Staring Death in the Face: The Financial Impact of Corporate Exposure to Prior Disasters

We examine how rms’ exposure to prior disastrous events can inuence their stock market footprint during the coronavirus crisis. While others have drawn comparisons between past pandemics and Covid-19, we argue that such comparisons are skewed due to the unprecedented reach and consequences of the latter. To better model the structural shock caused by Covid-19 in the USA, we look at the 9/11 terrorist attacks and specically examine how rms based in New York City back then reacted to the associated nan-cial turmoil. While 9/11 and Covid-19 are categorically different events, their short-term impacts on the stock market, and on New York exchanges in particular, are surprisingly similar. We nd rms that nancially ‘survived’ 9/11 also managed to do better – or suffer less – by about 7% in terms of stock returns during Covid-19, compared to control rms that were not exposed to 9/11. In a sense, we show that companies’ prior exposure to 9/11 partly ‘immunized’ them against the consequences of a similarly destabilizing event, albeit two decades later. Interestingly, the trading volume of exposed rms increased due to market buying pressures. Our analysis is robust to various nancial proxies, alternative denitions of control rms and varying estimation windows.


Introduction
It is difcult to imagine an event in the past few decades with similarly wide-reaching and catastrophic consequences for individuals, society and the economy as Covid-19.The aggregate economic damage caused by Covid) is already well documented by international organizations such as the IMF and the World Bank, as well as national statistics agencies around the world. 1 Indeed, a signicant part of the economic turmoil caused by Covid is intermediated through loss of productivity and lower or negative earn-1 See the various factsheets available at the IMF Covid database (https://www.imf.org/en/Topics/imf-and-covid19) and the World Bank Covid database (https://www.worldbank.org/en/who-we-are/news/coronavirus-covid19).
ings in the corporate sector.Research published in nance journals since the emergence of Covid similarly attests to its devastating impact on rms' corporate nancial variables.Acharya and Steffen (2020) show that rms drew down bank credit lines and consistent with the risk of becoming a fallen angel, the lowest-quality BBB-rated rms behaved more similarly to non-investment-grade rms.Halling, Yu and Zechner (2020) discuss how Covid affected rms' access to public capital markets.Bond issues increased substantially for all bond types, while equity issues slowed tremendously.Ramelli and Wagner (2020) argue that rms more exposed to trade with China underperformed.Further, corporate debt and cash holdings emerged as important value drivers.Salisu and Vo (2020) show that stock returns were very sensitive to any health news related to Covid.
Our study attempts to ll an important gap in this body of research.While certain sectors of the economy, such as travel, tourism and hospitality, suffered more heavily during Covid than others (Izzeldin ., 2021), it is not sufciently clear et al what cross-sectional variations exist among rms in each industry sector.In other words, after controlling for known factors such as industry characteristics and nancial position that often contribute to cross-sectional variations in corporate responses, it would be theoretically interesting, and practically of signicant importance, to explore which particular rms showed more nancial resilience during the Covid disruption.Hence, we use the Covid disruption as an empirical setting to examine corporate nancial resilience, and the common factors that contribute to it.
Crucially, we provide evidence suggesting that rms' exposure to prior disasters in the corporate world makes them more resilient in the face of new but fundamentally similar disruptions.Specifically, we examine companies headquartered in New York City (henceforth, NYC) and trading in one of the city's three stock exchanges.Among such rms, we focus on those that were active both during the 11 September 2001 terrorist attacks (henceforth, 9/11) and the 2020 Covid period.Importantly, we show such rms displayed more nancial resilience during the Covid turmoil compared to control groups.Our ndings show that the stock price losses of these companies during Covid were about 7% lower compared to rms that were not exposed to 9/11.This gure is both statistically and economically signicant, and represents billions of dollars of market value 'saved' compared to the control group.
In other words, the group of rms which were exposed to 9/11 are found to be, in some sense, immunized or more immune to the nancial hit of Covid compared to their peers.We argue that this may be due to such rms having learnt how to cope with a sudden shock to their workforce, management (given their central location in NYC), ofce space, supply chains (see e.g.Harland, 1996;Verbeke, 2020), need for urgent crisis management, stakeholder communication and, of course, shocks to their share price and associated nancial metrics.Thus, their resilience is a product of organizational learning and internalizing this learning in their organizational culture (e.g.Walsh and Ungson, 1991).It is also plausible that the investors of such rms may 'price in' the fact that they have learnt their lesson and, therefore, regard them as more resilient against systemic shocks of a comparable nature. 2  Organizational learning from disasters (Smith and Elliott, 2007) initially takes place at the level of senior managers who have to reght the disaster at hand, and then trickles down the organization.Disasters, by denition, are 'serious disruptions of the functioning of a community or a society at any scale due to hazardous events interacting with conditions of exposure, vulnerability and capacity, leading to one or more of the following: human, material, economic and environmental losses and impacts' (UNGA, 2017).Prior literature also refers to 'surprises' (e.g.Bechky and Okhuysen, 2011;Lampel and Shapira, 2001), 'rare events'(e.g.Lampel, Shamsie and Shapira, 2009;Starbuck, 2009), 'catastrophes' (e.g.Majchrzak, Jarvenpaa and Hollingshead, 2007) or 'crises' (e.g.Rerup, 2009).While not the direct focus of our study, the question of which managerial skills and attributes may facilitate more effective decision-making and subsequent learning in disaster situations (Akinci and Sadler-Smith, 2019) is an interesting research focus in and of itself (Amabile and Pratt, 2016).
Particularly with unprecedented disasters that require low-probability yet high-consequence decisions, and where situations with no similarities to previous experiences arise, adaptiveness and agility can be demonstrated in the form of initial situational assessment followed by mental simulation and consultation (Curnin, Brooks and Owen, 2020).The learning process is driven by two key cognitive functions.Firstly, expert intuition, domain-specic learning and experience (Salas, Rosen and DiazGranados, 2010) and secondly, rational, analytical thinking -see, inter alia, the 4Is organizational learning framework (Intuiting, Interpreting, Integrating, Institutionalizing) of Crossan, Lane and White (1999).
While Covid and 9/11 are categorically different disasters, for our purposes, the comparison between them is appropriate for several reasons.Both these events were exogenous, unforeseen and extremely destabilizing.As far as the USA is concerned, NYC was very severely hit and was in fact the epicentre of both these disasters.And the same goes for rms headquartered in NYC or trading in that city.Both events shocked investors and the stock markets, and both had devastating effects on supply chains, although far more short-lived in the case of 9/11.Air travel was similarly suspended during 9/11, albeit again for a much shorter period, and consequences for the tourism and hospitality sectors were similarly grave.As a recent Financial Times article puts it, 'the industry in late 2001 experienced many of the ills it is seeing now.Airlines bled cash.Their survival was threatened.Government stepped in with nancial support, as they are doing today' (Skapinker, 2020). 3urther, we nd that the difference between rms exposed to 9/11 and their peers is not limited to their stock price reaction.The immunized rms happen to outperform their peers in a clean difference-in-difference estimation of both stock returns and market-adjusted excess returns.In fact, controlling for overall risk, immunized rms earn 14% higher raw returns and 15% higher excess returns compared to the control group.When we control for Covid-specic risks, these gures go down only to 13%, which is still a considerable difference statistically and economically.In addition, the trading volume of the immunized rms increases due to buying pressures in the market, again compared to their peers.These results are consistent across various industry sectors, as explained in the main empirical section of the paper.
This study makes several important contributions to the nance and management literatures.Firstly, our ndings contribute to the literature on stock market reactions to the spread of diseases.For example, McTier, Tse and Wald (2013) explain how the US market reacts to inuenza through time but do not highlight any patterns in the crosssectional variation among rms.Our paper con-tributes to this body of work by highlighting a novel relationship between rms' prior disaster exposures and their market resilience during health pandemics.
Secondly, our ndings contribute to the literature on stock market reactions to terrorist eventssee, for example, Chesney, Reshetar andKaraman (2011), Karolyi (2006) and Nikkinen and Vähämaa (2010) -by showing that, at least in the case of 9/11, these shocks can make the surviving rms more resilient and their investors more forgiving or trusting once similar disasters strike again.
Thirdly, these results provide novel evidence that markets have long-term memory -see, for example, Lo (1991).This means that markets can 'price in' the success or failure of rms in extreme events, even after a couple of decades.
Fourthly, and equally importantly, we contribute to the management literature on organizational learning and memory by showing that prior exposure to unprecedented and traumatic events can have organizational learning and resilience benets over the long term.The impact of organizational learning capabilities on a company's prospects for survival are widely documented in the management literature (see e.g.Argyris and Schön, 1996;Camps and Luna-Arocas, 2012). 4 Our study contributes to this literature by highlighting the role of prior exposure to disasters in triggering organizational learning.While we cannot distinguish incremental from radical innovation in our rms, we nonetheless can best understand the key ndings through the lens of companies 'learning' how to respond to systemic shocks of a disastrous nature, even if they are few and far between.The organization's response to these systemic shocks, when initially confronted with them, becomes integrated in organizational culture and work processes that constitute its 'organizational memory' (e.g.Walsh and Ungson, 1991).
The remainder of the paper is organized as follows.The second section discusses the relevant 4 Examples from the nancial services industry include Morgan and Turnell (2003), who show that the market information processing and analytical capabilities of organizations improve when they exhibit more favourable learning values.Chiva, Ghauri and Alegre (2014)  literature and motivates our theoretical hypotheses.The third section presents the data sample, data sources and denitions of variables.In the fourth section, we present the empirical approach and headline ndings.The fth section conducts a range of robustness tests which mainly conrm our core results, and the sixth section concludes.

