between the days (i.e., “up” or “down”) based on the returns if higher or lower than the period average and then we compute their standard deviations (SD) to come up with a ratio of SD on “down” day returns to the “up” day returns as follows:

$${\mathit{DUVOL}}_{i,t}=\log \{(n_u-1)\sum _{\mathit{Down}}{W}_{i,t}^2/(n_d-1)\sum _{\mathit{Up}}{W}_{i,t}^2\}$$ (c)

where $n_u$ and $n_d$ indicates the “up” and “down” days respectively. A higher DUVOL value indicates a higher SPCR. Second, we employ an indicator variable (CRASH) that equals one for a firm-year that experiences one or more crash weeks during the fiscal year, and zero otherwise. All these measures are commonly used in the published literature (e.g., Chang et al., 2017; Chen et al., 2001; Hutton et al., 2009; Kim et al., 2011a; Kothari et al., 2009).

We classify a firm as TECH if it belongs to group 6 or 7 under Fama-French 12 (FF12) industry classifications. The remaining firms are classified as non-TECH firms. Our multivariate analysis employs a difference-in-differences research design in which we compare the changes in SPCR in the post-COVID year for TECH firms compared to non-TECH firms. We estimate the following ordinary least square (OLS) regression specification:

(NCSKEW or DUVOL or CRASH)_it = α + β POST_i x TECH_i + γX_it-1 + Firm FE + Year FE + ε_it (1)

where i and t indicate firm and year, respectively, POST is an indicator variable which equals one if the year is 2020 or 2021, and zero otherwise. TECH is a dichotomous variable which equals one if the firm is from FF12 industries in group 6 or 7 and zero otherwise. We expect a significant negative coefficient for the interaction term because the TECH sector has seen a great boom during the post-COVID period. Since we include firm and time fixed effects (i.e., Firm FE and Year FE), we do not include TECH or POST indicators separately (as they are absorbed by the fixed effects (see e.g., Huang et al., 2020). We cluster standard errors by firm.

In Eq. [1], X represents a vector of control variables. We use lagged control variables as is customary in the crash risk literature (Hutton et al., 2009, Kim et al., 2020). This means that our controls are from the immediately preceding year of our observation for crash risk measures. Taking lagged controls also attenuates endogeneity issues (e.g., Adams et al., 2010; Brodmann et al., 2021; Hossain et al., 2021). Following extant literature (e.g., Hutton et al., 2009; Kim et al., 2011a), we use firm size in the form of the natural log of the prior year market value of equity (L.SIZE); prior year debt ratio as in long term debt to total assets (L.LEVERAGE); lagged market value of equity to book value of equity (L.MTB); prior year income before extraordinary items scaled by total assets (L.ROA); the standard deviation of prior year firm-specific weekly returns over the prior fiscal year (L.SIGMA); prior year’s average weekly return for a firm over the prior fiscal year (L.RETURN); and the prior year negative coefficient of skewness of firm-specific daily returns (L.NCSKEW) as controls.

4. Discussion of results

4.1. Summary statistics and Pearson correlations

Table 1 presents some commonly used summary statistics for the variables used in our empirical exercise. The mean values for NCSKEW _t, DUVOL _t, and CRASH _t are 0.053, 0.048, and 0.244, respectively, and are comparable with earlier similar setup studies such as Huang and Liu (2021). The statistics for control variables are also comparable to standard empirical financial-economic literature (e.g., the mean lagged values of leverage, market-to-book ratio, and size, respectively, are 22.3%, 4.727, and 7.27).

Table 1.

Summary Statistics.

Variables	N	Mean	S.D.	p10	p25	p50	p75	p90
NCSKEW	7115	0.053	0.876	-0.937	-0.428	0.003	0.479	1.109
DUVOL	7115	0.048	0.525	-0.614	-0.299	0.032	0.369	0.717
CRASH	7115	0.244	0.429	0.000	0.000	0.000	0.000	1.000
L.SIZE	7115	7.270	1.991	4.587	5.894	7.302	8.620	9.904
L.LEVERAGE	7115	0.223	0.182	0.000	0.046	0.207	0.351	0.476
L.MTB	7115	4.727	5.511	0.959	1.537	2.782	5.368	10.514
L.ROA	7115	-0.032	0.227	-0.304	-0.047	0.032	0.078	0.133
L.SIGMA	7115	0.051	0.029	0.022	0.030	0.043	0.064	0.090
L.RETURN	7115	-0.169	0.217	-0.394	-0.200	-0.091	-0.044	-0.023
L.DETURN	7115	0.009	0.101	-0.075	-0.029	0.002	0.038	0.103
L.NCSKEW	7115	0.092	0.905	-0.954	-0.430	0.033	0.534	1.236

Notes: This table reports some summary statistics: the mean, standard deviation (S.D.), 10th, 25th, 50th (i.e., median), 75th, and 90th percentiles for our main variables. Variable definitions are provided in Appendix 1.