Related literature and hypothesis development
Despite the recent emergence of Covid, there is already a substantial and fast-growing body of work on its nancial aspects and implications.Altman (2020) shows that the non-nancial corporate debt market in the USA reached a record percentage of gross domestic product.Further, investor appetite grew for higher promised yields on risky xed-income assets.Baker et al. (2020aBaker et al. ( , 2020b) ) argue that Covid resulted in a year-on-year contraction in US real GDP of nearly 11% as of 2020 Q4.Guerrieri et al. (2020) illustrate that standard scal stimulus was less effective than usual, and monetary policy, unimpeded by the zero lower bound, had magnied effects by preventing rm exits.Chronopoulos, Lukas and Wilson (2020) nd that discretionary spending declined throughout the pandemic.Favero, Ichino and Rustichini (2020) show that policies of epidemic containment were efcient with respect to the number of fatalities and GDP loss.Gormsen and Koijen (2020) discuss that dividends shrank throughout the pandemic, and scal stimulus boosted the stock market and long-term growth but did little to increase short-term growth expectations.And Mamaysky (2020) argues that markets frequently reacted to uninformative news in the early stages of the pandemic.As far as pandemics go, the health impact and economic footprint of Covid are unprecedented, at least as far back as the inuenza pandemic of 1918.However, some prior studies have given warnings about these expected economic costs in foresight.For example, Bloom, Cadarette and Sevilla (2018) discuss the costs such pandemics incur to both public and private health organizations, as well as losses to workforce productivity and disruption caused by social distancing.
In a similar vein, Fan, Jamison and Summers (2018) estimate the expected annual losses from pandemics to be around 500 billion USD, which comes to about 0.6% of global income, which -with the benet of Covid hindsight -appears to be a great underestimation.Similar studies emphasizing the need to anticipate and manage the economic consequences of pandemics include Lewis (2001), Tam, Khan and Legido-Quigley (2016) and several others -see Goodell (2020) for a summary.Of notable mention is the World Health Organization's Global Preparedness Monitoring Board (2019) report, which warns, only 3 months before the outbreak of Covid, that the world is at imminent threat of a global pandemic with little or no precaution being undertaken.
Another strand of the epidemics and pandemics literature (Page, Song and Wu, 2012) compares their nancial hit to other forms of natural and man-made disasters.These can include various natural disasters (Toya and Skidmore, 2007), air crashes (Ho, Qiu and Tang, 2013) and acts of terrorism (Llussa and Tavares, 2011).In particular, research on the nancial market impact of terrorist attacks can provide some form of parallel.While terrorist events are localized in their initial manifestation, they are by their nature designed to create a widespread shift in public mood (Goodell, 2020) and by implication, investor sentiment.This is an angle through which one can compare the market impact of a pandemic in its early days with that of a terrorist attack.
The associated spillover effects of terrorist attacks are, , discussed by Karolyi (2006), inter alia who concludes that, with some caveats, spillovers indeed occur as evidenced in tests examining volatility or beta risks with asset-pricing models.As regards 9/11, Choudhry (2005) examines whether the attacks created a shift in market betas.Similarly, Hon, Strauss and Yong (2004) nd that the 9/11 attacks led to higher correlations in global markets, not unlike the Covid episode, with some geographic variations.Other studies with similar ndings include Boin (2004), Chesney, Reshetar and Karaman (2011) and Nikkinen and Vähämaa (2010).
Organizational resilience can be broken down into three components, namely, anticipation, coping and adaptation (Ducheck, 2020).We propose that while companies exposed to the 9/11 shock were not necessarily superior in anticipating the Covid shock, they were better at coping with its associated tremors -such as its effects on their supply chain, workforce and customers.It is also likely that such companies will adjust and adapt better once the Covid chapter is completely closed, although we cannot formally test this as Covid is still an unfolding though near-ending turmoil.
Therefore, we hypothesize that prior exposure to disastrous events (terrorist attacks in particular) can result in rms having a 'softer landing' when a new unforeseen event of similar nature hits them.In other words, we expect nancial markets to remember and 'price in' a company's prior successful experience with disastrous events that threaten their supply chains and workforce, among other things.Therefore, we form the following testable hypothesis: H1: Firms with prior exposure to disastrous events experience smaller stock market losses during the Covid pandemic.
Also, as explained above, we expect such rms to be in more demand by the market.This, we would anticipate, is because investors value the advantageous position of these immunized rms relative to their peers.Hence: H2: Firms with prior exposure to disastrous events have higher trading volumes due to larger buying pressures from the market.
Later, we test these hypotheses together with a range of alternative variations and a battery of robustness tests.

Data sample
The period of our analysis runs from 9 December 2019 to 30 April 2020.The rst ofcial Covid cases in NYC date back to 2 March 2020, and the deaths start soon after. 5Figure 1 shows that the daily new Covid cases in NYC have a slow progress until mid-March, when they jump from 620 to 2,122, and keep rising until 6 April to 6,367, the peak of the rst wave in NYC.The daily new cases drop to 1,004 by 30 April.The daily new deaths have a similar distribution.They start on 11 March and peak at 590 on 7 April.The number of daily new deaths drops to 56 by 30 April.The time interval for this study includes 2 months as the treatment period and 3 months before 2 March 2020 as the control period.
To determine the rms in our sample, immune we consider companies with headquarters (HQ) in NYC that were traded in three major stock markets in NYC, that is the New York Stock Exchange (NYSE), NYSE American and NASDAQ stock market, between 10 September 2001 and 30 November 2001 -a 3-month period.We aim to focus on rms that have experienced the 9/11 terrorist attacks at rst sight, with the highest exposure.Of the rms above, only 114 were still traded during the Covid period, thus resulting in a set of 114 immune rms.For the control sample, we include rms with HQ in NYC that were traded on the NYSE, NYSE American or NASDAQ during the Covid period but not during the 9/11 period.These are the control rms that have not experienced the 9/11 shock before but are exposed to Covid.We have 331 control rms in the sample.
Overall, the sample includes 445 rms and 43,124 rm-day observations.Figure 2 presents the breakdown of immune and control rms by four-digit SIC industries.Both groups have a similar distribution of rms across industries.The majority of immune and control rms operate in the Finance sector (60% and 57%, respectively).Services and Healthcare are the next big industries for both immune and control groups.These three sectors correspond to about 80% of immune (76% of control) rms.Remaining minor industries include Consumer Goods, Communication, Utilities and Others.