Table 2 reports the Pearson correlations between the variables used in this study. In the extant SPCR literature, we find that the alternative crash risk variables are highly correlated (e.g., the correlation between NCSKEW and DUVOL is 90%, and between NCSKEW and CRASH it is 65%, the p-value is <0.01 in both cases). Importantly, consistent with our primary hypothesis – H1, we find that our key variable of interest, POST X TECH, is negatively and significantly correlated with all three SPCR proxies. Moreover, we find evidence for the other standard correlations (e.g., the correlation between size and leverage is significantly positive, matching the extant empirical corporate finance literature).

Table 2.

Pearson Correlations.

Variables	NCSKEW	DUVOL	CRASH	POST x TECH	L.SIZE	L.LEVERAGE	L.MTB	L.ROA	L.SIGMA	L.RETURN	L.DETURN
NCSKEW	1.00
DUVOL	0.90 * **	1.00
CRASH	0.65 * **	0.55 * **	1.00
POST x TECH	-0.03 * *	-0.02 *	-0.04 * **	1.00
L.SIZE	0.11 * **	0.11 * **	0.03 * **	0.07 * **	1.00
L.LEVERAGE	0.01	0.00	-0.03 * *	-0.02	0.27 * **	1.00
L.MTB	0.08 * **	0.08 * **	0.04 * **	0.13 * **	0.31 * **	0.13 * **	1.00
L.ROA	0.01	0.00	0.00	0.03 * *	0.34 * **	0.16 * **	-0.10 * **	1.00
L.SIGMA	-0.04 * **	-0.03 * *	-0.03 * *	0.03 * **	-0.51 * **	-0.15 * **	0.01	-0.58 * **	1.00
L.RETURN	0.04 * **	0.02 *	0.03 * *	-0.02	0.40 * **	0.13 * **	-0.03 * **	0.56 * **	-0.95 * **	1.00
L.DETURN	0.00	0.00	-0.02 * *	0.00	-0.03 * *	0.03 * **	0.02	-0.09 * **	0.32 * **	-0.35 * **	1.00
L.NCSKEW	0.00	-0.01	0.02	-0.02	0.01	0.04 * **	-0.07 * **	0.03 * **	-0.01	0.05 * **	0.01

Notes: This table reports the Pearson correlations between variables in our main model. * , * *, and * * mean p < 0.10, p < 0.05, and p < 0.01, respectively. Variable definitions are provided in Appendix 1.

4.2. Baseline results

Table 3 presents our main multivariate analysis results. Here we examine whether COVID-19 brought a blessing in disguise for the TECH industry in the form of a lower stock price crash risk. We estimate Eq. [1] using all three measures of SPCR. The interaction term POST x TECH coefficient is negative and significant at the 1% level (i.e., the p-value is <0.01) for all three SPCR measures. These findings suggest a remarkable decrease in SPCR for the TECH industry firms in the post-COVID period compared to their non-TECH counterparts. Moreover, these declines are not economically trivial. For example, the decrease in NCSKEW represents 22.09% of one standard deviation in NCSKEW (i.e., 19.35% of 0.876); the decrease in DUVOL represents 19.14% of one standard deviation of DUVOL (i.e., 10.05% of 0.525); and the decrease in CRASH represents 29.02% of its standard deviation (i.e., 12.45% of 0.429). These results indicate that the post-COVID boom in the TECH industry firms (probably due to the higher systemic dependence on them by other firms) led to a significant decline in their SPCR relative to their non-TECH counterparts, consistent with our primary hypothesis – H1.

Table 3.

Baseline Results.