Firm and market variables
Table 1A describes the variables used in this paper.We obtain daily data on rm stock prices, traded volume and dollar volume for publicly traded US rms from CRSP.Excess returns are dened as the daily returns in excess of the risk-free rate that is proxied by the 1-month T-bill rate.Market activ-ity is measured using three different variables: the daily average traded volume of shares, dollar volume in US dollars and a signed version of traded volume calculated as the product of the realized daily returns and the daily average traded volume.While the former two represent a proxy for the aggregate fund ows that come into the marketplace, the latter one gives a sense of the direction of trading activity.The signed traded volume takes a positive (negative) value if there is buy (sell) pressure in the market (see e.g.Campbell, Grossman and Wang, 1993;Llorente et al., 2002;Tosun, 2021).As part of our robustness tests, we follow Mc-Tier, Tse and Wald (2013) and construct the daily change in natural logarithm of the traded volume and the dollar volume as dependent variables.To mitigate the inuence of outliers, all variables are winsorized at the 1st and 99th percentiles.
We use aggregate risk factors, such as market risk, size, value, investment opportunities and protability (see e.g.Fama and French, 2015) in our models.In unreported tests, we also control for momentum following Carhart (1997) and obtain virtually similar and robust results.We obtain factor-mimicking portfolios that proxy for these risk factors on a daily basis from the Kenneth R. French online library. 6The daily change in these The daily dummy variable equal to 1 between 2 March and 30 April 2020, and 0 between 9 December 2019 and 28 February 2020.

Return
The daily stock return as a percentage.

Excess Return
The daily stock return in excess of the risk-free rate that is proxied by the 1-month T-bill rate.

Traded Volume
The amount of shares traded daily, in millions.

Dollar Volume
The amount of shares traded multiplied by the daily closing price, in million USD.

Signed Volume
The amount of shares traded multiplied by the daily stock return, in tens of thousands.Delta Ln(Traded Volume) The daily change in natural logarithm of the traded volume.Delta Ln(Dollar Volume) The daily change in natural logarithm of the dollar volume.

Market Value
The daily closing price multiplied by common shares outstanding, in billion USD.

Mktrf
The daily market return in excess of the risk-free rate that is proxied by the 1-month T-bill rate (Fama and French, 2015).SMB Small minus big (SMB) is the average return on the nine small stock portfolios minus the average return on the nine big stock portfolios (Fama and French, 2015).HML High minus low (HML) is the average return on the two value portfolios minus the average return on the two growth portfolios (Fama and French, 2015).RMW Robust minus weak (RMW) is the average return on the two robust operating protability portfolios minus the average return on the two weak operating protability portfolios (Fama and French, 2015).CMA Conservative minus aggressive (CMA) is the average return on the two conservative investment portfolios minus the average return on the two aggressive investment portfolios (Fama and French, 2015).Overall Risk Following Hassan .( 2019), this risk measure relies on word counts that condition on proximity to et al the use of synonyms for 'risk' or 'uncertainty'.This measure counts the frequency of mentions of synonyms for risk or uncertainty, divided by the length of the transcript.Covid-19 Risk Following Hassan . (2020), this risk measure relies on word counts that condition on proximity to et al the use of synonyms for 'risk' or 'uncertainty'.This measure counts the frequency of mentions of synonyms for risk or uncertainty, particularly related to Covid-19, divided by the length of the transcript.
control variables is used in 'change regressions' as robustness tests.We also control for the aggregate market behaviour in return regressions by the total market value expressed in billions of USD.As part of the robustness tests, we follow Hassan et al. (2019Hassan et al. ( , 2020) ) and further control different rm-level risks. 7These risk measures rely on word counts that condition on proximity to the use of synonyms for 'risk' or 'uncertainty'.Overall Risk (Covid-19 Risk) is the frequency of mentions of synonyms for risk or uncertainty (related to Covid), divided by transcript length.
Table 1B provides the summary statistics for key variables of immune and control rms, in Panels A and B, respectively.Immune rms are larger in general ($11.9 billion) compared to control rms ($1.4 billion).The right-skewed distribution of Market Value suggests that there are few big rms in both samples of immune and control rms.While the average Return seems to be around −0.1% for both types of rms, highly right-skewed distributions of Traded and Dollar Volume imply that the stocks of certain rms are traded excessively more than others.The positive mean values for Signed Volume indicate that there is a buy pressure in the market.
Finally, Table 1C reports the correlations between the key variables.