DV=	NCSKEW	DUVOL	CRASH
Variable	(1)	(2)	(3)
POST x TECH	-0.1935 * **	-0.1005 * **	-0.1245 * **
	(−4.08)	(−3.74)	(−5.28)
L.SIZE	0.3998 * **	0.2464 * **	0.1185 * **
	(10.80)	(11.13)	(6.95)
L.LEVERAGE	-0.3038 *	-0.2915 * **	-0.1694 * *
	(−1.89)	(−3.09)	(−2.09)
L.MTB	0.0084	0.0058 *	0.0024
	(1.39)	(1.71)	(0.86)
L.ROA	0.3878 * **	0.1508 *	0.1371 *
	(2.61)	(1.72)	(1.90)
L.SIGMA	-8.5053 * **	-4.9799 * **	-5.3788 * **
	(−3.63)	(−3.58)	(−4.83)
L.RETURN	-0.3699	-0.2195	-0.3052 * *
	(−1.24)	(−1.26)	(−2.32)
L.DTURN	0.3607 * *	0.1918 * *	0.0881
	(2.50)	(2.23)	(1.23)
L.NCSKEW	-0.2188 * **	-0.1195 * **	-0.0623 * **
	(−16.93)	(−15.19)	(−9.38)
Constant	-2.4040 * **	-1.4638 * **	-0.3461 * **
	(−8.56)	(−8.64)	(−2.67)
Observations	7115	7115	7115
R2	0.426	0.417	0.374
Firm FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
Cluster	Firm	Firm	Firm

Notes: This table reports the baseline results of this study. The dependent variables in the three columns represent three of the conventional proxies of SPCR in the literature. The main variable of interest is POST x TECH. All our regressions employ firm fixed effects and year fixed effects. Standard errors are clustered at the firm-level. Control definitions appear in the Appendix 1. Values of t-statistics are in parentheses. * ** , * *, and * represent significance at 1% level, 5% level, and 10% level respectively.

4.3. Test for parallel trend assumption

Since we document a significant reduction in SPCR for tech firms relative to their non-tech counterparts by comparing pre- and post-COVID-19 periods following a quasi-natural experiment, it is crucial to keep in mind that this impact prevails only in the post-COVID-19 period. The extant literature often addresses such concerns by testing for a parallel trend assumption, i.e., if the impact from any event is true, there should be no trending difference preceding the event (see e.g., Chen et al., 2021; Fang et al., 2014; He and Tian, 2013; He et al., 2022; Huang et al., 2020; Ito, 2014; Lemmon and Roberts, 2010; Puri et al., 2011).

We closely follow Huang et al. (2020) for these tests. This set of tests explores the possibility that the post-COVID changes reflect trending differences between TECH and non-TECH firms prior to COVID, and do not result from COVID itself (i.e. it lacks causality). We report the results in Table 4. In these tests, 2018 (i.e. the first year of our sample which is year t-2) serves as the benchmark year. We find no trending differences between TECH and non-TECH firms before COVID (i.e. the year 2019). For example, the coefficient for D(t = 2019) x TECH is 0.0502 (t-stat. = 0.80) in Column (1). That means that between 2018 and 2019 there were no changes in SPCR between TECH and non-TECH firms. In other words, it shows that there are no preexisting differential trends in SPCR between TECH and non-TECH firms. The differences started to appear after COVID became a reality in the world (i.e. year 2020). For example, the coefficient for D(t = 2020) x TECH is − 0.1669 (t-stat = −2.68) in Column (1). This means that the differences in SPCR between TECH and non-TECH firms are only significant for post-COVID times compared with the pre-COVID year 2018. The same is true for second year of post-COVID (i.e. 2021).

Table 4.

Test for Parallel Trend Assumption.

DV=	NCSKEW	DUVOL	CRASH
Variable	(1)	(2)	(3)
D(t = 2019) x TECH	0.0502	0.0386	0.0244
	(0.80)	(1.08)	(0.78)
D(t = 2020) x TECH	-0.1669 * **	-0.0785 * *	-0.1246 * **
	(−2.68)	(−2.22)	(−3.97)
D(t = 2021) x TECH	-0.1657 * *	-0.0817 * *	-0.0887 * *
	(−2.29)	(−2.00)	(−2.50)
Constant	-2.4044 * **	-1.4668 * **	-0.3233 * *
	(−8.46)	(−8.57)	(−2.46)
Other controls as inTable 3	Yes	Yes	Yes
Observations	7115	7115	7115
R2	0.426	0.417	0.374
Firm FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
Cluster	Firm	Firm	Firm

Notes: This table reports the test for parallel trend assumption, i.e., looking for year by year trending difference in SPCR for tech firms. We anticipate insignificant SPCR coefficients of D(t = 2019) x TECH but significant SPCR coefficients of D(t = 2020) x TECH and D(t = 2021) x TECH. All our regressions employ firm fixed effects and year fixed effects. Standard errors are clustered at the firm-level. Control variables are as of Table 3 and the definitions appear in the Appendix 1. Values of t-statistics are in parentheses. ***, **, and * represent significance at 1% level, 5% level, and 10% level respectively.