Empirical approach
We rst examine whether markets react differently to immune rms than other control rms during the Covid period.Abnormal returns (AR) are measured using three different estimation windows: 3-month, 6-month and 9-month, which end 60 days before the Covid period, that is 2 March 2020.We estimate abnormal returns using two different event windows: 1-month and 2-month, starting on 2 March 2020.Expected returns are   1A.
estimated using the recent ve-factor specication outlined in Fama and French (2015).As discussed by Blitz .( 2018), we use the three-factor model et al and Carhart four-factor model in separate analyses to address the concerns around risk-return and momentum issues, as well as robustness concerns regarding the two additional factors in the ve-factor model.We estimate AR and obtain very similar results.As is common in event study analysis, the identifying assumption is that the Covid pandemic is not correlated with an immune rm's expected return after controlling for the tradable risk factors.Lastly, we construct the cumulative abnormal returns (CARs) for rms in the Covid period.To benchmark our results, we repeat this exercise for control rms as well.
To better understand the causal effect of immunization on rms during the Covid pandemic regarding returns and trading activity, we run a difference-in-difference (DID) analysis by estimating a set of panel regressions of the form where Market reaction i t , represents return, excess return, traded volume, dollar volume, signed volume for rm i on day t; Immune i is a dummy  variable for immunized rms as described before; Post t is a dummy that is equal to 1 for the 2-month period starting on 2 March 2020, and 0 for the 3month period before 2 March 2020; Controls i t , is a set of control variables, that is Mktrf, SMB, HML, RMW and CMA; δ i and µ t are rm and time xed effects, respectively.Immune and Post dummies enter into the model as the interaction term only because individually they are subsumed by the rm and time xed effects, respectively.Market Value is added as a control for all return regressions.In robustness tests, Overall Risk and Covid-19 Risk are included as additional controls.In further analyses, the Post t dummy is adjusted to cover only the rst month of the Covid period, to measure the most immediate reaction of the markets to immune rms.In other analyses, Delta Ln(Traded Volume) and Delta Ln(Dollar Volume) are used as dependent variables representing the daily changes.The associated control variables are also transformed into daily changes for those regressions.For all analyses in this paper, standard errors are clustered at the rm level.Meyer (1995) discusses the main advantages of the DID approach as its simplicity and potential to evade the endogeneity problems that arise when making comparisons between heterogeneous entities.Another key point of DID is that it accounts for change due to factors other than the treatment or intervention being studied.Also, since it focuses on change rather than the absolute levels, the groups being compared can start at different lev-els.Further, the DID method allows the treatment effect to be estimated, and when used in conjunction with a natural experiment, for example Covid, the shock provides the randomization which is essential for DID.Due to these reasons and advantages, we decided to use the DID approach in this study.
Considering the rapid increase in cases and deaths in Figure 1, Covid appears to be an unanticipated, random shock and its impact is immediate and sharp.Hence, endogeneity of this shock should not be an issue for our DID approach.Further, our analysis on daily data includes a 2month post-Covid period in NYC.Due to these reasons, we believe the DID model is adequate concerning any periodicity.However, the assumption is that the errors in the DID model are correlated and residuals are not independent.Hence, simple residual resampling to replicate the correlation in the data fails under a simple bootstrapping method.To validate our results, we conduct block bootstrapping as suggested by Bertrand, Duo and Mullainathan (2004).In particular we implement the overlapping (moving) block bootstrapping method, and our data are split into blocks of 50 observations,8 that is block length b, as by  denition we form n -b 1 blocks, where n is the + length of the time series in our sample.Observations 1 to 50 will be block #1, observations 2 to 51 will be block #2, and so on.Hence, the block bootstrapping approach can replicate the correlations by resampling inside blocks of data.In unreported analyses, we obtain robust results consistent with our original ndings.

Main ndings
In Table 2, we report the CARs for immune rms and control rms during our event periods.We start by discussing the results for CAR analyses  regarding how 1-month and 2-month event periods support our hypothesis that rms which survived the nancial distress associated with 9/11 (Sudarsanam and Lai, 2001) fared better during the Covid period.The results are robust to the choice of different estimation windows.We stress the statistically signicant differences in results between immune and control rms.When we use a 1-month event window, we observe that CARs are not signicantly different from zero and immune rms do not react to the Covid news.However, control rms have a negative reaction to the Covid news.The differences between the CARs of immune rms and those of control rms are  Panel A of this table presents estimates for Immune × Post along with Market Value, SMB, HML, RMW and CMA as control variables.Return, Excess Return, Traded Volume, Dollar Volume and Signed Volume are the dependent variables.For this analysis, the main model is the same, but the timeline is shifted 6 months backwards.Post is the daily dummy variable that is equal to 1 between 3 September 2019 and 31 October 2019, and 0 between 3 June and 30 August 2019.Immune is the dummy variable for rms with HQ in NYC and traded on the NYSE, NYSE American and NASDAQ during both 9/11 and the Covid period, and 0 for the peer (control) rms with HQ in NYC and traded only during the Covid period but not 9/11.Variable denitions are given in Table 1A.Time and rm xed effects are included.Standard errors are clustered by rms and given in parentheses.
* p < 0.10.* * p < 0.05.* * * p < 0.01.Panel B of this table gives t-test results on annual returns for immune rms and their peers (non-survivors).For this panel, peers (non-survivors) are rms with HQ in NYC trading during 9/11 but not Covid.Immune rms are dened the same as before.
economically signicant and vary between 6.8%, 6.6% and 7.1% when we use estimation windows of 3, 6 and 9 months, respectively.When we consider CARs during the extended 2-month event window, we observe again that the differences between the CARs of immune and control rms are economically signicant and vary between 5.2%, 4.5% and 5%, respectively.Table 3 presents results from our DID estimations for rms exposed to 9/11 and those that were not.The rst column reports results for stock returns and the second column reports results for excess returns.The ndings are robust to the choice of return denitions.The DID term is Immune × Post, where Immune takes the value 1 if the rm is exposed to 9/11 and survived, and 0 otherwise.Post is a dummy variable equal to 1 during Covid, and 0 otherwise.Control variables are signicant with expected signs.We observe that the rms exposed to 9/11 and survived have gained immunity   and learnt from such a disastrous experience, as they earn 14% (14.6%) more than their peers in stock returns (excess returns) during the Covid crisis.These ndings support H1. Table 4 presents results from our DID estimations on trading volume, dollar volume and signed volume for rms that faced 9/11 and those that did not during the Covid crisis.The results are robust to the choice of trading volume denitions.For all measures of trading activity, we observe the rms that were exposed to 9/11 and survived have learnt how to manage such crises as their trading activity was signicantly higher than that of their peers during Covid, with coefcient estimates of 0.5 for Traded Volume, 18.9 for Dollar Volume and 0.6 for Signed Volume.The positive coefcient for Signed Volume indicates that the higher trading activity is due to buy pressure by investors request-ing more shares of immune rms.These results support H2.