4.4. Propensity Score Matching (PSM) results

It could be argued that the firms in TECH and non-TECH industries are quite different, and hence we are getting the results due to covariate imbalance or sample selection bias. The use of the Propensity Score Matching (PSM) sample to address such concerns is widespread in the extant empirical financial accounting literature, given that PSM samples are arguably free from such biases (see e.g., Abadie and Imbens (2016); Hirano et al. (2003); Hossain et al. (2022); Jha and Chen (2015); Li and Zhao (2006); Rosenbaum and Rubin (1983); Shipman et al. (2017); Smith and Todd (2001)). Accordingly, we double-check our baseline findings using the PSM sample. In determining the PSM sample, we first divide our main sample into two groups: TECH and non-TECH. If a firm belongs to the TECH (non-TECH) industry, we place it into the treatment (control) group. For each of the observations, we calculate the propensity score, i.e., the probability of belonging to the TECH group - using a logit model using all of the firm-level controls, and firm- and year-fixed effects as in our primary model (see Eq. [1]) as covariates. Then for each observation from the treatment group, we determine one observation from the control group by finding the nearest neighbor with caliper 0.01 and without replacement. These results are presented in Table 5. We show the comparison of the pre- vs. post-match firm characteristics in Columns (a) and (b) and baseline results (as of Table 3) based on the PSM sample in Columns (1), (2), and (3). Notably, the firm characteristics in the post-match sample are insignificant,6 and our results remain quantitatively and qualitatively similar to our main findings in Table 3. We consistently find that SPCR for TECH industry firms declined significantly (p < 0.01 for all three SPCR proxies) post-COVID compared to non-TECH firms.

Table 5.

Regression analysis with propensity score matched (PSM) sample.

DV=	TECH	TECH	NCSKEW	DUVOL	CRASH
Variable	(a)	(b)	(1)	(2)	(3)
	Pre-match logit	Post-match logit	Baseline OLS	Baseline OLS	Baseline OLS
POST x TECH			-0.2081 * **	-0.1160 * **	-0.1289 * **
			(−4.19)	(−4.07)	(−5.13)
L.SIZE	0.0475	0.0297	0.4552 * **	0.2949 * **	0.1380 * **
	(1.33)	(0.84)	(7.51)	(8.38)	(4.75)
L.LEVERAGE	-1.7678 * **	-0.0775	-0.0304	-0.1854	-0.1014
	(−5.99)	(−0.26)	(−0.11)	(−1.19)	(−0.72)
L.MTB	0.0604 * **	-0.0058	0.0103	0.0054	0.0028
	(7.30)	(−0.67)	(1.35)	(1.20)	(0.72)
L.ROA	0.8978 * **	0.1161	0.9194 * **	0.3625 * *	0.1054
	(4.03)	(0.44)	(3.32)	(2.27)	(0.76)
L.SIGMA	19.3074 * **	2.5324	-8.7855 * *	-3.0214	-4.7601 * *
	(3.06)	(0.37)	(−2.01)	(−1.21)	(−2.19)
L.RETURN	2.8219 * **	0.1675	-0.3542	0.0537	-0.2053
	(3.45)	(0.19)	(−0.60)	(0.17)	(−0.76)
L.DTURN	0.0457	0.1365	0.3696	0.1306	0.0697
	(0.14)	(0.39)	(1.59)	(0.99)	(0.59)
L.NCSKEW	-0.0311	0.0133	-0.2309 * **	-0.1413 * **	-0.0586 * **
	(−0.98)	(0.35)	(−10.87)	(−11.23)	(−5.32)
Constant	-1.9011 * **	-0.2742	-2.8992 * **	-1.9230 * **	-0.5251 * *
	(−5.05)	(−0.71)	(−6.37)	(−7.23)	(−2.40)
Observations	7115	3432	3432	3432	3432
R2	0.039	0.001	0.556	0.562	0.512
Firm FE	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes
Cluster	Firm	Firm	Firm	Firm	Firm

Notes: This table reports the baseline results of this study using the PSM sample. Columns (a) and (b) compares the pre- and post-matching coefficients on the control variables. The dependent variables in the columns (1), (2), and (3) represent three of the conventional proxies of SPCR in the literature. The main variable of interest is POST x TECH. All our regressions employ firm fixed effects and year fixed effects. Standard errors are clustered at the firm-level. Control definitions appear in the Appendix 1. Values of t-statistics are in parentheses. * ** , * *, and * represent significance at 1% level, 5% level, and 10% level respectively.