Tests on resilience and peer rms
To analyse the resilience of immune rms, we conduct a placebo test where we keep our main model the same but shift the timeline 6 months (also 9 and 12 months in untabulated analyses) backwards.If our main results are driven by the resilience or survivorship of immune rms, then those rms should still perform better than their peers, that is control rms, in 'normal times' when they cannot benet from their prior 9/11 experience.Statistically in-signicant results in Panel A of Table 5 indicate that resilience (or survivorship) of immune rms  is not the reason leading our main results.Further, we compare the immune rms to their peers during 9/11 which did not survive until Covid, that is nonsurvivors.Considering a full period and 5-year period before and after 9/11, we conduct t tests on annual stock returns.The statistically insignicant results in Panel B of Table 5 show that immune rms do not perform differently from their peers, either pre-or post-9/11.

Additional risk factors
In this section we introduce additional risk factors.Panel A of Table 6 reports results for Overall Risk and Panel B for Covid-19 Risk, which we measure following the methodology as described in Hassan et al. (2019Hassan et al. ( , 2020)).We use word counts that condition on proximity using synonyms for 'risk' or 'uncertainty' overall or related to Covid scaled by the length of the transcript.Controlling for overall risk, rms exposed to the 9/11 shock earn 14.3% more in returns and 15% more in excess returns and have signicantly higher trading activity regardless of the measure we use.Similarly, when we control for Covid-19 Risk in particular, immune rms earn 12.6% and 13.2% higher returns and excess returns than control rms and have higher trading activity during Covid.

Focused time period
In this section we focus on the very rst outbreak period for Covid, that is the rst month, to capture the very initial reaction of the markets towards immune rms (see   Covid.Similarly, their trading activity is signicantly higher during this initial wave of Covid.

Different measures for trading activity
In this section we repeat the DID analysis using two different measures of trading activity (see Table 8).We follow the methodology in McTier, Tse and Wald (2013) and dene Delta Ln(Traded Volume) and Delta Ln(Dollar Volume) as the daily change in natural logarithm of the traded volume and dollar volume, respectively.This measure is t for our purpose as it introduces a daily change setup in the time frame of analysis, where information changes on a daily basis and there is ample uncertainly.Yet we observe that our main results are robust to the short-term denitions of trading volumes.Coefcient estimates are signicant and in the range of 1% and 1.2%, respectively, for Delta Ln(Traded Volume) and Delta Ln(Dollar Volume).

Analysis excluding the nance sector
The nance sector corresponds to about 60% of our sample for immune rms, therefore it could be argued that our results are driven by them.In this section we exclude rms in the nance sector and repeat the analysis for rms in other service and manufacturing industries (see Table 9).
Our results indicate that in fact those rms have higher levels of immunization.Firms that survived 9/11 have returns (excess returns) that are 36.4% (36.6%) higher than their peers.Similarly, all trading measures indicate trading activity is higher for such survivors of 9/11.

Analysis of shutdown industries
One of the main features of the Covid pandemic is the disproportionate treatment of immune rms in shutdown versus running industry classication.Following state regulation, the shutdown industries include Recreation, Entertainment, Textile, Mining, Construction, Restaurants and Hotels, and Others; while the running industries are Finance, Healthcare, Consumer Goods, Communication and Utilities.We estimate DID regressions separately for each group (see Table 10).We observe that immune rms in shutdown industries perform much better than their peers during the Covid crisis.They earn returns (excess returns) of 33.9% (33.1%) higher than their non-immune peers during the Covid crisis, and all trading activity measures indicate higher trading volumes.Immune rms in industries that continued to operate during the Covid crisis earn 10.6% (11.6%) more than their peers in stock returns (excess returns), yet marginally statistically significant despite trading activity measures indicating higher trading activity compared to their peers.3 and 4. Immune is the dummy variable for rms with HQ in NYC and traded on the NYSE, NYSE American and NASDAQ during both 9/11 and the Covid period, and 0 for rms with HQ in NYC and traded only during the Covid period but not 9/11.Post is the daily dummy variable that is equal to 1 between 2 March and 30 April 2020, and 0 between 9 December 2019 and 28 February 2020.Variable denitions are given in Table 1A.

Conclusions
Economic turmoil during Covid is unparalleled in recent history.Just as scientists are challenged by the very many unanswered questions posed by coronavirus, economists are similarly puzzled by its associated economic consequences and how nancial markets have reacted to this unprecedented disruption.In this paper, we show that to better understand how nancial markets capture and price corporate risks associated with the pandemic, it may be best to take a step back and look at events that posed a similar degree of shock to the nancial system.One such event with similar levels of shock imposed on nancial markets is the terrorist attacks of September 2001, which mainly targeted NYC.
Having examined rms that were headquartered in NYC and traded on one of its three stock exchanges during 2001, we zero in on those rms that managed to nancially survive 9/11 and were trading just before Covid hit them.We nd that such rms displayed more nancial resilience compared to their peer group during the Covid turmoil.Specically, their stock price losses during the Covid episode were lower by about 7% compared to rms that were not exposed to the 9/11 shock, a gure both statistically and economically signicant, and representing billions of dollars of market value 'saved' compared to the control group.
As explained in the empirical section of the paper, we ran various other tests to show that this nding is robust to alternative nancial proxies, different denitions of estimation windows and control rms.Interestingly, we show that such immunized rms that have learnt from similar struggles in the past also experienced higher trading volumes due to buying pressures from the market.In other words, there is strong evidence that nancial O. K. Tosun, A. Eshraghi, and G. Muradoglu   markets 'price in' and thus value corporate exposure to prior disasters, and by implication, the additional organizational resilience gained through such experiences.We argue that this organizational resilience is created through organizations 'learning' how to respond to systemic shocks of a disastrous nature, even if they are few and far between.The way organizations respond to these systemic shocks becomes part and parcel of their organizational culture and work processes that constitute their 'organisational memory' (Walsh and Ungson, 1991).Learning may also happen through fostering attitudes and processes that address error management within the organization following challenging times, particularly through a 'no blame' organizational learning approach (e.g.Provera, Montefusco and Canato, 2010).
In essence humans, and by extension organizations, categorize their knowledge around frames/schema.These frames typically enable organizations to interpret events through organiza-tional learning and memory processes.However, at times, they can also lead to a distorted construction of the accepted version of 'reality' (Goffman, 1974).This is particularly the case for traumatic and unusual events, where the sense-making process is more easily disturbed and thus opportunities for genuine learning arise (e.g.Smith and Elliott, 2007;Weick, 1995).In our context, this learning occurred (and proved useful later during Covid) in organizations that were exposed to the unprecedented and traumatic experience of 9/11.Thus, our ndings make contributions to (1) the literature on market reactions to pandemics and other public health events; (2) the body of work on market reactions to terrorist events and the associated spillover; (3) the literature on the long-term memory of stock markets; and (4) the literature on organizational learning facilitated by exposure to disastrous events.Furthermore, these results are of considerable importance to the corporate sector and to policymakers from a practical point of view, given that the likelihood of similar epidemics and pandemics occurring in the future is seen as nontrivial.
explore the interface of organizational learning and innovation, by categorizing organizations into those characterized by adaptive learning and incremental innovation and those characterized by generative learning and radical innovation.© 2021 The Authors.British Journal of Management published by John Wiley & Sons Ltd on behalf of British Academy of Management.

Figure 1 .
Figure 1.Ofcial Covid-19 cases and deaths in New York City This gure presents the daily new Covid-19 cases and deaths in New York City between 2 March 2020 and 30 April 2020.The very rst ofcial Covid-19 case was after 2 March.Data obtained from NYC government website.

Figure 2 .
Figure 2. Industry distribution for immune and control rms This gure shows the distribution of immune and control rms according to their industry classication.Industry aggregation is based on four-digit SIC codes.The 30 industry classication codes are used to construct the industries.They are obtained from Kenneth French's website.The period is between 9 December 2019 and 30 April 2020.Panel A: Distribution for 114 immune rms.Panel B: Distribution for 331 control rms.[Colour gure can be viewed at wileyonlinelibrary.com]

©
2021 The Authors.British Journal of Management published by John Wiley & Sons Ltd on behalf of British Academy of Management.

©
2021 The Authors.British Journal of Management published by John Wiley & Sons Ltd on behalf of British Academy of Management.
Immune × Post along with Mktrf, SMB, HML, RMW and CMA as control variables.Return and Excess Return are the dependent variables.Immune is the dummy variable for rms with HQ in NYC and traded on the NYSE, NYSE American and NASDAQ during both 9/11 and the Covid period, and 0 for rms with HQ in NYC and traded only during the Covid period but not 9/11.Post is the daily dummy variable that is equal to 1 between 2 March and 30 April 2020, and 0 between 9 December 2019 and 28 February 2020.Return is the daily stock return as a percentage.Excess Return is the daily stock return in excess of the risk-free rate that is proxied by the 1-month T-bill rate.Variable denitions are given in Table 1A.Time and rm xed effects are included.Standard errors are clustered by rms and given in parentheses.* p < 0.10.* * p < 0.05.* * * p < 0.01.
along with × SMB, HML, RMW and CMA as control variables.Traded Volume, Dollar Volume and Signed Volume are the dependent variables.Immune is the dummy variable for rms with HQ in NYC and traded on the NYSE, NYSE American and NASDAQ during both 9/11 and the Covid period, and 0 for rms with HQ in NYC and traded only during the Covid period but not 9/11.Post is the daily dummy variable that is equal to 1 between 2 March and 30 April 2020, and 0 between 9 December 2019 and 28 February 2020.Traded Volume is the amount of shares traded daily, in millions.Dollar Volume is the amount of shares traded multiplied by the daily closing price, in million USD.Signed Volume is the amount of shares traded multiplied by the daily stock return, in tens of thousands.Variable denitions are given in Table 1A.Time and rm xed effects are included.Standard errors are clustered by rms and given in parentheses.* p < 0.10.* * p < 0.05.* * * p < 0.01.

©
2021 The Authors.British Journal of Management published by John Wiley & Sons Ltd on behalf of British Academy of Management.