4.5. Cross-sectional analysis – role of external monitoring

Our first supporting hypothesis, H1a, suggests that our documented reduction in SPCR for tech firms during post-COVID-19 should be prominent with a higher external monitoring mechanism. To test this prediction, we investigate how the post-COVID decline in SPCR for the TECH industry fluctuates cross-sectionally with external monitoring by focusing on two critical aspects of external monitoring: institutional ownership and analyst following. Arguably, closer external monitoring (such as higher institutional ownership and analyst following) can deter managers from hoarding bad news and thus can reduce SPCR (Kim et al., 2016, 2019; Chowdhury et al., 2020). We divide the sample into high vs. low subsample groups based on the sample median of our respective moderating factors, i.e., the median percentage of top 5% institutional investors and the median number of analysts following. Specifically, observations with upper (lower) median values take the high (low) group. Data on institutional ownership comes from Thomson Reuters, and data on analyst following is from I/B/E/S. We present this set of cross-sectional results based on our main and PSM samples in Panels A and B of Table 6. Consistent with our predictions, high monitoring subgroups influence the decline in SPCR for the TECH industry firms. In other words, the coefficient of POST x TECH is significant only for high monitoring subgroups (i.e., those with a high level institutional ownership and analyst following for both regular and PSM subsamples).

Table 6.

Cross-sectional Analysis – Role of External Monitoring.

Panel A: Full sample
		Dependent Variable = NCSKEW
		Institutional Ownership				Analyst Following
		High		Low		High		Low
Variable		(1)		(2)		(3)		(4)
POST x TECH		-0.2076 * **		-0.0133		-0.1755 * *		-0.0074
		(−2.74)		(−1.01)		(−2.43)		(−0.67)
Constant		-2.5882 * **		-2.4327 * **		-2.7554 * **		-2.0551 * **
		(−5.53)		(−6.13)		(−4.94)		(−5.14)
Other controls as inTable 3		Yes		Yes		Yes		Yes
Observations		3500		3501		2880		3260
R2		0.466		0.515		0.456		0.485
Firm FE		Yes		Yes		Yes		Yes
Year FE		Yes		Yes		Yes		Yes
Cluster		Firm		Firm		Firm		Firm

Panel B: PSM sample
	Dependent Variable = NCSKEW
	Institutional Ownership				Analyst Following
	High		Low		High		Low
Variable	(1)		(2)		(3)		(4)
POST x TECH	-0.1716 * *		-0.0268		-0.1864 * *		-0.0259
	(−2.21)		(−0.72)		(−2.41)		(−0.77)
Constant	-2.9305 * **		-2.4111 * **		-2.9543 * **		-3.0405 * **
	(−3.91)		(−3.74)		(−3.45)		(−4.27)
Other controls as in Table 3	Yes		Yes		Yes		Yes
Observations	1687		1687		1458		1478
R2	0.620		0.617		0.561		0.632
Firm FE	Yes		Yes		Yes		Yes
Year FE	Yes		Yes		Yes		Yes
Cluster	Firm		Firm		Firm		Firm

Notes: This table reports results from our cross-sectional analyses where we examine the role of external monitoring (proxied by Institutional Ownership and Analyst Following). The variable of interest is POST x TECH. We anticipate the SPCR coefficients of POST x TECH to be negative and significant, and greater in magnitude for the High sub-groups than the Low sub-groups. For the sake of brevity, we only report results with NCSKEW as the proxy for SPCR. The unreported results using two other SPCR proxies (DUVOL and CRASH) are also consistent. Panel A presents results with the main sample, and Panel B presents the results with the PSM sample. All the high vs. low sub-sample splits are based on the respective median value of the sample. All our regressions employ firm fixed effects and year fixed effects. Standard errors are clustered at the firm-level. Control variables are as of Table 3 and the definitions appear in the Appendix 1. Values of t-statistics are in parentheses. * ** , * *, and * represent significance at 1% level, 5% level, and 10% level respectively.