©
2021 The Authors.British Journal of Management published by John Wiley & Sons Ltd on behalf of British Academy of Management.
along with Overall Risk (Panel A) and Covid-19 Risk (Panel B) as additional control × variables.Original control variables are also included in the model.Return, Excess Return, Traded Volume, Dollar Volume and Signed Volume are the dependent variables.Following Hassan .(2019, 2020), Overall Risk (Covid-19 Risk) relies on word counts that et al condition on proximity to the use of synonyms for 'risk' or 'uncertainty'.This measure counts the frequency of mentions of synonyms for risk or uncertainty divided by transcript length.Variable denitions are given in Table 1A.Time and rm xed effects are included.Standard errors are clustered by rms and given in parentheses.* p < 0.10.* * p < 0.05.* * * p < 0.01.
along with Market Value, SMB, HML, RMW and CMA as control variables.Return, Excess Return, Traded Volume, Dollar Volume and Signed Volume are the dependent variables.For this analysis, Post includes only the rst month of the Covid period.Particularly, Post is the daily dummy variable that is equal to 1 between 2 March and 31 March 2020, and 0 between 9 December 2019 and 28 February 2020.Immune is the dummy variable for rms with HQ in NYC and traded on the NYSE, NYSE American and NASDAQ during both 9/11 and Covid, and 0 for rms with HQ in NYC and traded only during Covid but not 9/11.Variable denitions are given in Table 1A.Time and rm xed effects are included.Standard errors are clustered by rms and given in parentheses.* p < 0.10.* * p < 0.05.* * * p < 0.01.
along with Delta × Mktrf, Delta SMB, Delta HML, Delta RMW and Delta CMA as control variables.Delta Ln(Traded Volume) and Delta Ln(Dollar Volume) are the new dependent variables.Following McTier, Tse and Wald (2013), Delta Ln(Traded Volume) and Delta Ln(Dollar Volume) are calculated as the daily change in natural logarithm of the traded volume and dollar volume, respectively.Immune is the dummy variable for rms with HQ in NYC and traded on the NYSE, NYSE American and NASDAQ during both 9/11 and the Covid period, and 0 for rms with HQ in NYC and traded only during the Covid period but not 9/11.Post is the daily dummy variable that is equal to 1 between 2 March and 30 April 2020, and 0 between 9 December 2019 and 28 February 2020.Variable denitions are given in Table 1A.Time and rm xed effects are included.Standard errors are clustered by rms and given in parentheses.* p < 0.10.* * p < 0.05.* * * p < 0.01.

©
2021 The Authors.British Journal of Management published by John Wiley & Sons Ltd on behalf of British Academy of Management.
along with control variables.Return, Excess Return, Traded Volume, Dollar Volume × and Signed Volume are the dependent variables.Analyses are conducted for shutdown and running industries during the Covid-19 pandemic, separately.Shutdown industries include Recreation, Entertainment, Textile, Mining, Construction, Restaurants and Hotels, and Others; while running industries are Finance, Healthcare, Consumer Goods, Communication and Utilities.Immune is the dummy variable for rms with HQ in NYC and traded on the NYSE, NYSE American and NASDAQ during both 9/11 and the Covid period, and 0 for rms with HQ in NYC and traded only during the Covid period but not 9/11.Post is the daily dummy variable that is equal to 1 between 2 March and 30 April 2020, and 0 between 9 December 2019 and 28 February 2020.Variable denitions are given in Table 1A.Time and rm xed effects are included.Standard errors are clustered by rms and given in parentheses.* p < 0.10.* * p < 0.05.* * * p < 0.01.

Table 1A .
Denition of variablesVariable DescriptionImmuneThe dummy variable for rms with HQ in NYC and traded on NYSE, NYSE American and NASDAQ during both 9/11 and the Covid period; 0 for rms with HQ in NYC and traded only during the Covid period not 9/11.Post

Table 1B .
Descriptive statistics of key variablesThis table presents mean, standard deviation, 25th percentile (P25), median and 75th percentile (P75) values of immune and control rms in the sample.While Panel A provides the statistics for 114 immune rms, Panel B gives the values for 331 control rms.The period is between 9 December 2019 and 30 April 2020.Market Value is daily closing price multiplied by common shares outstanding, in billion USD.Return is the daily stock return as a percentage.Traded Volume is the amount of shares traded daily for a stock in millions.Dollar Volume is the amount of shares traded multiplied by the daily closing price, in million USD.Signed Volume is the amount of shares traded multiplied by the daily stock return in tens of thousands.

Table 1C .
Correlation table This table presents the correlation between Return, Excess Return, Traded Volume, Dollar Volume, Signed Volume, Market Value, Mktrf, SMB, HML, RMW and CMA.Variable denitions are given in Table

Table 2 .
Cumulative abnormal returns of immune and control rms during CovidThis table presents the cumulative abnormal returns for immune and control rms during the Covid pandemic.Daily abnormal returns represent the return realized by an investor in excess of sources of systematic risk.The table reports the results using the Fama-French Five-Factor model for different estimation periods (3 months, 6 months and 9 months), 2 months before the rst ofcial Covid-19 case in NYC.The results are given for two different event windows (1 month and 2 months) after the rst ofcial Covid case recorded in NYC.The differences between CAR values of immune and control rms are also reported, along with the statistical signicance.Robust standard errors are reported in parentheses.

Table 3 .
Difference-in-difference analysis on returns

Table 4 .
Difference-in-difference analysis on trading activity

Table 5 .
Placebo test and t test for returns and trading activity

Table 6 . Additional risk factors Panel A: Controlling for overall risk
This table presents estimates for Immune Post

Table 7 .
First month of Covid pandemic This table presents estimates for Immune × Post

Table 7
British Journal of Management published by John Wiley & Sons Ltd on behalf of British Academy of Management.

Table 8 .
Difference-in-difference analysis with different measures for trading

Table 9 .
Analyses excluding nance sector This table presents estimates forImmune × Post along with Market Value, SMB, HML, RMW and CMA as control variables.Return, Excess Return, Traded Volume, Dollar Volume and Signed Volume are the dependent variables.The analyses are conducted excluding the major industry in the sample (i.e.Finance) to test the robustness of the original ndings in Tables

Table 10 .
Analysis of shutdown and running industries during the Covid period

Running industries during the Covid pandemic
This table presents estimates for Immune Post