4.6. Cross-sectional analysis – role of information environment

Our second supporting hypothesis, H1b, suggests that the post-COVID-19 decline in SPCR for tech firms would be prominent for firms having a better information environment. Since the concept of SPCR is central to the bad news hoarding behavior of corporate managers, if the information environment of a firm is transparent (i.e., less opaque), managers usually cannot easily conceal bad news from the market (Hutton et al., 2009, Kothari et al., 2009). Here, we examine the role of the information environment on the post-COVID decline in SPCR for the TECH industry by considering two famous metrics of information environment: (a) financial statement opacity as in Hutton et al. (2009), and (b) opacity proxied by analyst forecast error. Information opacity is one of the primary factors to aggravate SPCR (Ma et al., 2020). Therefore, we predict that the post-COVID decline in the TECH industry SPCR will be more pronounced for firms with low information opacity (i.e., a better information environment). In this analysis, we follow a similar empirical/methodological approach as in the previous Section (4.5) to examine the role of external monitoring in our context. Specifically, we first divide the sample into low vs. high opacity subgroups based on the median values of our moderating factors: Hutton Opacity and Analyst Forecast Error. Low sub-groups indicate lower information opacity and a better information environment for both opacity proxies. Therefore, we expect only the coefficients of POST x TECH for the low opacity sub-groups to be negative and significant. That is precisely what we find in this analysis. Notably, coefficients of POST x TECH are not significant for the high opacity sub-groups. We present the results with our regular and PSM samples in Panels A and B of Table 7. The Hutton Opacity Index is constructed following the Hutton et al. (2009) methodology, and Analyst Forecast Error is calculated as the absolute difference between forecast EPS and actual EPS as provided in I/B/E/S. Table 8.

Table 7.

Cross-sectional Analysis – Role of Information Environment.

Panel A: Full sample
	Dependent Variable = NCSKEW
	Hutton Opacity				Forecast Error
	Low	High			Low	High
Variable	(1)	(2)			(3)	(4)
POST x TECH	-0.2541 * **	-0.0579			-0.1643 *	-0.0939
	(−3.92)	(−0.72)			(−1.89)	(−1.56)
Constant	-2.6741 * **	-2.2677 * **			-2.8501 * **	-2.7195 * **
	(−6.30)	(−4.41)			(−5.19)	(−5.36)
Other controls as inTable 3	Yes	Yes			Yes	Yes
Observations	3525	3526			2897	3243
R2	0.504	0.551			0.588	0.558
Firm FE	Yes	Yes			Yes	Yes
Year FE	Yes	Yes			Yes	Yes
Cluster	Firm	Firm			Firm	Firm

Panel B: PSM sample
	Dependent Variable = NCSKEW
	Hutton Opacity			Forecast Error
	Low		High	Low			High
Variable	(1)		(2)	(3)			(4)
POST x TECH	-0.2056 * **		-0.0872	-0.1401 * **			-0.0216
	(−2.77)		(−1.14)	(−2.63)			(−1.36)
Constant	-2.3128 * **		-3.5760 * **	-3.5078 * **			-3.5035 * **
	(−3.54)		(−4.08)	(−3.33)			(−4.12)
Other controls as inTable 3	Yes		Yes	Yes			Yes
Observations	1701		1701	1383			1553
R2	0.595		0.679	0.728			0.674
Firm FE	Yes		Yes	Yes			Yes
Year FE	Yes		Yes	Yes			Yes
Cluster	Firm		Firm	Firm			Firm

Notes: This table reports results from our cross-sectional analyses where we examine the role of information environment (proxied by Hutton’s (2009) Opacity Index and Analyst Forecast Error). The variable of interest is POST x TECH. We anticipate the SPCR coefficients of POST x TECH to be negative and significant, and greater in magnitude for the Low sub-groups than the High sub-groups. For the sake of brevity, we only report results with NCSKEW as the proxy for SPCR. The unreported results using two other SPCR proxies (DUVOL and CRASH) are also consistent. Panel A presents results with the main sample, and Panel B presents the results with the PSM sample. All the high vs. low sub-sample splits are based on the respective median value of the sample. All our regressions employ firm fixed effects and year fixed effects. Standard errors are clustered at the firm-level. Control variables are as of Table 3 and the definitions appear in Appendix 1. Values of t-statistics are in parentheses. * ** , * *, and * represent significance at 1% level, 5% level, and 10% level respectively.

Table 8.

The Post-COVID-19 SPCR for Other Industries.

Panel A: Full sample
Dependent Variable = NCSKEW
Variable	(1)		(2)		(3)		(4)		(5)		(6)			(7)		(8)
	Consumer Non-Durables		Consumer Durables		Manufacturing		Energy		Chemicals		Retail			Healthcare		Miscellaneous
POST	-0.3981 * **		-0.1548		-0.1078		0.1574 *		-0.1702		-0.0609			-0.3453 * **		-0.2197 * **
	(−4.06)		(−1.33)		(−1.48)		(1.78)		(−1.43)		(−0.89)			(−5.15)		(−3.85)
Constant	-4.8688 * **		-1.3855		-1.5486 *		-2.3189 * *		-3.8047 * *		-1.3759 *			-3.3753 * **		-3.3887 * **
	(−3.16)		(−0.86)		(−1.92)		(−2.31)		(−2.20)		(−1.75)			(−5.48)		(−4.24)
Other controls as inTable 3	Yes		Yes		Yes		Yes		Yes		Yes			Yes		Yes
Observations	400		222		898		310		242		826			1379		1096
R2	0.475		0.437		0.401		0.416		0.374		0.416			0.477		0.456
Firm FE	Yes		Yes		Yes		Yes		Yes		Yes			Yes		Yes
Year FE	Yes		Yes		Yes		Yes		Yes		Yes			Yes		Yes
Cluster	Firm		Firm		Firm		Firm		Firm		Firm			Firm		Firm

Panel B: PSM sample
Dependent Variable = NCSKEW
Variable		(1)		(2)		(3)		(4)		(5)		(6)	(7)		(8)
		Consumer Non-Durables		Consumer Durables		Manufacturing		Energy		Chemicals		Retail	Healthcare		Miscellaneous
POST		-0.6551 * **		-0.0031		-0.1376		0.5179 *		-0.1830		-0.2000	-0.2024 * *		-0.1812
		(−3.54)		(−0.01)		(−1.13)		(1.91)		(−1.57)		(−1.03)	(−2.15)		(−1.13)
Constant		-7.6505 * **		-0.6307		-4.5999 * **		2.4310		-1.7902		0.3720	-1.9894		-5.2637 * *
		(−2.87)		(−0.13)		(−2.63)		(0.39)		(−0.60)		(0.17)	(−1.32)		(−2.36)
Other controls as inTable 3		Yes		Yes		Yes		Yes		Yes		Yes	Yes		Yes
Observations		135		70		277		83		73		262	484		332
R2		0.763		0.789		0.722		0.777		0.901		0.664	0.721		0.661
Firm FE		Yes		Yes		Yes		Yes		Yes		Yes	Yes		Yes
Year FE		Yes		Yes		Yes		Yes		Yes		Yes	Yes		Yes
Cluster		Firm		Firm		Firm		Firm		Firm		Firm	Firm		Firm

Notes: This table reports post-COVID-19 SPCR for various non-tech sectors (according to Fama-French-12 industry classification) except for the financials and utilities. The variable of interest is POST. For the sake of brevity, we only report results with NCSKEW as the proxy for SPCR. The unreported results using two other SPCR proxies (DUVOL and CRASH) are also consistent. Panel A presents results with the main sample, and Panel B presents the results with the PSM sample. All our regressions employ firm fixed effects and year fixed effects. Standard errors are clustered at the firm-level. Control variables are as of Table 3 and the definitions appear in the Appendix 1. Values of t-statistics are in parentheses. * ** , * *, and * represent significance at 1% level, 5% level, and 10% level respectively.

4.7. The Post-COVID-19 SPCR for other industries

As we document a decline in SPCR for tech industry firms during the post-COVID-19 period compared to their non-tech counterparts, readers may naturally be curious about the post-COVID-19 SPCR phenomena for the non-tech firms. In this study, we follow Fama-French-12 (FF-12) industry classifications where tech sector include categories 6 and 7. We do not consider financials and utilities for this study since they are highly regulated. In this section, we report the post-COVID-19 SPCR for the remaining eight industry categories under FF-12: Consumer – nondurables, Consumer – durables, Manufacturing, Energy, Chemicals, Retail, Healthcare, and Miscellaneous.

We find a significant decline in SPCR for firms in Consumer - nondurables and Healthcare sectors, but a significant increase in SPCR for firms in the energy sector using both our regular and PSM samples. These findings have at least two important implications. First, similar to the tech industry, firms from industries such as healthcare (e.g., hospitals, pharmaceuticals, etc.) and nondurable consumers (e.g., foods, beverages, etc.) also experienced a boom after the outbreak of COVID-19 because of an increasing economy-wide dependency on them. Second, in contrast to Huang and Liu (2021), we find that the post-COVID-19 SPCR for energy-sector companies decreases. The most plausible explanation for this interesting finding is that Huang and Liu (2021) focus on Chinese energy firms while our evidence comes from US energy firms. The use/demand of energy in the US and Chinese economies are different, and the policies undertaken to manage the post-COVID-19 situation were vastly different in both economies. Intuitively, China’s most significant share of energy is used for manufacturing purposes, which was only marginally interrupted during the post-COVID-19 period in China. However, a significant share of energy in the US is used for transportation and commercial purposes, and such sectors were largely interrupted by COVID-19.

5. A practitioner’s perspective

During our exercise, we spoke with a partner who is a leading IT consultant at a BIG4 accounting firm, and he agrees with the validity of our findings. When we asked him why the tech sector did well, his answer was: “The pandemic has been a catalyst for organizations to invest in digital channels, which had a flow-on effect for the tech sector. Effectively we saw a significant acceleration of plans and digital move from a supplementary channel for many organizations. In addition, organizations needed to move to new ways of working, again with the flow-on effects. The result is that we leaped forward many years in a short period of time.” When we asked if he thinks it is sustainable over the long term, he replied: “We’re clearly in a bubble right now, which I think is partially influenced by the availability of capital and wider economic factors. I don’t see us going backward in terms of adoption, but I also do expect to get to a more sustainable state in the medium term without necessary having a full-blown pullback.” Finally, we asked him if he believed that there is more transparency in the tech sector, and he answered: “No, I don’t. That’s not to say that I don’t wish that there was. Technology tends to reward scale and extraction of rents. Transparency doesn’t necessarily align with this.” This last answer was interesting. It indicates that the crash risk went down, not necessarily due to additional transparency, but because the tech sector was profitable and had no need bad news hoarding. Our cross-sectional findings also complement this. If the tech sector were very transparent, we would not have found any difference between those with a high versus low information environment. But that was not the case. We found a reduction in SPCR is more prominent for firms with a better information environment.

6. Conclusion

The COVID-19 pandemic has already taken the most extensive social, emotional, and economic toll in the history of humanity. Unlike many others, tech firms have enjoyed substantial growth during this pandemic, most likely because of an unusually high economic dependency of all other sectors on technology. Moreover, several other unique characteristics of tech firms have allowed them to be the systematic beneficiary of COVID-19. This study shows that Stock Price Crash Risk (SPCR) of tech firms declined during the COVID-19 pandemic compared to the prior years and to all other firms. In the cross-section, we show that a higher level of external monitoring and a better information environment help the further decline of SPCR for tech firms in the post-pandemic period. The design of our methodology is robust, and the outcome is credible, consistent, and supported by the extant literature. While most recent studies have highlighted the dark-side stories of COVID-19, our study reveals a situation in which COVID-19 proved a blessing in disguise for tech firms. Finally, our study contributes to the literature on several fronts with some important policy implications.

Author Statement

The authors declare that there is no conflict of interest.

Appendix 1. Variable definitions

Variables	Definitions	Sources
NCSKEW	NCSKEW is the negative skewness of the firm-specific weekly returns over the fiscal-year period.	CRSP
DUVOL	DUVOL is the log of the ratio of the standard deviations of down-week to up-week firm-specific returns.	CRSP
CRASH	An indicator variable that takes the value of one for a firm-year that experiences one or more firm-specific weekly returns falling 3.2 standard deviations below the mean firm-specific weekly returns over the fiscal year, with 3.2 chosen to generate frequencies of 0.1% in the normal distribution during the fiscal-year period, and zero otherwise.	CRSP
TECH	An indicator variable that takes the value of one for a firm if it belongs to group 6 or 7 under Fama-French 12 (FF12) industry classifications, and zero otherwise.	COMPUSTAT
POST	An indicator variable that takes the value of one for the years 2020 or 2021, and zero for the years 2019 or 2018.
SIZE	Natural logarithm of total assets.	COMPUSTAT
LEVERAGE	The ratio of total leverage and total assets.	COMPUSTAT
MTB	The ratio of the market value of equity to the book value of equity of a firm in a given fiscal year.	COMPUSTAT
ROA	The ratio of income before extra-ordinary income and total asset.	COMPUSTAT
SIGMA	The standard deviation of firm-specific daily returns over the fiscal year.	CRSP
RETURN	The mean of firm-specific daily returns over the fiscal year.	CRSP
DTURN	Changes in the average monthly share turnover over the fiscal year where monthly share turnover is the monthly share trading volume divided by the number of shares outstanding over the month.	CRSP

Appendix 2. Proof of PSM covariate balancing

Control Variables	Treatment TECH= 1 (n = 1716)	Control TECH= 0 (n = 1716)	Differences in means	t-stat
L.SIZE	7.4353	7.3814	0.0539	0.76
L.LEVERAGE	0.1949	0.1963	-0.0014	-0.23
L.MTB	5.7642	5.9495	-0.1853	-0.84
L.ROA	-.00779	-0.0136	0.0058	0.82
L.SIGMA	0.0486	0.0489	-0.0003	-0.38
L.RETURN	-0.1480	-0.1501	0.0021	0.36
L.DETURN	0.0051	0.0042	0.0009	0.28
L.NCSKEW	0.0750	0.0576	0.0174	0.57

Notes: This table reports the proof of covariate balancing for the PSM sample we use side-by-side with our primary sample to confirm all our main results.

Data availability

Data will be made available on request.